Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:89997
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain zort.net designates 96.241.205.2 as permitted sender)
Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 8.2 \(2104\))
In-Reply-To: <BD.65.07292.E46AA865@pb1.pair.com>
Date: Mon, 4 Jan 2016 13:34:10 -0500
Cc: internals@lists.php.net
Content-Transfer-Encoding: quoted-printable
Message-ID: <9D1F795D-DFDE-4B35-995B-4DE9EC5650E4@zort.net>
References: <070A1824-9ED7-4863-92AB-A4366AB92C98@zort.net> <BD.65.07292.E46AA865@pb1.pair.com>
To: Andrea Faulds <ajf@ajf.me>
Subject: Re: [PHP-DEV] Re: [RFC] array_key_(first|last|index) functions proposal
From: jbafford@zort.net (John Bafford)

Hi Andrea,

> On Jan 4, 2016, at 12:05, Andrea Faulds <ajf@ajf.me> wrote:
>=20
> Hi John,
>=20
> John Bafford wrote:
>> Happy New Year, everyone!
>>=20
>> I=E2=80=99d like to present the first new PHP RFC for this year, a =
proposal to add functions to easily get the first, last, or an arbitrary =
key (and value) by index from an array, taking advantage of PHP=E2=80=99s =
property that arrays are ordered maps.
>>=20
>> RFC: https://wiki.php.net/rfc/array_key_first_last_index
>> PR: https://github.com/php/php-src/pull/347
>=20
> How often would such functions be useful? Perhaps they fill a gap, but =
I'm not sure if it's one that needs filling. array_key_first and =
array_key_last can already be accomplished in two or so lines of code =
(four if you make a function), and array_key_index can be implemented in =
a few lines with a foreach() loop and a counter.

array_key_first() and array_key_last() can=E2=80=99t be implemented in =
userspace and maintain all three of fast, immutable, and =
doesn=E2=80=99t-look-weird.

The best you could do for array_key_first() is:

function array_key_first($arr) {
	foreach($arr as $k =3D> $v)
		return $k;
=09
	return null;
}

Which already looks kind of weird. For array_key_last, your best bet is

function array_key_last($arr) {
	$k =3D null;=09
	foreach($arr as $k =3D> $v)
		;
	//Rely on the fall-through of $k from the last iteration
	return $k;
}

which looks even weirder

or=20

function array_key_last($arr) {
	$keys =3D array_keys($arr);
	if($cnt =3D count($keys))
		return $keys[$cnt - 1];
=09
	return null;
}

Which adds the overhead of array_keys() iterating over the array and =
duplicating *all* of the array=E2=80=99s keys just to retrieve one.

Any solution that uses reset() + key() or end() + key() mutates the =
array, which means duplicating it if you pass it into a function. This =
takes a non-trivial time if you have a large array.


> array_key_first and array_key_last seem mostly harmless, at least. I'm =
not sure the same can be said for array_key_index, since it has O(n) =
complexity. I worry that it might end up used by people who think it is =
more efficient than iterating through an array, even though it is not. I =
would rather we not include this specific function, and avoid =
potentially disguising the time complexity of key lookup, especially as =
I can't think of a good use-case for it. As for array_key_first and =
array_key_last, well, maybe they're useful, I have no strong opinion on =
them. They might be handy to get the first and last key of an array =
without moving the internal pointer, so who knows, maybe we should add =
them.

I=E2=80=99ll admit that array_key_index() could be easily abused. If you =
need to use it more than a few times on a particular array, unless =
you=E2=80=99ve got a huge array and need to keep memory use constrained, =
you=E2=80=99re probably better off just calling array_keys() and using =
an index into the result. This would need to be included in the =
documentation as a use case caveat, though it wouldn=E2=80=99t be =
surprising if people ignored the warning and used it incorrectly anyway.

The reason array_key_index() is included in the RFC is because it kind =
of fell out of the original implementation and since it=E2=80=99s kind =
of useful, for a limited set of problems, I kept it in. It=E2=80=99s =
really more of a special-case tool with (very) limited general use, but =
it is a substantial improvement over the alternatives if you really do =
it need.

My primary goal with this RFC is array_key_first() and array_key_last(), =
so I would not be against having array_key_index() be a separate voting =
choice, if that would make people more comfortable.


> Thanks.
> --=20
> Andrea Faulds
> https://ajf.me/


-John