LaTeX2HTML

Used by PlanetMath for displaying the LaTeX in entries as Web pages.

But SteveCheng considers it broken and wants to scream at it after discovering that it has the following misfeature: it will mangle the ALT text of long formulae:

e.g.

 \displaystyle = \left\lvert\left( \frac{f'(\xi_x)}{g'(\xi_x)} - 
 m\right) + \fra... ...x)}{g'(\xi_x)} \left(  
 \frac{1-g(c)/g(x)}{1-f(c)/f(x) } - 1 \right) \right\rvert

Why!? It makes downstream processing impossible, and unreadable for text-based browsers. Never mind that trying to cleverly abbreviate a long formula is actually more effort to implement. And the HTML specification clearly says that the ALT text is to be the alternative text when the image is not available. "…" is not alternative text.

I bet this would be super-easy to hack out. --akrowne Mon Apr 3 03:29:37 UTC 2006

Yeah, but you probably wouldn't care to install the corrected version onto PlanetMath, would you? Who knows what things might break? That's why I wrote this rant. I can't do anything about it.

FYI:

RE ALT comments

If the mangling is a feature of Latex2Html then a new option could be added – ask at http://latex2html.org/ ?

Or if it is a bug, it might still be best fixed with an option in case the new, correct functionality would break existing uses.

The place to start is with the author/maintainers of the package – I would think…

--ocat 4-Apr-2006

Well, thanks to you! Perhaps I should not have been so hot-headed: it was after only reading the above-linked post that I discovered that LaTeX2HTML also puts in <!– MATH … -→ comments. I would hope these would not be truncated. That would solve the problem at least; I'm happy :) --SteveCheng

Other software related to LaTeX conversion

A real gem I found when researching this topic:

• PyTeX - The TeX daemon. Implementation of interactive TeX? typesetting.

Anyway, enough complaints about LaTeX2HTML. I’ve started writing a replacement, which I will soon make available. What it will have:

• A small Python module to translate LaTeX to an intermediate XML format (TeXML)
• Almost all of the LaTeX to HTML translation will be driven by an XSL stylesheet
• Another separate Python module to extract the math markup and turn them into PNG (using TeX? and dvipng)
• Or alternatively, direct LaTeX markup to MathML translation will be possible.

The first step is almost done.

Why do I write a new program instead of working on LaTeX2HTML instead? Because I think I can write a better program. Also:

• LaTeX2HTML is an eye-straining messy Perl script that is difficult to customize.
• LaTeX2HTML is too slow.
• LaTeX2HTML is just a pain in the ass. (I still have not gotten non-gray anti-aliased without-ugly-black-borders math images working yet, despite fiddling with the many combinations of switches available!)

— SteveCheng

I think this is a fine (actually good!) thing to work on, but do note that there are other latex-to-html converters on the market. It might be advisable to survey them all at some point. I think it would be swell to get a mathml display mode hacked into PlanetMath. --jcorneli

Besides LaTeX2HTML, there is also tex4ht which uses TeX? to parse its input. (There’s also TtH?, which is non-free, so I’m not even considering it for one more second.) I like neither, because I really want to get away from the TeX? system; the tool should be suitable for migrating from TeX?. TeX? is good typesetting, but its implementation sucks. Just even try to get good-quality PDF Chinese/Japanese/Korean output working from LaTeX and you’ll see what I mean. — SteveCheng

Progress

2006-05-18: The progress is going to be slow, unfortunately — there are lots of other things to work on. So for now, I’ll make the release of the first Python module I talked about earlier: http://gold-saucer.afraid.org/mathml/latex2xml.py

Notes:

• The code is totally undocumented.
• Not only that, the parser is messy (a hand-rolled push-down automaton thing). It really needs to be tamed by the literate programming treatment.
• This program comes with its own GTK+ GUI debugger ;) To use it, type

 $ python 
 >>> import latex2xml
 >>> latex2xml.debug("file.tex")

You can observe the state of the parser at each token, and manipulate it more with the interactive console.

• The program current does not parse (only pass through) LaTeX math environments. However, this will be bad when we need to implement equation numbering. Still searching for a solution.
• Need to investigate what needs to be done for round-tripping