1.4.6

[ondrejmirtes]

Improvements 🔧

• Implemented constant type inference for implode() function (#991), thanks @staabm!

Bugfixes 🐛

• Fix interaction between template type and union type (#992), Union Type + Nullable yields "Expects TEntity(A|B)|null, TEntity(A|B)|null given" #5591, thanks @arnaud-lb!
• Delegate most of TypeCombinator::remove() to Type::tryRemove() (#994), Incorrect expected return type with union type and generic #6438, thanks @arnaud-lb!
• TemplateType should accept a TemplateType derived from itself (#999), Recognize covariance when returning generic union types #5844, How to do simple generics #5706, thanks @arnaud-lb!
• Fixed intersecting and removal from TemplateUnionType (phpstan/phpstan-src@0d28835), False positive: Access to an undefined property T of A|B|C::$name (regression) #6566
• Fixed type seen by IterableInForeachRule with inline @var right above foreach (phpstan/phpstan-src@b6bbbaf), Can't override mixed to iterable #6564
• modulo 1 is always zero (#973), thanks @staabm!
• multiply by zero is always zero (#974), thanks @staabm!
• Fixed discovering backed enum without space around : (phpstan/phpstan-src@5162bcf), Classmap autoload for backed enum not working in all cases composer/composer#10498

Function signature fixes 🤖

• Function md5() always return non-empty-string (#987), thanks @janbarasek!
• Correct the case for parameters that accept an instance of Iterator. (#988), thanks @johnbillion!
• Fix the return types of round()/ceil()/floor() (#729), thanks @johnbillion!
• Correct the return type of SplFileObject::fstat() (#996), thanks @johnbillion!
• Improve the return type of functions that return an indexed array of strings (#997), thanks @johnbillion!

---

This discussion was created from the release 1.4.6.

[mvorisek]

starting with this release https://github.com/atk4/data/blob/191b20fd1e6e7025fd91991667cdbaba642981d5/src/Model/EntityFieldPair.php#L60 ignore is not longer needed, but I wonder why?

The required return type is TField (template type extending Field), but Field is returned (which can be different from TField).

Is this an expected change?

[ondrejmirtes]

@mvorisek Don't worry, all of that is by design.

[mvorisek]

a) returning Field for expected Field2 does emit an error since level 3
b) returning Field for expected template of Field2 does NOT emit an error until level 7
c) returning Field for expected template of Field does NOT emit an error until level 7

I agree the c) example needs a compromise, but the b) one is clearly not consistent vs. non-template a).

[ondrejmirtes]

b) does not emit an error until level 7 because Field is a parent of Field2.

[mvorisek]

Yes, but why is common parent checked instead of stripping the "template of " and checking it as a standard type?

Then behaviour will be unified and easy remember as such check is done in l3.

[ondrejmirtes]

You're welcome to investigate and try to come up with a fix.