Found here: http://www.pro-linux.de/cgi-bin/DBApp/check.cgi?ShowApp..15124.100
Google Translation:
'TXR implements a sophisticated rule-based query language that can retrieve data from multiple text files or Unix pipes and can write results to multiple files or Unix pipes. The queries search in whole files or in sections. TXR can output raw or template-driven reports. An outstanding feature is that, unlike parser generators or regular expressions, it is allowed to have multiple rules and bind variables - this is clearly described in the language. Functions can be programmed, and there is an exception handling, as is known from ordinary object-oriented programming languages. The syntax of Txr is reminiscent of xslt, but is much simpler and more meaningful. The handling of variables is similar to prolog (unification and comparison). For example, a Txr template could easily parse Python source code. So far, all the features of the language are described in a compact manpage."
(text/plain)