[DX] Remove *AnnotationIncorrectNullableRector rules as works with unreliable docblocks and can have 2 solutions

[TomasVotruba]

These rules work with unreliable docblocks and can have 2 solutions.

Since Rector 0.15 we declare type stricness and avoid docblock changes, as unreliable and could lead to incorrect assumptions: https://getrector.com/blog/new-in-rector-015-complete-safe-and-known-type-declarations

These are one of few forgotten rules. Instead handle manually and let PHPStan help you decide 👍

[samsonasik]

There are 3 rules that fix incorrect nullable:

ReturnAnnotationIncorrectNullableRector
ParamAnnotationIncorrectNullableRector
VarAnnotationIncorrectNullableRector

[TomasVotruba]

I see, I'll add them to this removal list 👍 Thanks

[samsonasik]

actually, I can see these rules are actually good, this example verify based on typed existing param, eg:

     /**
-     * @param \DateTime[] $dateTimes
+     * @param \DateTime[]|null $dateTimes
      */
     public function setDateTimes(?array $dateTimes): self

[samsonasik]

/cc @dorrogeray FYI

[samsonasik]

@TomasVotruba actually, I can see these rules are actually good, this example verify based on typed existing param, eg:

     /**
-     * @param \DateTime[] $dateTimes
+     * @param \DateTime[]|null $dateTimes
      */
     public function setDateTimes(?array $dateTimes): self

probably you can reconsider it?

[TomasVotruba]

I was thinking about the usefulness and they can preserve the state like example. The question is, often the nullability is not desired and code should be changed differently:

     /**
     * @param \DateTime[] $dateTimes
      */
-     public function setDateTimes(?array $dateTimes): self
+     public function setDateTimes(array $dateTimes): self

Same applies for var/return cases.

Instead of silently preserving wrong state, it should be handled by developer to fits exact needs. That's why I think it's better let PHPStan report these and people to decide what way should be taken.