Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:103740
MIME-Version: 1.0
References: <CAF+90c-V9vG8eAanQgujb+GhkLWtH2N0VRe500oocNQ7ryfb=Q@mail.gmail.com>
 <CADyq6s+qnFfZGMfr0HfyW6Qu9MC+_M_h3mQDOiDMJzWWTVHE-Q@mail.gmail.com>
In-Reply-To: <CADyq6s+qnFfZGMfr0HfyW6Qu9MC+_M_h3mQDOiDMJzWWTVHE-Q@mail.gmail.com>
Date: Tue, 15 Jan 2019 11:48:51 +0100
Message-ID: <CAOWwgpkk9WWQpn0EOmQD2A8XypCzGWJ1RpKSd5WWZ3kufq2Awg@mail.gmail.com>
To: Marco Pivetta <ocramius@gmail.com>, Nikita Popov <nikita.ppv@gmail.com>
Cc: PHP internals <internals@lists.php.net>
Content-Type: multipart/alternative; boundary="000000000000ed2eea057f7ce8a1"
Subject: Re: [PHP-DEV] [RFC] Reflection for references
From: nicolas.grekas+php@gmail.com (Nicolas Grekas)

--000000000000ed2eea057f7ce8a1
Content-Type: text/plain; charset="UTF-8"

Thank you Nikita, I will definitely need that RFC!

Answering to Marco:


 * `ReflectionReference` seems to be designed around arrays only: maybe
> `ReflectionArrayKeyReference` or such?
>

I agree with the wording in the RFC: the array is the primitive data
structure in PHP that holds references.
ALL other data structures can be turned to arrays. That includes objects
with `(array)` and scopes with `get_defined_vars()`.
Adding eg support for objects directly would open several questions, like:
what about magic methods?, etc.

So to me, we really need the simplest yet most powerful primite, and here
it is,  anything else can be built on top.



>  * Can the `getId()` return type be restricted to either `int` or `string`?
>

A string would be enough I suppose, but I'll let Nikita anwer :)



>  * what are possible scenarios for getting the identifier as a primitive,
>

Indexing. Eg VarCloner needs an index of already discovered references to
do its job.


>  * There seems to be a lot of design around `ReflectionReference#getId()`
> to avoid leaking internal pointer information: can it be completely
> dropped, if we have `$ref1->matches($ref2)` instead?
>

Nope, see previous answer. As such, I think `matches()` would be added
complexity for core with no extra benefit/power for users.

In short, I really like the RFC as is :)
(I removed your other comments Marco because I don't have anything special
to say on them)

Cheers,
Nicolas

--000000000000ed2eea057f7ce8a1--