Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:54235
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain gmail.com designates 209.85.161.42 as permitted sender)
MIME-Version: 1.0
Sender: ekneuss@gmail.com
In-Reply-To: <4E31328C.9030208@mageekbox.net>
References: <4E31328C.9030208@mageekbox.net>
Date: Thu, 28 Jul 2011 20:25:20 +0200
Message-ID: <CACQesk7OiWGkkKzHTY7BYW=LB_ERheVne42EnvpWhUQVqOUdyw@mail.gmail.com>
To: frederic.hardy@mageekbox.net
Cc: PHP Internals <internals@lists.php.net>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Subject: Re: [PHP-DEV] Exception throwed if parent::__construct() not called
 in \DirectoryIterator subclass
From: colder@php.net (Etienne Kneuss)

Hello,

2011/7/28 Fr=C3=A9d=C3=A9ric Hardy <frederic.hardy@mageekbox.net>:
> Hello !
>
> In PHP 5.4 alpha, an exception is throwing if a subclass of
> \DirectoryIterator not called the parent constructor.
> Moreover, this exception can not be catched in the constructor of the
> subclass.
> This behavious seems to be a good idea, because \DirectoryIterator may be=
 in
> an inconsistant state if the parent constructor is not called.
> But i think that it's not a good idea, because it's a BC break.
> It's not the case in 5.3 and i have unit test cases which fail by this ne=
w
> behavior, because moked \DirectoryIterator class does not call the parent
> constructor.
> In context of unit test, the fact that \DirectoryIterator is inconsistant
> may be not a problem and can be the desired behavior.

We have two ways internally of execute code that needs to be executed
as soon as we instantiate the object:

1) In the constructor
2) In the handler managing instantiation.

Doing it in (1) is problematic as PHP gives no implicit guarantee that
it will be called. Doing it in (2) is also problematic as it makes the
class "magic", by putting code with effects in a place where it
doesn't belong.
From a design perspective it is better to execute such code in the
constructor, especially if it takes argument, so (1) is "more correct"
than (2).

Now, as I said, PHP does not force you to call the parent constructor.
In almost every cases, not calling the overridden constructor is bad
practice and yields inconsistent results.
For internal classes, skipping code that often do mostly
initialization is dangerous and leads to crashes.

Now, for the issue at hand:
Even though it is a BC break, I believe that it is a very small one
and that it should be seen as a bug fix.
Note that the restriction is only here if the user tries to do
something terribly wrong.

IMO, it should remain impossible to extend DirectoryIterator's
__construct without calling the parent construct.

Best,

>
> There is a bug report : https://bugs.php.net/bug.php?id=3D55300
>
> Best regards,
> Fred
> --
>
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> Fr=C3=A9d=C3=A9ric Hardy : Architecte d'application/Admin. syst=C3=A8me/E=
rgonome
> =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 CV : http://blog.mageekbox.net/public/=
cv.frederic.hardy.pdf
> =C2=A0 =C2=A0 =C2=A0 =C2=A0 Blog : http://blog.mageekbox.net
> =C2=A0 =C2=A0 =C2=A0Twitter : http://twitter.com/mageekguy
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>
>
> --
> PHP Internals - PHP Runtime Development Mailing List
> To unsubscribe, visit: http://www.php.net/unsub.php
>
>



--=20
Etienne Kneuss
http://www.colder.ch