Josiah Carlson wrote:
Is mmap's inability to be subclassed a feature or bug?
It's a missing feature: it's not a bug, because nobody says this should work, and anybody trying will find out that it doesn't work, so nobody is tricked into believing it should work. The mmap type is not even documented; mmap.mmap is a function.
It's not a feature, because (atleast IMO) there would be nothing wrong with making mmap.mmap a subclassable type.
Is one's inability to use a = mmapinstance.__setslice__;a(1,2,'a') (and others) a feature or bug?
That is a bug. Slice assignments are supported, and so should __setslice__.