[PATCH 2 of 3 RFC] Remove all uses of hasattr(a, b) in favor of getattr(a, b, None)

[Augie Fackler]

On Mar 14, 2011, at 12:48 PM, Matt Mackall wrote:

> On Sun, 2011-03-13 at 12:11 +0100, Benoit Boissinot wrote:
>> On Tue, Mar 8, 2011 at 9:36 AM, Martin Geisler <mg at aragost.com> wrote:
>>> Augie Fackler <durin42 at gmail.com> writes:
>>> 
>>>> On Mar 3, 2011, at 8:03 AM, Michael Haggerty wrote:
>>>> 
>>>>> For the other cases, maybe it would make sense to define a
>>>>> 
>>>>> def safe_hasattr(o, name):
>>>>>   try:
>>>>>       getattr(o, name)
>>>>>   except AttributeError:
>>>>>       return False
>>>>>   else:
>>>>>       return True
>>>> 
>>>> Does anyone in the wider mercurial-devel have a feeling on this one?
>>>> It seems reasonable as a hedge against the day in about 2050 that we
>>>> can be Python3 only...
>> 
>> It seems it could be useful? Can you respin the patch with
>> safe_hasattr() and the "optimized defaults"?

Will do, hopefully later this week.

>>> Would it not make sense to overwrite __builtin__.hasattr with such a
>>> safe version or is that considered very bad style?
>>> 
>> 
>> Dunno, I don't think we monkey-patch python internals already, do we?
> 
> We do, but sparingly (see demandimport). In WSGI environments and
> elsewhere, messing with globals can be problematic.

This feels evil to me.

I'll try and re-roll this work later this week, once I'm caught up from pycon.