i18next-json-sync used in i18n:sync-keys script removes correct plurals from translations

[javnik36]

Not the bug with the app itself, but worth noting imo. Running i18n:sync-keys script removes correct plural forms from translations.

E.g. polish has 3 plural forms - so adding them manually to the translation json and running the script afterwards will remove them if they were never present in English original. Git diff as example:

Not entirely sure what is going wrong in this case - my manual additions, module used in the script or something else 😆 .

[fspoettel]

Good catch, seems like this is an issue with the cli we use: jwbay/i18next-json-sync#22 I'll check out the alternative mentioned in the thread and see if we can migrate to it.

[fspoettel]

The alternative is even worse in a brief test 😅 ... guess we need to roll our own here.

[fspoettel]

@javnik36 I created a little script to do this for us in #180. Feel free to give it a test and lmk if this works for you.

[javnik36]

Ok, I dug deep into this while checking your new script's output...maybe too deep :D Sorry for massive amount of text...

First of all, I will write from my perspective what I would expect while syncing original English file to my locale:

• script knows what plurals suffixes my locale uses and therefore:
  • adds keys for all suffixes my locale uses for given key (regardless of how many they are specified in original),
  • removes keys for suffixes my locale don't use - also removes key without a suffix (afaik used as fallback translation instead switching to English original - imo, it will be hard to spot that mistake later on)
• (bonus: throws warnings when there's missing plural form in original)

So, for example:

#en.json
"key1_one": "one",
"key2_one": "two",
"key2_other": "two_other",
"nonplural_key": "nonplkey"

#pl.json
"key1_one": "one_pl",
"key1_few": "one_few_pl",
"key1_many": "one_many_pl",
"key1_other": "one_other_pl",
"key2": "two_pl",
"key2_other": "two_other_pl"

I expect after syncing:

#pl.json
"key1_one": "one_pl",
"key1_few": "one_few_pl",
"key1_many": "one_many_pl",
"key1_other": "one_other_pl",
-"key2": "two_pl",
+"key2_one": "two",
+"key2_few": "two_other",
+"key2_many": "two_other",
"key2_other": "two_other_pl",
+"nonplural_key": "nonplkey"

I was looking for this behavior on locales there are right now available on the app on i18-locales-sync and script from #180. I didn't check original i18-json-sync script, as we already know it sucks xD

For me it looks like the i18-locales-sync almost fit my expectations (of course except bonus warning), but for some reason it doesn't like korean - it always removed all keys that have with plurals (according to this korean has 1 form - so I would expect either create/leave forms with "_other" suffix or without any suffix - reported this as felixmosh/i18next-locales-sync#40). Another downside is that list of synced languages must be specified in the command, so it will require a little attention while introducing a new translation.

#180 script as far I can tell, leaves all translated suffixes as is, but do not add them based on target locale.

Of course it is up to expectations - your script is undoubtedly better than the original i18-json-sync, as does not removes good stuff. It is up to you what's next - I would propose merging #180 immediately and maybe switch to i18-locales-sync long term - if my issue ends up being an issue and will be fixed 😄 .

[javnik36]

Oh shoot, I did took me so long that you already did that 😅 Hope that whatever I wrote is still somehow relevant xD

[fspoettel]

@javnik36 yeah it's relevant, thanks for the deep dive and opening an issue on the locales project. It mangled the ko locale for me as well, but I did not look into it further. 😅

I see your point re: creating the plurals automatically, that would be ideal. Can you ping back here once the issue gets solved upstream? I think the DX issue with adding locales we can solve quite easily with a wrapper script that reads the LOCALES constant.

[fspoettel]

btw in the mean-time, another PR was opened that relates to this #181 - yesterday I learned that russian plurals are quite special, so I think this is relevant for more locales than pl.

[javnik36]

Sure thing, will let you know.

Oh yeah, I discovered today that formally in pl we have more plurals than I though - in some cases plural for fraction numbers can be different than any other 😅 (though I have no idea if 18next supports that).