#23098 closed Bug (fixed)
ugettext_lazy gets translated too early in modelForm field choices
Reported by: | Mathieu Agopian | Owned by: | nobody |
---|---|---|---|
Component: | Forms | Version: | 1.7-rc-1 |
Severity: | Release blocker | Keywords: | |
Cc: | Florian Apolloner | Triage Stage: | Ready for checkin |
Has patch: | yes | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
This ticket first was submitted against the pytest_django repository: https://github.com/pelme/pytest_django/issues/135
However, I think the maintainer is right and it's a Django bug: because of the new feature allowing to customize the "empty choice", the lazy translation of named groups in a choice list for a modelForm field gets translated at import time.
The steps to reproduce are in the ticket linked above, and consists in:
- have a choice list for a model field which contains name groups, with the groups' names being translated (using ugettext_lazy)
- create a ModelForm for this model
- import the ModelForm from an external script that has the settings configured, but not called
django.setup()
yet
I understand the steps to reproduce are really a corner case and may seem irrelevant, but this bug means that we might have lazy translations that are translated way to early (at import time) instead of staying lazy.
With the normal Django usage, this will translate too early silently, and not raise any error exception.
From what I understand, this happens in the code that checks if the choice (in our case, the name of a group) is the empty choice: https://github.com/django/django/blob/master/django/db/models/fields/__init__.py#L732
I can imagine two solutions to this issue:
1/ only do the check if the "choice" isn't a group name
2/ do something like choice is None or isinstance(choice, six.text_type) and choice == ''
Change History (10)
comment:1 by , 10 years ago
Component: | Translations → Forms |
---|---|
Severity: | Normal → Release blocker |
Triage Stage: | Unreviewed → Accepted |
comment:2 by , 10 years ago
Has patch: | set |
---|
comment:5 by , 10 years ago
Resolution: | → invalid |
---|---|
Status: | new → closed |
This report seems invalid to me. The reproduction instructions include:
import the ModelForm from an external script that has the settings configured, but not called django.setup() yet
But the 1.7 release notes explain that you should call django.setup()
before doing anything else.
comment:6 by , 10 years ago
Resolution: | invalid |
---|---|
Status: | closed → new |
Mmm, regardless, if the translation is resolved at import time, that's bad. We should at least fix that.
comment:7 by , 10 years ago
Aymeric, I'm sorry that I confused things by talking about the pytest usage in the STR, while it actually has nothing to do with the issue. It's just that I discovered this issue in this particular case, while it was just "failing" (translating to early) silently otherwise.
As you said, the issue is that the translation is resolved at import time, and that's the point of this ticket (hence the title), sorry if I made it harder to understand by providing to much (or confusing) details.
comment:9 by , 10 years ago
Resolution: | → fixed |
---|---|
Status: | new → closed |
PR: https://github.com/django/django/pull/2959