Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:102874
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain gmail.com designates 209.85.208.176 as permitted sender)
MIME-Version: 1.0
Date: Mon, 16 Jul 2018 22:11:16 -0600
Message-ID: <CAFMT4NqPobFGvkcZ7nxaGHtpQsa+TOvdp13Ebab_6hH_859TCQ@mail.gmail.com>
To: internals <internals@lists.php.net>
Content-Type: text/plain; charset="UTF-8"
Subject: Replacing array_slice
From: levim@php.net (Levi Morrison)

Recently I've been scrutinizing array_slice. It's been painful.

For instance, `$preserve_keys` is totally ignored for string keys.
This was undocumented until Christoph integrated it 2-3 days ago
(thanks, cmb). However, this is really not a good design decision: why
have a parameter that is called `preserve_keys` if the data gets the
final say, especially since there isn't any technical requirement for
this?

Here's another "fun" behavior: negative offsets whose absolute values
are larger than the size of the array (eg the array has 3 elements and
you pass an offset of -5) will shift the window to begin at zero
instead of truncating the out-of-range values (which is what happens
on the positive side).

And another: `$length` is a length when it's a positive number, but
it's an offset from the end when it's negative. You could perhaps
convince me that when length is positive it's an offset from the
`$offset` parameter. However, if you look at the design of slicing
functions in other languages (eg Python, Ruby, JavaScript) they take
beginning and ending offsets, not a beginning offset and a length.

And another: it always iterates from the beginning of the array even
if you are only interested in the last half (somewhat common for
various algorithms that split things in half).

Fixing `array_slice` would probably do more harm than good at this
stage. Instead I would like to provide an alternative function that
does not have all this baggage, and will have decent performance much
of the time. The best ideas I have for names are not that great:

  - `array_real_slice`?
  - `array_slice2`?

We could also split it into 2 functions, one that preserves keys and
one that doesn't, instead of using a parameter. Any ideas there?