Re: [PHP-DOC] RE: [DOC-HU] PHP DOC BUILDS -- PLEASE READ

> > It seems that only 'ar' had been tried to be built. I have fixed the
> > error. BTW 'ar' does not need to be built anyway, it has only 11kb
> > translated content. Compared to the total 10-11MB in the English tree,
> > this is quite nothing. It was last modified more then a year ago. Or I
> > got some old files from CVS...
>
> Yes. Right now we're on 'es' (that's Spanish, Derick). I am pondering about
> the possibilities of making the build system more tailored to what we need
> it for (ie, I think it could be interesting to have a 'meta' build system
> that would know what languages to build. There is no point in building a
> language that doesn't have anything substantial translated.

Some criteria was establised for this. AFAIK it was said that it checks for 
language-snippets.ent precense, when deciding whether a translation should be 
built. BTW this file is the on translations are started with, so it may not 
be the best to check for... Maybe a sum translated size check?

> Also, as a consideration: how does translated languages get built? I have a
> sneaky suspicion that they get built fully. Can't we just start work in an
> en tree and just rebuild the files that got translated?

There were two things popping up about this:

 - build only what's translated for translations
 - build only what's changed for languages (assuming generated output
   from a previous build is available)

Both of these would enourmously speed up the generation, so all languages 
would be built in one day. Problems with this include inconsitencies 
appearing in:

 - TOC parts
 - Prev/next links
 - Any type of <xref> crosslinks

Imagine you translate a file. The translated title should show up in all TOCs 
on all levels, where that title was listed. It should also show up on the 
HTML page before the file and after the file (prev/next links), and also 
possibly in indexes... Xref links generate the link text from the endpoint's 
title. They are rarely used in phpdoc, and discouraged, but some authors tend 
to add such tags from time to time...

There is also no way I know to start the transformation on indicidual levels 
(transform chapters in sequence programatically instead of transform one 
book). So we probably would need to generate a temporary <book> for the 
translated content with only the translated file, and then try to replace the 
files in the EN build with the translated ones.

Then again, the above three problems appear, as well as that there may be 
files with generated IDs (ie. aen334423.html), which are also a problem if 
linked from the toc, but then translated. As the translated file will 
probably have a different generated ID. This is a rare case, AFAIK all files 
are generated after their ID (installation.apache.html) or tag names 
(copyright.html).

Hope that this shed some light on the issues involved...
Goba