[[I|tschwinge]] started writing a plugin to render
GNU Texinfo
inside the ikiwiki environment.
This plugin is not neccessarily meant to enable people to write arbitrary
wiki pages in the Texinfo format (even though that is possible, of course),
but rather to ease collaboration on existing Texinfo documents.
The plugin is available at http://www.schwinge.homeip.net/~thomas/tmp/texinfo.pm.
It's very basic at the moment, but will be improved over time.
Issues
N-to-M Mapping of Input and Output Files
Conventional ikiwiki [[htmlizeing|plugins/write#index6h3]] plugins
have a one-to-one mapping of input file and output file:
some/where/page.mdwn is rendered to some/where/page.html.
This can also be achieved for Texinfo files, but is somewhat
unusual there, when rendering them to HTML. In general, there
is a N-to-M mapping:
- N Texinfo input files (a main
.texi file,
several helper files (fdl.texi, version.texi, ...), and
additional text files which are included from the main .texi
file, e.g. history.texi, libfoo.texi, libbar.texi. --[[tschwinge]]
As far as multiple input files, you'd need to use add_depends()
to let ikiwiki know that a change to any of those files should cause a
rebuild of the "main" file. --[[Joey]]
(?) I'll see about a frob to get makeinfo provide me with a list of additional files
it used for rendering a given .texi file. --[[tschwinge]]
I guess you'd also have to somehow deal with
it wanting to render pages for each of the helper files. Not quite sure
what the best way would be to avoid that. --[[Joey]]
Might it be an option to simply not render the pages that are already
being used as an include file for another .texi file? --[[tschwinge]]
Ikiwiki is perfectly happy with a page creating other files (see eg, the
img and teximg plugins, as well as the inline plugin's rss generation).
The will_render() function supports that.
What hasn't been done though is a page creating more than one other page.
Perhaps you could call IkiWiki::genpage by hand for each additional page.
You might also want to manipulate each data structure that tracks info about
pages, adding the additional pages to them, so that they're first class
pages that work as pages everywhere in ikiwiki (ie, can be inlined,
appear in a site map, be linked to, etc). Not sure how to do that,
and perhaps you could get away without doing it actually. --[[Joey]]
Currently I use makeinfo --no-split and render to stdout, so that I can
easily capture the output and stuff it into the appropriate ikiwiki data structure.
If we want to have multiple output files (which we'll eventually want to have,
to avoid having such large single-file outputs), we won't be able to
do this anymore.
(?) Then we'll need a way to find the main output file, which
will be the one to be copied into what ikiwiki expects to be the main output
of the rendered .texi file.
Perhaps (again) parse the .texi file for a @setfilename statement?
The other generated files will also have to
copied somewhere (preferably into a subdirectory named alike the main file
to avoid name space collisions; but need to take care of links between the files then)
and need to be registed within the ikiwiki system.
--[[tschwinge]]
There needs to be some logic to establish a mapping between the N input files
and the M output files.
(At least for web-editing via CGI this is needed.)
Easiest would be either to leave M = 1 or to have
M = N and have a one-to-one mapping between input file n and output file m.
--[[tschwinge]]
makeinfo Output
makeinfo --html is being used for rendering. It creates stand-alone
HTML files, while ikiwiki only needs the files' <body>s.
(?) One possibility (which is what I'm doing at the moment) is to simply cut away
everythin until <body> is seen and after </body> has been seen. --[[tschwinge]]
|
