Community Translations

[Dastgir 1246]

We have introduced HULD(Hercules Ultimate Localization Deisgn) quite a few months ago, but there's no translation or anything yet...

I suggest to use https://poeditor.com/ , it have nice integration with github(you can see it in their web)

Also with OSI License , we get unlimited strings to translate.

@@Haru
@@Mysterious

[Haru 290]

I signed up for a test. Both poeditor.com and crowdin.com seem promising, both offer free accounts for OSI-licensed software. Crowdin integrates with the POEdit desktop app.

[4144 364]

Also exists https://www.transifex.com It even was open source, but latest versions look like is not.

Allow open source projects translation for free.

[Haru 290]

Just made an account to try transifex as well.

 

Here's my feedback so far:

 

poeditor

project url: https://poeditor.com/join/project/Dw7RGEwo1A

Good:

- it integrates with github, and it can pull .pot files (and changes to it) from there

Bad:

- Even with the account that should allow unlimited strings, it doesn't load more than 30,000 strings per file (we have about 120,000).

 

The integration idea is nice in theory, but considering that the .pot file we generate is over 20MB big, it's a bad idea to keep it in the repository. We'd need to make a separate repository for it. But the 30,000 strings limit makes it all useless.

 

 

crowdin

project url: https://crowdin.com/project/hercules

Good:

- Seems to load all 120,000 strings

- Offers an API (and an opensource, crossplatform API client), so we could upload the translations via Hercules bot automatically.

- It integrates with the desktop application POEdit, so people who don't want to use the web UI, can use the desktop application.

Bad:

- No automatic git integration

 

I like the API this service offers. It doesn't pull from a git repository, but keeping a 20MB+ translations file in the repository is out of the question anyways. With the API, we can automatically upload from the Hercules API bot after every commit that alters strings. I think it's an ideal situation. The POEdit integration is nice, although I haven't tried it out yet (I will).

 

 

transifex

Good:

- N/A (I was unable to test it)

Bad:

- It seems to have problems with our .pot file (it errors on the r, saying it's an invalid character - it can't handle carriage returns)

 

Unfortunately I was unable to try this out, because of the r error.

[4144 364]

about transifex, not sure abour r, but n works fine. It also allow update pot file from git.

 

Also i checked generated pot file, look like it broken. it write r as character 'r' but i think it must write it as two chars r.

Same issue with "n"

you can validate pot file by gettext:

msgfmt -v generated_translations.pot

[Haru 290]

Oh, you're right, that's fixed in the huld-improvements branch, I forgot about that.

 

The string it's choking on is

msgid "by pressing the '^4D4DFFCtrl^000000' and the '^4D4DFF - ^000000,' '^4D4DFF = ^000000,' or '^4D4DFF  ^000000' keys."

I'll fix that as well (and all other escape sequences)

[4144 364]

here error too.

gettext say error in "" i think need escape "" in text it must be ""

[Haru 290]

Yep... we're not passing the strings through sv_escape_c.

[Haru 290]

An update on transifex: now gettext validates the file correctly (see huld-improvements branch), but transifex still can't load it. The error message is now "'utf8' codec can't decode byte 0xa1 in position 4915429: invalid start byte". I believe it gets confused by the strings in CP949/EUC-KR encoding (which are invalid if the file is read as utf8)

[Dastgir 1246]

@@Haru

So PoEditor don't give unlimited string even for open source projects?

[4144 364]

@@Haru, yes in utf8 cant be any chars with set high bit. Probably before upload pot file need convert it from CP949 to utf8?

And for using with server convert it back?

[Haru 290]

@@Dastgir yes, strings are "unlimited" (meaning that you can add as many translations as you want - because translated strings count against your plan's limit), but it doesnt seem to be able to load more than 30,000 *source* strings.

 

@@4144 the problem is that Hercules doesn't care about the file encoding... We just parse the strings and send them as they are to the client (and the client prints them according to the selected langtype, as it used to happen in the windows 9x days). This means that the strings ee read aren't necessarily cp949, but they may be in any codepage or even in unicode already, depending on who wrote the scripts and for what client (I write them in utf8 for example, since I use an utf8 capable client).

Since we don't know what codepage the source files are, we can't reliably convert them. If we want to do this, we could perhaps specify it somewhere in the config, and implement some codepage-aware handling of strings. This might add some dependencies a bit too heavy though (iconv? x.x)

[4144 364]

if no codepage, this mean no editor or service what can show all lines in correct format.

Probably better store in utf8, because if store in cp949, for example russian or other locales not supported at all.

But client look like not support utf8...

 

Probably only solution convert pot file from any valid for it codepage to utf8 and store it in repository in utf8. services can translate it and create po files in utf8 too.

After convert tranlsations (po files) from utf8 to codepages what supported by client. for example russian from utf8 to cp1251. And store po files in repository in this codepages.

[maqc1 8]

How exactly does those websites work? I'd be glad to help with translations.

 

If this can help... In my experience with accentuated characters in RO, Windows-1252, Windows-1255 and ISO-8859-8 encodings are the only one displaying correctly in game (with roman characters at least ; no idea about russian and asian char).