Get rid of of Type::getArrays() usage in InvalidKeyInArrayDimFetchRule

[herndlm]

this thing just keeps annoying me and I want to do something about it. of course I dislike the instanceof MixedType, but maybe someone has a better idea?

[phpstan-bot]

This pull request has been marked as ready for review.

[ondrejmirtes]

The rest of this method should be:

if ($isSuperType-yes()) {
    return [];
}

return [
	RuleErrorBuilder::message(
		sprintf('%s array key type %s.', $isSuperType->no() ? 'Invalid' : 'Possibly invalid', $dimensionType->describe(VerbosityLevel::typeOnly())),
	)->build(),
];

[herndlm]

that looks better! adapted. but I had to keep the maybe-early-exit in to avoid the rule reporting maybes on level 3 already (the newly added levelstest case started failing)

[herndlm]

oh and I forgot to run make phpstan before pushing. need to check what's going on with mixed there which should be ignored still I guess

[ondrejmirtes]

It's a bit weird because RuleLevelHelper should have already taken care of the maybe case on lower levels.

[herndlm]

also had a feeling that the closure is maybe still wrong, but RuleLevelHelper doesn't seem to be dealing with checkMaybe? it deals with e.g. checkUnionTypes, but in my case the union is irrelevant, the int dim on a constantarray is a maybe :/

[ondrejmirtes]

Oh, you're right, the rule code needs to be a bit more complicated after all. Not sure what's with the mixed in make phpstan, PHPStan is running on level 8...

[herndlm]

the rule was explicitly ignoring mixed dims before as well when checking for maybe/potential invalid keys. maybe it can't be simplified after all. I was also looking forward to put all checks on top and into the closure

[herndlm]

the mixed dims that were reported seemed mostly legit, maybe this could even be turned on with bleedingEdge. but it's not related to this PR I think, I let the rule ignore them again. at least it's a bit simpler and makes use of RuleLevelHelper now hmm

[ondrejmirtes]

Thank you!