__year search trouble on 1st of january

[Dirk Datzert <dummy@…>]

Hi,

I detected in a data_hierarchy filter a strange behavior of dates at 1.1. of a year. I have a field birthday defined as a models.DateField(...). I'm working with a sqlite3 database, python2.5 and django-trunk.

Selecting in the admin view the year say 1990 all dates above 1.1.1990 are selected but not the 1.1.1990 himself.
Maybe this was introduced by the [4505] patch!?

Regards,
Dirk

[Dirk Datzert <dummy@…>]

this would fix my problem

[Malcolm Tredinnick]

I'm pretty sure this is SQLite-specific and we've had similar bugs in the past. The problem is the SQLite does not treat "2007-01-01 00:00:00" the same as "2007-01-01" and we really need to be querying with the latter. Ticket #1460 shows the previous occurrence.

I think the change in this patch should not restrict the change just to the start date. That feels a bit asymmetrical. Shouldn't the final test also be "< (year+1)-01-01" ?

[Nebojsa Djordjevic <nesh@…>]

This is not just a admin problem any foo__year query is affected. I hit it today in my non-admin related code.

Replying to mtredinnick:

I think the change in this patch should not restrict the change just to the start date. That feels a bit asymmetrical. Shouldn't the final test also be "< (year+1)-01-01" ?

Here is my stab at this, patch now checks <(year+1)-01-01 and I added some tests also (in tests/regressiontests/datatypes/models.py hope that's OK)

Also, I found another place with 00:00:00 in ​_sqlite_date_trunc -- should we change this also?

(btw. should I change any of the triage fields after submitting patch?)

[Nebojsa Djordjevic <nesh@…>]

patch + tests

[Nebojsa Djordjevic <nesh@…>]

whoops, old patch

[Nebojsa Djordjevic <nesh@…>]

Strange, now I got:

File "/Users/nesh/Documents/workspace/django/tests/modeltests/basic/models.py", line ?, in modeltests.basic.models.__test__.API_TESTS
Failed example:
    Article.objects.filter(pub_date__year=2004)
Expected:
    []
Got:
    [<Article: Area woman programs in Python>]

Looking into it ...

[Nebojsa Djordjevic <nesh@…>]

Strange, now is OK, maybe probem was with %s-1-1??

I added some future/past checks just to be sure.

[Nebojsa Djordjevic <nesh@…>]

latest ;)

[James Bennett]

#5365 was a duplicate.

[James Bennett]

See also #5504 which is covering discussion of testing for these cases.

[James Bennett]

#6141 was a duplicate and also had some patches.

[Malcolm Tredinnick]

(In [7150]) Fixed #3689, #5223 -- Fixed "1st of January" comparisons for SQLite without breaking the other backends.

Based on a patch from raminf and tests from Nebojsa Djordjevic.