Rector is wrongly relying on ObjectType::equals implementation

[VincentLanglet]

Hi @TomasVotruba @samsonasik,

I'm currently working on a Fix in phpstan-src on ObjectType::equals (phpstan/phpstan-src#5399)

• ObjectType->equals(GenericObjectType) was true (which is not correct)
• ObjectType->equals(TemplateObjectType) was true (which is worst since TemplateObjectType->equals(ObjectType) was false.

But, with our integration tests it shows that this fix will break

• DowngradeCovariantReturnTypeRectorTest (https://github.com/phpstan/phpstan-src/actions/runs/23980124279/job/69943196872?pr=5399)
• RenameClassRectorTest, FixtureAutoImportNames, AutoImportNamesTest (https://github.com/phpstan/phpstan-src/actions/runs/23980124279/job/69943196871?pr=5399)

I got confirmation from ondrej that my fix was correct and your implementation had some issues (phpstan/phpstan-src#5399 (comment))
After some investigations, it seems like this is because you rely a lot on Type::equals, for instance here:
https://github.com/rectorphp/rector-downgrade-php/blob/15578226e7d293f224a121dabaa11d8f6e7ab42a/rules/DowngradePhp74/Rector/ClassMethod/DowngradeCovariantReturnTypeRector.php#L239

This had the benefit BEFORE to return true when you compares a FullyQualifiedObjectType (

rector/src/StaticTypeMapper/ValueObject/Type/FullyQualifiedObjectType.php

Line 15 in 8e733db

final class FullyQualifiedObjectType extends ObjectType

) which the ObjectType of the same class. But as soon as the ObjectType::equals will be fixed it will returns false.

I'm not fully sure how to change your implementation, some ideas...

• Using describe method
• Using toPhpDocNode
• Using $child->isSuperTypeOf($parent)
• Overriding equals method in your custom ObjectType class

[samsonasik]

phpstan seems not using latest rector

https://github.com/phpstan/phpstan/blob/16babc0bfc4e9cde0136f20cf3ca30f4ecd50131/.github/workflows/integration-tests.yml#L217-L218

btw, could you create failing fixture test? Thank you.

[samsonasik]

@VincentLanglet is there a way to pull phpstan-src PR branch in rector-src to show the test error? If no, probably we need to wait the PR in phpstan-src to be merged so we can test the phpstan 2.1.x-dev :)

anyway, any such change imo should go to phpstan 2.2

[staabm]

@samsonasik you can download the PHPStan phar-file from the PRs github-action compile PHAR job and put it into your local rector repo

[VincentLanglet]

@VincentLanglet is there a way to pull phpstan-src PR branch in rector-src to show the test error? If no, probably we need to wait the PR in phpstan-src to be merged so we can test the phpstan 2.1.x-dev :)

I merged the PR on 2.1.x, so the PHAR is compiled and uploaded on https://github.com/phpstan/phpstan now.
It should be easier for you. (even if it's later decided to revert this commit to do it on only on 2.2)

anyway, any such change imo should go to phpstan 2.2

Technically that's not a new feature on phpstan side but a real bugfix which solves multiple issues on our side.
All equals method are meant to work in a way that A->equals B should gives the same result than B->equals A.
You can still try to challenge this with ondrej.

[samsonasik]

I am looking at it ...

• Make compatible with PHPStan 2.1.x-dev for ObjectType::equals() change behaviour rector-src#7956