Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:110879
Content-Type: multipart/alternative; boundary=Apple-Mail-5858FA66-9135-4422-B2CF-09FC83F60991
Content-Transfer-Encoding: 7bit
Mime-Version: 1.0 (1.0)
Message-ID: <7E896FDC-D3D3-49F3-9B64-84C5A605B3B6@joshbruce.dev>
Date: Wed, 8 Jul 2020 09:42:42 -0500
To: Rowan Tommins <rowan.collins@gmail.com>
Cc: internals@lists.php.net
Subject: Re:  [PHP-DEV] Proposal: A way for classes to define a response to any primitive type cast
From: josh@joshbruce.dev (Josh Bruce)

--Apple-Mail-5858FA66-9135-4422-B2CF-09FC83F60991
Content-Type: text/plain;
	charset=utf-8
Content-Transfer-Encoding: quoted-printable

=EF=BB=BF

> On Jul 7, 2020, at 3:24 PM, Rowan Tommins <rowan.collins@gmail.com> wrote:=

>=20
>> On 07/07/2020 18:13, Josh Bruce wrote:
>> Decided to put something together:https://github.com/joshbruce/external-p=
roject-proposals/blob/master/php-concepts/interact-with-instance-as-php.md
>=20
>=20
> Hi Josh,
>=20
> Reading through this, it seems to cover a few different concepts, which ar=
e certainly related, but not quite the same.
>=20
> Firstly, we have interfaces such as Countable and Serializable, which don'=
t make the object act like any particular primitive type, they just customis=
e its interaction with particular functions. Iterators could potentially be i=
ncluded in that category, too, because it's fairly easy to imagine the langu=
age having multiple primitive types that can be iterated (e.g. separate "lis=
t" and "dictionary" types).
>=20
> ArrayAccess feels more explicitly about making an object "act like" an arr=
ay, but more precisely it overloads the [] operator - it doesn't allow the o=
bject to be used in other contexts where arrays are allowed. More general op=
erator overloading was discussed recently, and the particular proposal put f=
orward did not pass its RFC vote. [1]
>=20
>  __toString, on the other hand, only allows the object to be *converted to=
* a string, not act like one. You can't, for instance, influence the behavio=
ur of the concatenation operator, and make "Hello" . $object . "!" return an=
 object instead of a string, or a string that doesn't begin "Hello". An __to=
Array method was proposed recently, to mixed reaction. [2]
>=20
> The other closely related topic is what is often discussed as "scalar meth=
ods" or "scalar objects" - the ability to call methods on non-object values,=
 and possibly to define your own methods that can be called in this way. Thi=
s would seem to fit your use case with Shoop quite well: rather than Shoop::=
array($array)->first()->unfold(), you would just call $array->first(). It's f=
airly high on a lot of people's wish lists, but there's a lot of details to g=
et right in both design and implementation, so I strongly recommend searchin=
g for previous discussions and proofs of concept before diving in.
>=20
>=20
> [1] https://wiki.php.net/rfc/userspace_operator_overloading
> [2] https://externals.io/message/105589
>=20
> --=20
> Rowan Tommins (n=C3=A9 Collins)
> [IMSoP]
>=20
> --=20
> PHP Internals - PHP Runtime Development Mailing List
> To unsubscribe, visit: https://www.php.net/unsub.php


All: Going to change threads now that I=E2=80=99m getting a better feel for t=
he list environment and conventions, I think.=20

Someniatko: Saw the pipes RFC a while back and really appreciate. Think both=
 could have their place and developed preference (switch and if-else, for ex=
ample). Added reference to document. See link below.

Rowan: Thanks for the feedback and links!

PHP Internals is definitely not my wheelhouse and I find myself speaking mai=
nly from the developer=E2=80=99s perspective because I don=E2=80=99t know en=
ough to speak from the PHP perspective. Thanks for your patience there.

First takeaway is that I should drastically reduce scope and cut-short the b=
ackground, lay of the land, sales pitch language I=E2=80=99ve seen in other R=
FCs and concepts.

Second, I should consider the audience more. Most of the people on this list=
 probably (read most definitely) know more about this than I do (or ever wil=
l) and y=E2=80=99all don=E2=80=99t need long-winded explanations and backgro=
und (could be wrong). So, thinking less text, more code.

Finally, as I see more of the landscape and given conversations are low-band=
width (long form, text-only), trying to cover more than one thing would most=
 likely distract at best and block at worst.

So, gonna go after defining a boolean return value for a class. (At least fo=
r now.)

Updated: https://bit.ly/php-0001

Thank you again for the feedback and links.

Cheers,
Josh

ps. The ability to call methods on non-object values is interesting and coul=
d definitely be helpful for Shoop. The main frustration for me when using Sh=
oop is the inability to just drop the pointer into a condition and have it w=
ork. :)

pps. The operator overloading feels like Swift=E2=80=A6nice feature...that w=
ould definitely be helpful. While it doesn=E2=80=99t do the array thing itse=
lf, I would be able to implement ArrayAccess and Iterator on it, yeah? At wh=
ich point it could feel like an array (if I skimmed properly) - throw in __t=
oString and we have the Pym primitive. :)=

--Apple-Mail-5858FA66-9135-4422-B2CF-09FC83F60991--