Python format check not applied to some plural forms

[VladimirSlavik]

Describe the issue

Hello, I have a source string which is a plural using python format (the percent kind). In weblate, I have turned on "python format" as a mandatory check. Somebody just translated the string, and did not use the percent format. There is nothing shown on the string page indicating this is a problem, and this translation is pushed into the repository.

Detailed information below.

I already tried

• I've read and searched the documentation.
• I've searched for similar issues in this repository.

Steps to reproduce the behavior

1. Go to https://translate.fedoraproject.org/translate/anaconda/rhel-9/ar/?q=%25d+_storage+device+selected&sort_by=-priority%2Cposition&checksum=
2. Observe Flags: python-format, lack of %d in some of the translation plurals, but no related warning or error.

Expected behavior

There should be some indication that the check failed.

Screenshots

No response

Exception traceback

No response

How do you run Weblate?

Other

Weblate versions

@yurchor can you please fill in the missing parts?

Weblate deploy checks

No response

Additional context

Detailed information:

• I am using the weblate instance at https://translate.fedoraproject.org, which runs version 4.18.2.
• The string is in Arabic, project anaconda, component rhel-9: singular %d _storage device selected, plural %d _storage devices selected
• Weblate correctly shows python-format in flags, and highlights the %d.
• The Arabic translation has these:

  لا جهازَ تخزينٍ مُحدَّدٌ
  جهازُ تخزينٍ واحدٍ مُحدَّدٌ
  جهازا تخزينٍ اثنان مُحددان
  %d‏ جهاز تخزين محدد
  %d‏ جهاز تخزين محدد
  %d‏ جهاز تخزين محدد
  

• When I run msgfmt -c, it detects correctly that this particular string's translation is wrong.
• The same string has been a problem already in a different component, the scenario was the same.

[yurchor]

Actually, I do not know what can be added here. It is hard to tackle with languages with many plural forms. It is allowed to skip %d for singular but I do not know how to treat this for dual and other forms in Arabic. It should be inspected by Weblate developers.

[VladimirSlavik]

Sorry, I meant the details the bug template asks about deployment (versions, deploy checks). But maybe that does not matter?

[nijel]

Weblate allows omitting the format string when there is only a single value for the plural form. So for Arabic, it is okay to have six plurals as:

• None storage device selected
• One storage device selected
• Two storage devices selected
• %d storage devices selected
• %d storage devices selected
• %d storage devices selected

This is inconsistent with msgfmt -c, but allows nicer translations, and most frameworks support this behavior.

[VladimirSlavik]

Thank you, that answers why this happens.

I suppose there is no way to adjust this on a component level?

Is this check done by something that I can swap with using msgfmt? I found PythonFormatCheck in https://github.com/WeblateOrg/weblate/blob/main/weblate/checks/format.py, which does not appear to be easy to isolate, so the answer there is again no?

[nijel]

I've just implemented flag to enforce such format strings in #9695

The Weblate checks are not easy to isolate because they depend on the database. You can use enforced checks in Weblate to make sure that no broken strings are committed as translations, see https://docs.weblate.org/en/latest/admin/projects.html#enforced-checks

[github-actions]

Thank you for your report; the issue you have reported has just been fixed.

• In case you see a problem with the fix, please comment on this issue.
• In case you see a similar problem, please open a separate issue.
• If you are happy with the outcome, don’t hesitate to support Weblate by making a donation.

[VladimirSlavik]

Thank you! That was unexpected...

I was looking at msgfmt instead, trying to see if I could make the relaxed mode for plurals an option there. Perhaps that would be a more reasonable thing in context of the whole ecosystem - as you say, many other parts allow this, and for some format kinds it's technically sound to omit things.