(🎁) should N804 enforce mcs on metaclass classmethods?

[KotlinIsland]

class M(type):
   def f(self):  # N804
       pass
   def m(cls):  # no error
       pass
   @classmethod
   def m(cls):  # no error
       pass
    @classmethod
    def c(mcs):  # N804
       pass

i would expect this to be N804, no error, N804, no error

[dhruvmanila]

There does seem to be some precedent for the "mcs" shorthand: https://grep.app/search?q=def%20%5B%5E(%5D+%5C(mcs%5B,%5C)%5D&regexp=true, @AlexWaygood any thoughts?

[eliegoudout]

Even though mcs itself maybe not be widely accepted, the fact that that it should be a one of mcs, mcls, metacls, <any other possible shorthand> to explicit difference in nature from cls (instance space in the case of metaclasses) seems more than reasonable (see #10656 (comment)).

I would love to see #13698 (comment) (planned to be) fixed!

Personnal opinion: I really love mcs for 3 reasons:

1. Three characters, like cls.
2. Hard to accidentally go from mcs to cls (or vice versa) than from mcls to cls.
3. Precedent point + no "tall letter" (e.g. the l from cls) also makes it easier to "read the difference" from a quick glance imo.

[eliegoudout]

Note that official library abc uses mcls for example:

https://github.com/python/cpython/blob/2a66dd33dfc0b845042da9bb54aaa4e890733f54/Lib/abc.py#L105-L108

[MichaReiser]

Related discussion about changing the rule not to enforce any naming convention for metaclass methods #6017