Paginator could detect if the object_list is a QuerySet or a list

[batiste@…]

If the Paginator is accidentally used to count a large QuerySet (instead of a QuerySetPaginator), the generated SQL request could be inefficient (all content will be retrieved just to make len on it). Why don't we automatically detect if object_list has a count attribute? The paginator API will be simpler and it will be more difficult to introduce performance issues.

    def _get_count(self):
        if self._count is None:
            if hasattr(self.object_list, 'count'):
                self._count = self.object_list.count()
            else:
                self._count = len(self.object_list)
        return self._count
    count = property(_get_count)

And by the way the Paginator is far way to simple for my needs so I have sub class it. But I can't derivate from both QuerySetPaginator and Paginator. It's antoher reason why I think it's not a good idea to have 2 classes.

[anonymous]

Woops count is not the good attribute to test because lists also have it. "all" seems a safe choice to me.

[Marc Fargas]

type(xxx) == QuerySet ... ?

[anonymous]

if isinstance(self.object_list, QuerySet):
    self._count = self.object_list.count()

[anonymous]

I am using this one and it works nicely:

    def _get_count(self):
        if self._count is None:
            if isinstance(self.object_list, QuerySet):
                self._count = self.object_list.count()
            else:
                self._count = len(self.object_list)
        return self._count
    count = property(_get_count)

[mrts]

This looks reasonable and in scope for 1.0.

[Adrian Holovaty]

(In [7865]) Fixed #7478 -- Rolled QuerySetPaginator into the Paginator class, to simplify things. QuerySetPaginator still exists as an alias, for backwards compatibility. Thanks for the suggestion, batiste@…

[Jacob]

Milestone 1.0 deleted