Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:115245
User-Agent: Cyrus-JMAP/3.5.0-alpha0-530-gd0c265785f-fm-20210616.002-gd0c26578
Mime-Version: 1.0
Message-ID: <8e6ad5a6-f627-4e3f-af04-6384ab6147f0@www.fastmail.com>
In-Reply-To: <17a5d374765.1289062a2301127.8166821672739325531@limesurvey.org>
References: <222b3921-3d9b-47f9-8d13-e6a123f36fad@www.fastmail.com>
 <CAC5V6ghFFjoP_JgUcJ14sFJ0DY-Q4z7qt40j9T5w0C3GHNsCrw@mail.gmail.com>
 <c16544c0-185a-43f9-853a-9941a018abe7@www.fastmail.com>
 <9f0c9565-22e1-4d94-bef5-cfc44ce02dc5@www.fastmail.com>
 <CAJpLVh1yMvjEbVaOsj3n7QLEhgJpPuchq6Y4t33X_8ACOehU5g@mail.gmail.com>
 <17a5d374765.1289062a2301127.8166821672739325531@limesurvey.org>
Date: Wed, 30 Jun 2021 12:10:37 -0500
To: "php internals" <internals@lists.php.net>
Content-Type: text/plain;charset=utf-8
Content-Transfer-Encoding: quoted-printable
Subject: Re: [PHP-DEV] [Vote] Partial Function Application
From: larry@garfieldtech.com ("Larry Garfield")

On Wed, Jun 30, 2021, at 8:59 AM, Olle H=C3=A4rstedt wrote:

>  > I've been pondering if a completely different approach with a prefi=
x=20
>  > symbol would be able to be less complex, and the simple answer is I=
=20
>  > have absolutely no idea.=C2=A0 But we are running low on symbols...=

> =20
>  Ah, I found the technical details that Joe gave me (right after I hit=
=20
> send, of course).=C2=A0 Quoting Joe:
> =20
>  "the engine expects certain things to happen, and is designed and the=
n=20
> optimized around those assumptions ... for example, a stream of INIT,=20=

> SEND, DO_FCALL is not meant to be interrupted, the first fundamental=20=

> change you have to make is making the engine aware that stream of INIT=
,=20
> SEND, + are not always followed by DO_FCALL "
> =20
>  So yes, it sounds like hooking into the function call process is wher=
e=20
> the complexity comes from.=C2=A0 Which suggests that an approach that =
works=20
> using a different syntax that desugars to a closure would avoid that=20=

> issue, but then we need a syntax that wouldn't be ambiguous, and that'=
s=20
> getting harder and harder to find.=C2=A0 (Nikita's first-class-callabl=
es RFC=20
> notes some of the issues with available symbols, and they're=20
> essentially the same for partials either way.)=C2=A0 And I've been tol=
d that=20
> creating closures in the AST compiler is Hard(tm)...
> =20
>  --Larry Garfield
>=20
>=20
> Wrapping stuff in lambdas is otherwise the obvious solution, no? Make=20=

> `strlen(?)` evaluate to `fn ($x) =3D> strlen($x)`. Does it depend on t=
he=20
> level of look-ahead in the compiler why it's so hard? Didn't work much=
=20
> with scripting language internals.
>=20
> Olle

The tricky part is that conversion has to happen entirely at runtime, be=
cause we need the type information from the function being partialed, an=
d at that point creating an actual closure is, apparently, rather hard. =
 It cannot be done up at the AST level where it would be conceptually mu=
ch easier.

We've been discussing this for the past several days in chat, and the ba=
sic conclusion is that the PHP engine does not offer any easy way to do =
this.  It's one hard-and-messy approach or another hard-and-messy approa=
ch.  So far no one has figured out a not hard-and-messy way to make it w=
ork, regardless of performance.

--Larry Garfield