Support narrowing a constant array to a list with count

[herndlm]

Implementing what was found/suggested in #3709 (comment) and improves very special cases with int ranges.

PR title is maybe misleading. it's making keys non-optional if it can for list-like cases.

[ondrejmirtes]

I believe this has nothing to do with lists. If you have array{a?: int, b?: string} and check count >= 2, you can make them required too. Please add more tests.

If I have array{a: int, b: string, c?: int} and check count > 2 or count >= 3, c should become required too.

[staabm]

If I have array{a: int, b: string, c?: int} and check count > 2 or count >= 3, c should become required too.

this sounds like sealed vs. unsealed array shape business

[ondrejmirtes]

Yes, in the future we will have to modify the condition to only apply to sealed shapes.

[ondrejmirtes]

You're right, we shouldn't change this now before sealed shapes are implemented.

[ondrejmirtes]

Thank you.

[ondrejmirtes]

We have another regression: https://phpstan.org/r/c97add7b-a241-45c9-88f1-b8e7c76daa09

Please, Martin or Markus, look into it. Thank you!

[herndlm]

weird, this stuff is supposed to deal only with lists or constant arrays, I'll try to take a look

[herndlm]

yeah, looks not related to recent changes at least. if I did not mess up massively right now, then this is *NEVER* already with phpstan 2.0.0

I'll see if I can find out why though..

[herndlm]

wait, everything is OK I think 😅 it is comparing a constant int with an array and that will always return false and make the assertion throw an exception. riiight? :) https://3v4l.org/k2LU5

I was already deep-debugging TypeSpecifier and my confusion grew by the second 😂

[ondrejmirtes]

Oh now I'm confused 😂 I need to look into what was I thinking.

[ondrejmirtes]

Alright, this is the right code sample: https://phpstan.org/r/032a2b0f-40bc-48e1-a5ee-a71859e78b04

This really changed before and after this PR. Sorry about the previous confusion.