Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:128042
Feedback-ID: ifab94697:Fastmail
Precedence: bulk
MIME-Version: 1.0
Date: Mon, 14 Jul 2025 20:22:34 +0200
To: "Dmitry Derepko" <xepozzd@gmail.com>
Cc: internals@lists.php.net
Message-ID: <d4da4b69-7f13-4e8f-a043-d610d178479b@app.fastmail.com>
In-Reply-To: <0D6532F3-6E95-48B9-B394-E9CC1EC00B56@gmail.com>
References: <d603f372-5f58-4308-aba1-151c35f3aa94@app.fastmail.com>
 <0D6532F3-6E95-48B9-B394-E9CC1EC00B56@gmail.com>
Subject: Re: [PHP-DEV] RFC: Records
Content-Type: multipart/alternative;
 boundary=e4060a689181452ca751c81c072c00d8
From: rob@bottled.codes ("Rob Landers")

--e4060a689181452ca751c81c072c00d8
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable

Hey Dmitry,

Please remember to bottom post!

On Mon, Jul 14, 2025, at 11:16, Dmitry Derepko wrote:
> Hi, Rob!
>=20
> I'm just wondering, why the implementation differs from a regular clas=
s?=20
>=20
> Record makes a "class" immutable and unique, using field value compari=
son instead of object refs. It has a custom `with` method which is simil=
ar to the new `clone` operator accepted recently. I'd suggest not introd=
ucing a new keyword "record" and new syntax to create records, but use r=
egular `class` and `new Class`. Class should have a modifier "data" like=
 readonly, so it should be:
>=20
> data class Record {
>     public function __construct(public $a){} // $a is mutable, because=
 why not?
> }
>=20
> Data class will compare with others using field value comparison. More=
over, if you want to make it readonly, use the existing keyword to achie=
ve this:
>=20
> readonly data class Record {
>     public function __construct(public $a){} // $a is immutable, becau=
se of "readonly" class modifier
> }

Data classes are basically structs (as we collectively discovered last D=
ecember when I tried implementing exactly that). Records are implemented=
 completely differently than structs, even though they seem very similar=
. This is largely why they=E2=80=99re completely different keywords. Fur=
ther, records were meant to be nested inside other classes, and I did tr=
y to actually pull that off with regular classes =E2=80=94 and it was de=
clined. This negates the short-syntax and nested aspect of records.

>=20
> 1. So record Record {} becomes readonly data class Record {}

A record is not a class just like an enum is not a class. :) Though, it =
is class-like semantics.

> 2. We introduce data classes, which are similar to records and are mut=
able by default

You=E2=80=99re referring to structs, not records.

> 3. No new constructions to create the new type of classes, no adjustme=
nts for autoloading and other things internally

Records aren=E2=80=99t created (aka, "new'd"); that=E2=80=99s an intenti=
onal design choice. Whether or not it is a good one is still up for deba=
te.

> 4. Eliminate with method, replace `with` method with the new `clone`

The new clone is not compatible with records or any other type defined v=
ia a constructor (including regular classes). This was a deliberate choi=
ce of that implementation. Here's some examples run on that branch:

<?php

final readonly class Response {
    public function __construct(
        public int $statusCode,
        public string $reasonPhrase,
        // ...
    ) {
        if($this->statusCode >=3D 600) {
            throw new LogicException();
        }
    }
}

$test =3D new Response(404, "Not Found");
var_dump($test);
$test =3D clone($test, ['statusCode' =3D> 900]);
var_dump($test);

--- output ---

PHP Fatal error:  Uncaught Error: Cannot modify protected(set) readonly =
property Response::$statusCode from global scope in /home/withinboredom/=
code/php-src/test.php:28
Stack trace:
#0 /home/withinboredom/code/php-src/test.php(28): clone()
#1 {main}
  thrown in /home/withinboredom/code/php-src/test.php on line 28
Fatal error: Uncaught Error: Cannot modify protected(set) readonly prope=
rty Response::$statusCode from global scope in /home/withinboredom/code/=
php-src/test.php:28
Stack trace:
#0 /home/withinboredom/code/php-src/test.php(28): clone()
#1 {main}
  thrown in /home/withinboredom/code/php-src/test.php on line 28

Then removing the readonly distinction:

object(Response)#1 (2) {
  ["statusCode"]=3D>
  int(404)
  ["reasonPhrase"]=3D>
  string(9) "Not Found"
}
object(Response)#2 (2) {
  ["statusCode"]=3D>
  int(900)
  ["reasonPhrase"]=3D>
  string(9) "Not Found"
}

---

As you can see, it doesn't work for readonly classes and allows invalid =
mutable objects to be created, requiring devs to rethink how validation =
will be structured as this will make any constructor validations entirel=
y bypassable in PHP 8.5+. Previously, you needed to use reflection to do=
 this, but now it is going to be incredibly easy and a footgun. I think =
this will be another weird quirk of PHP from now on and it is by design.

> 5. Inline constructor isn=E2=80=99t necessary and could be proposed se=
parately. I=E2=80=99ve thought recently about this feature

I will probably remove this, to be honest. It was for nesting records in=
side classes or other records, but this feature was declined, so there i=
sn't really a need for it.

>=20
> WDYT?
>=20
>=20
> (sorry for the duplicate in private mailbox)
>=20
> ----------
>=20
> Best regards,
> Dmitrii Derepko.
> @xepozz

Data classes are much more similar to structs than to records -- and are=
 even implemented completely differently. Records are effectively just w=
ell-defined arrays with behaviour attached. They're not classes, but off=
er similar semantics.

=E2=80=94 Rob
--e4060a689181452ca751c81c072c00d8
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE html><html><head><title></title></head><body><div>Hey Dmitry,<=
/div><div><br></div><div>Please remember to bottom post!<br></div><div><=
br></div><div>On Mon, Jul 14, 2025, at 11:16, Dmitry Derepko wrote:</div=
><blockquote type=3D"cite" id=3D"qt" style=3D"overflow-wrap:break-word;"=
><div>Hi, Rob!</div><div><br></div><div>I'm just wondering, why the impl=
ementation differs from a regular class?&nbsp;</div><div><br></div><div>=
Record makes a "class" immutable and unique, using field value compariso=
n instead of object refs. It has a custom `with` method which is similar=
 to the new `clone` operator accepted recently. I'd suggest not introduc=
ing a new keyword "record" and new syntax to create records, but use reg=
ular `class` and `new Class`. Class should have a modifier "data" like r=
eadonly, so it should be:</div><div><br></div><div>data class Record {</=
div><div>&nbsp; &nbsp; public function __construct(public $a){} // $a is=
 mutable, because why not?</div><div>}</div><div><br></div><div>Data cla=
ss will compare with others using field value comparison. Moreover, if y=
ou want to make it readonly, use the existing keyword to achieve this:</=
div><div><br></div><div>readonly data class Record {</div><div>&nbsp; &n=
bsp; public function __construct(public $a){} // $a is immutable, becaus=
e of "readonly" class modifier</div><div>}</div></blockquote><div><br></=
div><div>Data classes are basically structs (as we collectively discover=
ed last December when I tried implementing exactly that). Records are im=
plemented completely differently than structs, even though they seem ver=
y similar. This is largely why they=E2=80=99re completely different keyw=
ords. Further, records were meant to be nested inside other classes, and=
 I did try to actually pull that off with regular classes =E2=80=94 and =
it was declined. This negates the short-syntax and nested aspect of reco=
rds.</div><div><br></div><blockquote type=3D"cite" id=3D"qt" style=3D"ov=
erflow-wrap:break-word;"><div><br></div><div>1. So record Record {} beco=
mes readonly data class Record {}</div></blockquote><div><br></div><div>=
A record is not a class just like an enum is not a class. :) Though, it =
is class-like semantics.</div><div><br></div><blockquote type=3D"cite" i=
d=3D"qt" style=3D"overflow-wrap:break-word;"><div>2. We introduce data c=
lasses, which are similar to records and are mutable by default</div></b=
lockquote><div><br></div><div>You=E2=80=99re referring to structs, not r=
ecords.</div><div><br></div><blockquote type=3D"cite" id=3D"qt" style=3D=
"overflow-wrap:break-word;"><div>3. No new constructions to create the n=
ew type of classes, no adjustments for autoloading and other things inte=
rnally</div></blockquote><div><br></div><div>Records aren=E2=80=99t crea=
ted (aka, "new'd"); that=E2=80=99s an intentional design choice. Whether=
 or not it is a good one is still up for debate.</div><div><br></div><bl=
ockquote type=3D"cite" id=3D"qt" style=3D"overflow-wrap:break-word;"><di=
v>4. Eliminate with method, replace `with` method with the new `clone`</=
div></blockquote><div><br></div><div>The new clone is not compatible wit=
h records or any other type defined via a constructor (including regular=
 classes). This was a deliberate choice of that implementation. Here's s=
ome examples run on that branch:</div><div><br></div><div>&lt;?php</div>=
<div><br></div><div>final readonly class Response {</div><div>&nbsp;&nbs=
p;&nbsp; public function __construct(</div><div>&nbsp;&nbsp;&nbsp;&nbsp;=
&nbsp;&nbsp;&nbsp; public int $statusCode,</div><div>&nbsp;&nbsp;&nbsp;&=
nbsp;&nbsp;&nbsp;&nbsp; public string $reasonPhrase,</div><div>&nbsp;&nb=
sp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; // ...</div><div>&nbsp;&nbsp;&nbsp; ) =
{</div><div>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; if($this-&gt;stat=
usCode &gt;=3D 600) {</div><div>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbs=
p;&nbsp;&nbsp;&nbsp;&nbsp; throw new LogicException();</div><div>&nbsp;&=
nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; }</div><div>&nbsp;&nbsp;&nbsp; }<br>=
</div><div>}</div><div><br></div><div>$test =3D new Response(404, "Not F=
ound");</div><div>var_dump($test);</div><div>$test =3D clone($test, ['st=
atusCode' =3D&gt; 900]);</div><div>var_dump($test);</div><div><br></div>=
<div>--- output ---</div><div><br></div><div>PHP Fatal error:&nbsp; Unca=
ught Error: Cannot modify protected(set) readonly property Response::$st=
atusCode from global scope in /home/withinboredom/code/php-src/test.php:=
28</div><div>Stack trace:</div><div>#0 /home/withinboredom/code/php-src/=
test.php(28): clone()</div><div>#1 {main}</div><div>&nbsp; thrown in /ho=
me/withinboredom/code/php-src/test.php on line 28</div><div>Fatal error:=
 Uncaught Error: Cannot modify protected(set) readonly property Response=
::$statusCode from global scope in /home/withinboredom/code/php-src/test=
.php:28</div><div>Stack trace:</div><div>#0 /home/withinboredom/code/php=
-src/test.php(28): clone()</div><div>#1 {main}</div><div>&nbsp; thrown i=
n /home/withinboredom/code/php-src/test.php on line 28</div><div><br></d=
iv><div>Then removing the readonly distinction:</div><div><br></div><div=
>object(Response)#1 (2) {</div><div>&nbsp; ["statusCode"]=3D&gt;</div><d=
iv>&nbsp; int(404)</div><div>&nbsp; ["reasonPhrase"]=3D&gt;</div><div>&n=
bsp; string(9) "Not Found"</div><div>}</div><div>object(Response)#2 (2) =
{</div><div>&nbsp; ["statusCode"]=3D&gt;</div><div>&nbsp; int(900)</div>=
<div>&nbsp; ["reasonPhrase"]=3D&gt;</div><div>&nbsp; string(9) "Not Foun=
d"</div><div>}</div><div><br></div><div>---</div><div><br></div><div>As =
you can see, it doesn't work for readonly classes and allows invalid mut=
able objects to be created, requiring devs to rethink how validation wil=
l be structured as this will make any constructor validations entirely b=
ypassable in PHP 8.5+. Previously, you needed to use reflection to do th=
is, but now it is going to be incredibly easy and a footgun. I think thi=
s will be another weird quirk of PHP from now on and it is by design.</d=
iv><div><br></div><blockquote type=3D"cite" id=3D"qt" style=3D"overflow-=
wrap:break-word;"><div>5. Inline constructor isn=E2=80=99t necessary and=
 could be proposed separately. I=E2=80=99ve thought recently about this =
feature</div></blockquote><div><br></div><div>I will probably remove thi=
s, to be honest. It was for nesting records inside classes or other reco=
rds, but this feature was declined, so there isn't really a need for it.=
</div><div><br></div><blockquote type=3D"cite" id=3D"qt" style=3D"overfl=
ow-wrap:break-word;"><div><br></div><div>WDYT?</div><div><div dir=3D"aut=
o" style=3D"overflow-wrap:break-word;"><br></div></div><div><br></div><d=
iv>(sorry for the duplicate in private mailbox)</div><div><div dir=3D"au=
to" style=3D"color:rgb(0, 0, 0);letter-spacing:normal;text-align:start;t=
ext-indent:0px;text-transform:none;white-space-collapse:collapse;text-wr=
ap-mode:wrap;word-spacing:0px;-webkit-text-stroke-width:0px;text-decorat=
ion-line:none;text-decoration-thickness:initial;text-decoration-style:in=
itial;text-decoration-color:initial;overflow-wrap:break-word;"><div><div=
><br></div><div>----------</div><div><br></div><div>Best regards,</div><=
div>Dmitrii Derepko.</div><div>@xepozz</div></div></div></div></blockquo=
te><div><br></div><div>Data classes are much more similar to structs tha=
n to records -- and are even implemented completely differently. Records=
 are effectively just well-defined arrays with behaviour attached. They'=
re not classes, but offer similar semantics.</div><div><br></div><div id=
=3D"sig121229152">=E2=80=94 Rob</div></body></html>
--e4060a689181452ca751c81c072c00d8--