mark-todd

> I think this should be required reading for anyone working on this, as it is another language with gradual typing adding a full type system. https://elixir-lang.org/blog/2023/09/20/strong-arrows-gradual-typing/, running into the...

@mikeshardmind Thanks for breaking this down, appreciate it. So I'll park the discussion about Any for now - although I don't understand why expanding to Any where more information is...

> I determined that getting the special casing right on this would be harder and more fragile than it was worth. (specifically, it would be harder than getting the full...

I also think it's worth noting that waiting for a lot of pre-requisites to be resolved internally is quite unrealistic. Even if right now all of the pre-requisites appear on...

Ok, let's suppose we go with option 1 with the subtyping rules. If something is a subtype of `T & Any` it must be a subtype of `T` and a...

+1 on this issue - onwards of 7.54.1 introduces a breaking change on reset for us that breaks our tests, as we test to check the request data matches the...

> Is that actually useful though? A TypedDict is always exactly a dict, never a subclass. Yeah I'm inclined to agree that maybe it's not that helpful a subcase, and...

> While [PEP 705](https://peps.python.org/pep-0705/) could solve that maybe, I still would like to be able to "overwrite" the `__setitem__` Method with something with Never in order to tell the type...

> Another use case is pandas or numpy.ndarray which have a __getitem__ methods that relay to a array with defined types. I want to be able to write a TypeGuard...

> There are https://github.com/python/typing/issues/1458 and the actual allowed behavior of Never currently still varies between type checkers. I don't know that we need to solve this right this moment now...