Consider making implicit any the default for all generic type parameters

[RyanCavanaugh]

Now that we have generic defaults (:tada:), we should consider making all type parameters have an implicit any default. An unspecified uninferred type parameter would have type any and trigger a warning under noImplicitAny.

Benefits:

• We'll need to do this in Salsa anyway
• Improves JS migration story for codebases which inherit from generic base classes

Drawbacks:

• There may exist people who run with noImplicitAny off but still want to enforce type parameters to be provided?

[falsandtru]

trigger a warning under noImplicitAny.

It's a really great improvement.
I definitely want it.

[RyanCavanaugh]

@rbuckton interested in folding this into your current PR? We can also do it later instead

[mhegazy]

I would keep this in a separate PR.

[gcnew]

Wouldn't this be at odds with #1213 and #9949?

[DanielRosenwasser]

@gcnew could you elaborate on that thought?

[gcnew]

My hunch is that defaulting parameters to an implicit value is not a good idea in general. I don't have anything specific, it was more like throwing in areas where potential issues may reside.

Giving it a second thought, defaulting parameters to any will make the checker more permissive for users with noImplicitAny off. If #1213 and #9949 get implemented, the checker will become more restrictive again, which may be viewed as a backwards compatibility problem. On the other hand this proposal is beneficial for noImplicitAny users.

[DanielRosenwasser]

@gcnew I don't think that the implicit {} constraint type is much better though. If anything, {} is confusing for newcomers, and the behavior of getting it rather than any is actually more confusing for those who are familiar with noImplicitAny. I do think that the positives you listed are pretty compelling though, and explain much of the rationale for this proposal.

@RyanCavanaugh I just want to clarify the intent - is it all type parameters including those which have explicit defaults?

[RyanCavanaugh]

If there's an explicit default, we would obviously use that instead of any. Maybe I don't understand the question.

[DanielRosenwasser]

Sorry, I meant to ask about explicit constraints, not explicit defaults.

[RyanCavanaugh]

Somehow decided we should do #10571 as part of this discussion