Re: IMPORTANT: pls fix double names

On 2011-04-20 15:46, Rimas Kudelis wrote:
>
> Hi,
>
> 2011.04.20 16:37, Dwayne Bailey ra=C5=A1=C4=97:
>> I'd rather see us spending some time to create a way to inject and=20
>> store arbitrary errors for PO files stored on Pootle. That way anyone=20
>> can create a script that identifies problems, injects them into=20
>> Pootle and allows translators to fix them quickly on Pootle in an=20
>> environment that is familiar and safe.
>
> What about automatically marking invalid translations as fuzzy in the=20
> .po file and adding a translator comment above them? This would be=20
> generic enough, and would not depend on Pootle at all.
It's a hack.  It would mean that people writing checks must work on PO=20
files, the checks I'm referring to are working directly on SDF files. =20
You could script pogrep to find these, mark them fuzzy and inject them=20
into the original PO files - but still a hack in my mind.  Not all=20
failing checks should be fuzzy, I reserve that for checks that break thin=
gs.

The checks in Pootle are not actually written in Pootle but in the=20
Translate Toolkit which is why you can run them against your PO files=20
outside of Pootle.  We already use a scheme for marking these when we=20
use pofilter on the command line.  Maybe that is an approach that should=20
be examined, but it would still mean changing Pootle to read these=20
comments when it finds them in a PO file.

Marking something fuzzy doesn't push it into a class of error with a=20
specific agenda to fix.  So I can't go through errors of type X.  I will=20
get all my fuzzy strings together in one lump with only a comment (which=20
I must also now delete) to distinguish it from a proper fuzzy.

It has been my aim to make sure these checks cover gsicheck 100% as then=20
for a Pootle user Pootle becomes the right place to fix these errors. =20
They then don't need to decipher gsicheck output which is difficult. =20
Pootle has a special openoffice checker that it gets from the Translate=20
Tolkit so it is possible to write anything that is needed and then also=20
doesn't depend on Pootle.

So either the checks are written to be a proper Translate Toolkit check=20
so that it can show directly in Pootle, ideal (Unfortunately we haven't=20
had any Oracle devs contribute changes to the checks so don't hold your=20
breath). Or we have some way for outside checks to report their errors=20
to Pootle and mark units for review.

Programmers will do anything as quickly as they can.  Us localisers=20
accept what we get which is why we're reading SDF snippets and strange=20
error output.  We spend quite a bit of time interpreting the results, we=20
don't complain.  I think there is a better solution that would mean that=20
we can translate instead of pretend we're programmers as we interpret=20
SDF output.

--=20

regards
Dwayne

--=20
-----------------------------------------------------------------
To unsubscribe send email to [email protected]
For additional commands send email to [email protected]
with Subject: help