Pluralize localizable strings

[1ec5]

UI strings that include counts now use the appropriate plural form in English and (once translations are updated) in other languages. When looking up a translation and one of the tokens is numeric, look one level deeper in the translation file for the CLDR plural form corresponding to the numeric token.

In order to support this feature for all the languages that iD supports, with all their varying pluralization rules, this change introduces dependencies on:

• cldr-core, which supplies plural rules for a number of languages (part of cldr-data)
• CLDRPluralRuleParser, which parses a plural rule, determining whether it matches the given number (used by MediaWiki)

This works locally using npm start, but I’d like someone to double-check that I’m including these dependencies correctly. (All we need is a single JSON file from cldr-core, not the entire package.)

Keying off numbers in token replacements means any number formatting that occurs before calling t() will defeat the pluralization logic therein. Unlike the unit formatting that occurs in t()’s callers (as of #4672), number formatting needs to happen inside t().

More to come:

• Depend on cldr-core and CLDRPluralRuleParser
• Look up pluralized translations in t()
• Pluralize some strings in core.yaml
• Move number formatting into t()
• Combine manually pluralized strings like restriction.controls.via_up_to_one/restriction.controls.via_up_to_two
• Break up validations.many_deletions into strings with one pluralized word each
• Mention plural format in contributing documentation
• Update changelog

Fixes #597.

[bhousel]

This works locally using npm start, but I’d like someone to double-check that I’m including these dependencies correctly. (All we need is a single JSON file from cldr-core, not the entire package.)

Looks great! Rollup will include only the single JSON file 👍

Keying off numbers in token replacements means any number formatting that occurs before calling t() will defeat the pluralization logic therein. Unlike the unit formatting that occurs in t()’s callers (as of #4672), number formatting needs to happen inside t().

Ok, that makes sense. I don't know how often in the code we pass stringified numbers to t() but it is probably not too much work to change these.

[magol]

Any work on this PR?

[1ec5]

Sorry, I don’t know when I’ll get a chance to get back to this PR, but it’s certainly something I’d like to get in eventually. (At some point, browser support for Intl.PluralRules would improve enough that we could radically simplify this PR.)

[quincylvania]

To be clear, this was closed automatically when I renamed the master branch, but it was two years old.

I'd like to spend some time soon to get pluralization right as we work on expanding iD's global usability (see #7713).