Opened 13 years ago
Last modified 10 years ago
#18148 new New feature
django.contrib.sites.managers.CurrentSiteManager should be able to span multiple models
Reported by: | Rory Geoghegan | Owned by: | nobody |
---|---|---|---|
Component: | contrib.sites | Version: | 1.4 |
Severity: | Normal | Keywords: | |
Cc: | sfllaw@…, krzysiumed@…, Rory Geoghegan | Triage Stage: | Accepted |
Has patch: | yes | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | yes |
Easy pickings: | no | UI/UX: | no |
Description
The current implementation of CurrentSiteManager can only be used with models that have a ForeignKey or ManyToManyField reference to a Site model. Because get_query_set uses the built-in filter syntax, you could specify remote model references by using a double underscore, except that the validation code breaks.
For example, you should be able to do:
class Foo(model.Model): bar = models.ForeignKey('Bar') on_site = CurrentSiteManager('bar__site')
and it should do the right thing.
Attachments (3)
Change History (12)
by , 13 years ago
Attachment: | sites.diff added |
---|
comment:1 by , 13 years ago
Cc: | added |
---|
comment:2 by , 12 years ago
Component: | Uncategorized → contrib.sites |
---|---|
Triage Stage: | Unreviewed → Accepted |
Type: | Uncategorized → New feature |
comment:3 by , 12 years ago
Patch needs improvement: | set |
---|
The patch looks really well, but some error messages could be better:
If ArticleComment
model (from regressiontests.sites_framework.models module) is changed in this way:
on_site = CurrentSiteManager("parent_article__non_existing_field")
then the error message is
ValueError: CurrentSiteManager couldn't find a field named parent_article__non_existing_field in ExclusiveArticle.
The field name is non_existing_field
, but the message says it's parent_article__non_existing_field
.
The other situation when the error message is not meaningful enough is when the field exists but it's neither ForeignKey
nor ManyToManyField
. See example:
TypeError: invalid_field must be a ForeignKey or ManyToManyField.
The message doesn't say name of the model where invalid_field
was defined.
I also suggest using first_article
and second_article
instead of article
and article2
. It's not more meaningful, but it's easier to see the difference.
comment:4 by , 12 years ago
Cc: | added |
---|
comment:5 by , 12 years ago
Cc: | added |
---|
comment:6 by , 12 years ago
@krzysiumed: I updated the patch after your comments with the changes you requested.
by , 12 years ago
Attachment: | 18148_v3.diff added |
---|
comment:7 by , 12 years ago
The patch seems to be ready to check in, but I would like to suggest
some ideas. I attached patch including them.
The documentation is slightly unclear -- it's not obvious to which
paragraph refers the 'Changed in Django 1.5
'. I added subsections
which make it less ambiguous, but the titles (especially 'Spanning relationship
')
could be better. I also changed the section to which
the only link points. The previous target was 'Field lookups
' section
which describes lte
(e. g. pub_date__lte
), exact
(e. g. headline__exact
) and other features. I think that 'Lookups that span relationships
'
section is more appropriate. Finally, I
would be happy if the patch were reviewed by a native speaker.
The changes in code (and tests) are usually renamings and following
PEP8
. I renamed first_article
and second_article
to
article_inside
and article_outside
(I did the same about comment
and comment2
). The idea is to make the names more meaningful but I'm
not sure if the goal was achieved.
These ideas are propositions so feel free to reject them if you want.
comment:8 by , 12 years ago
Patch needs improvement: | unset |
---|
comment:9 by , 10 years ago
Patch needs improvement: | set |
---|
Patch for django-trunk revision 17913