Text format for translations

[puzrin]

For user, translations are expected to be in text format:

• Can be edited manually
• Can be exported/imported to web translation services
• Easy to track diffs

C sources will be generated by compiler utility, and to expected to be edited.

Default file type is YAML. Alternative - JSON (will be supported automatically as YAML subset).

Format

TBD

[puzrin]

Format. Something like this:

'en-GB':
  Nut: [ Nut, Nuts ]

'ru-RU':
  Settings: Настройки
  Nut: [ Гайка, Гайки, Гаек ]
  'LVGL is awesome!': ~ # null, added by extractor as mark to translate

Notes:

• for default locate plurals are enougth
• plural names (zero, ``one, two, few, many, other) are not used, because sequence is strict for each language, and optional (skippable) forms are at the end.

[kisvegabor]

Looks good to me. en_GB and ru_RU should be compiled to C file and array names as I suggested here (I used different naming but doesn't really matter)

[puzrin]

http://docs.translatehouse.org/projects/translate-toolkit/en/latest/formats/index.html - hear is nice list of available translations formats

My suggestion is close to Ruby, by they supports plurals in different way: https://github.com/translate/translate/issues/3615

Still not sure what is better:

• Arrays are comfortable for manual edits, when no web translators used.
• Named plurals are better for web translators, because standard format should be supported for import/export

Probably we just need to support both and add format option to scripts. Need to dig web translator docs to decide.

[puzrin]

For info: at web projects plurals are ~ 15% of phrases.

Ruby's way:

'en-GB':
  Nut:
    one: Nut
    many: Nuts

'ru-RU':
  Settings: Настройки
  Nut:
    one: Гайка
    few: Гайки
    many: Гаек
  'LVGL is awesome!': ~ # null, added by extractor as mark to translate

May be, not so bad...

[kisvegabor]

Can one, few, many or similar expressions describe complicated plural cases?

Have look at Slovenia here: https://www.gnu.org/software/gettext/manual/html_node/Plural-forms.html

[puzrin]

Can one, few, many or similar expressions describe complicated plural cases?

Yes. See https://github.com/nodeca/plurals-cldr#references

I did plurals codegen for js. It uses Unicode CLDR data to build code & tests automatically. Works with all languages.

We just will use a bit different template to generate C code instead of JS. Not a big deal.

[kisvegabor]

I see, fine!

[kisvegabor]

It seems to we still need a config file where we describe the languages. For example:

• en_GB
  • number of plurals
  • plural rule (one-line rules here )
  • and what comes later
• ru_RU
  • ...

The yaml files for the translations should be created based on the languages declared in this file.

[puzrin]

yaml is self-sufficient. It's created by user and extended by user/extractor. I can't imagine extra info, missed in yaml and required to update is.

• number of plurals
• plural rule (one-line rules here )

This is NOT for yaml file. This is for generated C source. Different abstraction layer. Info is generated automatically from locale name, CLDR data and so on.

[kisvegabor]

And where will be the mentioned info (number of plurals, plural rules) stored to automatically generated from the local name?

[puzrin]

Generated translations.c can contain any additional data you may wish for proper work.

• plural rules functions will be generated from CLDR
• plurals/any counters... - this can be added to compiler too, if needed.

I guess, this is more for #7.

[puzrin]

Close - landed to docs