Delivering PO files (Re: Volunteering for Translation - Tamil)

[Andrea Pescetti <pe...@apache.org>]

On 17/12/2012 Jürgen Schmidt wrote:
> On 12/15/12 12:13 AM, Andrea Pescetti wrote:
>> (note to Juergen: these are your same files, but I deleted Testtool and
>> the 135 files that are 100% translated, so volunteers won't be confused)
> ok, but I would prefer if we communicate the details and keep the po
> bundles untouched. It makes the reintegration easier for me and I have
> to do less.

OK, I won't produce the "easy" packages until we agree on this. But 
really, the overhead on us is minimal and the impact on a translator who 
has never seen PO files is large: if we only keep those PO files where 
there is something to translate, and ignore those that are fully 
translated, we reduce the number of files by 50-80% and things are much 
more manageable for volunteers who need to track/reassign translations.

> I think the used tools for offline translation can take care of already
> translated strings.

Note that I never mentioned strings. I removed 100% translated files, 
not strings. This means that doing so we save translators the effort to 
open (or assign) 100-200 files that actually need no work and would be 
delivered back exactly as they are.

And the overhead on us is:

- before delivering the files, run something like (e.g., if you have a 
subdirectory named "he" with all PO files for Hebrew)

LG=he; pocount --short-strings $LG | grep -v '100%t' | grep -v 
'/basic/source/app.po' | cut --delim=' ' -f1 | xargs tar jcvf 
aoo341_${LG}_po-files-diff.tar.bz2

- when we receive the files back, at worst one has to extract them over 
the "full" package (but I don't think this is even needed)

True, no volunteers so far complained that we sent them too many useless 
files. But, especially for minor releases, I would consider this 
approach again.

> But it is no big problem, I just want to mention that with the upcoming
> update for AOO 4.0 we will remove unused string anyway.

This is a different topic... but very good to do so, this will help 
volunteers too.

Regards,
   Andrea.

Re: Delivering PO files (Re: Volunteering for Translation - Tamil)

[Andrea Pescetti <pe...@apache.org>]

janI wrote:
> On 17 December 2012 13:33, Andrea Pescetti wrote:
>> LG=he; pocount --short-strings $LG | grep -v '100%t' | grep -v
>> '/basic/source/app.po' | cut --delim=' ' -f1 | xargs tar jcvf
>> aoo341_${LG}_po-files-diff.tar.bz2
> As far as I can judge, you have forgotten that files with "fuzzy" marked
> messages also needs to be looked at.

No, the output that is filtered is something like

he/desktop/source/deployment/misc.po strings: total: 3	| 2t	1f	0u	| 
66%t	33%f	0%u

so these files would still be included in the package, since I filter 
out only those that have 100% translated (and, consequently, 0% fuzzy 
and 0% untranslated).

> and hopefully we only have about 40 files for 4.0 (still work in progress).

Good!

Regards,
   Andrea.

Re: Delivering PO files (Re: Volunteering for Translation - Tamil)

[janI <ja...@apache.org>]

On 17 December 2012 13:33, Andrea Pescetti <pe...@apache.org> wrote:

> On 17/12/2012 Jürgen Schmidt wrote:
>
>> On 12/15/12 12:13 AM, Andrea Pescetti wrote:
>>
>>> (note to Juergen: these are your same files, but I deleted Testtool and
>>> the 135 files that are 100% translated, so volunteers won't be confused)
>>>
>> ok, but I would prefer if we communicate the details and keep the po
>> bundles untouched. It makes the reintegration easier for me and I have
>> to do less.
>>
>
> OK, I won't produce the "easy" packages until we agree on this. But
> really, the overhead on us is minimal and the impact on a translator who
> has never seen PO files is large: if we only keep those PO files where
> there is something to translate, and ignore those that are fully
> translated, we reduce the number of files by 50-80% and things are much
> more manageable for volunteers who need to track/reassign translations.
>
>  I think the used tools for offline translation can take care of already
>> translated strings.
>>
>
> Note that I never mentioned strings. I removed 100% translated files, not
> strings. This means that doing so we save translators the effort to open
> (or assign) 100-200 files that actually need no work and would be delivered
> back exactly as they are.
>
> And the overhead on us is:
>
> - before delivering the files, run something like (e.g., if you have a
> subdirectory named "he" with all PO files for Hebrew)
>
> LG=he; pocount --short-strings $LG | grep -v '100%t' | grep -v
> '/basic/source/app.po' | cut --delim=' ' -f1 | xargs tar jcvf
> aoo341_${LG}_po-files-diff.**tar.bz2
>

As far as I can judge, you have forgotten that files with "fuzzy" marked
messages also needs to be looked at.


>
> - when we receive the files back, at worst one has to extract them over
> the "full" package (but I don't think this is even needed)
>
> True, no volunteers so far complained that we sent them too many useless
> files. But, especially for minor releases, I would consider this approach
> again.
>
>  But it is no big problem, I just want to mention that with the upcoming
>> update for AOO 4.0 we will remove unused string anyway.
>>
>
> This is a different topic... but very good to do so, this will help
> volunteers too.
>
and hopefully we only have about 40 files for 4.0 (still work in progress).


>
> Regards,
>   Andrea.
>