Allow QuerySets to handle offsets without limits and negative slicing

Finishing that last sentence of the topic: "... are included in the patch"

That Oracle comment needs to be changed: the QuerySet? subclass that was in query.py at one point wound up back in base.py. I'll take a look at supplying an Oracle patch for this.

Negative indexing patch of django/db/backends/oracle/base.py

One of the test cases for this is failing:

In [63]: Article.objects.all()[-5:-2][-1]
---------------------------------------------------------------------------
<type 'exceptions.IndexError'>            Traceback (most recent call last)

/home/ikelly/projects/testproject/<ipython console> in <module>()

/home/ikelly/projects/django.trunk/django/db/models/query.py in __getitem__(self, k)
    216                 try:
    217                     return list(self._clone(_offset=k, _limit=1,
--> 218                                             **extra_clone_args))[0]
    219                 except self.model.DoesNotExist, e:
    220                     raise IndexError, e.args

<type 'exceptions.IndexError'>: list index out of range

The sql that the Oracle backend is generating for this case is:

SELECT * FROM
  (SELECT "TEST_ARTICLE"."ID","TEST_ARTICLE"."HEADLINE","TEST_ARTICLE"."PUB_DATE",
          ROW_NUMBER() OVER (ORDER BY "TEST_ARTICLE"."PUB_DATE" DESC, "TEST_ARTICLE"."HEADLINE" DESC ) AS rn
   FROM "TEST_ARTICLE" ORDER BY "TEST_ARTICLE"."PUB_DATE" DESC, "TEST_ARTICLE"."HEADLINE" DESC)
WHERE rn > -2 AND rn <= -1

This query will return no results due to the negative indices. Their presence in the SQL suggests that there's a bug in the way the negative indices are mapped in this case. The same test case also fails using the sqlite backend, but there it simply returns the wrong article rather than generating an error.

(In [7147]) queryset-refactor: Implemented slicing to end of querysets.

Refs #2150, #5012.

(In [7148]) queryset-refactor: Implemented the reverse() method on querysets.

Refs #5012.

Negative slicing isn't possible to implement in a way that is analogous with Python's behaviour without being fairly inefficient. The problem is you have to reverse the SQL query's order, limit the result set and then pull the results back into Python and reverse them again. That last step is the problem. We've just moved away from pulling results into memory until we need them (very important for large result sets), so adding it back just for this isn't worth it.

Instead, I've added a reverse() method to querysets so people can reverse and then slice off the front. It's close enough to negative slicing to provide all the same functionality, without being confusing by looking like Python slicing and not behaving precisely the same.

(In [7477]) Merged the queryset-refactor branch into trunk.

This is a big internal change, but mostly backwards compatible with existing code. Also adds a couple of new features.

Fixed #245, #1050, #1656, #1801, #2076, #2091, #2150, #2253, #2306, #2400, #2430, #2482, #2496, #2676, #2737, #2874, #2902, #2939, #3037, #3141, #3288, #3440, #3592, #3739, #4088, #4260, #4289, #4306, #4358, #4464, #4510, #4858, #5012, #5020, #5261, #5295, #5321, #5324, #5325, #5555, #5707, #5796, #5817, #5987, #6018, #6074, #6088, #6154, #6177, #6180, #6203, #6658