.\" t '\" vim:set syntax=groff:
.\"Copyright (C) 2009-2015 Kaz Kylheku <kaz@kylheku.com>.
.\"All rights reserved.
.\"
.\"BSD License:
.\"
.\"Redistribution and use in source and binary forms, with or without
.\"modification, are permitted provided that the following conditions
.\"are met:
.\"
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in
.\" the documentation and/or other materials provided with the
.\" distribution.
.\" 3. The name of the author may not be used to endorse or promote
.\" products derived from this software without specific prior
.\" written permission.
.\"
.\"THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
.\"IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
.\"WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
.\"
.\" Useful groff definitions.
.\"
.\" Some constants that depend on troff/nroff mode:
.ie n \{\
.ds vspc 1
.\}
.el \{\
.ds vspc 0.5
.\}
.\" Mount numeric fonts when not running under man2html
.if !\n(M2 \{\
. fp 4 CR
. fp 5 CI
.\}
.\" Base font
.nr fsav 1
.\" start of code block: switch to monospace font
.de cblk
. ft 4
..
.\" end of code block: restore font
.de cble
. ft 1
..
.\" typeset arguments in monospace
.\" .code x y z -> \f[CR]x y z\f[]
.de code
\f[4]\\$*\f[]
..
.\" like .code typesets meta-syntax
.\" which is done in angle brackets in nroff or oblique
.\" courier in PDF/HTML.
.de meta
. ie n \{\
<\\$*>
. \}
. el \{\
\f[5]\\$*\f[]
. \}
..
.\" like .meta but tack on second agument with no space.
.de metn
. ie n \{\
<\\$1>\\$2
. \}
. el \{\
\f[5]\\$1\f[]\\$2
. \}
..
.\" like .code but wraps in quotes
.\" .str x y z -> \f[CR]"x y z"\f[].
.de str
\f[4]"\\$*"\f[]
..
.\" wrap first argument in quotes, tack no second one with no space
.\" .strn x y -> \f[CR]"x"\f[]y.
.de strn
\f[4]"\\$1"\f[]\\$2
..
.\" like .IP but use monospace too
.de coIP
. IP "\\f[4]\\$*\\f[]"
..
.\" Directive heading
.de dir
. SS* The \f[4]\\$1\f[] directive
..
.\" Multiple directive heading
.de dirs
. ds s "
. while (\\n[.$]>2) \{\
. as s \f[4]\\$1\f[],
. shift
. \}
. if (\\n[.$]>1) \{\
. as s \f[4]\\$1\f[]
. shift
. \}
. if (\\n[.$]>0) \{\
. as s and \f[4]\\$1\f[]
. \}
. SS* The \\*s directives
..
.\" heading with code in position 1
.de c1SS
. ds s \\f[4]\\$1\\f[]
. shift
. as s " \\$*
. SS* \\*s
..
.\" utility macro for gathering material into "s" string.
.\" a pair of arguments "@ arg" becomes arg set in code
.\" a pair of arguments "@, arg" becomes "arg," where
.\" arg is set in code, followed by comma not in code.
.de gets
. ds s "
. while (\\n[.$]>0) \{\
. ie "\\$1"@" \{\
. shift
. as s \f[4]\\$1\f[]
. shift
. \}
. el \{\
. ie "\\$1"@," \{\
. shift
. as s \f[4]\\$1\f[],
. shift
. \}
. el \{\
. as s \\$1
. shift
. \}
. \}
. \}
..
.\" a macro for gathering material into "s"
.\" a pair of arguments "< arg" is typeset like
.\" .meta arg. "<< arg arg" is like .metn arg arg.
.ie n \{\
. de getm
. ds s "
. while (\\n[.$]>0) \{\
. ie "\\$1"<" \{\
. shift
. as s <\\$1>
. shift
. \}
. el \{\
. ie "\\$1"<<" \{\
. shift
. as s <\\$1>\\$2
. shift
. shift
. \}
. el \{\
. ie "\\$1">>" \{\
. shift
. as s \\$1<\\$2>
. shift
. shift
. \}
. el \{\
. ie "\\$1"<>" \{\
. shift
. as s \\$1<\\$2>\\$3
. shift
. shift
. shift
. \}
. el \{\
. as s \\$1
. shift
. \}
. \}
. \}
. \}
. \}
. .
.\}
.el \{\
. de getm
. ds s "
. while (\\n[.$]>0) \{\
. ie "\\$1"<" \{\
. shift
. as s \\f5\\$1\\f4
. shift
. \}
. el \{\
. ie "\\$1"<<" \{\
. shift
. as s \\f5\\$1\\f4\\$2
. shift
. shift
. \}
. el \{\
. ie "\\$1">>" \{\
. shift
. as s \\$1\\f5\\$2\\f4
. shift
. shift
. \}
. el \{\
. ie "\\$1"<>" \{\
. shift
. as s \\$1\\f5\\$2\f4\\$3
. shift
. shift
. shift
. \}
. el \{\
. as s \\$1
. shift
. \}
. \}
. \}
. \}
. \}
. .
.\}
.\" typeset all arguments by the last one in monospace, followed
.\" by the last one in previous font with no space
.\" .codn x y ... z -> \f[CR]x y ...\f[]z
.de codn
. ds s "
. while (\\n[.$]>2) \{\
. as s \\$1
. shift
. \}
\f[4]\\*s\\$1\f[]\\$2
..
.\" .cod1 a b c -> abc where a is typeset as code
.de cod1
\\$1\f[4]\\$2\f[]\$3
..
.\" .cod2 a b -> ab where b is typeset as code
.de cod2
\\$1\f[4]\\$2\f[]
..
.\" .cod3 a b c -> abc where a and c are typeset as code
.de cod3
\f[4]\\$1\f[]\\$2\f[4]\\$3\f[]
..
.\" Syntax section markup
.de synb
. TP* Syntax:
. cblk
..
.de syne
. cble
..
.\" Require section markup
.de reqb
. TP* Requires:
. cblk
..
.de reqe
. cble
..
.\" Used for meta-variables in syntax blocks
.de mets
. nr fsav \\n[.f]
. getm \\$*
. \"workaround for man2html:
. as s \\f\\n[fsav]
\\*s
. ft \\n[fsav]
..
.\" Used for meta-variables in inline blocks
.de meti
. nr fsav \\n[.f]
. getm \\$*
. \"workaround for man2html:
. as s \\f\\n[fsav]
\\*s
. ft \\n[fsav]
..
.\" Used for meta-variables in .coIP
.de meIP
. nr fsav \\n[.f]
. getm \\$*
. \"workaround for man2html:
. as s \\f\\n[fsav]
.coIP \\*s
. ft \\n[fsav]
..
.\" Description section
.de desc
. TP* Description:
..
.\" Section counters: heading, section, paragraph.
.nr shco 0 1
.nr ssco 0 1
.nr spco 0 1
.\" wrapper for .SH
.de SH*
. SH \\n+[shco] \\$*
. rs
. nr ssco 0
. nr spco 0
. sp \*[vspc]
. ns
..
.\" wrapper for .SS
.de SS*
. SS \\n[shco].\\n+[ssco] \\$*
. rs
. nr spco 0
. sp \*[vspc]
. ns
..
.\" wrapper for .TP
.de TP*
. ds s \\$1
. shift
. TP \\$*
\\*s
. sp \*[vspc]
. ns
..
.\" numbered paragraph
.de NP*
. ie \n(M2 \{\
. M2SS 2 H4 "\\n[shco].\\n[ssco].\\n+[spco] \\$*"
. \}
. el \{\
. TP* "\f[B]\\n[shco].\\n[ssco].\\n+[spco] \\$*\f[]"
. \}
. PP
..
.\" process arguments using .gets so that some material
.\" is typeset as code. Then pass to .SS* section macro.
.de coSS
. gets \\$*
. SS* \\*s
..
.\" like coSS but targetting NP*
.de coNP
. gets \\$*
. NP* \\*s
..
.\" like coSS but use monospace IP
.de ccIP
. gets \\$*
. IP "\\*s"
..
.\" TXR name
.ds TX \f[B]TXR\f[]
.ds TL \f[B]TXR Lisp\f[]
.\" Start of man page:
.TH TXR 1 2015-04-26 "Utility Commands" "TXR Data Processing Language" "Kaz Kylheku"
.SH* NAME
\*(TX \- text processing language (version 107)
.SH* SYNOPSIS
.cblk
.meti txr >> [ options ] < query-file < data-files ..
.cble
.sp
.SH* DESCRIPTION
\*(TX is a language oriented toward processing text from files or streams, using
multiple programming paradigms.
A \*(TX script is called a query, and it specifies a pattern which matches (a
prefix of) an entire file, or multiple files. Patterns can consists of large
chunks of multi-line free-form text, which is matched literally against
material in the input sources. Free variables occurring in the pattern
(denoted by the
.code @
symbol) are bound to the pieces of text occurring in the
corresponding positions. If the overall match is successful, then
\*(TX can do one of two things: it can report the list of variables which were
bound, in the form of a set of variable assignments which can be evaluated by
the
.B eval
command of the POSIX shell language, or generate a custom report according
to special directives in the query. Patterns can be arbitrarily complex,
and can be broken down into named pattern functions, which may be mutually
recursive. \*(TX patterns can work horizontally (characters within a line)
or vertically (spanning multiple lines). Multiple lines can be treated
as a single line.
In addition to embedded variables which implicitly match text, the
\*(TX query language supports a number of directives, for matching text using
regular expressions, for continuing a match in another file, for searching
through a file for the place where an entire sub-query matches, for collecting
lists, and for combining sub-queries using logical conjunction, disjunction and
negation, and numerous others.
Furthermore, embedded within \*(TX is a powerful Lisp dialect. \*(TL supports
functional and imperative programming, and provides data types such as symbols,
strings, vectors, hash tables with weak reference support, lazy lists, and
arbitrary-precision (bignum integers).
.SH* ARGUMENTS AND OPTIONS
Options which don't take an argument may be combined together.
The
.code -v
and
.code -q
options are mutually exclusive. Of these two, the one which
occurs in the rightmost position in the argument list dominates.
The
.code -c
and
.code -f
options are also mutually exclusive; if both are specified,
it is a fatal error.
.meIP >> -D var=value
Bind the variable
.meta var
to the value
.meta value
prior to processing the query. The name is in scope over the entire
query, so that all occurrence of the variable are substituted and
match the equivalent text. If the value contains commas, these
are interpreted as separators, which give rise to a list value.
For instance
.code -Da,b,c
creates a list of the strings
.strn "a" ,
.str "b"
and
.strn "c" .
(See Collect Directive bellow). List variables provide a multiple
match. That is to say, if a list variable occurs in a query, a successful
match occurs if any of its values matches the text. If more than one
value matches the text, the first one is taken.
.meIP >> -D var
Binds the variable
.meta var
to an empty string value prior to processing the query.
.coIP -q
Quiet operation during matching. Certain error messages are not reported on the
standard error device (but the if the situations occur, they still fail the
query). This option does not suppress error generation during the parsing
of the query, only during its execution.
.coIP -d
.coIP --debugger
Invoke the interactive \*(TX debugger. See the DEBUGGER section.
.coIP -n
.coIP --noninteractive
This option affects behavior related to \*(TX's
.code *std-input*
stream. Normally, if this stream is connected to a terminal device, it is
automatically marked as having the real-time property when \*(TX starts up (see
the functions .code stream-set-prop and
.codn real-time-stream-p ).
The
.code -n
option suppresses this behavior; the
.code *std-input*
stream remains ordinary.
The \*(TX pattern language reads standard input via
a lazy list, created by applying the
.code lazy-stream-cons
function to the
.code *std-input*
stream. If that stream is marked real-time, then the lazy list which is
returned by that function has behaviors that are better suited for scanning
interactive input. A more detailed explanation is given under the description
of this function.
.coIP -v
Verbose operation. Detailed logging is enabled.
.coIP -b
This is a deprecated option, which is silently ignored. In \*(TX versions
prior to 90, the printing of variable bindings (see
.code -B
option) was
implicit behavior which was automatically suppressed in certain situations.
The -b option suppressed it unconditionally.
.coIP -B
If the query is successful, print the variable bindings as a sequence
of assignments in shell syntax that can be
.IR eval -ed
by a POSIX shell.
II the query fails, print the word "false". Evaluation of this word
by the shell has the effect of producing an unsuccessful termination
status from the shell's
.I eval
command.
.coIP "-l or --lisp-bindings"
This option implies
.codn -B .
Print the variable bindings in Lisp syntax instead of
shell syntax.
.meIP -a < num
This option implies
.codn -B .
The decimal integer argument
.meta num
specifies the maximum
number of array dimensions to use for list-valued variable bindings.
The default is 1. Additional dimensions are expressed using numeric suffixes
in the generated variable names.
For instance, consider the three-dimensional list arising out of a triply
nested collect:
.cblk
((("a" "b") ("c" "d")) (("e" "f") ("g" "h"))).
.cble
Suppose this is bound to a variable V. With
.codn -a 1 ,
this will be
reported as:
.cblk
V_0_0[0]="a"
V_0_1[0]="b"
V_1_0[0]="c"
V_1_1[0]="d"
V_0_0[1]="e"
V_0_1[1]="f"
V_1_0[1]="g"
V_1_1[1]="h"
.cble
With
.codn -a 2 ,
it comes out as:
.cblk
V_0[0][0]="a"
V_1[0][0]="b"
V_0[0][1]="c"
V_1[0][1]="d"
V_0[1][0]="e"
V_1[1][0]="f"
V_0[1][1]="g"
V_1[1][1]="h"
.cble
The leftmost bracketed index is the most major index. That is to say,
the dimension order is:
.codn "NAME_m_m+1_..._n[1][2]...[m-1]" .
.meIP -c < query
Specifies the query in the form of a command line argument. If this option is
used, the query-file argument is omitted. The first non-option argument,
if there is one, now specifies the first input source rather than a query.
Unlike queries read from a file, (non-empty) queries specified as arguments
using -c do not have to properly end in a newline. Internally,
\*(TX adds the missing newline before parsing the query. Thus
.code -c
.str @a
is a valid query which matches a line.
Example:
Shell script which uses \*(TX to read two lines
.str 1
and
.str 2
from standard input,
binding them to variables
.code a
and
.codn b .
Standard
input is specified as
.code -
and the data
comes from shell "here document" redirection:
.RS
.IP code:
.cblk
\ #!/bin/sh
txr -B -c "@a
@b" - <<!
1
2
!
.cble
.IP output:
.cblk
\ a=1
b=2
.cble
.PP
The
.code @;
comment syntax can be used for better formatting:
.cblk
txr -B -c "@;
@a
@b"
.cble
.RE
.meIP -f < query-file
Specifies the file from which the query is to be read, instead of the
.meta query-file
argument. This is useful in
.code #!
("hash bang") scripts. (See Hash Bang Support below).
.meIP -e < expression
Evaluates a \*(TL expression for its side effects, without printing
its value. Can be specified more than once. The
.meta query-file
argument becomes optional if
.code -e
is used at least once. If the evaluation of every
.meta expression
evaluated this way terminates normally, and there is no
.meta query-file
argument, then \*(TX terminates with a successful status.
.meIP -p < expression
Just like
.code -e
but prints the value of
.meta expression
using the
.code prinl
function.
.meIP -P < expression
Like
.code -p
but prints using the
.code pprinl
function.
.meIP -t < expression
Like
.code -p
but prints using the
.code tprint
function.
.meIP -C < number
.meIP >> --compat= number
Requests \*(TX to behave in a manner that is compatible with the specified
version of \*(TX. This makes a difference in situations when a release of
\*(TX breaks backward compatibility. If some version N+1 deliberately introduces
a change which is backward incompatible, then
.code -C N
can be used to request the old behavior.
The requested value of N can be too low, in which case \*(TX will
complain and exit with an unsuccessful termination status. This indicates
that \*(TX refuses to be compatible with such an old version. Users requiring
the behavior of that version will have to install an older version of \*(TX which
supports that behavior, or even that exact version.
If the option is specified more than once, the behavior is not specified.
For more information, see the COMPATIBILITY section.
.meIP >> --gc-delta= number
The
.meta number
argument to this option must be a decimal integer. It represents
a megabyte value, the "GC delta": one megabyte is 1048576 bytes. The "GC
delta" controls an aspect of the garbage collector behavior.
See the
.code gc-set-delta
function for a description.
.coIP --help
Prints usage summary on standard output, and terminates successfully.
.coIP --license
Prints the software license. This depends on the software being
installed such that the LICENSE file is in the data directory.
Use of \*(TX implies agreement with the liability disclaimer in the license.
.coIP --version
Prints program version standard output, and terminates successfully.
.coIP --args
The
.code --args
option provides a way to encode multiple arguments as a single
argument, which is useful on some systems which have limitations in
their implementation of the "hash bang" mechanism. For details about
its special syntax, See Hash Bang Support below.
.coIP --
Signifies the end of the option list.
.coIP -
This argument is not interpreted as an option, but treated as a filename
argument. After the first such argument, no more options are recognized. Even
if another argument looks like an option, it is treated as a name.
This special argument
.code -
means "read from standard input" instead of a file.
The query file, or any of the data files, may be specified using this option.
If two or more files are specified as
.codn - ,
the behavior is system-dependent.
It may be possible to indicate EOF from the interactive terminal, and
then specify more input which is interpreted as the second file, and so forth.
.PP
After the options, the remaining arguments are files. The first file argument
specifies the query, and is mandatory. A file argument consisting of a single
.code -
means to read the standard input instead of opening a file. A file argument
which begins with an exclamation symbol means that the rest of the argument is
a shell command which is to be run as a coprocess, and its output read like a
file.
.PP
\*(TX begins by reading the query. The entire query is scanned, internalized
and then begins executing, if it is free of syntax errors. The reading of
data, on the other hand, is lazy. A file isn't opened until the query demands
material from that file, and then the contents are read on demand, not all at
once.
If no files arguments are specified on the command line, it is up to the
query to open a file, pipe or standard input via the
.code @(next)
directive
prior to attempting to make a match. If a query attempts to match text,
but has run out of files to process, the match fails.
.SH* STATUS AND ERROR REPORTING
\*(TX sends errors and verbose logs to the standard error device. The following
paragraphs apply when \*(TX is run without enabling verbose mode with
.codn -v ,
or the printing of variable
bindings with
.code -B
or
.codn -a .
If the command line arguments are incorrect, or the query has a malformed
syntax, \*(TX issues an error diagnostic and terminates with a failed status.
If the query fails due to a mismatch, \*(TX terminates
with a failed status. No diagnostics are issued.
If the query is well-formed, and matches, then \*(TX issues
no diagnostics, and terminates with a successful status.
In verbose mode (option
.codn -v ),
\*(TX issues diagnostics on the standard error device even in situations which
are not erroneous.
In bindings-printing mode (options
.code -B
or
.codn -a) ,
\*(TX prints the word
.code false
if the query fails, and exits with a failed
termination status. If the query succeeds, the variable bindings, if any,
are output on standard output.
.SH* BASIC QUERY SYNTAX AND SEMANTICS
.SS* Comments
A query may contain comments which are delimited by the sequence
.code @;
and extend to the end of the line. No whitespace can occur between the
.code @
and
.codn ; .
A comment which begins on a line swallows that entire line, as well as the
newline which terminates it. In essence, the entire comment line disappears.
If the comment follows some material in a line, then it does not consume
the newline. Thus, the following two queries are equivalent:
.IP 1.
.cblk
\ @a@; comment: match whole line against variable @a
@; this comment disappears entirely
@b
.cble
.IP 2.
.cblk
\ @a
@b
.cble
.PP
The comment after the
.code @a
does not consume the newline, but the
comment which follows does. Without this intuitive behavior,
line comment would give rise to empty lines that must match empty
lines in the data, leading to spurious mismatches.
Instead of the
.code ;
character, the
.code #
character can be used. This is an obsolescent feature.
.SS* Hash Bang Support
\*(TX has several features which support use of the "hash bang" convention
for creating apparently stand-alone executable programs.
If the first line of a query begins with the characters
.codn #! ,
that entire line is deleted from the query. This allows
for \*(TX queries to be turned into standalone executable programs in the POSIX
environment.
Shell example: create a simple executable program called
.str "twoline.txr"
and
run it. This assumes \*(TX is installed in
.codn /usr/bin .
.cblk
$ cat > hello.txr
#!/usr/bin/txr
@(bind a "Hey")
@(output)
Hello, world!
@(end)
$ chmod a+x hello.txr
$ ./hello.txr
Hello, world!
.cble
When this plain hash bang line is used, \*(TX receives the name of the script
as an argument. Therefore, it is not possible to pass additional options
to \*(TX. For instance, if the above script is invoked like this
.cblk
$ ./hello.txr -B
.cble
the -B option isn't processed by \*(TX, but treated as an additional argument,
just as if
.cblk
.meti txr < scriptname -B
.cble
had been executed directly.
This behavior is useful if the script author wants not to expose the
\*(TX options to the user of the script.
However, the hash bang line can use the
.code -f
option:
.cblk
#!/usr/bin/txr -f
.cble
Now, the name of the script is passed as an argument to the
.code -f
option, and \*(TX will look for more options after that, so that the resulting
program appears to accept \*(TX options. Now we can run
.cblk
$ ./hello.txr -B
Hello, world!
a="Hey"
.cble
The
.code -B
option is honored.
On some operating systems, it is not possible to pass more than one
argument through the hash bang mechanism. That is to say, this will
not work.
.cblk
#!/usr/bin/txr -B -f
.cble
To support systems like this, \*(TX supports the special argument
.codn --args .
With
.codn --args ,
it is possible to encode multiple arguments
into one argument. The
.code --args
option must be followed by a separator
character, chosen by the programmer. The characters after that are
split into multiple arguments on the separator character. The
.code --args
option is then removed from the argument list and replaced with these
arguments, which are processed in its place.
Example:
.cblk
#!/usr/bin/txr --args:-B:-f
.cble
The above has the same behavior as
.cblk
#!/usr/bin/txr -B -f
.cble
on a system which supports multiple arguments in hash bang.
The separator character is the colon, and so the remainder
of that argument,
.codn -B:-f ,
is split into the two arguments
.codn "-B -f" .
.SS* Whitespace
Outside of directives, whitespace is significant in \*(TX queries, and represents
a pattern match for whitespace in the input. An extent of text consisting of
an undivided mixture of tabs and spaces is a whitespace token.
Whitespace tokens match a precisely identical piece of whitespace in the input,
with one exception: a whitespace token consisting of precisely one space has a
special meaning. It is equivalent to the regular expression
.codn "@/[ ]+/" :
match an extent of one or more spaces (but not tabs!). Multiple consecutive
spaces do not have this meaning.
Thus, the query line
.str "a b"
(one space between
.code a
and
.codn b )
matches
.str "a b"
with any number of spaces between the two letters.
For matching a single space, the syntax
.code "@\e "
can be used (backslash-escaped space).
It is more often necessary to match multiple spaces than to exactly
match one space, so this rule simplifies many queries and adds inconvenience
to only few.
In output clauses, string and character literals and quasiliterals, a space
token denotes a space.
.SS* Text
Query material which is not escaped by the special character
.code @
is literal text, which matches input character for character. Text which occurs at
the beginning of a line matches the beginning of a line. Text which starts in
the middle of a line, other than following a variable, must match exactly at
the current position, where the previous match left off. Moreover, if the text
is the last element in the line, its match is anchored to the end of the line.
An empty query line matches an empty line in the input. Note that an
empty input stream does not contain any lines, and therefore is not matched
by an empty line. An empty line in the input is represented by a newline
character which is either the first character of the file, or follows
a previous newline-terminated line.
Input streams which end without terminating their last line with a newline are
tolerated, and are treated as if they had the terminator.
Text which follows a variable has special semantics, discussed in the
section Variables below.
A query may not leave a line of input partially matched. If any portion of a
line of input is matched, it must be entirely matched, otherwise a matching
failure results. However, a query may leave unmatched lines. Matching only
four lines of a ten line file is not a matching failure. The
.code eof
directive can be used to explicitly match the end of a file.
In the following example, the query matches the text, even though
the text has an extra line.
.IP code:
.cblk
\ Four score and seven
years ago our
.cble
.IP data:
.cblk
\ Four score and seven
years ago our
forefathers
.cble
.PP
In the following example, the query
.B fails
to match the text, because the text has extra material on one
line that is not matched:
.IP code:
.cblk
\ I can carry nearly eighty gigs
in my head
.cble
.IP data:
.cblk
\ I can carry nearly eighty gigs of data
in my head
.cble
.PP
Needless to say, if the text has insufficient material relative
to the query, that is a failure also.
To match arbitrary material from the current position to the end
of a line, the "match any sequence of characters, including empty"
regular expression
.code @/.*/
can be used. Example:
.IP code:
.cblk
\ I can carry nearly eighty gigs@/.*/
.cble
.IP data:
.cblk
\ I can carry nearly eighty gigs of data
.cble
.PP
In this example, the query matches, since the regular expression
matches the string "of data". (See Regular Expressions section below).
Another way to do this is:
.IP code:
.cblk
\ I can carry nearly eighty gigs@(skip)
.cble
.SS* Special Characters in Text
Control characters may be embedded directly in a query (with the exception of
newline characters). An alternative to embedding is to use escape syntax.
The following escapes are supported:
.meIP >> @\e newline
A backslash immediately followed by a newline introduces a physical line
break without breaking up the logical line. Material following this sequence
continues to be interpreted as a continuation of the previous line, so
that indentation can be introduced to show the continuation without appearing
in the data.
.meIP >> @\e space
A backslash followed by a space encodes a space. This is useful in line
continuations when it is necessary for some or all of the leading spaces to be
preserved. For instance the two line sequence
.cblk
abcd@\e
@\e efg
.cble
is equivalent to the line
.cblk
abcd efg
.cble
The two spaces before the
.code @\e
in the second line are consumed. The spaces after are preserved.
.coIP @\ea
Alert character (ASCII 7, BEL).
.coIP @\eb
Backspace (ASCII 8, BS).
.coIP @\et
Horizontal tab (ASCII 9, HT).
.coIP @\en
Line feed (ASCII 10, LF). Serves as abstract newline on POSIX systems.
.coIP @\ev
Vertical tab (ASCII 11, VT).
.coIP @\ef
Form feed (ASCII 12, FF). This character clears the screen on many
kinds of terminals, or ejects a page of text from a line printer.
.coIP @\er
Carriage return (ASCII 13, CR).
.coIP @\ee
Escape (ASCII 27, ESC)
.meIP @\ex < hex-digits
A
.code @\ex
immediately followed by a sequence of hex digits is interpreted as a hexadecimal
numeric character code. For instance
.code @\ex41
is the ASCII character A.
.meIP @\e < octal-digits
A
.code @\e
immediately followed by a sequence of octal digits (0 through 7) is interpreted
as an octal character code. For instance
.code @\e010
is character 8, same as
.codn @\eb .
.PP
Note that if a newline is embedded into a query line with
.code @\en,
this does not split the line into two; it's embedded into the line and thus
cannot match anything. However,
.code @\en
may be useful in the
.code @(cat)
directive and
in
.codn @(output) .
.SS* Character Handling and International Characters
\*(TX represents text internally using wide characters, which are used to
represent Unicode code points. The query language, as well as all data sources,
are assumed to be in the UTF-8 encoding. In the query language, extended
characters can be used directly in comments, literal text, string literals,
quasiliterals and regular expressions. Extended characters can also be
expressed indirectly using hexadecimal or octal escapes.
On some platforms, wide characters may be restricted to 16 bits, so that
\*(TX can only work with characters in the BMP (Basic Multilingual Plane)
subset of Unicode.
\*(TX does not use the localization features of the system library;
its handling of extended characters is not affected by environment variables
like
.code LANG
and
.codn L_CTYPE .
The program reads and writes only the UTF-8 encoding.
If
\*(TX encounters an invalid bytes in the UTF-8 input, what happens depends on
the context in which this occurs. In a query, comments are read without regard
for encoding, so invalid encoding bytes in comments are not detected. A comment
is simply a sequence of bytes terminated by a newline. In lexical elements
which represent text, such as string literals, invalid or unexpected encoding
bytes are treated as syntax errors. The scanner issues an error message,
then discards a byte and resumes scanning. Certain sequences pass through the
scanner without triggering an error, namely some UTF-8 overlong sequences.
These are caught when when the lexeme is subject to UTF-8 decoding, and treated
in the same manner as other UTF-8 data, described in the following paragraph.
Invalid bytes in data are treated as follows. When an invalid byte is
encountered in the middle of a multibyte character, or if the input
ends in the middle of a multibyte character, or if a character is extracted
which is encoded as an overlong form, the UTF-8 decoder returns to the starting
byte of the ill-formed multibyte character, and extracts just that byte,
mapping it to the Unicode character range U+DC00 through U+DCFF. The decoding
resumes afresh at the following byte, expecting that byte to be the start
of a UTF-8 code.
Furthermore, because \*(TX internally uses a null-terminated character
representation of strings which easily interoperates with C language
interfaces, when a null character is read from a stream, \*(TX converts it to
the code U+DC00. On output, this code converts back to a null byte,
as explained in the previous paragraph. By means of this representational
trick, \*(TX can handle textual data containing null bytes.
.SS* Regular Expression Directives
In place of a piece of text (see section Text above), a regular expression
directive may be used, which has the following syntax:
.cblk
@/RE/
.cble
where the RE part enclosed in slashes represents regular expression
syntax (described in the section Regular Expressions below).
Long regular expressions can be broken into multiple lines using a
backslash-newline sequence. Whitespace before the sequence or after the
sequence is not significant, so the following two are equivalent:
.cblk
@/reg \e
ular/
@/regular/
.cble
There may not be whitespace between the backslash and newline.
Whereas literal text simply represents itself, regular expression denotes a
(potentially infinite) set of texts. The regular expression directive
matches the longest piece of text (possibly empty) which belongs to the set
denoted by the regular expression. The match is anchored to the current
position; thus if the directive is the first element of a line, the match is
anchored to the start of a line. If the regular expression directive is the
last element of a line, it is anchored to the end of the line also: the regular
expression must match the text from the current position to the end of the
line.
Even if the regular expression matches the empty string, the match will fail if
the input is empty, or has run out of data. For instance suppose the third line
of the query is the regular expression
.codn @/.*/ ,
but the input is a file which has
only two lines. This will fail: the data has no line for the regular expression to
match. A line containing no characters is not the same thing as the absence of
a line, even though both abstractions imply an absence of characters.
Like text which follows a variable, a regular expression directive which
follows a variable has special semantics, discussed in the section Variables
below.
.SS* Variables
Much of the query syntax consists of arbitrary text, which matches file data
character for character. Embedded within the query may be variables and
directives which are introduced by a
.code @
character. Two consecutive
.code @@
characters encode a literal
.codn @ .
A variable matching or substitution directive is written in one of several
ways:
.cblk
.mets >> @ sident
.mets <> @{ bident }
.mets >> @* sident
.mets <> @*{ bident }
.mets >> @{ bident <> / regex /}
.mets >> @{ bident >> ( fun >> [ arg ... ])}
.mets >> @{ bident << number }
.cble
The forms with an
.code *
indicate a long match, see Longest Match below.
The last two three forms with the embedded regexp
.cblk
.meti <> / regex /
.cble
or
.meta number
or function
have special semantics; see Positive Match below.
The identifier
.code t
cannot be used as a name; it is a reserved symbol which
denotes the value true. An attempt to use the variable
.code @t
will result in an exception. The symbol
.code nil
can be used where a variable name is required syntacticaly,
but it has special semantics, described in a section below.
A
.meta sident
is a "simple identifier" form which is not delimited by
braces.
A
.meta sident
consists of any combination of
one or more letters, numbers, and underscores. It may not look like a number,
so that for instance
.code 123
is not a valid
.metn sident ,
but
.code 12A
is valid. Case is
sensitive, so that
.code FOO
is different from
.codn foo ,
which is different from
.codn Foo .
The braces around an identifier can be used when material which follows would
otherwise be interpreted as being part of the identifer. When a name is
enclosed in braces it is a
.metn bident .
The following additional characters may be used as part of
.meta bident
which are not allowed in a
.metn sident :
.cblk
! $ % & * + - < = > ? \e _ ~
.cble
The rule still holds that a name cannot look like a number so
.code +123
is not a valid
.meta bident
but these are valid:
.codn a->b ,
.codn *xyz* ,
.codn foo-bar .
The syntax
.code @FOO_bar
introduces the name
.codn FOO_bar ,
whereas
.code @{FOO}_bar
means the
variable named
.str FOO
followed by the text
.strn _bar .
There may be whitespace
between the
.code @
and the name, or opening brace. Whitespace is also allowed in the
interior of the braces. It is not significant.
If a variable has no prior binding, then it specifies a match. The
match is determined from some current position in the data: the
character which immediately follows all that has been matched previously.
If a variable occurs at the start of a line, it matches some text
at the start of the line. If it occurs at the end of a line, it matches
everything from the current position to the end of the line.
.SS* Negative Match
If a variable is one of the plain forms
.cblk
.mets >> @ sident
.mets <> @{ bident }
.mets >> @* sident
.mets <> @*{ bident }
.cble
then this is a "negative match". The extent of the matched text (the text
bound to the variable) is determined by looking at what follows the variable,
and ranges from the current position to some position where the following
material finds a match. This is why this is called a "negative match": the
spanned text which ends up bound to the variable is that in which the match for
the trailing material did not occur.
A variable may be followed by a piece of text, a regular expression directive,
a function call, a directive, another variable, or nothing (i.e. occurs at the
end of a line). These cases are discussed in detail below.
.NP* Variable Followed by Nothing
If the variable is followed by nothing, the negative match extends from the
current position in the data, to the end of the line. Example:
.IP code:
.cblk
\ a b c @FOO
.cble
.IP data:
.cblk
\ a b c defghijk
.cble
.IP result:
.cblk
\ FOO="defghijk"
.cble
.NP* Variable Followed by Text
For the purposes of determining the negative match, text is defined as a
sequence of literal text and regular expressions, not divided by a directive.
So for instance in this example:
.cblk
@a:@/foo/bcd e@(maybe)f@(end)
.cble
.PP
the variable @a is considered to be followed by
.strn ":@/foo/bcd e" .
If a variable is followed by text, then the extent of the negative match is
determined by searching for the first occurrence of that text within the line,
starting at the current position.
The variable matches everything between the current position and the matching
position (not including the matching position). Any whitespace which follows
the variable (and is not enclosed inside braces that surround the variable
name) is part of the text. For example:
.IP code:
.cblk
\ a b @FOO e f
.cble
.IP data:
.cblk
\ a b c d e f
.cble
.IP result:
.cblk
\ FOO="c d"
.cble
.PP
In the above example, the pattern text
.str "a b "
matches the
data
.strn "a b " .
So when the
.code @FOO
variable is processed, the data being
matched is the remaining
.strn "c d e f" .
The text which follows
.code @FOO
is
.strn " e f" .
This is found within the data
.str "c d e f"
at position 3 (counting from 0). So positions 0-2
.cblk
("c d")
.cble
constitute the matching text which is bound to FOO.
.NP* Variable Followed by a Function Call or Directive
If the variable is followed by a function call, or a directive, the extent is
determined by scanning the text for the first position where a match occurs
for the entire remainder of the line. (For a description of functions,
see FUNCTIONS.)
For example:
.cblk
@foo@(bind a "abc")xyz
.cble
Here,
.code foo
will match the text from the current position to where
.str "xyz"
occurs, even though there is a
.code @(bind)
directive. Furthermore, if
more material is added after the xyz, it is part of the search.
Note the difference between the following two:
.cblk
@foo@/abc/@(func)
@foo@(func)@/abc/
.cble
In the first example, the variable foo matches the text from the current
position until the match for the regular expression abc.
.code @(func)
is not
considered when processing
.codn @foo .
In the second example, the variable foo
matches the text from the current position until the position which matches
the function call, followed by a match for the regular expression.
The entire sequence
.code @(func)@/abc/
is considered.
.NP* Consecutive Variables
If an unbound variable specified a fixed-width match or a regular expression,
then the issue of consecutive variables does not arise. Such a variable
consumes text regardless of any context which follows it.
However, what if an unbound variable with no modifier is followed by another
variable? The behavior depends on the nature of the other variable.
If the other variable also has no modifier, this is a semantic error which
will cause the query to fail. A diagnostic message will be issued, unless
operating in quiet mode via
.codn -q .
The reason is that there is no way to bind two
consecutive variables to an extent of text; this is an ambiguous situation,
since there is no matching criterion for dividing the text between two
variables. (In theory, a repetition of the same variable, like
.codn @FOO@FOO ,
could find a solution by dividing the match extent in half, which would work
only in the case when it contains an even number of characters. This behavior
seems to have dubious value).
An unbound variable may be followed by one which is bound. The bound
variable is replaced by the text which it denotes, and the logic proceeds
accordingly.
It is possible for a variable to be bound to a regular expression.
If
.code x
is an unbound variable and
.code y
is bound to a regular expression
.codn RE ,
then
.code @x@y
means
.codn @x@/RE/ .
A variable
.code v
can be bound to a regular expression using, for example,
.codn @(bind v #/RE/) .
The
.code @*
syntax for longest match is available. Example:
.IP code:
.cblk
\ @FOO:@BAR@FOO
.cble
.IP data:
.cblk
\ xyz:defxyz
.cble
.IP result:
.cblk
\ FOO=xyz, BAR=def
.cble
.PP
Here,
.code FOO
is matched with
.strn "xyz" ,
based on the delimiting around the
colon. The colon in the pattern then matches the colon in the data,
so that
.code BAR
is considered for matching against
.strn "defxyz" .
.code BAR
is followed by
.codn FOO ,
which is already bound to
.strn "xyz" .
Thus
.str "xyz"
is located in the
.str "defxyz"
data following
.strn "def" ,
and so BAR is bound to
.strn "def" .
If an unbound variable is followed by a variable which is bound to a list, or
nested list, then each character string in the list is tried in turn to produce
a match. The first match is taken.
An unbound variable may be followed by another unbound variable which specifies
a regular expression or function call match. This is a special case called a
"double variable match". What happens is that the text is searched using the
regular expression or function. If the search fails, than neither variable is
bound: it is a matching failure. If the search succeeds, than the first
variable is bound to the text which is skipped by the search. The second
variable is bound to the text matched by the regular expression or function.
Examples:
.IP code:
.cblk
\ @foo@{bar /abc/}
.cble
.IP data:
.cblk
\ xyz@#abc
.cble
.IP result:
.cblk
\ foo="xyz@#", BAR="abc"
.cble
.PP
.NP* Consecutive Variables Via Directive
Two variables can be de facto consecutive in a manner shown in the
following example:
.cblk
@var1@(all)@var2@(end)
.cble
This is treated just like the variable followed by directive. No semantic
error is identified, even if both variables are unbound. Here,
.code @var2
matches everything at the current position, and so
.code @var1
ends up bound to the empty string.
Example 1:
.code b
matches at position 0 and
.code a
binds the empty string:
.IP code:
.cblk
\ @a@(all)@b@(end)
.cble
.IP data:
.cblk
\ abc
.cble
.IP result:
.cblk
\ a=""
b="abc"
.cble
.PP
Example 2:
.code *a
specifies longest match (see Longest Match below), and so it takes
everything:
.IP code:
.cblk
\ @*a@(all)@b@(end)
.cble
.IP data:
.cblk
\ abc
.cble
.IP result:
.cblk
\ a="abc"
b=""
.cble
.PP
.NP* Longest Match
The closest-match behavior for the negative match can be overridden to longest
match behavior. A special syntax is provided for this: an asterisk between the
.code @
and the variable, e.g:
.IP code:
.cblk
\ a @*{FOO}cd
.cble
.IP data:
.cblk
\ a b cdcdcdcd
.cble
.IP result:
.cblk
\ FOO="b cdcdcd"
.cble
.PP
.IP code:
.cblk
\ a @{FOO}cd
.cble
.IP data:
.cblk
\ a b cdcdcd
.cble
.IP result:
.cblk
\ FOO="b "
b=""
.cble
.PP
In the former example, the match extends to the rightmost occurrence of
.strn "cd" ,
and so
.code FOO
receives
.strn "b cdcdcd" .
In the latter example, the
.code *
syntax isn't used, and so a leftmost match takes place. The extent
covers only the
.strn "b " ,
stopping at the first
.str "cd"
occurrence.
.SS* Positive Match
There are syntactic variants of variable syntax which have an embedded expression
enclosed with the variable in braces:
.cblk
.mets >> @{ bident <> / regex /}
.mets >> @{ bident >> ( fun >> [args ...])}
.mets >> @{ bident << number }
.cble
These specify a variable binding that is driven by a positive match derived
from a regular expression, function or character count, rather than from
trailing material (which is regarded as a "negative" match, since the
variable is bound to material which is
.B skipped
in order to match the trailing material). In the
.cblk
.meti <> / regex /
.cble
form, the match
extends over all characters from the current position which match
the regular expression
.metn regex .
(see Regular Expressions section below).
In the
.cblk
.meti >> ( fun >> [ args ...])
.cble
form, the match extends over characters which
are matched by the call to the function, if the call
succeeds. Thus
.code @{x (y z w)}
is just like
.codn "@(y z w)" ,
except that the region of
text skipped over by
.code @(y z w)
is also bound to the variable
.codn x .
See FUNCTIONS below.
In the
.meta number
form, the match processes a field of text which
consists of the specified number of characters, which must be non-negative
number. If the data line doesn't have that many characters starting at the
current position, the match fails. A match for zero characters produces an
empty string. The text which is actually bound to the variable
is all text within the specified field, but excluding leading and
trailing whitespace. If the field contains only spaces, then an empty
string is extracted.
This syntax is processed without consideration of what other
syntax follows. A positive match may be directly followed by an unbound
variable.
.coSS Special Symbols @ nil and @ t
Just like in the Common Lisp language, the names
.code nil
and
.code t
are special.
.code nil
symbol stands for the empty
list object, an object which marks the end of a list, and boolean false. It is
synonymous with the syntax
.code ()
which may be used interchangeably with
.code nil
in most constructs.
In \*(TL,
.code nil
and
.code t
cannot be used as variables. When evaluated, they evaluate to themselves.
In the \*(TX pattern language,
.code nil
can be used in the variable binding syntax, but does not create a binding;
it has a special meaning. It allows the variable matching syntax to be used to
skip material, in ways similar to the
.codn skip
directive.
The
.code nil
symbol is also used as a
.code block
name, both in the \*(TX pattern language and in \*(TL.
A block named
.code nil
is considered to be anonymous.
.SS* Keyword Symbols
Names whose names begin with the
.code :
character are keyword symbols. These also
may not be used as variables either and stand for themselves. Keywords are
useful for labeling information and situations.
.SS* Regular Expressions
Regular expressions are a language for specifying sets of character strings.
Through the use of pattern matching elements, regular expression is
able to denote an infinite set of texts.
\*(TX contains an original implementation of regular expressions, which
supports the following syntax:
.coIP .
The period is a "wildcard" that matches any character.
.coIP []
Character class: matches a single character, from the set specified by
special syntax written between the square brackets.
This supports basic regexp character class syntax. POSIX
notation like
.code [:digit:]
is not supported.
The regex tokens
.codn \es ,
.code \ed
and
.code \ew
are permitted in character classes, but not their complementing counterparts.
These tokens simply contribute their characters to the class.
The class
.code [a-zA-Z]
means match an uppercase
or lowercase letter; the class
.code [0-9a-f]
means match a digit or
a lowercase letter; the class
.code [^0-9]
means match a non-digit, and so forth.
There are no locale-specific behaviors in \*(TX regular expressions;
.code [A-Z]
denotes an ASCII/Unicode range of characters.
The class
.code [\ed.]
means match a digit or the period character.
A
.code ]
or
.code -
can be used within a character class, but must be escaped
with a backslash. A
.code ^
in the first position denotes a complemented
class, unless it is escaped by backslash. In any other position, it denotes
itself. Two backslashes code for one backslash. So for instance
.code [\e[\e-]
means match a
.code [
or
.code -
character,
.code [^^]
means match any character other
than
.codn ^ ,
and
.code [\e^\e\e]
means match either a
.code ^
or a backslash. Regex operators
such as
.codn * ,
.code +
and
.code &
appearing in a character class represent ordinary
characters. The characters
.codn - ,
.code ]
and
.code ^
occurring outside of a character class
are ordinary. Unescaped
.code /
characters can appear within a character class. The
empty character class
.code []
matches no character at all, and its complement
.code [^]
matches any character, and is treated as a synonym for the
.code .
(period) wildcard operator.
.ccIP @, \es @ \ew and @ \ed
These regex tokens each match a single character.
The
.code \es
regex token matches a wide variety of ASCII whitespace characters
and Unicode spaces. The
.code \ew
token matches alphabetic word characters; it
is equivalent to the character class
.codn [A-Za-z_] .
The
.code \ed
token matches a digit, and is equivalent to
.codn [0-9] .
.ccIP @, \eS @ \eW and @ \eD
These regex tokens are the complemented counterparts of
.codn \es ,
.code \ew
and
.codn \ed .
The
.code \eS
token matches all those characters which
.code \es
does not match,
.code \eW
matches all characters that
.code \ew
does not match and
.code \eD
matches nondigits.
.coIP empty
An empty expression is a regular expression. It represents the set of strings
consisting of the empty string; i.e. it matches just the empty string. The
empty regex can appear alone as a full regular expression (for instance the
\*(TX syntax
.code @//
with nothing between the slashes)
and can also be passed as a subexpression to operators, though this
may require the use of parentheses to make the empty regex explicit. For
example, the expression
.code a|
means: match either
.codn a ,
or nothing. The forms
.code *
and
.code (*)
are syntax errors; though not useful, the correct way to match the
empty expression zero or more times is the syntax
.codn ()* .
.coIP nomatch
The nomatch regular expression represents the
empty set: it matches no strings at all, not even the empty string.
There is no dedicated syntax to directly express nomatch in the regex language.
However, the empty character class
.code []
is equivalent to nomatch, and may be
considered to be a notation for it. Other representations of nomatch are
possible: for instance, the regex
.code ~.*
which is the complement of the regex that
denotes the set of all possible strings, and thus denotes the empty set. A
nomatch has uses; for instance, it can be used to temporarily "comment out"
regular expressions. The regex
.code ([]abc|xyz)
is equivalent to
.codn (xyz) ,
since the
.code []abc
branch cannot match anything. Using
.code []
to "block" a subexpression allows
you to leave it in place, then enable it later by removing the "block".
.coIP (R)
If
.code R
is a regular expression, then so is
.code (R).
The contents of parentheses denote one regular expression unit, so that for
instance in
.codn (RE)* ,
the
.code *
operator applies to the entire parenthesized group.
The syntax
.code ()
is valid and equivalent to the empty regular expression.
.coIP R?
Optionally match the preceding regular expression
.codn R .
.coIP R*
Match the expression
.code R
zero or more times. This
operator is sometimes called the "Kleene star", or "Kleene closure".
The Kleene closure favors the longest match. Roughly speaking, if there are two
or more ways in which
.code R1*R2
can match, than that match occurs in which
.code R1*
matches the longest possible text.
.coIP R+
Match the preceding expression
.code R
one or more times. Like
.codn R* ,
this favors the longest possible match:
.code R+
is equivalent to
.codn RR* .
.coIP R1%R2
Match
.code R1
zero or more times, then match
.codn R2 .
If this match can occur in
more than one way, then it occurs such that
.code R1
is matched the fewest
number of times, which is opposite from the behavior of
.codn R1*R2 .
Repetitions of
.code R1
terminate at the earliest
point in the text where a non-empty match for
.code R2
occurs. Because
it favors shorter matches,
.code %
is termed a non-greedy operator. If
.code R2
is the empty expression, or equivalent to it, then
.code R1%R2
reduces to
. codn R1* .
So for
instance
.code (R%)
is equivalent to
.codn (R*) ,
since the missing right operand is
interpreted as the empty regex. Note that whereas the expression
.code (R1*R2)
is equivalent to
.codn (R1*)R2 ,
the expression
.code (R1%R2)
is
.B not
equivalent to
.codn (R1%)R2 .
.coIP ~R
Match the opposite of the following expression
.codn R ;
that is, match exactly
those texts that
.code R
does not match. This operator is called complement,
or logical not.
.coIP R1R2
Two consecutive regular expressions denote catenation:
the left expression must match, and then the right.
.coIP R1|R2
match either the expression
.code R1
or
.codn R2 .
This operator is known by
a number of names: union, logical or, disjunction, branch, or alternative.
.coIP R1&R2
Match both the expression
.code R1
and
.code R2
simultaneously; i.e. the
matching text must be one of the texts which are in the intersection of the set
of texts matched by
.code R1
and the set matched by
.codn R2 .
This operator is called intersection, logical and, or conjunction.
.PP
Any character which is not a regular expression operator, a backslash escape,
or the slash delimiter, denotes one-position match of that character itself.
Any of the special characters, including the delimiting
.codn / ,
and the backslash, can be escaped with a backslash to suppress its
meaning and denote the character itself.
Furthermore, all of the same escapes as are described in the section Special
Characters in Text above are supported - the difference is that in regular
expressions, the
.code @
character is not required, so for example a tab is coded as
.code \et
rather than
.codn @\et .
Octal and hex character escapes can be optionally
terminated by a semicolon, which is useful if the following characters are
octal or hex digits not intended to be part of the escape.
Only the above escapes are supported. Unlike in some other regular expression
implementations, if a backlash appears before a character which isn't a regex
special character or one of the supported escape sequences, it is an error.
This wasn't true of historic versions of \*(TX. See the Compatibility section.
.IP "Precedence table, highest to lowest:"
.TS
tab(!);
l l l.
Operators!Class!Associativity
\f[4](R) []\f[]!primary!
\f[4]R? R+ R* R%...\f[]!postfix!left-to-right
\f[4]R1R2\f[]!catenation!left-to-right
\f[4]~R ...%R\f[]\f[]\f[]!unary!right-to-left
\f[4]R1&R2\f[]!intersection!left-to-right
\f[4]R1|R2\f[]!union!left-to-right
.TE
.PP
The
.code %
operator is like a postfix operator with respect to its left
operand, but like a unary operator with respect to its right operand.
Thus
.code a~b%c~d
is
.cblk
a(~(b%(c(~d))))
.cble
, demonstrating right-to-left associativity,
where all of
.code b%
may be regarded as a unary operator being applied to
.codn c~d .
Similarly,
.code a?*+%b
means
.codn (((a?)*)+)%b ,
where the trailing
.code %b
behaves like a postfix operator.
In
\*(TX, regular expression matches do not span multiple lines. The regex
language has no feature for multi-line matching. However, the
.code @(freeform)
directive
allows the remaining portion of the input to be treated as one string
in which line terminators appear as explicit characters. Regular expressions
may freely match through this sequence.
It's possible for a regular expression to match an empty string.
For instance, if the next input character is
.codn z ,
facing a
the regular expression
.codn /a?/ ,
there is a zero-character match:
the regular expression's state machine can reach an acceptance
state without consuming any characters. Examples:
.IP code:
.cblk
\ @A@/a?/@/.*/
.cble
.IP data:
.cblk
\ zzzzz
.cble
.IP result:
.cblk
\ A=""
.cble
.PP
.IP code:
.cblk
\ @{A /a?/}@B
.cble
.IP data:
.cblk
\ zzzzz
.cble
.IP result:
.cblk
\ A="", B="zzzz"
.cble
.PP
.IP code:
.cblk
\ @*A@/a?/
.cble
.IP data:
.cblk
\ zzzzz
.cble
.IP result:
.cblk
\ A="zzzzz"
.cble
.PP
In the first example, variable
.code @A
is followed by a regular expression
which can match an empty string. The expression faces the letter
.code "z"
at position 0 in the data line. A zero-character match occurs there,
therefore the variable
.code A
takes on the empty string. The
.code @/.*/
regular expression then consumes the line.
Similarly, in the second example, the
.code /a?/
regular expression faces a
.codn "z" ,
and thus yields an empty string which is bound to
.codn A .
Variable
.code @B
consumes the entire line.
The third example requests the longest match for the variable binding.
Thus, a search takes place for the rightmost position where the
regular expression matches. The regular expression matches anywhere,
including the empty string after the last character, which is
the rightmost place. Thus variable
.code A
fetches the entire line.
For additional information about the advanced regular expression
operators, NOTES ON EXOTIC REGULAR EXPRESSIONS below.
.SS* Directives
The general syntax of a directive is:
.cblk
.mets >> @ expr
.cble
where
.meta expr
stands for a parenthesized list of subexpressions. A subexpression
is a symbol, number, string literal, character literal, quasiliteral, regular
expression, or a parenthesized expression. So, examples of syntactically valid
directives are:
.cblk
@(banana)
@(a b c (d e f))
@( a (b (c d) (e ) ))
@("apple" #\eb #\espace 3)
@(a #/[a-z]*/ b)
@(_ `@file.txt`)
.cble
A symbol has a slight more permissive lexical syntax than the
.meta bident
in the syntax
.cblk
.meti <> @{ bident }
.cble
introduced earlier. The
.code /
(slash) character may be part of an identifier, or even
constitute an entire identifier. In fact a symbol inside a
directive is a
.metn lident .
This is discussed in the Symbol Tokens section under TXR LISP.
A symbolmust not be a number; tokens that look like numbers are treated as
numbers and not symbols.
.SS* Character Literals
Character literals are introduced by the
.code #\e
syntax, which is either
followed by a character name, the letter
.code x
followed by hex digits,
the letter
.code o
followed by octal digits, or a single character. Valid character
names are:
.cblk
nul linefeed return
alarm newline esc
backspace vtab space
tab page pnul
.cble
For instance
.code #\eesc
denotes the escape character.
This convention for character literals is similar to that
of the Scheme language. Note that
.code #\elinefeed
and
.code #\enewline
are the same
character. The
.code #\epnul
character is specific to \*(TX and denotes the
.code U+DC00
code in Unicode; the name stands for "pseudo-null", which is related to
its special function. For more information about this, see the section
"Character Handling and International Characters".
.SS* String Literals
String literals are delimited by double quotes.
A double quote within a string literal is encoded using
.cblk
\e"
.cble
and a backslash is encoded as
.codn \e\e .
Backslash escapes like
.code \en
and
.code \et
are recognized, as are hexadecimal escapes like
.code \exFF
or
.code \exxabc
and octal
escapes like
.codn \e123 .
Ambiguity between an escape and subsequent
text can be resolved by using trailing semicolon delimiter:
.str "\exabc;d"
is a string consisting of the character
.code "U+0ABC"
followed by
.strn "d" .
The semicolon
delimiter disappears. To write a literal semicolon immediately after a hex
or octal escape, write two semicolons, the first of which will be interpreted
as a delimiter. Thus,
.str "\ex21;;"
represents
.strn "!;" .
If the line ends in the middle of a literal, it is an error, unless the
last character is a backslash. This backslash is a special escape which does
not denote a character; rather, it indicates that the string literal continues
on the next line. The backslash is deleted, along with whitespace which
immediately precedes it, as well as leading whitespace in the following line.
The escape sequence
.str "\e "
(backslash space) can be used to encode a significant space.
Example:
.cblk
"foo \e
bar"
"foo \e
\e bar"
"foo\e \e
bar"
.cble
The first string literal is the string
.strn "foobar" .
The second two are
.strn "foo bar" .
.SS* Word List Literals
A word list literal (WLL) provides a convenient way to write a list of strings
when such a list can be given as whitespace-delimited words.
There are two flavors of the WLL: the regular WLL which begins with
.cblk
#"
.cble
(hash, double-quote) and the splicing list literal which begins with
.cblk
#*"
.cble
(hash, star, double-quote).
Both types are terminated by a double quote, which may be escaped
as
.cblk
\e"
.cble
in order to include it as a character. All the escaping conventions
used in string literals can be used in word literals.
Unlike in string literals, whitespace (tabs, spaces and newlines) is not
significant in word literals: it separates words. Whitespace may be
escaped with a backslash in order to include it as a literal character.
Example:
.cblk
#"abc def ghi" --> notates ("abc" "def" "ghi")
#"abc def
ghi" --> notates ("abc" "def" "ghi")
#"abc\e def ghi" --> notates ("abc def" "ghi")
.cble
A splicing word literal differs from a word literal in that it does not
produce a list of string literals, but rather it produces a sequence of string
literals that is merged into the surrounding syntax. Thus, the following two
notations are equivalent:
.cblk
(1 2 3 #*"abc def" 4 5 #"abc def")
(1 2 3 "abc" "def" 4 5 ("abc" "def"))
.cble
The regular WLL produced a single list object, but the splicing
WLL expanded into multiple string literal objects.
.SS* String Quasiliterals
Quasiliterals are similar to string literals, except that they may
contain variable references denoted by the usual
.code @
syntax. The quasiliteral
represents a string formed by substituting the values of those variables
into the literal template. If
.code a
is bound to
.str "apple"
and
.code b
to
.strn "banana" ,
the quasiliteral
.code `one @a and two @{b}s`
represents the string
.strn "one apple and two bananas" .
A backquote escaped by a backslash represents
itself, and two consecutive
.code @
characters code for a literal
.codn @ .
There is no
.code \e@
escape. Quasiliterals support the full output variable
syntax. Expressions within variables substitutions follow the evaluation rules
of \*(TL when the quasiliteral occurs in \*(TL, and the rules of
the \*(TX pattern language when the quasiliteral occurs in the pattern language.
Quasiliterals can be split into multiple lines in the same way as ordinary
string literals.
.SS* Quasiword List Literals
The quasiword list literals (QLL-s) are to quasiliterals what WLL-s are to
ordinary literals. (See the above section Word List Literals.)
A QLL combines the convenience of the WLL
with the power of quasistrings.
Just as in the case of WLL-s, there are two flavors of the
QLL: the regular QLL which begins with
.code #`
\ (hash, backquote) and the splicing QLL which begins with
.code #*`
\ (hash, star, backquote).
Both types are terminated by a backquote, which may be escaped
as
.code \e`
\ in order to include it as a character. All the escaping conventions
used in quasiliterals can be used in QLL.
Unlike in quasiliterals, whitespace (tabs, spaces and newlines) is not
significant in QLL: it separates words. Whitespace may be
escaped with a backslash in order to include it as a literal character.
Note that the delimiting into words is done before the variable
substitution. If the variable a contains spaces, then
.code #`@a`
nevertheless
expands into a list of one item: the string derived from
.codn a .
Examples:
.cblk
#`abc @a ghi` --> notates (`abc` `@a` `ghi`)
#`abc @d@e@f
ghi` --> notates (`abc` `@d@e@f` `ghi`)
#`@a\e @b @c` --> notates (`@a @b` `@c`)
.cble
A splicing QLL differs from an ordinary QLL in that it does not produce a list
of quasiliterals, but rather it produces a sequence of quasiliterals that is
merged into the surrounding syntax.
.SS* Numbers
\*(TX supports integers and floating-point numbers.
An integer constant is made up of digits
.code 0
through
.codn 9 ,
optionally preceded by a
.code +
or
.code -
sign.
Examples:
.cblk
123
-34
+0
-0
+234483527304983792384729384723234
.cble
An integer constant can also be specified in hexadecimal using the prefix
.code #x
followed by an optional sign, followed by hexadecimal digits:
.code 0
through
.code 9
and the upper or lower case letters
.code A
through
.codn F :
.cblk
#xFF ;; 255
#x-ABC ;; -2748
.cble
Similarly, octal numbers are supported with the prefix
.code #o
followed by octal digits:
.cblk
#o777 ;; 511
.cble
and binary numbers can be written with a
.code #b
prefix:
.cblk
#b1110 ;; 14
.cble
A floating-point constant is marked by the inclusion of a decimal point, the
exponential "e notation", or both. It is an optional sign, followed
by a mantissa consisting of digits, a decimal point, more digits, and then an
optional exponential notation consisting of the letter
.code "e"
or
.codn "E" ,
an optional
.code +
or
.code -
sign, and then digits indicating the exponent value.
In the mantissa, the digits are not optional. At least one digit must either
precede the decimal point or follow. That is to say, a decimal point by itself
is not a floating-point constant.
Examples:
.cblk
.123
123.
1E-3
20E40
.9E1
9.E19
-.5
+3E+3
.cble
Examples which are not floating-point constant tokens:
.cblk
. ;; consing dot
123E ;; the symbol 123E
1.0E- ;; syntax error: invalid floating point constant
1.0E ;; syntax error: invalid floating point constant
1.E ;; 1; consing dot; symbol E
.e ;; consing dot followed by symbol e
.cble
In \*(TX there is a special "dotdot" token consisting of two consecutive periods.
An integer constant followed immediately by dotdot is recognized as such; it is
not treated as a floating constant followed by a dot. That is to say,
.code 123..
does not mean
.code 123. .
(floating point
.code 123.0
value followed by dot token). It means
.code 123 ..
(integer
.code 123
followed by
.code ..
token).
Dialect note: unlike in Common Lisp,
.code 123.
is not an integer, but the floating-point number
.codn 123.0 .
.SS* Comments
Comments of the form
.code @;
were already covered. Inside directives,
comments are introduced just by a
.code ;
character.
Example:
.cblk
@(foo ; this is a comment
bar ; this is another comment
)
.cble
This is equivalent to
.codn "@(foo bar)" .
.SS* Directives-driven Syntax
Some directives not only denote an expression, but are also involved in
surrounding syntax. For instance, the directive
.cblk
@(collect)
.cble
not only denotes an expression, but it also introduces a syntactic phrase which
requires a matching
.code @(end)
directive. In other words,
.code @(collect)
is not only
an expression, but serves as a kind of token in a higher level phrase structure
grammar.
Usually if this type of "syntactic directive" occurs alone in a line, not
preceded or followed by other material, it is involved in a "vertical" (or line
oriented) syntax.
If such a directive is embedded in a line (has preceding or trailing material)
then it is in a horizontal syntactic and semantic context (character-oriented).
There is an exception: the definition of a horizontal function looks like this:
.cblk
@(define name (arg))body material@(end)
.cble
Yet, this is considered one vertical item, which means that it does not match
a line of data. (This is necessary because all horizontal syntax matches
something within a line of data, which is undesirable for definitions.)
Many directives exhibit both horizontal and vertical syntax, with different but
closely related semantics. A few are vertical only, and some are
horizontal only.
A summary of the available directives follows:
.coIP @(eof)
Explicitly match the end of file. Fails if unmatched data remains in
the input stream.
.coIP @(eol)
Explicitly match the end of line. Fails if the current position is not the
end of a line. Also fails if no data remains (there is no current line).
.coIP @(next)
Continue matching in another file or other data source.
.coIP @(block)
Groups together a sequence of directives into a logical name block,
which can be explicitly terminated from within using
the
.code @(accept)
and
.code @(fail)
directives.
Blocks are discussed in the section BLOCKS below.
.coIP @(skip)
Treat the remaining query as a subquery unit, and search the lines (or
characters) of the input file until that subquery matches somewhere. A skip is
also an anonymous block.
.coIP @(trailer)
Treat the remaining query or subquery as a match for a trailing context. That
is to say, if the remainder matches, the data position is not advanced.
.coIP @(freeform)
Treat the remainder of the input as one big string, and apply the following
query line to that string. The newline characters (or custom separators) appear
explicitly in that string.
.coIP @(fuzz)
The fuzz directive, inspired by the patch utility, specifies a partial
match for some lines.
.ccIP @ @(line) and @ @(chr)
These directives match a variable or expression against the current line
number or character position.
.coIP @(some)
Multiple clauses are each applied to the same input. Succeeds if at least one
of the clauses matches the input. The bindings established by earlier
successful clauses are visible to the later clauses.
.coIP @(all)
Multiple clauses are applied to the same input. Succeeds if and only if each
one of the clauses matches. The clauses are applied in sequence, and evaluation
stops on the first failure. The bindings established by earlier successful
clauses are visible to the later clauses.
.coIP @(none)
Multiple clauses are applied to the same input. Succeeds if and only if none of
them match. The clauses are applied in sequence, and evaluation stops on the
first success. No bindings are ever produced by this construct.
.coIP @(maybe)
Multiple clauses are applied to the same input. No failure occurs if none of
them match. The bindings established by earlier successful clauses are visible
to the later clauses.
.coIP @(cases)
Multiple clauses are applied to the same input. Evaluation stops on the
first successful clause.
.coIP @(require)
The require directive is similar to the do directive: it evaluates one or more
\*(TL expressions. If the result of the rightmost expression is nil,
then require triggers a match failure. See the TXR LISP section far below.
.ccIP @, @(if) @, @(elif) and @ @(else)
The
.code if
directive with optional
.code elif
and
.code else
clauses is a syntactic sugar
which translates to a combination of
.code @(cases)
and
.codn @(require) .
.coIP @(choose)
Multiple clauses are applied to the same input. The one whose effect persists
is the one which maximizes or minimizes the length of a particular variable.
.coIP @(empty)
The
.code @(empty)
directive matches the empty string. It is useful in certain
situations, such as expressing an empty match in a directive that doesn't
accept an empty clause. The
.code @(empty)
syntax has another meaning in
.code @(output)
clauses, in conjunction with
.codn @(repeat) .
.meIP @(define < name >> ( args ...))
Introduces a function. Functions are discussed in the FUNCTIONS section below.
.coIP @(gather)
Searches text for matches for multiple clauses which may occur in arbitrary
order. For convenience, lines of the first clause are treated as separate
clauses.
.coIP @(collect)
Search the data for multiple matches of a clause. Collect the
bindings in the clause into lists, which are output as array variables.
The
.code @(collect)
directive is line oriented. It works with a multi-line
pattern and scans line by line. A similar directive called
.code @(coll)
works within one line.
A collect is an anonymous block.
.coIP @(and)
Separator of clauses for
.codn @(some) ,
.codn @(all) ,
.codn @(none) ,
.code @(maybe)
and
.codn @(cases) .
Equivalent to
.codn @(or) .
The choice is stylistic.
.coIP @(or)
Separator of clauses for
.codn @(some) ,
.codn @(all) ,
.codn @(none) ,
.code @(maybe)
and
.codn @(cases) .
Equivalent to
.codn @(and) .
The choice is stylistic.
.coIP @(end)
Required terminator for
.codn @(some) ,
.codn @(all) ,
.codn @(none) ,
.codn @(maybe) ,
.codn @(cases) ,
.codn @(if) ,
.codn @(collect) ,
.codn @(coll) ,
.codn @(output) ,
.codn @(repeat) ,
.codn @(rep) ,
.codn @(try) ,
.code @(block)
and
.codn @(define) .
.coIP @(fail)
Terminate the processing of a block, as if it were a failed match.
Blocks are discussed in the section BLOCKS below.
.coIP @(accept)
Terminate the processing of a block, as if it were a successful match.
What bindings emerge may depend on the kind of block: collect
has special semantics. Blocks are discussed in the section BLOCKS below.
.coIP @(try)
Indicates the start of a try block, which is related to exception
handling, discussed in the EXCEPTIONS section below.
.ccIP @ @(catch) and @ @(finally)
Special clauses within
.codn @(try) .
See EXCEPTIONS below.
.ccIP @ @(defex) and @ @(throw)
Define custom exception types; throw an exception. See EXCEPTIONS below.
.coIP @(assert)
The assert directive requires the following material to match, otherwise
it throws an exception. It is useful for catching mistakes or omissions
in parts of a query that are sure-fire matches.
.coIP @(flatten)
Normalizes a set of specified variables to one-dimensional lists. Those
variables which have scalar value are reduced to lists of that value.
Those which are lists of lists (to an arbitrary level of nesting) are converted
to flat lists of their leaf values.
.coIP @(merge)
Binds a new variable which is the result of merging two or more
other variables. Merging has somewhat complicated semantics.
.coIP @(cat)
Decimates a list (any number of dimensions) to a string, by catenating its
constituent strings, with an optional separator string between all of the
values.
.coIP @(bind)
Binds one or more variables against a value using a structural
pattern match. A limited form of unification takes place which can cause a
match to fail.
.coIP @(set)
Destructively assigns one or more existing variables using a structural
pattern, using syntax similar to bind. Assignment to unbound
variables triggers an error.
.coIP @(rebind)
Evaluates an expression in the current binding environment, and
then creates new bindings for the variables in the structural pattern.
Useful for temporarily overriding variable values in a scope.
.coIP @(forget)
Removes variable bindings.
.coIP @(local)
Synonym of
.codn @(forget) .
.coIP @(output)
A directive which encloses an output clause in the query. An output section
does not match text, but produces text. The directives above are not
understood in an output clause.
.coIP @(repeat)
A directive understood within an @(output) section, for repeating multi-line
text, with successive substitutions pulled from lists. The directive
.code @(rep)
produces iteration over lists horizontally within one line.
.coIP @(deffilter)
This directive is used for defining named filters, which are useful
for filtering variable substitutions in output blocks. Filters are useful
when data must be translated between different representations that
have different special characters or other syntax, requiring escaping
or similar treatment. Note that it is also possible to use a function
as a filter. See Function Filters below.
.coIP @(filter)
The filter directive passes one or more variables through a given
filter or chain or filters, updating them with the filtered values.
.coSS @ @(load) and @ @(include)
These directives allow \*(TX programs to be modularized. They bring in
code from a file, in two different ways.
.coIP @(do)
The do directive is used to evaluate \*(TL expressions, discarding their
result values. See the TXR LISP section far below.
.PP
.SH* INPUT SCANNING AND DATA MANIPULATION
.dir next
The
.code next
directive indicates that the remaining directives in the current block
are to be applied against a new input source.
It can only occur by itself as the only element in a query line,
and takes various arguments, according to these possibilities:
.cblk
.mets @(next)
.mets @(next << source )
.mets @(next < source :nothrow)
.mets @(next :args)
.mets @(next :env)
.mets @(next :list << expr )
.mets @(next :string << expr )
.cble
The lone
.code @(next)
without arguments specifies that subsequent directives
will match inside the next file in the argument list which was passed
to \*(TX on the command line.
If
.meta source
is given, it must be string-valued expression which denotes an
input source; it may be a string literal, quasiliteral or a string-valued
variable. For instance, if variable
.code A
contains the text
.strn "data" ,
then
.code @(next A)
means switch to the file called
.strn "data" ,
and
.code @(next `@A.txt`)
means to switch to the file
.strn "data.txt" .
If the input source cannot be opened for whatever reason,
\*(TX throws an exception (see EXCEPTIONS below). An unhandled exception will
terminate the program. Often, such a drastic measure is inconvenient;
if
.code @(next)
is invoked with the
.code :nothrow
keyword, then if the input
source cannot be opened, the situation is treated as a simple
match failure.
The variant
.code @(next :args)
means that the remaining command line arguments are to
be treated as a data source. For this purpose, each argument is considered to
be a line of text. The argument list does include that argument which specifies
the file that is currently being processed or was most recently processed.
As the arguments are matched, they are consumed. This means that if a
.code @(next)
directive without
arguments is executed in the scope of
.codn "@(next :args)" ,
it opens the file named
by the first unconsumed argument.
To process arguments, and then continue with the original file and argument
list, wrap the argument processing in a
.codn @(block) .
When the block terminates, the input source and argument list are restored
to what they were before the block.
The variant
.code @(next :env)
means that the list of process environment variables
is treated as a source of data. It looks like a text file stream
consisting of lines of the form
.strn "name=value" .
If this feature is not available
on a given platform, an exception is thrown.
The syntax
.cblk
.meti @(next :list << expr )
.cble
treats expression
.meta expr
as a source of
text. The value of
.meta expr
is flattened to a simple list in a way similar to the
.code @(flatten)
directive. The resulting list is treated as if it were the
lines of a text file: each element of the list must be a string,
which represents a line. If the strings happen contain embedded newline
characters, they are a visible constituent of the line, and do not act as line
separators.
The syntax
.cblk
.meti @(next :string << expr )
.cble
treats expression
.meta expr
as a source of text. The value of the expression must be a string. Newlines in
the string are interpreted as line terminators.
A string which is not terminated by a newline is tolerated, so that:
.cblk
@(next :string "abc")
@a
.cble
binds
.code a
to
.strn "abc" .
Likewise, this is also the case with input files and other
streams whose last line is not terminated by a newline.
However, watch out for empty strings, which are analogous to a correctly formed
empty file which contains no lines:
.cblk
@(next :string "")
@a
.cble
This will not bind
.code a
to
.strn "" ;
it is a matching failure. The behavior of
.code :list
is
different. The query
.cblk
@(next :list "")
@a
.cble
binds
.code a
to
.strn "" .
The reason is that under
.code :list
the string
.str ""
is flattened to
the list
.cblk
("")
.cble
which is not an empty input stream, but a stream consisting of
one empty line.
Note that the
.code @(next)
directive only redirect the source of input over the scope of subquery in which
the next directive appears, not necessarily all remaining directives. For
example, the following query looks for the line starting with
.str "xyz"
at the top of the file
.strn "foo.txt" ,
within a some
directive. After the
.code @(end)
which terminates the
.codn @(some) ,
the
.str "abc"
is matched in the previous input stream which was in effect before
the
.code
@(next)
directive:
.cblk
@(some)
@(next "foo.txt")
xyz@suffix
@(end)
abc
.cble
However, if the
.code @(some)
subquery successfully matched
.str "xyz@suffix"
within the
file
.codn foo.text ,
there is now a binding for the
.code suffix
variable, which
is visible to the remainder of the entire query. The variable bindings
survive beyond the clause, but the data stream does not.
The
.code @(next)
directive supports the file name conventions as the command
line. The name
.code -
means standard input. Text which starts with a
.code !
is
interpreted as a shell command whose output is read like a file. These
interpretations are applied after variable substitution. If the file is
specified as
.codn @a ,
but the variable a expands to
.strn "!echo foo" ,
then the output of
the
.str "echo foo"
command will be processed.
.dir skip
The
.code skip
directive considers the remainder of the query as a search
pattern. The remainder is no longer required to strictly match at the
current line in the current input stream. Rather, the current stream is searched,
starting with the current line, for the first line where the entire remainder
of the query will successfully match. If no such line is found, the
.code skip
directive fails. If a matching position is found, the remainder of
the query is processed from that point.
Of course, the remainder of the query can itself contain skip directives.
Each such directive performs a recursive subsearch.
Skip comes in vertical and horizontal flavors. For instance, skip and match the
last line:
.cblk
@(skip)
@last
@(eof)
.cble
Skip and match the last character of the line:
.cblk
@(skip)@{last 1}@(eol)
.cble
The skip directive has two optional arguments. If the first argument is a
number, its value limits the range of lines scanned for a match. Judicious use
of this feature can improve the performance of queries.
Example: scan until
.str "size: @SIZE"
matches, which must happen within
the next 15 lines:
.cblk
@(skip 15)
size: @SIZE
.cble
Without the range limitation skip will keep searching until it consumes
the entire input source. In a horizontal
.codn skip ,
the range-limiting numeric argument is expressed in characters, so that
.cblk
abc@(skip 5)def
.cble
means: there must be a match for
.str "abc"
at the start of the line, and then within the next five characters,
there must be a match for
.strn "def" .
Sometimes a skip is nested within a
.codn collect ,
or
following another skip. For instance, consider:
.cblk
@(collect)
begin @BEG_SYMBOL
@(skip)
end @BEG_SYMBOL
@(end)
.cble
The above
.code collect
iterates over the entire input. But, potentially, so does
the embedded
.codn skip .
Suppose that
.str "begin x"
is matched, but the data has no
matching
.strn "end x" .
The skip will search in vain all the way to the end of the
data, and then the collect will try another iteration back at the
beginning, just one line down from the original starting point. If it is a
reasonable expectation that an
.code "end x"
occurs 15 lines of a
.strn "begin x" ,
this can be specified instead:
.cblk
@(collect)
begin @BEG_SYMBOL
@(skip 15)
end @BEG_SYMBOL
@(end)
.cble
If the symbol
.code nil
is used in place of a number, it means to scan
an unlimited range of lines; thus,
.code @(skip nil)
is equivalent to
.codn @(skip) .
If the symbol
.code :greedy
is used, it changes the semantics of the skip
to longest match semantics. For instance, match the last three space-separated
tokens of the line:
.cblk
@(skip :greedy) @a @b @c
.cble
Without
.codn :greedy ,
the variable
.code @c
will can match multiple tokens,
and end up with spaces in it, because nothing follows
.code @c
and so it matches from any position which follows a space to the
end of the line. Also note the space in front of
.codn @a .
Without this
space,
.code @a
will get an empty string.
A line oriented example of greedy skip: match the last line without
using
.codn @eof :
.cblk
@(skip :greedy)
@last_line
.cble
There may be a second numeric argument. This specifies a minimum
number of lines to skip before looking for a match. For instance,
skip 15 lines and then search indefinitely for
.codn "begin ..." :
.cblk
@(skip nil 15)
begin @BEG_SYMBOL
.cble
The two arguments may be used together. For instance, the following
matches if, and only if, the 15th line of input starts with
.codn "begin " :
.cblk
@(skip 1 15)
begin @BEG_SYMBOL
.cble
Essentially,
.cblk
.meti @(skip 1 << n )
.cble
means "hard skip by
.meta n
lines".
.code @(skip 1 0)
is the same as
.codn "@(skip 1)" ,
which is a noop, because it means: "the remainder of the query must match
starting on the very next line", or, more briefly, "skip exactly zero lines",
which is the behavior if the skip directive is omitted altogether.
Here is one trick for grabbing the fourth line from the bottom of the input:
.cblk
@(skip)
@fourth_from_bottom
@(skip 1 3)
@(eof)
.cble
Or using greedy skip:
.cblk
@(skip :greedy)
@fourth_from_bottom
@(skip 1 3)
.cble
Nongreedy skip with the
.code @(eof)
has a slight advantage because the greedy skip
will keep scanning even though it has found the correct match, then backtrack
to the last good match once it runs out of data. The regular skip with explicit
.code @(eof)
will stop when the
.code @(eof)
matches.
.SS* Reducing Backtracking with Blocks
.code skip
can consume considerable CPU time when multiple skips are nested. Consider:
.cblk
@(skip)
A
@(skip)
B
@(skip)
C
.cble
This is actually nesting: the second a third skips occur within the body of the
first one, and thus this creates nested iteration. \*(TX is searching for the
combination of skips which find match the pattern of lines
.codn A ,
.code B
and
.codn C ,
with
backtracking behavior. The outermost skip marches through the data until it
finds
.codn A ,
followed by a pattern match for the second skip. The second skip
iterates within to find
.codn B ,
followed by the third skip, and the third skip
iterates to find
.codn C .
If there is only one line
.codn A ,
and one
.codn B ,
then this is reasonably fast. But suppose there are many lines matching
.code A
and
.codn B ,
giving rise to a large number combinations of skips which match
.code A
and
.codn B ,
and yet do not find a match for
.codn C ,
triggering backtracking. The nested stepping which tries
the combinations of
.code A
and
.code B
can give rise to a considerable running time.
One way to deal with the problem is to unravel the nesting with the help of
blocks. For example:
.cblk
@(block)
@ (skip)
A
@(end)
@(block)
@ (skip)
B
@(end)
@(skip)
C
.cble
Now the scope of each skip is just the remainder of the block in which it
occurs. The first skip finds
.codn A ,
and then the block ends. Control passes to the
next block, and backtracking will not take place to a block which completed
(unless all these blocks are enclosed in some larger construct which
backtracks, causing the blocks to be re-executed.
Of course, this rewrite is not equivalent, and cannot be used for instance
in backreferencing situations such as:
.cblk
@;
@; Find three lines anywhere in the input which are identical.
@;
@(skip)
@line
@(skip)
@line
@(skip)
@line
.cble
This example depends on the nested search-within-search semantics.
.dir trailer
The
.code trailer
directive introduces a trailing portion of a query or subquery
which matches input material normally, but in the event of a successful match,
does not advance the current position. This can be used, for instance, to
cause
.code @(collect)
to match partially overlapping regions.
Example:
.cblk
@(collect)
@line
@(trailer)
@(skip)
@line
@(end)
.cble
This script collects each line which has a duplicate somewhere later
in the input. Without the
.code @(trailer)
directive, this does not work properly
for inputs like:
.cblk
111
222
111
222
.cble
Without
.codn @(trailer) ,
the first duplicate pair constitutes a match which
spans over the
.codn 222 .
After that pair is found, the matching continues
after the second
.codn 111 .
With the
.code @(trailer)
directive in place, the collect body, on each
iteration, only consumes the lines matched prior to
.codn @(trailer) .
.dir freeform
The
.code freeform
directive provides a useful alternative to
\*(TX's line-oriented matching discipline. The freeform directive treats all
remaining input from the current input source as one big line. The query line
which immediately follows freeform is applied to that line.
The syntax variations are:
.cblk
@(freeform)
... query line ..
.mets @(freeform << number )
... query line ..
.mets @(freeform << string )
... query line ..
.mets @(freeform < number << string )
... query line ..
.cble
If
.meta number
and
.meta string
are both present, they may be given in either order.
If a numeric argument is given, it limits the range of lines which are combined
together. For instance
.code @(freeform 5)
means to only consider the next five lines
to to be one big line. Without a numeric argument, freeform is "bottomless". It
can match the entire file, which creates the risk of allocating a large amount
of memory.
If a string argument is given, it specifies a custom line terminator. The
default terminator is
.strn "\en" .
The terminator does not have to be one character long.
Freeform does not convert the entire remainder of the input into one big line
all at once, but does so in a dynamic, lazy fashion, which takes place as the
data is accessed. So at any time, only some prefix of the data exists as a flat
line in which newlines are replaced by the terminator string, and the remainder
of the data still remains as a list of lines.
After the subquery is applied to the virtual line, the unmatched remainder
of that line is broken up into multiple lines again, by looking for and
removing all occurrences of the terminator string within the flattened portion.
Care must be taken if the terminator is other than the default
.strn "\en" .
All occurrences of the terminator string are treated as line terminators in
the flattened portion of the data, so extra line breaks may be introduced.
Likewise, in the yet unflattened portion, no breaking takes place, even if
the text contains occurrences of the terminator string. The extent of data which
is flattened, and the amount of it which remains, depends entirely on the
query line underneath
.codn @(flatten) .
In the following example, lines of data are flattened using $ as the line
terminator.
.IP code:
.cblk
\ @(freeform "$")
@a$@b:
@c
@d
.cble
.IP data:
.cblk
\ 1
2:3
4
.cble
.IP "output (\f[4]-B\f[]):"
.cblk
\ a="1"
b="2"
c="3"
d="4"
.cble
.PP
The data is turned into the virtual line
.codn 1$2:3$4$ .
The
.code @a$@b:
subquery matches
the
.code 1$2:
portion, binding
.code a
to
.strn 1 ,
and
.code b
to
.strn 2 .
The remaining portion
.code 3$4$
is then split into separate lines again according to the line terminator
.codn $i :
.cblk
3
4
.cble
Thus the remainder of the query
.cblk
@c
@d
.cble
faces these lines, binding
.code c
to
.code 3
and
.code d
to
.codn 4 .
Note that since the data
does not contain dollar signs, there is no ambiguity; the meaning may be
understood in terms of the entire data being flattened and split again.
In the following example,
.code freeform
is used to solve a tokenizing problem. The
Unix password file has fields separated by colons. Some fields may be empty.
Using freeform, we can join the password file using
.str ":"
as a terminator.
By restricting freeform to one line, we can obtain each line of the password
file with a terminating
.strn ":" ,
allowing for a simple tokenization, because
now the fields are colon-terminated rather than colon-separated.
Example:
.cblk
@(next "/etc/passwd")
@(collect)
@(freeform 1 ":")
@(coll)@{token /[^:]*/}:@(end)
@(end)
.cble
.dir fuzz
The
.code fuzz
directive allows for an imperfect match spanning a set number of
lines. It takes two arguments, both expressions that should evaluate
to integers:
.cblk
@(fuzz m n)
...
.cble
This expresses that over the next n query lines, the matching strictness
is relaxed a little bit. Only m out of those n lines have to match.
Afterward, the rest of the query follows normal, strict processing.
In the degenerate situation that there are fewer than n query lines following
the
.code fuzz
directive, then m of them must succeed nevertheless. (If there
are fewer than m, then this is impossible.)
.dirs line chr
The
.code line
and
.code chr
directives perform binding between the current input line number or character
position within a line, against an expression or variable:
.cblk
@(line 42)
@(line x)
abc@(chr 3)def@(chr y)
.cble
The directive
.code @(line 42)
means "match the current input line number against the integer 42". If
the current line is 42, then the directive matches, otherwise it fails.
.code line
is a vertical directive which doesn't consume a line of input. Thus,
the following matches at the beginning of an input stream, and
.code x
ends up bound to the first line of input:
.cblk
@(line 1)
@(line 1)
@(line 1)
@x
.cble
The directive
.code @(line x)
binds variable
.code x
to the current input line number, if
.code x
is an unbound variable. If
.code x
is already bound, then the value of
.code x
must match the current line number, otherwise the directive fails.
The
.code chr
directive is similar to
.code line
except that it's a horizontal directive, and matches the character position
rather than the line position. Character positions are measured from zero,
rather than one.
.code chr
does not consume a character. Hence the two occurrences of
.code chr
in the following example both match, and
.code x
takes the entire line of input:
.cblk
@(chr 0)@(chr 0)@x
.cble
The argument of
.code line
or
.code chr
may be a
.codn @ -delimited
Lisp expression. This is useful for matching computed lines or
character positions:
.cblk
@(line @(+ a (* b c)))
.cble
.dirs some all none maybe cases choose
These directives, called the parallel directives, combine multiple subqueries,
which are applied at the same input position, rather than to consecutive input.
They come in vertical (line mode) and horizontal (character mode) flavors.
In horizontal mode, the current position is understood to be a character
position in the line being processed. The clauses advance this character
position by moving it to the right. In vertical mode, the current position is
understood to be a line of text within the stream. A clause advances the
position by some whole number of lines.
The syntax of these parallel directives follows this example:
.cblk
@(some)
subquery1
.
.
.
@(and)
subquery2
.
.
.
@(and)
subquery3
.
.
.
@(end)
.cble
And in horizontal mode:
.cblk
@(some)subquery1...@(and)subquery2...@(and)subquery3...@(end)
.cble
Long horizontal lines can be broken up with line continuations, allowing the
above example to be written like this, which is considered a single logical
line:
.cblk
@(some)@\e
subquery1...@\e
@(and)@\e
subquery2...@\e
@(and)@\e
subquery3...@\e
@(end)
.cble
The
.codn @(some) ,
.codn @(all) ,
.codn @(none) ,
.codn @(maybe) ,
.code @(cases)
or
.code @(choose)
must be followed
by at least one subquery clause, and be terminated by
.codn @(end) .
If there are two
or more subqueries, these additional clauses are indicated by
.code @(and)
or
.codn @(or) ,
which are interchangeable. The separator and terminator directives also must
appear as the only element in a query line.
The
.code choose
directive requires keyword arguments. See below.
The syntax supports arbitrary nesting. For example:
.cblk
QUERY: SYNTAX TREE:
@(all) all -+
@ (skip) +- skip -+
@ (some) | +- some -+
it | | +- TEXT
@ (and) | | +- and
@ (none) | | +- none -+
was | | | +- TEXT
@ (end) | | | +- end
@ (end) | | +- end
a dark | +- TEXT
@(end) *- end
.cble
nesting can be indicated using whitespace between
.code @
and the
directive expression. Thus, the above is an
.code @(all)
query containing a
.code @(skip)
clause which applies to a
.code @(some)
that is followed by the text line
.strn "a dark" .
The
.code @(some)
clause combines the text line
.strn it ,
and a
.code @(none)
clause which contains just one clause consisting of the line
.strn was .
The semantics of the parallel directives is:
.coIP @(all)
Each of the clauses is matched at the current position. If any of the
clauses fails to match, the directive fails (and thus does not produce
any variable bindings). Clauses following the failed directive are not
evaluated. Bindings extracted by a successful clause are visible to the clauses
which follow, and if the directive succeeds, all of the combined bindings
emerge.
.meIP @(some [ :resolve >> ( var ...) ])
Each of the clauses is matched at the current position. If any of the clauses
succeed, the directive succeeds, retaining the bindings accumulated by the
successfully matching clauses. Evaluation does not stop on the first successful
clause. Bindings extracted by a successful clause are visible to the clauses
which follow.
The
.code :resolve
parameter is for situations when the
.code @(some)
directive has
multiple clauses that need to bind some common variables to different
values: for instance, output parameters in functions. Resolve takes
a list of variable name symbols as an argument. This is called the
resolve set. If the clauses of
.code @(some)
bind variables in the resolve
set, those bindings are not visible to later clauses. However, those
bindings do emerge out of the
.code @(some)
directive as a whole.
This creates a conflict: what if two or more clauses introduce
different bindings for a variable in the resolve set?
This is why it is called the resolve set: conflicts for variables in the
resolve set are automatically resolved in favor of later directives.
Example:
.cblk
@(some :resolve (x))
@ (bind a "a")
@ (bind x "x1")
@(or)
@ (bind b "b")
@ (bind x "x2")
@(end)
.cble
Here, the two clauses both introduce a binding for
.codn x .
Without the
.code :resolve
parameter, this would mean that the second clause fails, because
.code x
comes in
with the value
.strn x1 ,
which does not bind with
.strn x2 .
But because
.code x
is placed
into the resolve set, the second clause does not see the
.str x1
binding. Both
clauses establish their bindings independently creating a conflict over
.codn x .
The conflict is resolved in favor of the second clause, and so the bindings
which emerge from the directive are:
.cblk
a="a"
b="b"
x="x2"
.cble
.coIP @(none)
Each of the clauses is matched at the current position. The
directive succeeds only if all of the clauses fail. If
any clause succeeds, the directive fails, and subsequent clauses are not
evaluated. Thus, this directive never produces variable bindings, only matching
success or failure.
.coIP @(maybe)
Each of the clauses is matched at the current position. The directive always
succeeds, even if all of the clauses fail. Whatever bindings are found in any
of the clauses are retained. Bindings extracted by any successful clause are
visible to the clauses which follow.
.coIP @(cases)
Each of the clauses is matched at the current position.
The clauses are matched, in order, at the current position.
If any clause matches, the matching stops and the bindings
collected from that clause are retained. Any remaining clauses
after that one are not processed. If no clause matches, the
directive fails, and produces no bindings.
.meIP @(choose [ :longest < var | :shortest < var ])
Each of the clauses is matched at the current position in order. In this
construct, bindings established by an earlier clause are not visible to later
clauses. Although any or all of the clauses can potentially match, the clause
which succeeds is the one which maximizes or minimizes the length of the
text bound to the specified variable. The other clauses have no effect.
For all of the parallel directives other than
.code @(none)
and
.codn @(choose) ,
the query
advances the input position by the greatest number of lines that match in any
of the successfully matching subclauses that are evaluated. The
.code @(none)
directive does not advance the input position.
For instance if there are two subclauses, and one of them matches three lines,
but the other one matches five lines, then the overall clause is considered to
have made a five line match at its position. If more directives follow, they
begin matching five lines down from that position.
.dir require
The syntax of
.code @(require)
is:
.cblk
.mets @(require << lisp-expression )
.cble
The require directive evaluates a \*(TL expression. (See TXR LISP far
below.) If the expression yields a true value, then it succeeds, and matching
continues with the directives which follow. Otherwise the directive fails.
In the context of the
.code require
directive, the expression should not be introduced by the
.code @
symbol; it is expected to be a Lisp expression.
Example:
.cblk
@; require that 4 is greater than 3
@; This succeeds; therefore, @a is processed
@(require (> (+ 2 2) 3))
@a
.cble
.dir if
The syntax of the
.code if
directive can be exemplified as follows:
.cblk
.mets @(if << lisp-expr )
.
.
.
.mets @(elif << lisp-expr )
.
.
.
.mets @(elif << lisp-expr )
.
.
.
@(else)
.
.
.
@(end)
.cble
The
.code @(elif)
and
.code @(else)
clauses are all optional. If
.code @(else)
is present, it must be
last, before
.codn @(end) ,
after any
.code @(elif)
clauses. Any of the clauses may be empty.
See the \*(TL section about \*(TL expressions. In this directive,
\*(TL expressions are not introduced by the
.code @
symbol, just like in the
.code require
directive.
For example:
.cblk
@(if (> (length str) 42))
foo: @a @b
@(else)
{@c}
@(end)
.cble
In this example, if the length of the variable
.code str
is greater than
.codn 42 ,
then matching continues with
.strn "foo: @a b" ,
otherwise it proceeds with
.codn {@c} .
The
.code if
directive is actually a syntactic sugar which is translated to
.code @(cases)
and
.codn @(require) .
That is to say, the following pattern:
.cblk
@(cases)
.mets @(require << lisp-expr-1 )
A
@(or)
.mets @(require << lisp-expr-2 )
B
@(or)
C
@(end)
.cble
corresponds to the somewhat shorter and clearer:
.cblk
.mets @(if << lisp-expr-1 )
A
.mets @(elsif << lisp-expr-2 )
B
@(else)
C
@(end)
.cble
.dir gather
Sometimes text is structured as items that can appear in an arbitrary order.
When multiple matches need to be extracted, there is a combinatorial explosion
of possible orders, making it impractical to write pattern matches for all
the possible orders.
The
.code gather
directive is for these situations. It specifies multiple clauses
which all have to match somewhere in the data, but in any order.
For further convenience, the lines of the first clause of the
.code gather
directive
are implicitly treated as separate clauses.
The syntax follows this pattern
.cblk
@(gather)
one-line-query1
one-line-query2
.
.
.
one-line-queryN
@(and)
multi
line
query1
.
.
.
@(and)
multi
line
query2
.
.
.
@(end)
.cble
Of course the multi-line clauses are optional. The
.code gather
directive takes
keyword parameters, see below.
.coNP The @ until / @ last clause in @ gather
Similarly to
.codn collect ,
.code gather
has an optional
.cod3 until / last
clause:
.cblk
@(gather)
...
@(until)
...
@(end)
.cble
How gather works is that the text is searched for matches for the single line
and multi-line queries. The clauses are applied in the order in which they appear.
Whenever one of the clauses matches, any bindings it produces are retained and
it is removed from further consideration. Multiple clauses can match at the
same text position. The position advances by the longest match from among the
clauses which matched. If no clauses match, the position advances by one line.
The search stops when all clauses are eliminated, and then the cumulative
bindings are produced. If the data runs out, but unmatched clauses remain, the
directive fails.
Example: extract several environment variables, which do not appear in a particular
order:
.cblk
@(next :env)
@(gather)
USER=@USER
HOME=@HOME
SHELL=@SHELL
@(end)
.cble
If the until or last clause is present and a match occurs, then the matches
from the other clauses are discarded and the gather terminates. The difference
between
.cod3 until / last
is that any bindings bindings established in last are
retained, and the input position is advanced past the matched material.
The
.cod3 until / last
clause has visibility to bindings established in the
previous clauses in that same iteration, even though those bindings
end up thrown away.
.coNP Keyword parameters in @ gather
The
.code gather
directive accepts the keyword parameter
.codn :vars .
The argument to vars is
a list of required and optional variables. Optional variables are denoted by
the specification of a default value. Example:
.cblk
@(gather :vars (a b c (d "foo")))
...
@(end)
.cble
Here,
.codn a ,
.codn b ,
.code c
and
.code e
are required variables, and
.code d
is optional. Variable
.code e
is
required because its default value is the empty list
.codn () ,
same as the symbol
.codn nil .
The presence of
.code :vars
changes the behavior in three ways.
Firstly, even if all the clauses in the gather match successfully and are
eliminated, the directive will fail if the required variables do not have
bindings. It doesn't matter whether the bindings are existing, or whether they
are established by the gather.
Secondly, if some of the clauses of the gather did not match, but all
of the required variables have bindings, then the directive succeeds.
Without the presence of
.codn :vars ,
it would fail in this situation.
Thirdly, if
.code gather
succeeds (all required variables have bindings),
then all of the optional variables which do not have bindings are given
bindings to their default values.
.dir collect
The syntax of the
.code collect
directive is:
.cblk
@(collect)
... lines of subquery
@(end)
.cble
or with an until or last clause:
.cblk
@(collect)
... lines of subquery: main clause
@(until)
... lines of subquery: until clause
@(end)
@(collect)
... lines of subquery: main clause
@(last)
... lines of subquery: last clause
@(end)
.cble
The
.code repeat
symbol may be specified instead of
.codn collect ,
which changes the meaning, see below:
.cblk
@(repeat)
... lines of subquery
@(end)
.cble
The subquery is matched repeatedly, starting at the current line.
If it fails to match, it is tried starting at the subsequent line.
If it matches successfully, it is tried at the line following the
entire extent of matched data, if there is one. Thus, the collected regions do
not overlap. (Overlapping behavior can be obtained: see the
.code @(trailer)
directive).
Unless certain keywords are specified, or unless the collection is explicitly
failed with
.codn @(fail) ,
it always succeeds, even if it collects nothing,
and even if the
.cod3 until / last
clause never finds a match.
If no
.cod3 until / last
last clause is specified, and the collect is not limited
using parameters, the collection is unbounded: it consumes the entire data
file. If any query material follows such the
.code collect
clause, it will fail if it
tries to match anything in the current file; but of course, it is possible to
continue matching in another file by means of
.codn @(next) .
.coNP The @ until / @ last clause in @ collect
If an
.cod3 until / last
last clause is specified, the collection stops when that clause
matches at the current position.
If an
.code until
clause terminates collect, no bindings are collected at that
position, even if the main clause matches at that position also. Moreover, the
position is not advanced. The remainder of the query begins matching at that
position.
If a last clause terminates collect, the behavior is different. Any bindings
captured by the main clause are thrown away, just like with the until clause.
However, the bindings in the last clause itself survive, and the position is
advanced to skip over that material.
Example:
.IP code:
.cblk
\ @(collect)
@a
@(until)
42
@b
@(end)
@c
.cble
.IP data:
.cblk
\ 1
2
3
42
5
6
.cble
.IP result:
.cblk
\ a[0]="1"
a[1]="2"
a[2]="3"
c="42"
.cble
.PP
The line
.code 42
is not collected, even though it matches
.con @a .
Furthermore,
the
.code @(until)
does not advance the position, so variable
.code c
takes
.codn 42 .
If the
.code @(until)
is changed to
.code @(last)
the output will be different:
.IP result:
.cblk
\ a[0]="1"
a[1]="2"
a[2]="3"
b="5"
c="6"
.cble
.PP
The
.code 42
is not collected into the a list, just like before. But now
the binding captured by
.code @b
emerges. Furthermore, the position advances
so variable now takes
.codn 6 .
The binding variables within the clause of a collect are treated specially.
The multiple matches for each variable are collected into lists,
which then appear as array variables in the final output.
Example:
.IP code:
.cblk
\ @(collect)
@a:@b:@c
@(end)
.cble
.IP data:
.cblk
\ John:Doe:101
Mary:Jane:202
Bob:Coder:313
.cble
.IP result:
.cblk
\ a[0]="John"
a[1]="Mary"
a[2]="Bob"
b[0]="Doe"
b[1]="Jane"
b[2]="Coder"
c[0]="101"
c[1]="202"
c[2]="313"
.cble
.PP
The query matches the data in three places, so each variable becomes
a list of three elements, reported as an array.
Variables with list bindings may be referenced in a query. They denote a
multiple match. The
.code -D
command line option can establish a one-dimensional
list binding.
The clauses of
.code collect
may be nested. Variable matches collated into lists in an
inner collect, are again collated into nested lists in the outer collect.
Thus an unbound variable wrapped in N nestings of
.code @(collect)
will
be an N-dimensional list. A one dimensional list is a list of strings;
a two dimensional list is a list of lists of strings, etc.
It is important to note that the variables which are bound within the main
clause of a collect. That is, the variables which are subject to
collection appear, within the collect, as normal one-value bindings. The
collation into lists happens outside of the collect. So for instance in the
query:
.cblk
@(collect)
@x=@x
@(end)
.cble
The left
.code @x
establishes a binding for some material preceding an equal sign.
The right
.code @x
refers to that binding. The value of
.cod @x
is different in each
iteration, and these values are collected. What finally comes out of the
collect clause is a single variable called
.code x
which holds a list containing each value that
was ever instantiated under that name within the collect clause.
Also note that the until clause has visibility over the bindings
established in the main clause. This is true even in the terminating
case when the until clause matches, and the bindings of the main clause
are discarded.
.coNP Keyword Parameters in @ collect
By default,
.code collect
searches the rest of the input indefinitely,
or until the
.cod3 until / last
clause matches. It skips arbitrary amounts of
nonmatching material before the first match, and between matches.
Within the
.code @(collect)
syntax, it is possible to specify some useful keyword
parameters for additional control of the behavior. For instance
.cblk
@(collect :maxgap 5)
.cble
means that the collect will terminate if it does not find a match within five
lines of the starting position, or if more than five lines are skipped since
any successful match. A
.code :maxgap
of
.code 0
means that the collected regions must be
adjacent. For instance:
.cblk
@(collect :maxgap 0)
M @a
@(end)
.cble
means: from here, collect consecutive lines of the form
.strn "M ..." .
This will not
search for the first such line, nor will it skip lines which do not match this
form.
Other keywords are
.codn :mingap ,
and
.codn :gap .
The
.code :mingap
keyword specifies a minimum
gap between matches, but has no effect on the distance to the first match. The
.code :gap
keyword effectively specifies
.code :mingap
and
.code :maxgap
at the same time, and can only be
used if these other two are not used. Thus:
.cblk
@(collect :gap 1)
@a
@(end)
.cble
means collect every other line starting with the current line. Several
other supported keywords are
.codn :times ,
.codn :mintimes ,
.code :maxtimes
and
.codn :lines .
The shorthand
.code :times N
means the same thing as
.codn ":mintimes N :maxtimes N" .
These specify how many matches should be collected. If there are fewer
than
.code :mintimes
matches, the collect fails. If
.code :maxtimes
matches are collected,
collect stops collecting immediately. Example:
.cblk
@(collect :times 3)
@a @b
@(end)
.cble
This will collect a match for
.str @a @b
exactly three times. If three
matches are not found, it will fail.
The
.code :lines
parameter specifies the upper bound on how many lines
should be scanned by collect, measuring from the starting position.
The extent of the collect body is not counted. Example:
.cblk
@(collect :lines 2)
foo: @a
bar: @b
baz: @c
@(end)
.cble
The above
.code collect
will look for a match only twice: at the current position,
and one line down.
There is one more keyword,
.codn :vars ,
discussed in the following section.
.coNP Specifying Variables in @ collect
Normally, any variable for which a new binding occurs in a
.code collect
block is collected. A collect clause may be "sloppy": it can neglect to collect
some variables on some iterations, or bind some variables which are intended to
behave like local temporaries, but end up collated into lists. Another issue is
that the collect clause might not match anything at all, and then none of the
variables are bound.
The
.code :vars
keyword allows the query writer to add discipline the
.code collect
body.
The argument to
.code :vars
is a list of variable specs. A variable spec is either a
symbol, or a
.cblk
.meti >> ( symbol << expression )
.cble
pair, where the expression specifies a
default value.
When a
.code :vars
list is specified, it means that only the given variables can
emerge from the successful collect. Any newly introduced bindings for other
variables do not propagate.
Furthermore, for any variable which is not specified with a default value, the
collect body, whenever it matches successfully, must bind that variable. If it
neglects to bind the variable, an exception of type query-error is thrown.
(If a
.code collect
body matches successfully, but produces no new bindings, then
this error is suppressed.)
For any variable which does have a default value, if the
.code collect
body neglects
to bind that variable, the behavior is as if
.code collect
did bind that variable
to that default value.
The default values are expressions, and so can be quasiliterals.
Lastly, if in the event that
.code collect
does not match anything, the variables
specified in vars (whether or not they have a default value) are all bound to
empty lists. (These bindings are established after the processing of the
.cod3 until / last
last clause, if present.)
Example:
.cblk
@(collect :vars (a b (c "foo")))
@a @c
@(end)
.cble
Here, if the body
.str @a @c
matches, an error will be thrown because one of the
mandatory variables is
.codn b ,
and the body neglects to produce a binding for
.codn b .
Example:
.cblk
@(collect :vars (a (c "foo")))
@a @b
@(end)
.cble
Here, if
.str @a @b
matches, only
.code a
will be collected, but not
.codn b ,
because
.code b
is not
in the variable list. Furthermore, because there is no binding for
.code c
in the
body, a binding is created with the value
.strn foo ,
exactly as if
.code c
matched
such a piece of text.
In the following example, the assumption is that
.code THIS NEVER MATCHES
is not found anywhere in the input but the line
.code THIS DOES MATCH
is found and has a successor which is bound to
.codn a .
Because the body did not
match, the
.code :vars
.code a
and
.cod b
should be bound to empty lists. But
.code a
is bound by the last clause to some text, so this takes precedence. Only
.code b
is bound to an empty list.
.cblk
@(collect :vars (a b))
THIS NEVER MATCHES
@(last)
THIS DOES MATCH
@a
@(end)
.cble
The following means: do not allow any variables to propagate out of any
iteration of the collect and therefore collect nothing:
.cblk
@(collect :vars nil)
...
@(end)
.cble
Instead of writing
.codn "@(collect :vars nil)" ,
it is possible to write
.codn @(repeat) .
.code @(repeat)
takes all collect keywords, except for
.codn :vars .
There is a
.code @(repeat)
directive used in
.code @(output)
clauses; that is a different directive.
.dir coll
The
.code coll
directive is the horizontal version of
.codn collect .
Whereas
.code collect
works with multi-line clauses on line-oriented
material,
.code coll
works within a single line. With
.codn coll ,
it is possible to
recognize repeating regularities within a line and collect lists.
Regular-expression based Positive Match variables work well with coll.
Example: collect a comma-separated list, terminated by a space.
.IP code:
.cblk
\ @(coll)@{A /[^, ]+/}@(until) @(end)@B
.cble
.IP data:
.cblk
\ foo,bar,xyzzy blorch
.cble
.IP result:
.cblk
\ A[0]="foo"
A[1]="bar"
A[2]="xyzzy"
B=blorch
.cble
.PP
Here, the variable
.code A
is bound to tokens which match the regular
expression
.code /[^, ]+/:
non-empty sequence of characters other than commas or
spaces.
Like
.codn collect ,
.code coll
searches for matches. If no match
occurs at the current character position, it tries at the next character
position. Whenever a match occurs, it continues at the character position which
follows the last character of the match, if such a position exists.
If not bounded by an until clause, it will exhaust the entire line. If the
until clause matches, then the collection stops at that position,
and any bindings from that iteration are discarded.
Like collect, coll also supports an
.cod3 until / last
clause, which propagates variable
bindings and advances the position.
.code coll
clauses nest, and variables bound within a coll are available to clauses
within the rest of the
.code coll
clause, including the
.cod3 until / last
clause, and appear as
single values. The final list aggregation is only visible after the
.code coll
clause.
The behavior of
.code coll
leads to difficulties when a delimited variable are used
to match material which is delimiter separated rather than terminated.
For instance, entries in a comma-separated files usually do
not appear as
.str a,b,c,
but rather
.strn a,b,c .
So for instance, the following result is not satisfactory:
.IP code:
.cblk
\ @(coll)@a @(end)
.cble
.IP data:
.cblk
\ 1 2 3 4 5
.cble
.IP result:
.cblk
\ a[0]="1"
a[1]="2"
a[2]="3"
a[3]="4"
.cble
.PP
The
.code 5
is missing because it isn't followed by a space, which the text-delimited
variable match
.str "@a "
looks for. After matching "4 ", coll continues to look for
matches, and doesn't find any. It is tempting to try to fix it like this:
.IP code:
.cblk
\ @(coll)@a@/ ?/@(end)
.cble
.IP data:
.cblk
\ 1 2 3 4 5
.cble
.IP result:
.cblk
\ a[0]=""
a[1]=""
a[2]=""
a[3]=""
a[4]=""
a[5]=""
a[6]=""
a[7]=""
a[8]=""
.cble
.PP
The problem now is that the regular expression
.code / ?/
(match either a space or nothing), matches at any position.
So when it is used as a variable
delimiter, it matches at the current position, which binds the empty string to
the variable, the extent of the match being zero. In this situation, the
.code coll
directive proceeds character by character. The solution is to use
positive matching: specify the regular expression which matches the item,
rather than a trying to match whatever follows. The collect directive will
recognize all items which match the regular expression:
.IP code:
.cblk
\ @(coll)@{a /[^ ]+/}@(end)
.cble
.IP data:
.cblk
\ 1 2 3 4 5
.cble
.IP result:
.cblk
\ a[0]="1"
a[1]="2"
a[2]="3"
a[3]="4"
a[4]="5"
.cble
.PP
The
.code until
clause can specify a pattern which, when recognized, terminates
the collection. So for instance, suppose that the list of items may
or may not be terminated by a semicolon. We must exclude
the semicolon from being a valid character inside an item, and
add an until clause which recognizes a semicolon:
.IP code:
.cblk
\ @(coll)@{a /[^ ;]+/}@(until);@(end);
.cble
.IP data:
.cblk
\ 1 2 3 4 5;
.cble
.IP result:
.cblk
\ a[0]="1"
a[1]="2"
a[2]="3"
a[3]="4"
a[4]="5"
.cble
.PP
Whether followed by the semicolon or not, the items are collected properly.
Note that the
.code @(end)
is followed by a semicolon. That's because
when the
.code @(until)
clause meets a match, the matching material
is not consumed.
This repetition can, of course, be avoided by using
.code @(last)
instad of
.code @(until)
since
.code @(last)
consumes the terminating material.
Instead of the above regular-expresion-based approach, this extraction problem
can also be solved with
.codn cases :
.IP code:
.cblk
\ @(coll)@(cases)@a @(or)@a@(end)@(end)
.cble
.IP data:
.cblk
\ 1 2 3 4 5
.cble
.IP result:
.cblk
\ a[0]="1"
a[1]="2"
a[2]="3"
a[3]="4"
a[4]="5"
.cble
.PP
.coNP Keyword parameters in @ coll
The
.code @(coll)
directive takes most of the same parameters as
.codn @(collect) .
See the section Collect Keyword Parameters above.
So for instance
.code @(coll :gap 0)
means that the collects must be
consecutive, and
.code @(coll :maxtimes 2)
means that at most two matches
will be collected. The
.code :lines
keyword does not exist, but there is
an analogous
.code :chars
keyword.
.dir flatten
The
.code flatten
directive can be used to convert variables to one dimensional
lists. Variables which have a scalar value are converted to lists containing
that value. Variables which are multidimensional lists are flattened to
one-dimensional lists.
Example (without
.codn @(flatten) )
.IP code:
.cblk
\ @b
@(collect)
@(collect)
@a
@(end)
@(end)
.cble
.IP data:
.cblk
\ 0
1
2
3
4
5
.cble
.IP result:
.cblk
\ b="0"
a_0[0]="1"
a_1[0]="2"
a_2[0]="3"
a_3[0]="4"
a_4[0]="5"
.cble
.PP
Example (with
.codn @(flatten) ):
.IP code:
.cblk
\ @b
@(collect)
@(collect)
@a
@(end)
@(end)
@(flatten a b)
.cble
.IP data:
.cblk
\ 0
1
2
3
4
5
.cble
.IP result:
.cblk
\ b="0"
a[0]="1"
a[1]="2"
a[2]="3"
a[3]="4"
a[4]="5"
.cble
.PP
.dir merge
The
.code merge
directive provides a way of combining two or more variables
in a somewhat complicated but very useful way.
To understand what merge does we first have to define a property called depth.
The depth of an atom such as a string is defined as
.codn 1 .
The depth of an empty
list is
.codn 0 .
The depth of a nonempty list is one plus the depth of its deepest
element. So for instance
.str foo
has depth 1,
.cblk
("foo")
.cble
has depth 2, and
.cblk
("foo" ("bar"))
.cble
has depth three.
We can now define the binary (two argument) merge operation as follows.
.IP 1
.code (merge A B)
first normalizes the values
.code A
and
.code B
such that they have equal depth.
.IP 2
A value which has depth zero is put into a one element list.
.IP 3
If either value has a smaller depth than the other, it is wrapped
in a list as many times as needed to give it equal depth.
Finally, the values are appended together.
.PP
Merge takes more than two arguments. These are merged by a left reduction. The
leftmost two values are merged, and then this result is merged with the third
value, and so on.
Merge is useful for combining the results from collects at different
levels of nesting such that elements are at the appropriate depth.
.dir cat
The
.code cat
directive converts a list variable into a single
piece of text. The syntax is:
.cblk
.mets @(cat < var <> [ sep ])
.cble
The
.meta sep
argument specifies a separating piece of text. If no separator
is specified, then a single space is used.
Example:
.IP code:
.cblk
\ @(coll)@{a /[^ ]+/}@(end)
@(cat a ":")
.cble
.IP data:
.cblk
\ 1 2 3 4 5
.cble
.IP result:
.cblk
\ a="1:2:3:4:5"
.cble
.PP
.dir bind
The syntax of the
.code bind
directive is:
.cblk
.mets @(bind < pattern < expression >> { keyword << value }*)
.cble
The
.code bind
directive is a kind of pattern match, which matches one or more
variables on the left hand side pattern to the value of a variable on the
right hand side. The right hand side variable must have a binding, or else the
directive fails. Any variables on the left hand side which are unbound receive
a matching piece of the right hand side value. Any variables on the left which
are already bound must match their corresponding value, or the bind fails. Any
variables which are already bound and which do match their corresponding value
remain unchanged (the match can be inexact).
The simplest bind is of one variable against itself, for instance bind
.code A
against
.codn A :
.cblk
@(bind A A)
.cble
This will fail if
.code A
is not bound, (and complain loudly). If
.code A
is bound, it
succeeds, since
.code A
matches
.codn A .
The next simplest bind binds one variable to another:
.cblk
@(bind A B)
.cble
Here, if
.code A
is unbound, it takes on the same value as
.codn B .
If
.code A
is bound, it has
to match
.codn B ,
or the bind fails. Matching means that either
.IP -
.code A
and
. code B
are the same text
.IP -
.code A
is text,
.code B
is a list, and
.code A
occurs within
.codn B .
.IP -
vice versa:
.code B
is text,
.code A
is a list, and
.code B
occurs within
.codn A .
.IP -
.code A
and
.code B
are lists and are either identical, or one is
found as substructure within the other.
.PP
The right hand side does not have to be a variable. It may be some other
object, like a string, quasiliteral, regexp, or list of strings, et cetera. For
instance
.cblk
@(bind A "ab\etc")
.cble
will bind the string
.str ab\etc
to the variable
.code A
if
.code A
is unbound. If
.code A
is bound, this will fail unless
.code A
already contains an identical string. However, the right hand side of a bind
cannot be an unbound variable, nor a complex expression that contains unbound
variables.
The left hand side of
.code bind
can be a nested list pattern containing variables.
The last item of a list at any nesting level can be preceded by a
.code .
(dot), which means that the variable matches the rest of the list from that
position.
Example: suppose that the list A contains
.cblk
("now" "now" "brown" "cow").
.cble
Then the
directive
.codn "@(bind (H N . C) A)" ,
assuming that
.codn H ,
.code N
and
.code C are unbound variables,
will bind
.code H
to
.strn how ,
code N
to
.strn now ,
and
.code C
to the remainder of the list
.cblk
("brown" "cow").
.cble
Example: suppose that the list
.code A
is nested to two dimensions and contains
.cblk
(("how" "now") ("brown" "cow")).
.cble
Then
.code @(bind ((H N) (B C)) A)
binds
.code H to
.strn how ,
.code N
to
.strn now ,
.code B
to
.str brown
and
.code C
to
.strn cow .
The dot notation may be used at any nesting level. it must be followed
by an item. The forms
.code (.)
and
.code (X .)
are invalid, but
.code (. X)
is valid and equivalent to
.codn X .
The number of items in a left pattern match must match the number of items in
the corresponding right side object. So the pattern
.code ()
only matches
an empty list. The notations
.code ()
and
.code nil
mean exactly the same thing.
The symbols
.codn nil ,
.code t
and keyword symbols may be used on either side.
They represent themselves. For example
.code @(bind :foo :bar)
fails,
but
.code @(bind :foo :foo)
succeeds since the two sides denote the same
keyword symbol object.
.coNP Keywords in the @ bind directive
The
.code bind
directive accepts these keywords:
.coIP :lfilt
The argument to
.code :lfilt
is a filter specification. When the left side pattern
contains a binding which is therefore matched against its counterpart from the
right side expression, the left side is filtered through the filter specified
by
.code :lfilt
for the purposes of the comparison. For example:
.cblk
@(bind "a" "A" :lfilt :upcase)
.cble
produces a match, since the left side is the same as the right after
filtering through the :upcase filter.
.coIP :rfilt
The argument to
.code :rfilt
is a filter specification. The specified filter is
applied to the right hand side material prior to matching it against
the left side. The filter is not applied if the left side is a variable
with no binding. It is only applied to determine a match. Binding takes
place the unmodified right hand side object.
For example, the following produces a match:
.cblk
@(bind "A" "a" :rfilt :upcase)
.cble
.coIP :filter
This keyword is a shorthand to specify both filters to the same value.
For instance
.code :filter :upcase
is equivalent to
.codn ":lfilt :upcase :rfilt :upcase" .
For a description of filters, see Output Filtering below.
Of course, compound filters like
.code (:from_html :upcase)
are supported with all these keywords. The filters apply across arbitrary
patterns and nested data.
Example:
.cblk
@(bind (a b c) ("A" "B" "C"))
@(bind (a b c) (("z" "a") "b" "c") :rfilt :upcase)
.cble
Here, the first bind establishes the values for
.codn a ,
.code b
and
.codn c ,
and the second bind
succeeds, because the value of a matches the second element of the list
.code ("z" "a")
if it is upcased, and likewise
.code b
matches
.str "b"
and
.code c
matches
.str c
if these are upcased.
.coNP Lisp forms in the @ bind directive
\*(TL forms, introduced by
.code @
may be used on either side of
.codn bind .
Example:
.cblk
@(bind a @(+ 2 2))
@(bind @(+ 2 2) @(* 2 2))
.cble
Here,
.code a
is bound to the integer
.codn 4 .
The second
.code bind
then succeeds because the forms
.code (+ 2 2)
and
.code (* 2 2)
produce equal values.
.dir set
The
.code set
directive syntactically resembles
.codn bind ,
but is not a pattern match. It overwrites
the previous values of variables with new values from the right hand side.
Each variable that is assigned must have an existing binding:
.code set
will not induce binding.
Examples follow.
Store the value of
.code A
back into
.codn A ,
an operation with no effect:
.cblk
@(set A A)
.cble
Exchange the values of
.code A
and
.codn B :
.cblk
@(set (A B) (B A))
.cble
Store a string into
.codn A :
.cblk
@(set A "text")
.cble
Store a list into
.codn A :
.cblk
@(set A ("line1" "line2"))
.cble
Destructuring assignment.
.code A
ends up with
.strn A ,
.code B
ends up with
.cblk
("B1" "B2")
.cble
and
.code C
binds to
.cblk
("C1" "C2").
.cble
.cblk
@(bind D ("A" ("B1" "B2") "C1" "C2"))
@(bind (A B C) (() () ()))
@(set (A B . C) D)
.cble
Note that
.code set
does not support a \*(TL expression on the left side, so the following
are invalid syntax:
.cblk
@(set @(+ 1 1) @(* 2 2))
@(set @b @(list "a"))
.cble
The second one is erroneous even though there is a variable on the left.
Because it is preceded by the
.code @
escape, it is a Lisp variable, and not a pattern variable.
.dir rebind
The
.code rebind
directive resembles
.code set
but it is not an assignment.
It combines the semantics of
.codn local ,
.code bind
and
.codn set .
The expression on the right hand side is evaluated in the current
environment. Then the variables in the pattern on the left are introduced
as new bindings, whose values come from the pattern.
.code rebind
makes it easy to create temporary bindings based on existing bindings.
.cblk
@(define pattern-function (arg))
@;; inside a pattern function:
@(rebind recursion-level @(+ recursion-level 1))
@;; ...
@(end)
.cble
When the function terminates, the previous value of recursion-level
is restored. The effect is like the following, but much easier
to write and faster to execute:
.cblk
@(define pattern-function (arg))
@;; inside a pattern function:
@(local temp)
@(set temp recursion-level)
@(local recursion-level)
@(set recursion-level @(+ temp 1))
@;; ...
@(end)
.cble
.dir forget
The
.code forget
has two spellings:
.code @(forget)
and
.codn @(local).
The arguments are one or more symbols, for example:
.cblk
@(forget a)
@(local a b c)
.cble
this can be written
.cblk
@(local a)
@(local a b c)
.cble
Directives which follow the forget or local directive no longer see
any bindings for the symbols mentioned in that directive, and
can establish new bindings.
It is not an error if the bindings do not exist.
It is strongly recommended to use the
.code @(local)
spelling in functions,
because the forgetting action simulates local variables:
for the given symbols, the machine forgets any earlier variables
from outside of the function, and consequently, any new bindings
for those variables belong to the function. (Furthermore,
functions suppress the propagation of variables that are not
in their parameter list, so these locals will be automatically
forgotten when the function terminates.)
.dir do
The syntax of
.code @(do)
is:
.cblk
.mets @(do << lisp-expression )
.cble
The do directive evaluates a \*(TL expression. (See TXR LISP far
below.) The value of the expression is ignored, and matching continues
continues with the directives which follow the
.code do
directive, if any.
In the context of the
.code do
directive, the expression should not be introduced by the
.code @
symbol; it is expected to be a Lisp expression.
Example:
.cblk
@; match text into variables a and b, then insert into hash table h
@(bind h (hash :equal-based))
@a:@b
@(do (set [h a] b))
.cble
.SH* BLOCKS
.SS* Overview
Blocks are sections of a query which are either denoted by a name,
or are anonymous. They may nest: blocks can occur within blocks
and other constructs.
Blocks are useful for terminating parts of a pattern matching search
prematurely, and escaping to a higher level. This makes blocks not only
useful for simplifying the semantics of certain pattern matches,
but also an optimization tool.
Judicious use of blocks and escapes can reduce or eliminate the amount of
backtracking that \*(TX performs.
.dir block
The
.cblk
.meti @(block << name )
.cble
directive introduces a named block, except when
.meta name
is the symbol
.codn nil .
The
.code @(block)
directive introduces an unnamed block, equivalent
to
.codn "@(block nil)" .
The
.code @(skip)
and
.code @(collect)
directives introduce implicit anonymous blocks,
as do function bodies.
.SS* Block Scope
The names of blocks are in a distinct namespace from the variable binding
space. So
.code @(block foo)
is unrelated to the variable
.codn @foo .
A block extends from the
.code @(block ...)
directive which introduces it,
until the matching
.codn @(end) ,
and may be empty. For instance:
.cblk
@(some)
abc
@(block foo)
xyz
@(end)
@(end)
.cble
Here, the block foo occurs in a
.code @(some)
clause, and so it extends to the
.code @(end)
which terminates the block. After that
.codn @(end) ,
the name foo is not
associated with a block (is not "in scope"). The second
.code @(end)
terminates
the
.code @(some)
block.
The implicit anonymous block introduced by
.code @(skip)
has the same scope
as the
.codn @(skip) :
it extends over all of the material which follows the skip,
to the end of the containing subquery.
.SS* Block Nesting
Blocks may nest, and nested blocks may have the same names as blocks in
which they are nested. For instance:
.cblk
@(block)
@(block)
...
@(end)
@(end)
.cble
is a nesting of two anonymous blocks, and
.cblk
@(block foo)
@(block foo)
@(end)
@(end)
.cble
is a nesting of two named blocks which happen to have the same name.
When a nested block has the same name as an outer block, it creates
a block scope in which the outer block is "shadowed"; that is to say,
directives which refer to that block name within the nested block refer to the
inner block, and not to the outer one.
.SS* Block Semantics
A block normally does nothing. The query material in the block is evaluated
normally. However, a block serves as a termination point for
.code @(fail)
and
.code @(accept)
directives which are in scope of that block and refer to it.
The precise meaning of these directives is:
.meIP @(fail << name )
Immediately terminate the enclosing query block called
.metn name ,
as if that block
failed to match anything. If more than one block by that name encloses
the directive, the inner-most block is terminated. No bindings emerge from
a failed block.
.coIP @(fail)
Immediately terminate the innermost enclosing anonymous block, as if
that block failed to match.
If the implicit block introduced by
.code @(skip)
is terminated in this manner,
this has the effect of causing
.code skip
itself to fail. I.e. the behavior
is as if skip search did not find a match for the trailing material,
except that it takes place prematurely (before the end of the available
data source is reached).
If the implicit block associated with a
.code @(collect)
is terminated this way,
then the entire
.code collect
fails. This is a special behavior, because a
collect normally does not fail, even if it matches nothing and collects nothing!
To prematurely terminate a collect by means of its anonymous block, without
failing it, use
.codn @(accept) .
.meIP @(accept << name )
Immediately terminate the enclosing query block called
.metn name ,
as if that block
successfully matched. If more than one block by that name encloses the
directive, the inner-most block is terminated. Any bindings established within
that block until this point emerge from that block.
.coIP @(accept)
Immediately terminate the innermost enclosing anonymous block, as if
that block successfully matched. Any bindings established within
that block until this point emerge from that block.
If the implicit block introduced by
.code @(skip)
is terminated in this manner,
this has the effect of causing the skip itself to succeed, as if
all of the trailing material had successfully matched.
If the implicit block associated with a
.code @(collect)
is terminated this way,
then the collection stops. All bindings collected in the current iteration of
the collect are discarded. Bindings collected in previous iterations are
retained, and collated into lists in accordance with the semantics of collect.
Example: alternative way to achieve
.code @(until)
termination:
.cblk
@(collect)
@ (maybe)
---
@ (accept)
@ (end)
@LINE
@(end)
.cble
This query will collect entire lines into a list called
.codn LINE .
However,
if the line
.code ---
is matched (by the embedded
.codn @(maybe)),
the collection
is terminated. Only the lines up to, and not including the
.code ---
line, are collected. The effect is identical to:
.cblk
@(collect)
@LINE
@(until)
---
@(end)
.cble
The difference (not relevant in these examples) is that the until clause has
visibility into the bindings set up by the main clause.
However, the following example has a different meaning:
.cblk
@(collect)
@LINE
@ (maybe)
---
@ (accept)
@ (end)
@(end)
.cble
Now, lines are collected until the end of the data source, or until a line is
found which is followed by a
.code ---
line. If such a line is found,
the collection stops, and that line is not included in the collection!
The
.code @(accept)
terminates the process of the collect body, and so the
action of collecting the last
.code @LINE
binding into the list is not performed.
.PP
.SS* Data Extent of Terminated Blocks
A query block may have matched some material prior to being terminated by
.codn accept .
In that case, it is deemed to have only matched that material,
and not any material which follows. This may matter, depending on the context
in which the block occurs.
Example:
.IP code:
.cblk
\ @(some)
@(block foo)
@first
@(accept foo)
@ignored
@(end)
@second
.cble
.IP data:
.cblk
\ 1
2
3
.cble
.IP result:
.cblk
\ first="1"
second="2"
.cble
.PP
At the point where the
.code accept
occurs, the foo block has matched the first line,
bound the text
.str 1
to the variable
.codn @first .
The block is then terminated.
Not only does the
.code @first
binding emerge from this terminated block, but
what also emerges is that the block advanced the data past the first line to
the second line. Next, the
.code @(some)
directive ends, and propagates the
bindings and position. Thus the
.code @second
which follows then matches the second
line and takes the text
.strn 2 .
In the following query, the foo block occurs inside a maybe clause.
Inside the foo block there is a
.code @(some)
clause. Its first subclause
matches variable
.code @first
and then terminates block foo. Since block foo is
outside of the
.code @(some)
directive, this has the effect of terminating the
.code @(some)
clause:
.IP code:
.cblk
\ @(maybe)
@(block foo)
@ (some)
@first
@ (accept foo)
@ (or)
@one
@two
@three
@four
@ (end)
@(end)
@second
.cble
.IP data:
.cblk
\ 1
2
3
4
5
.cble
.IP result:
.cblk
\ first="1"
second="2"
.cble
.PP
The second clause of the
.code @(some)
directive, namely:
.cblk
@one
@two
@three
@four
.cble
is never processed. The reason is that subclauses are processed in top
to bottom order, but the processing was aborted within the
first clause the
.codn "@(accept foo)" .
The
.code @(some)
construct never gets the
opportunity to match four lines.
If the
.code @(accept foo)
line is removed from the above query, the output
is different:
.IP code:
.cblk
\ @(maybe)
@(block foo)
@ (some)
@first
@# <-- @(accept foo) removed from here!!!
@ (or)
@one
@two
@three
@four
@ (end)
@(end)
@second
.cble
.IP data:
.cblk
\ 1
2
3
4
5
.cble
.IP result:
.cblk
\ first="1"
one="1"
two="2"
three="3"
four="4"
second="5"
.cble
.PP
Now, all clauses of the
.code @(some)
directive have the opportunity to match.
The second clause grabs four lines, which is the longest match.
And so, the next line of input available for matching is
.codn 5 ,
which goes
to the
.code @second
variable.
.coSS Interaction Between the @ trailer and @ accept Directives
If one of the clauses which follow a
.code @(trailer)
requests a successful
termination to an outer block via
.codn @(accept) ,
then
.code @(trailer)
intercepts the escape and adjusts the data extent to the position
that it was given.
Example:
.IP code:
.cblk
\ @(block)
@(trailer)
@line1
@line2
@(accept)
@(end)
@line3
.cble
.IP data:
.cblk
\ 1
2
3
.cble
.IP result:
.cblk
\ line1="1"
line2="2"
line3="1"
.cble
.PP
The variable
.code line3
is bound to
.str 1
because although
.code @(accept)
yields a data
position which has advanced to the third line, this is intercepted by
.code @(trailer)
and adjusted back to the first line. Neglecting to do this adjustment
would violate the semantics of
.codn trailer .
Directives other than
.code @(trailer)
have no such special interaction with accept.
.SH* FUNCTIONS
.SS* Overview
\*(TX functions allow a query to be structured to avoid repetition.
On a theoretical note, because
\*(TX functions support recursion, functions enable \*(TX to match some
kinds of patterns which exhibit self-embedding, or nesting,
and thus cannot be matched by a regular language.
Functions in \*(TX are not exactly like functions in mathematics or functional
languages, and are not like procedures in imperative programming languages.
They are not exactly like macros either. What it means for a
\*(TX function to take arguments and produce a result is different from
the conventional notion of a function.
A \*(TX function may have one or more parameters. When such a function is
invoked, an argument must be specified for each parameter. However, a special
behavior is at play here. Namely, some or all of the argument expressions may
be unbound variables. In that case, the corresponding parameters behave like
unbound variables also. Thus \*(TX function calls can transmit the "unbound"
state from argument to parameter.
It should be mentioned that functions have access to all bindings that are
visible in the caller; functions may refer to variables which are not
mentioned in their parameter list.
With regard to returning, \*(TX functions are also unconventional. If the
function fails, then the function call is considered to have failed. The
function call behaves like a kind of match; if the function fails, then the
call is like a failed match.
When a function call succeeds, then the bindings emanating from that function
are processed specially. Firstly, any bindings for variables which do not
correspond to one of the function's parameters are thrown away. Functions may
internally bind arbitrary variables in order to get their job done, but only
those variables which are named in the function argument list may propagate out
of the function call. Thus, a function with no arguments can only indicate
matching success or failure, but not produce any bindings. Secondly,
variables do not propagate out of the function directly, but undergo
a renaming. For each parameter which went into the function as an unbound
variable (because its corresponding argument was an unbound variable),
if that parameter now has a value, that value is bound onto the corresponding
argument.
Example:
.cblk
@(define collect-words (list))
@(coll)@{list /[^ \et]+/}@(end)
@(end)
.cble
The above function
.code collect-words
contains a query which collects words from a
line (sequences of characters other than space or tab), into the list variable
called
.codn list .
This variable is named in the parameter list of the function,
therefore, its value, if it has one, is permitted to escape from the function
call.
Suppose the input data is:
.cblk
Fine summer day
.cble
and the function is called like this:
.cblk
@(collect-words wordlist)
.cble
The result (with
.codn "txr -B" )
is:
.cblk
wordlist[0]=Fine
wordlist[1]=summer
wordlist[1]=day
.cble
How it works is that in the function call
.codn "@(collect-words wordlist)" ,
.code wordlist
is an unbound variable. The parameter corresponding to that
unbound variable is the parameter
.codn list .
Therefore, that parameter
is unbound over the body of the function. The function body collects the
words of
.str Fine summer day
into the variable
.codn list,
and then
yields the that binding. Then the function call completes by
noticing that the function parameter
.code list
now has a binding, and
that the corresponding argument
.code wordlist
has no binding. The binding
is thus transferred to the
.code wordlist
variable. After that, the
bindings produced by the function are thrown away. The only enduring
effects are:
.IP -
the function matched and consumed some input; and
.IP -
the function succeeded; and
.IP -
the
.code wordlist
variable now has a binding.
.PP
Another way to understand the parameter behavior is that function
parameters behave like proxies which represent their arguments. If an argument
is an established value, such as a character string or bound variable, the
parameter is a proxy for that value and behaves just like that value. If an
argument is an unbound variable, the function parameter acts as a proxy
representing that unbound variable. The effect of binding the proxy is
that the variable becomes bound, an effect which is settled when the
function goes out of scope.
Within the function, both the original variable and the proxy are
visible simultaneously, and are independent. What if a function binds both of
them? Suppose a function has a parameter called
.codn P ,
which is called with an argument
.codn A ,
which is an unbound variable, and then, in the function, both
.code A
and
.code P
bound. This is
permitted, and they can even be bound to different values. However, when the
function terminates, the local binding of A simply disappears (because
the symbol
.code A
is not among the paramters of the function).
Only the value bound to
.code P
emerges, and is bound to
.codn A ,
which still appears unbound at that point. The
.code P
binding disappears also, and the net effect is that
.code A
is now bound. The "proxy" binding of
.code A
thorugh the parameter
.code P
"wins" the conflict with the direct binding.
.SS* Definition Syntax
Function definition syntax comes in two flavors: vertical and horizontal.
Horizontal definitions actually come in two forms, the distinction
between which is hardly noticeable, and the need for which is
made clear below.
A function definition begins with a
.code @(define ...)
directive. For vertical
functions, this is the only element in a line.
The
.code define
symbol must be followed by a symbol, which is the name of the
function being defined. After the symbol, there is a parenthesized optional
argument list. If there is no such list, or if the list is specified as
.code ()
or
the symbol
.code nil
then the function has no parameters. Examples of valid
.code define
syntax are:
.cblk
@(define foo)
@(define bar ())
@(define match (a b c))
.cble
If the define directive is followed by more material on the same line, then
it defines a horizontal function:
.cblk
@(define match-x)x@(end)
.cble
If the define is the sole element in a line, then it
is a vertical function, and the function definition continues below:
.cblk
@(define match-x)
x
@(end)
.cble
The difference between the two is that a horizontal function matches
characters within a line, whereas a vertical function matches lines
within a stream. The former
.code match-x
matches the character
.codn x ,
advancing
to the next character position. The latter
.code match-x
matches a line consisting
of the character
.codn x ,
advancing to the next line.
Material between
.code @(define)
and
.code @(end)
is the function body. The define
directive may be followed directly by the
.code @(end)
directive, in which case the
function has an empty body.
Functions may be nested within function bodies. Such local functions have
dynamic scope. They are visible in the function body in which they are defined,
and in any functions invoked from that body.
The body of a function is an anonymous block. (See BLOCKS above).
.SS* Two Forms of The Horizontal Function
If a horizontal function is defined as the only element of a line,
it may not be followed by additional material. The following
construct is erroneous:
.cblk
@(define horiz (x))@foo:@bar@(end)lalala
.cble
This kind of definition is actually considered to be in the vertical context,
and like other directives that have special effects and that do not match
anything, it does not consume a line of input. If the above syntax were
allowed, it would mean that the line would not only define a function but also
match
.codn "lalala" .
This would, in turn, would mean that the
.code @(define)...@(end)
is
actually in horizontal mode, and so it matches a span of zero characters within
a line (which means that is would require a line of input to match: a
surprising behavior for a non-matching directive!)
A horizontal function can be defined in an actual horizontal context. This
occurs if its is in a line where it is preceded by other material.
For instance:
.cblk
X@(define fun)...@(end)Y
.cble
This is a query line which must match the text
.codn XY .
It also defines the function
.codn fun .
The main use of this form is for nested horizontal functions:
.cblk
@(define fun)@(define local_fun)...@(end)@(end)
.cble
.SS* Vertical-Horizontal Overloading
A function of the same name may be defined as both vertical and horizontal.
Both functions are available at the same time. Which one is used by
a call is resolved by context. See the section Vertical Versus Horizontal Calls
below.
.SS* Call Syntax
A function is invoked by compound directive whose first symbol is the name of
that function. Additional elements in the directive are the arguments.
Arguments may be symbols, or other objects like string and character
literals, quasiliterals ore regular expressions.
Example:
.IP code:
.cblk
\ @(define pair (a b))
@a @b
@(end)
@(pair first second)
@(pair "ice" cream)
.cble
.IP data:
.cblk
\ one two
ice milk
.cble
.IP result:
.cblk
\ first="one"
second="two"
cream="milk"
.cble
.PP
The first call to the function takes the line
.strn "one two" .
The parameter
.code a
takes
.str one
and parameter
.code b
takes
.strn two .
These are rebound to the arguments
.code first
and
.codn second .
The second call to the function binds the a parameter to the word
.strn "ice" ,
and the
.code b
is unbound, because the
corresponding argument
.code cream
is unbound. Thus inside the function,
.code a
is forced to match
.codn "ice" .
Then a space is matched and
.code b
collects the text
.strn milk .
When the function returns, the unbound
.str cream
variable gets this value.
If a symbol occurs multiple times in the argument list, it constrains
both parameters to bind to the same value. That is to say, all parameters
which, in the body of the function, bind a value, and which are all derived
from the same argument symbol must bind to the same value. This is settled when
the function terminates, not while it is matching. Example:
.IP code:
.cblk
\ @(define pair (a b))
@a @b
@(end)
@(pair same same)
.cble
.IP data:
.cblk
\ one two
.cble
.IP result:
.cblk
\ [query fails]
.cble
.PP
Here the query fails because
.code a
and
.code b
are effectively proxies for the same unbound variable
.code same
and are bound to different values, creating a conflict which
constitutes a match failure.
.SS* Vertical Versus Horizontal Calls
A function call which is the only element of the query line in
which it occurs is ambiguous. It can go either to a vertical
function or to the horizontal one. If both are defined, then
it goes to the vertical one.
Example:
.IP code:
.cblk
\ @(define which (x))@(bind x "horizontal")@(end)
@(define which (x))
@(bind x "vertical")
@(end)
@(which fun)
.cble
.IP result:
.cblk
\ fun="vertical"
.cble
.PP
Not only does this call go to the vertical function, but
it is in a vertical context.
If only a horizontal function is defined, then that is the one which is called,
even if the call is the only element in the line. This takes place in a
horizontal character-matching context, which requires a line of input which can
be traversed:
Example:
.IP code:
.cblk
\ @(define which (x))@(bind x "horizontal")@(end)
@(which fun)
.cble
.IP data:
.cblk
\ ABC
.cble
.IP result:
.cblk
\ [query fails]
.cble
.PP
The query fails because since
.code @(which fun)
is in horizontal mode,
it matches characters in a line. Since the function body consists
only of
.code @(bind ...)
which doesn't match any characters, the function
call requires an empty line to match. The line
.code ABC
is not empty,
and so there is a matching failure. The following
example corrects this:
Example:
.IP code:
.cblk
\ @(define which (x))@(bind x "horizontal")@(end)
@(which fun)
.cble
.IP data:
.cblk
\ [empty line]
.cble
.IP result:
.cblk
\ fun="horizontal"
.cble
.PP
A call made in a clearly horizontal context will prefer the
horizontal function, and only fall back on the vertical one
if the horizontal one doesn't exist. (In this fall-back case,
the vertical function is called with empty data; it is useful
for calling vertical functions which process arguments and
produce values.)
In the next example, the call is followed by trailing material,
placing it in a horizontal context. Leading material will
do the same thing:
Example:
.IP code:
.cblk
\ @(define which (x))@(bind x "horizontal")@(end)
@(define which (x))
@(bind x "vertical")
@(end)
@(which fun)B
.cble
.IP data:
.cblk
\ B
.cble
.IP result:
.cblk
\ fun="horizontal"
.cble
.PP
.SS* Local Variables
As described earlier, variables bound in a function body which are not
parameters of the function are discarded when the function returns. However,
that, by itself, doesn't make these variables local, because pattern functions
have visibility to all variables in their calling environment. If a variable
.code x
exists already when a function is called, then an attempt to bind it inside a
function may result in a failure. The
.code local
directive must be used in a
pattern function to list which variables are local.
Example:
.cblk
@(define path (path))@\e
@(local x y)@\e
@(cases)@\e
(@(path x))@(path y)@(bind path `(@x)@y`)@\e
@(or)@\e
@{x /[.,;'!?][^ \et\ef\ev]/}@(path y)@(bind path `@x@y`)@\e
@(or)@\e
@{x /[^ .,;'!?()\et\ef\ev]/}@(path y)@(bind path `@x@y`)@\e
@(or)@\e
@(bind path "")@\e
@(end)@\e
@(end)
.cble
This is a horizontal function which matches a path, which lands into four
recursive cases. A path can be parenthesized path followed by a path; it can be
a certain character followed by a path, or it can be empty
This function ensures that the variables it uses internally,
.code x
and
.codn y ,
do not have anything to do with any inherited bindings for
.code x
and
.codn y .
Note that the function is recursive, which cannot work without
.code x
and
.code y
being local, even if no such bindings exist prior to the top-level invocation of the
function. The invocation
.code @(path x)
causes
.code x
to be bound, which is
visible inside the invocation
.codn "@(path y)" ,
but that invocation needs to have its own binding of
.code x
for local use.
.SS* Nested Functions
Function definitions may appear in a function. Such definitions
are visible in all functions which are invoked from the body
(and not necessarily enclosed in the body). In other words, the
scope is dynamic, not lexical. Inner definitions shadow outer
definitions. This means that a caller can redirect the function
calls that take place in a callee, by defining local functions
which capture the references.
Example:
.IP code:
.cblk
\ @(define which)
@ (fun)
@(end)
@(define fun)
@ (output)
toplevel fun!
@ (end)
@(end)
@(define callee)
@ (define fun)
@ (output)
local fun!
@ (end)
@ (end)
@ (which)
@(end)
@(callee)
@(which)
.cble
.IP output:
.cblk
\ local fun!
toplevel fun!
.cble
.PP
Here, the function
.code which
is defined which calls
.codn fun .
A toplevel definition of
.code fun
is introduced which
outputs
.strn "toplevel fun!" .
The function
.code callee
provides its own local
definition of
.code fun
which outputs
.str "local fun!"
before calling
.codn which .
When
.code callee
is invoked, it calls
.codn which ,
whose
.code @(fun)
call is routed to callee's
local definition. When
.code which
is called directly from the top level, its
.code fun
call goes to the toplevel definition.
.SH* MODULARIZATION
.dirs load include
The syntax of the
.code load
and
.code include
directives is:
.cblk
.mets @(load << expr )
.mets @(include << expr )
.cble
Where
.meta expr
evaluates to a string giving the path of the file to load.
Unless the path is absolute, it is interpreted relative to the directory of the
source file from which the
.code @(load)
syntax was read. If there was no such
source file (for instance, the script was read from standard input),
then it is resolved relative to the current working directory.
If the file cannot be opened, then the
.code .txr
suffix is added and another
attempt is made. Thus load expressions need not refer to the suffix.
In the future, additional suffixes may be searched (compiled versions
of a file).
The two directives differ as follows. The action of
.code load
is not performed immediately but at evaluation time. Evaluation time
occurs after a \*(TX program is read from beginning to end and parsed.
The action of
.code include
is performed immediately, as the code is being scanned and parsed.
That is to say, as the \*(TX parser encounters
.code @(include)
it processes it immediately. The included material is read and parsed, and its
syntax tree is substituted in place of the
.code include
directive. The parser then
continues processing the original file after the
.code include
directive.
Note: the
.code include
directive is useful for loading \*(TX files which contain Lisp macros
which are needed by the parent program. The parent program cannot use
.code load
to bring in macros because macros are required during expansion, which
takes place prior to evaluation time, whereas
.code load
doesn't execute until evaluation time.
See also: the
.code *self-path*
and
.code stdlib
variables in \*(TL.
.SH* OUTPUT
.SS* Introduction
A \*(TX query may perform custom output. Output is performed by
.code output
clauses,
which may be embedded anywhere in the query, or placed at the end. Output
occurs as a side effect of producing a part of a query which contains an
.code @(output)
directive, and is executed even if that part of the query ultimately
fails to find a match. Thus output can be useful for debugging.
An
.code output
clause specifies that its output goes to a file, pipe, or (by
default) standard output. If any output clause is executed whose destination is
standard output, \*(TX makes a note of this, and later, just prior to
termination, suppresses the usual printing of the variable bindings or the word
false.
.dir output
The syntax of the
.code @(output)
directive is:
.cblk
.mets @(output [ < destination ] { < bool-keyword | < keyword < value }* )
.
. one or more output directives or lines
.
@(end)
.cble
The optional
.meta destination
is a string which gives the path name of
a file to open for output. If the name is
.code -
it instead denotes standard output, and if it begins with
.code !
then the rest of the shell is treated as a shell command
to which the output is piped.
The destination may be specified as a variable
which holds text, as a string literal or as a quasiliteral
The keyword list consists of a mixture of boolean keywords which
do not have an argument, or keywords with arguments.
The following boolean keywords are supported:
.coIP :nothrow
The output directive throws an exception if the output destination
cannot be opened, unless the
.code :nothrow
keyword is present, in which
case the situation is treated as a match failure.
Note that since command pipes are processes that report errors
asynchronously, a failing command will not throw an immediate exception that
can be suppressed with
.codn :nothrow .
This is for synchronous errors, like
trying to open a destination file, but not having permissions, etc.
.coIP :append
This keyword is meaningful for files, specifying append mode: the output is to
be added to the end of the file rather than overwriting the file.
The following value keywords are supported by
.codn :append :
.coIP :filter
The argument can be a symbol, which specifies a filter to be applied to
the variable substitutions occurring within the
.code output
clause.
The argument can also be a list of filter symbols, which specifies
that multiple filters are to be applied, in left to right order.
See the later sections Output Filtering below, and The Deffilter Directive.
.coIP :into
The argument of
.code :into
is a symbol which denotes a variable.
The output will go into that variable. If the variable is unbound,
it will be created. Otherwise, its contents are overwritten
unless the
.code :append
keyword is used. If
.code :append
is used, then
the new content will be appended to the previous content of
the variable, after flattening the content to a list,
as if by the
.code flatten
directive.
.coIP :named
The argument of
.code :named
is a symbol which denotes a variable.
The file or pipe stream which is opened for the output is
stored in this variable, and is not closed at the end of the
output block. This allows a subsequent output block to continue
output on the same stream, which is possible using the
next two keywords,
.code :continue
or
.codn :finish .
A new binding is established for the variable, even if it
already has an existing binding.
.coIP :continue
A destination should not be specified if
.code :continue
is used. The argument of
.code :continue
is an expression, such as a variable name, that evaluates to a
stream object. That stream object is used for the output block.
At the end of the output block, the stream is flushed, but not
closed. A usage example is given in the documentation for the Close Directive
below.
.coIP :finish
A destination should not be specified if
.code :finish
is used. The argument of
.code :finish
is an expression, such as a variable name, that evaluates to a
stream object. That stream object is used for the output block.
At the end of the output block, the stream is closed.
An example is given in the documentation for the Close Directive
below.
.SS* Output Text
Text in an output clause is not matched against anything, but is output
verbatim to the destination file, device or command pipe.
.SS* Output Variables
Variables occurring in an output clause do not match anything; instead
their contents are output.
A variable being output can be any object. If it is of a type other
than a list or string, it will be converted to a string as if by the
.code tostring
function in \*(TL.
A list is converted to a string in a special way: the elements are
individually converted to a string and then they are catenated together.
The default separator string is a single space: an alternate separation
can be specified as an argument in the brace substitution syntax.
Empty lists turn into an empty string.
Lists may be output within
.code @(repeat)
or
.code @(rep)
clauses. Each nesting of
these constructs removes one level of nesting from the list variables
that it contains.
In an output clause, the
.cblk
.meti >> @{ name << number }
.cble
variable syntax generates fixed-width
field, which contains the variable's text. The absolute value of the
number specifies the field width. For instance
.code -20
and
.code 20
both specify a field
width of twenty. If the text is longer than the field, then it overflows the
field. If the text is shorter than the field, then it is left-adjusted within
that field, if the width is specified as a positive number, and right-adjusted
if the width is specified as negative.
An output variable may specify a filter which overrides any filter established
for the output clause. The syntax for this is
.cblk
.meti @{NAME :filter << filterspec }.
.cble
The filter specification syntax is the same as in the output clause.
See Output Filtering below.
.SS* Output Variables: Indexing
Additional syntax is supported in output variables that does not appear
in pattern matching variables.
A square bracket index notation may be used to extract elements or
ranges from a variable, which works with strings, vectors and lists. Elements
are indexed from zero. This notation is only available in brace-enclosed
syntax, and looks like this:
.meIP <> @{name[ expr ]}
Extract the element at the position given by
.metn expr .
.meIP <> @{name[ expr1..expr2 ]}
Extract a range of elements from the position given by
.metn expr1 ,
up to
one position less than the position given by
.metn expr2 .
If the variable is a list, it is treated as a list substitution,
exactly as if it were the value of an unsubscripted list variable.
The elements of the list are converted to strings and catenated
together wit ha separator string between them, the default one being
a single space.
An alternate character may be given as a string argument in the brace
notation.
.PP
Example:
.cblk
@(bind a ("a" "b" "c" "d"))
@(output)
@{a[1..3] "," 10}
@(end)
.cble
The above produces the text
.str b,c
in a field
.code 10
spaces wide. The
.code [1..3]
argument extracts a range of
.codn a ;
the
.str ","
argument specifies an alternate
separator string, and
.code 10
specifies the field width.
.SS* Output Substitutions
The brace syntax has another syntactic and semantic extension in
.code output
clauses. In place
of the symbol, an expression may appear. The value of that expression
is substituted.
Example:
.cblk
@(bind a "foo")
@(output)
@{`@a:` -10}
.cble
Here, the quasiliteral expression
.code `@a:`
is evaluated, producing the string
.strn foo: .
This string is printed right-adjusted in a
.code 10
character field.
.dir repeat
The
.code repeat
directive generates repeated text from a "boilerplate",
by taking successive elements from lists. The syntax of repeat is
like this:
.cblk
@(repeat)
.
.
main clause material, required
.
.
special clauses, optional
.
.
@(end)
.cble
.code repeat
has four types of special clauses, any of which may be
specified with empty contents, or omitted entirely. They are described
below.
.code repeat
takes arguments, also described below.
All of the material in the main clause and optional clauses
is examined for the presence of variables. If none of the variables
hold lists which contain at least one item, then no output is performed,
(unless the repeat specifies an
.code @(empty)
clause, see below).
Otherwise, among those variables which contain non-empty lists, repeat finds
the length of the longest list. This length of this list determines the number
of repetitions, R.
If the
.code repeat
contains only a main clause, then the lines of this clause is
output R times. Over the first repetition, all of the variables which, outside
of the repeat, contain lists are locally rebound to just their first item. Over
the second repetition, all of the list variables are bound to their second
item, and so forth. Any variables which hold shorter lists than the longest
list eventually end up with empty values over some repetitions.
Example: if the list
.code A
holds
.strn 1 ,
.str 2
and
.strn 3 ;
the list
.code B
holds
.strn A ,
.strn B ;
and the variable
.code C
holds
.strn X ,
then
.cblk
@(repeat)
>> @C
>> @A @B
@(end)
.cble
will produce three repetitions (since there are two lists, the longest
of which has three items). The output is:
.cblk
>> X
>> 1 A
>> X
>> 2 B
>> X
>> 3
.cble
The last line has a trailing space, since it is produced by
.strn "@A @B" ,
where
.code B
has an empty value. Since
.code C
is not a list variable, it
produces the same value in each repetition.
The special clauses are:
.coIP @(single)
If the
.code repeat
produces exactly one repetition, then the contents of this clause
are processed for that one and only repetition, instead of the main clause
or any other clause which would otherwise be processed.
.coIP @(first)
The body of this clause specifies an alternative body to be used for the first
repetition, instead of the material from the main clause.
.coIP @(last)
The body of this clause is used instead of the main clause for the last
repetition.
.coIP @(empty)
If the repeat produces no repetitions, then the body of this clause is output.
If this clause is absent or empty, the repeat produces no output.
.coIP "@(mod n m)"
The forms
.code n
and
.code m
are expressions that evaluate to integers. The value of
.code m
should be nonzero. The clause denoted this way is active if the repetition
modulo
.code m
is equal to
.codn n .
The first repetition is numbered zero.
For instance the clause headed by
.code @(mod 0 2)
will be used on repetitions
0, 2, 4, 6, ... and
.code @(mod 1 2)
will be used on repetitions 1, 3, 5, 7, ...
.coIP "@(modlast n m)"
The meaning of
.code n
and
.code m
is the same as in
.codn "@(mod n m)" ,
but one more condition
is imposed. This clause is used if the repetition modulo
.code m
is equal to
.codn n ,
and if it is the last repetition.
.PP
The precedence among the clauses which take an iteration is:
.codn "single > first > mod > modlast > last > main" .
That is if two or more of these
clauses can apply to a repetition, then the leftmost one in this precedence
list applies. For instance, if there is just a single repetition, then any of
these special clause types can apply to that repetition, since it is the only
repetition, as well as the first and last one. In this situation, if there is a
.code @(single)
clause present, then the repetition is processed using that clause.
Otherwise, if there is a
.code @(first)
clause present, that clause is used. Failing
that,
.code @(mod)
is used if there is such a clause and its numeric conditions
are satisfied. If there isn't, then
.code @(modlast)
clauses are considered, and if there
are none, or none of them activate, then
.code @(last)
is considered. Finally if none
of all these clauses are present or apply, then the repetition is processed
using the main clause.
Repeat supports arguments.
.cblk
.mets @(repeat [:counter << symbol ] [:vars <> ( symbol *)])
.cble
The
.code :counter
argument designates a symbol which will behave as an integer
variable over the scope of the clauses inside the repeat. The variable provides
access to the repetition count, starting at zero, incrementing with each
repetition.
The
.code :vars
argument specifies a list of variables. The repeat directive
will pick out from this list those variables which have bindings.
It will assume that all these variables occur in the repeat block and
are to be iterated. This syntax is needed for situations in which
.code @(repeat)
is not able to deduce the existence of a variable in the block.
It does not dig very deeply to discover variables, and does not "see"
variables that are referenced via embedded \*(TL expressions.
For instance, the following produces no output:
.cblk
@(bind list ("a" "b" "c"))
@(output)
@(repeat)
@(format nil "<~a>" list)
@(end)
@(end)
.cble
Although the list variable appears in the repeat block, it is embedded
in a \*(TL construct. That construct will never be evaluated because
no repetitions take place: the repeat construct doesn't find any variables
and so doesn't iterate. The remedy is to provide a little help via
the :vars parameter:
.cblk
@(bind list ("a" "b" "c"))
@(output)
@(repeat :vars (list))
@(format nil "<~a>" list)
@(end)
@(end)
.cble
Now the repeat block iterates over list and the output is:
.cblk
<a>
<b>
<c>
.cble
.coSS Nested @ repeat directives
If a
.code repeat
clause encloses variables which hold multidimensional lists,
those lists require additional nesting levels of repeat (or rep).
It is an error to attempt to output a list variable which has not been
decimated into primary elements via a repeat construct.
Suppose that a variable
.code X
is two-dimensional (contains a list of lists).
.code X
must be twice nested in a
.codn repeat .
The outer repeat will traverse the lists
contained in
.codn X .
The inner repeat will traverse the elements of each of these
lists.
A nested repeat may be embedded in any of the clauses of a repeat,
not only the main clause.
.dir rep
The
.code rep
directive is similar to
.codn repeat .
Whereas
.code repeat
is line oriented,
.code rep
generates material within a line. It has all the same clauses,
but everything is specified within one line:
.cblk
@(rep)... main material ... .... special clauses ...@(end)
.cble
More than one
.code @(rep)
can occur within a line, mixed with other material.
A
.code @(rep)
can be nested within a
.code @(repeat)
or within another
.codn @(rep) .
Also,
.code @(rep)
accepts the same
.code :counter
and
.code :vars
arguments.
.coSS @ repeat and @ rep Examples
Example 1: show the list
.code L
in parentheses, with spaces between
the elements, or the word
.code EMPTY
if the list is empty:
.cblk
@(output)
@(rep)@L @(single)(@L)@(first)(@L @(last)@L)@(empty)EMPTY@(end)
@(end)
.cble
Here, the
.code @(empty)
clause specifies
.codn EMPTY .
So if there are no repetitions,
the text
.code EMPTY
is produced. If there is a single item in the list
.codn L ,
then
.code @(single)(@L)
produces that item between parentheses. Otherwise
if there are two or more items, the first item is produced with
a leading parenthesis followed by a space by
.code @(first)(@L
and the last item is produced with a closing parenthesis:
.codn @(last)@L) .
All items in between are emitted with a trailing space by
the main clause:
.codn @(rep)@L .
Example 2: show the list L like Example 1 above, but the empty list is
.codn () .
.cblk
@(output)
(@(rep)@L @(last)@L@(end))
@(end)
.cble
This is simpler. The parentheses are part of the text which
surrounds the
.code @(rep)
construct, produced unconditionally.
If the list
.code L
is empty, then
.code @(rep)
produces no output, resulting in
.codn () .
If the list
.code L
has one or more items, then they are produced with
spaces each one, except the last which has no space.
If the list has exactly one item, then the
.code @(last)
applies to it
instead of the main clause: it is produced with no trailing space.
.dir close
The syntax of the
.code close
directive is:
.cblk
.mets @(close << expr)
.cble
Where
.meta expr
evaluates to a stream. The
.code close
directive can be
used to explicitly close streams created using
.cblk
.meti @(output ... :named << var )
.cble
syntax, as an alternative to
.cblk
.meti @(output :finish << expr ).
.cble
Examples:
Write two lines to
.str foo.txt
over two output blocks using
a single stream:
.cblk
@(output "foo.txt" :named foo)
Hello,
@(end)
@(output :continue foo)
world!
@(end)
@(close foo)
.cble
The same as above, using
.code :finish
rather than
.code :continue
so that the stream is closed at the end of the second block:
.cblk
@(output "foo.txt" :named foo)
Hello,
@(end)
@(output :finish foo)
world!
@(end)
.cble
.SS* Output Filtering
Often it is necessary to transform the output to preserve its meaning
under the convention of a given data format. For instance, if a piece of
text contains the characters
.code <
or
.codn > ,
then if that text is being
substituted into HTML, these should be replaced by
.code <
and
.codn > .
This is what filtering is for. Filtering is applied to the contents of output
variables, not to any template text.
\*(TX implements named filters. Built-in filters are named by keywords, given
below. User-defined filters are possible, however. See notes on the deffilter
directive below.
Instead of a filter name, the syntax
.cblk
.meti (fun << name )
.cble
can be used. This
denotes that the function called
.meta name
is to be used as a filter.
This is discussed in the next section Function Filters below.
Built-in filters named by keywords:
.coIP :to_html
Filter text to HTML, representing special characters using HTML
ampersand sequences. For instance
.code >
is replaced by
.codn > .
.coIP :from_html
Filter text with HTML codes into text in which the codes are replaced by the
corresponding characters. For instance
.code >
is replaced by
.codn > .
.coIP :upcase
Convert the 26 lower case letters of the English alphabet to upper case.
.coIP :downcase
Convert the 26 upper case letters of the English alphabet to lower case.
.coIP :frompercent
Decode percent-encoded text. Character triplets consisting
of the
.code %
character followed by a pair of hexadecimal digits (case insensitive)
are are converted to bytes having the value represented by the hexadecimal
digits (most significant nybble first). Sequences of one or more such bytes are
treated as UTF-8 data and decoded to characters.
.coIP :topercent
Convert to percent encoding according to RFC 3986. The text is first converted
to UTF-8 bytes. The bytes are then converted back to text as follows.
Bytes in the range 0 to 32, and 127 to 255 (note: including the ASCII DEL),
bytes whose values correspond to ASCII characters which are listed by RFC 3986
as being in the "reserved set", and the byte value corresponding to the
ASCII
.code %
character are encoded as a three-character sequence consisting
of the
.code %
character followed by two hexadecimal digits derived from the
byte value (most significant nybble first, upper case). All other bytes
are converted directly to characters of the same value without any such
encoding.
.coIP :fromurl
Decode from URL encoding, which is like percent encoding, except that
if the unencoded
.code +
character occurs, it is decoded to a space character. Of course
.code %20
still decodes to space, and
.code %2B
to the
.code +
character.
.coIP :tourl
Encode to URL encoding, which is like percent encoding except that
a space maps to
.code +
rather than
.codn %20 .
The
.code +
character, being in the
reserved set, encodes to
.codn %2B .
.coIP :tonumber
Converts strings to numbers. Strings that contain a period,
.code e
or
.code E
are converted to floating point as if by the Lisp function
.codn flo-str .
Otherwise they are converted to integer as if using
.code int-str
with a radix of 10.
Non-numeric junk results in the object
.codn nil .
.coIP :tointeger
Converts strings to integers as if using
.code int-str
with a radix of 10.
Non-numeric junk results in the object
.codn nil .
.coIP :tofloat
Converts strings to floating-point values as if using the function
.codn flo-str .
Non-numeric junk results in the object
.codn nil.
.coIP :hextoint
Converts strings to integers as if using
.code int-str
with a radix of 16.
Non-numeric junk results in the object
.codn nil.
.PP
Examples:
To escape HTML characters in all variable substitutions occurring in an
output clause, specify
.code :filter :to_html
in the directive:
.cblk
@(output :filter :to_html)
...
@(end)
.cble
To filter an individual variable, add the syntax to the variable spec:
.cblk
@(output)
@{x :filter :to_html}
@(end)
.cble
Multiple filters can be applied at the same time. For instance:
.cblk
@(output)
@{x :filter (:upcase :to_html)}
@(end)
.cble
This will fold the contents of
.code x
to upper case, and then encode any special
characters into HTML. Beware of combinations that do not make sense.
For instance, suppose the original text is HTML, containing codes
like
.codn " .
The compound filter
.code (:upcase :from_html)
will not work
because
.code "
will turn to
.code "
which no longer be recognized by the
.code :from_html
filter, sonce the entity names in HTML codes
are case-sensitive.
Capture some numeric variables and convert to numbers:
.cblk
@date @time @temperature @pressure
@(filter :tofloat temperature pressure)
@;; temperature and pressure can now be used in calculations
.cble
.SS* Function Filters
A function can be used as a filter. For this to be possible, the function must
conform to certain rules:
.IP 1.
The function must take two special arguments, which may be followed
by additional arguments.
.IP 2.
When the function is called, the first argument will be bound to a string,
and the second argument will be unbound. The function must produce a
value by binding it to the second argument. If the filter is to be used
as the final filter in a chain, it must produce a string.
For instance, the following is a valid filter function:
.cblk
@(define foo_to_bar (in out))
@ (next :string in)
@ (cases)
foo
@ (bind out "bar")
@ (or)
@ (bind out in)
@ (end)
@(end)
.cble
This function binds the
.code out
parameter to
.str bar
if the in parameter
is
.strn foo ,
otherwise it binds the
.code out
parameter to a copy of the
.code in
parameter.
This is a simple filter.
To use the filter, use the syntax
.code (:fun foo_to_bar)
in place of a filter name.
For instance in the bind directive:
.cblk
@(bind "foo" "bar" :lfilt (:fun foo_to_bar))
.cble
The above should succeed since the left side is filtered from
.str foo
to
.strn bar ,
so that there is a match.
Of course, function filters can be used in a chain:
.cblk
@(output :filter (:downcase (:fun foo_to_bar) :upcase))
...
@(end)
.cble
Here is a split function which takes an extra argument
which specifies the separator:
.cblk
@(define split (in out sep))
@ (next :list in)
@ (coll)@(maybe)@token@sep@(or)@token@(end)@(end)
@ (bind out token)
@(end)
.cble
Furthermore, note that it produces a list rather than a string.
This function separates the argument in into tokens according to the
separator text carried in the variable
.codn sep .
Here is another function,
.codn join ,
which catenates a list:
.cblk
@(define join (in out sep))
@ (output :into out)
@ (rep)@in@sep@(last)@in@(end)
@ (end)
@(end)
.cble
Now here is these two being used in a chain:
.cblk
@(bind text "how,are,you")
@(output :filter (:fun split ",") (:fun join "-"))
@text
@(end)
.cble
Output:
.cblk
how-are-you
.cble
When the filter invokes a function, it generates the first two arguments
internally to pass in the input value and capture the output. The remaining
arguments from the
.code (:fun ...)
construct are also passed to the function.
Thus the string objects
.str ","
and
.str "-"
are passed as the
.code sep
argument to
.code split
and
.codn join .
Note that
.code split
puts out a list, which
.code join
accepts. So the overall filter
chain operates on a string: a string goes into split, and a string comes out of
join.
.dir deffilter
The
.code deffilter
directive allows a query to define a custom filter, which
can then be used in
.code output
clauses to transform substituted data.
This directive's syntax is illustrated in this example:
.IP code:
.cblk
\ @(deffilter rot13
("a" "n")
("b" "o")
("c" "p")
("d" "q")
("e" "r")
("f" "s")
("g" "t")
("h" "u")
("i" "v")
("j" "w")
("k" "x")
("l" "y")
("m" "z")
("n" "a")
("o" "b")
("p" "c")
("q" "d")
("r" "e")
("s" "f")
("t" "g")
("u" "h")
("v" "i")
("w" "j")
("x" "k")
("y" "l")
("z" "m"))
@(collect)
@line
@(end)
@(output :filter rot13)
@(repeat)
@line
@(end)
@(end)
.cble
.IP data:
.cblk
\ hey there!
.cble
.IP output:
.cblk
\ url gurer!
.cble
.PP
The
.code deffilter
symbol must be followed by the name of the filter to be defined,
followed by forms which evaluate to lists of strings. Each list must
be at least two elements long and specifies one or more texts which are mapped
to a replacement text. For instance, the following specifies a telephone keypad
mapping from upper case letters to digits.
.cblk
@(deffilter alpha_to_phone ("E" "0")
("J" "N" "Q" "1")
("R" "W" "X" "2")
("D" "S" "Y" "3")
("F" "T" "4")
("A" "M" "5")
("C" "I" "V" "6")
("B" "K" "U" "7")
("L" "O" "P" "8")
("G" "H" "Z" "9"))
@(deffilter foo (`@a` `@b`) ("c" `->@d`))
@(bind x ("from" "to"))
@(bind y ("---" "+++"))
@(deffilter sub x y)
.cble
The last deffilter above equivalent to
.codn "@(deffilter sub ("from" "to") ("---" "+++"))" .
Filtering works using a longest match algorithm. The input is scanned from left
to right, and the longest piece of text is identified at every character
position which matches a string on the left hand side, and that text is
replaced with its associated replacement text. The scanning then continues
at the first character after the matched text.
If none of the strings matches at a given character position, then that
character is passed through the filter untranslated, and the scan continues at
the next character in the input.
Filtering is not in-place but rather instantiates a new text, and so
replacement text is not re-scanned for more replacements.
If a filter definition accidentally contains two or more repetitions of the
same left hand string with different right hand translations, the later ones
take precedence. No warning is issued.
.dir filter
The syntax of the
.code filter
directive is:
.cblk
@(filter FILTER { VAR }+ )
.cble
A filter is specified, followed by one or more variables whose values
are filtered and stored back into each variable.
Example: convert
.codn a ,
.codn b ,
and
.code c
to upper case and HTML encode:
.cblk
@(filter (:upcase :to_html) a b c)
.cble
.SH* EXCEPTIONS
.SS* Introduction
The exceptions mechanism in \*(TX is another
disciplined form of non-local transfer, in addition to the blocks
mechanism (see BLOCKS above). Like blocks, exceptions provide a construct
which serves as the target for a dynamic exit. Both blocks and exceptions
can be used to bail out of deep nesting when some condition occurs.
However, exceptions provide more complexity. Exceptions are useful for
error handling, and \*(TX in fact maps certain error situations to exception
control transfers. However, exceptions are not inherently an error-handling
mechanism; they are a structured dynamic control transfer mechanism, one
of whose applications is error handling.
An exception control transfer (simply called an exception) is always identified
by a symbol, which is its type. Types are organized in a subtype-supertype
hierarchy. For instance, the
.code file-error
exception type is a subtype of the
.code error
type. This means that a file error is a kind of error. An exception
handling block which catches exceptions of type
.code error
will catch exceptions of
type
.codn file-error ,
but a block which catches
.code file-error
will not catch all
exceptions of type
.codn error .
A
.code query-error
is a kind of error, but not a kind of
.codn file-error .
The symbol
.code t
is the supertype of every type: every exception type
is considered to be a kind of
.codn t .
(Mnemonic:
.code t
stands for type, as in any type).
Exceptions are handled using
.code @(catch)
clauses within a
.code @(try)
directive.
In addition to being useful for exception handling, the
.code @(try)
directive
also provides unwind protection by means of a
.code @(finally)
clause,
which specifies query material to be executed unconditionally when
the try clause terminates, no matter how it terminates.
.dir try
The general syntax of the
.code try
directive is
.cblk
@(try)
... main clause, required ...
... optional catch clauses ...
... optional finally clause
@(end)
.cble
A
.code catch
clause looks like:
.cblk
@(catch TYPE [ PARAMETERS ])
.
.
.
.cble
and also this simple form:
.cblk
@(catch)
.
.
.
.cble
which catches all exceptions, and is equivalent
to
.codn "@(catch t)" .
A
.code finally
clause looks like:
.cblk
@(finally)
...
.
.
.cble
The main clause may not be empty, but the catch and finally may be.
A try clause is surrounded by an implicit anonymous block (see BLOCKS section
above). So for instance, the following is a no-op (an operation with no effect,
other than successful execution):
.cblk
@(try)
@(accept)
@(end)
.cble
The
.code @(accept)
causes a successful termination of the implicit anonymous block.
Execution resumes with query lines or directives which follow, if any.
.code try
clauses and blocks interact. For instance, an
.code accept
from within
a try clause invokes a
.codn finally .
.IP code:
.cblk
\ @(block foo)
@ (try)
@ (accept foo)
@ (finally)
@ (output)
bye!
@ (end)
@ (end)
.cble
.IP output:
.cblk
\ bye!
.cble
.PP
How this works: the
.code try
block's main clause is
.codn "@(accept foo)" .
This causes
the enclosing block named
.code foo
to terminate, as a successful match.
Since the
.cod try
is nested within this block, it too must terminate
in order for the block to terminate. But the try has a
.code finally
clause,
which executes unconditionally, no matter how the try block
terminates. The
.code finally
clause performs some output, which is seen.
.coSS The @ finally clause
A
.code try
directive can terminate in one of three ways. The main clause
may match successfully, and possibly yield some new variable bindings.
The main clause may fail to match. Or the main clause may be terminated
by a non-local control transfer, like an exception being thrown or a block
return (like the block foo example in the previous section).
No matter how the
.code try
clause terminates, the
.code finally
clause is processed.
The
.code finally
clause is itself a query which binds variables, which leads to
questions: what happens to such variables? What if the
.code finally
block fails
as a query? As well as: what if a
.code finally
clause itself initiates a
control transfer? Answers follow.
Firstly, a
.code finally
clause will contribute variable bindings only if the main
clause terminates normally (either as a successful or failed match).
If the main clause of the
.code try
block successfully matches, then the
.code finally
block continues
matching at the next position in the data, and contributes bindings.
If the main clause fails, then the
.code finally
block tries to match at the same position where the main clause failed.
The overall
.code try
directive succeeds as a match if either the main clause
or the
.code finally
clause succeed. If both fail, then the try directive is
a failed match.
Example:
.IP code:
.cblk
\ @(try)
@a
@(finally)
@b
@(end)
@c
.cble
.IP data:
.cblk
\ 1
2
3
.cble
.IP result:
.cblk
\ a="1"
b="2"
c="3"
.cble
.PP
In this example, the main clause of the
.code try
captures line
.str 1
of the data as
variable
.codn a ,
then the finally clause captures
.str 2
as
.codn b ,
and then the query continues with the
.code @c
line after try block, so that
.code c
captures
.strn "3" .
Example:
.IP code:
.cblk
\ @(try)
hello @a
@(finally)
@b
@(end)
@c
.cble
.IP data:
.cblk
\ 1
2
.cble
.IP result:
.cblk
\ b="1"
c="2"
.cble
.PP
In this example, the main clause of the
.code try
fails to match, because
the input is not prefixed with
.srn "hello " .
However, the
.code finally
clause
matches, binding
.code b
to
.strn "1" .
This means that the try block is a successful
match, and so processing continues with
.code @c
which captures
.strn "2" .
When
.code finally
clauses are processed during a non-local return,
they have no externally visible effect if they do not bind variables.
However, their execution makes itself known if they perform side effects,
such as output.
A
.code finally
clause guards only the main clause and the
.code catch
clauses. It does not
guard itself. Once the finally clause is executing, the
.code try
block is no
longer guarded. This means if a nonlocal transfer, such as a block accept
or exception, is initiated within the finally clause, it will not re-execute
the
.code finally
clause. The
.code finally
clause is simply abandoned.
The disestablishment of blocks and
.code try
clauses is properly interleaved
with the execution of
.code finally
clauses. This means that all surrounding
exit points are visible in a
.code finally
clause, even if the
.code finally
clause
is being invoked as part of a transfer to a distant exit point.
The finally clause can make a control transfer to an exit point which
is more near than the original one, thereby "hijacking" the control
transfer. Also, the anonymous block established by the
.code try
directive
is visible in the
.code finally
clause.
Example:
.cblk
@(try)
@ (try)
@ (next "nonexistent-file")
@ (finally)
@ (accept)
@ (end)
@(catch file-error)
@ (output)
file error caught
@ (end)
@(end)
.cble
In this example, the
.code @(next)
directive throws an exception of type
.codn file-error ,
because the given file does not exist. The exit point for this exception is the
.code @(catch file-error)
clause in the outer-most
.code try
block. The inner block is
not eligible because it contains no catch clauses at all. However, the inner
try block has a finally clause, and so during the processing of this
exception which is headed for
.codn "@(catch file-error)" ,
the
.code finally
clause performs an anonymous
.codn accept .
The exit point for that
.code accept
is the anonymous block
surrounding the inner
.codn try .
So the original
transfer to the
.code catch
clause is thereby abandoned. The inner
.code try
terminates
successfully due to the
.codn accept ,
and since it constitutes the main clause of the outer try,
that also terminates successfully. The
.str "file error caught"
message is never printed.
.c1SS catch clauses
.code catch
clauses establish their associated
.code try
blocks as potential exit points for
exception-induced control transfers (called "throws").
A
.code catch
clause specifies an optional list of symbols which represent
the exception types which it catches. The
.code catch
clause will catch
exceptions which are a subtype of any one of those exception types.
If a
.code try
block has more than one
.code catch
clause which can match a given
exception, the first one will be invoked.
When a
.code catch
is invoked, it is of course understood that the main clause did
not terminate normally, and so the main clause could not have produced any
bindings.
.code catch
clauses are processed prior to
.codn finally .
If a
.code catch
clause itself throws an exception, that exception cannot
be caught by that same clause or its siblings in the same try block.
The
.code catch
clauses of that block are no longer visible at that point.
Nevertheless, the
.code catch
clauses are still protected by the finally block.
If a catch clause throws, or otherwise terminates, the
.code finally
block is still processed.
If a
.code finally
block throws an exception, then it is simply aborted;
the remaining directives in that block are not processed.
So the success or failure of the
.code try
block depends on the behavior of the
.code catch
clause or the
.ocde finally
clause, if there is one. If either of them succeed, then the try
block is considered a successful match.
Example:
.IP code:
.cblk
\ @(try)
@ (next "nonexistent-file")
@ x
@ (catch file-error)
@a
@(finally)
@b
@(end)
@c
.cble
.IP data:
.cblk
\ 1
2
3
.cble
.IP result:
.cblk
\ a="1"
b="2"
c="3"
.cble
.PP
Here, the
.code try
block's main clause is terminated abruptly by a
.code file-error
exception from the
.code @(next)
directive. This is handled by the
.code catch
clause, which binds variable
.code a
to the input line
.strn 1 .
Then the
.code finally
clause executes, binding
.code b
to
.strn 2 .
The
.code try
block then terminates successfully, and so
.code @c
takes
.strn "3" .
.coSS @ catch Clauses with Parameters
A
.code catch
clause may have parameters following the type name, like this:
.cblk
@(catch pair (a b))
.cble
To write a catch-all with parameters, explicitly write the
master supertype t:
.cblk
@(catch t (arg ...))
.cble
Parameters are useful in conjunction with
.codn throw .
The built-in
.code error
exceptions carry one argument, which is a string containing
the error message. Using
.codn throw ,
arbitrary parameters can be passed
from the throw site to the catch site.
.dir throw
The
.code throw
directive generates an exception. A type must be specified,
followed by optional arguments. For example,
.cblk
@(throw pair "a" `@file.txt`)
.cble
throws an exception of type
.codn pair ,
with two arguments, being
.str a
and the expansion of the quasiliteral
.codn `@file.txt` .
The selection of the target
.code catch
is performed purely using the type
name; the parameters are not involved in the selection.
Binding takes place between the arguments given in
.code throw
and the target
.codn catch .
If any
.code catch
parameter, for which a
.code throw
argument is given, is a bound
variable, it has to be identical to the argument, otherwise the catch fails.
(Control still passes to the
.codn catch ,
but the catch is a failed match).
.IP code:
.cblk
\ @(bind a "apple")
@(try)
@(throw e "banana")
@(catch e (a))
@(end)
.cble
.IP result:
.cblk
\ [query fails]
.cble
.PP
If any argument is an unbound variable, the corresponding parameter
in the
.code catch
is left alone: if it is an unbound variable, it remains
unbound, and if it is bound, it stays as is.
.IP code:
.cblk
\ @(try)
@(trow e "honda" unbound)
@(catch e (car1 car2))
@car1 @car2
@(end)
.cble
.IP data:
.cblk
\ honda toyota
.cble
.IP result:
.cblk
\ car1="honda"
car2="toyota"
.cble
.PP
If a
.code catch
has fewer parameters than there are throw arguments,
the excess arguments are ignored:
.IP code:
.cblk
\ @(try)
@(throw e "banana" "apple" "pear")
@(catch e (fruit))
@(end)
.cble
.IP result:
.cblk
\ fruit="banana"
.cble
.PP
If a
.code catch
has more parameters than there are throw arguments, the excess
parameters are left alone. They may be bound or unbound variables.
.IP code:
.cblk
\ @(try)
@(trow e "honda")
@(catch e (car1 car2))
@car1 @car2
@(end)
.cble
.IP data:
.cblk
\ honda toyota
.cble
.IP result:
.cblk
\ car1="honda"
car2="toyota"
.cble
.PP
A
.code throw
argument passing a value to a
.code catch
parameter which is unbound causes
that parameter to be bound to that value.
.code throw
arguments are evaluated in the context of the
.codn throw ,
and the bindings
which are available there. Consideration of what parameters are bound
is done in the context of the catch.
.IP code:
.cblk
\ @(bind c "c")
@(try)
@(forget c)
@(bind (a c) ("a" "lc"))
@(throw e a c)
@(catch e (b a))
@(end)
.cble
.IP result:
.cblk
\ c="c"
b="a"
a="lc"
.cble
.PP
In the above example,
.code c
has a toplevel binding to the string
.strn "c" ,
but then becomes unbound
via
.code forget
within the
.code try
construct, and rebound to the value
.strn lc .
Since the
.code try
construct is terminated by a
.codn throw ,
these modifications of the
binding environment are discarded. Hence, at the end of the query, variable
.code c
ends up bound to the original value
.strn c .
The
.code throw
still takes place
within the scope of the bindings set up by the
.code try
clause, so the values of
.code a
and
.code c
that are thrown are
.str a
and
.strn lc .
However, at the
.code catch
site, variable
.code a
does not have a binding. At that point, the binding to
.str a
established in
the
.code try
has disappeared already. Being unbound, the
.code catch
parameter
.code a
can take
whatever value the corresponding throw argument provides, so it ends up with
.strn lci .
.dir defex
The
.code defex
directive allows the query writer to invent custom exception types,
which are arranged in a type hierarchy (meaning that some exception types are
considered subtypes of other types).
Subtyping means that if an exception type
.code B
is a subtype of
.codn A ,
then every
exception of type
.code B
is also considered to be of type
.codn A .
So a catch for type
.code A
will also catch exceptions of type
.codn B .
Every type is a supertype of itself: an
.code A
is a kind of
.codn A .
This of course implies that every type is a subtype of itself
also. Furthermore, every type is a subtype of the type
.codn t ,
which has no
supertype other than itself. Type
.code nil
is is a subtype of every type, including
itself. The subtyping relationship is transitive also. If
.code A
is a subtype
of
.codn B ,
and
.code B
is a subtype of
.codn C ,
then
.code A
is a subtype of
.codn C .
.code defex
may be invoked with no arguments, in which case it does nothing:
.cblk
@(defex)
.cble
It may be invoked with one argument, which must be a symbol. This introduces a
new exception type. Strictly speaking, such an introduction is not necessary;
any symbol may be used as an exception type without being introduced by
.codn @(defex) :
.cblk
@(defex a)
.cble
Therefore, this also does nothing, other than document the intent to use
a as an exception.
If two or more argument symbols are given, the symbols are all introduced as
types, engaged in a subtype-supertype relationship from left to right.
That is to say, the first (leftmost) symbol is a subtype of the next one,
which is a subtype of the next one and so on. The last symbol, if it
had not been already defined as a subtype of some type, becomes a
direct subtype of the master supertype
.codn t .
Example:
.cblk
@(defex d e)
@(defex a b c d)
.cble
The first directive defines
.code d
as a subtype of
.codn e ,
and
.code e
as a subtype of
.codn t .
The second defines
.code a
as a subtype of
.codn b ,
.code b
as a subtype of
.codn c ,
and
.code c
as a subtype of
.codn d ,
which is already defined as a subtype of
.codn e .
Thus
.code a
is now a subtype of
.codn e .
The the above can be condensed to:
.cblk
@(defex a b c d e)
.cble
Example:
.IP code:
.cblk
\ @(defex gorilla ape primate)
@(defex monkey primate)
@(defex human primate)
@(collect)
@(try)
@(skip)
@(cases)
gorilla @name
@(throw gorilla name)
@(or)
monkey @name
@(throw monkey name)
@(or)
human @name
@(throw human name)
@(end)@#cases
@(catch primate (name))
@kind @name
@(output)
we have a primate @name of kind @kind
@(end)@#output
@(end)@#try
@(end)@#collect
.cble
.IP data:
.cblk
\ gorilla joe
human bob
monkey alice
.cble
.IP output:
.cblk
\ we have a primate joe of kind gorilla
we have a primate bob of kind human
we have a primate alice of kind monkey
.cble
.PP
Exception types have a pervasive scope. Once a type relationship is introduced,
it is visible everywhere. Moreover, the
.code defex
directive is destructive,
meaning that the supertype of a type can be redefined. This is necessary so
that something like the following works right:
.cblk
@(defex gorilla ape)
@(defex ape primate)
.cble
These directives are evaluated in sequence. So after the first one, the
.code ape
type has the type
.code t
as its immediate supertype. But in the second directive,
.code ape
appears again, and is assigned the
.code primate
supertype, while retaining
.code gorilla
as a subtype. This situation could be diagnosed as an
error, forcing the programmer to reorder the statements, but instead
\*(TX obliges. However, there are limitations. It is an error to define a
subtype-supertype relationship between two types if they are already connected
by such a relationship, directly or transitively. So the following
definitions are in error:
.cblk
@(defex a b)
@(defex b c)
@(defex a c)@# error: a is already a subtype of c, through b
@(defex x y)
@(defex y x)@# error: circularity; y is already a supertype of x.
.cble
.dir assert
The
.code assert
directive requires the remaining query or sub-query which follows it
to match. If the remainder fails to match, the assert directive throws an
exception. If the directive is simply
.cblk
@(assert)
.cble
Then it throws an assertion of type assert, which is a subtype of error.
The assert directive also takes arguments similar to the throw
directive. The following assert directive, if it triggers, will throw
an exception of type
.codn foo ,
with arguments
.code 1
and
.strn 2 :
.cblk
@(assert foo 1 "2")
.cble
Example:
.cblk
@(collect)
Important Header
----------------
@(assert)
Foo: @a, @b
@(end)
.cble
Without the assertion in places, if the
.code Foo: @a, @b
part does not
match, then the entire interior of the
.code @(collect)
clause fails,
and the collect continues searching for another match.
With the assertion in place, if the text
.str "Important Header"
and its
underline match, then the remainder of the collect body must
match, otherwise an exception is thrown. Now the program will not
silently skip over any Important Header sections due to a problem
in its matching logic. This is particularly useful when the matching is varied
with numerous cases, and they must all be handled.
There is a horizontal
.code assert
directive also. For instance:
.cblk
abc@(assert)d@x
.cble
asserts that if the prefix
.str abc
is matched, then it must be
followed by a successful match for
.strn "d@x" ,
or else an exception is thrown.
.SH* TXR LISP
The \*(TX language contains an embedded Lisp dialect called \*(TL.
This language is exposed in \*(TX in several ways.
Firstly, in any situation that calls for an expression, a Lisp
expression can be used, if it is preceded by the
.code @
character. The Lisp expression
is evaluated and its value becomes the value of that expression.
Thus, \*(TX directives are embedded in literal text using
.codn @ ,
and Lisp expressions
are embedded in directives using
.code @
also.
Secondly, the
.code @(do)
directive can be used for evaluating one or more Lisp
forms, such that their value is thrown away. This is useful for evaluating some
Lisp code for the sake of its side effect, such as defining a variable,
updating a hash table, et cetera.
Thirdly, the
.code @(require)
directive can be used to evaluate Lisp expressions
as part of the matching logic of the \*(TX pattern language. The return value
of the rightmost expression is examined. If it is nil, then the
.code @(require)
directive triggers a match failure. Otherwise, matching proceeds.
Lastly, \*(TL expressions can be evaluated via the
command line, using the
.code -e
and
.code -p
options.
.B
Examples:
Bind variable
.code a
to the integer 4:
.cblk
@(bind a @(+ 2 2))
.cble
Bind variable
.code b
to the standard input stream:
.cblk
@(bind a @*stdin*)
.cble
Define several Lisp functions inside
.codn @(do) :
.cblk
@(do
(defun add (x y) (+ x y))
(defun occurs (item list)
(cond ((null list) nil)
((atom list) (eql item list))
(t (or (eq (first list) item)
(occurs item (rest list)))))))
.cble
Trigger a failure unless previously bound variable
.code answer
is greater than 42:
.cblk
@(require (> (int-str answer) 42)
.cble
.SS* Overview
\*(TL is a small and simple dialect, like Scheme, but much more similar to
Common Lisp than Scheme. It has separate value and function binding namespaces,
like Common Lisp (and thus is a Lisp-2 type dialect), and represents boolean
.B true
and
.B false
with the symbols
.code t
and
.code nil
(note the case sensitivity of
identifiers denoting symbols!) Furthermore, the symbol
.code nil
is also the empty list, which terminates nonempty lists.
\*(TL has lexically scoped local variables and dynamic global variables,
similarly to Common Lisp, including the convention that
.code defvar
marks symbols for dynamic binding in local scopes. Lexical closures
are supported. Funtions are lexically scoped in \*(TL; they can be
defined in pervasive global environment using
.code defun
or in local scopes using
.code flet
and
.codn labels .
.SS* Additional Syntax
Much of the \*(TL syntax has been introduced in the previous sections of the
manual, since directive forms are based on it. There is some additional syntax
that is useful in \*(TL programming.
.SS* Symbol Tokens
The symbol tokens in \*(TL,
called a
.meta lident
(Lisp identifier) has a similar syntax to the
.meta bident
(braced identifier) in the \*(TX pattern language. It may consist of
all the same characters, as well as the
.code /
(slash) character which may not be used in a
.metn bident .
Thus a
.meta lident
may consist of these characters, in addition to letters and numbers:
.cblk
! $ % & * + - < = > ? \e _ ~ /
.cble
and of course, may not look like a number. A lone
.code /
is a symbol in \*(TL. The token
.code /abc/
is also a symbol, and not a regular expression, like it is in the braced
variable syntax. Within \*(TL, regular expressions are written with
a leading
.codn # .
.SS* Consing Dot
Unlike other major Lisp dialects, \*(TL allows a consing dot with no forms
preceding it. This construct simply denotes the form which follows the dot.
That is to say, the parser implements the following transformation:
.cblk
(. expr) -> expr
.cble
This is convenient in writing function argument lists that only take
variable arguments. Instead of the syntax:
.cblk
(defun fun args ...)
.cble
the following syntax can be used:
.cblk
(defun fun (. args) ...)
.cble
When a
.code lambda
form is printed, it is printed in the following style.
.cblk
(lambda nil ...) -> (lambda () ...)
(lambda sym ...) -> (lambda (. sym) ...)
(lambda (sym) ...) -> (lambda (sym) ...)
.cble
In no other circumstances is
.code nil
printed as
.codn () ,
or an atom
.code sym
as
.codn "(. sym)" .
.SS* Quote and Quasiquote
.meIP >> ' expr
The quote character in front of an expression is used for suppressing evaluation,
which is useful for forms that evaluate to something other than themselves.
For instance if
.code '(+ 2 2)
is evaluated, the value is the three-element list
.codn "(+ 2 2)" ,
whereas if
.code (+ 2 2)
is evaluated, the value is
.codn 4 .
Similarly, the value of
.code 'a
is the symbol
.code a
itself, whereas the value of
.code a
is the contents of the variable
.codn a .
.meIP >> ^ qq-template
The caret in front of an expression is a quasiquote. A quasiquote is like
a quote, but with the possibility of substitution of material.
Under a quasiquote, form is considered to be a quasiquote template. The template
is considered to be a literal structure, except that it may contain
the notations
.cblk
.meti >> , expr
.cble
and
.cblk
.meti >> ,* expr
.cble
which denote non-constant parts.
A quasiquote gets translated into code which, when evaluated, constructs
the structure implied by
.metn qq-template ,
taking into account the unquotes and splices.
A quasiquote also processes nested quasiquotes specially.
If
.meta qq-template
does not contain any unquotes or splices (which match its
level of nesting), or is simply an atom, then
.cblk
.meti >> ^ qq-template
.cble
is equivalent to
.cblk
.meti >> ' qq-template .
.cble
in other words, it is like an ordinary quote. For instance
.code ^(a b ^(c ,d))
is equivalent to
.codn '(a b ^(c ,d)) .
Although there is an unquote ,d it
belongs to the inner quasiquote
.codn ^(c ,d) ,
and the outer quasiquote does not have
any unquotes of its own, making it equivalent to a quote.
Dialect note: in Common Lisp and Scheme,
.code ^form
is written
.codn `form ,
and
quasiquotes are also informally known as backquotes. In \*(TX, the backquote
character
.code `
used for quasi string literals.
.meIP >> , expr
The comma character is used within a
.meta qq-template
to denote an unquote. Whereas
the quote suppresses evaluation, the comma introduces an exception: an element
of a form which is evaluated. For example, list
.code ^(a b c ,(+ 2 2) (+ 2 2))
is the list
.codn (a b c 4 (+ 2 2)) .
Everything
in the quasiquote stands for itself, except for the
.code ,(+ 2 2)
which is evaluated.
Note: if a variable is called
.codn *x* ,
then the syntax
.code ,*x*
means
.codn ,* x* :
splice
the value of
.codn x* .
In this situation, whitespace between the comma and the
variable name should be used:
.codn , *x* .
.meIP >> ,* expr
The comma-star operator is used within quasiquote list to denote a splicing
unquote. The form which follows
.code ,*
must evaluate to a list. That list is spliced into
the structure which the quasiquote denotes. For example:
.code '(a b c ,*(list (+ 3 3) (+ 4 4) d))
evaluates to
.codn "(a b c 6 8 d)" .
The expression
.code (list (+ 3 3) (+ 4 4))
is evaluated to produce the list
.codn "(6 8)" ,
and this list is spliced into the quoted template.
Dialect note: in other Lisp dialects, the equivalent syntax is usually
.code ,@
(comma at). The
.code @
character already has an assigned meaning, so
.code *
is used.
.SS* Quasiquoting non-List Objects
Quasiquoting is supported over hash table and vector literals (see Vectors
and Hashes below). A hash table or vector literal can be quoted, like any
object, for instance:
.cblk
'#(1 2 3)
.cble
The
.code #(1 2 3)
literal is turned into a vector atom right in the \*(TX parser,
and this atom is being quoted: this is
.cblk
.meti (quote << atom )
.cble
syntactically, which evaluates to
.metn atom .
When a vector is quasi-quoted, this is a case of
.cblk
.meti >> ^ atom
.cble
which evaluates to
.metn atom .
A vector can be quasiquoted, for example:
.cblk
^#(1 2 3)
.cble
Of ourse, unquotes can occur within it.
.cblk
(let ((a 42))
^#(1 ,a 3)) ; value is #(1 42 3)
.cble
In this situation, the
.code ^#(...)
notation produces code which constructs a vector.
The vector in the following example is also a quasivector. It contains
unquotes, and though the quasiquote is not directly applied to it,
it is embedded in a quasiquote:
.cblk
(let ((a 42))
^(a b c #(d ,a))) ; value is (a b c #(d 42))
.cble
Hash table literals have two parts: the list of hash construction
arguments and the key-value pairs. For instance:
.cblk
#H((:equal-based) (a 1) (b 2))
.cble
where
.code (:equal-based)
is the list of construction arguments and the pairs
.code (a 1)
and
.code (b 2)
are the key/value entries. Hash literals may be quasiquoted. In
quasiquoting, the arguments and pairs are treated as separate syntax; it is not
one big list. So the following is not a possible way to express the above
hash:
.cblk
;; not supported: splicing across the entire syntax
(let ((hash-syntax '((:equal-based) (a 1) (b 2))))
^#H(,*hash-syntax))
.cble
This is correct:
.cblk
;; fine: splicing hash arguments and contents separately
(let ((hash-args '(:equal-based))
(hash-contents '((a 1) (b 2))))
^#H(,hash-args ,*hash-contents))
.cble
.SS* Quasiquoting combined with Quasiliterals
When a quasiliteral is embedded in a quasiquote, it is possible to use
splicing to insert material into the quasiliteral.
Example:
.cblk
(eval (let ((a 3)) ^`abc @,a @{,a} @{(list 1 2 ,a)}`))
-> "abc 3 3 1 2 3"
.cble
.SS* Vectors
.coIP "#(...)"
A hash token followed by a list denotes a vector. For example
.code #(1 2 a)
is a three-element vector containing the numbers
.code 1
and
.codn 2 ,
and the symbol
.codn a .
.SS* Hashes
.meIP <> #H(( hash-argument *) >> ( key << value )*)
The notation
.code #H
followed by a nested list syntax denotes a hash table literal.
The first item in the syntax is a list of keywords. These are the same
keywords as are used when calling the function hash to construct
a hash table. Allowed keywords are:
.codn :equal-based ,
.code :weak-keys
and
.codn :weak-values .
An empty list can be specified as
.code nil
or
.codn () ,
which defaults to a
hash table based on the
.code eql
function, with no weak semantics.
.coSS The @ .. notation
In \*(TL, there is a special "dotdot" notation consisting of a pair of dots.
This can be written between successive atoms or compound expressions, and is a
shorthand for cons.
That is to say,
.code A .. B
translates to
.codn "(cons A B)" ,
and so for instance
.code (a b .. (c d) e .. f . g)
means
.codn "(a (cons b (c d)) (cons e f) . g)" .
This is a syntactic sugar useful in certain situations in which a cons is used
to represent a pair of numbers or other objects. For instance, if
.code L
is a list, then
.code [L 1 .. 3]
computes a sublist of
.code L
consisting of
elements 1 through 2 (counting from zero).
Restrictions:
The notation must be enclosed in a list. For instance
.code a..b
is not, by itself, an
expression, but
.code (a..b)
is. This is important if Lisp data is being parsed from
a string or stream using the read function. If the data
.str "a..b"
is parsed, the symbol
.code a
will be extracted, leaving
.strn ..a ,
which, if parsed,
produces a syntax error since it consists of a "dotdot" token followed by
a symbol, which is not valid syntax.
The notation cannot occur in the dot position; that is, the syntax
.code (a . b .. c)
is invalid. The dotdot operator can only be used between the non-dot-position
elements of a list.
.SS* The DWIM Brackets
\*(TL has a square bracket notation. The syntax
.code [...]
is a shorthand
way of writing
.codn "(dwim ...)" .
The
.code []
syntax is useful for situations
where the expressive style of a Lisp-1 dialect is useful.
For instance if
.code foo
is a variable which holds a function object, then
.code [foo 3]
can be used to call it, instead of
.codn "(call foo 3)" .
If foo is a vector, then
.code [foo 3]
retrieves the fourth element, like
.codn "(vecref foo 3)" .
Indexing over lists,
strings and hash tables is possible, and the notation is assignable.
Furthermore, any arguments enclosed in
.code []
which are symbols are treated
according to a modified namespace lookup rule.
More details are given in the documentation for the
.code dwim
operator.
.SS* Compound Forms
In \*(TL, there are two types of compound forms: the Lisp-2 style
compound forms, denoted by ordinary lists that are expressed with parentheses.
There are Lisp-1 style compound forms denoted by the DWIM Brackets, discussed
in the previous section.
The first position of an ordinary Lisp-2 style compound form, is expected to
have a function or operator name. Then arguments follow. There may
also be an expression in the dotted position, if the form is a function call.
If the form is a function call then the arguments are evaluated. If any of the
arguments are symbols, they are treated according to Lisp-2 namespacing rules.
Additionally, if there is an expression in the dotted position, it is also
evaluated. It should evaluate to a sequence: a list, vector or string. The
elements of the sequence generate additional arguments for the function
call. Note, however, that a compound form cannot be used in the dot position,
for obvious reasons, namely that
.code (a b c . (foo z))
does not mean that there is
a compound form in the dot position, but denotes an alternate spelling for
.codn "(a b c foo z)" ,
where foo behaves as a variable.
The DWIM brackets are similar, except that the first position is an arbitrary
expression which is evaluated according to the same rules as the remaining
positions. The first expression must evaluate to a function, or else to some
other object for which the DWIM syntax is defined, such as a vector, string,
list or hash. Operators are not supported. The dotted syntax for application
of additional arguments from a list or vector is supported in the DWIM
brackets just like in the parentheses.
Examples:
.cblk
;; a contains 3
;; b contains 4
;; c contains #(5 6 7)
;; s contains "xyz"
(foo a b . c) ;; calls (foo 3 4 5 6 7)
(foo a) ;; calls (foo 3)
(foo . s) ;; calls (foo #\ex #\ey #\ez)
[foo a b . c] ;; calls (foo 3 4 5 6 7)
[c 1] ;; indexes into vector #(5 6 7) to yield 6
.cble
Dialect Note:
In some other Lisp dialects, the improper list syntax is not supported;
a function called apply (or similar) must be used for application even if
the expression which gives the trailing arguments is a symbol. Moreover,
applying sequences other than lists is not supported.
.SS* Regular Expressions
In \*(TL, the
.code /
character can occur in symbol names, and the
.code /
token
is a symbol. Therefore the
.code /regex/
syntax is not used for denoting regular expressions; rather, the
.code #/regex/
syntax is used.
.coSS Generalization of List Accessors
In ancient Lisp in the 1960's, it was not possible to apply the operations
.code car
and
.code cdr
to the
.code nil
symbol (empty list), because it is not a
.code cons
cell. In
the InterLisp dialect, this restriction was lifted: these operations were
extended to accept
.code nil
(and return
.codn nil ).
The convention was adopted in
other Lisp dialects such as MacLisp and eventually in Common Lisp. Thus there
exists an object which is not a cons, yet which takes
.code car
and
.codn cdr .
In \*(TL, this relaxation is extended further. For the sake of convenience,
the operations
.code car
and
.codn cdr ,
are made to work with strings and vectors:
.cblk
(cdr "") -> nil
(car "") -> nil
(car "abc") -> #\ea
(cdr "abc") -> "bc"
(cdr #(1 2 3)) -> #(2 3)
(car #(1 2 3)) -> 1
.cble
The
.code ldiff
function is also extended in a special way. When the right parameter
is a string or vector, then it uses the equal equality test rather than eq
for detecting the tail of the list.
.cblk
(ldiff "abcd" "cd") -> (#\ea #\eb)
.cble
The
.code ldiff
operation starts with
.str "abcd"
and repeatedly applies
.code cdr
to produce
.str "bcd"
and
.strn "cd" ,
until the suffix is equal to the second argument:
.cblk
(equal "cd" "cd")
.cble
yields true.
Operations based on
.codn car ,
.code cdr
and
.codn ldiff ,
such as
.code keep-if
and
.code remq
extend to
strings and vectors.
Most derived list processing operations such as
.code remq
or
.code mapcar
obey the following
rule: the returned object follows the type of the leftmost input list object.
For instance, if one or more sequences are processed by
.codn mapcar ,
and the
leftmost one is a character string, the function is expected to return
characters, which are converted to a character string. However, in the
event that the objects produced cannot be assembled into that type of
sequence, a list is returned instead.
For example
.cblk
[mapcar list "ab" "12"]
.cble
returns
.codn "((#\ea #\eb) (#\e1 #\e2))" ,
because a string cannot hold lists of characters. However
.cblk
[mappend list "ab" "12"]
.cble
returns
.strn "a1b2" .
The lazy versions of these functions such as
.code mapcar*
do not have this behavior;
they produce lazy lists.
.SS* Callable Objects
In \*(TL, sequences (strings, vectors and lists) and hashes can be used
as functions everywhere, not just with the DWIM brackets. Sequences work
as one or two-argument functions. With a single argument, an element is
selected by position and returned. With two arguments, a range is extracted and
returned. Hashes also work as one or two argument functions, corresponding
to the arguments of the gethash function.
Moreover, when a sequence is used as a function of one argument, and the
argument is a cons cell rather than an integer, then the call becomes a
two-argument call in which the car and cdr of the cell are passed as separate
arguments. This allows for syntax like
.cblk
(call "abc" 0..1).
.cble
.B Example 1:
.cblk
(mapcar "abc" '(2 0 1)) -> (#\ec #\ea #\eb)
.cble
Here,
.code mapcar
treats the string
.str abc
as a function of one argument (since there
is one list argument). This function maps the indices
.codn 0 ,
.code 1
and
.code 2
to the
corresponding characters of string
.strn abc .
Through this function, the list of
integer indices
.code (2 0 1)
is taken to the list of characters
.codn (#\ec #\ea #\eb) .
.B Example 2:
.cblk
(call '(1 2 3 4) 1..3) -> (2 3)
.cble
Here, the shorthand
.code 1 .. 3
denotes
.codn (cons 1 3) .
A cons cell as an argument
to a sequence performs range extraction: taking a slice starting at
index 1, up to and not including index 3, as if by the call
.codn (sub '(1 2 3 4) 1 3) .
.B Example 3:
.cblk
(call '(1 2 3 4) '(0 2)) -> (1 2)
.cble
A list of indices applied to a sequence is equivalent to using the
select function, as if
.code (select '(1 2 3 4) '(0 2))
were called.
.SS* Special Variables
Similarly to Common Lisp, \*(TL is lexically scoped by default, but
also has dynamically scoped (a.k.a "special") variables.
When a variable is defined with
.codn defvar ,
a binding for the symbol is
introduced in the global name space, regardless of in what scope the
.code defvar
form occurs.
Furthermore, at the time the defvar form is evaluated, the symbol which
names the variable is tagged as special.
When a symbol is tagged as special, it behaves differently when it is used
in a lexical binding construct like
.codn let ,
and all other such constructs
such as function parameter lists. Such a binding is not the usual lexical
binding, but a "rebinding" of the global variable. Over the dynamic scope
of the form, the global variable takes on the value given to it by the
rebinding. When the form terminates, the prior value of the variable
is restored. (This is true no matter how the form terminates; even if by
an exception.)
Because of this "pervasive special" behavior of a symbol that has been
used as the name of a global variable, a good practice is to make global
variables have visually distinct names via the "earmuffs" convention:
beginning and ending the name with an asterisk.
Certain variables in \*(TX's library break this convention; however, they at
least have distinct prefixes, examples being example s-ifmt, log-emerg and
sig-hup.
Example:
.cblk
(defvar *x* 42) ;; *x* has a value of 42
(defun print-x ()
(format t "~a\en" *x*))
(let ((*x* "abc")) ;; this overrides *x*
(print-x)) ;; *x* is now "abc" and so that is printed
(print-x) ;; *x* is 42 again and so "42" is printed
.cble
.SH* TXR LISP OPERATOR AND FUNCTION LIBRARY
A compound expression with a symbol as its first element, if
intended to be evaluated, denotes either an operator invocation or a function
call. This depends on whether the symbol names an operator or a function.
When the form is an operator invocation, the interpretation of the meaning of
that form is under the complete control of that operator.
If the compound form is a function call, the remaining forms, if any, denote
argument expressions to the function. They are evaluated in left to right
order to produce the argument values, which are passed to the function. An
exception is thrown if there are not enough arguments, or too many. Programs
can define named functions with the defun operator
Some operators are macros. There exist predefined macros in the library, and
macro operators can also be user-defined using the macro-defining operator
.codn defmacro .
Operators that are not macros are called special operators.
Macro operators work as functions which are given the source code of the form.
They analyze the form, and translate it to another form which is substituted in
their place. This happens during a code walking phase called the expansion
phase, which is applied to Lisp code prior to evaluation. All macros are
expanded in the expansion phase, resulting in code which contains only
function calls and the executable forms of the operators.
(Special operators can also perform code transformations during the expansion
phase, but that is not considered macroexpansion, but rather an adjustment
of the representation of the operator into an required executable form.)
The following sections list all of the special operators, macros
and functions in \*(TL.
In these sections Syntax is indicated using these conventions:
.ie n \{\
. coIP <word>
A symbol in angle brackets
.\}
.el \{\
. coIP \f[5]word\f[]
A symbol in
.meta fixed-width-italic
font
.\}
denotes some syntactic unit: it
may be a symbol or compound form. The syntactic unit is explained
in the Description section.
.ie n \{\
.coIP {syntax}* <word>*
.\}
.el \{\
.coIP {syntax}* \f[5]word\f[]*
.\}
This indicates a repetition of zero or more of the given
syntax enclosed in the braces or syntactic unit.
.ie n \{\
.coIP {syntax}+ <word>+
.\}
.el \{\
.coIP {syntax}+ \f[5]word\f[]+
.\}
This indicates a repetition of one or more of the given
syntax enclosed in the braces or syntactic unit.
.ie n \{\
.coIP [syntax] [<word>]
.\}
.el \{\
.coIP [syntax] [\f[5]word\f[]]
.\}
Square brackets indicate optional syntax.
.coIP alternative1 | alternative2 | ... | alternativeN
Multiple syntactic variations allowed in one place are
indicated as bar-separated items.
.SS* Control Flow and Sequencing
.coNP Operators @ progn and @ prog1
.synb
.mets (progn << form *)
.mets (prog1 << form *)
.syne
.desc
The
.code progn
operator evaluates forms in order, and returns the value
of the last form. The return value of the form
.code (progn)
is
.codn nil .
The
.code prog1
operator evaluates forms in order, and returns the value
of the first form. The return value of the form
.code (prog1)
is
.codn nil .
Various other operators such as
.code let
also arrange for the evaluation
of a body of forms, the value of the last of which is returned.
These operators are said to feature an implicit
.codn progn .
.coNP Operator @ cond
.synb
.mets (cond >> {( test << form *)}*)
.syne
.desc
The
.code cond
operator provides a multi-branching conditional evaluation of
forms. Enclosed in the cond form are groups of forms expressed as lists.
Each group must be a list of at least one form.
The forms are processed from left to right as follows: the first form,
.metn test ,
in each group is evaluated. If it evaluates true, then the remaining
forms in that group, if any, are also evaluated. Processing then terminates and
the result of the last form in the group is taken as the result of cond.
If
.meta test
is the only form in the group, then result of
.meta test
is taken
as the result of
.codn cond .
If the first form of a group yields
.codn nil ,
then processing continues with the
next group, if any. If all form groups yield
.codn nil ,
then the cond form yields
.codn nil .
This holds in the case that the syntax is empty:
.code (cond)
yields
.codn nil .
.coNP Macros @, caseq @ caseql and @ casequal
.synb
.mets (caseq < test-form << normal-clause * <> [ else-clause ])
.mets (caseql < test-form << normal-clause * <> [ else-clause ])
.mets (caseqqual < test-form << normal-clause * <> [ else-clause ])
.syne
.desc
These three macros arrange for the evaluation of of
.metn test-form ,
whose value is then compared against the key or keys in each
.meta normal-clause
in turn.
When the value matches a key, then the remaining forms of
.meta normal-clause
are evaluated, and the value of the last form is returned; subsequent
clauses are not evaluated. When the value doesn't match any of the keys
of a
.meta normal-clause
then the next
.meta normal-clause
is tested.
If all these clauses are exhausted, and there is no
.metn else-clause ,
then the value nil is returned. Otherwise, the forms in the
.meta else-clause
are evaluated, and the value of the last one is returned.
The syntax of a
.meta normal-clause
takes on these two forms:
.cblk
.mets >> ( key << form *)
.cble
where
.meta key
may be an atom which denotes a single key, or else a list
of keys. There is a restriction that the symbol
.code t
may not be used
as
.metn key .
The form
.code (t)
may be used as a key to match that symbol.
The syntax of an
.meta else-clause
is:
.cblk
.mets (t << form *)
.cble
which resembles a form that is often used as the final clause
in the
.code cond
syntax.
The three forms of the case construct differ from what type of
test they apply between the value of
.meta test-form
and the keys.
The
.code caseq
macro generates code which uses the
.code eq
function's
equality. The
.code caseql
macro uses
.codn eql ,
and
.code casequal
uses
.codn equal .
.TP* Example
.cblk
(let ((command-symbol (casequal command-string
(("q" "quit") 'quit)
(("a" "add") 'add)
(("d" "del" "delete") 'delete)
(t 'unknown))))
...)
.cble
.coNP Operator/function @ if
.synb
.mets (if < cond < t-form <> [ e-form ])
.mets [if < cond < then <> [ else ]]
.syne
.desc
There exist both an
.code if
operator and an
.code if
function. A list form with the symbol
.code if
in the fist position is interpreted as an invocation of the
.code if
operator.
The function can be accessed using the DWIM bracket notation and in other
ways.
The
.code if
operator provides a simple two-way-selective evaluation control.
The
.meta cond
form is evaluated. If it yields true then
.meta t-form
is evaluated, and that form's return value becomes the return value of the
.codn if .
If
.meta cond
yields false, then
.meta e-form
is evaluated and its return value is taken to be that of
.codn if .
If
.meta e-form
is omitted, then the behavior is as if
.meta e-form
were specified as
.codn nil .
The
.code if
function provides no evaluation control. All of arguments
are evaluated from left to right. If the
.meta cond
argument is true, then it
returns the
.meta then
argument, otherwise it returns the value of the
.meta else
argument if present, otherwise it returns
.codn nil .
.coNP Operator/function @ and
.synb
.mets (and << form *)
.mets [and << arg *]
.syne
.desc
There exist both an
.code and
operator and an
.code and
function. A list form with the
symbol
.code and
in the fist position is interpreted as an invocation of the
operator. The function can be accessed using the DWIM bracket notation and in
other ways.
The
.code and
operator provides three functionalities in one. It computes the
logical "and" function over several forms. It controls evaluation (a.k.a.
"short-circuiting"). It also provides an idiom for the convenient substitution
of a value in place of
.code nil
when some other values are all true.
The
.code and
operator evaluates as follows. First, a return value is
established and initialized to the value
.codn t .
The
.metn form s,
if any, are
evaluated from left to right. The return value is overwritten with
the result of each form. Evaluation stops when all forms are exhausted,
or when
.code nil
is stored in the return value.
When evaluation stops, the operator yields the return value.
The
.code and
function provides no evaluation control; it receives all of its
arguments fully evaluated. If it is given no arguments, it returns
.codn t .
If it is given one or more arguments, and any of them are
.codn nil ,
it returns
.codn nil .
Otherwise it returns the value of the last argument.
.TP* Examples:
.cblk
(and) -> t
(and (> 10 5) (stringp "foo")) -> t
(and 1 2 3) -> 3 ;; short-hand for (if (and 1 2) 3).
.cble
.coNP Operator/function @ or
.synb
.mets (or << form *)
.mets [or << arg *]
.syne
.desc
There exist both an
.code or
operator and an
.code or
function. A list form with the
symbol
.code or
in the fist position is interpreted as an invocation of the
operator. The function can be accessed using the DWIM bracket notation and in
other ways.
The or operator provides three functionalities in one. It computes the
logical "or" function over several forms. It controls evaluation (a.k.a.
"short-circuiting"). The behavior of
.code or
also provides an idiom for the selection of the first non-nil value from a
sequence of forms.
The
.code or
operator evaluates as follows. First, a return value is
established and initialized to the value
.codn nil .
The
.metn form s,
if any,
are evaluated from left to right. The return value is overwritten
with the result of each
.metn form .
Evaluation stops when all forms are
exhausted, or when a true value is stored into the return value.
When evaluation stops, the operator yields the return value.
The
.code or
function provides no evaluation control; it receives all of its
arguments fully evaluated. If it is given no arguments, it returns
.codn nil .
If all of its arguments are
.codn nil ,
it also returns
.codn nil .
Otherwise, it
returns the value of the first argument which isn't
.codn nil .
.TP* Examples:
.cblk
(or) -> nil
(or 1 2) -> 1
(or nil 2) -> 2
(or (> 10 20) (stringp "foo")) -> t
.cble
.coNP Macros @ when and @ unless
.synb
.mets (when < expression << form *)
.mets (unless < expression << form *)
.syne
.desc
The when macro operator evaluates
.metn expression .
If
.meta expression
yields
true, and there are additional forms, then each
.meta form
is evaluated.
The value of the last form is becomes the result value of the when form.
If there are no forms, then the result is
.codn nil .
The
.code unless
operator is similar to when, except that it reverses the
logic of the test. The forms, if any, are evaluated if, and only if
.meta expression
is false.
.coNP Macros @ while and @ until
.synb
.mets (while < expression << form *)
.mets (until < expression << form *)
.syne
.desc
The
.code while
macro operator provides a looping construct. It evaluates
.metn expression .
If
.meta expression
yields
.codn nil ,
then the evaluation of the
.code while
form
terminates, producing the value
.codn nil .
Otherwise, if there are additional forms,
then each
.meta form
is evaluated. Next, evaluation returns to
.metn expression ,
repeating all of the previous steps.
The
.code until
macro operator is similar to while, except that the until form
terminates when
.meta expression
evaluates true, rather than false.
These operators arrange for the evaluation of all their enclosed forms
in an anonymous block. Any of the
.metn form s,
or
.metn expression ,
may use
the
.code return
operator to terminate the loop, and optionally to specify
a result value for the form.
The only way these forms can yield a value other than
.code nil
is if the
.code return
operator is used to terminate the implicit anonymous block,
and is given an argument, which becomes the result value.
.coNP Macros @ while* and @ until*
.synb
.mets (while* < expression << form *)
.mets (until* < expression << form *)
.syne
.desc
The
.code while*
and
.code until*
macros are similar, respectively, to the macros
.code while
and
.codn until .
They differ in one respect: they begin by evaluating the
.metn form -s
one time unconditionally, without first evaluating
.metn expression .
After this evaluation, the subsequent behavior is
like that of
.code while
or
.codn until .
Another way to regard the behavior is that that these forms execute
one iteration unconditionally, without evaluating the termination test prior to
the first iteration. Yet another view is that these constructs relocate the
test from the "top of the loop" to the "bottom of the loop".
.coNP Macro @ whilet
.synb
.mets (whilet >> ({ sym | >> ( sym << init-form )}+)
.mets \ \ << body-form *)
.syne
.desc
The
.code whilet
macro provides a construct which combines iteration with variable
binding.
The evaluation of the form takes place as follows. First, fresh bindings are
established for
.metn sym -s
as if by the
.code let*
operator.
It is an error for the list of variable bindings to be empty.
After the establishment of the bindings, the the value of the
.meta sym
is tested. If the value is
.codn nil ,
then
.code whilet
terminates. Otherwise,
.metn body-form -s
are evaluated in the scope of the variable bindings, and then
.code whilet
iterates from the beginning, again establishing fresh bindings for the
.metn sym -s,
and testing the value of the last
.metn sym .
All evaluation takes place in an anonymous block, which can be
terminated with the
.code return
operator. Doing so terminates the loop.
If the
.code whilet
loop is thus terminated by an explicit
.codn return ,
a return value can be specified. Under normal termination, the return value is
.codn nil .
.TP* Examples:
.cblk
;; read lines of text from *std-input* and print them,
;; until the end-of-stream condition:
(whilet ((line (get-line)))
(put-line line))
;; read lines of text from *std-input* and print them,
;; until the end-of-stream condition occurs or
;; a line is identical to the character string "end".
(whilet ((line (get-line))
(more (and line (not (equal line "end")))))
(put-line line))
.cble
.coNP Macros @ iflet and @ whenlet
.synb
.mets (iflet >> ({ sym | >> ( sym << init-form )}+)
.mets \ \ < then-form <> [ else-form ])
.mets (whenlet >> ({ sym | >> ( sym << init-form )}+)
.mets \ \ << body-form *])
.syne
.desc
The
.code iflet
and
.code whenlet
macros combine the variable binding of
.code let*
with conditional evaluation of
.code if
and
.codn when ,
respectively.
The evaluation of these forms takes place as follows. First, fresh bindings are
established for
.metn sym -s
as if by the
.code let*
operator.
It is an error for the list of variable bindings to be empty.
Then, the last variable's value is tested. If it is not
.code nil
then the test is true, otherwise false.
In the case of the
.code iflet
operator, if the test is true, the operator evaluates
.meta then-form
and yields its value. Otherwise the test is false, and if the
optional
.meta else-form
is present, that is evaluated instead and its value is returned.
If this form is missing, then
.code nil
is returned.
In the case of the
.code whenlet
operator, if the test is true, then the
.metn body-form -s,
if any, are evaluated. The value of the last one is
returned, otherwise
.code nil
if the forms are missing.
If the test is false, then evaluation of
.metn body-form -s
is skipped, and
.code nil
is returned.
.TP* Examples:
.cblk
;; dispose of foo-resource if present
(whenlet ((foo-res (get-foo-resource obj)))
(foo-shutdown foo-res)
(set-foo-resource obj nil))
;; Contrast with: above, using when and let
(let ((foo-res (get-foo-resource obj)))
(when foo-res
(foo-shutdown foo-res)
(set-foo-resource obj nil)))
;; print frobosity value if it exceeds 150
(whenlet ((fv (get-frobosity-value))
(exceeds-p (> fv 150)))
(format t "frobosity value ~a exceeds 150\en" fv))
.cble
.coNP Macro @ dotimes
.synb
.mets (dotimes >> ( var < count-form <> [ result-form ]) << body-form *)
.syne
.desc
The
.code dotimes
macro implements a simple counting loop.
.meta var
is established as a variable, and initialized to zero.
.meta count-form
is evaluated one time to produce a limiting value, which should be a number.
Then, if the value of
.meta var
is less than the limiting value, the
.metn body-form -s
are evaluated,
.meta var
is incremented by one, and the process repeats with a new comparison
of
.meta var
against the limiting value possibly leading to another evaluation of
the forms.
If
.meta var
is found to equal or exceed the limiting value, then the loop terminates.
When the loop terminates, its return value is
.code nil
unless a
.meta result-form
is present, in which case the value of that form specifies the return value.
.metn body-form -s
as well as
.meta result-form
are evaluated in the scope in which the binding of
.meta var
is visible.
.coNP Operator @ unwind-protect
.synb
.mets (unwind-protect < protected-form << cleanup-form *)
.syne
.desc
The
.cod unwind-protect
operator evaluates
.meta protected-form
in such a way that no matter how the execution of
.meta protected-form
terminates, the
.metn cleanup-form s
will be executed.
The
.metn cleanup-form s,
however, are not protected. If a
.meta cleanup-form
terminates via
some non-local jump, the subsequent
.metn cleanup-form s
are not evaluated.
.metn cleanup-form s
themselves can "hijack" a non-local control transfer such
as an exception. If a
.meta cleanup-form
is evaluated during the processing of
a dynamic control transfer such as an exception, and that
.meta cleanup-form
initiates its own dynamic control transfer, the original control transfer
is aborted and replaced with the new one.
.TP* Example:
.cblk
(block foo
(unwind-protect
(progn (return-from foo 42)
(format t "not reached!\en"))
(format t "cleanup!\en")))
.cble
In this example, the protected
.code progn
form terminates by returning from
block
.codn foo .
Therefore the form does not complete and so the
output
.str not reached!
is not produced. However, the cleanup form
executes, producing the output
.strn cleanup! .
.coNP Operator @ block
.synb
.mets (block < name << body-form *)
.syne
.desc
The
.code block
operator introduces a named block around the execution of
some forms. The
.meta name
argument must be a symbol. Since a block name is not
a variable binding, keyword symbols are permitted, and so are the symbols
.code t
and
.codn nil .
A block named by the symbol nil is slightly special: it is
understood to be an anonymous block.
Blocks in \*(TL have dynamic scope. This means that the following
situation is allowed:
.cblk
(defun func () (return-from foo 42))
(block foo (func))
.cble
The function can return from the
.code foo
block even though the
.code foo
block
does not lexically surround
.codn foo .
Thus blocks in \*(TL provide dynamic non-local returns, as well
as returns out of lexical nesting.
.TP* "Dialect Note:"
In Common Lisp, blocks are lexical. A separate mechanism consisting of
catch and throw operators performs non-local transfer based on symbols.
The \*(TL example:
.cblk
(defun func () (return-from foo 42))
(block foo (func))
.cble
is not allowed in Common Lisp, but can be transliterated to:
.cblk
(defun func () (throw 'foo 42))
(catch 'foo (func))
.cble
Note that foo is quoted in CL. This underscores the dynamic nature of
the construct.
.code throw
itself is a function and not an operator.
.coNP Operators @ return and @ return-from
.synb
.mets (return <> [ value ])
.mets (return-from < name <> [ value ])
.syne
.desc
The
.code return
operator must be dynamically enclosed within an anonymous
block (a block named by the symbol
.codn nil ).
It immediately terminates the
evaluation of the innermost anonymous block which encloses it, causing
it to return the specified value. If the value is omitted, the anonymous
block returns
.codn nil .
The
.code return-from
operator must be dynamically enclosed within a named block
whose name matches the
.meta name
argument. It immediately terminates the
evaluation of the innermost such block, causing it to return the specified
value. If the value is omitted, that block returns
.codn nil .
.TP* Example:
.cblk
(block foo
(let ((a "abc\en")
(b "def\en"))
(pprint a *stdout*)
(return-from foo 42)
(pprint b *stdout*)))
.cble
Here, the output produced is
.strn "abc" .
The value of
.code b
is not printed
because.
.code return-from
terminates block
.codn foo ,
and so the second pprint form is not evaluated.
.SS* Evaluation
.coNP Operator @ dwim
.synb
.mets (dwim << argument *)
.mets <> [ argument *]
.syne
.desc
The
.code dwim
operator's name is an acronym: DWIM may be taken to mean
"Do What I Mean", or alternatively, "Dispatch, in a Way that is
Intelligent and Meaningful".
The notation
.code [...]
is a shorthand equivalent to
.code (dwim ...)
and is usually the preferred way for writing
.code dwim
expressions.
The
.code dwim
operator takes a variable number of arguments, which are
all evaluated in the same way: the first argument is not evaluated differently
from the remaining arguments.
This means that the first argument isn't a function name, but an ordinary
expression which can simply compute a function object (or, more generally,
a callable object).
Furthermore, for those arguments of
.code dwim
which are symbols (after all
macro-expansion is performed on the arguments), the evaluation rules are
altered. For the purposes of resolving symbols to values, the function and
variable binding namespaces are considered to be merged into a single space,
creating a situation that is very similar to a Lisp-1 style dialect.
This special Lisp-1 evaluation is not recursively applied. All arguments of
.code dwim
which, after macro expansion, are not symbols are evaluated using the
normal Lisp-2 evaluation rules. Thus, the DWIM operator must be used
in every expression where the Lisp-1 rules for reducing symbols to
values are desired.
After macro expansion, the first argument of
.code dwim
may not be an operator such
as
.codn let ,
or the name of a macro. Prior to macroexpansion, any argument of
.code dwim
may be a symbol macro.
If a symbol has bindings both in the variable and function namespace in scope,
and is referenced by a dwim argument, this constitutes a conflict which is
resolved according to two rules. When nested scopes are concerned, then an
inner binding shadows an outer binding, regardless of their kind. An inner
variable binding for a symbol shadows an outer or global function binding,
and vice versa.
If a symbol is bound to both a function and variable in the global namespace,
then the variable binding is favored.
Macros do not participate in the special scope conflation, with one
exception. What this means is that the space of symbol macros is not folded
together with the space of regular macros. An argument of
.code dwim
that is a symbol might be
symbol macro, variable or function, but it cannot be interpreted as the name of
a regular macro.
The exception is this: from the perspective of a
.code dwim
form, function bindings can shadow symbol macros. If a
function binding is defined in an inner scope relative to a symbol macro for
the same symbol,
using
.code flet
or
.codn labels ,
the function hides the symbol macro. In other words, when
macro expansion processes an argument of a
.code dwim
form, and that argument is a symbol, it is treated specially
in order to provide a consistent name lookup behavior. If the innermost
binding for that symbol is a function binding, it refers to that
function binding, even if a more outer symbol macro binding exists,
and so the symbol is not expanded using the symbol macro.
By contrast, in an ordinary form, a symbolic argument never resolves
to a function binding. The symbol refers to either a symbol macro or a
variable, whichever is nested closer.
How many arguments are required by the
.code dwim
operator depends on the type of
object to which the first argument expression evaluates. The possibilities
are:
.RS
.meIP >> [ function << argument *]
Call the given function object to the given arguments.
.meIP >> [ symbol << argument *]
If the first expression evaluates to a symbol, that symbol
is resolved in the function namespace, and then
the resulting function, if found, is called with the
given arguments.
.meIP >> [ sequence << index ]
Retrieve an element from
.metn sequence ,
given by the integer
.metn index .
The following equivalences hold:
.cblk
[seq index] <--> (ref seq index)
(set [seq index] new) <--> (refset seq index new)
.cble
This form accepts update operators like
.code inc
and
.codn flip .
.meIP >> [ sequence << from-index..to-below-index ]
Retrieve the specified range of elements.
The range of elements is specified in the
.code car
and
.code cdr
fields of a cons cell,
for which the
.code ..
(dotdot) syntactic sugar is useful.
See the section on Range Indexing below.
The following equivalences hold:
.cblk
[seq from..to] <--> (sub seq from to)
(set [seq from..to] new) <--> (refset seq new from to)
.cble
This form does not accept update operators like
.code inc
and
.codn flip .
.meIP >> [ sequence << index-list ]
Elements specified
by
.meta index-list
are extracted from
.meta sequence
and returned as a sequence
of the same kind as
.metn sequence .
This form is equivalent to
.cblk
.meti (select < sequence << where-index )
.cble
except when the target of an assignment operation.
The following equivalences hold between index-list-based indexing
and the
.code select
and
.code replace
functions:
.cblk
[seq idx-list] <--> (select seq idx-list)
(set [seq idx-list] new) <--> (replace seq new idx-list)
.cble
This form does not accept update operators like
.code inc
and
.codn flip .
Note that unlike the select function, this does not support
.cblk
.meti >> [ hash << index-list ]
.cble
because since hash keys may be lists, that syntax is
indistinguishable from a simple hash lookup where
.meta index-list
is the key.
.meIP >> [ hash < key << default-value ]
Retrieve a value from the hash table corresponding to
.metn key ,
or
.meta default-value
if there is no such entry.
The first argument may not be an operator such as
.codn let ,
only a function.
The places denoted by the
.code dwim
operator can be assigned. There are some
restrictions. List, string and vector ranges can only be replaced using
the set operator. The other operators like
.code push
do not apply.
Characters in a string can only be assigned with
.code set
or incremented with
.code inc
and dec.
The source of a range assignment can be a string, vector or list,
regardless of whether the target is a string, vector or list.
If the target is a string, the replacement sequence must be
a string, or a list or vector of characters.
.RE
.PP
.TP* "Range Indexing"
Vector and list range indexing is based from zero. The first element element
zero. Furthermore, the value
.code -1
refers to the last element of the vector or
list, and
.code -2
to the second last and so forth. So the range
.code 1 .. -2
means
"everything except for the first element and the last two".
The symbol
.code t
represents the position one past the end of the vector, string or
list, so
.code 0 .. t
denotes the entire list or vector, and the range
.code t .. t
represents the empty range just beyond the last element.
It is possible to assign to
.codn t .. t .
For instance:
.cblk
(defvar list '(1 2 3))
(set [list t .. t] '(4)) ;; list is now (1 2 3 4)
.cble
The value zero has a "floating" behavior when used as the end of a range.
If the start of the range is a negative value, and the end of the
range is zero, the zero is interpreted as being the position past the
end of the sequence, rather than the first element. For instance the range
.code -1..0
means the same thing as
.codn -1..t .
Zero at the start of a range
always means the first element, so that
.code 0..-1
refers to all the elements except for the last one.
.TP* Notes:
The dwim operator allows for a Lisp-1 flavor of programming in \*(TL,
which is principally a Lisp-2 dialect.
A Lisp-1 dialect is one in which an expression like
.code (a b)
treats both a and b
as expressions subject to the same evaluation rules\(emat least, when
.code a
isn't an operator or an operator macro. This means that the symbols
.code a
and
.code b
are resolved to values in the same namespace. The form denotes a function call
if the value of variable
.code a
is a function object. Thus in a Lisp-1, named functions do not exist as
such: they are just variable bindings. In a Lisp-1
.code (car 1 2)
means that there
is a variable called
.codn car ,
which holds a function. In a Lisp-2
.code (car 1 2)
means that there is a function called
.codn car ,
and so
.code (car car car)
is possible, because there can be also a variable called
.code car
which holds a cons cell object, rather than the
.code car
function.
The Lisp-1 approach is useful for functional programming, because it eliminates
cluttering occurrences of the call and fun operators. For instance:
.cblk
;; regular notation
(call foo (fun second) '((1 a) (2 b)))
;; [] notation
[foo second '((1 a) (2 b))]
.cble
Lisp-1 dialects can also provide useful extensions by giving a meaning
to objects other than functions in the first position of a form,
and the
.code dwim/[...]
syntax does exactly this.
However, Lisp-2 also has its advantages, whereas Lisp-1 also has disadvantages.
Lisp-2 integrates with macros and operators more naturally than Lisp-1. Macros
come in two flavors: symbol macros and operator macros. This distinction
integrates better into a two namespace model, in which an operator macro is
recognized in the first position of a form, and a symbol macro in other
positions. Moreover, operators and operator macros are not functions; they
cannot work by reducing a symbol to a macro function under ordinary
evaluation rules. A Lisp-1 implementation must inspect the leftmost expression
to determine whether or not it is an operator, which breaks the principle
that all expressions are to be evaluated in the same way. Moreover, macros
and operators are never indirected upon. If
.code x
is an operator or macro, it never makes sense for
.code x
to occur other than in the leftmost position of a form. In Lisp-1 dialects,
the situation of using the name of an operator as a variable other than
in the first position, as in
.code (list let if)
has no meaning, indicating that the single namespace model serves no purpose
for macros and operators.
\*(TL banishes operators and macros (except symbol macros) from Lisp-1
evaluation, thereby achieving a more pure model of Lisp-1 evaluation than
actual Lisp-1 languages, while retaining all the advantages of Lisp-2.
In short, a "best of both worlds" situation is achieved in \*(TL.
.coNP Function @ identity
.synb
.mets (identity << value )
.syne
.desc
The
.code identity
function returns its argument.
.TP* Notes:
The
.code identity
function is useful as a functional argument, when a transformation
function is required, but no transformation is actually desired.
.coNP Function @ eval
.synb
.mets (eval < form <> [ env ])
.syne
.desc
The
.code eval
function treats the
.meta form
object as a Lisp expression, which is
evaluated. The side effects implied by the form are performed, and the value
which it produces is returned. The optional
.meta env
object specifies an environment for
resolving the function and variable references encountered in the expression.
If this argument is omitted
.code nil
then evaluation takes place in the global environment.
See also: the
.code make-env
function.
.coNP Function @ make-env
.synb
.mets (make-env >> [ variable-bindings >> [ function-bindings <> [ next-env ]]])
.syne
.desc
The
.code make-env
function creates an environment object suitable as the
.code env
parameter.
The
.meta variable-bindings
and
.meta function-bindings
parameters, if specified,
should be association lists, mapping symbols to objects. The objects in
.meta function-bindings
should be functions, or objects callable as functions.
The
.meta next-env
argument, if specified, should be an environment.
Note: bindings can also be added to an environment using the
.code env-vbind
and
.code env-fbind
functions.
.coNP Functions @ env-vbind and @ env-fbind
.synb
.mets (env-vbind < env < symbol << value)
.mets (env-fbind < env < symbol << value)
.syne
.desc
These functions bind a symbol to a value in either the function or variable
space of environment
.codn env .
Values established in the function space should be functions or objects that
can be used as functions such as lists, strings, arrays or hashes.
If
.meta symbol
already exists in the environment, in the given space, then its
value is updated with
.codn value .
.coNP Function @ constantp
.synb
.mets (constantp < form >> [ env ])
.syne
.desc
The
.code constantp
function determines whether
.mode form
is a constant form, with respect to environment
.mode env .
If
.mode env
is absent, the global environment is used.
The
.mode env
argument is used for macro-expanding
.modn form .
Currently,
.code constantp
returns true for any form, which, after macro-expansion is a compound
form with the symbol
.code quote
in its first position, a non-symbolic atom, or one of the symbols
which evaluate to themselves and cannot be bound as variables.
These symbols are the keyword symbols, and the symbols
.code t
and
.codn nil .
In the future,
.code constantp
will be able to recognize more constant forms, such as calls to certain
functions whose arguments are constant forms.
.SS* Mutation
.coNP Operators @, inc @, dec @, set @, push @, pop @, flip @, zap and @ del
.synb
.mets (inc < place <> [ delta ])
.mets (dec < place <> [ delta ])
.mets (set < place << new-value )
.mets (push < item << place )
.mets (pop << place )
.mets (flip << place )
.mets (zap < place <> [ new-value ])
.mets (del << place )
.syne
.desc
These destructive operators update the value of a place. A place is a storage
location which is denoted by a form. Place forms are identical to value
accessing forms. That is to say, any form recognized as a place by these
operators can be evaluated by itself to retrieve the value of the storage
location. However, the converse is false: not all forms which access storage
location are recognized as places.
With are exceptions noted below, it is an error if a place does not exist.
For instance, a variable being assigned must exist.
Literal objects which are directly specified in the source code are
considered part of the program body. Modifying parts of these objects
therefore gives rise to self-modifying code. The behavior of self-modifying
code is not specified.
The
.code inc
and
.code dec
update the place by adding or subtracting, respectively, a
displacement to or from that number. If the
.meta delta
expression is
specified, then it is evaluated and its value is used as the increment.
Otherwise, a default increment of
.code 1
is used. The prior value of the place
and the delta must be suitable operands for the
.code +
and
.code -
functions.
.code (inc x)
is equivalent to
.codn (set x (+ 1 x)) ,
except that expression x
is evaluated only once to determine the storage location. The
.code inc
and
.code dec
operators return the new value that was stored.
The
.code set
operator overwrites the previous value of a place with a new value,
and also returns that value.
The
.code push
and
.code pop
operators operate on a place which holds a list. The
.code push
operator updates the list by replacing it with a new list which has a new item
at the front, followed by the previous list. The item is returned.
The
.code pop
operator performs the reverse operation: it removes the first item
from the list and returns it.
.code (push y x)
is similar to
.cblk
(let ((temp y)) (set x (cons temp x)) temp)
.cble
except that
.code x
is evaluated only once to determine the storage place, and no
such temporary variable is visible to the program. Similarly,
.code (pop x)
is much like
.cblk
(let ((temp (car x))) (set x (cdr x)) temp)
.cble
except that x is evaluated only once, and no such temporary variable
is visible to the program.
The
.code flip
operator toggles a place between true and false. If the place
contains a value other than
.codn nil ,
then its value is replaced with
.codn nil .
If it contains
.codn nil ,
it is replaced with
.codn t .
The
.code zap
operator is similar to
.code set
except that
.meta new-value
is optional, defaulting to
.codn nil ,
and the operator returns the previous value of the place.
This operator is useful in avoiding spurious retention of memory in the
face of garbage collection. The idiom
.code (function (zap variable))
passes the value of
.code variable
to
.code function
while at the same time clearing it to
.codn nil . This means that the caller no longer has a reference to the
contents of the variable as far as the garbage collector is concerned.
This approach is essential in some uses of infinite, lazy objects.
The
.code del
operator does not modify the value of a place, but rather deletes the
place itself. Index values and ranges of lists denoted using the
.code dwim
operator
indexing notation can be subject to a deletion, as can hash table entries
denoted using
.code dwim
or
.codn gethash .
It is an error to try to delete other kinds of
places such as simple variables. The
.code del
operator returns the value of the
place that was deleted. Deleting from a sequence means removing the element or
elements. Deleting a hash place means removing the corresponding entry from the
hash table.
Currently, these forms are recognized as places:
.meIP < symbol
A .meta symbol place denotes a variable. If the variable does not exist, it is
an error.
.meIP (car << cons )
.meIP (cdr << cons )
These places denote the corresponding slots
of a cons cell. The
.meta cons
form must be an expression which evaluates to a cons.
.meIP (gethash < hash < key << default-value )
The
.code gethash
place denotes a value stored in a hash table.
The form
.meta hash
must evaluate to a hash table. If the place does not exist
in the hash table under the given key, then the destructive operation
will create it. In that case, the
.meta default-value
form is evaluated to
determine the initial value of the place. Otherwise it is ignored.
.meIP (vecref < vector << index )
The
.code vecref
place denotes a vector element, allowing vector elements
to be treated as assignment places.
.meIP (dwim < obj ...)
.meIP >> [ obj ...] ;; equivalent
The
.code dwim/[]
place denotes a vector element, list element, string, or hash
table, depending on the type of
.metn obj .
.SS* Binding and Iteration
.coNP Operator @ defvar
.synb
.mets (defvar < sym << value )
.syne
.desc
The
.code defvar
operator binds a variable in the global environment.
If the variable named
.meta sym
already exists in the global environment, the
form has no effect; the
.meta value
form is not evaluated, and the value of the
variable is unchanged.
If the variable does not exist, then it is introduced, with a value given by
evaluating the
.meta value
form. The
.meta value
form is evaluated in the environment
in which the
.code defvar
form occurs, not necessarily in the global environment.
The symbols
.code t
and
.code nil
may not be used as variables, and neither
can be keyword symbols: symbols denoted by a leading colon.
In addition to creating a binding, the
.code defvar
operator also marks
.meta sym
as the name of a special variable. This changes what it means to bind
that symbol in a lexical binding construct such as the
.code let
operator, or a function parameter list. See the section "Special Variables" far
above.
.coNP Operators @ let and @ let*
.synb
.mets (let >> ({ sym | >> ( sym << init-form )}*) << body-form *)
.mets (let* >> ({ sym | >> ( sym << init-form )}*) << body-form *)
.syne
.desc
The.
.code let
and
.code let*
operators introduce a new scope with variables and
evaluate forms in that scope. The operator symbol, either
.code let
or
.codn let* ,
is followed by a list which can contain any mixture of variable
name symbols, or
.cblk
.meti >> ( sym << init-form )
.cble
pairs. A symbol
denotes the name of variable to be instantiated and initialized
to the value
.codn nil .
A symbol specified with an init-form denotes
a variable which is initialized from the value of the
.metn init-form .
The symbols
.code t
and
.code nil
may not be used as variables, and neither
can be keyword symbols: symbols denoted by a leading colon.
The difference between
.code let
and
.code let*
is that in
.codn let* ,
later
.codn init-form s
have visibility over the variables established by earlier variables
in the same let* construct. In plain
.codn let ,
the variables are not visible to any of the
.metn init-form s.
When the variables are established, then the
.metn body-form s
are evaluated in order. The value of the last
.meta body-form
becomes the return value of the
.codn let .
If there are no
.metn body-form s,
then the return value
.code nil
is produced.
The list of variables may be empty.
.TP* Examples:
.cblk
(let ((a 1) (b 2)) (list a b)) -> (1 2)
(let* ((a 1) (b (+ a 1))) (list a b (+ a b))) -> (1 2 3)
(let ()) -> nil
(let (:a nil)) -> error, :a and nil can't be used as variables
.cble
.coNP Operators @ for and @ for*
.synb
.mets ({for | for*} >> ({ sym | >> ( sym << init-form )}*)
.mets \ \ \ \ \ \ \ \ \ \ \ \ \ >> ([ test-form << result-form *])
.mets \ \ \ \ \ \ \ \ \ \ \ \ \ <> ( inc-form *)
.mets \ \ << body-form *)
.syne
.desc
The
.code for
and
.code for*
operators combine variable binding with loop iteration.
The first argument is a list of variables with optional initializers,
exactly the same as in the
.code let
and
.code let*
operators. Furthermore, the
difference between
.code for
and
.code for*
is like that between
.code let
and
.code let*
with regard to this list of variables.
The
.code for
and
.code for*
operators execute these steps:
.RS
.IP 1.
Establish bindings for the specified variables similarly to
.code let
and
.codn let* .
The variable bindings are visible over the
.metn test-form ,
each
.metn result-form ,
each
.meta inc-form
and each
.metn body-form .
.IP 2.
Establish an anonymous block over the remaining forms, allowing
the
.code return
operator to be used to terminate the loop.
.IP 3.
Evaluate
.metn test-form .
If
.meta test-form
yields
.codn nil ,
then the loop terminates. Each
.meta result-form
is evaluated, and the value of the last of these
forms is is the result value of the loop.
If there are no
.metn result-form s
then the result value is
.codn nil .
If the
.meta test-form
is omitted, then the test
is taken to be true, and the loop does not terminate.
.IP 4.
Otherwise, if
.meta test-form
yields true, then each
.meta body-form
is evaluated in turn. Then, each
.code inc-form
is evaluated in turn and processing resumes at step 2.
.RE
.IP
Furthermore, the
.code for
and
.code for*
operators establish an anonymous block,
allowing the
.code return
operator to be used to terminate at any point.
.coNP Operators @, each @, each* @, collect-each @, collect-each* @ append-each and @ append-each*
.synb
.mets (each >> ({( sym << init-form )}*) << body-form *)
.mets (each* >> ({( sym << init-form )}*) << body-form *)
.mets (collect-each >> ({( sym << init-form )}*) << body-form *)
.mets (collect-each* >> ({( sym << init-form )}*) << body-form *)
.mets (append-each >> ({( sym << init-form )}*) << body-form *)
.mets (append-each* >> ({( sym << init-form )}*) << body-form *)
.syne
.desc
These operators establish a loop for iterating over the elements of one or more
lists. Each
.meta init-form
must evaluate to a list. The lists are then iterated in
parallel over repeated evaluations of the
.metn body-form s,
with each
.meta sym
variable being assigned to successive elements of its list. The shortest list
determines the number of iterations, so if any of the
.metn init-form s
evaluate to
an empty list, the body is not executed.
The body forms are enclosed in an anonymous block, allowing the
.code return
operator to terminate the loop prematurely and optionally specify
the return value.
The
.code collect-each
and
.code collect-each*
variants are like
.code each
and
.codn each* ,
except that for each iteration, the resulting value of the body is collected
into a list. When the iteration terminates, the return value of the
.code collect-each
or
.code collect-each*
operator is this collection.
The
.code append-each
and
.code append-each*
variants are like
.code each
and
.codn each* ,
except that for each iteration other than the last, the resulting value of the
body must be a list. The last iteration may produce either an atom or a list.
The objects produced by the iterations are combined together as if they
were arguments to the append function, and the resulting value is the
value of the
.code append-each
or
.code append-each*
operator.
The alternate forms denoted by the adorned symbols
.codn each* ,
.code collect-each*
and
.codn append-each* ,
differ from
.codn each ,
.code collect-each
and
.code append-each*
in the following way. The plain forms evaluate the
.metn init-form s
in an environment in which none of the
.code sym
variables are yet visible. By contrast, the alternate
forms evaluate each
.meta init-form
in an environment in which bindings for the
previous
.meta sym
variables are visible. In this phase of evaluation,
.meta sym
variables are list-valued: one by one they are each bound to the list object
emanating from their corresponding
.metn init-form .
Just before the first loop
iteration, however, the
.meta sym
variables are assigned the first item from each
of their lists.
.TP* Example:
.cblk
;; print numbers from 1 to 10 and whether they are even or odd
(each* ((n (range 1 10)) ;; n list a list here!
(even (collect-each ((n m)) (evenp m))))
;; n is an item here!
(format t "~s is ~s\en" n (if even "even" "odd")))
.cble
.TP* Output:
.cblk
1 is odd
2 is even
3 is odd
4 is even
5 is odd
6 is even
7 is odd
8 is even
9 is odd
10 is even
.cble
.SS* Function Objects and Named Functions
.coNP Operator @ defun
.synb
.mets (defun < name <> ( param * [: << opt-param *] [. << rest-param ])
.mets \ \ << body-form )
.syne
.desc
The
.code defun
operator introduces a new function in the global function namespace.
The function is similar to a lambda, and has the same parameter syntax
and semantics as the
.code lambda
operator.
Unlike in
.codn lambda ,
the
.metn body-form s
of a
.code defun
are surrounded by a block.
The name of this block is the same as the name of the function, making it
possible to terminate the function and return a value using
.cblk
.meti (return-from < name << value ).
.cble
For more information, see the definition of the block operator.
A function may call itself by name, allowing for recursion.
.coNP Operator @ lambda
.synb
.mets (lambda <> ( param * [: << opt-param *] [. << rest-param ])
.mets \ \ << body-form )
.mets (lambda < rest-param
.mets \ \ << body-form )
.syne
.desc
The
.code lambda
operator produces a value which is a function. Like in most other
Lisps, functions are objects in \*(TL. They can be passed to functions as
arguments, returned from functions, aggregated into lists, stored in variables,
et cetera.
The first argument of
.code lambda
is the list of parameters for the function. It
may be empty, and it may also be an improper list (dot notation) where the
terminating atom is a symbol other than
.codn nil .
It can also be a single symbol.
The second and subsequent arguments are the forms making up the function body.
The body may be empty.
When a function is called, the parameters are instantiated as variables that
are visible to the body forms. The variables are initialized from the values of
the argument expressions appearing in the function call.
The dotted notation can be used to write a function that accepts
a variable number of arguments. There are two ways write a function that
accepts only a variable argument list and no required arguments:
.cblk
.mets (lambda (. << rest-param ) ...)
.mets (lambda < rest-param ...)
.cble
(These notations are syntactically equivalent because the list notation
.code (. X)
actually denotes the object
.code X
which isn't wrapped in any list).
The keyword symbol
.code :
(colon) can appear in the parameter list. This is
the symbol in the keyword package whose name is the empty string. This
symbol is treated specially: it serves as a separator between
required parameters and optional parameters. Furthermore, the
.code :
symbol has a role to play in function calls: it can be specified as an argument
value to an optional parameter by which the caller indicates that the
optional argument is not being specified. It will be processed exactly
that way.
An optional parameter can also be written in the form
.cblk
.meti >> ( name < expr <> [ sym ]).
.cble
In this situation, if the call does not specify a value for the parameter
(or specifies a value as the keyword
.code :
(colon)) then the parameter takes on the
value of the expression
.metn expr .
If
.meta sym
is specified, then
.meta sym
will be
introduced as an additional binding with a boolean value which indicates
whether or not the optional parameter had been specified by the caller.
The initializer expressions are evaluated an environment in which
all of the previous parameters are visible, in addition to the surrounding
environment of the lambda. For instance:
.cblk
(let ((default 0))
(lambda (str : (end (length str)) (counter default))
(list str end counter)))
.cble
In this
.codn lambda ,
the initializing expression for the optional parameter
end is
.codn (length str) ,
and the
.code str
variable it refers to is the previous
argument. The initializer for the optional variable counter is
the expression default, and it refers to the binding established
by the surrounding let. This reference is captured as part of the
.codn lambda 's
lexical closure.
.TP* Examples:
.IP "Counting function:"
This function, which takes no arguments, captures the
variable
.codn counter .
Whenever this object is called, it increments
.code counter
by
.code 1
and returns the incremented value.
.cblk
(let ((counter 0))
(lambda () (inc counter)))
.cble
.IP "Function that takes two or more arguments:"
The third and subsequent arguments are aggregated into a list passed as the
single parameter
.codn z :
.cblk
(lambda (x y . z) (list 'my-arguments-are x y z))
.cble
.IP "Variadic function:"
.cblk
(lambda args (list 'my-list-of-arguments args))
.cble
.IP "Optional arguments:"
.cblk
[(lambda (x : y) (list x y)) 1] -> (1 nil)
[(lambda (x : y) (list x y)) 1 2] -> (1 2)
.cble
.coNP Function @ call
.synb
.mets (call < function << argument *)
.syne
.desc
The
.code call
function invokes
.metn function ,
passing it the given arguments, if any.
.TP* Examples:
Apply arguments
.code 1 2
to a
.code lambda
which adds them to produce
.codn 3 :
.cblk
(call (lambda (a b) (+ a b)) 1 2)
.cble
Useless use of
.code call
on a named function; equivalent to
.codn (list 1 2) :
.cblk
(call (fun list) 1 2)
.cble
.coNP Operator @ fun
.synb
.mets (fun << function-name )
.syne
.desc
The
.code fun
operator retrieves the function object corresponding to a named
function in the current lexical environment.
The
.meta function-name
is a symbol denoting a named function: a built in
function, or one defined by
.codn defun .
Note: the
.code fun
operator does not see macro bindings. It is possible to
retrieve a global macro expander using
.codn symbol-function .
.TP* "Dialect Note:"
A lambda expression is not a function name in \*(TL. The
syntax
.code (fun (lambda ...))
is invalid.
.coNP Functions @ symbol-function and @ symbol-value
.synb
.mets (symbol-function << symbol )
.mets (symbol-value << symbol )
.syne
.desc
The
.code symbol-function
function retrieves the value of the global function binding of the
given
.meta symbol
if it has one: that is, the function object bound to the
.metn symbol .
If
.meta symbol
has no global function binding, then the value of the global macro
binding is returned. If that doesn't exist, then the value of a global special
operator binding is returned, and if that doesn't exist, then
.code nil
is returned.
The
.code symbol-value
function retrieves the value of a global variable, if it exists,
otherwise
.codn nil .
Note: a function binding is a function, but a macro or special operator binding
binding isn't. The value of a macro binding is a list of the following form:
.cblk
.mets (#<environment object> < macro-parameter-list << body-form *)
.cble
The value of a special operator binding is a "C pointer" object, whose
printed representation looks like:
.cblk
#<cptr: 808be4f>
.cble
These details may change in future version of \*(TX.
.TP* "Dialect note:"
Forms which call
.code symbol-function
or
.code symbol-value
are currently not assignable places. Only the
.code defun
operator defines functions.
.coNP Functions @ boundp and @ fboundp
.synb
.mets (boundp << symbol )
.mets (fboundp << symbol )
.syne
.desc
.code boundp
returns
.code t
if the symbol has a variable binding in the global
environment, otherwise
.codn nil .
.code fboundp
returns
.code t
if the symbol has a function or macro binding in the global
environment, or if it is an operator, otherwise
.codn nil .
.coNP Function @ func-get-form
.synb
.mets (func-get-form << func )
.syne
.desc
The
.code func-get-form
function retrieves a source code form of
.metn func ,
which must be an interpreted function. The source code form has the syntax
.cblk
.meti >> ( name < arglist << body-form *) .
.cble
.coNP Function @ func-get-env
.synb
.mets (func-get-env << func )
.syne
.desc
The
.code func-get-env
function retrieves the environment object associated with
function
.metn func .
The environment object holds the captured bindings of a
lexical closure.
.coNP Function @ functionp
.synb
.mets (functionp << obj )
.syne
.desc
The
.code functionp
function returns
.code t
if
.meta obj
is a function, otherwise it returns
.codn nil .
.coNP Function @ interp-fun-p
.synb
.mets (interp-fun-p << obj )
.syne
.desc
The
.code interp-fun-p
function returns
.code t
if
.meta obj
is an interpreted function, otherwise it returns
.codn nil .
.coNP Macros @ flet and @ labels
.synb
.mets (flet >> ({( name < param-list << function-body-form *)}*)
.mets \ \ << body-form *)
.mets (labels >> ({( name < param-list << function-body-form *)}*)
.mets \ \ << body-form *)
.syne
.desc
The
.code flet
and
.code labels
macros bind local, named functions in the lexical scope.
The difference between
.code flet
and
.code labels
is that a function defined by
.code labels
can see itself, and therefore recurse directly by name. Moreover, if multiple
functions are defined by the same labels construct, they all see each other.
By contrast, a
.codn flet -defined
function does not have itself in scope and cannot recurse.
Multiple functions in the same
.code flet
do not have each other's names in their scopes.
More formally, the
.metn function-body-form -s
and
.meta param-list
of the functions defined by
.code labels
are in a scope in which all of the function
names being defined by that same
.code labels
construct are visible.
Under both
.code labels
and
.codn flet ,
the local functions that are defined are
lexically visible to the main
.metn body-form -s.
Note that
.code labels
and
.code flet
are properly scoped with regard to macros.
During macro expansion, function bindings introduced by these
macro operators shadow macros defined by
.code macrolet
and
.codn defmacro .
Furthermore, function bindings introduced by
.code labels
and
.code flet
also shadow symbol macros defined by
.codn symacrolet ,
when those symbol macros occur as arguments of a
.code dwim
form.
See also: the
.code macrolet
operator.
.TP* Examples:
.cblk
;; Wastefully slow algorithm for determining even-ness.
;; Note:
;; - mutual recursion between labels-defined functions
;; - inner is-even bound by labels shadows the outer
;; one bound by defun so the (is-even n) call goes
;; to the local function.
(defun is-even (n)
(labels ((is-even (n)
(if (zerop n) t (is-odd (- n 1))))
(is-odd (n)
(if (zerop n) nil (is-even (- n 1)))))
(is-even n)))
.cblk
.SS* Object Type And Equivalence
.coNP Function @ typeof
.synb
.mets (typeof << value )
.syne
.desc
The
.code typeof
function returns a symbol representing the type of
.metn value .
The core types are identified by the following symbols:
.coIP cons
Cons cell.
.coIP str
String.
.coIP lit
Literal string embedded in the \*(TX executable image.
.coIP chr
Character.
.coIP fixnum
Fixnum integer: an integer that fits into the value word, not having to
be heap allocated.
.coIP sym
Symbol.
.coIP pkg
Symbol package.
.coIP fun
Function.
.coIP vec
Vector.
.coIP lcons
Lazy cons.
.coIP lstr
Lazy string.
.coIP env
Function/variable binding environment.
.coIP bignum
A bignum integer: arbitrary precision integer that is heap-allocated.
.PP
There are additional kinds of objects, such as streams.
.coNP Functions @, null @, not and @ false
.synb
.mets (null << value )
.mets (not << value )
.mets (false << value )
.syne
.desc
The
.codn null ,
.code not
and
.code false
functions are synonyms. They tests whether
.meta value
is
the object
.codn nil .
They return
.code t
if this is the case,
.code nil
otherwise.
.TP* Examples:
.cblk
(null '()) -> t
(null nil) -> t
(null ()) -> t
(false t) -> nil
(if (null x) (format t "x is nil!"))
(let ((list '(b c d)))
(if (not (memq 'a list))
(format t "list ~s does not contain the symbol a\en")))
.cble
.coNP Function @ true
.synb
.mets (true << value )
.syne
.desc
The
.code true
function is the complement of the
.codn null ,
.code not
and
.code false
functions.
It return
.code t
if the
.meta value
is any object other than
.codn nil .
If
.meta value
is
.codn nil ,
it returns
.codn nil .
Note: programs should avoid explicitly testing values with true.
For instance
.code (if x ...)
should be favored over
.codn (if (true x) ...) .
.TP* Example:
.cblk
;; Compute indices where the list '(1 nil 2 nil 3)
;; has true values:
[where '(1 nil 2 nil 3) true] -> (1 3)
.cble
.coNP Functions @, eq @ eql and @ equal
.synb
.mets (eq < left-obj << right-obj )
.mets (eql < left-obj << right-obj )
.mets (equal < left-obj << right-obj )
.syne
.desc
The principal equality test functions
.codn eq ,
.code eql
and
.code equal
test whether two objects are equivalent, using different criteria. They return
.code t
if the objects are equivalent, and
.code nil
otherwise.
The
.code eq
function uses the strictest equivalence test, called implementation
equality. The eq function returns
.code t
if, and only if,
.meta left-obj
and
.meta right-obj
are actually the same object. The
.code eq
test is is implemented
by comparing the raw bit pattern of the value, whether or not it is
an immediate value or a pointer to a heaped object.
Two character values are
.code eq
if they are the same character, and two fixnum integers
are
.code eq
if they have the same value. All other object representations are actually
pointers, and are
.code eq
if, and only, if they point to the same object in memory.
So, for instance, two bignum integers might not be
.code eq
even if they have the same numeric
value, two lists might not be
.code eq
even if all their corresponding elements are
.code eq
and two strings might not be eq even if they hold identical text.
The.
.code eql
function is slightly less strict than
.codn eq .
The difference between
.code eql
and
.code eq
is that if
.meta left-obj
and
.meta right-obj
are numbers which are of the same kind and have the same numeric value,
.code eql
returns
.metn t ,
even if they are different objects.
Note that an integers and a floating-point number are not
.code eql
even if one has a value which converts to the other: thus,
.code (eql 0.0 0)
yields
.codn nil ;
the comparison operation which finds these numbers equal is the
.codn (= 0.0 0) .
For all other object types,
.code eql
behaves like
.codn eq .
The
.code equal
function is less strict still than
.codn eql .
In general, it recurses into some
kinds of aggregate objects to perform a structural equivalence check.
Firstly, if
.meta left-obj
and
.meta right-obj
are
.code eql
then they are also
.codn equal ,
though of course the converse isn't necessarily the case.
If two objects are both cons cells, then they are equal if their
.code car
fields are
.code equal
and their
.code cdr
fields are
.codn equal .
If two objects are vectors, they are
.code equal
if they have the same length, and
their corresponding elements are
.codn equal .
If two objects are strings, they are equal if they are textually identical.
If two objects are functions, they are
.code equal
if they have
.code equal
environments,
and if they have
the same code. Two compiled functions are considered to have
the same code if and only if they are pointers to the same function.
Two interpreted functions are considered to have the same
code if their list
structure is
.codn equal .
Two hashes are
.code equal
if they use the same equality (both are
.codn :equal-based ,
or both are the default
.codn :eql-based ),
if their associated user data elements are equal (see the function
.codn get-hash-userdata ),
if their sets of keys are identical, and if the data items associated with
corresponding keys from each respective hash are
.code equal
objects.
For some aggregate objects, there is no special semantics. Two arguments
which are symbols, packages, or streams are
.code equal
if and only if they are the same object.
Certain object types have a custom
.code equal
function.
.SS* Basic List Library
.coNP Function @ cons
.synb
.mets (cons < car-value << cdr-value )
.syne
.desc
The
.code cons
function allocates, initializes and returns a single cons cell.
A cons cell has two fields called
.code car
and
.codn cdr ,
which are accessed by
functions of the same name, or by the functions
.code first
and
.codn rest ,
which are synonyms for these.
Lists are made up of conses. A (proper) list is either the symbol
.code nil
denoting an empty list, or a cons cell which holds the first item of
the list in its
.codn car ,
and the list of the remaining items in
.codn cdr .
The expression
.code (cons 1 nil)
allocates and returns a single cons cell which denotes the one-element
list
.codn (1) .
The
.code cdr
is
.codn nil ,
so there are no additional items.
A cons cell whose
.code cdr
is an atom other than
.code nil
is printed with the dotted
pair notation. For example the cell produced by
.code (cons 1 2)
is denoted
.codn (1 . 2) .
The notation
.code (1 . nil)
is perfectly valid as input, but the cell which it denotes
will print back as
.codn (1) .
The notations are equivalent.
The dotted pair notation can be used regardless of what type
of object is the cons cell's
.codn cdr .
so that for instance
.code (a . (b c))
denotes the cons cell whose
.code car
is the symbol a
.code a
and whose
.code cdr
is the list
.codn (b c) .
This is eactly the same thing as
.codn (a b c) .
In other words
.code (a b ... l m . (n o ... w . (x y z)))
is exactly the same as
.codn (a b ... l m n o ... w x y z).
Every list, and more generally cons cell tree structure, can be written
in a "fully dotted" notation, such that there are as many dots as there
are cells. For instance the cons structure of the nested list
.code (1 (2) (3 4 (5)))
can be made more explicit using
.codn (1 . ((2 . nil) . ((3 . (4 . ((5 . nil) . nil))) . nil)))) .
The structure contains eight conses, and so there are eight dots
in the fully dotted notation.
The number of conses in a linear list like
.code (1 2 3)
is simply the number of items, so that list in particular is
made of three conses. Additional nestings require additional conses,
so for instance
.code (1 2 (3))
requires four conses. A visual way to count the conses from the printed
representation is to count the atoms, then add the count of open parentheses,
and finally subtract one.
A list terminated by an atom other than
.code nil
is called an improper
list, and the dot notation is extended to cover improper lists.
For instance
.code (1 2 . 3)
is an improper list of two elements,
terminated by
.codn 3 ,
and can be constructed using
.codn (cons 1 (cons 2 3)) .
The fully dotted notation for this list is
.codn (1 . (2 . 3)) .
.coNP Function @ atom
.synb
.mets (atom << value )
.syne
.desc
The
.code atom
function tests whether
.meta value
is an atom. It returns
.code t
if this is the
case,
.code nil
otherwise. All values which are not cons cells are atoms.
.code (atom x)
is equivalent to
.codn (not (consp x)) .
.TP* Examples:
.cblk
(atom 3) -> t
(atom (cons 1 2)) -> nil
(atom "abc") -> t
(atom '(3)) -> nil
.cble
.coNP Function @ consp
.synb
.mets (consp << value )
.syne
.desc
The
.code consp
function tests whether
.meta value
is a cons. It returns
.code t
if this is the
case,
.code nil
otherwise.
.code (consp x)
is equivalent to
.codn (not (atom x)) .
Non-empty lists test positive under
.code consp
because a list is represented as a reference to the first cons in a chain of
one or more conses.
Note that a lazy cons is a cons and satisfies the
.code consp
test. See the function
.code make-lazy-cons
and the macro
.codn lcons .
.TP* Examples:
.cblk
(consp 3) -> nil
(consp (cons 1 2)) -> t
(consp "abc") -> nil
(consp '(3)) -> t
.cble
.coNP Functions @ car and @ first
.synb
.mets (car << object )
.mets (first << object )
.syne
.desc
The functions
.code car
and
.code first
are synonyms.
If
.meta obejct
is a cons cell, these functions retrieve the
.code car
field of that cons cell.
.code (car (cons 1 2))
yields
.codn 1 .
For programming convenience,
.meta object
may be of several other kinds in addition to conses.
.code (car nil)
is allowed, and returns
.codn nil .
.meta object
may also be a vector or a string. If it is an empty vector or
string, then
.code nil
is returned. Otherwise the first character of the string or
first element of the vector is returned.
.coNP Functions @ cdr and @ rest
.synb
.mets (cdr << object )
.mets (rest << object )
.syne
.desc
The functions
.code cdr
and
.code rest
are synonyms.
If
.meta obejct
is a cons cell, these functions retrieve the
.code cdr
field of that cons cell.
.code (cdr (cons 1 2))
yields
.codn 2 .
For programming convenience,
.meta object
may be of several other kinds in addition to conses.
.code (cdr nil)
is allowed, and returns
.codn nil .
.meta object
may also be a vector or a string. If it is a non-empty string or vector
containing at least two items, then the remaining part of the object is
returned, with the first element removed. For example
.code (cdr "abc")
yields
.codn "bc" .
If
.meta object
is is a one-element vector or string, or an empty vector or string,
then
.code nil
is returned. Thus
.code (cdr "a")
and
.code (cdr "")
both result in
.codn nil .
.TP*
Example:
Walk every element of the list
.code (1 2 3)
using a
.code for
loop:
.cblk
(for ((i '(1 2 3))) (i) ((set i (cdr i)))
(print (car i) *stdout*)
(print #\enewline *stdout*))
.cble
The variable
.code i
marches over the cons cells which make up the "backbone"
of the list. The elements are retrieved using the
.code car
function.
Advancing to the next cell is achieved using
.codn (cdr i) .
If
.code i
is the
last cell in a (proper) list,
.code (cdr i)
yields
.code nil
and so
.code i
becomes
.codn nil ,
the loop guard expression
.code i
fails and the loop terminates.
.coNP Functions @ rplaca and @ rplacd
.synb
.mets (rplaca < cons << new-car-value )
.mets (rplacd < cons << new-cdr-value )
.syne
.desc
The
.code rplaca
and
.code rplacd
functions assign new values into the
.code car
and
.code cdr
fields of the cell
.metn cons .
Note that
.code (rplaca x y)
is the same as the more generic
.codn (set (car x) y) ,
and likewise
.code (rplacd x y)
can be written as
.codn (set (cdr x) y) .
It is an error if
.meta cons
is not a cons or lazy cons. In particular,
whereas
.code (car nil)
is correct,
.code (rplaca nil ...)
is erroneous.
The
.code rplaca
and
.code rplacd
functions return
.metn cons .
Note: \*(TX versions 89 and earlier, these functions returned the new value.
The behavior was undocumented.
.coNP Functions @, second @, third @, fourth @ fifth and @ sixth
.synb
.mets (first << list )
.mets (second << list )
.mets (third << list )
.mets (fourth << list )
.mets (fifth << list )
.mets (sixth << list )
.syne
.desc
These functions access the elements of a proper list by position.
If the list is shorter than implied, these functions return
.codn nil .
.TP* Examples:
.cblk
(third '(1 2)) -> nil
(second '(1 2)) -> 2
(third '(1 2 . 3)) -> **error**
.cble
.coNP Functions @, append @ nconc and @ append*
.synb
.mets (append <> [ list * << last-arg ])
.mets (nconc <> [ list * << last-arg ])
.mets (append* <> [ list * << last-arg ])
.desc
The
.code append
function creates a new list which is a catenation of the
.meta list
arguments. All arguments are optional;
.code (append)
produces the empty list.
If a single argument is specified, then
.code append
simply returns the value of that
argument. It may be any kind of object.
If N arguments are specified, where N > 1, then the first N-1 arguments must be
proper lists. Copies of these lists are catenated together. The last argument
N, shown in the above syntax as
.metn last-arg ,
may be any kind of object. It is
installed into the
.code cdr
field of the last cons cell of the resulting list.
Thus, if argument N is also a list, it is catenated onto the resulting list,
but without being copied. Argument N may be an atom other than
.codn nil ;
in that case
.code append
produces an improper list.
The
.code nconc
function works like
.codn append ,
but avoids consing. It destructively
manipulates (that is to say, mutates) incoming lists to catenate them, and so
must be used with care.
The
.code append*
function works like
.codn append ,
but returns a lazy list which produces
the catenation of the lists on demand. If some of the arguments are
themselves lazy lists which are infinite, then
.code append*
can return immediately,
whereas append will get caught in an infinite loop trying to produce a
catenation and eventually exhaust available memory. (However, the last
argument to append may be an infinite lazy list, because append does not
traverse the last argument.)
.TP* Examples:
.cblk
;; An atom is returned.
(append 3) -> 3
;; A list is also just returned: no copying takes place.
;; The eq function can verify that the same object emerges
;; from append that went in.
(let ((list '(1 2 3)))
(eq (append list) list)) -> t
(append '(1 2 3) '(4 5 6) 7) -> '(1 2 3 4 5 6 . 7))
;; the (4 5 6) tail of the resulting list is the original
;; (4 5 6) object, shared with that list.
(append '(1 2 3) '(4 5 6)) -> '(1 2 3 4 5 6)
(append nil) -> nil
;; (1 2 3) is copied: it is not the last argument
(append '(1 2 3) nil) -> (1 2 3)
;; empty lists disappear
(append nil '(1 2 3) nil '(4 5 6)) -> (1 2 3 4 5 6)
(append nil nil nil) -> nil
;; atoms and improper lists other than in the last position
;; are erroneous
(append '(a . b) 3 '(1 2 3)) -> **error**
.cble
.coNP Function @ list
.synb
.mets (list << value *)
.syne
.desc
The
.code list
function creates a new list, whose elements are the
argument values.
.TP* Examples:
.cblk
(list) -> nil
(list 1) -> (1)
(list 'a 'b) -> (a b)
.cble
.coNP Function @ list*
.synb
.mets >> ( list* << value *)
.syne
.desc
The
.code list*
function is a generalization of cons. If called with exactly
two arguments, it behaves exactly like cons:
.code (list* x y)
is identical to
.codn (cons x y) .
If three or more arguments are specified,
the leading arguments specify additional atoms to be consed to the
front of the list. So for instance
.code (list* 1 2 3)
is the same as
.code (cons 1 (cons 2 3))
and produces the improper list
.codn (1 2 . 3) .
Generalizing in the other direction,
.code list*
can be called with just
one argument, in which case it returns that argument, and
can also be called with no arguments in which case it returns
.codn nil .
.TP* Examples:
.cblk
(list*) -> nil
(list* 1) -> 1
(list* 'a 'b) -> (a . b)
(list* 'a 'b 'c) -> (a b . c)
.cble
.TP* "Dialect Note:"
Note that unlike in some other Lisp dialects, the effect
of
.code (list* 1 2 x)
can also be obtained using
.codn (list 1 2 . x) .
However,
.code (list* 1 2 (func 3))
cannot be rewritten as
.code (list 1 2 . (func 3))
because the latter is equivalent to
.codn (list 1 2 func 3) .
.coNP Function @ sub-list
.synb
.mets (sub-list < list >> [ from <> [ to ]])
.syne
.desc
This function is like the
.code sub
function, except that it operates
strictly on lists.
For a description of the arguments and semantics, refer to the
.code sub
function.
.coNP Function @ replace-list
.synb
.mets (replace-list < list < item-sequence >> [ from <> [ to ]])
.syne
.desc
The
.code replace-list
function is like the replace function, except that the first
argument must be a list.
For a description of the arguments and semantics, refer to the
.code replace
function.
.coNP Functions @ listp and @ proper-listp
.synb
.mets (listp << value )
.mets (proper-listp << value )
.syne
.desc
The
.code listp
and
.code proper-listp
functions test, respectively, whether
.meta value
is a list, or a proper list, and return
.code t
or
.code nil
accordingly.
The
.code listp
test is weaker, and executes without having to traverse
the object.
.code (listp x)
is equivalent to
.codn (or (null x) (consp x)) .
The empty list
.code nil
is a list, and a cons cell is a list.
The
.code proper-listp
function returns
.code t
only for proper lists. A proper list is
either
.codn nil ,
or a cons whose
.code cdr
is a proper list.
.code proper-listp
traverses the
list, and its execution will not terminate if the list is circular.
.coNP Function @ length-list
.synb
.mets (length-list << list )
.syne
.desc
The
.code length-list
function returns the length of
.metn list ,
which may be
a proper or improper list. The length of a list is the number of conses in that
list.
.coNP Function @ copy-list
.synb
.mets (copy-list << list )
.syne
.desc
The
.code copy-list
function which returns a list similar to
.metn list ,
but with
a newly allocated cons cell structure.
If
.meta list
is an atom, it is simply returned.
Otherwise,
.meta list
is a cons cell, and
.code copy-list
returns the same object as the expression
.cblk
.meti (cons (car << list ) (copy-list (cdr << list ))).
.cble
Note that the object
.cblk
.meti (car << list )
.cble
is not deeply copied, but only
propagated by reference into the new list.
.code copy-list
produces
a new list structure out of the same items that are in
.metn list .
.TP* "Dialect Note:"
Common Lisp does not allow the argument to be an atom, except
for the empty list
.codn nil .
.coNP Function @ copy-cons
.synb
.mets (copy-cons << cons )
.syne
.desc
This function creates a fresh cons cell, whose
.code car
and
.code cdr
fields are copied from
.codn cons .
.coNP Functions @ reverse and @ nreverse
.synb
.mets (reverse << list )
.mets (nreverse << list )
.syne
.desc
Description:
The functions
.code reverse
and
.code nreverse
produce an object which contains
the same items as proper list
.metn list ,
but in reverse order.
If
.meta list
is
.codn nil ,
then both functions return
.codn nil .
The
.code reverse
function is non-destructive: it creates a new list.
The
.code nreverse
function creates the structure of the reversed list out of the
cons cells of the input list, thereby destructively altering it (if it contains
more than one element). How
.code nreverse
uses the material from the original list
is unspecified. It may rearrange the cons cells into a reverse order, or it may
keep the structure intact, but transfer the
.code car
values among cons cells into
reverse order. Other approaches are possible.
.coNP Function @ ldiff
.synb
.mets (ldiff < list << sublist )
.syne
.desc
The values
.meta list
and
.meta sublist
are proper lists.
The
.code ldiff
function determines whether
.meta sublist
is a structural suffix of
.meta list
(meaning that it actually is a suffix, and is not merely equal to one).
This is true if
.meta list
and
.meta sublist
are the same object, or else,
recursively, if
.meta sublist
is a suffix of
.cblk
.meti (cdr << list ).
.cble
The object
.code nil
is the sublist of every list, including itself.
The ldiff function returns a new list consisting of the elements of
the prefix of
.meta list
which come before the
.meta sublist
suffix. The elements
are in the same order as in
.metn list .
If
.meta sublist
is not a suffix of
.metn list ,
then a copy of
.meta list
is returned.
These functions also work more generally on sequences.
The
.meta list
and
.meta sublist
arguments may be strings or vectors. In this case, the suffixing
matching behavior is relaxed to one of structural equivalence.
See the relevant examples below.
.TP* Examples:
.cblk
;;; unspecified: the compiler could make
;;; '(2 3) a suffix of '(1 2 3),
;;; or they could be separate objects.
(ldiff '(1 2 3) '(2 3)) -> either (1) or (1 2 3)
;; b is the (1 2) suffix of a, so the ldiff is (1)
(let ((a '(1 2 3)) (b (cdr a)))
(ldiff a b))
-> (1)
;; string and vector behavior
(ldiff "abc" "bc") -> "a"
(ldiff "abc" nil) -> "abc"
(ldiff #(1 2 3) #(3)) -> #(1 2)
;; mixtures do not have above behavior
(ldiff #(1 2 3) '(3)) -> #(1 2 3)
(ldiff '(1 2 3) #(3)) -> #(1 2 3)
(ldiff "abc" #(#\eb #\ec)) -> "abc"
.cble
.coNP Function @ last
.synb
.mets (last << list )
.syne
.desc
If
.meta list
is a nonempty proper or improper list, the
.code last
function
returns the last cons cell in the list: that cons cell whose
.code cdr
field is a terminating atom.
If
.meta list
is
.codn nil ,
then
.code nil
is returned.
.coNP Functions @ flatten and @ flatten*
.synb
.mets (flatten << list )
.mets (flatten* << list )
.syne
.desc
The
.code flatten
function produces a list whose elements are all of the
.cod2 non- nil
atoms contained in the structure of
.metn list .
The
.code flatten*
function
works like
.code flatten
except that
.code flatten
creates and returns a complete
flattened list, whereas
.code flatten*
produces a lazy list which is
instantiated on demand. This is particularly useful when the input
structure is itself lazy.
.TP* Examples:
.cblk
(flatten '(1 2 () (3 4))) -> (1 2 3 4)
;; equivalent to previous, since
;; nil is the same thing as ()
(flatten '(1 2 nil (3 4))) -> (1 2 3 4)
(flatten nil) -> nil
(flatten '(((()) ()))) -> nil
.cble
.coNP Functions @, memq @ memql and @ memqual
.synb
.mets (memq < object << list )
.mets (memql < object << list )
.mets (memqual < object << list )
.syne
.desc
The
.codn memq ,
.code memql
and
.code memqual
functions search
.meta list
for a member
which is, respectively,
.codn eq ,
.code eql
or
.coe equal
to
.metn object .
(See the
.codn eq ,
.code eql
and
.code equal
functions above.)
If no such element found,
.code nil
is returned.
Otherwise, that suffix of
.meta list
is returned whose first element
is the matching object.
.coNP Functions @ member and @ member-if
.synb
.mets (member < key < sequence >> [ testfun <> [ keyfun ]])
.mets (member-if < predfun < sequence <> [ keyfun ])
.syne
.desc
The
.code member
and
.code member-if
functions search through
.meta sequence
for an item which
matches a key, or satisfies a predicate function, respectively.
The
.meta keyfun
argument specifies a function which is applied to the elements
of the sequence to produce the comparison key. If this argument is omitted,
then the untransformed elements of the sequence themselves are examined.
The
.code member
function's
.meta testfun
argument specifies the test function which is
used to compare the comparison keys taken from the sequence to the search key.
If this argument is omitted, then the
.code equal
function is used.
If
.code member
does not find a matching element, it returns
.codn nil .
Otherwise it
returns the suffix of
.meta sequence
which begins with the matching element.
The
.code member-if
function's
.meta predfun
argument specifies a predicate function
which is applied to the successive comparison keys pulled from the sequence
by applying the key function to successive elements. If no match is found,
then
.code nil
is returned, otherwise what is returned is the suffix of
.meta sequence
which begins with the matching element.
.coNP Functions @, remq @ remql and @ remqual
.synb
.mets (remq < object << list )
.mets (remql < object << list )
.mets (remqual < object << list )
.syne
.desc
The
.codn remq ,
.code remql
and
.code remqual
functions produce a new list based on
.metn list ,
removing the items which are
.codn eq ,
.code eql
or
.code equal
to
.metn object .
The input
.meta list
is unmodified, but the returned list may share substructure
with it. If no items are removed, it is possible that the return value
is
.meta list
itself.
.coNP Functions @, remq @ remql* and @ remqual*
.synb
.mets (remq* < object << list )
.mets (remql* < object << list )
.mets (remqual* < object << list )
.syne
.desc
The
.codn remq* ,
.code remql*
and
.code remqual*
functions are lazy versions of
.codn remq ,
.code remql
and
.codn remqual .
Rather than computing the entire new list
prior to returning, these functions return a lazy list.
Caution: these functions can still get into infinite looping behavior.
For instance, in
.codn (remql* 0 (repeat '(0))) ,
.code remql
will keep consuming
the
.code 0
values coming out of the infinite list, looking for the first item that
does not have to be deleted, in order to instantiate the first lazy value.
.TP* Examples:
.cblk
;; Return a list of all the natural numbers, excluding 13,
;; then take the first 100 of these.
;; If remql is used, it will loop until memory is exhausted,
;; because (range 1) is an infinite list.
[(remql* 13 (range 1)) 0..100]
.cble
.coNP Functions @, countqual @ countql and @ countq
.synb
.mets (countq < object << list )
.mets (countql < object << list )
.mets (countqual < object << list )
.syne
.desc
The
.codn countq ,
.code countql
and
.code countqual
functions count the number of objects
in
.meta list
which are
.codn eq ,
.code eql
or
.code equal
to
.metn object ,
and return the count.
.SS* Applicative List Processing
.coNP Functions @, remove-if @, keep-if @ remove-if* and @ keep-if*
.synb
.mets (remove-if < predicate-function < list <> [ key-function ])
.mets (keep-if < predicate-function < list <> [ key-function ])
.mets (remove-if* < predicate-function < list <> [ key-function ])
.mets (keep-if* < predicate-function < list <> [ key-function ])
.syne
.desc
The
.code remove-if
function produces a list whose contents are those of
.meta list
but with those elements removed which satisfy
.metn predicate-function .
Those elements which are not removed appear in the same order.
The result list may share substructure with the input list,
and may even be the same list object if no items are removed.
The optional
.meta key-function
specifies how each element from the
.meta list
is transformed to an argument to
.metn predicate-function .
If this argument is omitted
then the predicate function is applied to the elements directly, a behavior
which is identical to
.meta key-function
being
.codn (fun identity) .
The
.code keep-if
function is exactly like
.codn remove-if ,
except the sense of
the predicate is inverted. The function
.code keep-if
retains those items
which
.code remove-if
will delete, and removes those that
.code remove-if
will preserve.
The
.code remove-if*
and
.code keep-if*
functions are like
.code remove-if
and
.codn keep-if ,
but produce lazy lists.
.TP* Examples:
.cblk
;; remove any element numerically equal to 3.
(remove-if (op = 3) '(1 2 3 4 3.0 5)) -> (1 2 4 5)
;; remove those pairs whose first element begins with "abc"
[remove-if (op equal [@1 0..3] "abc")
'(("abcd" 4) ("defg" 5))
car]
-> (("defg" 5))
;; equivalent, without test function
(remove-if (op equal [(car @1) 0..3] "abc")
'(("abcd" 4) ("defg" 5)))
-> (("defg" 5))
.cble
.coNP Function @ count-if
.synb
.mets (count-if < predicate-function < list <> [ key-function ])
.syne
.desc
The
.code count-if
function counts the number of elements of
.meta list
which satisfy
.meta predicate-function
and returns the count.
The optional
.meta key-function
specifies how each element from the
.meta list
is transformed to an argument to
.metn predicate-function .
If this argument is omitted
then the predicate function is applied to the elements directly, a behavior
which is identical to
.meta key-function
being
.codn (fun identity) .
.coNP Functions @, posqual @ posql and @ posq
.synb
.mets (posq < object << list )
.mets (posql < object << list )
.mets (posqual < object << list )
.syne
.desc
The
.codn posq ,
.code posql
and
.code posqual
functions return the zero-based position of the
first item in
.meta list
which is, respectively,
.codn eq ,
.code eql
or
.code equal
to
.metn object .
.coNP Functions @, posqual @ posql and @ posq
.synb
.mets (posq < object << list )
.mets (posql < object << list )
.mets (posqual < object << list )
.syne
.desc
The
.codn posq ,
.code posql
and
.code posqual
functions return the zero-based position of the
first item in
.meta list
which is, respectively,
.codn eq ,
.code eql
or
.code equal
to
.metn object .
.coNP Functions @ pos and @ pos-if
.synb
.mets (pos < key < list >> [ testfun <> [ keyfun ]])
.mets (pos-if < predfun < list <> [ keyfun ])
.syne
.desc
The
.code pos
and
.code pos-if
functions search through
.meta list
for an item which matches
.metn key ,
or satisfies predicate function
.metn predfun ,
respectively.
They return the zero-based position of the matching item.
The
.meta keyfun
argument specifies a function which is applied to the elements
of
.meta list
to produce the comparison key. If this argument is omitted,
then the untransformed elements of
.meta list
are examined.
The
.code pos
function's
.meta testfun
argument specifies the test function which
is used to compare the comparison keys from
.meta list
to
.metn key .
If this argument is omitted, then the
.code equal
function is used.
The position of the first element
.meta list
whose comparison key (as
retrieved by
.metn keyfun )
matches the search (under
.metn testfun )
is
returned. If no such element is found,
.code nil
is returned.
The
.code pos-if
function's
.meta predfun
argument specifies a predicate function
which is applied to the successive comparison keys taken from
.meta list
by applying
.meta keyfun
to successive elements. The position of
the first element for which
.meta predfun
yields true is returned. If
no such element is found,
.code nil
is returned.
.coNP Functions @ pos-max and @ pos-min
.synb
.mets (pos-max < sequence >> [ testfun <> [ keyfun ]])
.mets (pos-min < sequence >> [ testfun <> [ keyfun ]])
.syne
.desc
The
.code pos-min
and
.code pos-max
functions implement exactly the same algorithm; they
differ only in their defaulting behavior with regard to the
.meta testfun
argument. If
.meta testfun
is not given, then the pos-max function defaults
.meta testfun
to the
.code greater
function, whereas
.code pos-min
defaults it to the
.code less
function.
If
.meta sequence
is empty, both functions return
.codn nil .
Without a
.meta testfun
argument, the
.code pos-max
function finds the zero-based
position index of the numerically maximum value occurring in
.metn sequence ,
whereas
.code pos-min
without a
.meta testfun
argument finds the index of the minimum
value.
If a
.meta testfun
argument is given, the two functions are equivalent.
The
.meta testfun
function must be callable with two arguments.
If
.meta testfun
behaves like a greater-than comparison, then
.code pos-max
and
.code pos-min
return the index of the maximum element. If
.meta testfun
behaves like a
.code less-than
comparison, then the functions return
the index of the minimum element.
The
.meta keyfun
argument defaults to the
.code identity
function. Each element
from
.meta sequence
is passed through this one-argument function, and
the resulting value is used in its place.
.coNP Function @ where
.synb
.mets (where < function << object )
.syne
.desc
If
.meta object
is a sequence, the
.code where
function returns
a list of the numeric indices of those of its elements which satisfy
.metn function .
The numeric indices appear in increasing order.
If
.meta object
is a hash, the
.code where
function returns an unordered list
of keys which have values which satisfy
.metn function .
.meta function
must be a function that can be called with one argument.
For each element of
.metn object ,
.meta function
is called with that element
as an argument. If a
.cod2 non- nil
value is returned, then the zero-based index of
that element is added to a list. Finally, the list is returned.
.coNP Function @ select
.synb
.mets (select < object >> { index-list <> | function })
.syne
.desc
The
.code select
function returns an object, of the same kind as
.metn object ,
which consists of those elements of
.meta object
which are identified by
the indices in
.metn index-list .
If
.meta function
is given instead of
.metn index-list ,
then
.meta function
is invoked with
.meta object
as its argument. The return value is then taken as
if it were the
.meta index-list
argument .
If
.meta object
is a sequence, then
.meta index-list
consists of numeric
indices. The
.code select
function stops processing
.meta object
upon encountering an
index inside
.meta index-list
which is out of range. (Rationale: without
this strict behavior,
.code select
would not be able to terminate if
.meta index-list
is infinite.)
If
.meta object
is a list, then
.meta index-list
must contain monotonically increasing
numeric values, even if no value is out of range, since the
.code select
function
makes a single pass through the list based on the assumption that indices
are ordered. (Rationale: optimization.)
If
.meta object
is a hash, then
.meta index-list
is a list of keys. A new hash is
returned which contains those elements of
.meta object
whose keys appear
in
.metn index-list .
All of
.meta index-list
is processed, even if it contains
keys which are not in
.metn object .
.coNP Funtion @ in
.synb
.mets (in < sequence < key >> [ testfun <> [ keyfun ]])
.mets (in < hash << key )
.syne
.desc
The
.code in
function tests whether
.meta key
is found inside
.meta sequence
or
.metn hash .
If the
.meta testfun
argument is specified, it specifies the function
which is used to comparison keys from the sequence
to
.metn key .
Otherwise the
.code equal
function is used.
If the
.meta keyfun
argument is specified, it specifies a function which
is applied to the elements of
.meta sequence
to produce the comparison keys. Without this
argument, the elements themselves are taken
as the comparison keys.
If the object being searched is a hash, then the
.meta keyfun
and
.meta testfun
arguments are ignored.
The
.code in
function returns
.code t
if it finds
.meta key
in
.meta sequence
or
.metn hash,
otherwise
.codn nil .
.coNP Funtion @ partition
.synb
.mets (partition < sequence >> { index-list >> | index <> | function })
.syne
.desc
If
.meta sequence
is empty, then
.code partition
returns an empty list, and the
second argument is ignored; if it is
.metn function ,
it is not called.
Otherwise,
.code partition
returns a lazy list of partitions of the sequence
.metn sequence .
Partitions are consecutive, non-empty sub-strings of
.metn sequence ,
of the same kind as
.metn sequence .
If the second argument is of the form
.metn index-list ,
it is a sequence of
strictly increasing, positive integers. First, any leading zeros in this sequence
are dropped. The
.code partition
function then divides
.meta sequence
according to the
indices in index list. The first partition begins with the first element of
.metn sequence .
The second partition begins at the first position in
.metn index-list ,
and so on. Indices beyond the length of the sequence are ignored.
If
.meta index-list
is empty then a one-element list containing the entire
.meta sequence
is returned.
If the second argument is a function, then this function is applied
to
.metn sequence ,
and the return value of this call is then used in place of the
second argument, which must be an
.meta index
or
.metn index-list .
If the second argument is an atom other than a function, it is assumed to be
an integer index, and is turned into an
.meta index-list
of one element.
.TP* Examples:
.cblk
(partition '(1 2 3) 1) -> ((1) (2 3))
;; split the string where there is a "b"
(partition "abcbcbd" (op where @1 (op eql #\eb))) -> ("a" "bc"
"bc" "bd")
.cble
.coNP Function @ partition*
.synb
.mets (partition* < sequence >> { index-list >> | index <> | function })
.syne
.desc
If
.meta sequence
is empty, then
.code partition* returns an empty list, and the
second argument is ignored; if it is
.metn function ,
it is not called.
If the second argument is of the form
.metn index-list ,
which is a sequence
of strictly increasing non-negative integers, then
.code partition*
produces
lazy list of pieces taken from
.metn sequence .
The pieces are formed by
deleting from
.meta sequence
the elements at the positions given
in
.metn index-list .
The pieces are the non-empty sub-strings between
the deleted elements.
If
.meta index-list
is empty then a one-element list containing the entire
.meta sequence
is returned.
If the second argument is a function, then this function is applied
to
.metn sequence ,
and the return value of this call is then used in place of the
second argument, which must be an
.meta index
or
.metn index-list .
If the second argument is an atom other than a function, it is assumed to be
an integer index, and is turned into an
.meta index-list
of one element.
.TP* Examples:
.cblk
(partition* '(1 2 3 4 5) '(0 2 4)) -> ((1) (3) (5))
(partition* "abcd" '(0 3)) -> "bc"
(partition* "abcd" '(0 1 2 3)) -> nil
.cble
.coNP Function @ tree-find
.synb
.mets (tree-find < obj < tree << test-function )
.syne
.desc
The
.code tree-find
function searches
.meta tree
for an occurrence of
.metn obj .
Tree can be
any atom, or a cons. If
.meta tree
it is a cons, it is understood to be a proper
list whose elements are also trees.
The equivalence test is performed by
.meta test-function
which must take two
arguments, and has conventions similar to
.codn eq ,
.code eql
or
.codn equal .
.code tree-find
works as follows. If
.meta tree
is equivalent to
.meta obj
under
.metn test-function ,
then
.code t
is returned to announce a successful finding.
If this test fails, and
.meta tree
is an atom,
.code nil
is returned immediately to
indicate that the find failed. Otherwise,
.meta tree
is taken to be a proper list,
and tree-find is recursively applied to each element of the list in turn, using
the same
.meta obj
and
.meta test-function
arguments, stopping at the first element
which returns a
.cod2 non- nil
value.
.coNP Functions @ find and @ find-if
.synb
.mets (find < key < sequence >> [ testfun <> [ keyfun ]])
.mets (find-if < predfun < sequence <> [ keyfun ])
.syne
.desc
The
.code find
and
.code find-if
functions search through a sequence for an item which
matches a key, or satisfies a predicate function, respectively.
The
.meta keyfun
argument specifies a function which is applied to the elements
of
.meta sequence
to produce the comparison key. If this argument is omitted,
then the untransformed elements of the
.meta sequence
are searched.
The
.code find
function's
.meta testfun
argument specifies the test function which
is used to compare the comparison keys from
.meta sequence
to the search key.
If this argument is omitted, then the
.code equal
function is used.
The first element from the list whose comparison key (as retrieved by
.metn keyfun )
matches the search (under
.metn testfun )
is returned. If no such element is found,
.code nil
is returned.
The
.code find-if
function's
.meta predfun
argument specifies a predicate function
which is applied to the successive comparison keys pulled from the list
by applying
.meta keyfun
to successive elements. The first element
for which
.meta predfun
yields true is returned. If no such
element is found,
.code nil
is returned.
.coNP Functions @ find-max and @ find-min
.synb
.mets (find-max < sequence >> [ testfun <> [ keyfun ]])
.mets (find-min < sequence >> [ testfun <> [ keyfun ]])
.syne
.desc
The
.code find-min
and
.code find-max
function implement exactly the same algorithm; they
differ only in their defaulting behavior with regard to the
.meta testfun
argument. If
.meta testfun
is not given, then the find-max function defaults it to
the
.code greater
function, whereas
.code find-min
defaults it to the
.code less
function.
Without a
.meta testfun
argument, the
.code find-max
function finds the numerically
maximum value occurring in
.metn sequence ,
whereas
.code pos-min
without a
.meta testfun
argument finds the minimum value.
If a
.meta testfun
argument is given, the two functions are equivalent.
The
.meta testfun
function must be callable with two arguments.
If
.meta testfun
behaves like a greater-than comparison, then
.code find-max
and
.code find-min
both return the maximum element. If
.meta testfun
behaves like a less-than comparison, then the functions return
the minimum element.
The
.meta keyfun
argument defaults to the
.code identity
function. Each element
from
.meta sequence
is passed through this one-argument function, and
the resulting value is used in its place for the purposes of the
comparison. However, the original element is returned.
.coNP Function @ set-diff
.synb
.mets (set-diff < seq1 < seq2 >> [ testfun <> [ keyfun ]])
.syne
.desc
The
.code set-diff
function treats the sequences
.meta seq1
and
.meta seq2
as if they were sets
and computes the set difference: a sequence which contains those elements in
.meta seq1
which do not occur in
.metn seq2 .
.code set-diff
returns a sequence of the same kind as
.metn seq1 .
Element equivalence is determined by a combination of
.meta testfun
and
.metn keyfun .
Elements are compared pairwise, and each element of a pair is passed through
.meta keyfun
function to produce a comparison value. The comparison values
are compared using
.metn testfun .
If
.meta keyfun
is omitted, then the
untransformed elements themselves are compared, and if
.meta testfun
is omitted,
then the
.code equal
function is used.
If
.meta seq1
contains duplicate elements which do not occur in
.meta seq2
(and thus are preserved in the set difference) then these duplicates appear
in the resulting
sequence. Furthermore, the order of the items from
.meta seq1
is preserved.
.coNP Functions @, mapcar @ mappend @ mapcar* and @ mappend*
.synb
.mets (mapcar < function << sequence *)
.mets (mappend < function << sequence *)
.mets (mapcar* < function << sequence *)
.mets (mappend* < function << sequence *)
.syne
.desc
When given only one argument, the
.code mapcar
function returns
.codn nil .
.meta function
is never called.
When given two arguments, the
.code mapcar
function applies
.meta function
to each elements of
.meta sequence
and returns a sequence of the resulting values
in the same order as the original values.
The returned sequence is the same kind as
.metn sequence ,
if possible. If the accumulated values cannot be
elements of that type of sequence, then a list is returned.
When additional sequences are given as arguments, this filtering behavior is
generalized in the following way:
.code mapcar
traverses the sequences in parallel,
taking a value from each sequence as an argument to the function. If there
are two lists,
.meta function
is called with two arguments and so forth.
The traversal is limited by the length of the shortest sequence.
The return values of the function are collected into a new sequence which is
returned. The returned sequence is of the same kind as the leftmost
input sequence, unless the accumulated values cannot be elements of that type of
sequence, in which case a list is returned.
The
.code mappend
function works like
.codn mapcar ,
with the following difference.
Rather than accumulating the values returned by the function into a sequence,
mappend expects the items returned by the function to be sequences which
are catenated with
.codn append ,
and the resulting sequence is returned. The returned sequence is of the same
kind as the leftmost input sequence, unless the values cannot be elements
of that type of sequence, in which case a list is returned.
The
.code mapcar*
and
.code mappend*
functions work like
.code mapcar
and
.codn mappend ,
respectively.
However, they return lazy lists rather than generating the entire
output list prior to returning.
.TP* Caveats:
Like
.codn mappend ,
.code mappend*
must "consume" empty lists. For instance,
if the function being mapped puts out a sequence of
.codn nil s,
then the result must be the empty list
.codn nil ,
because
.code (append nil nil nil nil ...)
is
.codn nil .
But suppose that
.code mappend*
is used on inputs which are infinite lazy
lists, such that the function returns
.code nil
values indefinitely.
For instance:
.cblk
;; Danger: infinite loop!!!
(mappend* (fun identity) (repeat '(nil)))
.cble
The
.code mappend*
function is caught in a loop trying to consume
and squash an infinite stream of
.codn nil s,
and so doesn't return.
.TP* Examples:
.cblk
;; multiply every element by two
(mapcar (lambda (item) (* 2 item)) '(1 2 3)) -> (4 6 8)
;; "zipper" two lists together
(mapcar (lambda (le ri) (list le ri)) '(1 2 3) '(a b c)) '((1 a) (2 b) (3 c)))
;; like append, mappend allows a lone atom or a trailing atom:
(mappend (fun identity) 3) -> (3)
(mappend (fun identity) '((1) 2)) -> (1 . 2)
;; take just the even numbers
(mappend (lambda (item) (if (evenp x) (list x))) '(1 2 3 4 5))
-> (2 4)
.cble
.coNP Function @ mapdo
.synb
.mets (mapdo < function << sequence *)
.syne
.desc
The
.code mapdo
function is similar to
.codn mapcar ,
but always returns
.codn nil .
It is useful
when
.meta function
performs some kind of side effect, hence the "do" in the name,
which is a mnemonic for the execution of imperative actions.
When only the
.meta function
argument is given,
.meta function
is never called,
and
.code nil
is returned.
If a single
.meta sequence
argument is given, then
.code mapdo
iterates over
.metn sequence ,
invoking
.meta function
on each element.
If two or more
.meta sequence
arguments are given, then
.code mapdo
iterates over
the sequences in parallel, extracting parallel tuples of items. These
tuples are passed as arguments to
.metn function,
which must accept as many
arguments as there are sequences.
.coNP Functions @ transpose and @ zip
.synb
.mets (transpose << sequence )
.mets (zip << sequence *)
.syne
.desc
The
.code transpose
function performs a transposition on
.metn sequence .
This means that the
elements of
.meta sequence
must be sequences. These sequences are understood to be
columns; transpose exchanges rows and columns, returning a sequence of the rows
which make up the columns. The returned sequence is of the same kind as
.metn sequence ,
and the rows are also the same kind of sequence as the first column
of the original sequence. The number of rows returned is limited by the
shortest column among the sequences.
All of the input sequences (the elements of
.metn sequence)
must have elements
which are compatible with the first sequence. This means that if the first
element of
.meta sequence
is a string, then the remaining sequences must be
strings, or else sequences of characters, or of strings.
The
.code zip
function takes variable arguments, and is equivalent to calling
.code transpose
on a list of the arguments. The following equivalences hold:
.synb
(zip . x) <--> (transpose x)
[apply zip x] <--> (transpose x)
.syne
.TP* Examples:
.cblk
;; transpose list of lists
(transpose '((a b c) (c d e))) -> ((a c) (b d) (c e))
;; transpose vector of strings:
;; - string columns become string rows
;; - vector input becomes vector output
(transpose #("abc" "def" "ghij")) -> #("adg" "beh" "cfi")
;; error: transpose wants to make a list of strings
;; but 1 is not a character
(transpose #("abc" "def" '(1 2 3))) ;; error!
;; String elements are catenated:
(transpose #("abc" "def" ("UV" "XY" "WZ"))) -> #("adUV" "beXY" "cfWZ")
(zip '(a b c) '(c d e)) -> ((a c) (b d) (c e))
.cble
.coNP Function @ interpose
.synb
.mets (interpose < sep << sequence )
.syne
.desc
The
.code interpose
function returns a sequence of the same type as
.metn sequence ,
in which the elements from
.meta sequence
appear with the
.meta sep
value inserted
between them.
If
.meta sequence
is an empty sequence or a sequence of length 1, then a
sequence identical to
.meta sequence
is returned. It may be a copy of
.meta sequence
or it may be
.meta sequence
itself.
If
.meta sequence
is a character string, then the value
.meta sep
must be a character.
It is permissible for
.metn sequence ,
or for a suffix of
.meta sequence
to be a lazy
list, in which case interpose returns a lazy list, or a list with a lazy
suffix.
.TP* Examples:
.cblk
(interpose #\e- "xyz") -> "x-y-z"
(interpose t nil) -> nil
(interpose t #()) -> #()
(interpose #\ea "") -> ""
(interpose t (range 0 0)) -> (0)
(interpose t (range 0 1)) -> (0 t 1)
(interpose t (range 0 2)) -> (0 t 1 t 2)
.cble
.coNP Functions @ conses and @ conses*
.synb
.mets (conses << list )
.mets (conses* << list )
.syne
.desc
These functions return a list whose elements are the conses which make
up
.metn list .
The
.code conses*
function does this in a lazy way, avoiding the
computation of the entire list: it returns a lazy list of the conses of
.metn list .
The
.code conses
function computes the entire list before returning.
The input
.meta list
may be proper or improper.
The first cons of
.meta list
is that
.meta list
itself. The second cons is the rest
of the list, or
.cblk
.meti (cdr << list ).
.cble
The third cons is
.cblk
.meti (cdr (cdr << list ))
.cble
and so on.
.TP* Example:
.cblk
(conses '(1 2 3)) -> ((1 2 3) (2 3) (3))
.cble
.TP* "Dialect Note:"
These functions are useful for simulating the
.code maplist
function found in other dialects like Common Lisp.
\*(TL's
.code (conses x)
can be expressed in Common Lisp as
.codn (maplist #'identity x) .
Conversely, the Common Lisp operation
.code (maplist function list)
can be computed in \*(TL as
.codn (mapcar function (conses list)) .
More generally, the Common Lisp operation
.cblk
(maplist function list0 list1 ... listn)
.cble
can be expressed as:
.cblk
(mapcar function (conses list0)
(conses list1) ... (conses listn))
.cble
.coNP Functions @ apply and @ iapply
.synb
.mets (apply < function <> [ arg * << trailing-args ])
.mets (iapply < function <> [ arg * << trailing-args ])
.syne
.desc
The
.code apply
function invokes
.metn function ,
optionally passing to it an argument
list. The return value of the
.code apply
call is that of
.metn function .
If no arguments are present after
.metn function ,
then
.meta function
is invoked without arguments.
If one argument is present after
.metn function ,
then it is interpreted as
.metn trailing-args .
If this is a sequence (a list, vector or string),
then the elements of the sequence are passed as individual arguments to
.metn function .
If
.meta trailing-args
is not a sequence, then
.meta function
is invoked
with an improper argument list, terminated by the
.meta trailing-args
atom.
If two or more arguments are present after
.metn function ,
then the last of these arguments is interpreted as
.metn trailing-args .
The previous arguments represent leading arguments which are applied to
.metn function ,
prior to the arguments taken from
.metn trailing-args .
The
.code iapply
function ("improper apply") is similar to
.codn apply ,
except with regard to the treatment of
.metn trailing-args .
Firstly, under
.codn iapply ,
if
.meta trailing-args
is an atom other than
.code nil
(possibly a sequence, such as a vector or string),
then it is treated as an ordinary argument:
.meta function
is invoked with a proper argument list, whose last element is
.metn trailing-args .
Secondly, if
.meta trailing-args
is a list, but an improper list, then the terminating atom of
.meta trailing-args
becomes an ordinary argument. Thus, in all possible cases,
.code iapply
treats an extra
.cod2 non- nil
atom as an argument, and never calls
.meta function
with an improper argument list.
.TP* Examples:
.cblk
;; '(1 2 3) becomes arguments to list, thus (list 1 2 3).
(apply (fun list) '(1 2 3)) -> (1 2 3)
;; this effectively invokes (list 1 2 3 4)
(apply (fun list) 1 2 '(3 4)) -> (1 2 3)
;; this effectively invokes (list 1 2 . 3)
(apply (fun list) 1 2 3)) -> (1 2 . 3)
;; "abc" is separated into characters which become arguments of list
(apply (fun list) "abc") -> (#\ea #\eb #\ec)
.cble
.TP* "Dialect Note:"
Note that some uses of this function that are necessary in other Lisp dialects
are not necessary in \*(TL. The reason is that in \*(TL, improper list
syntax is accepted as a compound form, and performs application:
.cblk
(foo a b . x)
.cble
Here, the variables
.code a
and
.code b
supply the first two arguments for
.codn foo .
In
the dotted position,
.code x
must evaluate to a list or vector. The list or
vector's elements are pulled out and treated as additional arguments for
.codn foo .
Of course, this syntax can only be used if
.code x
is a symbolic form or an atom. It
cannot be a compound form, because
.code (foo a b . (x))
and
.code (foo a b x)
are equivalent structures.
.coNP Functions @ reduce-left and @ reduce-right
.synb
.mets (reduce-left < binary-function < list
.mets \ \ \ \ \ \ \ \ \ \ \ \ >> [ init-value <> [ key-function ]])
.mets (reduce-right < binary-function < list
.mets \ \ \ \ \ \ \ \ \ \ \ \ \ >> [ init-value <> [ key-function ]])
.syne
.desc
The
.code reduce-left
and
.code reduce-right
functions reduce lists of operands specified
by
.meta list
and
.meta init-value
to a single value by the repeated application of
.metn binary-function .
An effective list of operands is formed by combining
.meta list
and
.metn init-value .
If
.meta key-function
is specified, then the items of
.meta list
are
mapped to a new values through
.metn key-function .
If
.meta init-value
is supplied,
then in the case of
.codn reduce-left ,
the effective list of operands is formed by
prepending
.meta init-value
to
.metn list .
In the case of
.codn reduce-right ,
the effective operand list is produced by appending
.meta init-value
to
.metn list .
The production of the effective list can be expressed like this,
though this is not to be understood as the actual implementation:
.cblk
(append (if init-value-present (list init-value))
[mapcar (or key-function identity) list]))))
.cble
In the
.code reduce-right
case, the arguments to
.code append
are reversed.
If the effective list of operands is empty, then
.meta binary-function
is called
with no arguments at all, and its value is returned. This is the only
case in which
.meta binary-function
is called with no arguments; in all
remaining cases, it is called with two arguments.
If the effective list contains one item, then that item is returned.
Otherwise, the effective list contains two or more items, and is decimated as
follows.
Note that an
.meta init-value
specified as
.code nil
is not the same as a missing
.metn init-value ;
this means that the initial value is the object
.codn nil .
Omitting
.meta init-value
is the same as specifying a value of
.code :
(the colon symbol).
It is possible to specify
.meta key-function
while omitting an
.meta init-value
argument. This is achieved by explicitly specifying
.code :
as the
.meta init-value
argument.
Under
.codn reduce-left ,
the leftmost pair of operands is removed
from the list and passed as arguments to
.metn binary-function ,
in the same order
that they appear in the list, and the resulting value initializes an
accumulator. Then, for each remaining item in the list,
.meta binary-function
is invoked on two arguments: the current accumulator value, and the next element
from the list. After each call, the accumulator is updated with the return
value of
.metn binary-function .
The final value of the accumulator is returned.
Under
.codn reduce-right ,
the list is processed right to left. The rightmost
pair of elements in the effective list is removed, and passed as arguments to
.metn binary-function ,
in the same order that they appear in the list. The
resulting value initializes an accumulator. Then, for each remaining item in
the list,
.meta binary-function
is invoked on two arguments: the
next element from the list, in right to left order, and the current
accumulator value. After each call, the accumulator is updated with the return
value of
.metn binary-function .
The final value of the accumulator is returned.
.TP* Examples:
.cblk
;;; effective list is (1) so 1 is returned
(reduce-left (fun +) () 1 nil) -> 1
;;; computes (- (- (- 0 1) 2) 3)
(reduce-left (fun -) '(1 2 3) 0 nil) -> -6
;;; computes (- 1 (- 2 (- 3 0)))
(reduce-right (fun -) '(1 2 3) 0 nil) -> 2
;;; computes (* 1 2 3)
(reduce-left (fun *) '((1) (2) (3)) nil (fun first)) -> 6
;;; computes 1 because the effective list is empty
;;; and so * is called with no arguments, which yields 1.
(reduce-left (fun *) nil)
.cble
.coNP Function @, some @ all and @ none
.synb
.mets (some < sequence >> [ predicate-fun <> [ key-fun ]])
.mets (all < sequence >> [ predicate-fun <> [ key-fun ]])
.mets (none < sequence >> [ predicate-fun <> [ key-fun ]])
.syne
.desc
The
.codn some ,
.code all
and
.code none
functions apply a predicate test function
.meta predicate-fun
over a list of elements. If the argumenti
.meta key-fun
is
specified, then elements of
.meta sequence
are passed into
.metn key-fun ,
and
.meta predicate-fun
is
applied to the resulting values. If
.meta key-fun
is omitted, the behavior is
as if
.meta key-fun
is the identity function. If
.meta predicate-fun
is omitted,
the behavior is as if
.meta predicate-fun
is the identity function.
These functions have short-circuiting semantics and return conventions similar
to the and and or operators.
The some function applies
.meta predicate-fun
to successive values
produced by retrieving elements of
.meta list
and processing them through
.metn key-fun .
If the list is empty, it returns
.codn nil .
Otherwise it returns the
first
.cod2 non- nil
return value returned by a call to
.meta predicate-fun
and
stops evaluating more elements. If
.meta predicate-fun
returns
.code nil
for all
elements, it returns
.metn nil .
The
.code all
function applies
.meta predicate-fun
to successive values
produced by retrieving elements of
.meta list
and processing them through
.metn key-fun .
If the list is empty, it returns
.codn t .
Otherwise, if
.meta predicate-fun
yields
.code nil
for any value, the
.cod all
function immediately
returns without invoking
.meta predicate-fun
on any more elements.
If all the elements are processed, then the all function returns
the value which
.meta predicate-fun
yielded for the last element.
The
.code none
function applies
.meta predicate-fun
to successive values
produced by retrieving elements of
.meta list
and processing them through
.metn key-fun .
If the list is empty, it returns
.codn t .
Otherwise, if
.meta predicate-fun
yields
.cod2 non- nil
for any value, the none function
immediately returns nil. If
.meta predicate-fun
yields nil for all
values, the none function returns
.codn t .
.TP* Examples:
.cblk
;; some of the integers are odd
[some '(2 4 6 9) oddp] -> t
;; none of the integers are even
[none '(1 3 4 7) evenp] -> t
.cble
.coNP Function @ multi
.synb
.mets (multi < function << list *)
.syne
.desc
The
.code multi
function distributes an arbitrary list processing function
.meta multi
over multiple lists given by the
.meta list
arguments.
The
.meta list
arguments are first transposed into a single list of tuples. Each
successive element of this transposed list consists of a tuple of the
successive items from the lists. The length of the transposed list is that
of the shortest
.meta list
argument.
The transposed list is then passed to
.meta function
as an argument.
The
.meta function
is expected to produce a list of tuples, which are transposed
again to produce a list of lists which is then returned.
Conceptually, the input lists are columns and
.meta function
is invoked on
a list of the rows formed from these columns. The output of
.meta function
is a transformed list of rows which is reconstituted into a list of columns.
.TP* Example:
.cblk
;; Take three lists in parallel, and remove from all of them
;; them the element at all positions where the third list
;; has an element of 20.
(multi (op remove-if (op eql 20) @1 third)
'(1 2 3)
'(a b c)
'(10 20 30))
-> ((1 3) (a c) (10 30))
;; The (2 b 20) "row" is gone from the three "columns".
;; Note that the (op remove if (op eql 20) @1 third)
;; expression can be simplified using the ap operator:
;;
;; (op remove-if (ap eql @3 20))
.cble
.SS* Association Lists
Association lists are ordinary lists formed according to a special convention.
Firstly, any empty list is a valid association list. A non-empty association
list contains only cons cells as the key elements. These cons cells are
understood to represent key/value associations, hence the name "association
list".
.coNP Function @ assoc
.synb
.mets (assoc < key << alist )
.syne
.desc
The
.code assoc
function searches an association list
.meta alist
for a cons cell whose
car field is equivalent to
.meta key
(with equality determined by the equal
function). The first such cons is returned. If no such cons is found,
.code nil
is returned.
.coNP Function @ assql
.synb
.mets (assql < key << alist )
.syne
.desc
The
.code assql
function is just like
.codn assoc ,
except that the equality test
is determined using the
.code eql
function rather than
.codn equal .
.coNP Function @ acons
.synb
.mets (acons < car < cdr << alist )
.syne
.desc
The
.code acons
function constructs a new alist by consing a new cons to the
front of
.metn alist .
The following equivalence holds:
.cblk
(acons car cdr alist) <--> (cons (cons car cdr) alist)
.cble
.coNP Function @ acons-new
.synb
.mets (acons-new < car < cdr << alist )
.syne
.desc
The
.code acons-new
function searches
.metn alist ,
as if using the assoc function,
for an existing cell which matches the key provided by the car argument.
If such a cell exists, then its cdr field is overwritten with the
.meta cdr
argument, and then the
.meta alist
is returned. If no such cell exists, then
a new list is returned by adding a new cell to the input list consisting
of the
.meta car
and
.meta cdr
values, as if by the
.code acons
function.
.coNP Function @ aconsql-new
.synb
.mets (aconsql-new < car < cdr << alist )
.syne
.desc
This function is like
.codn acons-new ,
except that the
.code
eql
function is used
for equality testing. Thus, the list is searched for an existing cell
as if using the
.code assql
function rather than
.codn assoc .
.coNP Function @ alist-remove
.synb
.mets (alist-remove < alist << keys )
.syne
.desc
The
.code alist-remove
function takes association list
.meta alist
and produces a
duplicate from which cells matching the specified keys have been removed. The
.meta keys
argument is a list of the keys not to appear in the output list.
.coNP Function @ alist-nremove
.synb
.mets (alist-nremove < alist << keys )
.syne
.desc
The
.code alist-nremove
function is like
.codn alist-remove ,
but potentially destructive.
The input list
.meta alist
may be destroyed and its structural material re-used to
form the output list. The application should not retain references to the input
list.
.coNP Function @ copy-alist
.synb
.mets (copy-alist << alist )
.syne
.desc
The
.code copy-alist
function duplicates
.codn alist .
Unlike
.codn copy-list ,
which only duplicates list structure,
.code copy-alist
also duplicates each cons
cell of the input alist. That is to say, each element of the output list
is produced as if by the
.code copy-cons
function applied to the corresponding
element of the input list.
.SS* Property Lists
.coNP Function @ prop
.synb
.mets (prop < plist << key )
.syne
.desc
A property list a flat list of even length consisting of interleaved
pairs of property names (usually symbols) and their values (arbitrary
objects). An example property list is (:a 1 :b "two") which contains
two properties, :a having value 1, and :b having value "two".
The
.code prop
function searches property list
.meta plist
for key
.metn key .
If the key is found, then the value next to it is returned. Otherwise
.code nil
is returned.
It is ambiguous whether
.code nil
is returned due to the property not being
found, or due to the property being present with a
.code nil
value.
.SS* List Sorting
.coNP Function @ merge
.synb
.mets (merge < seq1 < seq2 >> [ lessfun <> [ keyfun ]])
.syne
.desc
The
.code merge
function merges two sorted sequences
.meta seq1
and
.meta seq2
into a single
sorted sequence. The semantics and defaulting behavior of the
.meta lessfun
and
.meta keyfun
arguments are the same as those of the sort function.
The sequence which is returned is of the same kind as
.metn seq1 .
This function is destructive of any inputs that are lists. If the output
is a list, it is formed out of the structure of the input lists.
.coNP Function @ multi-sort
.synb
.mets (multi-sort < columns < less-funcs <> [ key-funcs ])
.syne
.desc
The
.code multi-sort
function regards a list of lists to be the columns of a
database. The corresponding elements from each list constitute a record.
These records are to be sorted, producing a new list of lists.
The
.meta columns
argument supplies the list of lists which comprise the columns of
the database. The lists should ideally be of the same length. If the lists are
of different lengths, then the shortest list is taken to be the length of the
database. Excess elements in the longer lists are ignored, and do not appear in
the sorted output.
The
.meta less-funcs
argument supplies a list of comparison functions which are
applied to the columns. Successive functions correspond to successive
columns. If
.meta less-funcs
is an empty list, then the sorted database will
emerge in the original order. If
.meta less-funcs
contains exactly one function,
then the rows of the database is sorted according to the first column. The
remaining columns simply follow their row. If
.meta less-funcs
contains more than
one function, then additional columns are taken into consideration if the items
in the previous columns compare
.codn equal .
For instance if two elements from column
one compare
.codn equal ,
then the corresponding second column elements are compared
using the second column comparison function.
The optional
.meta key-funcs
argument supplies transformation functions through
which column entries are converted to comparison keys, similarly to the single
key function used in the sort function and others. If there are more key
functions than less functions, the excess key functions are ignored.
.SS* Lazy Lists and Lazy Evaluation
.coNP Function @ make-lazy-cons
.synb
.mets (make-lazy-cons << function )
.syne
.desc
The function
.code make-lazy-cons
makes a special kind of cons cell called a lazy
cons, or lcons. Lazy conses are useful for implementing lazy lists.
Lazy lists are lists which are not allocated all at once. Rather,
their elements materialize when they are accessed, like
magic stepping stones appearing under one's feet out of thin air.
A lazy cons has
.code car
and
.code cdr
fields like a regular cons, and those
fields are initialized to
.code nil
when the lazy cons is created. A lazy cons also
has an update function, the one which is provided as the
.meta function
argument to
.codn make-lazy-cons .
When either the
.code car
and
.code cdr
fields of a cons are accessed for the first time,
the function is automatically invoked first. That function has the opportunity
to initialize the
.code car
and
.code cdr
fields. Once the function is called, it is removed
from the lazy cons: the lazy cons no longer has an update function.
To continue a lazy list, the function can make another call to
.code make-lazy-cons
and install the resulting cons as the
.code cdr
of the lazy cons.
.TP* Example:
.cblk
;;; lazy list of integers between min and max
(defun integer-range (min max)
(let ((counter min))
;; min is greater than max; just return empty list,
;; otherwise return a lazy list
(if (> min max)
nil
(make-lazy-cons
(lambda (lcons)
;; install next number into car
(rplaca lcons counter)
;; now deal wit cdr field
(cond
;; max reached, terminate list with nil!
((eql counter max)
(rplacd lcons nil))
;; max not reached: increment counter
;; and extend with another lazy cons
(t
(inc counter)
(rplacd lcons (make-lazy-cons
(lcons-fun lcons))))))))))
.cble
.coNP Function @ lconsp
.synb
.mets (lconsp << value )
.syne
.desc
The
.code lconsp
function returns
.code t
if
.meta value
is a lazy cons cell. Otherwise
it returns
.codn nil ,
even if
.meta value
is an ordinary cons cell.
.coNP Function @ lcons-fun
.synb
.mets (lcons-fun << lazy-cons )
.syne
.desc
The
.code lcons-fun
function retrieves the update function of a lazy cons.
Once a lazy cons has been accessed, it no longer has an update function
and
.code lcons-fun
returns
.codn nil .
While the update function of a lazy cons is
executing, it is still accessible. This allows the update function
to retrieve a reference to itself and propagate itself into
another lazy cons (as in the example under
.codn make-lazy-cons ).
.coNP Macro @ lcons
.synb
.mets (lcons < car-expression << cdr-expression )
.syne
.desc
The
.code lcons
macro simplifies the construction of structures based on lazy conses.
Syntactically, it resembles the
.code cons
function. However, the arguments are expressions rather than values.
The macro generates code which, when evaluated, immediately produces
a lazy cons. The expressions
.meta car-expression
and
.meta cdr-expression
are not immediately evaluated. Rather, when either the
.code car
or
.code cdr
field of the lazy cons cell is accessed, these expressions are both
evaluated at that time, in the order that they appear in the
.code lcons
expression, and in the original lexical scope in which that
expression was evaluated. The return values of these expressions
are used, respectively, to initialize the corresponding fields
of the lazy cons.
Note: the
.code lcons
macro may be understood in terms of the following reference
implementation, as a syntactic sugar combining the
.code make-lazy-cons
constructor with a lexical closure provided by a
.code lambda
function:
.cblk
(defmacro lcons (car-form cdr-form)
(let ((lc (gensym)))
^(make-lazy-cons (lambda (,lc)
(rplaca ,lc ,car-form)
(rplacd ,lc ,cdr-form)))))
.cble
.TP* Example:
.cblk
;; Given the following function ...
(defun fib-generator (a b)
(lcons a (fib-generator b (+ a b))))
;; ... the following function call generates the Fibonacci
;; sequence as an infinite lazy list.
(fib-generator 1 1) -> (1 1 2 3 5 8 13 ...)
.cble
.coNP Functions @ lazy-stream-cons and @ get-lines
.synb
.mets (lazy-stream-cons << stream )
.mets (get-lines <> [ stream ])
.syne
.desc
The
.code lazy-stream-cons
and
.code get-lines
functions are synonyms, except that the
.meta stream
argument is optional in
.code get-lines
and defaults to
.codn *stdin* .
Thus, the following
description of
.code lazy-stream-cons
also applies to
.codn get-lines .
The
.code lazy-stream-cons
returns a lazy cons which generates a lazy list based on
reading lines of text from input stream
.metn stream ,
which form the elements of
the list. The
.code get-line
function is called on demand to add elements to the
list.
The
.code lazy-stream-cons
function itself makes the first call to
.code get-line
on the stream. If this returns
.codn nil ,
then the stream is closed and
.code nil
is
returned. Otherwise, a lazy cons is returned whose update function will install
that line into the
.code car
field of the lazy cons, and continue the lazy list
by making another call to
.codn lazy-stream-cons ,
installing the result into the
.code cdr
field.
.code lazy-stream-cons
inspects the real-time property of a stream
as if by the
.code real-time-stream-p
function. This determines which of two
styles of lazy list are returned. For an ordinary (non-real-time) stream,
the lazy list treats the end-of-file condition accurately: an empty
file turns into the empty list
.codn nil ,
a one line file into a one-element
list which contains that line and so on. This accuracy requires one
line of lookahead which is not acceptable in real-time streams, and
so a different type of lazy list is used, which generates an extra
.code nil
item after the last line. Under this type of lazy list, an empty input stream
translates to the list
.codn (nil) ;
a one-line stream translates to
.code ("line" nil)
and so forth.
.coNP Macro @ delay
.synb
.mets (delay << expression )
.syne
.desc
The delay operator arranges for the delayed (or "lazy") evaluation of
.metn expression .
This means that the expression is not evaluated immediately.
Rather, the delay expression produces a promise object.
The promise object can later be passed to the
.code force
function (described
later in this document). The force function will trigger the evaluation
of the expression and retrieve the value.
The expression is evaluated in the original scope, no matter where
the
.code force
takes place.
The expression is evaluated at most once, by the first call to
.codn force .
Additional calls to
.code force
only retrieve a cached value.
.TP* Example:
.cblk
@(do
;; list is popped only once: the value is computed
;; just once when force is called on a given promise
;; for the first time.
(defun get-it (promise)
(format t "*list* is ~s\en" *list*)
(format t "item is ~s\en" (force promise))
(format t "item is ~s\en" (force promise))
(format t "*list* is ~s\en" *list*))
(defvar *list* '(1 2 3))
(get-it (delay (pop *list*))))
Output:
*list* is (1 2 3)
item is 1
item is 1
*list* is (2 3)
.cble
.coNP Function @ force
.synb
.mets (force << promise )
.syne
.desc
The
.code force
function accepts a promise object produced by the
.code delay
macro.
The first time
.code force
is invoked, the
.meta expression
which was wrapped inside
.meta promise
by the
.code delay
macro is evaluated (in its original lexical environment, regardless of where in
the program the
.code force
call takes place). The value of
.meta expression
is
cached inside
.meta promise
and returned, becoming the return value of the
.code force
function call. If the
.code force
function is invoked additional times on
the same promise, the cached value is retrieved.
.coNP Macro @ mlet
.synb
.mets (mlet >> ({ sym | >> ( sym << init-form )}*) << body-form *)
.syne
.desc
The
.code mlet
macro ("magic let" or "mutual let") implements a variable binding construct
similar to
.code let
and
.codn let* .
Under
.codn mlet ,
the scope of the bindings of the
.meta sym
variables extends over the
.metn init-form -s,
as well as the
.metn body-form -s.
Unlike the
.code let*
construct, each
.meta init-form
has each
.meta sym
in scope. That is to say, an
.metn init-form
can refer not only to previous variables, but also to later variables
as well as to its own variable.
The variables are not initialized until their values are accessed for
the first time. Any
.meta sym
whose value is not accessed is not initialized.
Furthermore, the evaluation of each
.meta init-form
does not take place until the time when its value is needed
to initialize the associated
.metn sym .
This evaluation takes place once. If a given
.meta sym
is not accessed during the evaluation of the
.code mlet
construct, then its
.meta init-form
is never evaluated.
The bound variables may be assigned. If, before initialization, a variable is
updated in such a way that its prior value is not needed, it is unspecified
whether initialization takes place, and thus whether its
.meta init-form
is evaluated.
Direct circular references erroneous and are diagnosed.
.TP* Examples:
.cblk
;; Dependent calculations in arbitrary order
(mlet ((x (+ y 3))
(z (+ x 1))
(y 4))
(+ z 4)) --> 12
;; Error: circular reference:
;; x depends on y, y on z, but z on x again.
(mlet ((x (+ y 1))
(y (+ z 1))
(z (+ x 1)))
z)
;; Okay: lazy circular reference because lcons is used
(mlet ((list (lcons 1 list)))
list) --> (1 1 1 1 1 ...) ;; circular list
.cble
In the last example, the
.code list
variable is accessed for the first time in the body of the
.code mlet
form. This causes the evaluation of the
.code lcons
form. This form evaluates its arguments lazily, which means that it
is not a problem that
.code list
is not yet initialized. The form produces a lazy cons, which is then used
to initialize
.code list.
When the
.code car
or
.code cdr
fields of the lazy cons are accessed, the
.code list
expression in the
.code lcons
argument is accessed. By that time, the variable is initialized
and holds the lazy cons itself, which creates the circular reference,
and a circular list.
.coNP Functions @ generate and @ giterate
.synb
.mets (generate < while-fun << gen-fun )
.mets (giterate < while-fun < gen-fun <> [ value ])
.syne
.desc
The
.code generate
function produces a lazy list which dynamically produces items
according to the following logic.
The arguments to
.code generate
are functions which do not take any arguments. The
return value of generate is a lazy list.
When the lazy list is accessed, for instance with the functions car and cdr, it
produces items on demand. Prior to producing each item,
.meta while-fun
is
called. If it returns a true boolean value (any value other than
.codn nil ),
then
the
.meta gen-fun
function is called, and its return value is incorporated as
the next item of the lazy list. But if
.meta while-fun
yields
.codn nil ,
then the lazy list immediately terminates.
Prior to returning the lazy list, generate invokes the
.meta while-fun
one time.
If
.code while-fun
yields
.codn nil ,
then
.code generate
returns the empty list
.code nil
instead of a lazy list. Otherwise, it instantiates a lazy list, and invokes the
.code gen-func
to populate it with the first item.
The
.code giterate
function is similar to
.codn generate ,
except that
.meta while-fun
and
.meta gen-fun
are functions of one argument rather than functions of
no arguments. The optional
.meta value
argument defaults to
.code nil
and is threaded through the function calls. That is to say, the lazy
list returned is
.cblk
.meti >> ( value >> [ gen-fun << value ] >> [ gen-fun >> [ gen-fun << value ]] ...).
.cble
The lazy list terminates when a value fails to satsify
.metn while-fun .
That is to say, prior to generating each value, the lazy list tests
the value using
.metn while-fun .
If that function returns
.codn nil ,
then the item is not added, and the sequence terminates.
Note:
.code giterate
could be written in terms of
.code generate
like this:
.cblk
(defun giterate (w g v)
(generate (lambda () [w v])
(lambda () (prog1 v (set v [g v])))))
.cble
.TP* Example:
.cblk
(giterate (op > 5) (op + 1) 0) -> (0 1 2 3 4)
.cble
.coNP Function @ repeat
.synb
.mets (repeat < list <> [ count ])
.syne
.desc
If
.meta list
is empty, then repeat returns an empty list.
If
.meta count
is omitted, the
.code repeat
function produces an infinite lazy list
formed by catenating together copies of
.metn list .
If
.meta count
is specified and is zero or negative, then an empty list is
returned.
Otherwise a list is returned consisting of
.meta count
repetitions of
.meta list
catenated together.
.coNP Function @ pad
.synb
.mets (pad < sequence < object <> [ count ])
.syne
.desc
The
.code pad
function produces a lazy list which consists of all of the
elements of
.meta sequence
followed by repetitions of
.meta object .
If
.meta count
is omitted, then the repetition of
.meta object
is infinite. Otherwise the specified number of repetitions
occur.
Note that
.meta sequence
may be a lazy list which is infinite. In that case, the repetitions
of
.meta object
will never occur.
.coNP Function @ weave
.synb
.mets (weave <> { sequence }*)
.syne
.desc
The
.code weave
function interleaves elements from the sequences given as arguments.
If called with no arguments, it returns the empty list.
If called with a single sequence, it returns the elements of that sequence
as a new lazy list.
When called with two or more sequences,
.code
weave
returns a lazy list which draws elements from the sequeces in a round-robin
fashion, repeatedly scanning the sequences from left to right, and
taking an item from each one, removing it from the sequence.
Whenever a sequence runs out of items, it is deleted; the weaving then
continues with the remaining sequences. The weaved sequence terminates
when all sequences are eliminated. (If at least one of the sequences
is an infinite lazy list, then the weaved sequence is infinite.)
.TP* Examples:
.cblk
;; Weave negative integers with positive ones:
(weave (range 1) (range -1 : -1)) -> (1 -1 2 -2 3 -3 ...)
(weave "abcd" (range 1 3) '(x x x x x x x))
--> (#\ea 1 x #\eb 2 x #\ec 3 x #\ed x x x x)
.cble
.coNP Macros @ gen and @ gun
.synb
.mets (gen < while-expression << produce-item-expression )
.mets (gun << produce-item-expression )
.syne
.desc
The
.code gen
macro operator produces a lazy list, in a manner similar to the
.code generate
function. Whereas the
.code generate
function takes functional arguments,
the
.code gen
operator takes two expressions, which is often more convenient.
The return value of
.code gen
is a lazy list. When the lazy list is accessed, for
instance with the functions
.code car
and
.codn cdr ,
it produces items on demand. Prior to
producing each item, the
.meta while-expression
is evaluated, in its original
lexical scope. If the expression yields a
.cod2 non- nil
value, then
.meta produce-item-expression
is evaluated, and its return value is incorporated as
the next item of the lazy list. If the expression yields
.codn nil ,
then the lazy list immediately terminates.
The
.code gen
operator itself immediately evaluates
.meta while-expression
before
producing the lazy list. If the expression yields
.codn nil ,
then the operator
returns the empty list
.codn nil .
Otherwise, it instantiates the lazy list and
invokes the
.meta produce-item-expression
to force the first item.
The
.code gun
macro similarly creates a lazy list according to the following
rules. Each successive item of the lazy list is obtained as a result of
evaluating
.metn produce-item-expression .
However, when
.meta produce-item-expression
yields
.codn nil ,
then the list terminates (without adding that
.code nil
as an item).
Note 1: the form
.code gun
can be implemented as a macro-expanding to
an instance of the
.code gen
operator, like this:
.cblk
(defmacro gun (expr)
(let ((var (gensym)))
^(let (,var)
(gen (set ,var ,expr)
,var))))
.cble
This exploits the fact that the
.code set
operator returns the value that is
assigned, so the set expression is tested as a condition by
.codn gen ,
while having the side effect of storing the next item temporarily
in a hidden variable.
In turn,
.code gen
can be implemented as a macro expanding to some
.code lambda
functions which are passed to the
.code generate
function:
.cblk
(defmacro gen (while-expr produce-expr)
^(generate (lambda () ,while-expr) (lambda () ,produce-expr)))
.cble
Note 2:
.code gen
can be considered as an acronym for Generate, testing Expression
before Next item, whereas
.code gun
stands for Generate Until Null.
.TP* Example:
.cblk
@(do
;; Make a lazy list of integers up to 1000
;; access and print the first three.
(let* ((counter 0)
(list (gen (< counter 1000) (inc counter))))
(format t "~s ~s ~s\en" (pop list) (pop list) (pop list))))
Output:
1 2 3
.cble
.coNP Functions @ range and @ range*
.synb
.mets (range >> [ from >> [ to <> [ step ]]])
.mets (range* >> [ from >> [ to <> [ step ]]])
.syne
.desc
The
.code range
and
.code range*
functions generate a lazy sequence of integers, with a
fixed step between successive values.
The difference between
.code range
and
.code range*
is that
.code range*
excludes the endpoint.
For instance
.code (range 0 3)
generates the list
.codn (0 1 2 3) ,
whereas
.code (range* 0 3)
generates
.codn (0 1 2) .
All arguments are optional. If the
.meta step
argument is omitted, then it defaults
to
.codn 1 :
each value in the sequence is greater than the previous one by
.codn 1 .
Positive or negative step sizes are allowed. There is no check for a step size
of zero, or for a step direction which cannot meet the endpoint.
The
.meta to
argument specifies the endpoint value, which, if it occurs in the
sequence, is excluded from it by the
.code range*
function, but included by the range
function. If
.meta to
is missing, or specified as
.codn nil ,
then there is no endpoint,
and the sequence which is generated is infinite, regardless of
.metn step .
If
.meta from
is omitted, then the sequence begins at zero, otherwise
.meta from
must be an integer which specifies the initial value.
The sequence stops if it reaches the endpoint value (which is included in the
case of
.codn range ,
and excluded in the case of
.codn range *).
However, a sequence with a stepsize greater than
.code 1
or less than
.code -1
might step over the endpoint value, and
therefore never attain it. In this situation, the sequence also stops, and the
excess value which surpasses the endpoint is excluded from the sequence.
.coNP Function @ perm
.synb
.mets (perm < seq <> [ len ])
.syne
.desc
The
.code rperm
function returns a lazy list which consists of all
length
.meta len
permutations of formed by items taken from
.metn seq .
The permutations do not use any element of
.meta seq
more than once.
Argument
.metn len ,
if present, must be a positive integer, and
.meta seq
must be a sequence.
If
.meta len
is not present, then its value defaults to the length of
.metn seq :
the list of the full permutations of the entire sequence is returned.
The permutations in the returned list are sequences of the same kind as
.codn seq .
If
.meta len
is zero, then a list containing one permutation is returned, and that
permutations is of zero length.
If
.meta len
exceeds the length of
.metn seq ,
then an empty list is returned,
since it is impossible to make a single non-repeating permutation that
requires more items than are available.
The permutations are lexicographically ordered.
.coNP Function @ rperm
.synb
.mets (rperm < seq << len )
.syne
.desc
The
.code rperm
function returns a lazy list which consists of all the repeating
permutations of length
.meta len
formed by items taken from
.metn seq .
"Repeating" means that the items from
.meta seq
can appear more than
once in the permutations.
The permutations which are returned are sequences of the same kind as
.metn seq .
Argument
.meta len
must be a nonnegative integer, and
.meta seq
must be a sequence.
If
.meta len
is zero, then a single permutation is returned, of zero length.
This is true regardless of whether
.meta seq
is itself empty.
If
.meta seq
is empty and
.meta len
is greater than zero, then no permutations are
returned, since permutations of a positive length require items, and the
sequence has no items. Thus there exist no such permutations.
The first permutation consists of
.meta le
repetitions of the first element of
.metn seq .
The next repetition, if there is one, differs from the first
repetition in that its last element is the second element of
.metn seq .
That is to say, the permutations are lexicographically ordered.
.TP* Examples:
.cblk
(rperm "01" 4) -> ("000" "001" "010" "011"
"100" "101" "110" "111")
(rperm #(1) 3) -> (#(1 1 1))
(rperm '(0 1 2) 2) -> ((0 0) (0 1) (0 2) (1 0)
(1 1) (1 2) (2 0) (2 1) (2 2))
.cble
.coNP Function @ comb
.synb
.mets (comb < seq << len )
.syne
.desc
The
.code comb
function returns a lazy list which consists of all
length
.meta len
non-repeating combinations formed by taking items taken from
.metn seq .
"Non-repeating combinations" means that the combinations do not use any
element of
.meta seq
more than once. If
.meta seq
contains no duplicates, then
the combinations contain no duplicates.
Argument
.meta len
must be a nonnegative integer, and
.meta seq
must be a sequence or a hash table.
The combinations in the returned list are objects of the same kind as
.metn seq .
If
.meta len
is zero, then a list containing one combination is returned, and that
permutations is of zero length.
If
.meta len
exceeds the number of elements in
.metn seq ,
then an empty list is returned, since it is impossible to make a single
non-repeating combination that requires more items than are available.
If
.meta seq
is a sequence, the returned combinations are lexicographically ordered.
This requirement is not applicable when
.meta seq
is a hash table.
.TP* Example:
.cblk
;; powerset function, in terms of comb.
;; Yields a lazy list of all subsets of s,
;; expressed as sequences of the same type as s.
(defun powerset (s)
(mappend* (op comb s) (range 0 (length s))))
.cble
.coNP Function @ rcomb
.synb
.mets (rcomb < seq << len )
.syne
.desc
The
.code comb
function returns a lazy list which consists of all
length
.meta len
repeating combinations formed by taking items taken from
.metn seq .
"Repeating combinations" means that the combinations can use
an element of
.meta seq
more than once.
Argument
.meta len
must be a nonnegative integer, and
.meta seq
must be a sequence.
The combinations in the returned list are sequences of the same kind as
.metn seq .
If
.meta len
is zero, then a list containing one combination is returned, and that
permutations is of zero length. This is true even if
.meta seq
is empty.
If
.meta seq
is empty, and
.meta len
is nonzero, then an empty list is returned.
The combinations are lexicographically ordered.
.SS* Characters and Strings
.coNP Function @ mkstring
.synb
.mets (mkstring < length << char )
.syne
.desc
The
.code mkstring
function constructs a string object of a length specified
by the
.meta length
parameter. Every position in the string is initialized
with
.metn char ,
which must be a character value.
.coNP Function @ copy-str
.synb
.mets (copy-str << string )
.syne
.desc
The
.code copy-str
function constructs a new string whose contents are identical
to
.metn string .
.coNP Function @ upcase-str
.synb
.mets (upcase-str << string )
.syne
.desc
The
.code upcase-str
function produces a copy of
.meta string
such that all lower-case
characters of the English alphabet are mapped to their upper case counterparts.
.coNP Function @ downcase-str
.synb
.mets (downcase-str << string )
.syne
.desc
The
.code downcase-str
function produces a copy of
.meta string
such that
all upper case characters of the English alphabet are mapped to their
lower case counterparts.
.coNP Function @ string-extend
.synb
.mets (string-extend < string << tail )
.syne
.desc
The
.code string-extend
function destructively increases the length of
.metn string ,
which must be an ordinary dynamic string. It is an error to invoke this
function on a literal string or a lazy string.
The
.meta tail
argument can be a character, string or integer. If it is a string or
character, it specifies material which is to be added to the end of the string:
either a single character or a sequence of characters. If it is an integer, it
specifies the number of characters to be added to the string.
If
.meta tail
is an integer, the newly added characters have indeterminate contents.
The string appears to be the original one because of an internal terminating
null character remains in place, but the characters beyond the terminating zero
are indeterminate.
.coNP Function @ stringp
.synb
.mets (stringp << obj )
.syne
.desc
The
.code stringp
function returns t if
.meta obj
is one of the several
kinds of strings. Otherwise it returns
.codn nil .
.coNP Function @ length-str
.synb
.mets (length-str << string )
.syne
.desc
The
.code length-str
function returns the length
.meta string
in characters. The argument must be a string.
.coNP Function @ search-str
.synb
.mets (search-str < haystack < needle >> [ start <> [ from-end ]])
.syne
.desc
The
.code search-str
function finds an occurrence of the string
.meta needle
inside
the
.meta haystack
string and returns its position. If no such occurrence exists,
it returns
.codn nil .
If a
.meta start
argument is not specified, it defaults to zero. If it is
a non-negative integer, it specifies the starting character position for
the search. Negative values of
.meta start
indicate positions from the end of the
string, such that
.code -1
is the last character of the string.
If the
.meta from-end
argument is specified and is not
.codn nil ,
it means
that the search is conducted right-to-left. If multiple matches are possible,
it will find the rightmost one rather than the leftmost one.
.coNP Function @ search-str-tree
.synb
.mets (search-str-tree < haystack < tree >> [ start <> [ from-end ]])
.syne
.desc
The
.code search-str-tree
function is similar to
.codn search-str ,
except that instead of
searching
.meta haystack
for the occurrence of a single needle string, it searches
for the occurrence of numerous strings at the same time. These search strings
are specified, via the
.meta tree
argument, as an arbitrarily structured tree whose
leaves are strings.
The function finds the earliest possible match, in the given search direction,
from among all of the needle strings.
If
.meta tree
is a single string, the semantics is equivalent to
.codn search-str .
.coNP Function @ match-str
.synb
.mets (match-str < bigstring < littlestring <> [ start ])
.syne
.desc
Without the
.meta start
argument, the
.code match-str
function determines whether
.meta littlestring
is a prefix of
.metn bigstring ,
returning a
.code t
or
.code nil
indication.
If the
.meta start
argument is specified, and is a non-negative integer, then the
function tests whether
.meta littlestring
matches a prefix of that portion of
.meta bigstring
which starts at the given position.
If the
.meta start
argument is a negative integer, then
.code match-str
determines
whether
.meta littlestring
is a suffix of
.metn bigstring ,
ending on that position
of bigstring, where
.code -1
denotes the last character of
.metn bigstring ,
.code -2
the second last one and so on.
If
.meta start
is
.codn -1 ,
then this corresponds to testing whether
.meta littlestring
is a suffix of
.metn bigstring .
.coNP Function @ match-str-tree
.synb
.mets (match-str-tree < bigstring < tree <> [ start ])
.syne
.desc
The
.code match-str-tree
function is a generalization of match-str which matches
multiple test strings against
.meta bigstring
at the same time. The value
reported is the longest match from among any of the strings.
The strings are specified as an arbitrarily shaped tree structure which has
strings at the leaves.
If
.meta tree
is a single string atom, then the function behaves
exactly like match-str.
.coNP Function @ sub-str
.synb
.mets (sub-str < string >> [ from <> [ to ]])
.syne
.desc
The
.code sub-str
function is like the more generic function
.codn sub ,
except that it
operates only on strings. For a description of the arguments and semantics,
refer to the
.code sub
function.
.coNP Function @ replace-str
.synb
.mets (replace-str < string < item-sequence >> [ from <> [ to ]])
.syne
.desc
The
.code replace-str
function is like the
.code replace
function, except that the first
argument must be a string.
For a description of the arguments and semantics, refer to the
.code replace
function.
.coNP Function @ cat-str
.synb
.mets (cat-str < string-list <> [ sep-string ])
.syne
.desc
The
.code cat-str
function catenates a list of strings given by
.meta string-list
into a
single string. The optional
.meta sep-string
argument specifies a separator string
which is interposed between the catenated strings.
.coNP Function @ split-str
.synb
.mets (split-str < string << sep )
.syne
.desc
The
.code split-str
function breaks the
.meta string
into pieces, returning a list
thereof. The
.meta sep
argument must be either a string or a regular expression.
It specifies the separator character sequence within
.metn string .
All non-overlapping matches for
.meta sep
within
.meta string
are identified in left
to right order, and are removed from
.metn string .
The string is broken into pieces
according to the gaps left behind by the removed separators, and a list
of the remaining pieces is returned.
If
.meta sep
is the empty string, then the separator pieces removed from the
string are considered to be the empty strings between its
characters. In this case, if
.meta string
is of length one or zero, then it is considered to have no such pieces, and a
list of one element is returned containing the original string.
If a match for
.meta sep
is not found in the string at all, then the string is not
split at all: a list of one element is returned containing the original
string.
If
.meta sep
matches the entire string, then a list of two empty strings is
returned, except in the case that the original string is empty, in which case a
list of one element is returned, containing the empty string.
Whenever two adjacent matches for
.meta sep
occur, they are considered separate
cuts with an empty piece between them.
This operation is nondestructive:
.meta string
is not modified in any way.
Note: To split a string into pieces of length one such that an empty string
produces
.code nil
rather than
.codn ("") ,
use the
.cblk
.meti (tok-str < string #/./)
.cble
pattern.
.coNP Function @ split-str-set
.synb
.mets (split-str-set < string << set )
.syne
.desc
The
.code split-str-set
function breaks the
.meta string
into pieces, returning a list
thereof. The
.meta set
argument must be a string. It specifies a set of
characters. All occurrences of any of these characters within
.meta string
are
identified, and are removed from
.metn string .
The string is broken into pieces
according to the gaps left behind by the removed separators.
Adjacent occurrences of characters from
.meta set
within
.meta string
are considered to
be separate gaps which come between empty strings.
This operation is nondestructive:
.meta string
is not modified in any way.
.coNP Functions @ tok-str and @ tok-where
.synb
.mets (tok-str < string < regex <> [ keep-between ])
.mets (tok-where < string << regex )
.syne
.desc
The
.code tok-str
function searches
.meta string
for tokens, which are defined as
substrings of
.mta string
which match the regular expression
.meta regex
in the
longest possible way, and do not overlap. These tokens are extracted from the
string and returned as a list.
Whenever
.meta regex
matches an empty string, then an empty token is returned, and
the search for another token within
.meta string
resumes after advancing by one
character position. So for instance,
.code (tok-str "abc" #/a?/)
returns the
.cblk
("a" "" "" "").
.cble
After the token
.str "a"
is extracted from a non-empty match
for the regex, the regex is considered to match three more times: before the
.strn "b" ,
between
.str "b"
and
.strn "c" ,
and after the
.strn "c" .
If the
.meta keep-between
argument is specified, and is not
.codn nil ,
then the behavior
of
.code tok-str
changes in the following way. The pieces of
.meta string
which are
skipped by the search for tokens are included in the output. If no token is
found in
.metn string ,
then a list of one element is returned, containing
.metn string .
Generally, if N tokens are found, then the returned list consists of 2N + 1
elements. The first element of the list is the (possibly empty) substring which
had to be skipped to find the first token. Then the token follows. The next
element is the next skipped substring and so on. The last element is the
substring of
.meta string
between the last token and the end.
The
.code tok-where
function works similarly to
.codn tok-str ,
but instead of returning
the extracted tokens themselves, it returns a list of the character position
ranges within
.meta string
where matches for
.meta regex
occur. The ranges
are pairs of numbers, represented as cons cells, where the first number
of the pair gives the starting character position, and the second number
is one position past the end of the match. If a match is empty, then the
two numbers are equal.
The tok-where function does not support the
.meta keep-between
parameter.
.coNP Function @ list-str
.synb
.mets (list-str << string )
.syne
.desc
The
.code list-str
function converts a string into a list of characters.
.coNP Function @ trim-str
.synb
.mets (trim-str << string )
.syne
.desc
The
.code trim-str
function produces a copy of
.meta string
from which leading and
trailing whitespace is removed. Whitespace consists of spaces, tabs,
carriage returns, linefeeds, vertical tabs and form feeds.
.coNP Function @ chrp
.synb
.mets (chrp << obj )
.syne
.desc
Returns
.code t
if
.meta obj
is a character, otherwise nil.
.coNP Function @ chr-isalnum
.synb
.mets (chr-isalnum << char )
.syne
.desc
Returns
.code t
if
.meta char
is an alpha-numeric character, otherwise nil. Alpha-numeric
means one of the upper or lower case letters of the English alphabet found in
ASCII, or an ASCII digit. This function is not affected by locale.
.coNP Function @ chr-isalpha
.synb
.mets (chr-isalpha << char )
.syne
.desc
Returns
.code t
if
.meta char
is an alphabetic character, otherwise
.codn nil .
Alphabetic
means one of the upper or lower case letters of the English alphabet found in
ASCII. This function is not affected by locale.
.coNP Function @ chr-isascii
.synb
.mets (chr-isalpha << char )
.syne
.desc
This function returns
.code t
if the code of character
.meta char
is in the range 0 to 127 inclusive. For characters outside of this range, it
returns
.codn nil .
.coNP Function @ chr-iscntrl
.synb
.mets (chr-iscntrl << char )
.syne
.desc
This function returns
.code t
if the character
.meta char
is a character whose code
ranges from 0 to 31, or is 127. In other words, any non-printable ASCII
character. For other characters, it returns
.codn nil .
.coNP Function @ chr-isdigit
.synb
.mets (chr-isdigit << char )
.syne
.desc
This function returns
.code t
if the character
.meta char
is is an ASCII digit.
Otherwise, it returns
.codn nil .
.coNP Function @ chr-isgraph
.synb
.mets (chr-isgraph << char )
.syne
.desc
This function returns
.code t
if
.meta char
is a non-space printable ASCII character.
It returns nil if it is a space or control character.
It also returns nil for non-ASCII characters: Unicode characters with a code
above 127.
.coNP Function @ chr-islower
.synb
.mets (chr-islower << char )
.syne
.desc
This function returns
.code t
if
.meta char
is an ASCII lower case letter. Otherwise it returns
.codn nil .
.coNP Function @ chr-isprint
.synb
.mets (chr-isprint << char )
.syne
.desc
This function returns
.code t
if
.meta char
is an ASCII character which is not a
control character. It also returns
.code nil
for all non-ASCII characters: Unicode
characters with a code above 127.
.coNP Function @ chr-ispunct
.synb
.mets (chr-ispunct << char )
.syne
.desc
This function returns
.code t
if
.meta char
is an ASCII character which is not a
control character. It also returns nil for all non-ASCII characters: Unicode
characters with a code above 127.
.coNP Function @ chr-isspace
.synb
.mets (chr-isspace << char )
.syne
.desc
This function returns
.code t
if
.meta char
is an ASCII whitespace character: any of the
characters in the set
.codn #\espace ,
.codn #\etab ,
.codn #\elinefeed ,
.codn #\enewline ,
.codn #\ereturn ,
.code #\evtab
and
.codn #\epage .
For all other characters, it returns
.codn nil .
.coNP Function @ chr-isblank
.synb
.mets (chr-isblank << char )
.syne
.desc
This function returns
.code t
if
.meta char
is a space or tab: the character
.code #\espace
or
.codn #\etab .
For all other characters, it returns
.codn nil .
.coNP Function @ chr-isunisp
.synb
.mets (chr-isunisp << char )
.syne
.desc
This function returns
.code t
if
.meta char
is a Unicode whitespace character. This the case for
all the characters for which
.code chr-isspace
returns
.codn t.
It also returns
.code t
for these additional characters:
.codn #\exa0 ,
.codn #\ex1680 ,
.codn #\ex180e ,
.codn #\ex2000 ,
.codn #\ex2001 ,
.codn #\ex2002 ,
.codn #\ex2003 ,
.codn #\ex2004 ,
.codn #\ex2005 ,
.codn #\ex2006 ,
.codn #\ex2007 ,
.codn #\ex2008 ,
.codn #\ex2009 ,
.codn #\ex200a ,
.codn #\ex2028 ,
.codn #\ex2029 ,
.codn #\ex205f ,
and
.codn #\ex3000 .
For all other characters, it returns
.codn nil .
.coNP Function @ chr-isupper
.synb
.mets (chr-isupper < char )
.syne
.desc
This function returns
.code t
if
.meta char
is an ASCII upper case letter. Otherwise it returns
.codn nil .
.coNP Function @ chr-isxdigit
.synb
.mets (chr-isxdigit << char )
.syne
.desc
This function returns
.code t
if
.meta char
is a hexadecimal digit. One of the ASCII
letters
.code A
through
.codn F ,
or their lower-case equivalents, or an ASCII digit
.code 0
through
.codn 9 .
.coNP Function @ chr-toupper
.synb
.mets (chr-toupper << char )
.syne
.desc
If character
.meta char
is a lower case ASCII letter character, this function
returns the upper case equivalent character. If it is some other
character, then it just returns
.metn char .
.coNP Function @ chr-tolower
.synb
.mets (chr-tolower << char )
.syne
.desc
If character
.meta char
is an upper case ASCII letter character, this function
returns the lower case equivalent character. If it is some other
character, then it just returns
.metn char .
.coNP Functions @ num-chr and @ chr-num
.synb
.mets (num-chr << char )
.mets (chr-num << num )
.syne
.desc
The argument
.meta char
must be a character. The
.code num-chr
function returns that
character's Unicode code point value as an integer.
The argument
.meta num
must be a fixnum integer in the range
.code 0
to
.codn #\ex10FFFF .
The argument is taken to be a Unicode code point value and the
corresponding character object is returned.
.coNP Function @ chr-str
.synb
.mets (chr-str < str << idx )
.syne
.desc
The
.code chr-str
function performs random access on string
.meta str
to retrieve
the character whose position is given by integer
.metn idx ,
which must
be within range of the string.
The index value 0 corresponds to the first (leftmost) character of the string
and so non-negative values up to one less than the length are possible.
Negative index values are also allowed, such that -1 corresponds to the
last (rightmost) character of the string, and so negative values down to
the additive inverse of the string length are possible.
An empty string cannot be indexed. A string of length one supports index 0 and
index -1. A string of length two is indexed left to right by the values 0 and
1, and from right to left by -1 and -2.
.TP* Notes:
Direct use of
.code chr-str
is equivalent to the DWIM bracket notation except
that
.code str
must be a string. The following relation holds:
.cblk
(chr-str s i) --> [s i]
.cble
since
.codn [s i] <--> (ref s i) ,
this also holds:
.cblk
(chr-str s i) --> (ref s i)
.cble
.coNP Function @ chr-str-set
.synb
.mets (chr-str-set < str < idx << char )
.syne
.desc
The
.code chr-str
function performs random access on string
.meta str
to overwrite
the character whose position is given by integer
.metn idx ,
which must
be within range of the string. The character at
.meta idx
is overwritten
with character
.metn char .
The
.meta idx
argument works exactly as in
.codn chr-str .
The
.meta str
argument must be a modifiable string.
.TP* Notes:
Direct use of
.code chr-str
is equivalent to the DWIM bracket notation except
that
.meta str
must be a string. The following relation holds:
.cblk
(chr-str-set s i c) --> (set [s i] c)
.cble
since
.codn (set [s i] c) <--> (refset s i c) ,
this also holds:
.cblk
(chr-str s i) --> (refset s i c)
.cble
.coNP Function @ span-str
.synb
.mets (span-str < str << set )
.syne
.desc
The
.code span-str
function determines the longest prefix of string
.meta str
which
consists only of the characters in string
.metn set ,
in any combination.
.coNP Function @ compl-span-str
.synb
.mets (compl-span-str < str << set )
.syne
.desc
The
.code compl-span-str
function determines the longest prefix of string
.meta str
which
consists only of the characters which do not appear in
.metn set ,
in any combination.
.coNP Function @ break-str
.synb
.mets (break-str < str << set )
.syne
.desc
The
.code break-str
function returns an integer which represents the position of the
first character in string
.meta str
which appears in string
.metn set .
If there is no such character, then
.code nil
is returned.
.SS* Lazy Strings
Lazy strings are objects that were developed for the \*(TX pattern matching
language, and are exposed via \*(TL. Lazy strings behave much like strings,
and can be substituted for strings. However, unlike regular strings, which
exist in their entirety, first to last character, from the moment they are
created, lazy strings do not exist all at once, but are created on demand. If
character at index N of a lazy string is accessed, then characters 0 through N
of that string are forced into existence. However, characters at indices
beyond N need not necessarily exist.
A lazy string dynamically grows by acquiring new text from a list of strings
which is attached to that lazy string object. When the lazy string is accessed
beyond the end of its hitherto materialized prefix, it takes enough strings
from the list in order to materialize the index. If the list doesn't have
enough material, then the access fails, just like an access beyond the end of a
regular string. A lazy string always takes whole strings from the attached
list.
Lazy string growth is achieved via the
.code lazy-str-force-upto
function which
forces a string to exist up to a given character position. This function is
used internally to handle various situations.
The
.code lazy-str-force
function forces the entire string to materialize. If the
string is connected to an infinite lazy list, this will exhaust all memory.
Lazy strings are specially recognized in many of the regular string functions,
which do the right thing with lazy strings. For instance when
.code sub-str
is invoked on a lazy string, a special version of the
.code sub-str
logic is
used which handles various lazy string cases, and can potentially return
another lazy string. Taking a
.code sub-str
of a lazy string from a given character position
to the end does not force the entire lazy string to exist,
and in fact the operation will work on a lazy string that is infinite.
Furthermore, special lazy string functions are provided which allow programs to
be written carefully to take better advantage of lazy strings. What carefully
means is code that avoids unnecessarily forcing the lazy string. For instance,
in many situations it is necessary to obtain the length of a string, only to
test it for equality or inequality with some number. But it is not necessary to
compute the length of a string in order to know that it is greater than some
value.
.coNP Function @ lazy-str
.synb
.mets (lazy-str < string-list >> [ terminator <> [ limit-count ]])
.syne
.desc
The
.code lazy-str
function constructs a lazy string which draws material from
.meta string-list
which is a list of strings.
If the optional
.meta terminator
argument is given, then it specifies a string
which is appended to every string from
.metn string-list ,
before that string is
incorporated into the lazy string. If
.meta terminator
is not given,
then it defaults to the string
.strn "\en" ,
and so the strings from
.meta string-list
are effectively treated as lines which get terminated by newlines
as they accumulate into the growing prefix of the lazy string.
To avoid the use of a terminator string, a null string
.meta terminator
argument
must be explicitly passed. In that case, the lazy string grows simply
by catenating elements from
.metn string-list .
If the
.meta limit-count
argument is specified, it must be a positive integer. It
expresses a maximum limit on how many elements will be consumed from
.meta string-list
in order to feed the lazy string. Once that many elements are
drawn, the string ends, even if the list has not been exhausted.
.coNP Function @ lazy-stringp
.synb
.mets (lazy-stringp << obj )
.syne
.desc
The
.code lazy-stringp
function returns
.code t
if
.meta obj
is a lazy
string. Otherwise it returns
.codn nil .
.coNP Function @ lazy-str-force-upto
.synb
.mets (lazy-str-force-upto < lazy-str << index )
.syne
.desc
The
.code lazy-str-force-upto
function tries to instantiate the lazy string such that
the position given by
.meta index
materializes. The
.meta index
is a character
position, exactly as used in the
.code chr-str
function.
Some positions beyond
.meta index
may also materialize, as a side effect.
If the string is already materialized through to at least
.metn index ,
or if it is
possible to materialize the string that far, then the value
.code t
is returned to indicate success.
If there is insufficient material to force the lazy string through to the
.meta index
position, then nil is returned.
It is an error if the
.meta lazy-str
argument isn't a lazy string.
.coNP Function @ lazy-str-force
.synb
.mets (lazy-str-force << lazy-str )
.syne
.desc
The
.meta lazy-str
argument must be a lazy string. The lazy string is forced
to fully materialize.
The return value is an ordinary, non-lazy string equivalent to the fully
materialized lazy string.
.coNP Function @ lazy-str-get-trailing-list
.synb
.mets (lazy-str-get-trailing-list < string << index )
.syne
.desc
The
.code lazy-str-get-trailing-list
function can be considered, in some way, an inverse operation to
the production of the lazy string from its associated list.
First,
.meta string
is forced up through the position
.metn index .
That is the only extent to which
.meta string
is modified by this function.
Next, the suffix of the materialized part of the lazy string starting at
position
.metn index ,
is split into pieces on occurrences of the
terminator character (which had been given as the
.meta terminator
argument in the
.code lazy-str
constructor, and defaults to newline). If the
.meta index
position is beyond the part of the string which can be materialized
(in adherence with the lazy string's
.meta limit-count
constructor parameter), then the list of pieces is considered
to be empty.
Finally, a list is returned consisting of the pieces produced by the split,
to which is appended the remaining list of the string which has not yet been
forced to materialize.
.coNP Functions @, length-str-> @, length-str->= @ length-str-< and @ length-str-<=
.synb
.mets (length-str-> < string << len )
.mets (length-str->= < string << len )
.mets (length-str-< < string << len )
.mets (length-str-<= < string << len )
.syne
.desc
These functions compare the lengths of two strings. The following
equivalences hold, as far as the resulting value is concerned:
.cblk
(length-str-> s l) <--> (> (length-str s) l)
(length-str->= s l) <--> (>= (length-str s) l)
(length-str-< s l) <--> (< (length-str s) l)
(length-str-<= s l) <--> (<= (length-str s) l)
.cble
The difference between the functions and the equivalent forms is that if the
string is lazy, the
.code length-str
function will fully force it in order to
calculate and return its length.
These functions only force a string up to position
.metn len ,
so they are not
only more efficient, but on infinitely long lazy strings they are usable.
.code length-str
cannot compute the length of a lazy string with an unbounded
length; it will exhaust all memory trying to force the string.
These functions can be used to test such as string whether it is longer
or shorter than a given length, without forcing the string beyond
that length.
.coNP Function @ cmp-str
.synb
.mets (cmp-str < left-string << right-string )
.syne
.desc
The
.code cmp-str
function returns a negative integer if
.meta left-string
is lexicographically prior to
.metn right-string ,
and a positive integer
if the reverse situation is the case. Otherwise the strings are equal
and zero is returned.
If either or both of the strings are lazy, then they are only forced to the
minimum extent necessary for the function to reach a conclusion and return the
appropriate value, since there is no need to look beyond the first character
position in which they differ.
The lexicographic ordering is naive, based on the character code point
values in Unicode taken as integers, without regard for locale-specific
collation orders.
.coNP Functions @, str= @, str< @, str> @ str>= and @ str<=
.synb
.mets (str= < left-string << right-string )
.mets (str< < left-string << right-string )
.mets (str> < left-string << right-string )
.mets (str<= < left-string << right-string )
.mets (str>= < left-string << right-string )
.syne
.desc
These functions compare
.meta left-string
and
.meta right-string
lexicographically,
as if by the
.code cmp-str
function.
The
.code str=
function returns
.code t
if the two strings are exactly the same, character
for character, otherwise it returns
.codn nil .
The
.code str<
function returns
.code t
if
.meta left-string
is lexicographically before
.metn right-string ,
otherwise nil.
The
.code str>
function returns
.code t
if
.meta left-string
is lexicographically after
.metn right-string ,
otherwise
.codn nil .
The
.code str<
function returns
.code t
if
.meta left-string
is lexicographically before
.metn right-string ,
or if they are exactly the same, otherwise
.codn nil .
The
.code str<
function returns
.code t
if
.meta left-string
is lexicographically after
.metn right-string ,
or if they are exactly the same, otherwise
.codn nil .
.coNP Function @ string-lt
.synb
.mets (string-lt < left-str << right-str )
.syne
.desc
The
.code string-lt
is a deprecated alias for
.codn str< .
.SS* Vectors
.coNP Function @ vector
.synb
.mets (vector < length <> [ initval ])
.syne
.desc
The
.code vector
function creates and returns a vector object of the specified
length. The elements of the vector are initialized to
.metn initval ,
or to nil if
.meta initval
is omitted.
.coNP Function @ vec
.synb
.mets (vec << arg *)
.syne
.desc
The
.code vec
function creates a vector out of its arguments.
.coNP Function @ vectorp
.synb
.mets (vectorp << obj )
.syne
.desc
The
.code vectorp
function returns t if
.meta obj
is a vector, otherwise it returns
.codn nil .
.coNP Function @ vec-set-length
.synb
.mets (vec-set-length < vec << len )
.syne
.desc
The
.code vec-set-length
modifies the length of
.metn vec ,
making it longer or
shorter. If the vector is made longer, then the newly added elements
are initialized to nil. The
.meta len
argument must be nonnegative.
The return value is
.metn vec .
.coNP Function @ vecref
.synb
.mets (vecref < vec << idx )
.syne
.desc
The
.code vecref
function performs indexing into a vector. It retrieves
an element of
.meta vec
at position
.metn idx ,
counted from zero.
The
.meta idx
value must range from 0 to one less than the
length of the vector. The specified element is returned.
.coNP Function @ vec-push
.synb
.mets (vec-push < vec << elem )
.syne
.desc
The
.code vec-push
function extends the length of a vector
.meta vec
by one element, and
sets the new element to the value
.metn elem .
The previous length of the vector (which is also the position of
.metn elem )
is returned.
This function performs similarly to the generic function
.codn ref ,
except that the
first argument must be a vector.
.coNP Function @ length-vec
.synb
.mets (length-vec << vec )
.syne
.desc
The
.code length-vec
function returns the length of vector
.metn vec .
It performs
similarly to the generic
.code length
function, except that the argument must
be a vector.
.coNP Function @ size-vec
.synb
.mets (size-vec << vec )
.syne
.desc
The
.code size-vec
function returns the number of elements for which storage
is reserved in the vector
.metn vec .
.TP* Notes:
The
.code length
of the vector can be extended up to this size without any memory
allocation operations having to be performed.
.coNP Function @ vector-list
.synb
.mets (vector-list << list )
.syne
.desc
This function returns a vector which contains all of the same elements
and in the same order as list
.metn list .
.coNP Function @ list-vector
.synb
.mets (list-vector << vec )
.syne
.desc
The
.code list-vector
function returns a list of the elements of vector
.metn vec .
.coNP Function @ copy-vec
.synb
.mets (copy-vec << vec )
.syne
.desc
The
.code copy-vec
function returns a new vector object of the same length
as
.meta vec
and containing the same elements in the same order.
.coNP Function @ sub-vec
.synb
.mets (sub-vec < vec >> [ from <> [ to ]])
.syne
.desc
The
.code sub-vec
function is like the more generic function
.codn sub ,
except that it
operates only on vectors.
For a description of the arguments and semantics, refer to the
.code sub
function.
.coNP Function @ replace-vec
.synb
.mets (replace-vec < vec < item-sequence >> [ from <> [ to ]])
.syne
.desc
The
.code replace-vec
is like the
.code replace
function, except that the first argument
must be a vector.
For a description of the arguments and semantics, refer to the
.code replace
function.
.coNP Function cat-vec
.synb
.mets (cat-vec << vec-list )
.syne
.desc
The
.meta vec-list
argument is a list of vectors. The
.code cat-vec
function
produces a catenation of the vectors listed in
.metn vec-list .
It returns
a single large vector formed by catenating those vectors together in
order.
.SS* Sequence Manipulation
.coNP Function @ seqp
.synb
.mets (seqp << object )
.syne
.desc
The function
.code seqp
returns
.code t
if
.meta object
is a sequence, otherwise
.codn nil .
A sequence is defined as a list, vector or string. The object
.code nil
denotes
the empty list and so is a sequence.
.coNP Function @ length
.synb
.mets (length << sequence )
.syne
.desc
The
.code length
function returns the number of items in
.metn sequence ,
and
returns it.
.meta sequence
may be a hash, in which case (hash-count
.metn sequence )
is returned.
.coNP Function @ empty
.synb
.mets (empty << sequence )
.syne
.desc
Returns
.code t
if
.cblk
.meti (length << sequence )
.cble
is zero, otherwise
.codn nil .
.coNP Function @ copy
.synb
.mets (copy << object )
.syne
.desc
The
.code copy
function duplicates objects of various supported types: sequences, hashes and random states. If
.meta object
is
.codn nil ,
it
returns
.codn nil .
If
.meta object
is a list, it returns
.cblk
.meti (copy-list << object ).
.cble
If
.meta object
is a string, it returns
.cblk
.meti (copy-str << object ).
.cble
If
.meta object
is a vector, it returns
.cblk
.meti (copy-vec << object ).
.cble
If
.meta object
is a hash, it returns
.cblk
.meti (copy-hash << object )
.cble
Lastly, if
.meta object
is a random state, it returns
.cblk
.meti (make-random-state << object ).
.cble
Except in the case when
.meta sequence
is
.codn nil ,
.code copy
returns a value that
is distinct from (not
.code eq
to)
.metn sequence .
This is different from
the behavior of
.cblk
.meti >> [ sequence 0..t]
.cblk
or
.cblk
.meti (sub < sequence 0 t)
.cble
which recognize
that they need not make a copy of
.metn sequence ,
and just return it.
Note however, that the elements of the returned sequence may be
eq to elements of the original sequence. In other words, copy is
a deeper copy than just duplicating the
.code sequence
value itself,
but it is not a deep copy.
.coNP Function @ sub
.synb
.mets (sub < sequence >> [ from <> [ to ]])
.syne
.desc
The
.code sub
function extracts a slice from input sequence
.metn sequence .
The slice is
a sequence of the same type as
.metn sequence .
If the
.meta from
argument is omitted, it defaults to
.codn 0 .
If the
.meta to
parameter is
omitted, it defaults to
.codn t .
Thus
.code (sub a)
means
.codn (sub a 0 t) .
The following equivalence holds between the
.code sub
function and
the DWIM-bracket syntax:
.cblk
;; from is not a list
(sub seq from to) <--> [seq from..to]
.cble
The description of the
.code dwim
operator\(emin particular, the section
on Range Indexing\(emexplains the semantics of the range specification.
If the sequence is a list, the output sequence may share substructure
with the input sequence.
.coNP Function @ replace
.synb
.mets (replace < sequence < replacement-sequence >> [ from <> [ to ]])
.mets (replace < sequence < replacement-sequence << index-list )
.syne
.desc
The
.code replace
function has two invocation styles, distinguished by the
type of the third argument. If the third argument is a list, then it is deemed to be the
.meta index-list
parameter of the second form.
Otherwise, if the third argument is missing, or is not a list, then
it is deemed to be the
.meta from
argument of the first form.
The first form of the replace function replaces a contiguous subsequence of the
.meta sequence
with
.metn replacement-sequence .
The replaced subsequence may be empty,
in which case an insertion is performed. If
.meta replacement-sequence
is empty
(for example, the empty list
.codn nil ),
then a deletion is performed.
If the
.meta from
and
.meta to
arguments are omitted, their values default
to
.code 0
and
.code t
respectively.
The following equivalence holds between assignment to a place denoted by
DWIM bracket syntax and first form of the replace function:
.cblk
(set seq (replace seq new fr to)) <--> (set [seq fr..to] new)
.cble
The description of the dwim operator\(emin particular, the section
on Range Indexing\(emexplains the semantics of the range specification.
This operation is destructive: it may work "in place" by modifying
the original sequence. The caller should retain the return value
and stop relying on the original input sequence.
The second form of the replace function replaces a subsequence of
elements from
.meta sequence
given by
.metn index-list ,
with their counterparts
from
.metn replacement-sequence .
This form of the replace function does not insert
or delete; it simply overwrites elements. If
.meta replacement-sequence
and
.meta idex-list
are of different lengths, then the shorter of the two determines
the maximum number of elements which are overwritten.
Furthermore, similar restrictions apply on
.meta index-list
as under the
select function. Namely, the replacement stops when an index value
in
.meta index-list
is encountered which is out of range for
.metn sequence .
furthermore, if
.meta sequence
is a list, then
.meta index-list
must
be monotonically increasing.
The following equivalence holds between assignment to a place denoted by
DWIM bracket syntax and second form of the
.code replace
function:
.cblk
(set seq (replace seq new ix-list)) <--> (set [seq ix-list] new)
.cble
.coNP Function @ search
.synb
.mets (search < haystack < needle >> [ testfun <> [ keyfun ])
.syne
.desc
The
.code search
function determines whether the sequence
.meta needle
occurs as substring
within
.metn haystack ,
under the given comparison function
.meta testfun
and
key function
.metn keyfun .
If this is the case, then the zero-based position of
the leftmost occurrence of
.meta key
within
.meta haystack
is returned. Otherwise
.code nil
is returned to indicate that
.meta key
does not occur within
.metn haystack .
If
.meta key
is empty, then zero is always returned.
The arguments
.meta haystack
and
.meta needle
are sequences: lists, vectors
or strings, in any combination.
If
.meta needle
is not empty, then occurs at some position N within
.meta haystack
if
the first element of
.meta needle
matches the element at position N of
.metn haystack ,
the second element of
.meta needle
matches the element at position N+1 of
.meta haystack
and so forth, for all elements of
.metn needle .
A match between elements
is determined by passing each element through
.metn keyfun ,
and then comparing the resulting values using
.metn testfun .
If
.meta testfun
is supplied, it must be a function which can be
called with two arguments. If it is not supplied, it defaults to
.codn eql .
If
.meta keyfun
is supplied, it must be a function which can be called
with one argument. If it is not supplied, it defaults to
.codn identity .
.TP* Examples:
.cblk
;; fails because 3.0 doesn't match 3
;; under the default eql function
[search #(1.0 3.0 4.0 7.0) '(3 4)] -> nil
;; occurrence found at position 1:
;; (3.0 4.0) matches (3 4) under =
[search #(1.0 3.0 4.0 7.0) '(3 4) =] -> 1
;; "even odd odd odd even" pattern
;; matches at position 2
[search #(1 1 2 3 5 7 8) '(2 1 1 1 2) : evenp] -> 2
;; Case insensitive string search
[search "abcd" "CD" : chr-toupper] -> 2
;; Case insensitive string search
;; using vector of characters as key
[search "abcd" #(#\eC #\eD) : chr-toupper] -> 2
.cble
.coNP Functions @ ref and @ refset
.synb
.mets (ref < seq << index )
.mets (refset < seq < index << new-value )
.syne
.desc
The
.code ref
and
.code refset
functions perform array-like indexing into sequences.
The
.code ref
function retrieves an element of
.metn seq ,
whereas
.code refset
overwrites an
element of
.meta seq
with a new value.
The
.meta index
argument is based from zero, and negative values are permitted,
with a special meaning as described in the Range Indexing section under the
description of the
.code dwim
operator.
The
.code refset
function returns the new value.
The following equivalences hold between
.code ref
and
.codn refset ,
and the DWIM bracket syntax:
.cblk
(ref seq idx) <--> [seq idx]
(refset seq idx new) <--> (set [seq idx] new)
.cble
The difference is that
.code ref
and
.code refset
are first class functions which
can be used in functional programming as higher order functions, whereas the
bracket notation is syntactic sugar, and
.code set
is an operator, not a function.
Therefore the brackets cannot replace all uses of
.code ref
and
.codn refset .
.coNP Function @ update
.synb
.mets (update < sequence-or-hash << function )
.syne
.desc
The
.code update
function replaces each elements in a sequence, or each value
in a hash table, with the value of
.meta function
applied to that element
or value.
The sequence or hash table is returned.
.coNP Function @ less
.synb
.mets (less < left-obj << right-obj )
.mets (less < obj << obj *)
.syne
.desc
The
.code less
function, when called with two arguments, determines whether
.meta left-obj
compares less than
.meta right-obj
in a generic way which handles arguments of various types.
The argument syntax of
.code less
is generalized. It can accept one argument, in which case it unconditionally
returns
.code t
regardless of that argument's value. If more than two arguments are
given, then
.code less
generalizes in a way which can be described by the following equivalence
pattern, with the understanding that each argument expression
is evaluated exactly once:
.cblk
(less a b c) <--> (and (less a b) (less b c))
(less a b c d) <--> (and (less a b) (less b c) (less c d))
.cble
The
.code less
function is used as the default for the
.meta lessfun
argument of the functions
.code sort
and
.codn merge ,
as well as the
.meta testfun
argument of the
.code pos-min
and
.codn find-min .
The
.code less
function is capable of comparing numbers, characters, symbols, strings,
as well as lists and vectors of these.
If both arguments are the same object so that
.cblk
.meti (eq < left-obj << right-obj )
.cble
holds true, then the function returns
.code nil
regardless of the type of
.metn left-obj ,
even if the function doesn't handle comparing different instances
of that type. In other words, no object is less than itself, no matter
what it is.
If both arguments are numbers or characters, they are compared as if using the
.code < function.
If both arguments are strings, they are compared as if using the
.code string-lt
function.
If both arguments are symbols, then their names are compared in
their place, as if by the
.code string-lt
function.
If both arguments are conses, then they are compared as follows:
.RS
.IP 1.
The
.code less
function is recursively applied to the
.code car
fields of both arguments. If it yields true, then
.meta left-obj
is deemed to be less than
.metn right-obj .
.IP 2.
Otherwise, if the
.code car
fields are unequal under
the
.code equal
function,
.code less
returns
.codn nil.
.IP 3.
If the
.code car
fields are
.code equal
then
.code less
is recursively applied to the
.code cdr
fields of the arguments, and the result of that comparison is returned.
.RE
.IP
This logic performs a lexicographic comparison on ordinary lists such
that for instance
.code (1 1)
is less than
.code (1 1 1)
but not less than
.code (1 0)
or
.codn (1) .
Note that the empty
.code nil
list nil compared to a cons is handled by type-based precedence, described
below.
If the arguments are vectors, they are compared lexicographically, similar
to strings. Corresponding elements, starting with element 0, of the
vectors are compared until an index position is found where the vectors
differ. If this differing position is beyond the end of one of the two vectors,
then the shorter vector is considered to be lesser. Otherwise, the result
of
.code less
is the outcome of comparing those differing elements themselves
with
.codn less .
If the two arguments are of the above types, but of mutually different types,
then
.code less
resolves the situation based on the following precedence: numbers and
characters are less than strings, which are less than symbols,
which are less than conses, which are less than vectors.
Note that since
.code nil
is a symbol, it is ranked lower than a cons. This interpretation ensures
correct behavior when
.code nil
is regarded as an empty list, since the empty list is lexicographically prior to
a nonempty list.
Finally, if either of the arguments has a type other than the above discussed
types, the situation is an error.
.coNP Function @ greater
.synb
.mets (greater < left-obj << right-obj )
.mets (greater < obj << obj *)
.syne
.desc
The
.code
greater
function is equivalent to
.code less
with the arguments reversed. That is to say, the following
equivalences hold:
.cblk
(greater a <--> (less a) <--> t
(greater a b) <--> (less b a)
(greater a b c ...) <--> (less ... c b a)
.cble
The
.code greater
function is used as the default for the
.meta testfun
argument of the
.code pos-max
and
.code find-max
functions.
.coNP Functions @ lequal and @ gequal
.synb
.mets (lequal < obj << obj *)
.mets (gequal < obj << obj *)
.syne
.desc
The functions
.code lequal
and
.code gequal
are similar to
.code less
and
.code greater
respectively, but differ in the following respect:
when called with two arguments which compare true under the
.code equal
function, the
.code lequal
and
.code gequal
functions return
.codn t .
When called with only one argument, both functions return
.code t
and both functions generalize to three or more arguments
in the same way as do
.code less
and
.codn greater .
.coNP Function @ sort
.synb
.mets (sort < sequence >> [ lessfun <> [ keyfun ]])
.syne
.desc
The
.code sort
function destructively sorts
.metn sequence ,
producing a sequence
which is sorted according to the
.meta lessfun
and
.meta keyfun
arguments.
The
.meta keyfun
argument specifies a function which is applied to elements
of the sequence to obtain the key values which are then compared
using the lessfun. If
.meta keyfun
is omitted, the identity function is used
by default: the sequence elements themselves are their own sort keys.
The
.meta lessfun
argument specifies the comparison function which determines
the sorting order. It must be a binary function which can be invoked
on pairs of keys as produced by the key function. It must
return a
.cod2 non- nil
value if the left argument is considered to be lesser
than the right argument. For instance, if the numeric function
.code <
is used
on numeric keys, it produces an ascending sorted order. If the function
.code >
is used, then a descending sort is produced. If
.meta lessfun
is omitted, then it defaults to the generic
.code less
function.
The
.code sort
function is stable for sequences which are lists. This means that the
original order of items which are considered identical is preserved.
For strings and vectors,
.code sort
is not stable.
.coNP Function @ sort-group
.synb
.mets (sort-group < sequence >> [ keyfun <> [ lessfun ]])
.syne
.desc
The
.code sort-group
function sorts
.meta sequence
according to the
.meta keyfun
and
.meta lessfun
arguments, and then breaks the resulting sequence into groups,
based on the equivalence of the elements under
.metn keyfun .
The following equivalence holds:
.cblk
(sort-group sq lf kf) <--> (partition-by kf (sort (copy sq) kf lf))
.cble
Note the reversed order of
.meta keyfun
and
.meta lessfun
arguments between
.code sort
and
.codn sort-group .
.coNP Function @ uniq
.synb
.mets (uniq << sequence )
.syne
.desc
The
.code uniq
function returns a sequence of the same kind as
.metn sequence ,
but with
duplicates removed. Elements of
.meta sequence
are considered equal under
the
.code equal
function. The first occurrence of each element is retained,
and the subsequent duplicates of that element, of any, are suppressed,
such that the order of the elements is otherwise preserved.
The following equivalence holds between
.code uniq
and
.codn unique :
.cblk
(uniq s) <--> [unique s : :equal-based]
.cble
That is,
.code uniq
is like
.code unique
with the default
.meta keyfun
argument (the
.code identity
function) and an
.codn equal -based
hash table.
.coNP Function @ unique
.synb
.mets (uniq < sequence >> [ keyfun <> { hash-arg }* ])
.syne
.desc
The
.code unique
function is a generalization of
.codn uniq .
It returns a sequence of the same kind as
.metn sequence ,
but with duplicates removed.
If neither
.meta keyfun
nor
.metn hash-arg -s
are specified, then elements of sequence are considered equal under the
.code eql
function. The first occurrence of each element is retained,
and the subsequent duplicates of that element, of any, are suppressed,
such that the order of the elements is otherwise preserved.
If
.meta keyfun
is specified, then that function is applied to each element,
and the resulting values are compared for equality.
In other words, the behavior is as if
.meta keyfun
were the
.code identity
function.
If one or more
.metn hash-arg -s
are present, these specify the arguments for the construction of
the internal hash table used by
.codn unique .
The arguments are like those of the
.code hash
function. In particular, the argument
.code :equal-based
causes
.code unique
to use
.code equal
equality.
.coNP Function @ tuples
.synb
.mets (tuples < length < sequence <> [ fill-value ])
.syne
.desc
The
.code tuples
function produces a lazy list which represents a reorganization
of the elements of
.meta sequence
into tuples of
.metn length ,
where
.meta length
must be a positive integer.
The length of the sequence might not be evenly divisible by the tuple length.
In this case, if a
.meta fill-value
argument is specified, then the last tuple
is padded with enough repetitions of
.meta fill-value
to make it have
.meta length
elements. If
.meta fill-value
is not specified, then the last tuple is left
shorter than
.metn length .
The output of the function is a list, but the tuples themselves are sequences
of the same kind as
.metn sequence .
If
.meta sequence
is any kind of list, they
are lists, and not lazy lists.
.TP* Examples:
.cblk
(tuples 3 #(1 2 3 4 5 6 7 8) 0) -> (#(1 2 3) #(4 5 6) #(7 8 0))
(tuples 3 "abc") -> ("abc")
(tuples 3 "abcd") -> ("abc" "d")
(tuples 3 "abcd" #\ez) -> ("abc" "dzz")
(tuples 3 (list 1 2) #\ez) -> ((1 2 #\ez))
.cble
.coNP Function @ partition-by
.synb
.mets (partition-by < function << sequence )
.syne
.desc
If
.meta sequence
is empty, then
.code partition-by
returns an empty list,
and
.meta function
is never called.
Otherwise,
.code partition-by
returns a lazy list of partitions of the sequence
.metn sequence .
Partitions are consecutive, non-empty sub-strings of
.metn sequence ,
of the same kind as
.metn sequence .
The partitioning begins with the first element of
.meta sequence
being placed into a partition.
The subsequent partitioning is done according to
.metn function ,
which is applied
to each element of
.metn sequence .
Whenever, for the next element, the function
returns the same value as it returned for the previous element, the
element is placed into the same partition. Otherwise, the next element
is placed into, and begins, a new partition.
The return values of the calls to
.meta function
are compared using the
.code equal
function.
.TP* Examples:
.cblk
[partition-by identity '(1 2 3 3 4 4 4 5)] -> ((1) (2) (3 3)
(4 4 4) (5))
(partition-by (op = 3) #(1 2 3 4 5 6 7)) -> (#(1 2) #(3)
#(4 5 6 7))
.cble
.coNP Function @ make-like
.synb
.mets (make-like < list << ref-sequence )
.syne
.desc
The
.meta list
argument must be a list. If
.meta ref-sequence
is a sequence type,
then
.meta list
is converted to the same type of sequence and returned.
Otherwise the original
.meta list
is returned.
Note: the
.code make-like
function is a helper which supports the development of
unoptimized versions of a generic function that accepts any type of
sequence as input, and produces a sequence of the same type as output.
The implementation of such a function can internally accumulate a list, and
then convert the resulting list to the same type as an input value
by using
.codn make-like .
.coNP Function @ nullify
.synb
.mets (nullify << sequence )
.syne
.desc
The
.code nullify
function returns
.code nil
if
.meta sequence
is an empty sequence.
Otherwise it returns
.meta sequence
itself.
Note: the
.code nullify
function is a helper to support unoptimized generic
programming over sequences. Thanks to the generic behavior of
.codn cdr ,
any sequence can be traversed using
.code cdr
functions, checking for the
.code nil
value as a terminator. This, however, breaks for empty sequences which are not
lists, because they are not equal to
.codn nil :
to
.code car
and
.code cdr
they look like
a one-element sequence containing
.codn nil .
The
.code nullify
function reduces all empty
sequences to
.codn nil ,
thereby correcting the behavior of code which traverses
sequences using
.codn cdr ,
and tests for termination with
.codn nil .
.SS* Math Library
.coNP Functions @ + and @ -
.synb
.mets (+ << number *)
.mets (- < number << number *)
.mets (* << number *)
.syne
.desc
The
.codn + ,
.code -
and
.code *
functions perform addition, subtraction and multiplication,
respectively. Additionally, the
.code -
function performs additive inverse.
The
.code +
function requires zero or more arguments. When called with no
arguments, it produces 0 (the identity element for addition), otherwise it
produces the sum over all of the arguments.
Similarly, the
.code *
function requires zero or more arguments. When called
with no arguments, it produces 1 (the identity element for multiplication).
Otherwise it produces the product of all the arguments.
The semantics of
.code -
changes from subtraction to additive inverse
when there is only one argument. The argument is treated as a subtrahend,
against an implicit minuend of zero. When there are two or more
argument, the first one is the minuend, and the remaining are subtrahends.
When there are three or more operands, these operations are performed as if by
binary operations, in a left-associative way. That is to say,
.code (+ a b c)
means
.codn (+ (+ a b) c) .
The sum of
.code a
and
.code b
is computed first, and then this is added to
.codn c .
Similarly
.code (- a b c)
means
.codn (- (- a b) c) .
First,
.code b
is subtracted from
.codn a ,
and then
.code c
is subtracted from that result.
The arithmetic inverse is performed as if it were subtraction from integer 0.
That is,
.code (- x)
means the same thing as
.codn (- 0 x) .
The operands of
.codn + ,
.code -
and
.code *
can be characters, integers (fixnum and bignum), and
floats, in nearly any combination.
If two operands have different types, then one of them is converted to the
type of the one with the higher rank, according to this ranking:
character < integer < float. For instance if one operand is integer, and the
other float, the integer is converted to a float.
.TP* Restrictions:
Characters are not considered numbers, and participate in these operations in
limited ways. Subtraction can be used to computed the displacement between the
Unicode values of characters, and an integer displacement can be added to a
character, or subtracted from a character. For instance
.codn (- #\e9 #\e0) is 9 .
The Unicode value of a character
.code C
can be found using
.codn (- C #\ex0) :
the displacement from the NUL character.
The rules can be stated as a set of restrictions:
.RS
.IP 1
Two characters may not be added together.
.IP 2
A character may not be subtracted from an integer (which also rules out
the possibility of computing the additive inverse of a character).
.IP 3
A character operand may not be opposite to a floating point operand
in any operation.
.IP 4
A character may not be an operand of multiplication.
.RE
.PP
.coNP Functions @, / @ trunc, @ mod and @ trunc-rem
.synb
.mets (/ <> [ dividend ] << divisor )
.mets (trunc < dividend << divisor )
.mets (mod < dividend << divisor )
.mets (trunc-rem < dividend << divisor )
.syne
.desc
The arguments to these functions are numbers. Characters are not permitted.
The
.code /
function performs floating-point division. Each operands is first
converted to floating-point type, if necessary. If
.meta dividend
is omitted,
then it is taken to be
.code 1.0
and the function calculates the reciprocal.
The
.code trunc
function performs a division of
.meta dividend
by
.meta divisor
whose result
is truncated to integer toward zero. If both operands are integers, then an
integer division is performed and the result is an integer. If either operand
is a floating point value, a floating point division occurs, and the result is
truncated toward zero to a floating-point integral value.
The
.code mod
function performs a modulus operation. Firstly, the absolute value
of
.meta divisor
is taken to be a modulus. Then a residue of
.meta dividend
with respect to
.meta modulus
is calculated. The residue's sign follows
that of the sign of
.metn divisor .
That is, it is the smallest magnitude
(closest to zero) residue of
.meta dividend
with respect to the absolute
value of
.metn divisor ,
having the same sign as
.metn divisor .
If the operands are integer, the result is an integer. If either operand
is of type float, then the result is a float. The modulus operation is
then generalized into the floating point domain. For instance the expression
.code (mod 0.75 0.5)
yields a residue of 0.25 because 0.5 "goes into" 0.75 only
once, with a "remainder" of 0.25.
The
.code trunc-rem
function returns a list of two values: a
.meta quotient
and a
.metn remainder.
The
.meta quotient
is exactly the same value as what
.code trunc
would return for the same inputs.
The
.meta remainder
obeys the following identity:
.cblk
.mets (eql < remainder (- < dividend >> (* divisor << quotient )))
.cble
.coNP Functions @ wrap and @ wrap*
.synb
.mets (wrap < start < end << number )
.mets (wrap < start < end << number )
.syne
.desc
The
.code wrap
and
.code wrap*
functions reduce
.meta number
into the range specified by
.meta start
and
.metn end .
Under
.code wrap
the range is inclusive of the
.meta end
value, whereas under
.code wrap*
it is exclusive.
The following equivalence holds
.cblk
(wrap a b c) <--> (wrap* a (succ b) c)
.cble
The expression
.code (wrap* x0 x1 x)
performs the following calculation:
.cblk
.mets (+ (mod (- x x0) (- x1 x0)) x0)
.cble
In other words, first
.meta start
is subtracted from
.metn number .
Then the result is reduced modulo the displacement
between
.code start
and
.codn end .
Finally,
.meta start
is added back to that result, which is returned.
.TP* Example:
.cblk
;; perform ROT13 on the string "nop"
[mapcar (opip (+ 13) (wrap #\ea #\ez)) "nop"] -> "abc"
.cble
.coNP Functions @ gcd and @ lcm
.synb
.mets (gcd << number *)
.mets (lcm << number *)
.syne
.desc
The
.code gcd
function computes the greatest common divisor: the largest positive
integer which divides each
.metn number .
The
.code lcm
function computes the lowest common multiple: the smallest positive
integer which is a multiple of
each
.metn number .
Each
.meta number
must be an integer.
Negative integers are replaced by their absolute values, so
.code (lcm -3 -4)
is
.code 12
and
.code (gcd -12 -9)
yields
.codn 3 .
The value of
.code (gcd)
is
.code 0
and that of
.code (lcm)
is 1 .
The value of
.code (gcd x)
and
.code (lcm x)
is
.codn (abs x) .
Any arguments of
.code gcd
which are zero are effectively ignored so that
.code (gcd 0)
and
.code (gcd 0 0 0)
are both the same as
.code (gcd)
and
.code (gcd 1 0 2 0 3)
is the same as
.codn (gcd 1 2 3) .
If
.code lcm
has any argument which is zero, it yields zero.
.coNP Function @ abs
.synb
.mets (abs << number )
.syne
.desc
The
.code abs
function computes the absolute value of
.metn number .
If
.meta number
is positive, it is returned. If
.meta number
is negative, its additive inverse is
returned: a positive number of the same type with exactly the same magnitude.
.coNP Functions @ floor and @ ceil
.synb
.mets (floor << number )
.mets (ceil << number )
.syne
.desc
The
.code floor
function returns the highest integer which does not exceed
the value of
.metn number .
The ceiling function returns the lowest integer which
does not exceed the value of
.metn number .
If
.meta number
an integer, it is simply returned.
If the argument is a float, then the value returned is a float.
For instance
.code (floor 1.1)
returns 1.0 rather than 1.
.coNP Functions @, sin @, cos @, tan @, asin @, acos @ atan and @ atan2
.synb
.mets (sin << radians )
.mets (cos << radians )
.mets (tan << radians )
.mets (atan << slope )
.mets (atan2 < y << x )
.mets (asin << num )
.mets (acos << num )
.syne
.desc
These trigonometric functions convert their argument to floating point and
return a float result. The
.codn sin ,
.code cos
and
.code tan
functions compute the sine and
cosine and tangent of the
.meta radians
argument which represents an angle
expressed in radians. The
.codn atan ,
.code acos
and
.code asin
are their respective inverse
functions. The
.meta num
argument to
.code asin
and
.code acos
must be in the
range -1.0 to 1.0. The
.code atan2
function converts the rectilinear coordinates
.meta x
and
.meta y
to an angle in polar coordinates in the range [0, 2\(*p).
.coNP Functions @, exp @, log @ log10 and @ log2
.synb
.mets (exp << arg )
.mets (log << arg )
.mets (log10 << arg )
.mets (log2 << arg )
.syne
.desc
The
.code exp
function calculates the value of the transcendental number e raised to
the exponent
.metn arg .
The
.code log
function calculates the base e logarithm of
.metn arg ,
which must be a positive value.
The
.code log10
function calculates the base 10 logarithm of
.metn arg ,
which must be a positive value.
The
.code log2
function calculates the base 2 logarithm of
.metn arg ,
which must be a positive value.
.coNP Functions @, expt @ sqrt and @ isqrt
.synb
.mets (expt < base << exponent *)
.mets (sqrt << arg )
.mets (isqrt << arg )
.syne
.desc
The
.code expt
function raises
.meta base
to zero or more exponents given
by the
.meta exponent
arguments.
.code (expt x)
is equivalent to
.codn (expt x 1) ,
and yields
.code x
for all
.codn x .
For three or more arguments, the operation is right-associative.
That is to say,
.code (expt x y z)
is equivalent to
.codn (expt x (expt y z)) ,
similarly to the way nested exponents work in standard algebraic
notation.
Exponentiation is done pairwise using a binary operation.
If both operands to this binary operation are integers, then the
result is an integer. If either operand is a float, then the other
operand is converted to a float, and a floating point exponentiation
is performed. Exponentiation that would produce a complex number is
not supported.
The
.code sqrt
function produces a floating-point square root of
.metn arg ,
which is converted from integer to floating-point if necessary. Negative
operands are not supported.
The
.code isqrt
function computes the integer square root of
.metn arg ,
which must be an integer.
The integer square root is a value which is the
greatest integer that is no greater than the real square root of
.metn arg .
The input value must be an integer.
.coNP Function @ exptmod
.synb
.mets (exptmod < base < exponent << modulus )
.syne
.desc
The
.code exptmod
function performs modular exponentiation and accepts only integer
arguments. Furthermore,
.meta exponent
must be a non-negative and
.meta modulus
must be positive.
The return value is
.meta base
raised to
.metn exponent ,
and reduced to the
least positive residue modulo
.metn modulus .
.coNP Function @ cum-norm-dist
.synb
.mets (cum-norm-dist << argument )
.syne
.desc
The
.code cum-norm-dist
function calculates an approximation to the cumulative normal
distribution function: the integral, of the normal distribution function, from
negative infinity to the
.metn argument .
.coNP Functions @ n-choose-k and @ n-perm-k
.synb
.mets (n-choose-k < n << k )
.mets (n-perm-k < n << k )
.syne
.desc
The
.code n-choose-k
function computes the binomial coefficient nCk which
expresses the number of combinations of
.meta k
items that can be chosen from
a set of
.metn n ,
where combinations are subsets.
The
.code n-perm-k
function computes nPk: the number of permutations of size
.meta k
that can be drawn from a set of
.metn n ,
where permutations are sequences,
whose order is significant.
The calculations only make sense when
.meta n
and
.meta k
are nonnegative integers, and
.meta k
does not exceed
.metn n .
The behavior is not specified if these conditions
are not met.
.coNP Functions @, fixnump @, bignump @, integerp @ floatp and @ numberp
.synb
.mets (fixnump << object )
.mets (bignump << object )
.mets (integerp << object )
.mets (floatp << object )
.mets (numberp << object )
.syne
.desc
These functions test the type of
.metn object ,
returning
.code t
if it is an object
of the implied type,
.code nil
otherwise. The
.codn fixnump ,
.code bignump
and
.code floatp
functions return
.code t
if the object is of the basic type
.codn fixnum ,
.code bignum
or
.codn float .
The function
.code integerp
returns true of
.meta object
is either a
.code fixnum
or
a
.codn bignum .
The function
.code numberp
returns
.code t
if
.meta object
is either
a
.codn fixnum ,
.code bignum
or
.codn float .
.coNP Function @ zerop
.synb
.mets (zerop << number )
.syne
.desc
The
.code zerop
function tests
.meta number
for equivalence to zero. The argument must be
a number or character. It returns
.code t
for the integer value
.code 0
and for the floating-point
value
.codn 0.0 .
For other numbers, it returns
.codn nil .
It returns
.code t
for the null character
.code #\enul
and
.code nil
for all other characters.
.coNP Functions @ plusp and @ minusp
.synb
.mets (plusp << number )
.mets (minusp << number )
.syne
.desc
These functions test whether a number is positive or negative,
returning
.code t
or
.codn nil ,
as the case may be.
The argument may also be a character. All characters other than
the null character
.code #\enul
are positive. No character is negative.
.coNP Functions @ evenp and @ oddp
.synb
.mets (evenp << integer )
.mets (oddp << integer )
.syne
.desc
The
.code evenp
and
.code oddp
functions require integer arguments.
.code evenp
returns
.code t
if
.meta integer
is even (divisible by two), otherwise it returns
.codn nil .
.code oddp
returns
.code t
if
.meta integer
is not divisible by two (odd), otherwise
it returns
.codn nil .
.coNP Functions @, succ @, ssucc @, sssucc @, pred @, ppred @ and pppred
.synb
.mets (succ << number )
.mets (ssucc << number )
.mets (sssucc << number )
.mets (pred << number )
.mets (ppred << number )
.mets (pppred << number )
.syne
.desc
The
.code succ
function adds 1 to its argument and returns the resulting value.
If the argument is an integer, then the return value is the successor
of that integer, and if it is a character, then the return value
is the successor of that character according to Unicode.
The
.code pred
function subtracts 1 from its argument, and under similar considerations
as above, the result represents the predecessor.
The
.code ssucc
and
.code sssucc
functions add 2 and 3, respectively. Similarly,
.code ppred
and
.code pppred
subtract 2 and 3 from their argument.
.coNP Functions @, > @, < @, >= @ <= and @ =
.synb
.mets (> < number << number *)
.mets (> < number << number *)
.mets (>= < number << number *)
.mets (<= < number << number *)
.mets (= < number << number *)
.syne
.desc
These relational functions compare characters and numbers for numeric equality
or inequality. The arguments must be one or more numbers or characters.
If just one argument is given, then these functions all return
.codn t .
If two arguments are given then, they are compared as follows.
First, if the numbers do not have the same type, then the one
which has the lower ranking type is converted to the type of
the other, according to this ranking: character < integer < float.
For instance if a character and integer are compared, the character
is converted to integer. Then a straightforward numeric comparison
is applied.
Three or more arguments may be given, in which case the comparison proceeds
pairwise from left to right. For instance in
.codn (< a b c) ,
the comparison
.code (< a b)
is performed in isolation. If the comparison is false, then
.code nil
is returned, otherwise
the comparison
.code (< b c)
is performed in isolation, and if that is false,
.code nil
is returned, otherwise
.code t
is returned. Note that it is possible for
.code b
to
undergo two different conversions. For instance in the
.cblk
.meti (< < float < character << integer )
.cble
comparison,
.meta character
will first convert to a floating-point representation
of its Unicode value so that it can be compared to
.metn float ,
and if that comparison succeeds, then in the second comparison,
.meta character
will be converted to integer so that it can be compared to
.metn integer .
.coNP Function @ /=
.synb
.mets (/= << number *)
.syne
.desc
The arguments to
.code /=
may be numbers or characters. The
.code /=
function returns
.code t
if no two of its arguments are numerically equal. That is to say, if there
exist some
.code a
and
.code b
which are distinct arguments such that
.code (= a b)
is true, then
the function returns
.codn nil .
Otherwise it returns
.codn t .
.coNP Functions @ max and @ min
.synb
.mets (max < first-arg << arg *)
.mets (min < first-arg << args *)
.syne
.desc
The
.code max
and
.code min
functions determine and return the highest or lowest
value from among their arguments.
If only
.meta first-arg
is given, that value is returned.
These functions are type generic, since they compare arguments
using the same semantics as the
.code less
function.
If two or more arguments are given, then
.code (max a b)
is equivalent to
.codn (if (less a b) b a) ,
and
.code (min a b)
is equivalent to
.codn (if (less a b) a b) .
If the operands do not
have the same type, then one of them is converted to the type of the other;
however, the original unconverted values are returned. For instance
.code (max 4 3.0)
yields the integer
.codn 4 ,
not
.codn 4.0 .
If three or more arguments are given,
.code max
and
.code min
reduce the arguments in a left-associative manner.
Thus
.code (max a b c)
means
.codn (max (max a b) c) .
.coNP Functions @, int-str @ flo-str and @ num-str
.synb
.mets (int-str < string <> [ radix ])
.mets (flo-str << string )
.mets (num-str << string )
.syne
.desc
These functions extract numeric values from character string
.metn string .
Leading whitespace in
.metn string ,
if any, is skipped. If no digits can be successfully extracted, then
.code nil
is returned. Trailing material which does not contribute to the number is
ignored.
The
.code int-str
function converts a string of digits in the specified
radix to an integer value. If the radix isn't specified, it defaults to 10.
Otherwise it must be an integer in the range 2 to 36.
For radices above 10, letters of the alphabet
are used for digits:
.code A
represent a digit whose value is 10,
.code B
represents 11 and
so forth until
.codn Z .
For values of radix above 36, the returned value is
unspecified. Upper and lower case letters are recognized.
Any character which is not a digit of the specified radix is regarded
as the start of trailing junk at which the extraction of the digits stops.
The
.code flo-str
function converts a floating-point decimal notation to a nearby
floating point value. The material which contributes to the value
is the longest match for optional leading space, followed by a
mantissa which consists of an optional sign followed by a mixture of at least
one digit, and at most one decimal point, optionally followed by an exponent
part denoted by the letter
.code E
or
.codn e ,
an optional sign and one or more optional exponent digits.
The
.code num-str
function converts a decimal notation to either an integer as if by
a radix 10 application of
.codn int-str ,
or to a floating point value as if by
.codn flo-str .
The floating point interpretation is chosen if the possibly empty
initial sequence of digits (following any whitespace and optional sign) is
followed by a period, or by
.code e
or
.codn E .
.coNP Functions @ int-flo and @ flo-int
.synb
.mets (int-flo << float )
.mets (flo-int << integer )
.syne
.desc
These functions perform numeric conversion between integer and floating point
type. The
.code int-flo
function returns an integer by truncating toward zero.
The
.code flo-int
function returns an exact floating point value corresponding to
.metn integer ,
if possible, otherwise an approximation using a nearby
floating point value.
.coNP Functions @ tofloat and @ toint
.synb
.mets (tofloat << value )
.mets (toint < value <> [ radix ])
.syne
.desc
These convenience functions convert
.meta value
to floating-point or integer, respectively.
If a floating-point value is passed into tofloat, or an integer value into
toint, then the value is simply returned.
If
.meta value
is a character, then it is treated as a string of length one
containing that character.
If
.meta value
is a string, then it is converted by
.code tofloat
as if by the function
.metn flo-str ,
, and by
.code toint
as if by the function
.codn int-str .
If
.meta value
is an integer, then it is converted by
.code tofloat
as if by the function
.codn flo-int .
If
.meta value
is a floating-point number, then it is converted by
.code toint
as if by the function
.codn int-flo .
.coNP Variables @, *flo-min* @, *flo-max* and @ *flo-epsilon*
.desc
These variables hold, respectively: the smallest positive floating-point
value; the largest positive floating-point value; and the difference
between 1.0 and the smallest representable value greater than 1.0.
.code *flo-min*
and
.code *flo-max*
define the floating-point range, which consists
of three regions: values from
.code (- *flo-max*)
to
.codn (- *flo-min*) ;
the value 0.0, and values from
.code *flo-min*
to
.codn *flo-max* .
.coNP Variable @ *flo-dig*
.desc
This variable holds an integer representing the number of decimal digits
in a decimal floating-point number such that this number can be converted
to a \*(TX floating-point number, and back to decimal, without a change in any of
the digits. This holds regardless of the value of the number, provided that it
does not exceed the floating-point range.
.coNP Variables @ *pi* and @ *e*
.desc
These variables hold an approximation of the mathematical constants \(*p and e.
To four digits of precision, \(*p is 3.142 and e is 2.718. The
.code *pi*
and
.code *e*
approximations are accurate to
.code *flo-dig*
decimal digits.
.SS* Bit Operations
In \*(TL, similarly to Common Lisp, bit operations on integers are based
on a concept that might be called "infinite two's-complement".
Under infinite two's complement, a positive number is regarded as having
a binary representation prefixed by an infinite stream of zero digits (for
example
.code 1
is
.codn ...00001 ).
A negative number
in infinite two's complement is the biwise negation of its positive counterpart,
plus one: it carries an infinite prefix of 1 digits. So for instance the number
.code -1
is represented by
.codn ...11111111 :
an infinite sequence of
1
bits. There
is no specific sign bit; any operation which produces such an infinite sequence
of 1 digits on the left gives rise to a negative number. For instance, consider the
operation of computing the bitwise complement of the number
.codn 1 .
Since the
number
.code 1
is represented as
.codn ...0000001 ,
its complement is
.codn ...11111110 .
Each one of the
.cod 0
digits in the infinite sequence is replaced by
.codn 1 ,
And this leading sequence means that the number
is negative, in fact corresponding to the two's-complement representation of
the value
.codn -2 .
Hence, the infinite digit concept corresponds to an arithmetic
interpretation.
In fact \*(TL's bignum integers do not use a two's complement
representation internally. Numbers are represented as an array which holds a
pure binary number. A separate field indicates the sign: negative,
or non-negative. That negative numbers appear as two's-complement under the
bit operations is merely a carefully maintained illusion (which makes bit
operations on negative numbers more expensive).
The
.code logtrunc
function, as well as a feature of the
.code lognot
function, allow bit
manipulation code to be written which works with positive numbers only, even if
complements are required. The trade off is that the application has to manage a
limit on the number of bits.
.coNP Functions @, logand @, logior and @ logxor
.synb
.mets (logand << integer *)
.mets (logior << integer *)
.mets (logxor < int1 << int2 )
.syne
.desc
These operations perform the familiar bitwise and, inclusive or, and exclusive
or operations, respectively. Positive values inputs are treated as
pure binary numbers. Negative inputs are treated as infinite-bit
two's-complement.
For example
.code (logand -2 7)
produces
.codn 6 .
This is because
.code -2
is
.code ...111110
in infinite-bit two's-complement. And-ing this value with
.code 7
(or
.codn ...000111 )
produces
.codn 110 .
The
.code logand
and
.code logior
functions are variadic, and may be called with zero, one,
two, or more input values. If
.code logand
is called with no arguments, it produces
the value -1 (all bits 1). If
.code logior
is called with no arguments it produces
zero. In the one-argument case, the functions just return their argument value.
.coNP Function @ logtest
.synb
.mets (logtest < int1 << int2 )
.syne
.desc
The
.code logtest
function returns true if
.meta int1
and
.meta int2
have bits in
common. The following equivalence holds:
.cblk
(logtest a b) <--> (not (zerop (logand a b)))
.cble
.coNP Functions @ lognot and @ logtrunc
.synb
.mets (lognot < value <> [ bits ])
.mets (logtrunc < value << bits )
.syne
.desc
The
.code lognot
function performs a bitwise complement of
.metn value .
When the one-argument form of lognot is used, then if
.meta value
is nonnegative,
then the result is negative, and vice versa, according to the infinite-bit
two's complement representation. For instance
.code (lognot -2)
is
.codn 1 ,
and
.code (lognot 1)
is
.codn -2 .
The two-argument form of
.code lognot
produces a truncated complement. Conceptually,
a bitwise complement is first calculated, and then the resulting number is
truncated to the number of bits given by
.metn bits ,
which must be a nonnegative integer. The following equivalence holds:
.cblk
(lognot a b) <--> (logtrunc (lognot a) b)
.cble
The
.code logtrunc
function truncates the integer
.meta value
to the specified number
of bits. If
.meta value
is negative, then the two's-complement representation
is truncated. The return value of
.code logtrunc
is always a non-negative integer.
.coNP Function @ sign-extend
.synb
.mets (sign-extend < value << bits )
.syne
.desc
The
.code sign-extend
function first truncates the infinite-bit two's complement representation of
the integer
.meta value
to the specified number of bits, similarly to the
.code logtrunc
function. Then, this truncated value is regarded as a
.meta bits
wide two's complement integer. The value of this integer is
calculated and returned.
.TP* Examples:
.cblk
(sign-extend 127 8) -> 127
(sign-extend 128 8) -> -128
(sign-extend 129 8) -> -127
(sign-extend 255 8) -> -1
(sign-extend 256 8) -> 0
(sign-extend -1 8) -> -1
(sign-extend -255 8) -> 0
.cble
.coNP Function @ ash
.synb
.mets (ash < value << bits )
.syne
.desc
The
.code ash
function shifts
.meta value
by the specified number of
.meta bits
producing a
new value. If
.meta bits
is positive, then a left shift takes place. If
.meta bits
is negative, then a right shift takes place. If
.meta bit
is zero, then
.meta value
is returned unaltered. For positive numbers, a left shift by n bits is
equivalent to a multiplication by two to the power of n, or
.codn (expt 2 n) .
A right shift by n bits of a positive integer is equivalent to integer
division by
.codn (expt 2 n) ,
with truncation toward zero.
For negative numbers, the bit shift is performed as if on the two's-complement
representation. Under the infinite two's-complement representation,
a right shift does not exhaust the infinite sequence of
.code 1
digits which
extends to the left. Thus if
.code -4
is shifted right it becomes
.code -2
because
the bitwise representations of these values are
.code ...111100
and
.codn ...11110 .
.coNP Function @ bit
.synb
.mets (bit < value << bit )
.syne
.desc
The
.code bit
function tests whether the integer
.meta value
has a 1 in bit position
.metn bit .
The
.meta bit
argument must be a non-negative integer. A value of zero of
.meta bit
indicates the least significant bit position of
.metn value .
The
.code bit
function has a boolean result, returning the symbol
.code t
if bit
.meta bit
of
.meta value
is set, otherwise
.codn nil .
If
.meta value
is negative, it is treated as if it had an infinite-bit two's
complement representation. For instance, if value is
.codn -2 ,
then the bit
function returns
.code nil
for a
.meta bit
value of zero, and
.code t
for all other values,
since the infinite bit two's complement representation of
.code -2
is
.codn ...11110 .
.coNP Function @ mask
.synb
.mets (mask << integer *)
.syne
.desc
The
.code mask
function takes zero or more integer arguments, and produces an integer
value which corresponds a bitmask made up of the bit positions specified by the
integer values.
If
.code mask
is called with no arguments, then the return value is zero.
If
.code mask
is called with a single argument
.meta integer
then the return value is the same as
that of the expression
.codn (ash 1 <integer>) :
the value 1 shifted left by
.meta integer
bit positions. If
.meta integer
is zero, then the result is
.codn 1 ;
if
.meta integer
is
.codn 1 ,
the
result is
.code 2
and so forth. If
.meta value
is negative, then the result is zero.
If
.code mask
is called with two or more arguments, then the result is a bitwise of
the masks individually computed for each of the values.
In other words, the following equivalences hold:
.cblk
(mask) <--> 0
(mask a) <--> (ash 1 a)
(mask a b c ...) <--> (logior (mask a) (mask b) (mask c) ...)
.cble
.coNP Function @ width
.synb
.mets (width << integer *)
.syne
.desc
A two's complement representation of an integer consists of a sign bit and a
manitssa field.
The
.code width
function computes the minimum number of bits required for the mantissa portion
of the two's complement representation of the
.meta integer
argument.
For a nonnegative argument, the width also corresponds to the number of bits
required for a natural binary representation of that value.
Two integer values have a width of zero, namely 0 and -1. This means that these
two values can be represented in a one-bit two's complement, consisting of only
a sign bit: the one-bit two's complement bitfield 1 denotes -1, and 0 denotes
0.
Similarly, two integer values have a width of 1: 1 and -2. The two-bit
two's complement bitfield 01 denotes 1, and 10 denotes -2.
The argument may be a character.
.SS* Exceptions
.coNP Functions @, throw @ throwf and @ error
.synb
.mets (throw < symbol << arg *)
.mets (throwf < symbol < format-string << format-arg *)
.mets (error < format-string << format-arg *)
.syne
.desc
These functions generate an exception. The
.code throw
and
.code throwf
functions generate
an exception identified by
.metn symbol ,
whereas
.code error
throws an exception of
type
.codn error .
The call
.code (error ...)
can be regarded as a shorthand for
.codn (throwf 'error ...) .
The
.code throw
function takes zero or more additional arguments. These arguments
become the arguments of a
.code catch
handler which takes the exception. The
handler will have to be capable of accepting that number of arguments.
The
.code throwf
and
.code error
functions generate an exception which has a single
argument: a character string created by a formatted print to a string stream
using the
.code format
string and additional arguments.
.coNP Operator @ catch
.synb
.mets (catch < try-expression
.mets \ \ >> {( symbol <> ( arg *) << body-form *)}*)
.syne
.desc
The
.code catch
operator establishes an exception catching block around
the
.metan try-expression .
The
.meta try-expression
is followed by zero or more
catch clauses. Each catch clause consists of a symbol which denotes
an exception type, an argument list, and zero or more body forms.
If
.meta try-expression
terminates normally, then the catch clauses
are ignored. The catch itself terminates, and its return value is
that of the
.metn try-expression .
If
.meta try-expression
throws an exception which is a subtype of one or more of
the type symbols given in the exception clauses, then the first (leftmost) such
clause becomes the exit point where the exception is handled.
The exception is converted into arguments for the clause, and the clause
body is executed. When the clause body terminates, the catch terminates,
and the return value of the catch is that of the clause body.
If
.meta try-expression
throws an exception which is not a subtype of any of
the symbols given in the clauses, then the search for an exit point for
the exception continues through the enclosing forms. The catch clauses
are not involved in the handling of that exception.
When a clause catches an exception, the number of arguments in the catch must
match the number of elements in the exception. A catch argument list
resembles a function or lambda argument list, and may be dotted. For instance
the clause
.code (foo (a . b))
catches an exception subtyped from
.codn foo ,
with one or
more elements. The first element binds to parameter
.codn a ,
and the rest, if any,
bind to parameter
.codn b .
If there is only one element,
.code b
takes on the value
.codn nil .
Also see: the
.code unwind-protect
operator, and the functions
.codn throw ,
.code throwf
and
.codn error .
.coNP Macro @ ignerr
.synb
.mets (ignerr << form *)
.syne
.desc
The
.code ignerr
macro operator evaluates each
.meta form
similarly to the
.code progn
operator. If no forms are present, it returns
.codn nil .
Otherwise it evaluates each
.meta form
in turn, yielding the value of the last one.
If the evaluation of any
.meta form
is abandoned due to an exception of type
.codn error ,
the code generated by the
.code ignerr
macro catches this exception. In this situation,
the execution of the
.code ignerr
form terminates without evaluating the remaining
forms, and yields
.codn nil .
.coNP Variable @ *unhandled-hook*
The
.code *unhandled-hook*
variable is initialized with
.code nil
by default.
It may instead be assigned a function which is capable of taking
three arguments.
When an exception occurs which has no handler, this function is called,
with the following arguments: the exception type symbol, the exception object,
and a third value which is either
.code nil
or else the form which was being evaluated the exception was thrown.
Otherwise, if the variable is
.code nil
some informational messages are printed about the exception, and the process
exits with a failed termination status.
In the same situation, if the variable contains an object which is not a
function, the process terminates abnormally as if by a call to the
.code abort
function.
Prior to the function being called, the
.code *unhandled-hook*
variable is reset to
.codn nil .
If the function registered in
.code *unhandled-hook*
returns, the process exits with a failed termination status.
Note: the functions
.code source-loc
or
.code source-loc-str
may be applied to the third argument of the
.code *unhandled-hook*
function to obtain more information about the form.
.SS* Regular Expression Library
.coNP Functions @ search-regex and @ range-regex
.synb
.mets (search-regex < string < regex >> [ start <> [ from-end ]])
.mets (range-regex < string < regex >> [ start <> [ from-end ]])
.mets (range-regst < string < regex >> [ start <> [ from-end ]])
.syne
.desc
The
.code search-regex
function searches through
.meta string
starting
at position
.meta start
for a match for
.metn regex .
If
.meta start
is omitted, the search starts at position 0. If
.meta from-end
is specified and has a
.cod2 non- nil
value, the search
proceeds in reverse, from the last position in the string, toward
.metn start .
This function returns
.code nil
if no match is found, otherwise it returns
a cons, whose
.code car
indicates the position of the match, and whose
.code cdr
indicates the length of the match.
The
.code range-regex
function is similar to
.codn search-regex ,
except that when
a match is found, it returns a position range, rather than a position
and length. A cons is returned whose
.code car
indicates the position
of the match, and whose
.code cdr
indicates the position one element past the
last character of the match. If the match is empty, the two integers
are equal.
The
.code search-regst
differs from
.code search-regex
in the representation of the return value in the matching case.
Rather than returning the position and length of the match,
it returns the matching substring of
.metn string .
.coNP Functions @ match-regex and @ match-regst
.synb
.mets (match-regex < string < regex <> [ position ])
.mets (match-regst < string < regex <> [ position ])
.syne
.desc
The
.code match-regex
function tests whether
.meta regex
matches at
.meta position
in
.metn string .
If
.meta position
is not specified, it is taken to be zero.
If the regex matches, then the length of the match is returned.
If it does not match, then
.code nil
is returned.
The
.code match-regst
differs from
.code match-regex
in the representation of the return value in the matching case.
Rather than returning the length of the match, it returns
matching substring of
.metn string .
.coNP Functions @ match-regex-right and @ match-regst-right
.synb
.mets (match-regex-right < string < regex <> [ end-position ])
.mets (match-regst-right < string < regex <> [ end-position ])
.syne
.desc
The
.code match-regex
function tests whether
.meta string
contains a match which ends
precisely on the character just before
.metn end-position .
If
.meta end-position
is not specified, it defaults to the length of the string, and the function
performs a right-anchored regex match.
If a match is found, then the length of the match is returned.
The match must terminate just before
.meta end-position
in the sense that
additional characters at
.meta end-position
and beyond can no longer satisfy the
regular expression. More formally, the function searches, starting from
position zero, for positions where there occurs a match for the regular
expression, taking the longest possible match. The length of first such a match
which terminates on the character just before
.meta end-position
is returned.
If no such a match is found, then
.code nil
is returned.
The
.code match-regst-right
differs from
.code match-regst-right
in the representation of the return value in the matching case.
Rather than returning the length of the match, it returns
matching substring of
.metn string .
.TP* Examples:
.cblk
;; Return matching portion rather than length thereof.
(defun match-regex-right-substring (str reg : end-pos)
(set end-pos (or end-pos (length str)))
(let ((len (match-regex-right str reg end-pos)))
(if len
[str (- end-pos len)..end-pos]
nil)))
(match-regex-right-substring "abc" #/c/) -> ""
(match-regex-right-substring "acc" #/c*/) -> "cc"
;; Regex matches starting at multiple positions, but all
;; the matches extend past the limit.
(match-regex-right-substring "acc" #/c*/ 2) -> nil
;; If the above behavior is not wanted, then
;; we can extract the string up to the limiting
;; position and do the match on that.
(match-regex-right-substring ["acc" 0..2] #/c*/) -> "c"
;; Equivalent of above call
(match-regex-right-substring "ac" #/c*/) -> "c"
.cble
.coNP Function @ regsub
.synb
.mets (regsub < regex < replacement << string )
.syne
.desc
The
.code regsub
function searches
.meta
string
for multiple occurrences of
non-overlapping matches for
.metn regex .
A new string is constructed
similar to
.meta string
but in which each matching region is replaced
with using
.meta replacement
as follows.
The
.meta replacement
object may be a character or a string, in which
case it is simply taken to be the replacement for each match
of the regular expression.
The
.meta replacement
object may be a function of one argument, in
which case for every match which is found, this function is invoked,
with the matching piece of text as an argument. The function's
return value is then taken to be the replacement text.
.TP* Examples:
.cblk
;; match every lower case e or o, and replace by filtering
;; through the upcase-str function:
[regsub #/[eo]/ upcase-str "Hello world!"] -> "HEllO wOrld!"
;; Replace Hello with Goodbye:
(regsub #/Hello/ "Goodbye" "Hello world!") -> "Goodbye world!"
.cble
.coNP Function @ regexp
.synb
.mets (regexp << obj )
.syne
.desc
The
.code regexp
function returns
.code t
if
.meta obj
is a compiled regular expression
object. For any other object type, it returns
.codn nil .
.coNP Function @ regex-compile
.synb
.mets (regex-compile < form-or-string <> [ error-stream ])
.syne
.desc
The
.code regex-compile
function takes the source code of a regular expression,
expressed as a Lisp data structure representing an abstract syntax tree, or
else a regular expression specified as a character string, and compiles it to a
regular expression object.
If
.meta form-or-string
is a character string, it is parsed to an
abstract syntax tree first, if by the
.code regex-parse
function.
If the parse is successful (the result is not
.codn nil )
then
the resulting tree structure is compiled by a recursive call to
.codn regex-compile .
The optional
.meta error-stream
argument is passed down to
.code regex-parse
as well as in the recursive call to
.codn regex-compile ,
if that call takes place.
If
.meta error-stream
is specified, it must be a stream. Any error diagnostics are sent to that
stream.
.TP* Examples:
.cblk
;; the equivalent of #/[a-zA-Z0-9_/
(regex-compile '(set (#\ea . #\ez) (#\eA . #\eZ)
(#\e0 . #\e9) #\e_))
;; the equivalent of #/.*/ and #/.+/
(regex-compile '(0+ wild))
(regex-compile '(1+ wild))
;; #/a|b|c/
(regex-compile '(or (or #\ea #\eb) #\ec))
;; string
(regex-compile "a|b|c")
.cble
.coNP Function @ regex-parse
.synb
.mets (regex-parse < string <> [ error-stream ])
.syne
.desc
The
.code regex-parse
function parses a character string which contains a regular expression
(without any surrounding / characters) and turns it into a Lisp data structure
(the abstract syntax tree representation of the regular expression).
The regular expression syntax
.code #/RE/
produces the same structure, but as a
literal which is processed at the time \*(TX source code is read; the
.code regex-parse
function performs this parsing at run-time.
If there are parse errors, the function returns
.codn nil .
The optional
.meta error-stream
argument specifies a stream to which error messages
are sent from the parser. By default, diagnostic output goes to the
.code *stdnull*
stream, which discards it. If
.meta error-stream
is specified as
.codn t ,
then the diagnostic output goes to the
.code *stdout*
stream.
If
.code regex-parse
returns a
.cod2 non- nil
value, that structure is then something
which is suitable as input to
.codn regex-compile .
.SS* Hashing Library
.coNP Functions @, make-hash and @ hash
.synb
.mets (make-hash < weak-keys < weak-vals << equal-based )
.mets (hash { :weak-keys | :weak-vals | :equal-based }*)
.syne
.desc
These functions construct a new hash table.
A hash table is an object which retains an association between pairs of
objects. Each pair consists of a key and value. Given an object which is
similar to a key in the hash table, it is possible to retrieve the
corresponding value. Entries in a hash table are not ordered in any way, and
lookup is facilitated by hashing: quickly mapping a key object to a numeric
value which is then used to index into one of many buckets where the matching
key will be found (if such a key is present in the hash table).
.code make-hash
takes three mandatory boolean arguments. The
.meta weak-keys
argument specifies whether the hash table shall have weak keys. The
.meta weak-vals
argument specifies whether it shall have weak values, and
.meta equal-based
specifies whether it is
.codn equal- based.
The hash function defaults
all three of these properties to false, and allows them to be overridden to
true by the presence of keyword arguments.
It is an error to attempt to construct an
.codn equal -based
hash table which has weak keys.
The hash function provides an alternative interface. It accepts optional
arguments which are keyword symbols. Any combination of the three symbols
.codn :weak-keys ,
.code :weak-vals
and
.code :equal-based
can be specified in any order
to turn on the corresponding properties in the newly constructed hash table.
If any of the keywords is not specified, the corresponding property defaults to
.codn nil .
If a hash table has weak keys, this means that from the point of view
of garbage collection, that table holds only weak references to the keys
stored in it. Similarly, if a hash table has weak values, it means that it
holds a weak reference to each value stored. A weak reference is one
which does not prevent the reclamation of an object by the garbage
collector. That is to say, when the garbage collector discovers that the only
references to some object are weak references, then that object is considered
garbage, just as if it had no references to it. The object is reclaimed, and
the weak references "lapse" in some way, which depends on what kind they are.
Hash table weak references lapse by entry removal: if either a key or a value
object is reclaimed, then the corresponding key-value entry is erased from the
hash table.
Important to the operation of a hash table is the criterion by which keys are
considered same. By default, this similarity follows the eql function. A hash
table will search for a stored key which is
.code eql
to the given search key.
A hash table constructed with the
.codn equal -based
property compares keys using
the
.code equal
function instead.
In addition to storing key-value pairs, a hash table can have a piece of
information associated with it, called the user data.
A hash table can be traversed to visit all of the keys and data. The order of
traversal bears no relation to the order of insertion, or to any properties of
the key type.
During an open traversal, new keys can be inserted into a hash table or deleted
from it while a a traversal is in progress. Insertion of a new key during
traversal will not cause any existing key to be visited twice or to be skipped;
however, it is not specified whether the new key will be traversed. Similarly,
if a key is deleted during traversal, and that key has not yet been visited, it
is not specified whether it will be visited during the remainder of the
traversal.
An open traversal of a hash table is performed by the
.code maphash
function and the
.code dohash
operator. The traversal is open because code supplied by the program
is evaluated for each entry.
The functions
.codn hash-keys ,
.codn hash-values ,
.codn hash-pairs ,
and
.code hash-alist also perform an open traversal, because they return
lazy lists. The traversal isn't complete until the returned lazy list
is fully instantiated. In the meanwhile, the
\*(TX program can mutate the hash table from which the lazy list
is being generated.
.coNP Function @ hash-construct
.synb
.mets (hash-construct < hash-args << key-val-pairs )
.syne
.desc
The
.code hash-construct
function constructs a populated hash in one step. The
.meta hash-args
argument specifies a list suitable as an argument list in a call to the hash
function. The
.meta key-val-pairs
is a sequence of pairs, which are two-element
lists representing key-value pairs.
A hash is constructed as if by a call to
.cblk
.meti [apply hash << hash-args ],
.cble
then populated
with the specified pairs, and returned.
.coNP Function @ hash-update
.synb
.mets (hash-update < hash << function )
.syne
.desc
The
.code hash-update
function replaces each values in
.metn hash ,
with the value of
.meta function
applied to that value.
The return value is
.metn hash .
.coNP Function @ hash-update-1
.synb
.mets (hash-update-1 < hash < key < function <> [ init ])
.syne
.desc
The
.code hash-update-1
function operates on a single entry in the hash table.
If
.meta key
exists in the hash table, then its corresponding value is passed
into
.metn function ,
and the return value of
.meta function
is then installed
in place of the key's value. The value is then returned.
If
.meta key
does not exist in the hash table, and no
.meta init
argument is given,
then
.code hash-update-1
does nothing and returns
.codn nil .
If
.meta key
does not exist in the hash table, and an
.meta init
argument is given,
then
.meta function
is applied to
.metn init ,
and then
.meta key
is inserted into
.meta hash
with the value returned by
.meta function
as the datum. This value
is also returned.
.coNP Function @ group-by
.synb
.mets (group-by < func < sequence << option *)
.syne
.desc
The
.code group-by
function produces a hash table from
.metn sequence ,
which is a
list or vector. Entries of the hash table are not elements of
.metn sequence ,
but lists of elements of
.metn sequence .
The function
.meta func
is applied to
each element of
.meta sequence
to compute a key. That key is used to determine
which list the item is added to in the hash table.
The trailing arguments
.cblk
.meti << option *
.cble
if any, consist of the same keywords
that are understood by the hash function, and determine the properties
of the hash.
.TP* Example:
Group the integers from 0 to 10 into three buckets keyed on 0, 1 and 2
according to the modulo 3 congruence:
.cblk
(group-by (op mod @1 3) (range 0 10)))
-> #H(() (0 (0 3 6 9)) (1 (1 4 7 10)) (2 (2 5 8)))
.cble
.coNP Functions @ make-similar-hash and @ copy-hash
.synb
.mets (make-similar-hash << hash )
.mets (copy-hash << hash )
.syne
.desc
The
.code make-similar-hash
and copy-hash functions create a new hash object based on
the existing
.meta hash
object.
.code make-similar-hash
produces an empty hash table which inherits all of the
attributes of
.metn hash .
It uses the same kind of key equality, the
same configuration of weak keys and values, and has the same user data (see
the
.code set-hash-userdata
function).
The
.code copy-hash
function is like
.codn make-similar-hash ,
except that instead of
producing an empty hash table, it produces one which has all the same elements
as
.metn hash :
it contains the same key and value objects.
.coNP Function @ inhash
.synb
.mets (inhash < hash < key <> [ init ])
.syne
.desc
The
.code inhash
function searches hash table
.meta hash
for
.metn key .
If
.meta key
is found, then it return the hash table's cons cell which
represents the association between
.meta hash
and
.metn key .
Otherwise, it returns
.codn nil .
If argument
.meta init
is specified, then the function will create
an entry for
.meta key
in
.meta hash
whose value is that of
.metn init .
The cons cell representing that association is returned.
Note: for as long as the
.meta key
continues to exist inside
.metn hash .
modifying the
.code car
field of the returned cons has ramifications for the logical integrity of
the hash. Modifying the
.code cdr
field has the effect of updating the association with a new value.
.coNP Function @ gethash
.synb
.mets (gethash < hash < key <> [ alt ])
.syne
.desc
The
.code gethash
function searches hash table
.meta hash
for key
.metn key .
If the
key is found then the associated value is returned. Otherwise, if
the
.meta alt
argument was specified, it is returned. If the
.meta alt
argument
was not specified,
.code nil
is returned.
.coNP Function @ sethash
.synb
.mets (sethash < hash < key << value )
.syne
.desc
The
.code sethash
function places a value into
.meta hash
table under the given
.metn key .
If a similar key already exists in the hash table, then that key's
value is replaced by
.metn value .
Otherwise, the
.meta key
and
.meta value
pair is
newly inserted into
.metn hash .
.coNP Function @ pushhash
.synb
.mets (pushhash < hash < key << element )
.syne
.desc
The
.code pushhash
function is useful when the values stored in a hash table
are lists. If the given
.meta key
does not already exist in
.metn hash ,
then a list of
length one is made which contains
.metn element ,
and stored in
.meta hash
table under
.metn key .
If the
.meta key
already exists in the hash table, then the corresponding
value must be a list. The
.meta element
value is added to the front of that list,
and the extended list then becomes the new value under
.metn key .
.coNP Function @ remhash
.synb
.mets (remhash < hash << key )
.syne
.desc
The
.code remhash
function searches
.meta hash
for a key similar to the
.metn key .
If that key is found, then that key and its corresponding value are
removed from the hash table.
.coNP Function @ hash-count
.synb
.mets (hash-count << hash )
.syne
.desc
The
.code hash-count
function returns an integer representing the number of
key-value pairs stored in
.metn hash .
.coNP Function @ get-hash-userdata
.synb
.mets (get-hash-userdata << hash )
.syne
.desc
This function retrieves the user data object associated with
.metn hash .
The user data object of a newly-created hash table is initialized to
.codn nil .
.coNP Function @ set-hash-userdata
.synb
.mets (set-hash-userdata < hash << object )
.syne
.desc
The
.code set-hash-userdata
replaces, with the
.metn object ,
the user data object
associated with
.metn hash .
.coNP Function @ hashp
.synb
.mets (hashp << object )
.syne
.desc
The
.code hashp
function returns
.code t
if the
.meta object
is a hash table,
otherwise it returns
.codn nil .
.coNP Function @ maphash
.synb
.mets (maphash < hash << binary-function )
.syne
.desc
The
.code maphash
function successively invokes
.meta binary-function
for each entry stored in
.metn hash .
Each entry's key and value are passed as arguments
to
.codn binary-function .
The function returns
.codn nil .
.coNP Functions @ hash-eql and @ hash-equal
.synb
.mets (hash-eql << object )
.mets (hash-equal << object )
.syne
.desc
These functions each compute an integer hash value from the internal
representation of
.metn object ,
which satisfies the following properties.
If two objects
.code A
and
.code B
are the same under the
.code eql
function, then
.code (hash-eql A)
and
.code (hash-eql B)
produce the same integer hash value. Similarly,
if two objects
.code A
and
.code B
are the same under the
.code equal
function, then
.code (hash-equal A)
and
.code (hash-equal B)
each produce the same integer hash value. In all other
circumstances, the hash values of two distinct objects are unrelated, and
may or may not be the same.
.coNP Functions @, hash_keys @, hash_values @ hash_pairs and @ hash_alist
.synb
.mets (hash-keys << hash )
.mets (hash-values << hash )
.mets (hash-pairs << hash )
.mets (hash-alist << hash )
.syne
.desc
These functions retrieve the bulk key-value data of hash table
.meta hash
in various ways.
.code hash-keys
retrieves a list of the keys.
.code hash-values
retrieves a list of the values.
.code hash-pairs
retrieves a list of pairs,
which are two-element lists consisting of the key, followed by the value.
Finally,
.code hash-alist
retrieves the key-value pairs as a Lisp association list:
a list of cons cells whose
.code car
fields are keys, and whose
.code cdr
fields are the values. Note that
.code hash-alist
returns the actual entries from the hash table, which are
conses. Modifying the
.code cdr
fields of these conses constitutes modifying the hash values
in the original hash table. Modifying the
.code car
fields interferes with the integrity of the hash table.
These functions all retrieve the keys and values in the
same order. For example, if the keys are retrieved with
.codn hash-keys ,
and the values with
.codn hash-values ,
then the corresponding entries from
each list pairwise correspond to the pairs in
.metn hash .
The list returned by each of these functions is lazy, and hence constitutes
an open traversal of the hash table.
.coNP Operator @ dohash
.synb
.mets (dohash ( < key-var < value-var < hash-form <> [ result-form ])
.mets \ \ << body-form *)
.syne
.desc
The
.code dohash
operator iterates over a hash table. The
.meta hash-form
expression must
evaluate to an object of hash table type. The
.meta key-var
and
.meta value-var
arguments must be symbols suitable for use as variable names.
Bindings are established for these variables over the scope of the
.metn body-form s
and the optional
.metn result-form .
For each element in the hash table, the
.meta key-var
and
.meta value-var
variables are set to the key and value of that entry, respectively,
and each
.metn body-form ,
if there are any, is evaluated.
When all of the entries of the table are thus processed, the
.meta result-form
is evaluated, and its return value becomes the return value of the dohash form.
If there is no
.metn result-form ,
the return value is
.codn nil .
The
.meta result-form
and
.metn body-form s
are in the scope of an implicit anonymous
block, which means that it is possible to terminate the execution of
dohash early using
.cblk
.meti (return << value )
.cble
or
.codn (return) .
.coNP Functions @, hash-uni @ hash-diff and @ hash-isec
.synb
.mets (hash-uni < hash1 < hash2 <> [ join-func ])
.mets (hash-diff < hash1 << hash2 )
.mets (hash-isec < hash1 < hash2 <> [ join-func ])
.syne
.desc
These functions perform basic set operations on hash tables in a nondestructive
way, returning a new hash table without altering the inputs. The arguments
.meta hash1
and
.meta hash2
must be compatible hash tables. This means that their keys
must use the same kind of equality.
The resulting hash table inherits attributes from
.metn hash1 ,
as if created by the
.code make-similar-hash
function. If
.meta hash1
has userdata, the resulting hash table
has the same userdata. If
.meta hash1
has weak keys, the resulting table has weak
keys, and so forth.
The
.code hash-uni
function performs a set union. The resulting hash contains all of
the keys from
.meta hash1
and all of the keys from
.metn hash2 ,
and their corresponding
values. If a key occurs both in
.meta hash1
and
.metn hash2 ,
then it occurs only once
in the resulting hash. In this case, if the
.meta join-func
argument is not given,
the value associated with this key is the one from
.metn hash1 .
If
.meta join-func
is specified then it is called with two arguments: the respective
data items from
.meta hash1
and
.metn hash2 .
The return value of this function is used
as the value in the union hash.
The
.code hash-diff
function performs a set difference. First, a copy of
.meta hash1
is made as if by the
.code copy-hash
function. Then from this copy, all keys which occur
in
.code hash2
are deleted.
The
.code hash-isec
function performs a set intersection. The resulting hash contains
only those keys which occur both in
.meta hash1
and
.metn hash2 .
If
.meta join-func
is not
specified, the values selected for these common keys are those from
.metn hash1 .
If
.meta join-func
is specified, then for each key which occurs in both
.meta hash1
and
.metn hash2 ,
it is called with two arguments: the respective data items. The return
value is then used as the data item in the intersection hash.
.SS* Partial Evaluation and Combinators
.coNP Macros @ op and @ do
.synb
.mets (op << form +)
.mets (do << form +)
.syne
.desc
The
.code op
and
.code do
macro operators are similar.
Like the lambda operator, the
.code op
operator creates an anonymous function based on its syntax.
The difference is that the arguments of the function are implicit, or
optionally specified within the function body, rather than as a formal
parameter list before the body.
Also, the
.meta form
arguments of
.code op
are implicitly turned into a DWIM expression,
which means that argument evaluation follows Lisp-1 rules. (See the
.code dwim
operator).
The.
.code do
operator is like the
.code op
operator with the following difference:
the
.meta form
arguments of
.code op
are not implicitly treated as a DWIM expression,
but as an ordinary expression. In particular, this means that operator
syntax is permitted. Note that the syntax
.code (op @1)
makes sense, since
the argument can be a function, which will be invoked, but
.code (do @1)
doesn't
make sense because it will produce a Lisp-2 form like
.code (#:arg1 ...)
referring
to nonexistent function
.codn #:arg1 .
Because it accepts operators,
.code do
can be used with imperative constructs
which are not functions, like set: like set: for instance
.code (do set x)
produces
an anonymous function which, if called with one argument, stores that argument
into
.codn x .
The argument forms are arbitrary expressions, within which a special
convention is permitted:
.RS
.meIP >> @ num
A number preceded by a
.code @
is a metanumber. This is a special syntax
which denotes an argument. For instance
.code @2
means that the second argument of
the anonymous function is to be substituted in place of the
.codn @2 .
.code op
generates a function which has a number of required arguments equal to the
highest value of
.meta num
apearing in a
.cblk
.mati >> @ num
.cble
construct in the body. For instance
.code (op car @3)
generates a three-argument function (which passes its third
argument to
.codn car ,
returning the result, and ignores its first two arguments).
There is no way to use
.code op
to generate functions which have optional arguments.
.meIP < @rest
If the meta-symbol
.meta @rest
appears in the
.code op
syntax, it explicitly denotes the list of trailing arguments,
allowing them to be placed anywhere in the expression.
.RE
.IP
Functions generated by
.code op
are always variadic; they always take additional arguments after
any required ones, whether or not the
.meta @rest
syntax is used.
If the body does not contain
any
.meta @num
or
.meta @rest
syntax, then
.code @rest
is implicitly inserted. What this means is that, for example, since
the form
.code (op foo)
does not contain any numeric positional arguments like
.codn @1 ,
and does not contain
.codn @rest ,
it is actually a shorthand for
.codn (op foo . @rest) :
a function which applies all of its arguments to
.codn foo .
The actions of
.code op
be understood by these examples, which show
how
.code op
is rewritten to lambda. However, note that the real translator
uses generated symbols for the arguments, which are not equal to any
symbols in the program.
.cblk
(op) -> invalid
(op +) -> (lambda rest [+ . rest])
(op + foo) -> (lambda rest [+ foo . rest])
(op @1 @2) -> (lambda (arg1 arg2 . rest) [arg1 arg2])
(op @1 . @rest) -> (lambda (arg1 . rest) [arg1 . @rest])
(op @1 @rest) -> (lambda (arg1 . rest) [arg1 @rest])
(op @1 @2) -> (lambda (arg1 arg2 . rest) [arg1 arg2])
(op foo @1 (@2) (bar @3)) -> (lambda (arg1 arg2 arg3 . rest)
[foo arg1 (arg2) (bar arg3)])
(op foo @rest @1) -> (lambda (arg1 . rest) [foo rest arg1])
(do + foo) -> (lambda rest (+ foo . rest))
(do @1 @2) -> (lambda (arg1 arg2 . rest) (arg1 arg2))
(do foo @rest @1) -> (lambda (arg1 . rest) (foo rest arg1))
.cble
Note that if argument
.meta @n
appears, it is not necessary
for arguments
.meta @1
through
.meta @n-1
to appear. The function will have
.code n
arguments:
.cblk
(op @3) -> (lambda (arg1 arg2 arg3 . rest) [arg3])
.cble
The
.code op
and
.code do
operators can be nested, in any combination. This raises the
question: if a metanumber like
.code @1
or
.code @rest
occurs in an
.code op
that is nested
within an
.codn op , what is the meaning?
A metanumber always belongs with the inner-most op or do operator. So for
instance
.code (op (op @1))
means that an
.code (op @1)
expression is nested
within an
.code op
expression which itself contains no meta-syntax.
The
.code @1
belongs with the inner op.
There is a way for an inner
.code op
to refer to an outer op metanumber argument. This is
expresed by adding an extra
.code @
prefix for every level of escape. For example in
.code (op (op @@1))
the
.code @@1
belongs to the outer
.codn op :
it is the same as
.code @1
appearing in the outer
.codn op .
That is to say,
in the expression
.codn (op @1 (op @@1)) ,
the
.code @1
and
.code @@1
are the same thing:
both are parameter 1 of the lambda function generated by the outer
.codn op .
By contrast, in the expression
.code (op @1 (op @1))
there are two different parameters:
the first
.code @1
is argument of the outer function, and the second
.code @1
is the first argument of the inner function. Of course, if there
are three levels of nesting, then three
.code @
meta-prefixes are needed to insert
a parameter from the outermost
.code op
into the innermost
.codn op .
.TP* Examples:
.cblk
;; Take a list of pairs and produce a list in which those pairs
;; are reversed.
(mapcar (op list @2 @1) '((1 2) (a b))) -> ((2 1) (b a))
.cble
.coNP Macros @, ap @, ip @ ado and @ ido.
.synb
.mets (ap << form +)
.mets (ip << form +)
.mets (ado << form +)
.mets (ido << form +)
.syne
.desc
The
.code ap
macro is based on the
.code op
macro and has identical argument
conventions.
The
.code ap
macro analyzes its arguments and produces a function
.metn f ,
in exactly the same same way as the
.code op
macro. However, instead of returning
.metn f ,
directly, it returns a different function
.metn g ,
which is a one-argument function which accepts a list,
and then applies the list as arguments to
.metn f .
In other words, the following equivalence holds:
.cblk
(ap form ...) <--> (apf (op form ...))
.cble
The
.code ap
macro nests properly with
.code op
and
.codn do ,
in any combination, in regard to the
.meta ...@@n
notation.
The
.ode ip
macro is very similar to the
.code ap
macro, except that it is based
on the semantics of the function
.code iapply
rather than
.codn apply ,
according
to the following equivalence:
.cblk
(ip form ...) <--> (ipf (op form ...))
.cble
The
.code ado
and
.code ido
macros are related to do macro in the same way that
.code ap
and
.code ip
are related to
.codn op .
They produce a one-argument function which works
as if by applying its arguments to the function generated by do,
according to the following equivalence:
.cblk
(ado form ...) <--> (apf (do form ...))
(ido form ...) <--> (ipf (do form ...))
.cblk
See also: the
.code apf
and
.code ipf
functions.
.coNP Macros @ opip and @ oand
.synb
.mets (opip << clause *)
.mets (oand << clause *)
.syne
.desc
The
.code opip
and
.code oand
operators make it possible to chain together functions which are expressed
using the
.code op
syntax. (See the
.code op
operator for more information).
Both macros perform the same transformation except that
.code opip
translates its arguments to a call to the
.code chain
function, whereas
.code oand
transaltes its arguments in the same way to a call to the
.code chand
function.
More precisely, these macros perform the following rewrites:
.cblk
(opip arg1 arg2 ... argn) -> [chain {arg1} {arg2} ... {argn}]
(oand arg1 arg2 ... argn) -> [chand {arg1} {arg2} ... {argn}]
.cble
where the above
.code {arg}
notation denotes the following transformation applied to each argument:
.cblk
(function ...) -> (op function ...)
(operator ...) -> (do operator ...)
(macro ...) -> (do macro ...)
(dwim ...) -> (dwim ...)
[...] -> [...]
atom -> atom
.cble
In other words, compound forms whose leftmost symbol is a macro or operator
are translated to the
.code do
notation. Compound forms denoting function calls are translated to the
.code op
notation. Compound forms which are
.code dwim
invocations, either explicit or via the DWIM brackets notation, are
preserved, as are any forms which are atoms.
Note: the
.code opip
and
.code oand
macros use their macro environment in determining whether a form is a
macro call, thereby respecting lexical scoping.
.TP* Example:
Take each element from the list
.code (1 2 3 4)
and multiply it by three, then add 1.
If the result is odd, collect that into the resulting list:
.cblk
(mappend (opip (* 3)
(+ 1)
[iff oddp list])
(range 1 4))
.cble
The above is equivalent to:
.cblk
(mappend (chain (op * 3)
(op + 1)
[iff oddp list])
(range 1 4))
.cble
The
.code (* 3)
and
.code (+ 1)
terms are rewriten to
.code (op * 3)
and
.codn (op + 1) ,
respectively, whereas
.code [iff oddp list]
is passed through untransformed.
.coNP Macro @ ret
.synb
.mets (ret << form )
.syne
.desc
The
.code ret
macro's
.meta form
argument is treated similarly to the second and subsequent arguments of the
.code op
operator.
The
.code ret
macro produces a function which takes any number of arguments,
and returns the value specified by
.metn form .
.meta form
can contain
.code op
meta syntax like
.meta @n
and
.codn @rest .
The following equivalence holds:
.cblk
(ret x) <--> (op identity x))
.cble
Thus the expression
.code (ret @2)
returns a function similar to
.codn (lambda (x y . z) y) ,
and the expression
.code (ret 42)
returns a function similar to
.codn (lambda (. rest) 42) .
.coNP Macro @ aret
.synb
.mets (aret << form )
.syne
.desc
The
.code ret
macro's
.meta form
argument is treated similarly to the second and subsequent arguments of the
.code op
operator.
The
.code aret
macro produces a function which takes any number of arguments,
and returns the value specified by
.metn form .
.meta form
can contain
.code ap
meta syntax like
.meta @n
and
.codn @rest .
The following equivalence holds:
.cblk
(aret x) <--> (ap identity x))
.cble
Thus the expression
.code (aret @2)
returns a function similar to
.codn (lambda (. rest) (second rest)) ,
and the expression
.code (aret 42)
returns a function similar to
.codn (lambda (. rest) 42) .
.coNP Function @ dup
.synb
.mets (dup << func )
.syne
.desc
The
.code dup
function returns a one-argument function which calls the two-argument
function
.metn func
by duplicating its argument.
.TP* Example:
.cblk
;; square the elements of a list
(mapcar [dup *] '(1 2 3)) -> (1 4 9)
.cble
.coNP Function @ flip
.synb
.mets (flip << func )
.syne
.desc
The
.code flip
function returns a two-argument function which calls the two-argument
function
.metn func
with reversed arguments.
.coNP Functions @ chain and @ chand
.synb
.mets (chain << func *)
.mets (chand << func *)
.syne
.desc
The
.code chain
function accepts zero or more functions as arguments, and returns
a single function, called the chained function, which represents the chained
application of those functions, in left to right order.
If
.code chain
is given no arguments, then it returns a variadic function which
ignores all of its arguments and returns
.codn nil .
Otherwise, the first function may accept any number of arguments. The second
and subsequent functions, if any, must accept one argument.
The chained function can be called with an argument list which is acceptable
to the first function. Those arguments are in fact passed to the first
function. The return value of that call is then passed to the second
function, and the return value of that call is passed to the third function
and so on. The final return value is returned to the caller.
The
.code chand
function is similar, except that it combines the functionality of
.code andf
into chaining. The difference between
.code chain
and
.code chand
is that
.code chand
immediately terminates and returns
.code nil
whenever any of the functions returns
.codn nil ,
without evaluating the remainder of the functions.
.TP* Example:
.cblk
(call [chain + (op * 2)] 3 4) -> 14
.cble
In this example, a two-element chain is formed from the
.code +
function
and the function produced by
.code (op * 2)
which is a one-argument
function that returns the value of its argument multiplied by two.
(See the definition of the
.code op
operator).
The chained function is invoked using the
.code call
function, with the arguments
.code 3
and
.codn 4 .
The chained evaluation begins by passing
.code 3
and
.code 4
to
.codn + ,
which yields
.codn 7 .
This
.code 7
is then passed to the
.code (op * 2)
doubling function, resulting in
.codn 14 .
A way to write the above example without the use of the DWIM brackets and the
op operator is this:
.cblk
(call (chain (fun +) (lambda (x) (* 2 x))) 3 4)
.cble
.coNP Function @ juxt
.synb
.mets (juxt << func *)
.syne
.desc
The
.code juxt
function accepts a variable number of arguments which are functions. It
combines these into a single function which, when invoked, passes its arguments
to each of these functions, and collects the results into a list.
Note: the juxt function can be understood in terms of the following reference
implementation:
.cblk
(defun juxt (funcs)
(lambda (. args)
(mapcar (lambda (fun)
(apply fun args))
funcs)))
.cble
.TP* Example:
.cblk
;; separate list (1 2 3 4 5 6) into lists of evens and odds,
;; which end up juxtaposed in the output list:
[(op [juxt keep-if remove-if] evenp)
'(1 2 3 4 5 6)] -> ((2 4 6) (1 3 5))
;; call several functions on 1, collecting their results:
[[juxt (op + 1) (op - 1) evenp sin cos] 1]'
-> (2 0 nil 0.841470984807897 0.54030230586814)
.cble
.coNP Functions @ andf and @ orf
.synb
.mets (andf << func *)
.mets (orf << func *)
.syne
.desc
The
.code andf
and
.code orf
functions are the functional equivalent of the
.code and
and
.code or
operators. These functions accept multiple functions and return a new function
which represents the logical combination of those functions.
The input functions should have the same arity. Failing that, there should
exist some common argument arity with which each of these can be invoked. The
resulting combined function is then callable with that many arguments.
The
.code andf
function returns a function which combines the input functions with
a short-circuiting logical conjunction. The resulting function passes its
arguments to the functions successively, in left to right order. As soon as any
of the functions returns
.codn nil ,
then nil is returned immediately, and the
remaining functions are not called. Otherwise, if none of the functions return
.codn nil ,
then the value returned by the last function is returned. If the list of
functions is empty, then
.code t
is returned. That is,
.code (andf)
returns a function
which accepts any arguments, and returns
.codn t .
The
.code orf
function combines the input functions with a short-circuiting logical
disjunction. The function produced by
.code orf
passes its arguments down to the
functions successively, in left to right order. As soon as any function
returns a
.cod2 non- nil
value, that value is returned and the remaining functions are
not called. If all functions return
.codn nil ,
then
.code nil
is returned. The expression
.code (orf)
returns a function which accepts any arguments and returns
.codn nil .
.coNP Function @ notf
.synb
.mets (notf << function )
.syne
.desc
The
.code notf
function returns a function which is the boolean negation
of
.metn function .
The returned function takes a variable number of arguments. When
invoked, it passes all of these arguments to
.meta function
and then inverts the result as if by application of the
.codn not .
.coNP Functions @ iff and @ iffi
.synb
.mets (iff < cond-func >> [ then-func <> [ else-func ]])
.mets (iffi < cond-func < then-func <> [ else-func ])
.syne
.desc
The
.code iff
function is the functional equivalent of the
.code if
operator. It accepts
functional arguments and returns a function.
The resulting function takes its arguments, if any, and applies them to
.metn cond-func .
If
.meta cond-func
yields true, then the arguments are passed to
.meta then-func
and the
resulting value is returned. Otherwise the arguments are passed to
.meta else-func
and the resulting value is returned.
If
.meta then-func
is omitted then
.code identity
is used as default. This omission is not permitted by
.codn iffi ,
only
.codn iff .
If
.meta else-func
needs to be called, but is omitted, then
.code nil
is returned.
The
.code iffi
function differs from
.code iff
only in the defaulting behavior with respect
to the
.meta else-func
argument. If
.meta else-func
is omitted in a call to
.code iffi
then the deafult function is
.codn identity .
This is useful in situations when one value is to be
replaced with another one when the condition is true, otherwise
preserved.
The following equivalences hold between
.code iffi
and
.codn iff :
.cblk
(iffi a b c) <--> (iff a b c)
(iffi a b) <--> (iff a b identity)
(iffi a b false) <--> (iff a b)
(iffi a identity false) <--> (iff a)
.cble
The following equivalence illustrates
.code iff
with both optional arguments omitted:
.cblk
(iff a) <---> (iff a identity false)
.cble
.coNP Functions @ tf and @ nilf
.synb
.mets (tf << arg *)
.mets (nilf << arg *)
.syne
.desc
The
.code tf
and
.code nilf
functions take zero or more arguments, and ignore them.
The
.code tf
function returns
.codn t ,
and the
.code nilf
function returns
.codn nil .
Note: the following equivalences hold between these functions and the
.code ret
operator, and
.code retf
function.
.cblk
(fun tf) <--> (ret t) <--> (retf t)
(fun nilf) <--> (ret nil) <--> (ret) <--> (retf nil)
.cble
In Lisp-1-style code,
.code tf
and
.code nilf
behave like constants which can replace uses of
.code (ret t)
and
.codn (ret nil) :
.cblk
[mapcar (ret nil) list] <--> [mapcar nilf list]
.cble
.TP* Example:
.cblk
;; tf and nilf are useful when functions are chained together.
;; test whether (trunc n 2) is odd.
(defun trunc-n-2-odd (n)
[[chain (op trunc @1 2) [iff oddp tf nilf]] n])
.cble
In this example, two functions are chained together, and
.code n
is passed
through the chain such that it is first divided by two via the
function denoted by
.code (op trunc @1 2)
and then the result is passed into the
function denoted by
.codn [iff oddp tf nilf] .
The
.code iff
function passes its argument into
.codn oddp ,
and if
.code oddp
yields true, it passes the same argument to
.codn tf .
Here
.code tf
proves its utility by ignoring that value and returning
.codn t .
If the argument (the divided value) passed into
.code iff
is even, then iff passes it into the
.code nilf
function, which ignores the value and returns
.codn nil .
.coNP Function retf
.synb
.mets (retf << value )
.syne
.desc
The
.code retf
function returns a function. That function can take zero or
more arguments. When called, it ignores its arguments and returns
.metn value .
See also: the
.code ret
macro.
.TP* Example:
.cblk
;; the function returned by (retf 42)
;; ignores 1 2 3 and returns 42.
(call (retf 42) 1 2 3) -> 42
.cble
.coNP Functions @ apf and @ ipf
.synb
.mets (apf << function )
.mets (ipf << function )
.syne
.desc
The
.code apf
function returns a one-argument function which accepts
a list. When the function is called, it treats the list as
arguments which are applied to
.meta function
as if by apply. It returns whatever
.meta function
returns.
The
.code ipf
function is similar to
.codn apf ,
except that the returned
function applies arguments as if by
.code iapply
rather than
.codn apply .
See also: the
.code ap
macro.
.TP* Example:
.cblk
;; Function returned by [apf +] accepts the
;; (1 2 3) list and applies it to +, as
;; if (+ 1 2 3) were called.
(call [apf +] '(1 2 3)) -> 6
.cble
.coNP Function @ callf
.synb
.mets (callf < main-function << arg-function *)
.syne
.desc
The
.code callf
function returns a function which applies its arguments to each
.metn arg-function ,
juxtaposing the return values of these calls to form arguments
which are then passed to
.metn main-function .
The return value of
.meta main-function
is returned.
The following equivalence holds, except for the order of evaluation of
arguments:
.cblk
(callf fm f0 f1 f2 ...) <--> (chain (juxt f0 f1 f2 ...) (apf fm))
.cble
.TP* Example:
.cblk
;; Keep those pairs which are two of a kind
(keep-if [callf eql first second] '((1 1) (2 3) (4 4) (5 6)))
-> ((1 1) (4 4))
.cble
.coNP Function @ mapf
.synb
.mets (mapf < main-function << arg-function *)
.syne
.desc
The
.code mapf
function returns a function which distributes its arguments
into the
.metn arg-function -s.
That is to say, each successive argument of the returned
function is associated with a successive
.metn arg-function .
Each
.metn arg-function
is called, passed the corresponding argument. The return
values of these functions are then passd as arguments
to
.meta main function
and the resulting value is returned.
If the returned function is calle with fewer arguments than there
are
.metn arg-function -s,
then only that many functions are used. Conversely, if the function is
called with more arguments than there are
.metn arg-function -s, then those arguments are ignored.
The following equivalence holds:
.cblk
(mapf fm f0 f1 ...) <--> (lambda (. rest)
[apply fm [mapcar call (list f0 f1 ...) rest]])
.cble
.TP* Example:
.cblk
;; Keep those pairs which are two of a kind
(keep-if [callf eql first second] '((1 1) (2 3) (4 4) (5 6)))
-> ((1 1) (4 4))
.cble
.SS* Input and Output (Streams)
\*(TL supports input and output streams of various kinds, with
generic operations that work across the stream types.
In general, I/O errors are usually turned into exceptions. When the description
of error reporting is omitted from the description of a function, it can be
assumed that it throws an error.
.coNP Special variables @, *stdout* @, *stddebug* @, *stdin* @ *stderr* and @ *stdnull*
.desc
These variables hold predefined stream objects. The
.codn *stdin* ,
.code *stdout*
and
.code *stderr*
streams closely correspond to the underlying operating system streams.
Various I/O functions require stream objects as arguments.
The
.code *stddebug*
stream goes to the same destination as
.codn *stdout* ,
but is a separate object which can be redirected independently, allowing
debugging output to be separated from normal output.
The
.code *stdnull*
stream is a special kind of stream called a null stream.
This stream is not connected to any device or file. It is similar to
the
.code /dev/null
device on Unix, but does not involve the operating system.
.coNP Function @ format
.synb
.mets (format < stream-designator < format-string << format-arg *)
.syne
.desc
The
.code format
function performs output to a stream given by
.metn stream-designator ,
by interpreting the actions implicit in a
.metn format-string ,
incorporating material pulled from additional arguments given by
.cblk
.meti << format-arg *.
.cble
Though the function is simple to invoke, there is complexity in format string
language, which is documented below.
The
.meta stream-designator
argument can be a stream object, or one of the values
.code t
or
.codn nil .
The value
.code t
serves as a shorthand for
.codn *stdout* .
The value
.code nil
means that the function will send output into a newly instantiated string
output stream, and then return the resulting string.
.TP* "Format string syntax:"
Within
.metn format-string ,
most characters represent themselves. Those
characters are simply output. The character
.code ~
(tilde) introduces formatting
directives, which are denoted by a single character, usually a letter.
The special sequence
.code ~~
(tilde-tilde) encodes a single tilde. Nothing is
permitted between the two tildes.
The syntax of a directive is generally as follows:
.cblk
.mets <> ~[ width ] <> [, precision ] < letter
.cble
In other words, the
.code ~
(tilde) character, followed by a
.meta width
specifier, a
.meta precision
specifier introduced by a comma,
and a
.metn letter ,
such that
.meta width
and
.meta precision
are independently optional: either or both may be omitted.
No whitespace is allowed between these elements.
The
.meta letter
is a single alphabetic character which determines the
general action of the directive. The optional width and precision
are specified as follows:
.RS
.meIP < width
The width specifier consists of an optional
.code <
(left angle bracket) character,
followed by an optional width specification.
If the leading
.code <
character is present, then the printing will be left-adjusted within
this field. Otherwise it will be right-adjusted by default.
The width can be specified as a decimal integer, or as the character
.codn * .
The
.code *
notation means that instead of digits, the value of the next argument is
consumed, and expected to be an integer which specifies the width. If that
integer value is negative, then the field will be left-adjusted.
If the value is positive, but the
.code <
character is present in the width
specifier, then the field is left adjusted.
.meIP < precision
The precision specifier is introduced by a leading comma. If this comma appaers
immediately after the directive's
.code ~
character, then it means that
.meta width
is being omitted; there is only a precision field.
The precision specifier may begin with these optional characters:
.RS
.coIP 0
(the "leading zero flag"),
.coIP +
(print a sign for positive values")
.IP space
(print a space in place of a positive sign).
.RE
The precision specifier itself is either a decimal integer that does not
begin with a zero digit, or the
.code *
character.
The precision field's components have a meaning which depends on the type of
object printed and the conversion specifier.
For integer arguments, the precision value specifies the minimum number of digits
to print. If the precision field has a leading zero flag, then the integer is
padded with zeros to the required number of digits, otherwise the number is
padded with spaces instead of zeros. If zero or space padding is present, and
a leading positive or negative sign must be printed, then it is placed before
leading zeros, or after leading spaces, as the case may be.
For floating-point values, the meaning of the precision value depends on which
specific conversion specifier
.cod1 ( f ,
.codn e ,
.code a
or
.codn s )
is used. The details are
documented in the description of each of these, below. The leading zero flag is
ignored for floating-point values regardless of the conversion specifier.
For integer or floating-point arguments, if the precision specifier has a
.code +
sign
among the special characters, then a
.code +
sign is printed for positive numbers. If
the precision specifier has a leading space instead of a
.code +
sign, then the
.ocde +
sign is rendered as a space for positive numbers. If there is no leading space
or
.codn + ,
then a sign character is omitted for positive numbers. Negative
numbers are unconditionally prefixed with a
.code -
sign.
For all other objects, the precision specifies the maximum number of characters
to print. The object's printed representation is crudely truncated at that
number of characters.
.RE
.TP* "Format directives:"
.RS
Format directives are case sensitive, so that for example
.code ~x
and
.code ~X
have a
different effect, and
.code ~A
doesn't exist whereas
.code ~a
does. They are:
.coIP a
Prints any object in an aesthetic way, as if by the
.code pprint
function.
The aesthetic notation violates read-print consistency: this notation
is not necessarily readable if it is implanted in \*(TX source code.
The field width specifier is honored, including the left-right adjustment
semantics.
When this specifier is used for floating-point values, the precision specifies
the maximum number of total significant figures, which do not include any
digits in the exponent, if one is printed. Numbers are printed in exponential
notation if their magnitude is small, or else if their exponent exceeds their
precision. (If the precision is not specified, then it defaults to the
system-dependent number of digits in a floating point value, derived from the C
language
.code DBL_DIG
constant.) Floating point values which are integers are
printed without a trailing
.code .0
(point zero).
.coIP s
Prints any object in a standard way, as if by the print function. Objects for
which read-print consistency is possible are printed in a way such that
if their notation is implanted in \*(TX source, they are readable.
The field width specifier is honored, including the left-right adjustment
semantics. The precision field is treated very similarly to the
.code ~a
format directive, except that non-exponentiated floating point numbers that
would be mistaken for integers include a trailing
.code .0
for the sake of read-print
consistency. Objects truncated by precision may not have read-print
consistency. For instance, if a string object is truncated, it loses its
trailing closing quote, so that the resulting representation is no longer
a properly formed string object.
.coIP x
Requires an argument of character or integer type. The integer value or
character code is printed in hexadecimal, using lower-case letters
for the digits
.code a
through
.codn f .
Width and precision semantics
are as described for the
.code a
format directive, for integers.
.coIP X
Like the
.code x
directive, but the hexadecimal digits
.code a
through
.code f
are rendered in upper case.
.coIP o
Like the
.code x
directive, but octal is used instead of hexadecimal.
.coIP f
The
.code f
directive prints numbers in a fixed point decimal notation, with
a fixed number of digits after the decimal point. It requires a numeric
argument. (Unlike
.codn x ,
.code X
and
.codn o ,
it does not allow an argument of character type).
The precision specifier gives the number of digits past the decimal point.
The number is rounded off to the specified precision, if necessary.
Furthermore, that many digits are always printed, regardless of the actual
precision of the number or its type. If it is omitted, then the default value
is three: three digits past the decimal point. A precision of zero means no
digits pas the decimal point, and in this case the decimal point is suppressed
(regardless of whether the numeric argument is floating-point or integer).
.coIP e
The
.code e
directive prints numbers in exponential notation. It requires
a numeric argument. (Unlike
.codn x ,
.code X
and
.codn o ,
it does not allow an argument of character type).
The precision specifier gives the number of digits past the decimal point
printed in the exponential notation, not counting the digits in the exponent.
Exactly that many digits are printed, regardless of the precision of the
number. If the precision is omitted, then the number of digits after the
decimal point is three. If the precision is zero, then a decimal portion is
truncated off entirely, including the decimal point.
.coIP p
The
.code p
directive prints a numeric representation in hexadecimal of the bit pattern
of the object, which is meaningful to someone familiar with the internals
of \*(TX. If the object is a pointer to heaped data, that value
has a correspondence to its address.
.RE
.PP
.coNP Functions @, print @, pprint @, prinl @, pprinl @ tostring and @ tostringp
.synb
.mets (print < obj <> [ stream ])
.mets (pprint < obj <> [ stream ])
.mets (prinl < obj <> [ stream ])
.mets (pprinl < obj <> [ stream ])
.mets (tostring << obj )
.mets (tostringp << obj )
.syne
.desc
The
.code print
and
.code pprint
functions render a printed character representation of the
.meta obj
argument into
.metn stream .
If a stream argument is not supplied, then
the destination is the stream currently stored in the
.code *stdout*
variable. The
.code print
function renders in a way which strives for read-print
consistency: an object is printed in a notation which is recognized as
a similar object of the same kind when it appears in \*(TX source code.
The
.code pprint
function ("pretty print") does not strive for read-print consistency.
For instance it prints a string object simply by dumping its characters, rather
than by adding the surrounding quotes and rendering escape syntax for
special characters. Both functions return
.metn obj .
The
.code prinl
and
.code pprinl
functions are like
.code print
and
.codn pprint ,
except that they issue a newline character after printing the object.
These functions also return
.metn obj .
The
.code tostring
and
.code tostringp
functions are like
.code print
and
.codn pprint ,
but they do not accept a stream argument. Instead they print to a freshly
instantiated string stream, and return the resulting string.
The following equivalences hold between calls to the
.code format
function and calls to the above functions:
.cblk
(format stream "~s" obj) <--> (print obj stream)
(format t "~s" obj) <--> (print obj)
(format t "~s\en" obj) <--> (prinl obj)
(format nil "~s" obj) <--> (tostring obj)
.cble
For
.codn pprint ,
.code tostringp
and
.codn pprinl ,
the equivalence is produced by using
.code ~a
in format rather than
.codn ~s .
Note: for characters, the print function behaves as follows: most control
characters in the Unicode
.code C0
and
.code C1
range are rendered using the
.code #\ex
notation,
using two hex digits. Codes in the range
.code D800
to
.codn DFFF ,
and the codes
.code FFFE
and
.code FFFF
are printed in the
.code #\exNNNN
with four hexadecimal digits, and
character above this range are printed using the same notation, but with six
hexadecimal digits. Certain characters in the
.code C0
range are printed using
their names such as
.code #\enul
and
.codn #\ereturn ,
which are documented
in the Character Literals section.
The
.code DC00
character is printed as
.codn #\epnul .
All other characters are printed as
.cblk
.meti >> #\e char
.cble
where
.meta char
is the actual character.
Caution: read-print consistency is affected by trailing material. If additional
digits are printed immediately after a number without intervening whitespace,
they extend that number. If hex digits are printed after the character
.codn x ,
which is rendered as
.codn #\ex ,
they look like a hex character code.
.coNP Function @ tprint
.synb
.mets (tprint < obj <> [ stream ])
.syne
.desc
The
.codn tprint
function prints a represntation of
.meta obj
on
.metn stream .
If the stream argument is not supplied, then
the destination is the stream currently stored in the
.code *stdout*
variable.
For all object types except lists and vectors,
.code tprint
behaves like
.codn pprinl .
If
.code obj
is a list or vector, then
.code tprint
recurses: the
.code tprint
function is applied to each element. An empty list or vector
results in no output at all. This effectively means that an arbitrarily nested
structure of lists and vectors is printed flattened, with one element on each
line.
.coNP Function @ streamp
.synb
.mets (streamp << obj )
.syne
.desc
The
.code streamp
function returns
.code t
if
.meta obj
is any type of stream. Otherwise it returns
.codn nil .
.coNP Function @ real-time-stream-p
.synb
.mets (real-time-stream-p << obj )
.syne
.desc
The
.code real-time-streamp-p
function returns
.code t
if
.meta obj
is a stream marked as
"real-time". If
.meta obj
is not a stream, or not a stream marked as "real-time",
then it returns
.codn nil .
Only certain kinds of streams accept the real-time attribute: file streams and
tail streams. This attribute controls the semantics of the application of
.code lazy-stream-cons
to the stream. For a real-time stream,
.code lazy-stream-cons
returns a stream with "naive" semantics which returns data as soon as it is
available, at the cost of generating spurious
.code nil
item when the stream
terminates. The application has to recognize and discard that
.code nil
item.
The ordinary lazy streams read ahead by one line and suppress this extra
item, so their representation is more accurate.
When \*(TX starts up, it automatically marks the
.code *std-input*
stream as real-time, if it is connected to a TTY device (a device for which
the POSIX function
.code isatty
reports true). This is only supported on platforms that have this function.
The behvior is overridden by the
.code -n
command line option.
.coNP Function @ make-string-input-stream
.synb
.mets (make-string-input-stream << string )
.syne
.desc
This function produces an input stream object. Character read operations on the
stream object read successive characters from
.metn string .
Output operations and byte operations are not supported.
.coNP Function @ make-string-byte-input-stream
.synb
.mets (make-string-byte-input-stream << string )
.syne
.desc
This function produces an input stream object. Byte read operations on
this stream object read successive byte values obtained by encoding
.meta string
into UTF-8. Character read operations are not supported, and neither
are output operations.
.coNP Function @ make-string-output-stream
.synb
(make-string-output-stream)
.syne
.desc
This function, which takes no arguments, creates a string output stream.
Data sent to this stream is accumulated into a string object.
String output streams supports both character and byte output operations.
Bytes are assumed to represent a UTF-8 encoding, and are decoded in order
to form characters which are stored into the string.
If an incomplete UTF-8 code is output, and a character output operation then
takes place, that code is assumed to be terminated and is decoded as invalid
bytes. The UTF-8 decoding machine is reset and ready for the start of a new
code.
The
.code get-string-from-stream
function is used to retrieve the accumulated string.
If the null character is written to a string output stream, either via
a character output operation or as a byte operation, the resulting string
will appear to be prematurely terminated. \*(TX strings cannot contain null
bytes.
.coNP Function @ get-string-from-stream
.synb
.mets (get-string-from-stream << stream )
.syne
.desc
The
.meta stream
argument must be a string output stream. This function finalizes
the data sent to the stream and retrieves the accumulated character string.
If a partial UTF-8 code has been written to
.metn stream ,
and then this
function is called, the byte stream is considered complete and the partial
code is decoded as invalid bytes.
After this function is called, further output on the stream is not possible.
.coNP Function @ make-strlist-output-stream
.synb
(make-strlist-output-stream)
.syne
.desc
This function is very similar to
.codn make-string-output-stream .
However, the stream object produced by this function does not produce a string,
but a list of strings. The data is broken into multiple strings by newline
characters written to the stream. Newline characters do not appear in the
string list. Also, byte output operations are not supported.
.coNP Function @ get-list-from-stream
.synb
.mets (get-list-from-stream << stream )
.syne
.desc
This function returns the string list which has accumulated inside
a string output stream given by
.metn stream .
The string output stream is
finalized, so that further output is no longer possible.
.coNP Function @ close-stream
.synb
.mets (close-stream < stream <> [ throw-on-error-p ])
.syne
.desc
The
.code close-stream
function performs a close operation on
.metn stream ,
whose meaning is depends on the type of the stream. For some types of streams,
such as string streams, it does nothing. For streams which are connected
to operating system files or devices, will perform a close of the underlying
file descriptor, and dissociate that descriptor from the stream. Any buffered
data is flushed first.
.code close-stream
returns a boolean true value if the close has occurred without
errors, otherwise
.codn nil .
For most streams, "without errors" means that any buffered output data is
flushed successfully.
For command and process pipes (see open-command and open-process), success also
means that the process terminates normally, with a successful error code, or an
unsuccessful one. An abnormal termination is considered an error, as
as is the inability to retrieve the termination status, as well as the situation
that the process continues running in spite of the close attempt.
Detecting these situations is platform specific.
If the
.meta throw-on-error-p
argument is specified, and isn't
.codn nil ,
then the
function throws an exception if an error occurs during the close operation
instead of returning
.codn nil .
.coNP Functions @, get-error @ get-error-str and @ clear-error
.synb
.mets (get-error < stream )
.mets (get-error-str < stream )
.mets (clear-error < stream )
.syne
.desc
When a stream operation fails, the
.code get-error
and
.code get-error-str
functions may be used to inquire about a more detailed cause of the error.
Not all streams support these functions to the same extent. For instance,
string input streams have no persistent state. The only error which occurs
is the condition when the string has no more data.
The
.code get-error
inquires
.meta stream
about its error condition.
The function returns
.code nil
to indicate there is no error condition,
.code t
to indicate an end-of-data condition,
or else a value which is specific to the stream type indicating the
specific error type.
Note: for some streams, it is possible for the
.code t
value to be returned even though no operation has failed; that is to say, the
streams "know" they are at the end of the data even though no read operation
has failed. Code which depends on this will not work with streams which
do not thus indicate the end-of-data
.I a priori,
but by means of a read operation which fails.
The
.code get-error-str
function returns a text representation of the error code. The
.code nil
error code is represented as the string
.codn "no error" ;
the
.code t
error code as
.code "eof"
and other codes have a stream-specific representation.
The
.code clear-error
function removes the error situation from a stream. On some streams, it does
nothing. If an error has occurred on a stream, this function should be called
prior to re-trying any I/O or positioning operations.
.coNP Functions @, get-line @ get-char and @ get-byte
.synb
.mets (get-line <> [ stream ])
.mets (get-char <> [ stream ])
.mets (get-byte <> [ stream ])
.syne
.desc
These fundamental stream functions perform input. The
.meta stream
argument
is optional. If it is specified, it should be an input stream which supports
the given operation. If it is not specified, then the
.code *stdin*
stream is used.
The
.code get-char
function pulls a character from a stream which supports character
input. Streams which support character input also support the
.code get-line
function which extracts a line of text delimited by the end of the stream or a
newline character and returns it as a string. (The newline character does not
appear in the string which is returned).
Character input from streams based on bytes requires UTF-8 decoding, so that
get-char actually may read several bytes from the underlying low level
operating system stream.
The
.code get-byte
function bypasses UTF-8 decoding and reads raw bytes from
any stream which supports byte input. Bytes are represented as integer
values in the range 0 to 255.
Note that if a stream supports both byte input and character input, then mixing
the two operations will interfere with the UTF-8 decoding.
These functions return
.code nil
when the end of data is reached. Errors are
represented as exceptions.
See also:
.code get-lines
.coNP Function @ get-string
.synb
.mets (get-string >> [ stream >> [ count <> [ close-after-p ]]])
.syne
.desc
The
.code get-string
function reads characters from a stream, and assembles them into
a string, which is returned. If the
.meta stream
argument is omitted, then the
.code *stdin*
stream is used.
The stream is closed after extracting the data, unless
.meta close-after-p
is specified as
.codn nil .
The default value of this argument is
.codn t .
If the
.meta count
argument is missing, then all of the characters from the
stream are read and assembled into a string.
If present, the
.meta count
argument should be a positive integer indicating
a limit on how many characters to read. The returned string will be no
longer than
.metn count ,
but may be shorter.
.coNP Functions @ unget-char and @ unget-byte
.synb
.mets (unget-char < char <> [ stream ])
.mets (unget-byte < byte <> [ stream ])
.syne
.desc
These functions put back, into a stream, a character or byte which was
previously read. The character or byte must match the one which was most
recently read. If the
.meta stream
argument is omitted, then the
.code *stdin*
stream is used.
If the operation succeeds, the byte or character value is returned.
A
.code nil
return indicates that the operation is unsupported.
Some streams do not support these operations; some support
only one of them. In general, if a stream supports
.codn get-char ,
it supports
.codn unget-char ,
and likewise for
.code get-byte
and
.codn unget-byte .
Space is available for only one character or byte of pushback.
Pushing both a byte and a character, in either order, is also unsupported.
Pushing a byte and then reading a character, or pushing a character and
reading a byte, are unsupported mixtures of operations.
If the stream is binary, then pushing back a byte decrements its position,
except if the position is already zero. At that point, the position becomes
indeterminate.
.coNP Functions @, put-string @, put-line @ put-char and @ put-byte
.synb
.mets (put-string < string <> [ stream ])
.mets (put-line >> [ string <> [ stream ]])
.mets (put-char < char <> [ stream ])
.mets (put-byte < byte <> [ stream ])
.syne
.desc
These functions perform output on an output stream. The
.meta stream
argument
must be an output stream which supports the given operation. If it is omitted,
then
.code *stdout*
is used.
The
.code put-char
function writes a character given by
.code char
to a stream. If the
stream is based on bytes, then the character is encoded into UTF-8 and multiple
bytes are written. Streams which support
.code put-char
also support put-line, and
.codn put-string .
The
.code put-string
function writes the characters of a string out to
the stream as if by multiple calls to put-char. The
.meta string
argument
may be a symbol, in which case its name is used as the string.
The
.code put-line
function is like
.codn put-string ,
but also writes an additional newline
character. The string is optional in
.codn put-line ,
and defaults to the empty string.
The
.code put-byte
function writes a raw byte given by the
.meta byte
argument
to
.metn stream ,
if
.meta stream
supports a byte write operation. The byte
value is specified as an integer value in the range 0 to 255.
.coNP Functions @ put-strings and @ put-lines
.synb
.mets (put-strings < sequence <> [ stream ]])
.mets (put-lines < sequence <> [ stream ]])
.syne
.desc
These functions assume
.meta sequence
to be a sequence of strings, or of
symbols, or a mixture thereof. These strings are sent to the stream. The
.meta stream
argument must be an output stream. If it is omitted, then
.code *stdout*
is used.
The
.code put-strings
function iterates over
.meta sequence
and writes each element
to the stream as if using the
.code put-string
function.
The
.code put-lines
function iterates over
.code sequence
and writes each element
to the stream as if using the
.code put-line
function.
Both functions return
.codn t .
.coNP Function @ flush-stream
.synb
.mets (flush-stream << stream )
.syne
.desc
This function is meaningful for output streams which accumulate data
which is passed on to the operating system in larger transfer units.
Calling
.code flush-stream
causes all accumulated data inside
.meta stream
to be passed
to the operating system. If called on streams for which this function is not
meaningful, it does nothing.
.coNP Function @ seek-stream
.synb
.mets (seek-stream < stream < offset << whence )
.syne
.desc
The
.code seek-stream
function is meaningful for file streams. It changes the
current read/write position within
.metn stream .
It can also be used to determine the current position: see the notes about the
return value below.
The
.meta offset
argument is a positive or negative integer which gives a
displacement that is measured from the point identified by the
.meta whence
argument.
Note that for text files, there isn't necessarily a 1:1 correspondence between
characters and positions due to line-ending conversions and conversions
to and from UTF-8.
The
.meta whence
argument is one of three keywords: :from-start, :from-current
and :from-end. These denote the start of the file, the current position
and the end of the file.
If
.meta offset
is zero, and
.meta whence
is
.codn :from-current ,
then
.code seek-stream
returns the current absolute position within the
stream, if it can successfully obtain it. Otherwise, it
returns
.code t
if it is successful.
If a character has been successfully put back into a text stream with
.code unget-char
and is still pending, then the position value is unspecified. If a
byte has been put back into a binary stream with
.codn unget-byte ,
and the previous position wasn't zero, then the position is decremented by one.
On failure, it throws an exception of type
.codn stream-error .
.coNP Functions @ stream-get-prop and @ stream-set-prop
.synb
.mets (stream-get-prop < stream << indicator )
.mets (stream-set-prop < stream < indicator << value )
.syne
.desc
These functions get and set properties on a stream. Only certain properties
are meaningful with certain kinds of streams, and the meaning depends on
the stream. If two or more stream types support a property of the same name, it
is expected that the property has the same or very similar meaning for both
streams to the maximum extent that similarity is possible.
The
.code stream-set-prop
function sets a property on a stream. The
.meta indicator
argument is a symbol, usually a keyword symbol, denoting the property,
and
.meta value
is the property value. If the stream understands and accepts the
property, the function returns
.codn t .
Otherwise it returns
.codn nil .
The
.code stream-get-prop
function inquires about the value of a property on a
stream. If the stream understands the property, then it returns its current
value. If the stream does not understand a property, nil is returned, which is
also returned if the property exists, but its value happens to be
.codn nil .
Properties are currently used for marking certain streams as "real-time" (see
the
.code real-time-stream-p
function above), and also for setting the priority at
which messages are reported to syslog by the
.code *stdlog*
stream (see
.code *stdlog*
in the UNIX SYSLOG section).
If
.meta stream
is a catenated stream (see the function
.codn make-catenated-stream )
then these functions transparently operate on the current head stream of the
catenation.
.coNP Functions @ make-catenated-stream and @ cat-streams
.synb
.mets (make-catenated-stream << stream *)
.mets (cat-streams << stream-list )
.syne
.desc
The
.code make-catenated-stream
function takes zero or more arguments which
are input streams of the same type, and combines
them into a single virtual stream called a catenated stream.
The
.code cat-streams
function takes a single list of input streams of
the same type, and similarly combines them into a catenated stream.
A catenated stream does not support seeking operations or output,
regardless of the capabilities of the streams in the list.
If the stream list is not empty, then the leftmost element of the
list is called the head stream.
The
.codn get-char ,
.codn get-byte ,
.codn get-line ,
.code unget-char
and
.code unget-byte
functions delegate
to the corresponding operations on the head stream, if it exists.
If the stream list is empty, they return
.code nil
to the caller.
If the
.codn get-char ,
.code get-byte
or
.code get-line
operation on the head stream yields
.codn nil ,
and there are more lists in the stream, then the stream is closed, removed from
the list, and the next stream, if any, becomes the head list. The operation is
then tried again. If any of these operations fail on the last list, it is not
removed from the list, so that a stream remains in place which can take the
.code unget-char
or
.code unget-byte
operations.
In this manner, the catenated streams appear to be a single stream.
Note that the operations can fail due to being unsupported. It is
the caller's responsibility to make sure all of the streams in the list
are compatible with the intended operations.
If the stream list is empty then an empty catenated stream is produced.
Input operations on this stream yield
.codn nil ,
and the
.code unget-char
and
.code unget-byte
operations throw an exception.
.coNP Functions @ open-files and @ open-files*
.synb
.mets (open-files < path-list <> [ alternative-stream ])
.mets (open-files* < path-list <> [ alternative-stream ])
.syne
.desc
The
.code open-files
and
.code open-files*
functions create a list of streams by invoking
the open-file function on each element of
.metn path-list .
These streams are turned
into a catenated stream as if applied as arguments to
.codn make-catenated-stream .
The effect is that multiple files appear to be catenated together into a single
input stream.
If the optional
.meta alternative-stream
argument is supplied, then if
.meta path-list
is empty,
.meta alternative-stream
is returned instead of an empty catenated stream.
The difference between
.code open-files
and
.code open-files*
is that
.code open-files
creates all of the
streams up-front. So if any of the paths cannot be opened, the operation throws.
The
.code open-files*
variant is lazy: it creates a lazy list of streams out of the
path list. The streams are opened as needed: before the second stream is opened,
the program has to read the first stream to the end, and so on.
.TP* Example:
Collect lines from all files that are given as arguments on the command line. If
there are no files, then read from standard input:
.cblk
@(next @(open-files *args* *stdin*))
@(collect)
@line
@(end)
.cble
.coNP Function @ abs-path-p
.synb
.mets (abs-path-p << path )
.syne
.desc
The
.code abs-path-function
tests whether the argument
.meta path
is an absolute path, returning a
.code t
or
.code nil
indication.
An absolute path is a string which either begins with a slash or backslash
character, or which begins with an alphanumeric word, followed by a colon,
followed by a slash or backslash.
Examples of absolute paths:
.cblk
/etc
c:/tmp
ftp://user@server
disk0:/home
Z:\eUsers
.cble
Examples of strings which are not absolute paths.
.cblk
.mets < (the < empty < string)
.
abc
foo:bar/x
$:\eabc
.cble
.coNP Function @ read
.synb
.mets (read >> [ source >> [ error-stream <> [ error-return-value ]]])
.syne
.desc
The
.code read
function converts text denoting \*(TL structure, into the
corresponding data structure. The
.meta source
argument may be either a character
string, or a stream. If it is omitted, then
.code *stdin*
is used as the stream.
The source must provide the text representation of one complete \*(TL object.
Multiple calls to read on the same stream will extract successive objects
from the stream. To parse successive objects from a string, it is necessary
to convert it to a string stream.
The optional
.meta error-stream
argument can be used to specify a stream to which
parse errors diagnostics are sent. If absent, the diagnostics are suppressed.
If there are no parse errors, the function returns the parsed data
structure. If there are parse errors, and the
.meta error-return-value
parameter is
present, its value is returned. If the
.meta error-return-value
parameter
is not present, then an exception of type
.code syntax-error
is thrown.
.SS* Filesystem Access
.coNP Function @ open-directory
.synb
.mets (open-directory << path )
.syne
.desc
The
.code open-directory
function tries to create a stream which reads the
directory given by the string argument
.metn path .
If a filesystem object exists
under the path, is accessible, and is a directory, then the function
returns a stream. Otherwise, a file error exception is thrown.
The resulting stream supports the get-line operation. Each call to the
.code get-line
operation retrieves a string representing the next directory
entry. The value
.code nil
is returned when there are no more directory entries.
The
.code .
and
.code ..
entries in Unix filesystems are not skipped.
.coNP Function @ open-file
.synb
.mets (open-file < path <> [ mode-string ])
.syne
.desc
The
.code open-file
function creates a stream connected to the file
which is located at the given
.metn path ,
which is a string.
The
.meta mode-string
argument is a string which uses the same
conventions as the mode argument of the C language
.code fopen
function, with some extensions.
The mode string determines whether the stream is an input stream
or output stream. Note that the
.str b
mode is passed through to the C library, but has no special meaning to \*(TX.
Whether a stream is text or binary depends on which operations
are invoked on it.
If the
.meta mode-string
argument is omitted, mode
.str r
is used.
The option letter
.str i
is supported. If present, it will create a stream which has the real-time
property set.
.coNP Function @ open-tail
.synb
.mets (open-tail < path >> [ mode-string <> [ seek-to-end-p ]])
.syne
.desc
The
.code open-tail
function creates a tail stream connected to the file which is
located at the given
.metn path .
The
.meta mode-string
argument is a string which uses
the same conventions as the mode argument of the C language
.code fopen
function. If this argument is omitted, then
.str r
is used.
See the
.code open-file
function for a discussion of modes.
The
.code seek-to-end-p
argument is a boolean which determines whether the initial
read/write position is at the start of the file, or just past the end.
It defaults to
.codn nil .
This argument only makes a difference if the file exists
at the time
.code open-tail
is called. If the file does not exist, and is later
created, then the tail stream will follow that file from the beginning. In
other words,
.meta seek-to-end-p
controls whether the tail stream reads all the
existing data in the file, if any, or whether it reads only newly added data
from approximately the time the stream is created.
A tail stream has special semantics with regard to reading at the end
of file. A tail stream never reports an end-of-file condition; instead
it polls the file until more data is added. Furthermore, if the file
is truncated, or replaced with a smaller file, the tail stream follows
this change: it automatically opens the smaller file and starts reading from
the beginning (the
.meta seek-to-end-p
flag only applies to the initial open).
In this manner, a tail stream can dynamically growing rotating log files.
Caveat: since a tail stream can re-open a new file which has the same
name as the original file, it behave incorrectly if the program
changes the current working directory, and the path name is relative.
.coNP Function @ remove-path
.synb
.mets (remove-path << path )
.syne
.desc
The
.code remove-path
function tries to remove the filesystem object named
by
.metn path ,
which may be a file, directory or something else.
If successful, it returns
.codn t .
A failure to remove the object results in an exception of type
.codn file-error .
.coNP Function @ rename-path
.synb
.mets (rename-path < from-path << to-path )
.syne
.desc
The
.code remove-path
function tries to rename filesystem path
.metn from-path ,
which may refer to a file, directory or something else, to the path
.metn to-path .
If successful, it returns
.codn t .
A failure results in an exception of type
.codn file-error .
.SS* Coprocesses
.coNP Functions @ open-command and @ open-process
.synb
.mets (open-command < system-command <> [ mode-string ])
.mets (open-process < program < mode-string <> [ argument-list ])
.syne
.desc
These functions spawn external programs which execute concurrently
with the \*(TX program. Both functions return a unidirectional stream for
communicating with these programs: either an output stream, or an input
stream, depending on the contents of
.metn mode-string .
In
.codn open-command ,
the
.meta mode-string
argument is optional, defaulting to
the value
.str r
if it is missing. See the
.code open-file
function for a discussion of modes.
The
.code open-command
function accepts, via the
.meta system-command
string parameter, a
system command, which is in a system-dependent syntax. On a POSIX system, this
would be in the POSIX Shell Command Language.
The
.code open-process
function specifies a program to invoke via the
.meta command
argument. This is subject to the operating system's search strategy.
On POSIX systems, if it is an absolute or relative path, it is treated as
such, but if it is a simple base name, then it is subject to searching
via the components of the PATH environment variable. If open-process
is not able to find
.metn program ,
or is otherwise unable to execute
the program, the child process will exit, using the value of the C variable
.code errno
as its exit status. This value can be retrieved via
.codn close-stream .
The
.meta mode-string
argument follows the convention used by the POSIX
.code popen
function.
The
.meta argument-list
argument is a list of strings which specifies additional
optional arguments to be passed passed to the program. The
.meta program
argument
becomes the first argument, and
.meta argument-string
become the second and
subsequent arguments. If
.meta argument-strings
is omitted, it defaults to empty.
If a coprocess is open for writing
.cblk
.meti >> ( mode-string
.cble
is specified as
.strn w ),
then
writing on the returned stream feeds input to that program's standard input
file descriptor. Indicating the end of input is performed by closing the
stream.
If a coprocess is open for reading
.cblk
.meti >> ( mode-string
.cble
is specified as
.strn r ),
then
the program's output can be gathered by reading from the returned stream.
When the program finishes output, it will close the stream, which can be
detected as normal end of data.
If a coprocess terminates abnormally or unsuccessfully, an exception is raised.
.SS* Symbols and Packages
A package is an object which serves as a container of symbols.
A symbol which exists inside a package is said to be interned in that package.
A symbol can be interned in at most one package at a time.
Eacy symbol has a name, which is a string. It is not necessarily unique:
two distinct symbols can have the sama name. However, a symbol name is unique
within a package, because it serves as the key which associates the
symbol with the package. Two symbols cannot be in the same package if they
have the same name. Moreover, a symbol cannot exist in more than one package
at at time, although it can be relocated from one package to another. A
symbols exist which is not entered into any package: such a symbol is
called "uninterned".
Packages are held in a global list which can be used to search for a package by
name. The
.code find-package
function performs this lookup. A package may be
deleted from the list with the
.code delete-package
function, but it continues
to exist until the program loses the last reference to that package.
.coNP Special variables @, *user-package* @, *keyword-package* and @ *system-package*
.desc
These variables hold predefined packages. The
.code *user-package*
is the one
in which symbols are read when a \*(TX program is being scanned.
The
.code *keyword-package*
holds keyword symbols, which are printed with
a leading colon. The
.code *system-package*
is for internal symbols, helping
the implementation avoid name clashes with user code in some situations.
.coNP Function @ make-sym
.synb
.mets (make-sym << name )
.syne
.desc
The
.code make-sym
function creates and returns a new symbol object. The argument
.metn name ,
which must be a string, specifies the name of the symbol. The symbol
does not belong to any package (it is said to be "uninterned").
Note: an uninterned symbol can be interned into a package with the
.code rehome-sym
function. Also see the
.code intern
function.
.coNP Function @ gensym
.synb
.mets (gensym <> [ prefix ])
.syne
.desc
The
.code gensym
function is similar to make-sym. It creates and returns a new
symbol object. If the
.meta prefix
argument is omitted, it defaults to
.strn g .
Otherwise it must be a string.
The difference between
.code gensym
and
.code make-sym
is that
.code gensym
creates the name
by combining the prefix with a numeric suffix.
The numeric suffix is a decimal digit string, taken from the value of
the variable
.codn *gensym-counter* ,
after incrementing it.
Note: the variation in name is not the basis of the uniqueness assurance
offered by
.code make-sym
and
.codn gensym ;
the basis is that the returned symbol is a freshly instantiated object.
.code make-sym
still returns unique symbols even if repeatedly called with the same
string.
.coNP Special variable @ *gensym-counter*
.desc
This variable is initialized to 0. Each time the
.code gensym
function is called,
it is incremented. The incremented value forms the basis of the numeric
suffix which gensym uses to form the name of the new symbol.
.coNP Function @ make-package
.synb
.mets (make-package << name )
.syne
.desc
The
.code make-package
function creates and returns a package named
.metn name ,
where
.meta name
is a string. It is an error if a package by that name exists already.
.coNP Function @ packagep
.synb
.mets (packagep << obj )
.syne
.desc
The
.code packagep
function returns
.codet
if
.meta obj
is a package, otherwise it returns
.codn nil .
.coNP Function @ find-package
.synb
.mets (find-package << name )
.syne
.desc
The argument
.meta name
should be a string. If a package called
.meta name
exists,
then it is returned. Otherwise
.code nil
is returned.
.coNP Function @ intern
.synb
.mets (intern < name <> [ package ])
.syne
.desc
The argument
.meta name
should be a symbol. The optional argument
.meta package
should be a package. If
.meta package
is not supplied, then the value
taken is that of
.codn *user-package* .
The
.code intern
function searches
.meta package
for a symbol called
.metn name .
If that symbol is found, it is returned. If that symbol is not found,
then a new symbol called
.meta name
is created and inserted into
.metn package ,
and that symbol is returned. In this case, the package becomes the
symbol's home package.
.coNP Function @ rehome-sym
.synb
.mets (rehome-sym < symbol <> [ package ])
.syne
.desc
The arguments
.meta symbol
and
.meta package
must be a symbol and package object,
respectively. If
.meta package
is not given, then it defaults to the value of
.codn *user-package* .
The
.code rehome-sym
function moves
.meta symbol
into
.metn package .
If
.meta symbol
is already in a package, it is first removed from that package.
If a symbol of the same name exists in
.meta package
that symbol is first removed
from
.metn package .
.coNP Function @ symbolp
.synb
.mets (symbolp << obj )
.syne
.desc
The
.code symbolp
function returns
.code t
if
.meta obj
is a symbol, otherwise it returns
.codn nil .
.coNP Function @ symbol-name
.synb
.mets (symbol-name << symbol )
.syne
.desc
The
.code symbol-name
function returns the name of
.metn symbol .
.coNP Function @ symbol-package
.synb
.mets (symbol-package << symbol )
.syne
.desc
The
.code symbol-package
function returns the home package of
.metn symbol .
.coNP Function @ packagep
.synb
.mets (packagep << obj )
.syne
.desc
The
.code packagep
function returns
.code t
if
.meta obj
is a package, otherwise it returns
.codn nil .
.coNP Function @ keywordp
.synb
.mets (keywordp << obj )
.syne
.desc
The
.code keywordp
function returns
.code t
if
.meta obj
is a keyword symbol, otherwise it
returns
.codn nil .
.SS* Pseudo-random Numbers
.coNP Special variable @ *random-state*
.desc
The
.code *random-state*
variable holds an object which encapsulates the state
of a pseudo-random number generator. This variable is the default argument
value for the
.code random-fixnum
and
.codn random functions ,
for the convenience of writing programs which are not concerned about the
management of random state.
On the other hand, programs can create and manage random states, making it
possible to obtain repeatable sequences of pseudo-random numbers which do not
interfere with each other. For instance objects or modules in a program can
have their own independent streams of random numbers which are repeatable,
independently of other modules making calls to the random number functions.
When \*(TX starts up, the
.code *random-state*
variable is initialized with
a newly created random state object, which is produced as if by
the call
.codn (make-random-state 42) .
.coNP Function @ make-random-state
.synb
.mets (make-random-state <> [ seed ])
.syne
.desc
The
.code make-random-state
function creates and returns a new random state,
an object of the same kind as what is stored in the
.code *random-state*
variable.
The seed, if specified, must be either an integer value, or an
existing random state object.
Note that the sign of the seed is ignored, so that negative seed
values are equivalent to their additive inverses.
If the seed is not specified, then
.code make-random-state
produces a seed based
on some information in the process environment, such as current
time of day. It is not guaranteed that two calls to
.code (make-random-state)
that are separated by less than some minimum increment of real time produce
different seeds. The minimum time increment depends on the platform.
On a platform with a millisecond-resolution real-time clock, the minimum
time increment is a millisecond. Calls to make-random-state less than
a millisecond apart may predictably produce the same seed.
If an integer seed is specified, then the integer value is mapped to a
pseudo-random sequence, in a platform-independent way.
If a random state is specified as a seed, then it is duplicated. The
returned random state object is a distinct object which is in the same
state as the input object. It will produce the same remaining pseudo-random
number sequence, as will the input object.
.coNP Function @ random-state-p
.synb
.mets (random-state-p << obj )
.syne
.desc
The
.code random-state-p
function returns
.code t
if
.meta obj
is a random state, otherwise it
returns
.codn nil .
.coNP Functions @, random-fixnum @ random and @ rand
.synb
.mets (random-fixnum <> [ random-state ])
.mets (random < random-state << modulus )
.mets (rand < modulus <> [ random-state ])
.syne
.desc
All three functions produce pseudo-random numbers, which are positive integers.
The numbers are obtained from a WELLS 512 pseudo-random number generator, whose
state is stored in the random state object.
The
.code random-fixnum
function produces a random fixnum integer: a reduced range
integer which fits into a value that does not have to be heap-allocated.
The
.code random
and
.code rand
functions produce a value in the range [0,
.metn modulus ).
They differ only in the order of arguments. In the
.code rand
function, the random state
object is the second argument and is optional. If it is omitted, the global
.code *random-state*
object is used.
.SS* Time
.coNP Functions @ time and @ time-usec
.synb
(time)
(time-usec)
.syne
.desc
The
.code time
function returns the number of seconds that have elapsed since
midnight, January 1, 1970, in the UTC timezone.
The
.code time-usec
function returns a cons cell whose
.code car
field holds the seconds measured in the same way, and whose
.code cdr
field extends the precision by giving
number of microseconds as an integer value between 0 and 999999.
.coNP Functions @ time-string-local and @ time-string-utc
.synb
.mets (time-string-local < time << format )
.mets (time-string-utc < time << format )
.syne
.desc
These functions take the numeric time returned by the
.code time
function, and convert it to a textual representation in a flexible way,
according to the contents of the
.meta format
string.
The
.code time-string-local
function converts the time to the local timezone of
the host system. The
.code time-string-utc
function produces time in UTC.
The
.meta format
argument is a string, and follows exactly the same conventions as
the format string of the C library function
.codn strftime .
The
.meta time
argument is an integer representing seconds obtained from the
time function or from the
.code car
field of the cons returned by the
.code time-usec
function.
.coNP Functions @ time-fields-local and @ time-fields-utc
.synb
.mets (time-fields-local << time )
.mets (time-fields-utc << time )
.syne
.desc
These functions take the numeric time returned by the time function,
and convert it to a list of seven fields.
The
.code time-string-local
function converts the time to the local timezone of
the host system. The
.code time-string-utc
function produces time in UTC.
The fields returned as a list consist of six integers, and a boolean value.
The six integers represent the year, month, day, hour, minute and second.
The boolean value indicates whether daylight savings time is in effect
(always
.code nil
in the case of
.codn time-fields-utc ).
The
.meta time
argument is an integer representing seconds obtained from the
.code time
function or from the
.code time-usec
function.
.coNP Functions @ make-time and @ make-time-utc
.synb
.mets (make-time < year < month < day < hour < minute < second << dst-advice )
.mets (make-time-utc < year < month < day < hour < minute < second << dst-advice )
.syne
.desc
The
.code make-time
function returns a time value, similar to the one returned by the
.code time
function. The
.code time
value is constructed not from the system clock, but
from a date and time specified as arguments. The
.meta year
argument is a calendar year, like 2014.
The
.meta month
argument ranges from 1 to 12.
The
.meta hour
argument is a 24-hour time, ranging from 0 to 23.
These arguments represent a local time, in the current time zone.
The
.meta dst-advice
argument specifies whether the time is expressed in
daylight savings time (DST). It takes on three possible values:
.codn nil ,
the keyword
.codn :auto ,
or else the symbol
.codn t .
Any other value has the same interpretation as
.codn t .
If
.meta dst-advice
is
.codn t ,
then the time is assumed to be expressed in DST.
If the argument is
.codn nil ,
then the time is assumed not to be in DST.
If
.meta dst-advice
is
.codn :auto ,
then the function tries to determine whether
DST is in effect in the current time zone for the specified date and time.
The
.code make-time-utc
function is similar to
.codn make-time ,
except that
it treats the time as UTC rather than in the local time zone.
The
.meta dst-advice
argument is supported by
.code make-time-utc
for function
call compatibility with
.codn make-time .
It may or may not have any effect
on the output (since the UTC zone by definition doesn't have daylight
savings time).
.SS* Environment Variables and Command Line
Note that environment variable names, their values, and command line
arguments are all regarded as being externally encoded in UTF-8. \*(TX performs
the encoding and decoding automatically.
.coNP Special variables @ *args* and @ *args-full*
.desc
The
.code *args*
variable holds a list of strings representing the remaining
arguments which follow any options processed by the \*(TX executable,
and the script name.
The
.code *args-full*
variable holds the original, complete list of arguments passed
from the operating system.
Note: the
.code *args*
variable is
.code nil
during the processing of the command line,
so \*(TL expressions invoked using the
.code -p
or
.code -e
option cannot use it.
.coNP Function @ env
.synb
(env)
.syne
.desc
The
.code env
function retrieves the list of environment variables. Each
variable is represented by a single entry in the list: a string which
contains an
.code =
(equal) character somewhere, separating the variable name
from its value.
See also: the
.code env-hash
function.
.coNP Function @ env-hash
.synb
(env-hash)
.syne
.desc
The
.code env-hash
function constructs and returns an
.code :equal-based
hash. The hash is
populated with the environment variables, represented as key-value pairs.
.coNP Functions @, getenv @, setenv and @ unsetenv
.synb
.mets (getenv << name )
.mets (setenv < name < value <> [ overwrite-p ])
.mets (unsetenv << name )
.syne
.desc
These functions provide access to, as well as manipulation of, environment
variables. Of these three,
.code setenv
and
.code unsetenv
might not be available on some platforms, or
.code unsetenv
might be be present in a simulated form which sets the variable
.meta name
to the empty string rather than deleting it.
The
.code getenv
function searches the environment for the environment variable whose name
is
.metn name .
If the variable is found, its value is returned. Otherwise
.code nil
is returned.
The
.code setenv
function creates or modifies the environment variable indicated by
.metn name .
The
.meta value
string argument specifies the new value for the variable.
If the
.meta overwrite-p
argument is specified, and is true,
then the variable is overwritten if it already exists.
If the argument is false, then the variable is not modified if it
already exists. If the argument is not specified, it defaults
to the value
.metn t,
effectively giving rise to a two-argument form of
.code setenv
which creates or overwrites environment variables.
The
.code setenv
function unconditionally returns
.meta value
regardless of whether or not it overwrites an existing variable.
The
.code unsetenv
function removes the enviornment variable
specified by
.metn name ,
if it exists. On some platforms, it instead sets the environment variable
to the empty string.
.SS* System Programming
.coNP Function @ errno
.synb
.mets (errno <> [ new-errno ])
.syne
.desc
The
.code errno
function retrieves the current value of the C library error variable
.codn errno .
If the argument
.meta new-errno
is present and is not
.codn nil ,
then it
specifies a value which is stored into
.codn errno .
The value returned is the prior value.
.coNP Function @ exit
.synb
.mets (exit << status )
.syne
.desc
The
.code exit
function terminates the entire process (running \*(TX image), specifying
the termination status to the operating system. Values of
.meta status
may be
.codn nil ,
.codn t ,
or an integer value. The value
.code nil
corresponds to the C constant
.codn EXIT_FAILURE ,
and
.code t
corresponds to
.codn EXIT_SUCCESS .
These are platform-independent
indicators of failed or successful termination. The numeric value 0 also
indicates success.
.coNP Function @ abort
.synb
.mets (abort)
.syne
.desc
The
.code abort
function terminates the entire process (running \*(TX image), specifying
an abnormal termination status to the process.
Note:
.code abort
calls the C library function
.code abort
which works by raising the
.code SIG_ABRT
signal, known in \*(TX as the
.code sig-abrt
variable. Abnormal termination of the process is this signal's
default action.
.coNP Function @ usleep
.synb
.mets (usleep << usec )
.syne
.desc
The
.code usleep
function suspends the execution of the program for at least
.meta usec
microseconds.
The return value is
.code t
if the sleep was successfully executed. A
.code nil
value indicates premature wakeup or complete failure.
Note: the actual sleep resolution is not guaranteed, and depends on granularity
of the system timer. Actual sleep times may be rounded up to the nearest 10
millisecond multiple on a system where timed suspensions are triggered by a 100
Hz tick.
.coNP Functions @ mkdir and @ ensure-dir
.synb
.mets (mkdir < path <> [ mode ])
.mets (ensure-dir < path <> [ mode ])
.syne
.desc
.code mkdir
tries to create the directory named
.meta path
using the POSIX
.code mkdir
function.
An exception of type
.code file-error
is thrown if the function fails. Returns
.code t
on success.
The
.meta mode
argument specifies the request numeric permissions
for the newly created directory. If omitted, the requested permissions are
.code #o777
(511): readable and writable to everyone. The requested permissions
are subject to the system
.codn umask .
The function
.code ensure-dir
is similar to
.code mkdir
except that it attempts to create all the missing parent directories
as necessary, and does not throw an error if the directory exists.
.coNP Function @ chdir
.synb
.mets (chdir << path )
.syne
.desc
.code chdir
changes the current working directory to
.metn path ,
and returns
.metn t ,
or else throws an exception of type
.codn file-error .
.coNP Function @ pwd
.synb
(pwd)
.syne
.desc
The
.code pwd
function retrieves the current working directory.
If the underlying
.code getcwd
C library function fails with an
.code errno
other than
.codn ERANGE ,
an exception will be thrown.
.coNP Functions @ sh and @ run
.synb
.mets (sh << system-command )
.mets (run < program <> [ argument-list ])
.syne
.desc
The
.code sh
function executes
.meta system-command
using the system command interpreter.
The run function spawns a
.metn program ,
searching for it using the
system PATH. Using either method, the executed process receives environment
variables from the parent.
\*(TX blocks until the process finishes executing. If the program terminates
normally, then its integer exit status is returned. The value zero indicates
successful termination.
The return value
.code nil
indicates an abnormal termination, or the inability
to run the process at all.
In the case of the
.code run
function, if the child process is created successfully
but the program cannot be executed, then the exit status will be an
.code errno
value from the failed
.code exec
attempt.
.SS* Unix Filesystem Manipulation
.coNP Function @ stat
.synb
.mets (stat << path )
.syne
.desc
The
.code stat
function inquires the filesystem about the existence of an object
denoted by the string
.metn path .
If the object is not found or cannot be
accessed, an exception is thrown.
Otherwise, information is retrieved about the object. The information takes the
form of a property list in which keyword symbols denote numerous properties.
An example such property list is:
.cblk
(:dev 2049 :ino 669944 :mode 16832 :nlink 23
:uid 500 :gid 500 :rdev 0
:size 12288 :blksize 4096 :blocks 24
:atime 1347933533 :mtime 1347933534 :ctime 1347933534)
.cble
These properties correspond to the similarly-named entries of the
.code struct stat
structure in POSIX. For instance, the
.code :dev
property has the same value as the
.code st_dev
field.
.coNP Special variables @, s-ifmt @, s-iflnk @, s-ifreg @, s-ifblk ... , @ s-ixoth
The following variables exist, having integer values. These are bitmasks
which can be applied against the value given by the
.code :mode
property
in the property list returned by the function stat:
.codn s-ifmt ,
.codn s-ifsock ,
.codn s-iflnk ,
.codn s-ifreg ,
.codn s-ifblk ,
.codn s-ifdir ,
.codn s-ifchr ,
.codn s-ififo ,
.codn s-isuid ,
.codn s-isgid ,
.codn s-isvtx ,
.codn s-irwxu ,
.codn s-irusr ,
.codn s-iwusr ,
.codn s-ixusr ,
.codn s-irwxg ,
.codn s-irgrp ,
.codn s-iwgrp ,
.codn s-ixgrp ,
.codn s-irwxo ,
.codn s-iroth ,
.code s-iwoth
and
.codn s-ixoth .
These variables correspond to the C language constants from POSIX:
.codn S_IFMT ,
.codn S_IFLNK ,
.code S_IFREG
and so forth.
The
.code logtest
function can be used to test these against values of mode.
For example
.code (logtest mode s-irgrp)
tests for the group read permission.
.coNP Functions @, makedev @ minor and @ major
.synb
.mets (makedev < minor << major )
.mets (minor << dev )
.mets (major << dev )
.syne
.desc
The parameters
.metn minor ,
.meta major
and
.meta dev
are all integers. The
.code makedev
function constructs a combined device number from a minor and major pair (by
calling the Unix
.code makedev
function). This device number is suitable as an
argument to the
.codee mknod
function (see below). Device numbers also appear the
.code :dev
property returned by the
.code stat
function.
The
.code minor
and
.code major
functions extract the minor and major device number
from a combined device number.
.coNP Function @ mknod
.synb
.mets (chmod < path << mode )
.syne
.desc
The
.code chmod
function changes the permissions of the filesystem objects
specified by
.metn path.
It is a direct wrapper for the POSIX C library function of the same name.
The permissions are specified by
.metn mode ,
an integer argument.
The existing permissions may be obtained using the
.code stat
function.
The function throws a
.code file-error
exception if an error occurs, otherwise it returns
.codn t.
.TP* Example:
.cblk
;; Set permissions of foo.txt to "rw-r--r--"
;; (owner can read and write; group owner
;; and other users can only read).
;; numerically:
(chmod "foo.txt" #o644)
;; symbolically:
(chmod "foo.txt" (logior s-irusr s-iwusr
s-irgrp
s-iroth))
.cble
.coNP Function @ mknod
.synb
.mets (mknod < path < mode <> [ dev ])
.syne
.desc
The
.code mknod
function tries to create an entry in the filesystem: a file,
FIFO, or a device special file, under the name
.metn path .
If it is successful,
it returns
.codn t ,
otherwise it throws an exception of type
.codn file-error .
The
.meta mode
argument is a bitwise or combination of the requested permissions,
and the type of object to create: one of the constants
.codn s-ifreg ,
.codn s-ififo ,
.codn s-ifchr ,
.code s-ifblk
or
.codn s-ifsock .
The permissions are subject to the system
.codn umask .
If a block or character special device
.cod2 ( s-ifchr
or
.codn s-ifblk )
is being
created, then the
.meta dev
argument specifies the major and minor numbers
of the device. A suitable value can be constructed from a major and minor
pair using the
.code makedev
function.
.TP* Example:
.cblk
;; make a character device (8, 3) called /dev/foo
;; requesting rwx------ permissions
(mknod "dev/foo" (logior #o700 s-ifchr) (makedev 8 3))
.cble
.coNP Functions @ symlink and @ link
.synb
.mets (symlink < target << path )
.mets (link < target << path )
.syne
.desc
The
.code symlink
function creates a symbolic link called
.meta path
whose contents
are the absolute or relative path
.metn target .
.meta target
does not actually have to exist.
The link function creates a hard link. The object at
.meta target
is installed
into the filesystem at
.meta path
also.
If these functions succeed, they return
.codn t .
Otherwise they throw an exception
of type
.codn file-error .
.coNP Function @ readlink
.synb
.mets (readlink << path )
.syne
.desc
If
.meta path
names a filesystem object which is a symbolic link, the
.code readlink
function reads the contents of that symbolic link and returns it
as a string. Otherwise, it fails by throwing an exception of type
.codn file-error .
.SS* Unix Signal Handling
On platforms where certain advanced features of POSIX signal handling are
available at the C API level, \*(TX exposes signal-handling functionality.
A \*(TX program can install a \*(TL function (such as an anonymous.
.codn lambda ,
or the function object associated with a named function) as the handler for
a signal.
When that signal is delivered, \*(TX will intercept it with its own safe,
internal handler, mark the signal as deferred (in a \*(TX sense) and then
dispatch the registered function at a convenient time.
Handlers currently are not permitted to interrupt the execution of most
\*(TX internal code. Immediate, asynchronous execution of handlers is
currently enabled only while \*(TX is blocked on I/O operations or sleeping.
Additionally, the
.code sig-check
function can be used to dispatch and clear deferred
signals. These handlers are then safely called if they were subroutines of
.codn sig-check ,
and not asynchronous interrupts.
.coNP Special variables @, sig-hup @, sig-int @, sig-quit @, sig-ill @, sig-trap @, sig-abrt @, sig-bus @, sig-fpe @, sig-kill @, sig-usr1 @, sig-segv @, sig-usr2 @, sig-pipe @, sig-alrm @, sig-term @, sig-chld @, sig-cont @, sig-stop @, sig-tstp @, sig-ttin @, sig-ttou @, sig-urg @, sig-xcpu @, sig-xfsz @, sig-vtalrm @, sig-prof @, sig-poll @, sig-sys @, sig-winch @, sig-iot @, sig-stkflt @, sig-io @ sig-lost and @ sig-pwr
.desc
These variables correspond to the C signal constants
.codn SIGHUP ,
.code SIGINT
and so forth.
The variables
.codn sig-winch ,
.codn sig-iot ,
.codn sig-stk flt,
.codn sig-io ,
.code sig-lost
and
.code sig-pwr
may not be available since a system may lack the corresponding signal
constants. See notes for the function
.codn log-authpriv .
The highest signal number is 31.
.coNP Functions @ set-sig-handler and @ get-sig-handler
.synb
.mets (set-sig-handler < signal-number << handling-spec )
.mets (get-sig-handler << signal-number )
.syne
.desc
The
.code set-sig-handler
function is used to specify the handling for a signal, such
as the installation of a handler function. It updates the signal handling for
a signal whose number is
.meta signal-number
(usually one of the constants like
.codn sig-hup ,
.code sig-int
and so forth), and returns the previous value. The
.code get-sig-handler
function returns the current value.
The
.meta signal-number
must be an integer the range 1 to 31.
Initially, all 31 signal handling specifications are set to the value
.codn t .
The
.meta handling-spec
parameter may be a function. If a function is specified,
then the signal is enabled and connected to that function until another
call to
.code set-sig-handler
changes the handling for that signal.
If
.meta handling-spec
is the symbol
.codn nil ,
then the function previously associated
with the signal, if any, is removed, and the signal is disabled. For a signal
to be disabled means that the signal is set to the
.code SIG_IGN
disposition (refer to the C API).
If
.meta handling-spec
is the symbol
.codn t ,
then the function previously associated
with the signal, if any, is removed, and the signal is set to its default
disposition. This means that it is set to
.code SIG_DFL
(refer to the C API).
Some signals terminate the process if they are generated while the
handling is configured to the default disposition.
Note that the certain signals like
.code sig-quit
and
.code sig-kill
cannot be ignored or handled.
Please observe the signal documentation in the IEEE POSIX standard, and your
platform.
A signal handling function must take two arguments. It is of the form:
.cblk
.mets (lambda >> ( signal << async-p ) ...)
.cble
The
.meta signal
argument is an integer indicating the signal number for which the
handler is being invoked. The
.meta asyncp-p
argument is a boolean value.
If it is
.codn t ,
it indicates that the handler is being invoked
asynchronously\(emdirectly in a signal handling context. If it is
.codn nil ,
then it
is a deferred call. Handlers may do more things in a deferred call, such
as terminate by throwing exceptions, and perform I/O.
The return value of a handler is normally ignored. However if it invoked
asynchronously (the
.meta async-p
argument is true), then if the handler returns
a
.cod2 non- nil
value, it is understood that the handler
requesting that it be deferred. This means that the signal will be marked
as deferred, and the handler will be called again at some later
time in a deferred context, whereby
.meta async-p
is
.codn nil .
This is not guaranteed, however;
it's possible that another signal will arrive before that happens,
possibly resulting in another async call, so the handler must
be prepared to deal with an async call at any time.
If a handler is invoked synchronously, then its return value is ignored.
In the current implementation, signals do not queue. If a signal is delivered
to the process again, while it is marked as deferred, it simply stays deferred;
there is no counter associated with a signal, only a boolean flag.
.coNP Function @ sig-check
.synb
(sig-check)
.syne
.desc
The
.code sig-check
function tests whether any signals are deferred, and for each
deferred signal in turn, it executes the corresponding handler. For a signal to
be deferred means that the signal was caught by an internal handler in
\*(TX and the event was recorded by a flag. If a handler function is removed
while a signal is deferred, the deferred flag is cleared for that signal.
Calls to the
.code sig-check
function may be inserted into CPU-intensive code that
has no opportunity to be interrupted by signals, because it doesn't invoke any
I/O functions.
.coNP Function @ kill
.synb
.mets (kill < process-id <> [ signal ])
.syne
.desc
The
.code kill
function is used for sending a signal to a process group or process.
It is a wrapper for the POSIX
.code kill
function.
If the
.meta signal
argument is omitted, it defaults to the same value as
.codn sig-term .
.SS* Unix Processes
.coNP Functions @ fork and @ wait
.synb
.mets (fork)
.mets (wait >> [ pid <> [ flags ]])
.syne
.desc
The
.code fork
and
.code wait
functions are interfaces to the Unix functions
.code fork
and
.codn waitpid .
The
.code fork
function creates a child process which is a replica of the parent. Both
processes return from the function. In the child process, the return value is
zero. In the parent, it is an integer representing the process ID of the child.
If the function fails to create a child, it returns
.code nil
rather than an integer. In this case, the
.code errno
function can be used to inquire about the cause.
The
.code wait
function, if successful, returns a cons cell consisting of a pair of integers.
The
.code car
of the cons is the process ID of the process or group which was successfully
waited on, and the
.code cdr
is the status. If
.code wait
fails, it returns
.codn nil .
The
.code errno
function can be used to inquire about the cause.
The
.meta process-id
argument, if not supplied, defaults to -1, which means that
.code wait
waits for any process, rather than a specific process. Certain other
values have special meaning, as documented in the POSIX standard
for the
.code waitpid
function.
The
.meta flags
argument defaults to zero. If it is specified as nonzero, it should be
a bitwise combination (via the
.code logior
function) of the constants
.codn w-nohang ,
.codn w-untraced
and
.codn w-continued .
If
.code w-nohang
is used, then
.code wait
returns a cons cell whose
.code car
specifies a process ID value of zero in the situation that at least
one of the processes designated by
.code process-id
exist and are children of the calling process, but have not changed state.
In this case, the status value in the
.code cdr
is unspecified.
Status values may be inspected with the functions
.codn w-ifexited ,
.codn w-exitstatus ,
.codn w-ifsignaled ,
.codn w-termsig ,
.codn w-coredump ,
.codn w-ifstopped ,
.code w-stopsig
and
.codn w-ifcontinued .
.coNP Functions @, w-ifexited @, w-exitstatus @, w-ifsignaled @, w-termsig @, w-coredump @ w-ifstopped and @ w-stopsig
.synb
.mets (w-ifexited << status )
.mets (w-exitstatus << status )
.mets (w-ifsignaled << status )
.mets (w-termsig << status )
.mets (w-coredump << status )
.mets (w-ifstopped << status )
.mets (w-stopsig << status )
.mets (w-ifcontinued << status )
.syne
.desc
These functions analyze process exit values produced by the
.code wait
function.
They are closely based on the
POSIX macros
.codn WIFEXITED ,
.code WEXITSTATUS ,
and so on.
The
.meta status
value is either an integer, or a cons cell. In this case, the cons
cell is expected to have an integer in its
.code cdr
which is used as the status.
The
.codn w-ifexited ,
.codn w-ifsignaled ,
.codn w-coredump ,
.code w-ifstopped
and
.code w-ifcontinued
functions have Lisp boolean return semantics, unlike their C language
counterparts: they return
.code t
or
.codn nil ,
rather than zero or nonzero. The others return integer values.
.coNP Function @ exec
.synb
.mets (exec < file <> [ args ])
.syne
.desc
The exec function replaces the process image with the executable specified
by string argument
.metn file .
The executable is found by searching the system path.
The
.meta file
argument becomes the first argument of the executable, argument zero.
If
.meta args
is specified, it is a list of strings. These are passed as the additional
arguments of the executable.
If
.code exec
fails, an exception of type
.code file-error
is thrown.
.coNP Function @ exit*
.synb
.mets (exit* << status )
.syne
.desc
The
.code exit*
function terminates the entire process (running \*(TX image), specifying
the termination status to the operating system. The
.meta status
argument is treated exactly like that of the
.code exit
function. Unlike that function, this one exits the process immediately,
cleaning up only low-level operating system resources such as closing file
descriptors and releasing memory mappings, without performing user-space
cleanup.
.code exit*
is implemented using a call to the POSIX function
.codn _exit .
.coNP Functions @ getpid and @ getppid
.synb
(getpid)
(getppid)
.syne
.desc
These functions retrieve the current proces ID and the parent process ID
respectively. They are wrappers for the POSIX functions
.code getpid
and
.codn getppid .
.coNP Function @ daemon
.synb
.mets (daemon < nochdir-p << noclose-p )
.syne
.desc
This is a wrapper for the function
.code daemon
which originated in BSD Unix.
It returns
.code t
if successful,
.code nil
otherwise, and the
.code errno
variable is set in that case.
.SS* Unix File Descriptors
.coNP Function @ open-fileno
.synb
.mets (open-fileno < file-descriptor <> [ mode-string ])
.syne
The
.code open-fileno
function creates a \*(TX stream over a file descriptor. The
.meta file-descriptor
argument must be an integer denoting a valid file descriptor.
For a description of
.metn mode-string ,
see the
.code open-file
function.
.coNP Function @ fileno
.synb
.mets (fileno << stream )
.syne
.desc
The
.code fileno
function returns the underlying file descriptor of
.metn stream ,
if it has one. Otherwise, it returns
.codn nil.
This is equivalent to querying the stream using
.code stream-get-prop
for the
.code :fd
property.
.coNP Function @ dupfd
.synb
.mets (dupfd < old-fileno <> [ new-fileno ])
.syne
.desc
The
.code dupfd
function provides an interface to the POSIX functions
.code dup
or
.codn dup2 ,
when called with one or two arguments, respectively.
.coNP Function @ pipe
.synb
(pipe)
.syne
.desc
The
.code pipe
function, if successful, returns a pair of integer file descriptors
as a cons cell pair. The descriptor in the
.code car
field of the pair is the read end of the pipe.
The
.code cdr
holds the write end.
If the function fails, it throws an exception of type
.codn file-error .
.coNP Function @ poll
.synb
.mets (poll < poll-list <> [ timeout ])
.syne
.desc
The
.code poll
function suspends execution while monitoring one or more file descriptors
for specified events. It is a wrapper for the same-named POSIX function.
The
.meta poll-list
argument is a list of
.code cons
pairs. The
.code car
of each pair is either an integer file descriptor, or else a stream
object which has a file descriptor (the
.code fileno
function can be applied to that stream to retrieve a descriptor).
The
.code cdr
of each pair is an integer bit mask specifying the events, whose
occurrence the file descriptor is to be monitored for. The variables
.codn poll-in ,
.codn poll-out ,
.code poll-err
and several others are available which hold bitmask values corresponding
to the constants
.codn POLLIN ,
.codn POLLOUT ,
.code POLLERR
used with the C language
.code poll
function.
The
.meta timeout
argument, if absent, defaults to the value -1, which specifies an indefinite
wait. A nonnegative value specifies a wait with a timeout, measured in
milliseconds.
The function returns a list of pairs representing the descriptors or streams
which were successfully polled. If the function times out, it returns an
empty list. If an error occurs, an exception is thrown.
The returned list is similar in structure to the input list. However, it holds
only entries which polled positive. The
.code cdr
of every pair now holds a bitmask of the events which were to have occurred.
.SS* Unix Itimers
Itimers ("interval timers") can be used in combination with signal handling to
execute asynchronous actions. Itimers deliver delayed, one-time signals,
and also periodically recurring signals. For more information, consult the
POSIX specification.
.coNP Variables @, itimer-real @, itimer-virtual and @ itimer-prof
.desc
These variables correspond to the POSIX constants
.codn ITIMER_REAL ,
.code ITIMER_VIRTUAL
and
.codn ITIMER_PROF .
Their values are suitable as the
.meta timer
argument of the
.code getitimer
and
.code setitimer
functions.
.coNP Functions @ getitimer and @ setitimer
.synb
.mets (getitimer << timer )
.mets (setitimer < timer < interval << value )
.syne
.desc
The
.code getitimer
function returns the current value of the specified timer,
which must be
.codn itimer-real ,
.code itimer-virtual
or
.codn itimer-prof .
The current value consists of a list of two integer values, which
represents microseconds. The first value is the timer interval,
and the second value is the timer's current value.
Like
.codn getitimer ,
the
.code setitimer
function also retrieves the specified timer.
In addition, it stores a new value in the timer,
which is given by the two arguments, expressed in microseconds.
.SS* Unix Syslog
On platforms where a Unix-like syslog API is available, \*(TX exports this
interface. \*(TX programs can configure logging via the
.code openlog
function,
control the logging mask via
.code setlogmask
and generate logs via
.codn syslog ,
or using special syslog streams.
.coNP Special variables @, log-pid @, log-cons @, log-ndelay @, log-odelay @ log-nowait and @ log-perror
.desc
These variables take on the values of the corresponding C preprocessor
constants from the
.code <syslog.h>
header:
.codn LOG_PID ,
.codn LOG_CON S, etc.
These integer values represent logging options used in the option argument to
the
.code openlog
function.
Note:
.code LOG_PERROR
is not in POSIX, and so
.code log-perror
might not be available.
See notes about
.code LOG_AUTHPRIV
in the documentation for
.codn log-authpriv .
.coNP Special variables @, log-user @, log-daemon @ log-auth and @ log-authpriv
.desc
These variables take on the values of the corresponding C preprocessor
constants from the
.code <syslog.h>
header:
.codn LOG_USER ,
.codn LOG_DAEMON ,
.code LOG_AUTH
and
.codn LOG_AUTHPRIV .
These are the integer facility codes specified in the
.code openlog
function.
Note:
.code LOG_AUTHPRIV
is not in POSIX, and so
.code log-authpriv
might not be available.
For portability use code like
.code (or (symbol-value 'log-authpriv) 0)
to evaluate to 0 if
.code log-authpriv
doesn't exist, or else check for its existence
using
.codn (boundp 'log-authpriv) .
.coNP Special variables @, log-emerg @, log-alert @, log-crit @, log-err @, log-warning @, log-notice @ log-info and @ log-debug
These variables take on the values of the corresponding C preprocessor
constants from the
.code <syslog.h>
header:
.codn LOG_EMERG ,
.codn LOG_ALERT ,
etc.
These are the integer priority codes specified in the
.code syslog
function.
.coNP The @ *stdlog* special variable
.desc
The
.code *stdlog*
variable holds a special kind of stream: a syslog stream. Each
newline-terminated line of text sent to this stream becomes a log message.
The stream internally maintains a priority value that is applied
when it generates messages. By default, this value is that of
.codn log-info .
The stream holds the priority as the value of the
.code :prio
stream property, which may be changed with the
.code stream-set-prop
function.
The latest priority value which has been configured on the stream is used
at the time the newline character is processed and the log message
is generated, not necessarily the value which was in effect at the time the
accumulation of a line began to take place.
Messages sent to
.code *stdlog*
are delimited by newline characters. That is to say, each line of
text written to the stream is a new log.
.coNP Function @ openlog
.synb
.mets (openlog < id-string >> [ options <> [ facility ]])
.syne
.desc
The
.code openlog
function is a wrapper for the
.code openlog
C function, and the
arguments have the same semantics. It is not necessary to use
.code openlog
in order
to call the
.code syslog
function or to write data to
.codn *stdlog* .
The call is necessary in order to override the default identifying string, to
set options, such as having the PID (process ID) recorded in log messages, and
to specify the facility.
The
.meta id-string
argument is mandatory.
The
.meta option
argument is a bitwise mask (see the logior function) of option
values such as
.code log-pid
and
.codn log-cons .
If it is missing, then a value of 0 is
used, specifying the absence of any options.
The
.meta facility
argument is one of the values
.codn log-user ,
.code log-daemon
or
.codn log-auth .
If it is missing, then
.code log-user
is assumed.
.coNP Function @ closelog
.synb
(closelog)
.syne
.desc
The
.code closelog
function is a wrapper for the C function
.codn closelog .
.coNP Function @ setlogmask
.synb
.mets (setlogmask << bitmask-integer )
.syne
.desc
The
.code setlogmask
function interfaces to the corresponding C function, and has the
same argument and return value semantics. The
.meta bitmask-integer
argument is a mask of priority
values to enable. The return value is the prior value. Note that if the
argument is zero, then the function doesn't set the mask to zero; it only
returns the current value of the mask.
Note that the priority values like
.code log-emerg
and
.code log-debug
are integer
enumerations, not bitmasks. These values cannot be combined directly to create
a bitmask. Rather, the
.code mask
function should be used on these values.
.TP* Example:
.cblk
;; Enable LOG_EMERG and LOG_ALERT messages,
;; suppressing all others
(setlogmask (mask log-emerg log-alert))
.cble
.coNP Function @ syslog
.synb
.mets (syslog < priority < format << format-arg *)
.syne
.desc
This function is the interface to the
.code syslog
C function. The
.code printf
formatting capabilities of the function are not used;
the
.meta format
argument follows the conventions of the \*(TL
.code format
function instead. Note in particular that
the
.code %m
convention for interpolating the value of strerror(errno) which is
available in some versions of the
.code syslog
C function is currently not supported.
Note that syslog messages are not newline-terminated.
.SS* Unix Path Globbing
On platforms where the POSIX
.code glob
function is available \*(TX provides this functionality in
the form of a like-named function, and some numeric constants.
.coNP Special variables @, glob-err @, glob-mark @, glob-nosort @, glob-nocheck @, glob-noescape @, glob-period @, glob-altdirfunc @, glob-brace @, glob-nomagic @, glob-tilde @ glob-tilde-check and @ glob-onlydir
These variables take on the values of the corresponding C preprocessor
constants from the
.code <glob.h>
header:
.codn GLOB_ERR ,
.codn GLOB_MARK ,
.codn GLOB_NOSORT ,
etc.
These values are passed as the optional second argument of the
.code glob
function. They are bitmasks and so multiple values can be combined
using the
.code logior
function.
Note that the
.codn glob-period ,
.codn glob-altdirfunc ,
.codn glob-brace ,
.codn glob-nomagic ,
.codn glob-tilde ,
.code glob-tilde-check
and
.code glob-onlydir
variables may not be avaiable. They are extensions in the GNU C library
implementation of
.codn glob .
.coNP Function @ glob
.synb
.mets (glob < pattern >> [ flags <> [ error-func ]])
.syne
.desc
The
.code glob
function is a interface to the Unix function of the same name.
The
.meta pattern
argument must be a string, which holds a glob pattern: a pattern which
matches zero or more path names, similar to a regular expression.
The function tries to expand the pattern and return a list of strings
representing the matching path names in the file system.
If there are no matches, then an empty list is returned.
The optional
.meta flags
argument defaults to zero. If given, it may be a bitwise combination of the
values of the variables
.codn glob-err ,
.codn glob-mark ,
.code glob-nosort
and others.
If the
.meta error-func
argument is specified, it gives a callback function which is invoked
when
.code glob
encounters errors accessing paths. The function takes two arguments:
the pathname and the
.code errno
value which occurred for that pathname. The function's return value is
boolean. If the function returns true, then
.code glob
will terminate.
Details of the semantics of the
.code glob
function, and the meaning of all the
.meta flags
arguments are given in the documentation for the C function.
.SS* Web Programming Support
.coNP Functions @ url-encode and @ url-decode
.synb
.mets (url-encode < string <> [ space-plus-p ])
.mets (url-decode < string <> [ space-plus-p ])
.syne
.desc
These functions convert character strings to and from a form which is suitable
for embedding into the request portions of URL syntax.
Encoding a string for URL use means identifying in it certain characters that
might have a special meaning in the URL syntax and representing it using
"percent encoding": the percent character, followed by the ASCII value of the
character. Spaces and control characters are also encoded, as are all byte
values greater than or equal to 127 (7F hex). The printable ASCII characters
which are percent-encoded consist of this set:
.cblk
:/?#[]@!$&'()*+,;=%
.cble
More generally, strings can consists of Unicode characters, but the URL
encoding consists only of printable ASCII characters. Unicode characters in the
original string are encoded by expanding into UTF-8, and applying
percent-encoding the UTF-8 bytes, which are all in the range
.codn \exx80-\exxFF .
Decoding is the reverse process: reconstituting the UTF-8 byte sequence
specified by the URL-encoding, and then decoding the UTF-8 sequence into the
string of Unicode characters.
There is an additional complication: whether or not to encode spaces as plus,
and to decode plus characters to spaces. In encoding, if spaces are not encoded
to the plus character, then they are encoded as
.codn %20 ,
since spaces are reserved
characters that must be encoded. In decoding, if plus characters are not
decoded to spaces, then they are left alone: they become plus characters in the
decoded string.
The
.code url-encode
function performs the encoding process. If the
.code space-plus-p
argument is omitted or specified as
.codn nil ,
then spaces are encoded as
.codn %20 .
If the argument is a value other than
.codn nil ,
then spaces are encoded as the
character
.code +
.codn (plus) .
The
.code url-decode
function performs the decoding process. If the
.code space-plus-p
argument is omitted or specified as
.codn nil ,
then
.code +
.code (plus)
characters in the
encoded data are retained as
.code +
characters in the decoded strings. Otherwise,
plus characters are converted to spaces.
.coNP Functions @ html-encode and @ html-decode
.synb
.mets (html-encode << text-string )
.mets (html-decode << html-string )
.syne
.desc
The
.code html-encode
and
.code html-decode
functions convert between an HTML and raw
representation of of text.
The
.code html-encode
function returns a string which is based on the content of
.metn text-string ,
but in which all characters which have special meaning in HTML
have been replaced by HTML codes for representing those characters literally.
The returned string is the HTML-encoded verbatim representation of
.metn text-string .
The
.code html-decode
function converts
.metn html-string ,
which may contain HTML
character encodings, into a string which contains the actual characters
represented by those encodings.
The function composition
.code (html-decode (html-encode text))
returns a string which is equal to
.codn text .
The reverse composition
.code (html-encode (html-decode html))
does not necessarily return a string equal to
.codn html .
For instance if html is the string
.strn "<p>Hello, world!</p>" ,
then
.code html-decode
produces
.strn "<p>Hello, world!</p>" .
From this,
.code html-encode
produces
.strn "<p>Hello, world!</p>" .
.SS* Filter Module
The filter module provides a trie (pronounced "try") data structure,
which is suitable for representing dictionaries for efficient filtering.
Dictionaries are unordered collections of keys, which are strings, which
have associated values, which are also strings. A trie can be used to filter
text, such that keys appearing in the text are replaced by the corresponding
values. A trie supports this filtering operation by providing an efficient
prefix-based lookup method which only looks at each input character ones, and
which does not require knowledge of the length of the key in advance.
.coNP Function @ make-trie
.synb
(make-trie)
.syne
.desc
The
.code make-trie
function creates an empty trie. There is no special data type for
a trie; a trie is some existing type such as a hash table.
.coNP Function @ trie-add
.synb
.mets (trie-add < trie < key << value )
.syne
.desc
The
.code trie-add
function adds the string
.meta key
to the trie, associating
it with
.metn value .
If
.meta key
already exists in
.metn trie ,
then the value is updated with
.metn value .
The
.meta trie
must not have been compressed with
.metn trie-compress .
A trie can contain keys which are prefixes of other keys. For instance
it can contain
.str dog
and
.strn dogma .
When a trie is used for matching
and substitution, the longest match is used. If the input presents
the text
.strn doggy ,
then the match is
.strn dog .
If the input is
.strn dogmatic ,
then
.str dogma
matches.
.coNP Function @ trie-compress
.synb
.mets (trie-compress << trie )
.syne
.desc
The
.code trie-compress
function changes the representation of
.meta trie
to a representation which occupies less space and supports faster lookups.
The new representation is returned.
The compressed representation of a trie does not support the
.code trie-add
function.
This function destructively manipulates
.metn trie ,
and may return an object
that is the same object as
.codn trie ,
or it may return a different object,
while at the same time still modifying the internals of
.metn trie .
Consequently, the program should not retain the input object
.codn trie ,
but use the returned object in its place.
.coNP Function @ trie-lookup-begin
.synb
.mets (trie-lookup-begin << trie )
.syne
.desc
The
.code trie-lookup-begin
function returns a context object for performing
an open-coded lookup traversal of a trie. The
.meta tri
argument
is expected to be a trie that was created by the
.code make-trie
function.
.coNP Function @ trie-lookup-feed-char
.synb
.mets (trie-lookup-feed-char < trie-context << char )
.syne
.desc
The
.code trie-lookup-feed-char
function performs a one character step in a trie
lookup. The
.meta trie-context
argument must be a trie contxt returned
by
.metn trie-lookup-begin ,
or by some previous call to
.codn trie-lookup-feed-char .
The
.meta char
argument is the next character to match.
If the lookup is successful (the match through the trie can continue
with the given character) then a new trie context object is returned.
The old trie context remains valid.
If the lookup is unsuccessful,
.code nil
is returned.
Note: determining whether a given string is stored in a trie can be
performed looking up every character of the string successively
with
.codn trie-lookup-feed-char ,
using the newly returned context
for each successive operation. If every character is found, it means
that either that exact string is found in the trie, or a prefix.
The ambiguity can be resolved by testing whether the trie has a value
at the last node using
.codn tree-value-at .
For instance, if
.str catalog
is inserted into an empty trie with value
.strn foo ,
then
.str cat
will look up successfully, being a prefix of
.strn catalog ;
however, the value at
.str cat
is
.codn nil ,
indicating that
.str cat
is only a prefix of one or more entries in the trie.
.coNP Function tree-value-at
.synb
.mets (trie-value-at << trie-context )
.syne
.desc
The
.code trie-value-at
function returns the value stored at the node in
in the trie given by
.metn trie-context .
Nodes which have not been given
a value hold the value
.codn nil .
.coNP Function @ filter-string-tree
.synb
.mets (filter-string-tree < filter << obj )
.syne
.desc
The
.code filter-string-tree
a tree structure similar to
.metn obj ,
in which all of the
string atoms have been filtered through
.metn filter .
The
.meta obj
argument is a string tree structure: either the symbol
.codn nil ,
denoting an empty structure; a string; or a list of tree structures. If
.meta obj
is
.codn nil ,
then
.code filter-string-tree
returns
.codn nil .
The
.meta filter
argument is a filter: it is either a trie, a function, or nil.
If
.meta filter
is
.codn nil ,
then
.code filter-string-trie
just returns
.metn obj .
If
.meta filter
is a function, it must be a function that can be called
with one argument. The strings of the string tree are filtered by passing
each one into the function and substituting the return value into the
corresponding place in the returned structure.
Otherwise if
.meta filter
is a trie, then this trie is used for filtering,
the string elements similarly to a function. For each string, a new
string is returned in which occurrences of the keys in the trie are
replaced by the values in the trie.
.coNP Function @ filter-equal
.synb
.mets (filter-equal < filter-1 < filter-2 < obj-1 << obj-2 )
.syne
.desc
The
.code filter-equal
function tests whether two string trees are equal
under the given filters.
The precise semantics can be given by this expression:
.cblk
.mets (equal (filter-string-tree < filter-1 << obj-1 )
.mets \ \ \ \ \ \ (filter-string-tree < filter-2 << obj-2 ))
.cble
The string tree
.meta obj-1
is filtered through
.metn filter-1 ,
as if by the
.code filter-string-tree
function, and similarly,
.meta obj-2
is
filtered through
.metn filter-2 .
The resulting structures are compared
using
.codn equal ,
and the result of that is returned.
.SS* Access To TXR Pattern Language From Lisp
It is useful to be able to invoke the abilities of the \*(TX pattern Language
from \*(TL. An interface for doing this provided in the form of the
.code match-fun
function, which is used for invoking a \*(TX pattern function.
The
.code match-fun
function has a cumbersome interface which requires the \*(TL program to
explicitly deal with the variable bindings emerging from the pattern match
in the form of an association list.
To make it the interface easier to use, \*(TX provides a library of macros:
the macros
.codn txr-if ,
.codn txr-when
and
.codn txr-case .
These macros are not in the \*(TX image; they must be included from the
.code stdlib
directory.
.coNP Function match-fun
.synb
.mets (match-fun < name < args < input << files )
.syne
.desc
The
.code match-fun
function invokes a \*(TX pattern function whose name is
given by
.metn name ,
which must be a symbol.
The
.meta args
argument is a list of expressions. The expressions may be symbols
which will be interpreted as pattern variables, and may be bound or unbound.
If they are not symbols, then they are treated as expressions (of the
pattern language, not \*(TL) and evaluated accordingly.
The
.meta input
argument is a list of strings, which may be lazy. It represents the
lines of the text stream to be processed.
The
.meta file
argument is a list of filename specifications, which follow
the same conventions as files given on the \*(TX command line. If the pattern
function uses the
.code @(next)
directive, it can process these additional files.
The
.code match-fun
function's return value falls into three cases. If there is a
match failure, it returns
.codn nil .
Otherwise it returns a cons cell. The
.code car
field
of the cons cell holds the list of captured bindings. The
.code cdr
of the cons cell is one of two values. If the entire input was processed, the
cdr field holds the symbol
.codn t .
Otherwise it holds another cons cell whose
.code car
is the remainder of the list of lines which were not matched, and whose
.code cdr
is the line number.
.TP* Example:
.cblk
@(define foo (x y))
@x:@y
@line
@(end)
@(do
(format t "~s\en"
(match-fun 'foo '(a b)
'("alpha:beta" "gamma" "omega") nil)))
Output:
(((a . "alpha") (b . "beta")) ("omega") . 3)
.cble
In the above example, the pattern function
.code foo
is called with arguments
.codn (a b) .
These are unbound variables, so they correspond to parameters
.code x
and
.code y
of the function. If
.code x
and
.code y
get bound, those values propagate to
.code a
and
.codn b .
The data being matched consists of the lines
.strn alpha:beta ,
.str gamma
and
.strn omega .
Inside
.codn foo,
.code x
and
.code y
bind to
.str alpha
and
.strn beta ,
and then the line variable binds to
.strn gamma .
The input stream is left with
.strn omega .
Hence, the return value consists of the bindings of
.code x
and
.code y
transferred to
.code a
and
.codn b ,
and the second cons cell which gives information about the rest of the
stream: it is the part starting at
.strn omega ,
which is line 3. Note that the binding for the
.code line
variable does not propagate
out of the pattern function
.codn foo ;
it is local inside it.
.coNP Macro txr-if
.reqb
.mets @(include `@stdlib/txr-case`)
.reqe
.synb
.mets (txr-if < name <> ( argument *) < input < then-expr <> [ else-expr ])
.syne
.desc
The
.code txr-if
macro invokes the \*(TX pattern matching function
.metn name
on some input given by the
.meta input
parameter, which is a list of strings, or a single string.
If
.meta name
succeeds, then
.meta then-expr
is evaluated, and if it fails,
.meta else-expr
is evaluated instead.
In the successful case,
.meta then-expr
is evaluated in a scope in which the bindings emerging from the
.meta name
function are turned into \*(TL variables.
The result of
.code txr-if
is that of
.metn then-expr .
In the failed case,
.meta else-expr
is evaluated in a scope which does not have any new bindings.
The result of
.code txr-if
is that of
.metn else-expr .
If
.meta else-expr
is missing, the result is
.codn nil .
The
.meta argument
forms supply arguments to the pattern function
.metn name .
There must be as many of these arguments as the function
has parameters.
Any argument which is a symbol is treated, for the purposes
of calling the pattern function, as an unbound pattern variable.
The function may or may not produce a binding for that variable.
Also, every argument which is a symbol also denotes a local variable
that is established around
.meta then-expr
if the function suceeds. For any such pattern variable for which the function
produces a binding, the corresponding local variable will be initialized
with the value of that pattern variable. For any such pattern variable
which is left unbound by the function, the corresponding local variable
will be set to
.codn nil .
Any
.meta argument
can be a form other than a symbol. In this situation, the argument is
evaluated, and will be passed to the pattern function as the value of
the binding for the corresponding argument.
.TP* Example:
.cblk
@(include `@stdlib/txr-case`)
@(define date (year month day))
@{year /\ed\ed\ed\ed/}-@{month /\ed\ed/}-@{day /\ed\ed/}
@(end)
@(do
(each ((date '("09-10-20" "2009-10-20" "July-15-2014" "foo")))
(txr-if date (y m d) date
(put-line `match: year @y, month @m, day @d`)
(put-line `no match for @date`))))
Output:
no match for 09-10-20
match: year 2009, month 10, day 20
no match for July-15-2014
no match for foo
.cble
.coNP Macro @ txr-when
.reqb
.mets @(include `@stdlib/txr-case`)
.reqe
.synb
.mets (txr-when < name <> ( argument *) < input << form *)
.syne
.desc
The
.code txr-when
macro is based on
.codn txr-if .
It is equivalent to
.code
.cblk
.meti \ \ (txr-if < name <> ( argument *) < input (progn << form *))
.cble
If the pattern function
.meta name
produces a match, then each
.meta form
is evaluated in the scope of the variables established by the
.meta argument
expressions. The result of the
.code txr-when
form is that of the last
.metn form .
If the pattern function fails then the forms are not evaluated,
adn the result value is
.codn nil .
.coNP Macro @ txr-case
.reqb
.mets @(include `@stdlib/txr-case`)
.reqe
.synb
.mets (txr-case < input-form
.mets \ \ >> {( name <> ( argument *) << form *)}*
.mets \ \ >> [( t << form *)])
.syne
.desc
The
.code txr-case
macro evaluates
.meta input-form
and then uses the value as an input to zero or more test clauses.
Each test clause invokes the pattern function named by that clause's
.meta name
argument.
If the function succeeds, then each
.meta form
is evaluated, and the value of the last
.meta form
is taken to be the result value of
.codn txr-case ,
which terminates. If there are no forms, then
.code txr-case
terminates with a
.code nil
result.
The forms are evaluated in an environment in which variables are bound
based on the
.meta argument
forms, with values depending on the result of the
invocation of the
.meta name
pattern function, in the same manner as documented in detail for the
.code txr-if
macro.
If the function fails, then the forms are not evaluated, and control passes to
the next clause.
A clause which begins with the symbol
.code t
executes unconditionally and causes
.code txr-case
to terminate. If it has no forms, then
.code txr-case
yields
.codn nil ,
otherwise the forms are evaluated in order and the value of the last
one specifies the result of
.codn txr-case .
.SS* Quote/Quasiquote Operator Syntax
.coNP Operator @ quote
.synb
.mets (quote << form )
.syne
.desc
The
.code quote
operator, when evaluated, suppresses the evaluation of
.metn form ,
and instead returns
.meta form
itself as an object. For example, if
.meta form
is a symbol, then
.meta form
is not evaluated to the symbol's value; rather
the symbol itself is returned.
Note: the quote syntax
.cblk
.meti >> ' <form>
.cble
is translated to
.cblk
.meti (quote << form ).
.cble
.TP* Example:
.cblk
(quote a) ;; yields a
(quote (+ 2 2)) ;; yields (+ 2 2), not 4.
.cble
.coNP Macro @ qquote
.synb
.mets (qquote << form )
.syne
.desc
The
.code qquote
(quasi-quote) macro operator implements a notation for convenient
list construction. If
.meta form
is an atom, or a list structure which
does not contain any
.code unquote
or
.code splice
operators, then
.cblk
.meti (qquote << form )
.cble
is equivalent to
.cblk
.meti (qquote << form ).
.cble
If
.metn form ,
however, is a list structure which contains
.code unquote
or
.code splice
operators, then the substitutions implied by those operators are performed
on
.metn form ,
and the
.code qquote
operator returns the resulting structure.
Note: how the qquote operator actually works is that it is compiled into
code. It becomes a Lisp expression which, when evaluated, computes the
resulting structure.
A
.code qquote
can contain another
.codn qquote .
If an
.code unquote
or
.code splice
operator occurs
within a nested
.codn qquote ,
it belongs to that
.codn qquote ,
and not to the outer one.
However, an unquote operator which occurs inside another one belongs one level
higher. For instance in
.cblk
(qquote (qquote (unquote (unquote x))))
.cble
the leftmost
.code qquote
belongs with the rightmost unquote, and the inner
.code qquote
and
.code unquote
belong together. When the outer
.code qquote
is evaluated,
it will insert the value of
.codn x ,
resulting in the object
.codn (qquote (unquote [value-of-x])) .
If this resulting qquote value is evaluated again as Lisp syntax, then it will
yield
.codn [value-of-value-of-x] ,
the value of
.code [value-of-x]
when treated as a Lisp expresion and evaluated.
.TP* Examples:
.cblk
(qquote a) -> a
(qquote (a b c)) -> (a b c)
(qquote (1 2 3 (unquote (+ 2 2)) (+ 2 3))) -> (1 2 3 4 (+ 2 3))
(qquote (unquote (+ 2 2))) -> 4
.cble
In the second-to-last example, the
.code 1 2 3
and the
.code (+ 2 3)
are quoted verbatim.
Whereas the
.code (unquote (+ 2 2))
operator caused the evaluation of
.code (+ 2 2)
and the substitution of the resulting value.
The last example shows that
.meta form
can itself (the entire argument of
.codn qquote )
can be an unquote operator.
However, note:
.code (quote (splice form))
is not valid.
Note: a way to understand the nesting behavior is a via a possible model of
quasi-quote expansion which recursively compiles any nested quasi quotes first,
and then treats the result of their expansion. For instance, in the processing
of
.cblk
(qquote (qquote (unquote (unquote x))))
.cble
the
.code qquote
operator first encounters the
embedded
.code (qquote ...)
and compiles it to code. During that recursive
compilation, the syntax
.code (unquote (unquote x))
is encountered. The inner quote
processes the outer unquote which belongs to it, and the inner
.code (unquote x)
becomes material that is embedded verbatim in the compilation, which will then
be found when the recursion pops back to the outer quasiquote, which will
then traverse the result of the inner compilation and find the
.codn (unquote x) .
.TP* "Dialect note:"
In Lisp dialects which have a published quasiquoting operator syntax, there is
the expectation that the quasiquote read syntax corresponds to it. That is to
say, that for instance the read syntax
.code ^(a b ,c)
is expected translated to
.codn (qquote b (unquote c)) .
In \*(TL, this is not true! Although
.code ^(b b ,c)
is translated to a
quasiquoting macro, it is an internal one, not based on the public
.codn qquote ,
.code unquote
and
.code splice
symbols being documented here.
This idea exists for hygiene. The quasiquote read syntax is not confused
by the presence of the symbols
.codn qquote ,
.code unquote
or
.code splice
in the template, since it doesn't treat them specially.
This also allows programmers to use the quasiquote read syntax to construct
quasiquote macros. For instance
.cblk
^(qquote (unquote ,x)) ;; does not mean ^^,x
.cble
To the quasiquote reader, the
.code qquote
and
.code unquote
symbols mean nothing special,
and so this syntax simply means that if the value of
.code x
is
.codn foo ,
the result will be
.codn (qquote (unquote foo)) .
The form's expansion is actually this:
.cblk
(sys:qquote (qquote (unquote (sys:unquote x))))
.cble
the
.code sys:qquote
macro recognizes
.code sys:unquote
embedded in the form, and
the other symbols not in the
.code sys:
package are just static template material.
The
.code sys:quote
macro and its associated
.code sys:unquote
and
.code sys:splice
operators work exactly like their ordinary counterparts. So in effect, \*(TX has
two nearly identical, independent quasi-quote implementations, one of which is
tied to the read syntax, and one of which isn't. This is useful for writing
quasiquotes which write quasiquotes.
.coNP Operator @ unquote
.synb
.mets (qquote (... (unquote << form ) ...))
.mets (qquote (unquote << form ))
.syne
.desc
The
.code unquote
operator is not an operator
.I per
.IR se .
The
.code unquote
symbol has no
binding in the global environment. It is a special syntax that is recognized
within a
.code qquote
form, to indicate forms within the quasiquote which are to be
evaluated and inserted into the resulting structure.
The syntax
.cblk
.meti (qquote (unquote << form ))
.cblk
is equivalent to
.metn form :
the
.code qquote
and
.code unquote
"cancel out".
.coNP Operator splice
.synb
.mets (qquote (... (splice << form ) ...))
.syne
.desc
The
.code splice
operator is not an operator
.I per
.IR se .
The
.code splice
symbol has no
binding in the global environment. It is a special syntax that is recognized
within a
.code qquote
form, to indicate forms within the quasiquote which are to be
evaluated and inserted into the resulting structure.
The syntax
.cblk
.meti (qquote (splice << form ))
.cble
is not permitted and raises an exception if evaluated. The
.code splice
syntax must occur within a list, and not in the dotted position.
The
.code splice
form differs from unquote in that
.cblk
.meti (splice << form )
.cble
requires that
.meta form
must evaluate to a list. That list is
integrated into the surrounding list.
.SS* Macros
\*(TL supports structural macros. \*(TX's model of macroexpansion is that
\*(TL code is processed in two phases: the expansion phase and the
evaluation phase. The expansion phase is invoked on Lisp code early during the
processing of source code. For instance when a \*(TX file containing a
.code @(do ...)
directive
is loaded, expansion of the Lisp forms are its arguments takes place during the
parsing of the entire source file, and is complete before any of the code in
that file is executed. If the
.code @(do ...)
form is later executed,
the expanded forms are then evaluated.
\*(TL also supports symbol macros, which are symbolic forms that stand
for forms, with which they are replaced at macro expansion time.
When Lisp data is processed as code by the
.code eval
function, it is first expanded,
and so processed in its entirety in the expansion phase. Then it is processed
in the evaluation phase.
.coNP Macro parameter lists
\*(TX macros support destructuring, similarly to Common Lisp macros.
This means that macro parameter lists are like function argument lists,
but support nesting. A macro parameter list can specify a nested parameter
list in every place where a function argument list allows only a parameter
name. For instance, consider this macro parameter list:
.cblk
((a (b c)) : (c frm) ((d e) frm2 de-p) . g)
.cble
The top level of this list has four elements: the mandatory parameter
.codn (a (b c)) ,
the optional parameter
.code c
(with default init form
.codn frm ),
the optional parameter
.code (d e)
(with default init form
.code frm2
and presence-indicating variable
.codn de-p ),
and the dot-position parameter
.code g
which captures trailing arguments.
Note that some of the parameters are compounds:
.code (a (b c))
and
.codn (d e) .
These compounds express nested macro parameter lists.
Macro parameter lists match a similar tree structure to their own.
For instance a mandatory parameter
.code (a (b c))
matches a structure like
.codn (1 (2 3)) ,
such that the parameters
.codn a ,
.code b
and
.code c
will end up bound
to
.codn 1 ,
.code 2
and
.codn 3 ,
respectively.
The binding strictness is relaxed for optional parameters. If
.code (a (b c))
is optional, and the argument is, say,
.codn (1) ,
then
.code a
gets
.codn 1 ,
and
.code b
and
.code c
receive
.codn nil .
Macro parameter lists also support two special keywords, namely
.code :env
and
.codn :whole .
The parameter list
.code (:whole x :env y)
will bind parameter
.code x
to the entire
macro form, and bind parameter
.code y
to the macro environment.
The
.code :whole
and
.code :env
notation can occur anywhere in a macro parameter list.
.coNP Operator @ macro-time
.synb
.mets (macro-time << form *)
.syne
.desc
The
.code macro-time
operator has a syntax similar to the
.code progn
operator. Each
.meta form
is evaluated from left to right, and the resulting value is that of the last
form.
The special behavior of
.code macro-time
is that the evaluation takes place during
the expansion phase, rather than during the evaluation phase.
During the expansion phase, all
.code macro-time
expressions which occur in a context
that calls for evaluation are evaluated, and replaced by their quoted values.
For instance
.code (macro-time (list 1 2 3))
evaluates
.code (list 1 2 3)
to the object
.code (1 2 3)
and the entire
.code macro-time
form is replaced by that value, quoted:
.codn '(1 2 3) .
If the form is evaluated again at evaluation-time, the resulting value will be
that of the quote, in this case
.codn (1 2 3) .
.code macro-time
forms do not see the surrounding lexical environment; the see only
global function and variable bindings and macros.
Note 1:
.code macro-time
is intended for defining helper functions and variables that
are used by macros. A macro cannot "see" a
.code defun
function or
.code defvar
variable
because
.code defun
and
.code defvar
forms are evaluated at evaluation time, which occurs
after expansion time. The macro-time operator hoists the evaluation of these
forms to macro-expansion time.
Note 2:
.code defmacro
forms are implicitly in macro-time; they do not have to
be wrapped with the
.code macro-time
operator. The
.code macro-time
operator doesn't have
to be used in programs whose macros do not make references to variables
or functions.
.coNP Operator @ defmacro
.synb
.mets (defmacro < name <> ( param * [: << opt-param * ] [. < rest-param ])
.mets \ \ << body-form *)
.syne
.desc
The
.code defmacro
operator is evaluated at expansion time. It defines a
macro-expander function under the name
.metn name ,
effectively creating a new operator.
Note that the parameter list is a macro parameter list, and not a
function parameter list. This means that each
.meta param
and
.meta opt-param
can be not only a symbol, but it can itself be a parameter list.
The corresponding argument is then treated as a structure which
matches that parameter list. This nesting of parameter lists
can be carried to an arbitrary depth.
A macro is called like any other operator, and resembles a function. Unlike in
a function call, the macro receives the argument expressions themselves, rather
than their values. Therefore it operates on syntax rather than on values.
Also, unlike a function call, a macro call occurs in the expansion phase,
rather than the evaluation phase.
The return value of the macro is the macro expansion. It is substituted in
place of the entire macro call form. That form is then expanded again;
it may itself be another macro call, or contain more macro calls.
.TP* Examples:
.cblk
;; A macro to swap two places, such as variables.
;; (swap x y) will now exchange the contents
;; of x and y.
(defmacro swap (a b)
(let ((temp (gensym)))
^(let ((,temp ,a))
(set ,a ,b)
(set ,b ,temp))))
;; dolist macro similar to Common Lisp's:
;;
;; The following will print 1, 2 and 3
;; on separate lines:
;; and return 42.
;;
;; (dolist (x '(1 2 3) 42)
;; (format t "~s\en"))
(defmacro dolist ((var list : result) . body)
(let ((i (my-gensym)))
^(for ((i ,list)) (i ,result) ((set i (cdr i)))
(let ((,var (car i)))
,*body))))
.cble
.coNP Operator @ macrolet
.synb
.mets (macrolet >> ({( name < macro-style-params << macro-body-form *)}*)
.mets \ \ << body-form *)
.syne
.desc
The
.code macrolet
binding operator extends the macro-time lexical environment
by making zero or more new local macros visible.
The
.code macrolet
symbol is followed by a list of macro definitions.
Each definition is a form which begins with a
.metn name ,
followed by
.meta macro-style-params
which is a macro parameter list, and zero or more
.metn macro-body-form s.
These macro definitions are similar
to those globally defined by the
.code defmacro
operator, except that they
are in a local environment.
The macro definitions are followed by optional
.metn body-forms .
The macros specified in the definitions are visible to these
forms.
Forms inside the macro definitions such as the
.metn macro-body-form s,
and initializer forms appearing in the
.meta macro-style-params
are subject
to macro-expansion in a scope in which none of the new macros being
defined are yet visible. Once the macro definitions are themselves
macro-expanded, they are placed into a new macro environment, which
is then used for macro expanding the
.metn body-form s.
A
.code macrolet
form is fully processed in the expansion phase of a form, and is
effectively replaced by
.code progn
form which contains expanded versions of
.metn body-form s.
This expanded structure shows no evidence that any
macrolet forms ever existed in it. Therefore, it is impossible for the code
evaluated in the bodies and parameter lists of
.code macrolet
macros to have any visibility to any surrounding lexical variable bindings,
which are only instantiated in the evaluation phase, after expansion is done
and macros no longer exist.
.coNP Function @ macro-form-p
.synb
.mets (macro-form-p < obj <> [ env ])
.syne
.desc
The
.code macro-form-p
function returns
.code t
if
.meta obj
represents the syntax of
a form which is a macro form: either a compound macro or a symbol macro.
Otherwise it returns
.codn nil .
A macro form is one that will transform under
.code macroexpand-1
or
.codn macroexpand ;
an object which isn't a macro form will not undergo expansion.
The optional
.meta env
parameter is a macroexpansion environment.
A macroexpansion environment is passed down to macros and can be received
via their special
.code :env
parameter.
.meta env
is used by
.code macro-form-p
to determine whether
.meta obj
is a macro in a lexical macro environment.
If
.meta env
is not specified or is
.codn nil ,
then
.code macro-form-p
only recognizes global macros.
.TP* Example:
.cblk
;; macro which translates to 'yes if its
;; argument is a macro from, or otherwise
;; transforms to the form 'no.
(defmacro ismacro (:env menv form)
(if (macro-form-p form menv)
''yes ''no))
(macrolet ((local ()))
(ismacro (local))) ;; yields yes
(ismacro (local)) ;; yields no
(ismacro (ismacro foo)) ;; yields yes
.cble
During macro expansion, the global macro
.code ismacro
is handed the macro-expansion environment
via
.codn :env menv .
When the macro is invoked within the macrolet,
this environment includes the macro-time lexical scope in which the
.code local
macro is defined. So when global checks whether the argument form
.code (local)
is a macro, the conclusion is yes: the (local) form is a macro
call in that environment:
.code macro-form-p
yields
.codn t .
When
.code (global (local))
is invoked outside of the macrolet, no local macro is visible is
there, and so
.code macro-form-p
yields
.codn nil .
.coNP Functions @ macroexpand-1 and @ macroexpand
.synb
.mets (macroexpand-1 < obj <> [ env ])
.mets (macroexpand < obj <> [ env ])
.syne
.desc
If
.meta obj
is a macro form (an object for which
.code macro-form-p
returns
.codn t ),
these functions expand the macro form and return the expanded form.
Otherwise, they return
.metn obj .
.code macroexpand-1
performs a single expansion, expanding just the macro
that is referenced by the symbol in the first position of
.metn obj ,
and returns the expansion. That expansion may itself be a macro form.
.code macroexpand
performs an expansion similar to
.codn macroexpand-1 .
If the result is
a macro form, then it expands that form, and keeps repeating this process
until the expansion yields a non-macro-form. That non-macro-form is then
returned.
The optional
.meta env
parameter is a macroexpansion environment.
A macroexpansion environment is passed down to macros and can be received
via their special
.code :env
parameter. The environment they receive is their
lexically apparent macro-time environment in which local macros may be
visible. A macro can use this environment to "manually" expand some
form in the context of that environment.
.TP* Example:
.cblk
;; (foo x) expands x, and if x begins with a number,
;; it removes the number and returns the resulting
;; form. Otherwise, it returns the entire form.
(defmacro rem-num (:env menv some-form)
(let ((expanded (macroexpand some-form menv)))
(if (numberp (car expanded))
(cdr expanded)
some-form)))
;; The following yields (42 a).
(macrolet ((foo () '(1 list 42))
(bar () '(list 'a)))
(list (rem-num (foo)) (rem-num (bar)))))
.cble
The
.code rem-num
macro is able to expand the
.code (foo)
and
.code (bar)
forms it receives as
the
.code some-form
argument, even though these forms use local macro that are only
visible in their local scope. This is thanks to the macro
environment passed to
.codn rem-num .
It is correctly able to work with the
expansions
.code (1 list 42)
and
.code (list 'a)
to produce
.code (list 42)
and
.code (list 'a)
which evaluate to
.code 42
and
.code a
respectively.
.coNP Functions @ lexical-var-p and @ lexical-fun-p
.synb
.mets (lexical-var-p < env << form)
.mets (lexical-fun-p < env << form)
.syne
.desc
These two functions are useful to macro writers. They are intended
to be called from the bodies of macro expanders, such as the bodies of
.code defmacro
or
.code macrolet
forms. The
.meta env
argument is a macro-time environment, which is available to macros
via the special
.code :env
parameter. Using these functions, a macro can enquire whether
a given
.meta form
is a symbol which has a variable binding or a function binding
in the lexical environment.
This information is known during macro expansion. The macro expander
recognizes lexical function and variable bindings, because these
bindings can shadow macros.
.TP* Example:
.cblk
;;
;; this macro replaces itself with :lexical-var if its
;; argument is a lexical variable, :lexical-fun if
;; its argument is a lexical function, or with
;; :not-lex-fun-var if neither is the case.
;;
(defmacro classify (sym :env e)
(cond
((lexical-var-p e expr) :lexical-var)
((lexical-fun-p e expr) :lexical-fun)
(t :not-lex-fun-var)))
;;
;; This returns:
;;
;; (:lexical-var :not-lex-fun-var :lexical-fun)
;;
(let ((x 1) (y 2))
(symacrolet ((y x))
(flet ((f () (+ 2 2)))
(list (classify x) (classify y) (classify z)))))
.cble
.TP* Note:
These functions do not call
.code macroexpand
on the form. In most cases, it is necessary for the macro writers
to do so. Not that in the above example, symbol
.code y
is classified as neither a lexical function nor variable.
However, it can be macro-expanded to
.code x
which is a lexical variable.
.coNP Function @ lexical-lisp1-binding
.synb
.mets (lexical-lisp1-binding < env << symbol)
.syne
.desc
The
.code lexical-lisp1-binding
function inspects the macro-time environment
.meta env
to determine what kind of binding, if any, does
.meta symbol
have in that environment, from a Lisp-1 perspective.
That is to say, it considers function bindings, variable bindings
and symbol macro bindings to be in a single name space and finds
the innermost binding of one of these types for
.metn symbol .
If such a binding is found, then the function returns one of
the three keyword symbols
.codn :var ,
.codn :fun ,
or
.codn :symacro .
If no such lexical binding is found, then the function
returns
.codn nil .
Note that a
.code nil
return doesn't mean that the symbol doesn't have a lexical binding. It could
have an operator macro lexical binding (a macro binding in the function
namespace established by
.codn macrolet ).
.coNP Operator @ defsymacro
.synb
.mets (defsymacro < sym << form )
.syne
.desc
A
.code defsymacro
form introduces a symbol macro: a symbol macro is a parameterless
substitution keyed to a symbol. In contexts where a symbol macro
definition of
.meta sym
is visible, if the form
.meta sym
appears such that its evaluation is called for, it is subject
to replacement by
.metn form .
After replacement takes place,
.meta form
itself is then processed for further replacement of macros and
symbol macros.
Symbol macros are also recognized in contexts
where
.meta sym
denotes a place which is the target of an assignment operation
like
.code set
and similar.
A
.code defsymacro
form is implicitly executed at expansion time, and thus need
not be wrapped in a
.code macro-time
form, just like
.codn defmacro .
Note: if a symbol macro expands to itself directly, expansion stops. However,
if a symbol macro expands to itself through a chain of expansions,
an infinite expansion time loop results.
.coNP Operator @ symacrolet
.synb
.mets (symacrolet >> ({( sym << form )}*) << body-form *)
.syne
.desc
The
.code symacrolet
operator binds local, lexically scoped macros that are
similar to the global symbol macros introduced by
.codn defsymacro .
Each
.meta sym
in the bindings list is bound to its corresponding form, creating a
new extension of the expansion-time lexical macro environment.
Each
.meta body-form
is subsequently macro-expanded in this new environment
in which the new symbol macros are visible.
Note: ordinary lexical bindings such as those introduced by let or by
function parameters lists shadow symbol macros. If a symbol
.code x
is bound by nested instances of
.code macrolet
and a
.codn let ,
then the scope enclosed by both
constructs will see whichever of the two bindings is more inner,
even though the bindings are active in completely separate phases of
processing.
From the perspective of the arguments of a
.code dwim
form, lexical function bindings also shadow symbol macros.
This is consistent with the Lisp-1-style name resolution which
applies inside a
.code dwim
form. Of course, lexical operator macros do not shadow
symbol macros under any circumstances.
.coNP Operator @ tree-bind
.synb
.mets (tree-bind < macro-style-params < expr << form *)
.syne
.desc
The
.code tree-bind
operator evaluates
.codn expr ,
and then uses the
resulting value as a counterpart to a macro-style parameter list.
If the value has a tree structure which matches the parameters,
then those parameters are established as bindings, and the
.metn form s,
if any, are evaluated in the scope of those bindings. The value
of the last
.meta form
is returned. If there are no forms,
.code nil
is returned.
Note: this operator throws an exception if there is a
structural mismatch between the parameters and the value of
.codn expr .
One way to avoid this exception is to use
.codn tree-case .
.coNP Operator @ tree-case
.synb
.mets (tree-case < expr >> {( macro-style-params << form *)}*)
.syne
.desc
The
.code tree-case
operator evaluates
.meta expr
and matches it against a succession
of zero or more cases. Each case defines a pattern match, expressed as a macro
style parameter list
.metn macro-style-params .
If the object produced by
.meta expr
matches
.metn macro-style-params ,
then the parameters are bound, becoming local variables, and the
.metn form s,
if any, are evaluated in order in the environment in which those variables are
visible. If there are forms, the value of the last
.meta form
becomes the result
value of the case, otherwise the result value of the case is nil.
If the result value of a case is the object
.code :
(the colon symbol), then processing continues with the next case. Otherwise the
evaluation of
.code tree-case
terminates, returning the result value.
If the value of
.meta expr
does not match the
.meta macro-style-params
parameter list of a case, processing continues with the next case.
If no cases match, then
.code tree-case
terminates, returning
.codn nil .
.TP* Example:
.cblk
;; reverse function implemented using tree-case
(defun tb-reverse (obj)
(tree-case obj
(() ()) ;; the empty list is just returned
((a) obj) ;; one-element list returned (unnecessary case)
((a . b) ^(,*(tb-reverse b) ,a)) ;; car/cdr recursion
(a a))) ;; atom is just returned
.cble
Note that in this example, the atom case is placed last, because an
argument list which consists of a symbol is a "catch all" match
that matches any object. We know that it matches an atom, because
the previous
.code (a . b)
case matches conses. In general, the order of the cases in
.code tree-case
is important: even more so than the order of cases in a
.code cond
or
.codn caseql .
.coNP Macro @ tb
.synb
.mets (tb < macro-style-params << form *)
.syne
.desc
The
.code tb
macro is similar to the
.code lambda
operator but its argument binding is based on a macro-style parameter list.
The name is an abbreviation of
.codn tree-bind .
A
.code tb
form evaluates to a function which takes a variable number of
arguments.
When that function is called, those arguments are taken as a list object which
is matched against
.meta macro-style-params
as if by
.metn tree-bind .
If the match is successful, then the parameters are bound to the
corresponding elements from the argument structure and each successive
.meta form
is evaluated an environment in which those bindings are visible.
The value of the last
.meta form
is the return value of the function. If there are no forms,
the function's return value is
.codn nil .
The following equivalence holds, where
.code args
should be understood to be a globally unique symbol:
.cblk
(tb pattern body ...) <--> (lambda (. args)
(tree-bind pattern args body ...))
.cble
.coNP Macro @ tc
.synb
.mets (tc >> {( macro-style-params << form *)}*)
.syne
.desc
The
.code tc
macro produces an anonymous function whose behavior is closely
based on the
.code tree-case
operator. Its name is an abbreviation of
.codn tree-case .
The anonymous function takes a variable number of arguments.
Its argument list is taken to be the value macro is tested
against the multiple pattern clauses of an implicit
.codn tree-bind .
The return value of the function is that of the implied
.codn tree-bind .
The following equivalence holds, where
.code args
should be understood to be a globally unique symbol:
.cblk
(tc clause1 clause2 ...) <--> (lambda (. args)
(tree-bind args
clause1 clause2 ...))
.cble
.SS* Debugging Functions
.coNP Functions source-loc and source-loc-str
.synb
.mets (source-loc << form )
.mets (source-loc-str << form )
.syne
.desc
These functions map an expression in a \*(TX program to the file name and
line number of the source code where that form came from.
The
.code source-loc
function returns the raw information as a cons cell
whose
.cod3 car / cdr
consist of the line number, and file name.
The
.code source-loc-str
function formats the information as a string.
If
.meta form
is not a piece of the program source code that was constructed by the
\*(TX parser, then
.code source-loc
returns
.codn nil ,
and
.code source-loc-str
returns a string whose text says that source location is not available.
.coNP Function @ rlcp
.synb
.mets (rlcp < dest-form << source-form )
.syne
.desc
The
.code rlcp
function copies the source code location info ("rl" means "read location")
from the
.meta source-form
object to the
.meta dest-form
object. These objects
are pieces of list-based syntax.
Note: the function is intended to be used in macros. If a macro transforms
.meta source-form
to
.metn dest-form ,
this function can be used to propagate the
source code location info also, so that when the \*(TL evaluator
encounters errors in transformed code, it can give diagnostics which refer
to the original untransformed source code.
.SS* Profiling
.coNP Operator @ prof
.synb
.mets (prof << form *)
.syne
.desc
The
.code prof
operator evaluates the enclosed forms from left to right similarly
to
.codn progn ,
while determining the memory allocation requests and time
consumed by the evaluation of the forms.
If there are no forms, the prof operator measures the smallest measurable
operation of evaluating nothing and producing
.codn nil .
If the evaluation terminates normally (not abruptly by a non-local
control transfer), then
.code prof
yields a list consisting of:
.cblk
.mets >> ( value < malloc-bytes < gc-bytes << milliseconds )
.cble
where
.meta value
is the value returned by the rightmost
.metn form ,
or
.code nil
if there are no forms,
.meta malloc-bytes
is the total number of bytes of all memroy allocation
requests (or at least those known to the \*(TX runtime, such as those of all
internal objects),
.meta gc-bytes
is the total number of bytes drawn from the
garbage-collected heaps, and
.meta milliseconds
is the total processor time
consumed over the execution of those forms.
Notes:
The bytes allocated by the garbage collector from the C function
.code malloc
to create
heap areas are not counted as
.metn malloc-bytes .
.meta malloc-bytes
includes storage
such as the space used for dynamic strings, vectors and bignums (in addition to
their gc-heap allocated nodes), and the various structures used by the
.code cobj
type objects such as streams and hashes. Objects in external libraries that use
uninstrumented allocators are not counted: for instance the C
.code FILE *
streams.
.coNP Macro @ pprof
.synb
.mets (pprof << form *)
.syne
.desc
The
.code pprof
(pretty-printing
.codn prof )
macro is similar to
.codn progn .
It evaluates
.metn form s,
and returns the rightmost one, or
.code nil
if there are no forms.
Over the evaluation of
.metn form s,
it counts memory allocations, and measures
CPU time. If
.metn form s
terminate normally, then just prior to returning,
.code pprof
prints these statistics in a concise report on the
.codn *std-output* .
The
.code pprof
macro relies on the
.code prof
operator.
.SS* Garbage Collection
.coNP Function @ sys:gc
.synb
(sys:gc)
.syne
.desc
The
.code gc
function triggers garbage collection. Garbage collection means
that unreachable objects are identified and reclaimed, so that their
storage can be re-used.
The function returns
.code nil
if garbage collection is disabled (and consequently nothing is done), otherwise
.codn t .
.coNP Function @ sys:gc-set-delta
.synb
.mets (sys:gc-set-delta << bytes )
.syne
.desc
The
.code gc-set-delta
function sets the GC delta parameter.
Note: This function may disappear in a future release of \*(TX or suffer
a backward-incompatible change in its syntax or behavior.
When the amount of new dynamic memory allocated since the last garbage
collection equals or exceeds the GC delta, a garbage collection pass is
triggered. From that point, a new delta begins to be accumulated.
Dynamic memory is used for allocating heaps of small garbage-collected objects
such as cons cells, as well as the satellite data attached to some objects:
like the storage arrays of vectors, strings or bignum integers. Most garbage
collector behaviors are based on counting objects in the heaps.
Sometimes a program works with a small number of objects which are very large,
frequently allocating new, large objects and turning old ones into garbage.
For instance a single large integer could be many megabytes long. In such a
situation, a small number of heap objects therefore control a large amount of
memory. This requires garbage collection to be triggered much more often than
when working with small objects, such as conses, to prevent runaway allocation
of memory. It is for this reason that the garbage collector uses the GC delta.
There is a default GC delta of 64 megabytes. This may be overridden in
special builds of \*(TX for small systems.
.coNP Function @ finalize
.synb
.mets (finalize < object << function )
.syne
.desc
The
.code finalize
function registers
.meta function
to be invoked in the situation when
.meta object
is identified by the garbage collector as unreachable.
This function is called a finalizer.
If and when this situation occurs, the finalizer
.meta function
will be called with
.meta object
as its only argument.
Finalizers are called in the same order in which they are registered:
newer registrations are called after older registrations.
If
.meta object
is registered multiple times by multiple calls to
.codn finalize ,
then if those finalizers are called, they are all called, in the order
of registration.
After a finalization call takes place, its registration is removed;
.meta object
and
.meta function
are no longer associated. However, neither
.meta object
nor
.meta function
are reclaimed immediately; they are treated as if they were reachable objects
until at least the next garbage collection pass.
It is therefore safe for
.meta function
to store somewhere a persistent reference to
.meta object
or to itself, thereby reinstanting these objects as reachable.
.meta function
is itself permitted to call
.code finalize
to register the original
.code object
or any other object for finalization. Such registrations made during
finalization execution are not eligible for the current phase of finalization
processing; they will be processed in a later garbage collection pass.
.SS* Modularization
.coNP Special variable @ *self-path*
.desc
This variable holds the invocation path name of the \*(TX program.
.coNP Special variable @ stdlib
The
.code stdlib
variable expands to the directory where the \*(TX standard library
is installed. It can be referenced in
.code @(load)
and
.code @(include)
directives via quasiliteral substitution, as in, for example:
.cblk
@(include `@stdlib/extract`)
.cble
.SS* Debugger
\*(TX has a simple, crude, built-in debugger. The debugger is invoked by adding
the
.code -d
command line option to an invocation of \*(TX.
In this debugger it is possible to step through code, set breakpoints,
and examine the variable binding environment.
Prior to executing any code, the debugger waits at the
.code txr>
prompt, allowing for the opportunity to set breakpoints.
Help can be obtained with the
.code h
or
.code ?
command.
Whenever the program stops at the debugger, it prints the Lisp-ified
piece of syntax tree that is about to be interpreted.
It also shows the context of the input being matched.
The s command can be used to step into a form; n to step over.
Sometimes the behavior seems counter-intuitive. For instance stepping
over a
.code @(next)
directive actually means skipping everything which follows
it. This is because the query material after a
.code @(next)
is actually child
nodes in the abstract syntax tree node of the
.code next
directive, whereas the surface syntax appears flat.
.coNP Sample Session
Here is an example of the debugger being applied to a web scraping program
which connects to a US NAVY clock server to retrieve a dynamically-generated
web page, from which the current time is extracted, in various time zones.
The handling of the web request is done by the wget command; the
\*(TX query opens a wget command as and scans the body of the HTTP response
containing
HTML. This is the code, saved in a file called navytime.txr:
.cblk
@(next `!wget -c http://tycho.usno.navy.mil/cgi-bin/timer.pl -O - 2> /dev/null`)
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2 Final"//EN>
<html>
<body>
<TITLE>What time is it?</TITLE>
<H2> US Naval Observatory Master Clock Time</H2> <H3><PRE>
@(collect :vars (MO DD HH MM SS (PM " ") TZ TZNAME))
<BR>@MO. @DD, @HH:@MM:@SS @(maybe)@{PM /PM/} @(end)@TZ@/\et+/@TZNAME
@ (until)
</PRE>@/.*/
@(end)
</PRE></H3><P><A HREF="http://www.usno.navy.mil"> US Naval Observatory</A>
</body></html>
@(output)
@ (repeat)
@MO-@DD @HH:@MM:@SS @PM @TZ
@ (end)
@(end)
.cble
This is the debug session:
.cblk
$ txr -d navytime.txr
stopped at line 1 of navytime.txr
form: (next (sys:quasi "!wget -c http://tycho.usno.navy.mil/cgi-bin/timer.pl -O - 2> /dev/null"))
depth: 1
data (nil):
nil
.cble
The user types
.code s
to step into the
.code (next ...)
form.
.cblk
txr> s
stopped at line 2 of navytime.txr
form: (sys:text "<!DOCTYPE" (#<sys:regex: 95e4590> 1+ #\espace) "HTML" (#<sys:regex: 95e4618> 1+ #\espace) "PUBLIC" (#<sys:regex: 95e46a8> 1+ #\espace) "\e"-//W3C//DTD" (#<sys:regex: 95e4750> 1+ #\espace) "HTML" (#<sys:regex: 95e47d8> 1+ #\espace) "3.2" (#<sys:regex: 95e4860> 1+ #\espace) "Final\e"//EN>")
depth: 2
data (1):
"<!DOCTYPE HTML PUBLIC \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
txr> s
.cble
The current form now is a
.code sys:text
form which is an internal representation of
a block of horizontal material. The pattern matching is in vertical mode at
this point, and so the line of data is printed without an indication of
character position.
.cblk
stopped at line 2 of navytime.txr
form: (sys:text "<!DOCTYPE" (#<sys:regex: 95e4590> 1+ #\espace) "HTML" (#<sys:regex: 95e4618> 1+ #\espace) "PUBLIC" (#<sys:regex: 95e46a8> 1+ #\espace) "\e"-//W3C//DTD" (#<sys:regex: 95e4750> 1+ #\espace) "HTML" (#<sys:regex: 95e47d8> 1+ #\espace) "3.2" (#<sys:regex: 95e4860> 1+ #\espace) "Final\e"//EN>")
depth: 3
data (1:0):
"" . "<!DOCTYPE HTML PUBLIC \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
.cble
The user types
.code s
to step in.
.cblk
txr> s
stopped at line 2 of navytime.txr
form: "<!DOCTYPE"
depth: 4
data (1:0):
"" . "<!DOCTYPE HTML PUBLIC \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
.cble
Now, the form about to be processed is the first item of the
.codn (sys:text ...) ,
the string
.strn <!DOCTYPE .
The input is shown broken into two quoted strings with a dot in between.
The dot indicates the current position. The left string is empty, meaning
that this is the leftmost position. The programmer steps:
.cblk
txr> s
stopped at line 2 of navytime.txr
form: (#<sys:regex: 95e4590> 1+ #\espace)
depth: 4
data (1:9):
"<!DOCTYPE" . " HTML PUBLIC \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
.cble
Control has now passed to the second element of the
.codn (sys:text ...) ,
a regular expression which matches one or more spaces, generated by
a single space in the source code according to the language rules.
The input context shows that
.str <!DOCTYPE
was matched in the input, and the position moved past it.
.cblk
txr> s
stopped at line 2 of navytime.txr
form: "HTML"
depth: 4
data (1:10):
"<!DOCTYPE " . "HTML PUBLIC \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
.cble
Now, the regular expression has matched and moved the position past
the space; the facing input is now
.strn "HTML ..." .
The programmer then repeats the
.code s
command by hitting Enter.
.cblk
txr>
stopped at line 2 of navytime.txr
form: (#<sys:regex: 95e4618> 1+ #\espace)
depth: 4
data (1:14):
"<!DOCTYPE HTML" . " PUBLIC \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
txr>
stopped at line 2 of navytime.txr
form: "PUBLIC"
depth: 4
data (1:15):
"<!DOCTYPE HTML " . "PUBLIC \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
txr>
stopped at line 2 of navytime.txr
form: (#<sys:regex: 95e46a8> 1+ #\espace)
depth: 4
data (1:21):
"<!DOCTYPE HTML PUBLIC" . " \e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
txr>
stopped at line 2 of navytime.txr
form: "\e"-//W3C//DTD"
depth: 4
data (1:22):
"<!DOCTYPE HTML PUBLIC " . "\e"-//W3C//DTD HTML 3.2 Final\e"//EN>"
txr>
stopped at line 2 of navytime.txr
form: (#<sys:regex: 95e4750> 1+ #\espace)
depth: 4
data (1:34):
"<!DOCTYPE HTML PUBLIC \e"-//W3C//DTD" . " HTML 3.2 Final\e"//EN>"
.cble
It is not evident from the session transcript, but during interactive use,
the input context appears to be animated. Whenever the programmer hits
Enter, the new context is printed and the dot appears to advance.
Eventually the programmer becomes bored and place a breakpoint on line 15,
where the
.code @(output)
block begins, and invokes the
.code c
command to continue the execution:
.cblk
txr> b 15
txr> c
stopped at line 15 of navytime.txr
form: (output (((repeat nil (((sys:var MO nil nil) "-" (sys:var DD nil nil) " " (sys:var HH nil nil) ":" (sys:var MM nil nil) ":" (sys:var SS nil nil) " " (sys:var PM nil nil) " " (sys:var TZ nil nil))) nil nil nil nil nil nil))))
depth: 2
data (16):
""
.cble
The programmer issues a
.code v
command to take a look at the variable bindings,
which indicate that the
.code @(collect)
has produced some lists:
.cblk
txr> v
bindings:
0: ((PM " " "PM" "PM" "PM" "PM" "PM" "PM") (TZNAME "Universal Time" "Eastern Time" "Central Time" "Mountain Time" "Pacific Time" "Alaska Time" "Hawaii-Aleutian Time") (TZ "UTC" "EDT" "CDT" "MDT" "PDT" "AKDT" "HAST") (SS "35" "35" "35" "35" "35" "35" "35") (MM "32" "32" "32" "32" "32" "32" "32") (HH "23" "07" "06" "05" "04" "03" "01") (DD "30" "30" "30" "30" "30" "30" "30") (MO "Mar" "Mar" "Mar" "Mar" "Mar" "Mar" "Mar"))
.cble
Then a continue command, which finishes the program, whose output appears:
.cblk
txr> c
Mar-30 23:22:52 UTC
Mar-30 07:22:52 PM EDT
Mar-30 06:22:52 PM CDT
Mar-30 05:22:52 PM MDT
Mar-30 04:22:52 PM PDT
Mar-30 03:22:52 PM AKDT
Mar-30 01:22:52 PM HAST
.cble
.SS* Compatibility
New \*(TX versions are usually intended to be backward-compatible with prior
releases in the sense that documented features will continue to work in
the same way. Due to new features, new versions of \*(TX will supply new
behaviors where old versions of \*(TX would have produced an error, such as a
syntax error. Though, strictly speaking, this means that something is working
differently in a new version, replacing an error situation with functionality
is usually not considered a deviation from backward-compatibility.
When a change is introduced which is not backward compatible, \*(TX's
.code -C
option can be used to request emulation of old behavior.
The option was introduced in \*(TX 98, and so the oldest \*(TX version which
can be emulated is \*(TX 97.
Here are version values which have a special meaning as arguments to the
.code -C
option, along with a description of what behaviors are affected. For each
of these version values, the described behaviors are provided if
.code -C
is given an argument which is equal or lower. For instance
.code -C 103
selects the behaviors described below for verison 105, but not those for 102.
.IP 105
Provides the behavior that the
.code open-file
function automatically marks a stream open on a TTY devices as a real-time stream
(subject to the availability of the POSIX
.code isatty
function).
Also allows unrecognized backslash escape sequences in regular
expression syntax to simply denote the escaped character literally,
as was historically the case prior to \*(TX 106, so that
.code \ez
for instance denotes
.codn z .
As of \*(TX 106, these are diagnosed as errors.
.IP 102
Up to \*(TX 102, the
.code get-string
function did not close the stream. This old behavior is emulated.
.IP 101
Up to \*(TX 101, the
.code make-like
function incorrectly returned
.code nil
when converting the empty list
.code nil
to string type. This affects numerous generic sequence functions,
causing their result to be
.code nil
instead of an empty string.
.IP 100
Up to \*(TX 100, the
.code split-str
function had an undocumented behavior. When the
.code sep
argument was an empty string, it split the string into
individual characters as if by calling
.codn list-str .
This behavior changed to the currently
documented behavior starting in \*(TX 101.
Also, the arguments of the
.code where
function, which introduced in \*(TX 91, were reversed starting
in \*(TX 101.
.IP 99
Up to \*(TX 99, the substitution of TXR Lisp expressions in
.code @(output)
directives and in the quasistrings of the pattern language
exhibited the buggy behavior that if the TXR Lisp expression
produced a list, the list was rendered as a parenthesized
representation, or the text
.code nil
in the empty list case. Moreover, in the
.code @(output)
case, the value of TXR Lisp expressions was not subject to filtering.
Starting with \*(TX 100, these issues
are fixed, making the the behavior is consistent with
the behavior of TXR Lisp quasiliterals.
.IP 97
Up to \*(TX 97, the error exception symbols such as
.code file-error
were named with underscores, as in
.codn file_error .
These error symbols existed:
.codn type_error ,
.codn internal_error ,
.codn numeric_error ,
.codn range_error ,
.codn query_error ,
.code file_error
and
.codn process_error .
.SH* Appendix
.SS* A. NOTES ON EXOTIC REGULAR EXPRESSIONS
Users familiar with regular expressions may not be familiar with the complement
and intersection operators, which are often absent from text processing tools
that support regular expressions. The following remarks are offered in hope
that they are of some use.
.TP* "Equivalence to Sets"
Regexp intersection is not essential; it may be obtained from complement and
union as follows, since De Morgan's law applies to regular expression algebra:
.code (R1)&(R2)
=
.codn ~(~(R1)|~(R2)) .
(The complement of the union of the complements of
R1 and R2 constitutes the intersection.) This law works because the regular
expression operators denote set operations in a straightforward way. A regular
expression denotes a set of strings (a potentially infinite one) in a condensed
way. The union of two regular expressions
.code R1|R2
denotes the union of the set
of texts denoted by
.code R1
and that denoted by
.codn R2 .
Similarly
.code R1&R2
denotes a set intersection, and
.code ~R
denotes a set complement. Thus algebraic laws
that apply to set operations apply to regular expressions. It's useful to keep
in mind this relationship between regular expressions and sets in understanding
intersection and complement.
Given a finite set of strings, like the set
.cblk
{ "abc", "def" }
.cble
which corresponds to the regular expression
.codn (abc|def) ,
the complement is the set which contains
an infinite number of strings: it consists of all possible strings except
.str abc
and
.strn def .
It includes the empty string, all strings of length 1, all strings
of length 2, all strings of length 3 other than
.str abc
and
.strn def ,
all strings of
length 4, etc. This means that a "harmless looking" expression like
.code ~(abc|def)
can actually match arbitrarily long inputs.
.TP* "Set Difference"
How about matching only three-character-long strings other than
.str abc
or
.strn def ?
To express this, regex intersection can be used: these strings are the
intersection of the set of all three-character strings, and the set of all
strings which are not
.str abc
or
.strn def .
The straightforward set-based reasoning
leads us to this:
.codn ...&~(abc|def) .
This
.code A&~B
idiom is also called set
difference, sometimes notated with a minus sign:
.code A-B
(which is not supported in \*(TX regular expression syntax). Elements which
are in the set
.codn A ,
but not
.codn B ,
are those elements which are in the intersection of
.code A
with the complement of
.codn B .
This is similar to the arithmetic rule
.codn A - B = A + -B :
subtraction is
equivalent to addition of the additive inverse. Set difference is a useful
tool: it enables us to write a positive match which captures a more general set
than what is intended (but one whose regular expression is far simpler
than a positive match for the exact set we want), then we can
intersect this over-generalized set with the complemented set of
another regular expression which matches the particulars that we wish excluded.
.TP* "Expressivity versus Power"
It turns out that regular expressions which do not make use of the
complement or intersection operators are just as powerful as expressions
that do. That is to say, with or without these operators, regular expressions
can match the same sets of strings (all regular languages). This means that
for a given regular expression which uses intersection and complement, it is
possible to find a regular expression which doesn't use these operators, yet
matches the same set of strings. But, though they exist, such equivalent
regular expressions are often much more complicated, which makes them difficult
to design. Such expressions do not necessarily
. B express
what it is they match; they merely capture the equivalent set. They
perform a job, without making it obvious what it is they do. The use of
complement and intersection leads to natural ways of expressing many kinds of
matching sets, which not only demonstrate the power to carry out an operation,
but also easily express the concept.
.TP* "Example: Matching C Language Comments"
For instance, using complement, we can write a straightforward regular
expression which matches C language comments. A C language
comment is the digraph
.codn /* ,
followed by any string which does not contain the
closing sequence
.codn */ ,
followed by that closing sequence.
Examples of valid comments are
.codn /**/ ,
.code /* abc */
or
.codn /***/ .
But C
comments do not nest (cannot contain comments), so that
.code /* /* nested */ */
actually consists of the comment
.codn /* /* nested */ ,
which is followed by the trailing junk
.codn */ .
Our simple characterization of interior part of a C comment as a string
which does not contain the terminating digraph makes use of the
complement, and can be expressed using the complemented regular expression like
this:
.codn (~.*[*][/].*) .
That is to say, strings which contain
.code */
are matched by
the expression
.codn .*[*][/].* :
zero or more arbitrary characters, followed by
.codn */ ,
followed by zero or more arbitrary characters. Therefore, the complement of
this expression matches all other strings: those which do not contain
.codn */ .
These strings make up the inside of a C comment between the
.code /*
and
.codn */ .
The equivalent simple regex is quite a bit more complicated.
Without complement, we must somehow write a positive match for all strings such
that we avoid matching
.codn */ .
Obviously, sequences of characters other than
.code *
are included:
.codn [^*]* .
Occurrences of
.code *
are also allowed, but only if followed
by something other than a slash, so let's include this via union:
.cblk
([^*]|[*][^/])*.
.cble
Alas, already, we have a bug in this expression. The
subexpression
.code [*][^/]
can match
.codn ** ,
since a
.code *
is not a
.codn / .
If the next character in the input is
.codn / ,
we missed a comment close. To fix the problem we
revise to this:
.cblk
([^*]|[*][^*/])*
.cble
(The interior of a C language comment is a any
mixture of zero or more non-asterisks, or digraphs consisting of an asterisk
followed by something other than a slash or another asterisk). Oops, now we
have a problem again. What if two asterisks occur in a comment? They are not
matched by
.codn [^*] ,
and they are not matched by
.codn [*][^*/] .
Actually, our regex must not simply match asterisk-non-asterisk digraphs, but
rather sequences of one or more asterisks followed by a non-asterisk:
.cblk
([^*]|[*]*[^*/])*
.cble
This is still not right, because, for instance, it fails to match the interior
of a comment which is terminated by asterisks, including the simple test cases
where the comment interior is nothing but asterisks. We have no provision in
our expression for this case; the expression requires all runs of asterisks to
be followed by something which is not a slash or asterisk. The way to fix this
is to add on a subexpression which optionally matches a run of zero or more
interior asterisks before the comment close:
.cblk
([^*]|[*]*[^*/])*[*]*
.cble
Thus our the semi-final regular expression is
.cblk
[/][*]([^*]|[*]*[^*/])*[*]*[*][/]
.cble
(Interpretation: a C comment is an interior string enclosed in
.codn /* */ ,
where this interior part
consists of a mixture of non-asterisk characters, as well as runs of asterisk
characters which are terminated by a character other than a slash, except for
possibly one rightmost run of asterisks which extends to the end of the
interior, touching the comment close. Phew!) One final simplification is
possible: the tail part
.code [*]*[*][/]
can be reduced to
.code [*]+[/]
such that the
final run of asterisks is regarded as part of an extended comment terminator
which consists of one or more asterisks followed by a slash. The regular
expression works, but it's cryptic; to someone who has not developed it, it
isn't obvious what it is intended to match. Working out complemented matching
without complement support from the language is not impossible, but it may be
difficult and error-prone, possibly requiring multiple iterations of
trial-and-error development involving numerous test cases, resulting in an
expression that doesn't have a straightforward relationship to the original
idea.
.TP* "The Non-Greedy Operator"
The non-greedy operator
.code %
is actually defined in terms of a set difference,
which is in turn based on intersection and complement. The uninteresting case
.code (R%)
where the right operand is empty reduces to
.codn (R*) :
if there is no trailing
context, the non-greedy operator matches
.code R
as far as possible, possibly to the
end of the input, exactly like the greedy operator. The interesting case
.code (R%T)
is defined as a "syntactic sugar" which expands to the expression
.code ((R*)&(~.*(T&.+).*))T
which means: match the longest string which is matched
by
.codn R* ,
but which does not contain a non-empty match for
.codn T ;
then, match
.codn T .
This is a useful and expressive notation. With it, we can write the regular
expression for matching C language comments simply like this:
.code [/][*].%[*][/]
(match the opening sequence
.codn /* ,
then match a sequence of zero or more
characters non-greedily, and then the closing sequence
.codn */ .
With the non-greedy
operator, we don't have to think about the interior of the comment as set of
strings which excludes
.codn */ .
Though the non-greedy operator appears expressive,
its apparent simplicity may be deceptive. It looks as if it works "magically"
by itself; "somehow" this
.code .%
part "knows" only to consume enough characters so that
it doesn't swallow an occurrence of the trailing context. Care must be taken
that the trailing context passed to the operator really is the correct text
that should be excluded by the non-greedy match. For instance, take the
expression
.codn .%abc .
If you intend the trailing context to be merely
.codn a ,
you must be careful to write
.codn (.%a)bc .
Otherwise, the trailing context is
.codn abc ,
and this means that the
.code .%
match will consume the longest string that does not contain
.codn abc ,
when in fact what was intended was to consume the longest string that
does not contain
.codn a .
The change in behavior of the
.code %
operator upon modifying the
trailing context is not as intuitive as that of the * operator, because the
trailing context is deeply involved in its logic.
On a related note, for single-character trailing contexts, it may be a good
idea to use a complemented character class instead. That is to say, rather than
.codn (.%a)bc ,
consider
.codn [^a]*abc .
The set of strings which don't contain the
character a is adequately expressed by
.codn [^a]* .