Improve pluralization of translated strings

[rmol]

Status

Ready for review

Description of Changes

Switch a few strings containing numeric variables to ngettext from gettext, to ensure they can be properly translated. Use "num" more consistently as the placeholder for numeric variables in strings marked with ngettext, to make things easier for translators.

Fixes #5566.

Testing

• git checkout -b more-plurals origin/more-plurals
• make translate
• Choose a test language, e.g. de_DE and edit its .po file (securedrop/translations/de_DE/LC_MESSAGES/messages.po) to find the updated source strings, remove the fuzzy markers, and correct the placeholder names.
• In that .po file's directory, run msgfmt messages.po.
• Run make dev.
• Visit the journalist interface, switch to your test language, and ensure that the strings are being translated correctly:
  • Visit the admin page to add a new user.
    • Enter a HOTP secret a few characters long.
    • Enter less than three characters in the username field.
    • Enter more than 100 characters in the first and last name fields.
    • Pressing add user should result in properly-translated error messages.
  • Visit the source list and verify that the document, message, and unread counts are being properly translated. You'll need to submit a single file and document as a new source to test singular translations.
  • Delete a single source from the source list. The {num} collection deleted message should be properly translated.
  • Delete more than one source from the source list. The {num} collections deleted message should be properly translated.
  • Delete a single source submission. The Submission deleted flash message should be properly translated.
  • Delete more than one source submission. The {num} submissions deleted flash message should be properly translated.

Deployment

The updated strings will need to be translated at the next release. Until then, they'll always appear in English.

Checklist

If you made changes to the server application code:

• Linting (make lint) and tests (make test) pass in the development container

If you made non-trivial code changes:

• I have written a test plan and validated it for this PR

[emkll]

Thanks @rmol test plan as described is passing, however encountered an error locally prior to updating the translations. Without the translations generated, an error is thrown (likely due to the name of the variables changing) but is being caught by the API error handler in https://github.com/freedomofpress/securedrop/blob/more-plurals/securedrop/journalist_app/__init__.py#L92:

  File "/home/user/src/securedrop/securedrop/journalist_app/__init__.py", line 92, in _handle_http_exception
    handler = list(app.error_handler_spec['api'][error.code].values())[0]
AttributeError: 'KeyError' object has no attribute 'code'

Error handling here might warrant further investigation and/or we we consider strings as as part of this PR to support multiple languages in develop (prior to merging string changes). Right now, on this branch, using any other language than English on the source list of the journalist interface will result in error 500.

[rmol]

I've updated the translations. There should be no more errors because of the changed strings, and in the next translation period, translators should only have to update the strings newly marked with ngettext, and only in the languages with multiple plural forms.

[emkll]

(rebased on latest develop)

[emkll]

Looks like translation tests are failing after the rebase on latest develop, it is likely due to the Tor Browser update introduced in #5578

[rmol]

I think it was actually my fix for Redis dump file version incompatibility. The translation-test script never sets REPOROOT, breaking the run-redis function and causing the Redis connection errors and subsequent Selenium timeouts. I added a commit to make sure it does get set in dev-deps before it's used, and so far so good.

[rmol]

Nope, after rebasing, also had problems running the newly-type-hinted i18n_tool.py under Python 2 (ahem), so updated the translation test job to use Python 3. About time. Translation tests are now passing.

[emkll]

Thanks @rmol looks good to me, went through the diff and the functional test of the translations:

• Visit the journalist interface, switch to your test language, and ensure that the strings are being translated correctly:
• Enter a HOTP secret a few characters long.
• Enter less than three characters in the username field.
• Enter more than 100 characters in the first and last name fields.
• Pressing add user should result in properly-translated error messages.
• Visit the source list and verify that the document, message, and unread counts are being properly translated. You'll need to submit a single file and document as a new source to test singular translations.
• Delete a single source from the source list. The {num} collection deleted message should be properly translated.
• Delete more than one source from the source list. The {num} collections deleted message should be properly translated.
• Delete a single source submission. The Submission deleted flash message should be properly translated.
• Delete more than one source submission. The {num} submissions deleted flash message should be properly translated.

I've noticed that for cs, ro, ru and sk, the fuzzy attribute is preserved in the translations of the strings that were touched. was this intentional?

[emkll]

Approving based on previous review, as discussed fuzzy translation for those four languages is expected.