Fix wrong sorting of parsed language header #2696
Conversation
There was a problem hiding this comment.
Apart from the code change request, this bug PR is missing a test.
I'd suggest using openslides_backend/tests/system/action/meeting/test_create.py lines 525 (just use the header_language string that causes the bug as a value) and 544 (_(name)) as a reference for how translation can be set up and used and writing a simple test in a completely new test file in a central place somewhere in the tests folder.
| @@ -56,7 +56,7 @@ def parse_language_header(self, lang_header: str) -> list[str]: | |||
| code = code.split("-")[0] | |||
| result.append((q, code)) | |||
| # sort by quality value and return only the codes | |||
| return [t[1] for t in sorted(result)] | |||
| return [t[1] for t in result] | |||
There was a problem hiding this comment.
| return [t[1] for t in result] | |
| return [t[1] for t in sorted(result, key=lambda tup: tup[0])] |
The sort did not only sort the languages in alphabetical order: It sorted them by the weight value q first. If that wasn't given or equal to another value, it sorted them in alphabetical order.
I will agree that the latter is probably a bug (in this case caused by the way python sorts tuples), but the former definitely isn't.
The above code should sort the languages by weight first and leave empty values in the same order as before.
If, during testing, it turns out to wrongly prioritize or wrongly not-prioritize languages with no q-value over those with q-value, I'd suggest trying to set q to "0.0" instead of "1.0" in line 50.
Fixed partly #2695
Prefered language settings of browser (weighted by user, e.g. [de,en,cs]) should not sorted! In this example: [cs,de,en] -> cs translation is empty for password reset strings -> backend send empty email strings!