Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:125998
Feedback-ID: i8414410d:Fastmail
Precedence: bulk
MIME-Version: 1.0
Date: Wed, 20 Nov 2024 00:01:29 -0600
To: "php internals" <internals@lists.php.net>
Message-ID: <1fd335a7-8764-4b26-be95-041d0bf94bbe@app.fastmail.com>
In-Reply-To: 
 <CABdc3WrWjH2zTTTtbbaS43-C+kJUeA5dpp_NcfosPfar04t4Pw@mail.gmail.com>
References: <d603f372-5f58-4308-aba1-151c35f3aa94@app.fastmail.com>
 <68eba37e-1bfe-4509-af41-112bb196415c@app.fastmail.com>
 <CABdc3WrWjH2zTTTtbbaS43-C+kJUeA5dpp_NcfosPfar04t4Pw@mail.gmail.com>
Subject: Re: [PHP-DEV] RFC: Records
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
From: larry@garfieldtech.com ("Larry Garfield")

On Tue, Nov 19, 2024, at 5:56 AM, Micha=C5=82 Marcin Brzuchalski wrote:
> Hi Larry,
>
> niedz., 17 lis 2024 o 08:24 Larry Garfield <larry@garfieldtech.com> na=
pisa=C5=82(a):
>> ...
>> I can see the benefit of an inline constructor.  Kotlin has something=
 similar.  But I can see the benefit of it for all classes, even service=
 classes, not just records.  (In Kotlin, it's used for service classes a=
ll the time.)
>
> What visibility would you expect for inline constructor properties in=20
> service classes?

Good question.  If we follow Kotlin's lead[1] (Kotlin has like 3-4 diffe=
rent places to do constructor-ish logic, it's kinda weird), then they'd =
default to public, but you can optionally specify a visibility. So, the =
same as any other property.  They do not imply readonly, but Kotlin has =
per-arg val/var markers on everything anyway to handle that.

[1] https://kotlinlang.org/docs/classes.html#constructors

>> I would far prefer assembling record-ish behavior myself, using the s=
maller parts above.  Eg:
>>=20
>> final readonly data class Point(int $x, int $y);
>>=20
>> "final" prevents extension.  "readonly" makes it immutable.  "data" g=
ives it value-passing semantics.  Any class can use an inline constructo=
r.  "with" is designed to work automatically on all objects.  Boom, I've=
 just assembled a Record out of its constituent parts, which also makes =
it easier for others to learn what I'm doing, because the features opted=
-in to are explicit, not implicit.
>
> It opens a bunch of concerns, questions like why would you use "data"=20
> keyword without having "final"?
> Although I like the "data" keyword very much.=20

To borrow a little syntax from Kotlin again, and randomly spitball:

data class Rectangle(int $h, int $w) {
    public int $area { get =3D> $this->h * $this->w; }
}

data class Square(int $side) extends Rectangle($side, $side);

(That calls the constructor of Rectangle with the provided values.)

I'm generally not in favor of final-by-default. final has its place, cer=
tainly, but I am firmly on team "no, you should not be making every clas=
s final by default, that's just silly."

Another thing to consider is ADTs.  (cf https://wiki.php.net/rfc/tagged_=
unions, though the text there is quite old and out dated at this point; =
do not take literally.)  The intent for enum cases with associated data =
was that you would specify properties and types, and those would be publ=
ic-readonly-enforced. That's the only option.  Ilija also had plans (and=
 I think implementation, but I'm not sure) for reusing instances with ma=
tching properties so they would point to the same object in memory, maki=
ng =3D=3D=3D work.  So:

enum Move {
  case Left(int $distance);
  case Right(int $distance);
}

$step1 =3D Move::Left(5);
$step2 =3D Move::Left(5);
$step1 =3D=3D=3D $step2; // true

That sounds somewhat similar to the restrictions proposed for Records, t=
hough in context I think they make more sense on enums.  I can see there=
 being overlap, though, so I mention it here for completeness.  I'm not =
sure how this all dovetails together.

To reiterate, my main issue with the Records concept is that it couples =
too many features together into an all-or-nothing package.  readonly did=
 that with only 2 features, and it caused Ilija and I a lot of heartburn=
 and helped keep aviz from passing in 8.3.  Records would couple 4 featu=
res together.  Once bitten, twice as cautious.

--Larry Garfield