GetClassToInstanceOfRector invalid fix to instanceof for strict type check

[jeremyVignelles]

Bug Report

Subject	Details
Rector version	v1.0.4

I'm using getclass($object) === A::class to test if my variable is exactly of type A (and not a subclass).
Rector tries to edit this into an instanceof check, due to the GetClassToInstanceOfRector rule

Minimal PHP Code Causing Issue

<?php

class A {}
class B extends A {}

$c = new B();

if (get_class($c) === A::class) {
    echo 'KO - it is exactly of type A';
} else {
    echo 'OK - it is not exactly of type A';
}

Code gets "fixed" to $c instanceof \A

Expected Behaviour

I should be able to express that I am willing to keep that get_class call to check that exact type.

[jeremyVignelles]

Seems like $c::class === A::class is OK for rector, is that the recommended thing to do ?

[samsonasik]

@TomasVotruba @jeremyVignelles I am thinking that this rule should only cover !== instead of ===, as it seems valid? it requires some verification tho...

<?php

class A {}
class B extends A {}

$c = new B();

-if (get_class($c) !== A::class) {
+if (! $c instanceof A) {
   
}

wdyt?

[samsonasik]

Oh, the result seems different https://3v4l.org/7cbU6

[samsonasik]

If the type is exactly of what it defined (sub type), it can be checked by something like:

        $varType = $this->nodeTypeResolver->getNativeType($varNode);
        $objectType = new ObjectType($className);

        if ($varType instanceof ObjectType
            && ! $this->typeComparator->areTypesEqual($varType, $objectType)
            && $this->typeComparator->isSubtype($varType, $objectType)) {
            return null;
        }

some verification on dynamic variable may be needed. I guess for your use case, you can skip the rule:

https://getrector.com/documentation/ignoring-rules-or-paths

as it may be usefull to improve type check, but on some very specific use case, it may cause a bug.