Consider the following code:

>>> class badstr(str):
...     def __len__(self):
...         return 6
...     def __getitem__(self, index):
...         return "a"
...
>>> filter(None, badstr("bbb"))
'aaa'

I would have expected the answer to be 'aaaaaa'.

The cause for this is that
Python/bltinmodule.c:filter{string,unicode,tuple} all
call PyString_Size (or the appropriate equivalent),
even if the sequence is a subclass of one of these
types. This bypasses any overloading of __len__ done by
the subclass, as demonstrated above.

If filter() is going to respect the subclass's
__getitem__ overload, it should also respect
overloading of __len__. The attached patch corrects this.

Logged In: YES 
user_id=80475

This isn't a bug.  You're making assumptions about
undocumented implementation details.  The is no shortage of
cases where the C code makes optimized direct internal calls
that bypass method overrides on subclasses of builtin types.