Re: [RFC] Transform void into an alias for null

On 4.6.2025 22:39:28, Daniel Scherzer wrote:

On Wed, Jun 4, 2025 at 1:35 PM Bob Weinand <bobwei9@hotmail.com> wrote:

    On 4.6.2025 16:54:05, Bob Weinand wrote:

On 2.6.2025 18:27:51, Gina P. Banyard wrote:

Hello internals, This is the second RFC out of a set of type system related RFCs I want to propose for PHP 8.5. The objective is to fix a weird quirk of PHP's type system, where void lives in its own type hierarchy. This is visible mainly in that a lack of return type is not isomorphic to a function that has a return type of mixed. Let me know what you think about it. RFC: https://wiki.php.net/rfc/void-as-null Best regards, Gina P. Banyard

I have to agree with other posters here that the distinction

    between

null and void is an useful one. In particular I'd consider the null returned by void to be

    incidental

rather than intentional. I consider the return value of void

    functions

"some arbitrary value". It just happens to be null. Like every function has to return something. But returning null

    is not

an intrinsic property of a void function. It's an extrinsic one.

    You

observe void functions to generally return null. But that null in itself is meaningless. So, my counter-proposal would be allowing covariance with void and allowing everything, including non-nullable types as child type of void functions. I.e. effectively giving void and never the same semantics,

    except that

never also indicates that it never returns. Additionally I'd be in favour of disallowing (e.g. E_WARNING) consuming the return value of _direct_ calls to void functions

    (with

the exception of standalone direct calls in short closures, because consuming that value is intrinsic rather than necessarily intentional). (Disallowing indirect calls would be detrimental for usage as callback.) Bob

    Clarification: *opposite* semantics to never (which is the bottom
    type).
    Void would be effectively the top type (only inferior to untyped).

    So, it allows child classes to then return a meaningful value when
    the
    interface was just "void" (= no significant return type). As an
    example,
    when the interface says "set($val): void", the child class can
    specify
    "set($val): mixed" and return the old stored value.

    Basically, an interface can now say without further clarification "I
    have no real return value" = "void", rather than having to say
    "mixed"
    and then explaining "this is not really mixed, but whatever you want".

    (I have seen interface method return values being "upgraded" from
    void
    to mixed (or just untyped) in the past, just so that a specific child
    class can now return a meaningful value.)

    Bob

MediaWiki's hook system (https://www.mediawiki.org/wiki/Manual:Hooks) has two different kinds of hooks - those that can be aborted, for one hook handler to say that no other hook handlers should run - those that cannot be aborted MediaWiki uses void return types to help enforce this system, where hooks that cannot be aborted must have void returns. See https://www.mediawiki.org/wiki/Manual:Hooks#Hook_handler_return_values. Making it so that any interface function with a void return can be implemented by a function returning anything would seem to be a huge B/C break. If you want to use the top type, why not just use mixed? -Daniel

Hey Daniel, where's the BC break? Nothing which worked today will stop working (except you won't get exceptions in some cases). That's not a BC break. The only thing which stops working is if it's intentionally used as a guard. However, in the case of MediaWiki they do actually _care_ about the return type (and the caller of these hooks will actually check for null/true/false). So it should be annotated ": null". And not ": void". Explicit intentions are important. They probably still use ": void" as to be compatible with PHP 8.1 and older. ": null" is only supported starting PHP 8.2. I'd assume as they upgrade their required PHP version (8.1 currently) they'll shift to ": null". So, yeah, the guard will lose its guarding functionality (but we don't consider that a BC break). Regarding why not mixed? Because the intention with mixed is that the value is something meaningful. With void it's meaningless. There's a semantic distinction (and it forbids returning). And, as proposed, you could forbid direct calls of void functions giving runtime / static analysis hints. With void being covariant with respect to child functions now. Bob