Explicit generic function arguments about tl HOT 3 OPEN

There might be some cases where this workaround isn't sufficient but I can't find it at the moment.

I honestly don't see it as a workaround, since it amounts to the same number of annotations. I understand that one gets the feeling that ideally no variables should need to be annotated, but if anything, I find that not needing a new (pseudo-?)operator is a cleaner approach, language-wise.

Explicit variable annotations and bidirectional inference do a lot of heavy lifting to resolve the types of things like literal tables (e.g. local t: MyComplexRecord = { ... { ... { ... deeply_nested_thing = ... } ... } ... }), and this was the primary reason why we added bidirectional inference. Being able to use the same machinery to avoid the need for the "turbofish annotation" felt like a good way to go. The present rule is that variables for function calls are resolved either by the expected types of its return values in the given call-site context, or the types of its given arguments.

from tl.

Teal does bidirectional type inference to try to avoid the need for this.

In your example above, doing local foo: integer = get_foo() should make the expected type flow into the function call.

Do you have a more complex example where that doesn't work and the explicit generic function argument would still be necessary? I've been trying to see how far we can go without adding more syntax for function calls.

from tl.

So I went back to the case that prompted this request, and found that I was able to avoid the need for an explicit generic by adding more locals with explicit types. There might be some cases where this workaround isn't sufficient but I can't find it at the moment.

from tl.