Re: create compendium automatically

[Danilo Segan <dsegan gmx net>]

недељa, 28. септембар 2003. 22:12:21 CEST — Keld Jørn Simonsen написа:
> I think the compendia are a good idea, but it seems like we need to  
> do some
> optimization on the code, eg by employing database techniques.

Yes, of course it's probably possible to do it, but it would involve  
rewriting of a lot of standard gettext tools to be more suitable for  
the task (or writing the same tools from scratch, with different  
objectives).

Quite similar approach could be taken in regard to status pages  
optimizations, like caching the results of one msgmerge (which original  
messages were fuzzied based on which previous translations), and  
reusing this information in all other translations, and doing  
incremental updates instead of rerunning msgmerge for each translation.

(For instance, we do a msgmerge on sv.po for gnumeric, and cache which  
english messages are no longer present, and if there are some new  
english messages, from what was a fuzzy match gotten [which english  
string]; then, we just do the same job for sr.po, without searching all  
the stuff, and all the messages that were not present in sv.po but are  
in sr.po are checked in the same way and added to database if they were  
fuzzied.)

Yet, there are big flaws to any of these approaches: dedication of time  
and resources to accomplish them is quite big, at least at the start.  
At the moment, I think it would be far better to dedicate any developer  
time we've got left to creating tools for DocBook-to-PO transformation  
tools. Of course, my priorities might be skewed, so I hope someone will  
give another point of view soon, with all the reasoning behind it.


Cheers,
Danilo