We've been using epydoc on our FiPy project for a couple of years and are very happy with it. The one area where we thought there was room for improvement was in separating the content from the LaTeX markup (I recall somebody expressing a similar desire re: HTML/CSS expressed at PyCon?'04; we don't use HTML output (too much math), so I don't know where that side stands). We got this working well enough for our own needs about two years ago, but never cleaned it up enough to pass back. My reward for that lethargy is that I got to merge the changes from 2.1 to 3.0 myself instead of letting Ed do it 8^). I've now managed to pull things together so that they're generally usable and brought them up to date with r1588 from the epydoc svn repository.

The files that we've added or changed are:

epydoc.sty

A LaTeX style file that produces a reasonable facsimile of what comes out of stock epydoc. The remaining discrepancies are either due to bugs in r1588 or had a poor cost/benefit until I knew whether this was even of interest to the epydoc developers. I'm happy to try to fix any remaining issues.

latex.py (diff vs. r1588)

Abstracted the markup, leaving the formatting details to a LaTeX .sty file.

restructuredtext.py (diff vs. r1588)

Markup for reST admonitions.

cli.py (diff vs. r1588)

• Added --[no-]sub-modules options to control output of write_module_list() (we prefer to use a minitoc for this information).

• Added --pdflatex option (latex -> dvips -> ps2pdf is so 20^th century)

I wanted to move most of the formatting out of ParsedEpytextDocstring._to_latex() as well, but my efforts there caused more problems than they solved.

A few examples to compare the output:

attachment:api-r1588.pdf:

epdoc APIs processed with epydoc r1588

attachment:api-epydoc_sty.pdf:

epdoc APIs processed with hacked epydoc and formatted with epydoc.sty

attachment:api-fipy_sty.pdf:

epdoc APIs processed with hacked epydoc and formatted with fipy.sty (this isn't a petition for a change to what epydoc-LaTeX output should look like; just an example of what's possible when content is separated from markup).