Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:99046
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: softfail (pb1.pair.com: domain golemon.com does not designate 74.125.82.44 as permitted sender)
MIME-Version: 1.0
Sender: php@golemon.com
In-Reply-To: <CAOWwgpkcDqWUX=KAYXC73bB_4k66rG+AjPpCX8kEnUaWnJ69vA@mail.gmail.com>
References: <CAOWwgpm3Sq0nLeVcgXdGAjtcaQGuLgK=oTJRWFvQay4dDUWiXw@mail.gmail.com>
 <CAEZPtU5i5+LEmQkHgh1MAJ4XQi3uzaoeiQRJrNvp5Z26sE0kZA@mail.gmail.com> <CAOWwgpkcDqWUX=KAYXC73bB_4k66rG+AjPpCX8kEnUaWnJ69vA@mail.gmail.com>
Date: Mon, 15 May 2017 10:21:23 -0500
Message-ID: <CAESVnVrmPxg2cQrLNKXT1y4Pdg7LAEtHKZRa2+UtLFvSkPfCvQ@mail.gmail.com>
To: Nicolas Grekas <nicolas.grekas@gmail.com>
Cc: Pierre Joye <pierre.php@gmail.com>, 
	"internals@lists.php.net" <internals@lists.php.net>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Subject: Re: [PHP-DEV] BC break in master (7.2) - Distinguish between
 unmatched subpatterns and empty matches in preg_*()
From: pollita@php.net (Sara Golemon)

On Mon, May 15, 2017 at 6:36 AM, Nicolas Grekas
<nicolas.grekas@gmail.com> wrote:
>> I agree that a minimum 5 years old possible bug, quite small, causing
>> BC breaks is not good. As far as the fix is critical or justified, I
>> think it is sometimes ok to break BC for edge cases. However I do not
>> see this is not the case here.
>>
>
> Another possibility would be to make this opt-in, with a new flag.
> FYI, I quickly spotted at least two places in the code base where we chec=
k
> the matches with something like:
>
> if ('' !=3D=3D $matches[2]) {...} else {...}
>
> This should be a pretty common check so the BC break will impact more
> userland code when released.
>
I agree with the original bug report that being able to distinguish
between not-matched and empty-match is a useful and meaningful
distinction, and since most users won't care, then the decision to
just do it was *probably* right at the time.  Given the observation
about strict equality checks (which I don't have the data for, but it
rings true) it does seem optimistic in *handsight*.

Given that, I'd recommend a middle-ground such as the opt-in behavior
described here.  The extra effort to add that option flag in the rare
handful of cases where it matters is miniscule compared to the cost of
adjusting equality checks in the myriad places where it doesn't
matter, but shows up.

My 2 =C2=A2
-Sara