Le lun. 27 janv. 2020 à 10:19, Nikita Popov <nikita....@gmail.com> a écrit :
> On Wed, Jan 22, 2020 at 4:47 PM Nicolas Grekas < > nicolas.grekas+...@gmail.com> wrote: > >> Hello everyone, >> >> as announced last week, I'm officially opening a discussion for adding a >> "Stringable" interface to PHP 8. >> >> The RFC and its rationale are presented here as required: >> https://wiki.php.net/rfc/stringable >> >> The patch is found on GitHub, where some discussions happened already, you >> might be interested in having a look: >> https://github.com/php/php-src/pull/5083 >> >> TL;DR, I think we need an interface to allow passing objects with a >> __toString() method in a type-safe way, using the string|Stringable type. >> Here is the stub of the proposal: >> interface Stringable { public function __toString(): string; } >> >> I invite everyone to review the RFC before commenting here as I tried to >> sum up considerations that some already had before and that you might have >> too. >> >> Cheers, >> Nicolas >> > > I'm wondering whether it could make sense to automatically implement this > interface for all classes that define __toString(). Explicitly implementing > the interface is useful for the transition period (where old PHP versions > also need to be supported), but when considering only PHP 8 support, I > think it may make sense to implicitly implement the interface. If the > __toString() functionality also works fine without implementing the > interface, and there's probably only going to be relatively few consumers > of the Stringable type, it will likely be quite common that the interface > will not get implemented... > > Automatically implementing interfaces based on implemented methods could > also be a general language feature (see Go interfaces for example), but > likely not one we're interested in having in it's full generality. In this > particular case, it seems useful though. > This looks like a nice improvement to the proposal, it'd be OK on my side. For the implementation, I might need your help to implement this logic, especially for extensions, since I suppose we'd like them to also have this behavior of implicitly declaring the return-type + interface when they declare the method. Could you please help me achieve that? Nicolas
