Fix non-existent offset when guarding

[SplotyCode]

This pr is trying to fix a false positive.

/**
 * @param list<int> $array
 * @param positive-int $index
 */
public function guardNotSafeLowerBound(array $array, int $index) {
    if ($index < count($array)) { // index is within 0..count(array)-1
        return $array[$index]; // phpstan reports a potentially unsafe array access
    }
    return null;
}

Note that 1) $array is a list 2) $index is positive (or at least >= 0)

[staabm]

is this PR trying to fix phpstan/phpstan#13773 ?
(since the PR does not at all contain a description.. I can only guess)

if so, you should either adjust the logic added with #4403 or maybe even delete it in case you want to re-implement it in a different place

[SplotyCode]

Thanks for pointing this out! I updated the description of this pr.

I have not opened a issue for this pr. But i will also try and fix phpstan/phpstan#13773 later in a diffrent pr. Here i am trying to fix array access on lists if we explicitly guard the index.

[SplotyCode]

There seems to be a regression

$var = null;
if ($index < count($array) {
    $var = 2;
}
if ($index < count($array) {
    /* phpstan does not recognize that $var is always 2 now */
}

This did work before this pr. I am wondering if this is also true for the other checks in the method i changed.

[staabm]

this type does not make sense, because non-empty-list<stdClass> already implies that the list has a offset 0 with value stdClass

[staabm]

instanceof Expr\Variable is usually a bug as most logic works for other expressions (e.g. a pure function call can also be remembered).

most of the time this works just by dropping this line

[staabm]

usually we pass thru the $context, instead of creating a new one.

[ondrejmirtes]

This PR looks abandonded. Feel free to reopen on latest 2.1.x if you still want to fix it.