[Python-Dev] About [].append == [].append

[Serhiy Storchaka]

23.06.18 05:21, Guido van Rossum пише:
> A bound method is a fairly complicated object, and for builtin bound 
> methods, the == comparison has the following definition:
> - if the `__self__` objects are not the same object, return False
> - otherwise, return True iff it's the same method (i.e. the same name / 
> the same underlying C function)
> 
>     I think it is more important that builtin methods and Python methods
>     behave the same. Should Python methods be changed to compare self with
>     "is" or are we too late to make that change?
> 
> 
> I am not sure. It's surprising, but I fear it may be too late to change. 
> Are there tests in the stdlib for this behavior?

Two tests are failed if change the current behavior. Both were added by 
fd01d7933bc3e9fd64d81961fbb7eabddcc82bc3 in issue1350060 together with 
changing comparisons for methods.

https://github.com/python/cpython/commit/fd01d7933bc3e9fd64d81961fbb7eabddcc82bc3
https://bugs.python.org/issue1350060

It changed the behavior "for consistency". But unfortunately it made it 
less consistent (and broke your definition). There are different kind of 
methods, and currently they have different behavior.

    >>> [].append == [].append
    False
    >>> [].__iadd__ == [].__iadd__
    True
    >>> UserList().append == UserList().append
    True

Seems the last two examples returned False before issue1350060.

I think changes made in issue1350060 should be reverted. See 
https://github.com/python/cpython/pull/7848 . But perhaps only in the 
master branch for minimizing possible breakage.