GH-16317: allow __debugInfo() overrides in mysqli classes#16543
GH-16317: allow __debugInfo() overrides in mysqli classes#16543DanielEScherzer wants to merge 1 commit intophp:masterfrom
__debugInfo() overrides in mysqli classes#16543Conversation
For classes that extend `mysqli`, `mysqli_result`, or `mysqli_statement`, and define a `__debugInfo()` magic method, use it.
| class subclass_mysqli extends mysqli { | ||
| public function __construct() {} | ||
| public function __debugInfo(): array { | ||
| return ['a' => 'b']; |
There was a problem hiding this comment.
Perhaps a test for seeing that you can call parent::__debugInfo() might be useful to add here as well for the different subclasses
There was a problem hiding this comment.
Except the parent classes don't define those methods? There isn't a parent method to call
There was a problem hiding this comment.
Hmm interesting if the internals of mysqli fills out an implementation for __debugInfo, then I would almost argue that is a secondary bug and it should properly expose such /or any other internal class that does similar).
Just tested and confirmed the patch with without my above comment too.
There was a problem hiding this comment.
If you want to file that bug I might be able to work on it
There was a problem hiding this comment.
Internal classes faking methods that are supposed to exists in userland is indeed a long standing issue. E.g. Internal classes don't have a default constructor... i.e. https://3v4l.org/SIMVn#v8.4.2
So I think the best and most consistent solution would be to add the debugInfo() method instead of patching the handler. As far as I see, the get_debug_info object handler of mysqli classes are just performance optimizations (the __debugInfo()method doesn't have to be called). Since performance is not a concern in case of a debug messages, it's not a problem at all if we use the magic method instead of the handler.
In any case, as this idea may be controversial, other maintainers should also express their preference how to best solve the problem before continuing the implementation...
BTW: I suspected that we have similar issues with other object handlers, but I was not able to find any other issues (initially I thought that at least I can find issues with the *_dimension handlers)
3 participants
For classes that extend
mysqli,mysqli_result, ormysqli_statement, and define a__debugInfo()magic method, use it.