It is important to keep in mind that pyfa translations are not the same as the translations that come from EVE data. These translations are dumped directly from the game, and the pyfa team has no control over them. This includes, but is not limited to:

Market browser
Ship browser
Item names, description, traits, attributes

If there is a tranlation issue in EVE data, you must submit a ticket to CCP instead.

Getting involved

Translations are community-driven. The pyfa team isn't going to

Maintain individual language packs as a part of general development work, or
Delay a release if translations aren't available

This is because the pyfa team is only versed in a few languages, at best. If you are interested in getting involved, then either start a discussion in the language thread (linked above) or, if you're interested in contributing directly, please read the short synopsis of how it works below and submit a Pull Request.

How it works

A quick introduction to GNU gettext translations! There is no programming knowledge required to help with most of the translations. Each langauge that we support has a LC_MESSAGES/lang.po file, and in this file there are multiple groups of msgid and msgstr. msgid is usually the English version that is displayed in pyfa, whereas msgstr would be the translated version.

msgid "Click to toggle between effective HP and raw HP"
msgstr "点击切换有效HP和原始HP"

90% of translations are as simple as that. The other 10% may require you to take a quick dip into the code to tweak the formatting of the string, assign context prefixs (for translations that may translate differently for the same work due to context), or to even add the annotation to the string that will expose it to the translation engine. If you're not comfortable with that, you can always request it!

POEdit

Poedit offers a nice GUI for updating translations.

To update PO file for existing translation

open a existing locale/ll_CC/LC_MESSAGES/lang.po
Catalog -> Update from POT file
select pre-prepared lang.pot file

To translate and generate MO file

edit the translation and hit Save :)

FAQ

Q: The English text isn't in the .po file for me to translate
A: This is probably one of two things:

Missing annotations in the source code. All text that needs to be translated needs to be wrapped with _t() to make it locale-aware
Out of date .po file. As pyfa development continues, the .po file may fall behind. See next question.

Q: How do I update the .po file for my language?
A: See Commands section below for a number of useful commands

Q: I run pyfa in Linux but the translations don't work
A: If you're running from source / your own method, this is because the .mo files aren't checked into the repo and thus aren't available by default. Running python3 scripts\compile_lang.py should compile all language files. If you're running from a package from a third-party repository, YMMV - please contact the maintainer of that package.

Commands

Below is a summary of GNU gettext manual, adapted for Pyfa i18n workflow.

Poedit offers a nice GUI for same GNU gettext translation workflow.

i18n with command line

Windows users can get these tools via Git for windows, Msys2 or Cygwin; or just use WSL / WSL2. For Linux and macOS users these tools might be available out-of-box.

To generate new template for translation:

$ find gui/ *.py -name "*.py" | xgettext --from-code=UTF-8 -o locale/lang.pot -d lang -k_t -f - -s

explanation:

find gui/ *.py -name "*.py": collect all .py file path in gui folder and all sub-folders, write it to stdout
xgettext: a utility looking for keyword and put string literals in a specific format for human translation
- --from-code=UTF-8: designates encoding of files
- -o locale/lang.pot: let xgettext write to locale/lang.pot
- -d lang: default language domain is lang
- -k_t: besides default keyword (including _, see info xgettext for detail), also look for _t
- -f -: let xgettext to read from stdin, which is connected to find stdout
- -s: sort output according to msgid

this locale/lang.pot is called PO template, which is discarded once actual ll_CC/LC_MESSAGES/lang.po is ready for use.

To initialize PO file for new language

$ msginit -i locale/lang.pot -l ll_CC -o locale/ll_CC/LC_MESSAGES/lang.po -s

explanation:

-i locale/lang.pot: input file location
-l ll_CC: target locale. ll should be a language code, and CC should be a country code
-o locale/ll_CC/LC_MESSAGES/lang.po: output file
- ll_CC: same as above
- LC_MESSAGES: GNU gettext conventional path to search for localized messages
- lang.po: language domain and file format

this locale/ll_CC/LC_MESSAGES/lang.po should be checked into VCS, later it will be converted into mechine readable format (.mo).

To update PO file for existing translation

$ msgmerge -s locale/ll_CC/LC_MESSAGES/lang.po locale/lang.pot

To do actual translation

just edit the lang.po file, either manually or via GUI applications like POEdit

To generate machine readable MO file

For a single locale:

$ msgfmt locale/ll_CC/LC_MESSAGES/lang.po -o locale/ll_CC/LC_MESSAGES/lang.mo

For all available locales:

for f in locale/*/; do 
    msgfmt $f/LC_MESSAGES/lang.po -o $f/LC_MESSAGES/lang.mo
done

Since compiling .po files is useful to everyone, we also have a script in the repo that can do it without the need for the normal gettext tools: python3 scripts/compile_lang.py

To merge 2 or more PO file

$ msgcat -s path/to/old.po [path/to/another.po] -o path/to/new.po

Note that msgcat cannot perform a 3-way merge, it will simply stack translations with same msgid on top of each other. If you use msgcat to merge multiple PO file, please check and fix the output before commit to Git.