Since around August 2009, I've been developing a language called TXR
which consists of:
* a pattern language (the TXR Pattern Language) for matching whole text
documents/streams and scraping their contents.
* an original Lisp dialect called TXR Lisp.
TXR is offered under the two-clause BSD license.
A few days ago, I put out version 195, which is what it looks like it
means: the one hundred ninety-fifth public release. (Semantic
Originally, TXR was just the pattern language. It sprouted the Lisp
dialect, and much of the development has been in that area.
TXR Lisp's basic political alignment is with ancient tradition: lists
are terminated by the symbol nil, which is also Boolean false. It is a
Lisp-2. However, it supports Lisp-1-style evaluation also under the dwim
operator, which sugared by [ ... ] notation. This shift to an
alternative evaluation is deeply integrated into the expander.
TXR Lisp has various goodies such as: exception handling, an OOP system
(not CLOS-like; single dispatch, single inheritance), syntactic places,
reasonably rich I/O with a format function, string interpolation,
built-in lazy lists, buffers, delimited continuations, a Lisp-ified
implementation of Awk, a package system, and a really great FFI.
Numerous POSIX interfaces are included in the library.
There is an interactive listener, with command persistent command
history, decent editing (visual copy and paste, undo) with a great
Recently I developed a virtual machine and compiler for TXR Lisp.
TXR Lisp makes list operations generic over vectors, strings and
user-defined sequences. For instance (car "abc") returns #\a, and (cdr
"z") yields nil.
It has numerous primitives for working with lazy lists. There are lazy
strings and The OOP system supports lazy objects.
There are nice ways to index sequences. For instance (swap [a 2..5] [b
10..19]) will do what it looks like: exchange the given sub-ranges of
sequences a and b in place, even though they are of different length.
Objects other than functions are callable, with useful semantics.
The OOP system supports a mechanism called "equivalence substitution"
which allows objects to be used as custom hash keys without the
programmer having to write a hashing function.
TXR is documented by a thorough user manual which takes the form of one
giant man page that converts to about a 640 page PDF as well as a
hyper-linked HTML document hosted online.
gnu-misc-discuss mailing list