Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:123786
Precedence: bulk
MIME-Version: 1.0
References: <2a6b92eb-d5e9-4a1a-9548-a068ac42ebd2@app.fastmail.com>
 <CAG4FTXz6cH26C5H51VmTGnpFK5XxROFbW4yDzwAw2v1gcQWGgA@mail.gmail.com>
 <CAPzBOBOTSCCKDR4pDocr4xW+oEXJ27YizgPLkn1PGR9nyPU7ww@mail.gmail.com> <1E295280-619B-4490-B53C-0899B64F9215@chaz.works>
In-Reply-To: <1E295280-619B-4490-B53C-0899B64F9215@chaz.works>
Date: Mon, 24 Jun 2024 21:53:43 +0200
Message-ID: <CAPzBOBPSSEmEbPj1zLcfGOOG7voo7YMUO16NjXzpTy7Wfx4ycw@mail.gmail.com>
Subject: Re: [PHP-DEV] [Early Feedback] Pattern matching
To: chaz@chaz.works
Cc: php internals <internals@lists.php.net>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
From: landers.robert@gmail.com (Robert Landers)

On Mon, Jun 24, 2024 at 7:28=E2=80=AFPM Chuck Adams <chaz@chaz.works> wrote=
:
>
>
> > On Jun 24, 2024, at 3:31=E2=80=AFAM, Robert Landers <landers.robert@gma=
il.com> wrote:
> >
> >
> > There's no need to use `?` to check for existence on a key, so this:
> >
> > $arr is ['a' =3D> string, ?'b' =3D> string, ...];
> >
> > should be this:
> >
> > $arr is ['a' =3D> string, 'b' =3D> ?string, ...];
> >
> > because $arr['non-existent-key'] is NULL.
> >
>
>
> The first means b is an optional key, but if it=E2=80=99s there, can only=
 be a string. The second says b is a required key, but it may be a string o=
r null. If there were a binding involved, that determines the type of the b=
inding in incompatible ways.  I=E2=80=99m fine with requiring bindings to b=
e nullable for optional keys, but it strikes me as strictly less flexible a=
nd not consistent with the rest of PHP=E2=80=99s behavior, at least not und=
er E_ALL.
>

Hi Chuck,

To be honest, this is one of the smaller concerns I have with the new
syntax. There might be some misunderstanding here, though. A
non-existent key is NULL, always has been, and always will be. No
current RFCs are stating otherwise, to my knowledge. While a warning
is emitted when accessing a non-existent key, and some people trigger
exceptions or fatal errors on warnings, this isn=E2=80=99t universal, and i=
t
isn=E2=80=99t always an error (after all, the value is NULL, which might be=
 a
properly handled case). If you see the warning, it=E2=80=99s usually a hint
that something might be wrong, but often, a quick ?? is all you need
to make the warning go away, further reinforcing that the value of a
non-existent key is NULL.

$arr =3D ['a' =3D> 'a string'];
$arr is ['a' =3D> string, ?'b' =3D> $value, ...];

This syntax implies that a non-existent key is a special case, and if
it passes as-is, it will be. If there is a binding and the key is
missing, what happens to that binding? Is it left in an undefined
state or is it assigned NULL? If it is assigned NULL, what is the
value of checking for non-existence; why not just check for NULL in
the value? If it is left "undefined", then you still have to handle it
as though it is a NULL value (perhaps more so because simply trying to
detect its value may cause the very warning you are trying to avoid --
and will fatally error in PHP 9: https://3v4l.org/Rbadk. see
https://wiki.php.net/rfc/undefined_variable_error_promotion).

Thus, this "key existence check" is inconsistent with the language and
possibly redundant, as you'd have to check that resulting value was
null anyway:

$arr is ['a' =3D> string, 'b' =3D> ?string, ...];

Hope that clarifies things.