[Python-Dev] Bad interaction of __index__ and sequence repeat

[Nick Coghlan]

Nick Coghlan wrote:
> Tim Peters wrote:
>> OTOH, in the long discussion about PEP 357, I'm not sure anyone except
>> Travis was clear on whether nb_index was meant to apply only to
>> sequence /slicing/ or was meant to apply "everywhere an object gets
>> used in an index-like context".  Clipping makes sense only for the
>> former, but it looks like the implementation treats it more like the
>> latter.  This was probably exacerbated by:
>>
>>     http://mail.python.org/pipermail/python-dev/2006-February/060663.html
>>
>>     [Travis]
>>     There are other places in Python that check specifically for int objects
>>     and long integer objects and fail with anything else.  Perhaps all of
>>     these should aslo call the __index__ slot.
>>
>>     [Guido]
>>     Right, absolutely.
>>
>> This is a mess :-)
> 
> I've been trawling through the code a bit, and I don't think it's as bad as 
> all that.

Damn, it really is a mess. . . nb_index returns the Pyssize_t directly, and a 
whole heap of the code expects errors to be signalled via returning -1 before 
checking PyErr_Occurred().

To get it to work without clipping everywhere, wrap_lenfunc (typeobject.c), 
_PyEval_SliceIndex (ceval.c), PyNumber_Index (abstract.c) and sequence_repeat 
(abstract.c) all had to be modified to recognize PY_SSIZE_T_MIN and 
PY_SSIZE_T_MAX as potential error flags (in order to clear the overflow error 
for _PyEval_SliceIndex, and in order to propagate the exception for the other 
three).

And using this approach still means that (2**100).__index__() raises an 
OverflowError.

It would probably be cleaner to change the signature of nb_index to return a 
PyObject *, and let the code that uses it worry about how (or even whether!) 
to convert PyLong results to a Py_ssize_t.

Cheers,
Nick.


-- 
Nick Coghlan   |   ncoghlan at gmail.com   |   Brisbane, Australia
---------------------------------------------------------------
             http://www.boredomandlaziness.org