This is great: thanks for the report, Christian. :) On 20/12/2006, at 5:14 PM, Christian Perrier wrote:
This is a long report but I hope it will be useful for Pootle developers. CC'ed to the Toolkit development list because some issues could rather pertain to the TT.
Extremely useful: exactly what is needed.
I'm still fighting with Pootle on i18n.debian.net. I try nowto
integrate all Debian packages with "debconf" translations ("debconf"
is the Debian configuration system, ie the screens used for prompting
users when installing some packages..."debconf" translation is one of
the most achieved i18n part in Debian).
For people not accustomed to the Debian i18n project, the key thing to remember about debconf in processing terms is that the files are very small, ranging from 1-2 strings to perhaps 40 strings, with most containing only a handful of strings. But there are a LOT of them.
This sharply increases the probability of error, because you have so many different files, processed by different editors and people. One large file from one person/project is much less likely to create as many different errors as 100 debconf templates could.
So trying to process 700+ debconf templates is like dealing with a swarm of wasps. You can get bitten all over before you know what's happening. ;)
What I want to import is a tree with over 700 directories, named after
Debian source packages. Each of these directories contains a variable
number of PO files, named after the languages ISO-639 code and *one*
POT file always named "templates.pot".
When I put this tree in my "test" project, I face the following
problems:
- files that are not encoded in UTF-8 AND contain non ASCII characters
in their headers break Pootle because of bug #193
workaraound: re-encode files in UTF-8 but that's not exactly
satisfying in the long term. It will take ages before
we enforce UTF-8 on all Debian l10n teams (why would
we?)
Because it's the standard? :)
expected fix: fix bug #193..:-)
- invalid PO files stop the import by Pootle (see my posts about
the wrong Portuguese translation with a header missing a "\n")
workaround: remove these files and report a bug against the package
expected fix: have the Tookit *ignore* invalid files and Pootle
continue the import/indexing
Or we could have an option: ignore invalid files, OR display an interactive error.
- when one of my 700+ directories has a two-letter or three-letter
name, Pootle seems to switch itself to "let's assume that this
project arranges file with language directories and many PO files
for the same language in the same directory" layout mode
workaround: do not try to import two-letter and three-letter
directories which means ignore packages such as "nn",
"jed", "cvs", etc.
expected fix: allow the project admin to enforce the style of layout
between:
- many directories containing files named after
languages ISO codes and ONE POT file each
- one directory per language and many pairs of
PO/POT files in each
Yes, Pootle has to be able to handle that.
- when a language has only complete files but is missing translations in some directories, it shows up as 100% in overall stats while, IMHO, it is below 100% because of the missing files. Extreme case: if French has only ONE translation in my 700+directories, it will show up as 100% while it should appear as very lowworkaround: create empty PO files for missing files but that clutters the tree a *lot* (56 lanbuages x 700 directories) expected fix: have pootle automagically consider that a missing file is a 0% file (using the number of strings in the POT to calculate the number of strings).
Again, I think that needs to be an stats option. Some projects will want to see stats relative to all existing POT files (how many original strings have been translated?), while others will want to see stats relative to all existing PO files (how many translation files are complete?).
We need both of those capacities.from Clytie (vi-VN, Vietnamese free-software translation team / nhóm Việt hóa phần mềm tự do)
http://groups-beta.google.com/group/vi-VN
PGP.sig
Description: This is a digitally signed message part
------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________ Translate-pootle mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/translate-pootle
