On Mon, Nov 26, 2018 at 7:56 AM Chris Angelico <ros...@gmail.com> wrote: > > On Mon, Nov 26, 2018 at 7:55 AM Iwo Herka <iwohe...@gmail.com> wrote: > > > > > class Foo: > > > def setup(self): ... > > > __init__ = lambda self: self.setup() > > > > Sorry, didn't address this. This is fine too, since I'm assuming that > > only methods named __init__ are allowed to mutate the object. > > Because 'setup' is not '__init__' it's disqualified. > > > > Ahh, okay. In that case, yeah, you can safely ignore lambda > functions... until Python 3.8, when assignment expressions become a > thing. (See PEP 572.) So if you want to do something hacky like using > the __name__, go for it. >
Actually... it kinda depends on the point of the optimization. If you want a quick check, look at __name__. If it's okay to spend a bit more time figuring out whether to optimize or not, look for a STORE_ATTR opcode. If there aren't any, there can't be any "self.foo = bar". (You can't be certain of the converse, as "other_object.foo = bar" will also show up as STORE_ATTR.) This is a bit more work, but it's fundamentally non-hacky, because you're actually looking for the thing you care about. ChrisA -- https://mail.python.org/mailman/listinfo/python-list
