Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:125295
Feedback-ID: id5114917:Fastmail
Date: Mon, 26 Aug 2024 23:06:49 +0100
To: internals@lists.php.net
Subject: Re: [PHP-DEV] [RFC] Default expression
User-Agent: K-9 Mail for Android
In-Reply-To: <CAJp_myXPpgmxkJautxUiADLz_fgVNv4RJSkpz_duAao9acSojA@mail.gmail.com>
References: <CAJp_myV3Riup2DyDWjgpHypDE6qP03Ogzdr6mZo6wWUmCtx16g@mail.gmail.com> <E1458589-B95C-43D0-A654-F8DB75782AB7@getmailspring.com> <CAJp_myXPpgmxkJautxUiADLz_fgVNv4RJSkpz_duAao9acSojA@mail.gmail.com>
Message-ID: <BFA9B3D1-6B5E-4AD0-98D4-26EE5A258EBE@rwec.co.uk>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable
From: imsop.php@rwec.co.uk ("Rowan Tommins [IMSoP]")



On 26 August 2024 22:27:55 BST, Matthew Weier O'Phinney <mweierophinney@gm=
ail=2Ecom> wrote:
>How exactly is this worrisome? Consider this:
>
>    class A {
>        public function __construct(private LogInterface $logger =3D new
>DefaultLogger()) { }
>    }
>
>    class ProxiedLogger implements LogInterface { =2E=2E=2E }
>
>    $a =3D new A(new ProxyLogger(default));
>
>If class A is programming to the `LogInterface` contract, the fact that i=
t
>gets a proxied version of the default should not matter in the least=2E B=
eing
>able to proxy like this means that a _consumer_ of class A does not need =
to
>know or care what the default implementation is; they can assume it follo=
ws
>the same contract, and proxy to it regardless=2E The upstream developer
>doesn't need to care, because they are programming to the interface, not
>the implementation=2E This doesn't violate the separation of concerns
>principle, nor covariance=2E


On a technicality, it doesn't violate any variance rules, because PHP does=
n't treat constructors as substitutable; although rules of API stability be=
tween versions generally do=2E

For any other method, it would violate variance principles because they gu=
arantee that the method accepts *at least* LogInterface, not *exactly* LogI=
nterface=2E A child class can change the parameter type to any superset - a=
 parent interface, a nullable type, a union, or mixed - and choose a defaul=
t to match=2E

As Bruce Weirdan astutely pointed out [https://externals=2Eio/message/1251=
83#125274] the caller is treating the default value as though it's an *outp=
ut*, which would need to be *covariant* (allows narrowing), but it's actual=
ly an *input* so is *contravariant* (allows widening)=2E

If we state that default values, or their types, can reliably be inspected=
 and used by callers in this way, we would logically need to declare any op=
tional parameter *invariant* (must match exactly in all child classes), whi=
ch would be a huge loss of functionality=2E

Regards,
Rowan Tommins
[IMSoP]