#2415 bytes() should respect __bytes__

This patch allows classes to define __bytes__ to customize what bytes(obj) returns. I added two new C functions PyObject_Bytes, and PyBytes_FromObject.

[barry, 2008-08-26 09:05:15]

http://codereview.appspot.com/3245/diff/1/4
File Objects/object.c (right):

http://codereview.appspot.com/3245/diff/1/4#newcode469
Line 469: return PyBytes_FromString("<NULL>");
Should this raise an exception instead?  What's the use case for allowing v ==
NULL?

http://codereview.appspot.com/3245/diff/1/4#newcode491
Line 491: PyErr_Clear();
Should this swallow all lookup errors or just more specific ones indicating
__bytes__ doesn't exist?

[Benjamin, 2008-08-26 13:30:04]

On 2008/08/26 09:05:15, warsaw wrote:
> http://codereview.appspot.com/3245/diff/1/4
> File Objects/object.c (right):
> 
> http://codereview.appspot.com/3245/diff/1/4#newcode469
> Line 469: return PyBytes_FromString("<NULL>");
> Should this raise an exception instead?  What's the use case for allowing v ==
> NULL?

I did this because this is what PyObject_Str does if the value is NULL.
> 
> http://codereview.appspot.com/3245/diff/1/4#newcode491
> Line 491: PyErr_Clear();
> Should this swallow all lookup errors or just more specific ones indicating
> __bytes__ doesn't exist?

This is also the behavior of 2.6's PyObject_Unicode if the lookup returns NULL.
I personally think that it could check for an attribute error, but I guess that
might be expensive.