fix bug #74478 null coalescing operator failing with SplFixedArray

[jhdxr]

fix bug #74478 null coalescing operator failing with SplFixedArray

[nikic]

This should only be done if the method has been overridden.

[nikic]

Though you might want to simply delegate to spl_fixedarray_object_has_dimension here.

[jhdxr]

fixed. thank you.

[jhdxr]

I don't choose to call spl_fixedarray_object_has_dimension because I think the call to spl_fixedarray_object_has_dimension_helper inside that function is not needed here.

[nikic]

Might it not be necessary in the (admittedly rather odd) case where offsetGet has been overridden but offsetHas not? In that case we should still perform the isset based on default semantics (or not?)

[jhdxr]

Good point. IMHO we should follow current logic, simply call offsetGet, if offsetGet is overridden while offsetHas not.
The reason why bug #74478 is a bug is $data['foo'] ?? 42; equals to isset($data['foo']) ? $data['foo'] : 42;. We override offsetHas because we hope we can define how isset() works, and this bug ruled out the ability. If we don't override offsetHas, it indicates I'm pretty satisfied with how isset() works now, so we just need to follow current logic.

[nikic]

So what I had in mind here is something like this:

<?php
class MyArray extends SplFixedArray {
    public function offsetGet($key) {
        return "prefix_" . parent::offsetGet($key);
    }
}

$arr = new MyArray(1);
var_dump(isset($arr[0]) ? $arr[0] : 42); // Returns 42
var_dump($arr[0] ?? 42); // According to your patch (I think) this will throw rather than returning 42

Due to a typo only extending offsetGet didn't actually work, just fixed this in 1967950.

[jhdxr]

yes. In fact I have fixed this in my patch as well.

[keradus]

@jhdxr thanks for picking up my bug report !

[jhdxr]

@keradus my pleasure.

ping @krakjoe .

[nikic]

Sorry for the delay, now merged as 872e43d into 7.0+. Thanks!