Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:129947
Date: Wed, 28 Jan 2026 21:25:09 +0100
To: Larry Garfield <larry@garfieldtech.com>, internals@lists.php.net
Message-ID: <5c291a56-baee-48db-93a8-27c8dd903bc0@Spark>
In-Reply-To: <56dce794-a512-421a-a8d2-91a9692c1ab4@app.fastmail.com>
References: <e25ea0a6-bb82-4762-9e51-a543d0e6efe5@app.fastmail.com>
 <56dce794-a512-421a-a8d2-91a9692c1ab4@app.fastmail.com>
Subject: Re: [PHP-DEV] [RFC] Pattern Matching
Precedence: list
MIME-Version: 1.0
Content-Type: multipart/alternative; boundary="697a70aa_6a5f7029_d0e3"
From: jordikroon@me.com (Jordi Kroon)

--697a70aa_6a5f7029_d0e3
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

Hi,
> 1. match() statements. There's two ways we could do this: All-branches,=
 or individual branches. This is particularly important as, in practice, =
we expect match() to be the most common use of patterns.
I prefer to have one syntax. The most flexible one is individual arms. An=
d to me it also looks more clean from the functional side.

If we support both we should also disallow the use of mixing all arms and=
 individual arms. Thus this syntax should return an error.
=24result =3D match (=24somevar) is =7B // Error=21
   is =46oo =3D> 'foo',
   is Bar =3D> 'bar',
   is Baz=7CBeep =3D> 'baz',
=7D;

> Positional array patterns. How picky should the pattern matching be for=
 list-esque arrays=3F There's 3 options listed in the R=46C as possible w=
ays forward, so I won't repeat them here but just provide a link:
>
> https://wiki.php.net/rfc/pattern-matching=23positional=5Farray=5Fenforc=
ement

=46rom a cost perspective it would be best to keep it to a minimum. But a=
rray=E2=80=99s exist in different styles where it matters to have at leas=
t a few edge-cases covered. Especially when converting between data forma=
ts (Serialise, JSON, ..)

I would personally discourage to return true with the following cases:

Given:
=24a =3D =5B'a', 'b', 'c'=5D;
=24a is =5B2 =3D> 'c', 0 =3D> 'a', 1 =3D> 'b'=5D;
Should return =60false=60, because when converting to other data formats.=
 The style will change from numerical to key'ed (associative style). The =
same applies to the inverse.

Though given:
=24a =3D =5B'a', 'b', 'c'=5D;
=24a is =5B0 =3D> 'a', 1 =3D> 'b', 2 =3D> 'c'=5D;
Should return =60true=60, because they keep the numerical order. And when=
 converted it will be the same.

> 3. Do we want to have a pattern for =22match an object's properties wit=
hout specifying what type the object has to be=3F=22 If so, what syntax s=
hould it be=3F Just omitting the class name (producing =60=24o is =7Bx: 1=
=7D=60 ) is not possible. We tried. :-) Options include:
>
> =60=24o is =5F(x: 1)=60
> =60=24o is object(x: 1)=60
I am not a favour of using shorthands. They tend to make code more diffic=
ult to read and understand.=C2=A0=C2=A0Thus =60=24o is object(x: 1)=60 is=
 what I suggest as well.


4.
> =C2=A0The variable pinning syntax. There's been some concerns about it =
being non-obvious, which is valid. The main argument for it is =22it's wh=
at Ruby does=22, which is not the most compelling argument, but it's not =
invalid.
I strongly agree that the the =60=5E=60 is not a good choice. It is diffi=
cult to understand,I personally think pinning at all is not important. Th=
ough if we proceed forward on this =60=7B=24x=7D=60 would make most sense=
 since it is more in line with string interpolation syntax. And like you =
mentioned it opens more possibilities.

Regards,
Jordi



--697a70aa_6a5f7029_d0e3
Content-Type: text/html; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

<html xmlns=3D=22http://www.w3.org/1999/xhtml=22>
<head>
<title></title>
</head>
<body>
<div name=3D=22messageBodySection=22>
<div dir=3D=22auto=22>Hi,</div>
<blockquote style=3D=22border-left-color: rgb(26, 188, 156); margin: 5px;=
 padding-left: 10px; border-left-width: thin; border-left-style: solid;=22=
>1. match() statements. There's two ways we could do this: All-branches, =
or individual branches. This is particularly important as, in practice, w=
e expect match() to be the most common use of patterns.</blockquote>
<div dir=3D=22auto=22>I prefer to have one syntax. The most flexible one =
is individual arms. And to me it also looks more clean from the functiona=
l side.<br />
<br />
If we support both we should also disallow the use of mixing all arms and=
 individual arms. Thus this syntax should return an error.<br /></div>
<pre>=24result =3D match (=24somevar) is =7B // Error=21
    is =46oo =3D&gt; 'foo',
    is Bar =3D&gt; 'bar',
    is Baz=7CBeep =3D&gt; 'baz',
=7D;</pre>
<div dir=3D=22auto=22><br /></div>
<blockquote style=3D=22border-left-color: rgb(26, 188, 156); margin: 5px;=
 padding-left: 10px; border-left-width: thin; border-left-style: solid;=22=
>Positional array patterns. How picky should the pattern matching be for =
list-esque arrays=3F There's 3 options listed in the R=46C as possible wa=
ys forward, so I won't repeat them here but just provide a link:<br />
<br />
https://wiki.php.net/rfc/pattern-matching=23positional=5Farray=5Fenforcem=
ent</blockquote>
<div dir=3D=22auto=22><br />
=46rom a cost perspective it would be best to keep it to a minimum. But a=
rray=E2=80=99s exist in different styles where it matters to have at leas=
t a few edge-cases covered. Especially when converting between data forma=
ts (Serialise, JSON, ..)<br />
<br />
I would personally discourage to return true with the following cases:<br=
 />
<br />
Given:</div>
<pre>=24a =3D =5B'a', 'b', 'c'=5D;
=24a is =5B2 =3D&gt; 'c', 0 =3D&gt; 'a', 1 =3D&gt; 'b'=5D;</pre>
<div dir=3D=22auto=22>Should return =60false=60, because when converting =
to other data formats. The style will change from numerical to key'ed (as=
sociative style). The same applies to the inverse.<br />
<br />
Though given:</div>
<pre>=24a =3D =5B'a', 'b', 'c'=5D;
=24a is =5B0 =3D&gt; 'a', 1 =3D&gt; 'b', 2 =3D&gt; 'c'=5D;
Should return =60true=60, because they keep the numerical order. And when=
 converted it will be the same.</pre>
<div dir=3D=22auto=22><br /></div>
<blockquote style=3D=22border-left-color: rgb(26, 188, 156); margin: 5px;=
 padding-left: 10px; border-left-width: thin; border-left-style: solid;=22=
>3. Do we want to have a pattern for =22match an object's properties with=
out specifying what type the object has to be=3F=22 If so, what syntax sh=
ould it be=3F Just omitting the class name (producing =60=24o is =7Bx: 1=7D=
=60 ) is not possible. We tried. :-) Options include:<br />
<br />
=60=24o is =5F(x: 1)=60<br />
=60=24o is object(x: 1)=60&=23160;</blockquote>
<div dir=3D=22auto=22>I am not a favour of using shorthands. They tend to=
 make code more difficult to read and understand.&=23160;&=23160;Thus =60=
=24o is object(x: 1)=60 is what I suggest as well.<br />
<br />
<br />
4.</div>
<blockquote style=3D=22border-left-color: rgb(26, 188, 156); margin: 5px;=
 padding-left: 10px; border-left-width: thin; border-left-style: solid;=22=
>&=23160;The variable pinning syntax. There's been some concerns about it=
 being non-obvious, which is valid. The main argument for it is =22it's w=
hat Ruby does=22, which is not the most compelling argument, but it's not=
 invalid.</blockquote>
<div dir=3D=22auto=22>I strongly agree that the the =60=5E=60 is not a go=
od choice. It is difficult to understand,I personally think pinning at al=
l is not important. Though if we proceed forward on this =60=7B=24x=7D=60=
 would make most sense since it is more in line with string interpolation=
 syntax. And like you mentioned it opens more possibilities.<br />
<br />
Regards,<br />
Jordi</div>
</div>
<div name=3D=22messageSignatureSection=22><br />
<br /></div>
</body>
</html>

--697a70aa_6a5f7029_d0e3--