Supporting Generator with parameters

[mmarcinmichal]

Hi,
it will be greate if the package supports Generator with types. Currently, it is working only on typing.Generator, i.e.

class Foo2:
    @multimethod
    def bar(self, x1: str):
        print("str " + x1)

    @bar.register
    def _(self, x1: typing.Generator):
        print("Generator ")


tt = Foo2()
tt.bar("ala")
my_gen = (i for i in range(10))
tt.bar(my_gen)

When I try specidied generator type, i.e.

class Foo2:
    @multimethod
    def bar(self, x1: str):
        print("str " + x1)

    @bar.register
    def _(self, x1: typing.Generator[int, None, None]):
        print("Generator ")


tt = Foo2()
tt.bar("ala")
my_gen = (i for i in range(10))
tt.bar(my_gen)

raise DispatchError(msg, types, keys)
multimethod.DispatchError: ('bar: 0 methods found', (<class '__main__.Foo2'>, <class 'generator'>), [])

Best,

[coady]

I don't see any way of doing it though, for iterators in general. Sequence[...] checks the first value - which is already inaccurate if the sequence has mixed types - but it simply can't check Iterator[...] without altering it. Even if it cached the first value and constructed a new iterator, the function would receive a different object.

It could partially support Iterator[...] by only checking for Iterator, but I think that would be surprising too. I'm not sure that's better, unless most users think it's better.