Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:76474
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain gmail.com designates 209.85.216.178 as permitted sender)
MIME-Version: 1.0
In-Reply-To: <942D2AF6-BDC2-4214-BD58-4B5945C28CE3@ajf.me>
References: <CAESVnVr0J5qyAc-L0Yt_3Ahab5LVgtUErgoKsSPvgwe=QzppTw@mail.gmail.com>
	<CAH-PCH4t1bnGwOmitU4cS4W--==8sZRhw1X-DkV0Y7zNj8E3kw@mail.gmail.com>
	<942D2AF6-BDC2-4214-BD58-4B5945C28CE3@ajf.me>
Date: Wed, 13 Aug 2014 12:57:40 +0200
Message-ID: <CAH-PCH64zsLDB4uQpRRaZ2JN5B8O_m08nt1rKp6H0wUJCofFEA@mail.gmail.com>
To: Andrea Faulds <ajf@ajf.me>
Cc: James <james@notjam.es>, Sara Golemon <pollita@php.net>, 
	PHP internals <internals@lists.php.net>
Content-Type: multipart/alternative; boundary=001a1132e2b623616a050080a89c
Subject: Re: [PHP-DEV] [RFC] Disallow multiple default blocks in a single
 switch statement
From: tyra3l@gmail.com (Ferenc Kovacs)

--001a1132e2b623616a050080a89c
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On Wed, Aug 13, 2014 at 11:42 AM, Andrea Faulds <ajf@ajf.me> wrote:

> Hi!
>
> On 13 Aug 2014, at 08:47, Ferenc Kovacs <tyra3l@gmail.com> wrote:
>
> > and I also think that this isn't an important enough issue to warrant a
> BC
> > break (albeit this is the better kind of BC: probably doesn't effect to=
o
> > many people, and they will be clearly notified about the error at compi=
le
> > time) so I voted no based on this two thing.
>
> This isn=E2=80=99t really a BC break. Multiple default blocks didn=E2=80=
=99t actually work
> anyway, we just silently ignored extra ones.
>

not sure what do you mean here, multiple default cases can be reached the
same way as any other duplicated case branch(as shown in my snippet in the
mail you replied to).
even if those were always "silently ignored" as you falsely claim, it would
still be a BC break, as code working before this change would stop working
with a fatal error.


>
> On 13 Aug 2014, at 09:13, James <james@notjam.es> wrote:
>
> > I entirely believe this behavior is weird and should be removed. Howeve=
r,
> > breaking backwards
> > compatibility in a minor release because the incomplete spec says so is
> > kind of odd.  A BC break
> > is a BC break, which doesn't belong in a minor revision.
>
> It isn=E2=80=99t a BC break that will affect anyone. It fixes a parser bu=
g.
>

it is a BC break, but I tend to agree that there isn't much code
intentionally using this behavior (but I'm also think that nobody really
looked for examples and just guessing about the usage of this feature), but
we had explicit phpt test for this behavior so I would call it a bad design
decision or lazy coding than an actual bug or hindsight.


>
> > This has been known to at least one person for many years if 034.phpt i=
s
> to
> > be believed.
>
> Just because it=E2=80=99s tested doesn=E2=80=99t mean anyone relies on it=
. We have plenty
> of tests which ensure PHP contains bugs and will error if they don=E2=80=
=99t.
>

sure, and those which we consider bugs are usually tests which expect the
correct behavior and marked as XFAIL so we won't forget that they should be
fixed.
those which are testing and expecting the "bad" behavior is for us to not
change those unintentionally, alas introduce unintentional BC breaks.


>
> > I could see
> > where people would use it - there are reasons to, even if they are poor
> in
> > choice to do so.
>
> How, exactly, could there ever be a use for having multiple default:
> sections and ignoring all but one?


I think you forget the fall-through cases (eg. a case/default without a
break/continue), in those cases all of the matching cases will be executed
(even the duplicated ones as they are all matching)


> This =E2=80=9Cfeature=E2=80=9D is completely and utterly useless, and I=
=E2=80=99ll eat my hat if
> anyone intentionally relied on it.
>

not sure about the multiple defaults, but I'm fairly sure that there are
code out there which uses switch blocks as state-machines, and changes the
switch variable in the case blocks, and have duplicated case blocks for
executing code after the switch variable reached it's final form.
and I consider the default block as a glorified wildcard case pattern, so I
don't think that it is a good idea to prevent having multiple default
blocks while allowing the normal case blocks to check for the same value
multiple times.

--=20
Ferenc Kov=C3=A1cs
@Tyr43l - http://tyrael.hu

--001a1132e2b623616a050080a89c--