python-markdown-discuss

[Python-markdown-discuss] Adding an extension to convert LaTeX(ish) to MathML

[Andrew S. <and...@ma...>]

Dear Python Markdowners,

Is anyone interested in helping to write an extension that would convert
LaTeX(ish) syntax to MathML?

There is a c++ library called itex2MML which converts a subset of LaTeX
mathematics syntax to MathML.  Using swig, it can be compiled with the
appropriate "hooks" to allow it to be called from a python script.  Using
a bit of "cargo cult" programming (a term I've just heard about and which
describes my approach to programming perfectly), I've written a CGI script
that uses this to convert this LaTeXish language (known as iTeX) to MathML.
Combining it with the existent python program SVGMath, and the libcairo
extensions, it is possible to convert it further to SVG and thence to PNG, if
required (thus meaning that it can _internally_ cope with different browser
requirements).

It strikes me that this would be an extremely useful thing to add to the
python markdown suite.  This already exists in one of the Ruby variants of
Markdown (Maruku) and is used in the Instiki wiki engine.  I've written the
PHP code necessary to do this and am using it in a PHP forum.  Having a python
implementation would enable things like the OSQA platform to be used for
proper mathematics.  Indeed, because python already has the SVGMath program,
there are considerable advantages in having this for python than any of the
other languages.

The only drawback is that I've no experience of programming in python, beyond
that little hackery needed for the CGI script.  So I wondered if there was
anyone here who would be interested in doing the python bit of it.

To be clear: I already have the extensions, and from doing the PHP version
I already know some of the potential pitfalls that may be encountered.  So it
really is just the ability to program in python that I'm lacking!  And much as
I would benefit from learning another language, it would be far quicker and
produce a far more robust extension if someone who knew about python already
was prepared to help!

To see the CGI script in action, go to:

http://www.mathforge.org/itex/itex.html

Thanks,

Andrew Stacey

Re: [Python-markdown-discuss] Adding an extension to convert LaTeX(ish) to MathML

[Andrew S. <and...@ma...>]

Hi!

Thanks for the quick initial response.

On Mon, Aug 02, 2010 at 12:45:55PM -0400, Waylan Limberg wrote:
> Hmm, I'm not a real big fan of Swig myself (mostly through my own lack of
> success in compiling a few packages). It's not that Swig is bad (it
> certainly has it's merits), but this just feels like too heavy of
> a dependency for it to be shipped with Python-Markdown.

Understood.

> Of course, extensions can certainly live as third party packages, so
> it may be appropriate that way.

For me, this would be absolutely fine.  I wasn't angling for more.

> > required (thus meaning that it can _internally_ cope with different
> > browser requirements).
> 
> While this may be appropriate on a case by case basis for certain
> projects, this seems like a bad idea to me for a general use
> extension. Browser sniffing and making images available for a web
> server to serve separate from the markdown document just doesn't
> belong here.  Of course, if an individual project needs those
> features, they could certainly hack up the basic extension to fit
> their specific needs.

Absolutely!  I misspoke a little.  I would propose that the extension have
a switch by which it can be told to produce MathML, SVG, or PNG.  Then the
calling script can do the browser selection or just choose one and so pass the
appropriate option to the extension.  Alternatively, the extension can just do
the conversion to MathML and then the calling script can further pass the
whole XHTML document through SVGMath to do the rest.

I mentioned this because one of the major issues with MathML is its lack of
browser support.  But it's chicken-and-egg: people don't want to adopt MathML
until it's supported widely, but the push to get it supported widely won't
come until lots of people use it and see how good it is.  So being able to
offer a choice is a way of getting people to use MathML without losing their
audience.

> On the other hand, I haven't looked at the solutions offered in other
> languages. Maybe those issues have already been worked out and a
> default behavior has been established, in which case we could simply
> copy it.

As I said, I've done the PHP coding and the guy who currently maintains
itex2MML has done the ruby version.  A perl version is no doubt not far off
(that one I could do myself but don't have an application to hand right now in
which to use it!).

> Any chance we could see the source of your CGI script?

Yes of course!  I should have put that in my earlier email.  My apologies.  My
additions to itex2MML are available at:

http://www.math.ntnu.no/~stacey/code/itexToMML

The python cgi script is in cgi-bin/itex.py (relative to that link).  The
whole thing is a bzr repository.  Note that this isn't a live version but just
the code.

> Again, thanks for taking the time to get this started.

No problem!  Thanks for the quick initial reply,

Andrew