Re: [Texmacs-dev] TeXmacs server / filesystem

[Joris van der Hoeven]

Hi Felix,

On Sat, Jun 17, 2006 at 07:24:43PM +0200, Felix Breuer wrote:
> until now the TeXmacs server has been discussed with a focus on
> chat/wiki capabilities. I am more interested in the texmacs filesystem
> and version control aspects Joris mentioned. So: Joris, what do you plan
> in that department?

I mainly intend to implement what is necessary for myself in order to
manage my own articles. This includes version control, hyperlinking and
some simple facilities for searching.

> A "TeXmacs server" would be most useful to me as a system for storing
> and managing texmacs trees (or, better even, graphs) that can be
> accessed across processes and across a network. Arbitrary programs could
> interface with such a server in several ways:
> 
> a) via messages passed over a socket (as is the case ATM)
> b) via a virtual file system ([1],[2])
> c) via the command line
> d) by linking with a texmacs server library
> 
> Especially d) would be of interest to me. As mentioned before [3] I
> would very much like to see texmacs decomposed into several libraries (a
> texmacs-tree-lib, a typsetting-lib, etc.) that I could link my apps
> against.

Sure, these possibilities are all very interesting. However, as I said before,
I do not have much to spend on these issues (even though I would like to).
Nevertheless, the server is written in a very modular way, so it should be
easy to add new services which do the things that you want.

Recall that the server is currently written in scheme,
so it is not really a library to which you can link.
On the other hand, it is very easy to add new services as a module.

Another thing which may interest you is to add literate programming
or conversion services, which would simply act like daemons which
convert part of the server files into compilable source code or a web site.

> Motivation: Recently I wanted to have my computer run some special
> purpose computation using a custom algorithm. To this end I needed to
> generate some data (which was convenient to do with Haskell) and run my
> algorithm (which was convenient to write in C++). I did not have the
> time to write a UI and ended up 
> 
> * creating a mini-file-format for my data,
> * writing serialization/deserialization code in both Haskell and C++,
> * staring at the plain-string output of my algorithm dumped into some
> text-file and
> * visualizing the whole thing using pen and paper.
> 
> Now, it would have been so much nicer to have both of my apps talk to a
> texmacs server to do all the data-wrangling and then write a texmacs
> editor stylesheet to do the visualization for me! (And if I could have
> used a texmacs typsetting/graphics library I might even have coded a
> GUI.)

Yes, that would be the idea. As for mini-file-formats,
I systematically recommend scheme, because it is easiest to parse.

> This might also be useful promotion-wise: if texmacs is not as popular
> as it should be with the LaTeX / Gnome+KDE / XML crowds because texmacs
> does everything in a different way, then texmacs has to make it easy for
> programmers to do everything the texmacs way!  :)

Sure, I try to work on that, but time is currently lacking...

By the way, did you make progress with the litterate programming stuff?
If you have some working code, then I am still interested to include it
into TeXmacs. The approach which just uses comments with TeXmacs text
should not be hard to implement.

Best wishes, Joris