[Texmacs-dev] Re: [axiom] TeXmacs and Literate Programming (fwd)

[Joris van der Hoeven]

More literate programming...


---------- Forwarded message ----------
Date: Sat, 1 Jun 2002 16:22:20 +0200 (MET DST)
From: Joris van der Hoeven <address@hidden>
To: root <address@hidden>
Cc: address@hidden, address@hidden,
    address@hidden, address@hidden, address@hidden,
    address@hidden, address@hidden, address@hidden,
    address@hidden, address@hidden,
    Gilbert Baumslag <address@hidden>,
    Al Thaler <address@hidden>,
    Manuel Bronstein <address@hidden>,
    Richard Fateman <address@hidden>,
    William Sit <address@hidden>,
    Richard Paul <address@hidden>
Subject: Re: [axiom] TeXmacs and Literate Programming

Tim,

> I had an extended discussion with Gilbert Baumslag and Al Thaler (of
> Magnus fame).  Gilbert heads a department at City College of NY. They
> have a project to make computers more useful for mathematics and the
> whole topic of literate programming came up. They are going to have
> discussions this summer about future directions for research.  I'll
> pass along your comments.

Thanks.

> I think that TeXmacs, as a direction for algebra system front-ends, is
> a good path to follow. Indeed, I hope to have a closer integration
> between TeXmacs and Axiom. The ability to do 2D I/O has long been
> needed. I used to be able to do 2D I/O on a pen tablet years ago at
> IBM Research but it never made it out of the labs. I'd also like to
> integrate Axiom's asq abilities into some menu and key-binding driven
> system to make the algebra more accessible.

Thanks.

> As to the idea of a special TeXmacs language I have only one
> hesitation.  TeX is so widely favored among researchers that virtually
> all of the work I'm aware of is written in TeX.

We have written import filters for LaTeX. Because of the lack of semantics
of TeX/LaTeX, this converter cannot be perfect, but we will keep improving
it in the future and it is reasonable to assume that all the texts that
you are thinking about can be imported correctly with a bit of work.

> That means that I can combine existing research papers with code easily.

Yes, but in TeXmacs this might become even easier and better:

First, as I said before, the high-level style rewriter engine that
we are currently implementing may be used to dynamically display
ASCII texts in a pritty printed way (some kind of syntactic highlighting,
but better).

Secondly, the literate programs may become more dynamic,
using folding/unfolding, automatically generated hyperlinks,
plugging the programs into a computer algebra session when
this makes sense, etc.

Thirdly, extracting a classical ASCII program from a TeXmacs document
is very easy: remember that TeXmacs documents are essentially trees,
which is way better than ASCII documents. Consequently, it is very easy
to perform operations like extractions on TeXmacs documents.

Fourthly, remember that TeXmacs documents carry more semantics,
something which is very useful in a mathematical environment.

Etc.

> I'm using Richard
> Paul's work as an example for the Denavit-Hartenberg Matrix domain.
> I'll send you the "pamphlet" (as I've taken to calling a literate
> programming document of an algebra domain) I develop to play with if
> you like. Noweb is available online (www.eecs.harvard.edu/~nr/noweb).

The site does not reply. Anyway, once again: why are there so many people
out there who are eager to invest into an n-th dirty TeX/LaTeX hack,
and virtually nobody who wants to invest in a really modern approach
to typesetting / editing / programming / etc.?
I understand that TeX/LaTeX is a standard and blah, blah,
but this is not sufficient to explain this lack of courage.
Also, in my opinion, only a few successful man-years would
be sufficient to turn TeXmacs into the killer-app that
many scientists have been waiting for.

> Fateman and Bronstein have raised some issues to think thru and 
> they probably also apply to your effort. They are quoted in the
> latest digest at (home.earthlink.net/~jgg964/axiom.html).

Yes, I know about this kind of arguments, but I do not want to
launch another discussion about the merits of free software.
I also recall that free software is not anti-commercial software and
that I continue to believe that it should be possible to create
efficient support/subscription services around free scientific software.

> Noweb is my current tool of choice though I find it doesn't fully
> cover the directions I'd like (such as generating examples and test
> cases from the pamphlet as well as code). I want to raise the level of
> language interface for the original researcher so I can extract the
> maximum knowledge from him while the subject is still considered his
> hot topic. Knuth seems not to have covered the "systems issues" in the
> original idea but I find they are important.  Currently Axiom is only
> used to develop running code but the test and example cases get lost
> and the reseach paper gets entombed in a library. In a place with
> great research facilities this was not an issue but it will be for
> open source developers.
> 
> Having all the literate programs in TeX also means that I can combine
> literate programs into "booklets" easily.  The booklet idea is a
> mechanism I plan to use to take various "slices" thru the algebra
> abilities of Axiom (say, all matrix types) by combining "pamphlets"
> with additional tutorial text for teaching purposes.  A special
> TeXmacs language would leave me with the task of format conversions of
> existing documents which would be yet another challenge. One upside
> you might provide is the ability to translate embedded examples into
> any computer algebra system. Currently my thoughts on modifications to
> Noweb have only concerned generating Axiom specific examples and test
> cases from the pamphlet.

See ^^^.

The only potential difficulty I see is the conversion of existing
documents. But this will not be automatic anyway, even in your scheme,
because you will need to merge the text with the code in some way.

As to the extraction of code from TeXmacs documents and
conversion to other formats, I repeat that this is easy
because of the rich tree-structured document format of TeXmacs.

> I'd like to have further discussion of pamphlet and booklet support
> in TeXmacs if you think this is in line with your current interests.

We just implemented a (still buggy) presentation mode, a first version of
folding/unfolding and switching slides. Next on the list is folding/unfolding
of existing structures like sections and computer algebra sessions.
Notice also that one may imagine other dynamic markup in an editor like TeXmacs.

Yours, Joris