Implement time filters

[Vincent Foley]

As of r8442, it is possible to filter a queryset by year, month or day using the usual syntax,
However, filters for hours, minutes and seconds were missing. I have added those into a patch for your consideration along with unit tests and documentation.

Meeting.objects.filter(meeting_time__hour=12)

[Thomas Kerpe]

Duplicate of #8423 other closed.

[h3]

I can understand why this patch as been scheduled for post-1.0, to respect the API freeze, but it should really be included in the 1.0 release.

Being able to filter by time is not a trivial or candy feature, it *should* be possible, we don't use dateTIME fields just for the fun of it.

Furthermore, in pre-1.0-beta was possible to "hack" a time filter like this:

 @register.filter
 def filter_by_hour(obj, time):
     return obj.filter(start_time__contains=' %02d' % time.hour)

Since the beta is out (I don't know the exact revision sorry), it can't be done. It looks like a problem with the database wrapper.. I don't know if it's intentional or not.

Anyway, unless someone can point me out another easy way to filter a datetime field by time, I think this patch is primordial.

Regards

[Malcolm Tredinnick]

Feature freeze has passed. This isn't a bug. It's an enhancement request.

[(none)]

Milestone post-1.0 deleted

[David Grant]

Any chance this could get in to 1.1?

[Russell Keith-Magee]

I've needed these more than once. I can't see any reason to not have them.

Patch needs to be updated; also needs some real tests. Three queries that return nothing doesn't constitute at test of anything.

[Ulrich Petri]

(djangocon.eu 2010 sprint)
I'll give this one a shot

[Timothée Peignier]

Updated patch

[Timothée Peignier]

I've added an updated patch with more tests (and I'll happily add more if needed).

P.S: Patch is in git diff format.

[Ramiro Morales]

Last thunk of latest patch

--- a/tests/regressiontests/null_queries/tests.py
+++ b/tests/regressiontests/null_queries/tests.py
@@ -53,17 +53,17 @@ class NullQueriesTests(TestCase):
         """
         obj = OuterA.objects.create()
         self.assertQuerysetEqual(
-            OuterA.objects.filter(inner__second=None),
+            OuterA.objects.filter(inner__secondary=None),
             ['<OuterA: OuterA object>']
         )
         self.assertQuerysetEqual(
-            OuterA.objects.filter(inner__second__data=None),
+            OuterA.objects.filter(inner__secondary__data=None),
             ['<OuterA: OuterA object>']
         )
 
         inner_obj = Inner.objects.create(first=obj)
         self.assertQuerysetEqual(
-            Inner.objects.filter(first__inner__second=None),
+            Inner.objects.filter(first__inner__secondary=None),
             ['<Inner: Inner object>']
         )
  

reveals something potentially dangerous. FTR, from a django-dev IRC channel discussion about it:

cramm: hrm, looking at the patch for #8424.
cramm: Specifically the last thunk. Adding new filtering lookups like 'hour', 'second', ...  means there could be clashes with existing identically named model fields?
cramm: I mean, in user apps
Alex_Gaynor cramm: yeah, it would
Alex_Gaynor: I'd much rather wait for us to have a real solution to alternate filtering, but that won't be 1.4 at the earliest
carljm: cramm, Alex_Gaynor: dunno, I'm inclined to think these are obvious omissions.
carljm: i.e. "wait till we have custom filtering" doesn't make sense to me.
Alex_Gaynor: carljm: unless it breaks everyone's fields named hour, second, minute
carljm: yes. that, of course, is icky.
carljm: that shouldn't be a problem, though. can't the splitter be smarter?
carljm: i.e. only DateTimeFields can have these lookups
carljm: and only related fields can be traversed.
carljm: you can't traverse across a DateTimeField to a related object.
Alex_Gaynor: carljm: it could, ATM it's not
carljm: right. so it's a "patch needs improvement" situation, then.

[anonymous]

Docs in the patch: "This will match every record with a pub_date that occurred between 3:00 AM and 3:59 AM."

3:59:15 AM is not between 3:00 AM and 3:59 AM but it should be matched.

[Ulrich Petri]

Related: #16187 Refactor of the lookup system

[anonymous]

#18049 was a duplicate.

[Aymeric Augustin]

Fixed in e74e207cce54802f897adcb42149440ee154821e.

[anonymous]

why not directly expose __time as well as __hour, __minute, __second?