Fix ClassConstFetch Identical specification with Yoda conditions

[herndlm]

Closes phpstan/phpstan#9542

Is this how you would have done it? Might be the simplest way I guess. It's just an adapted version to handle the left/right mixed variant of the block directly before it.

UPDATE: I had the urge to somehow simplify/cleanup things, like the way left and right types are fetched from the scope, assigned to variables and used, in Identical in a follow-up but I failed. Better not touch it too much 😅

[ondrejmirtes]

Hi, I'd be happier with NodeScopeResolverTest. We can leave the rule test but it'd be best to check the type inference (truthy/falsy context too). Thanks!

[herndlm]

sure. added it.
UPDATE: missed the falsy context assertions. addem them too now..

[ondrejmirtes]

There's a bad bug here. So you've eliminated that $translatable isn't TranslatableMessage. But it can still be a subclass of TranslatableMessage.

Most likely we should only specify the type for $context->true().

Same thing is done for get_class:

phpstan-src/src/Analyser/TypeSpecifier.php

Lines 194 to 210 in 2df1a59

	if (
	$context->true()
	&& $unwrappedLeftExpr instanceof FuncCall
	&& $unwrappedLeftExpr->name instanceof Name
	&& strtolower($unwrappedLeftExpr->name->toString()) === 'get_class'
	&& isset($unwrappedLeftExpr->getArgs()[0])
	) {
	if ($rightType->getClassStringObjectType()->isObject()->yes()) {
	return $this->create(
	$unwrappedLeftExpr->getArgs()[0]->value,
	$rightType->getClassStringObjectType(),
	$context,
	false,
	$scope,
	$rootExpr,
	)->unionWith($this->create($leftExpr, $rightType, $context, false, $scope, $rootExpr));
	}

[staabm]

could we be "smarter" if TranslatableMessage is a reference to a final class (or a interface)?

[ondrejmirtes]

We could, but first we need to get rid of the bug, then we can improve it :)

[herndlm]

adapted. better? I didn't even notice this.. :)

[herndlm]

oh sorry, I was lazy and didn't run the full tests :/

[herndlm]

thx for the fix!