assert that some type is a list

[orklah]

will fix an item of #6395

[weirdan]

I wonder whether a stub would not suffice here. E.g.

/**
 * @psalm-assert-if-true list<mixed> $array
 */
function array_is_list(array $array): bool {}

[orklah]

Possibly, but string, bool, array, object are all handled here. I think having it here is less confusing.

We could do another PR if we want to migrate everything to stubs though... Or if you want I can migrate all of them in this PR

[weirdan]

Consistency is good, so it should be ok the way it is. Just curious, how hard would it be to use FQNs instead of local names? Because right now Psalm doesn't understand namespace-local overrides of global predicates: https://psalm.dev/r/8d7345506e c.f. https://3v4l.org/4Pko7

[psalm-github-bot]

I found these snippets:

https://psalm.dev/r/8d7345506e

<?php

namespace NS;

function is_string(mixed $_var): bool { return true; }

function f(mixed $v): string {
    if (is_string($v)) {
        return $v;
    }
    return '';
}

var_dump(f(new \stdClass));

Psalm output (using commit 4eadb73):

ERROR: ForbiddenCode - 14:1 - Unsafe var_dump

[orklah]

Oh, this is kinda bad (I'd not feel bad about the fool who did override those with another meaning though!). I wonder if this is fixable in assertions, I'll check

[orklah]

There's probably a better way to check whether we're using the base function, but I couldn't find it!

[orklah]

I'll have to check why CI is failing, I'd not expected that

[orklah]

the CI was failing because I didn't handle the namespace fallback. It's now OK