Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:11652
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Message-ID: <5.1.0.14.2.20040729232459.02f31418@127.0.0.1>
Date: Thu, 29 Jul 2004 23:43:13 -0700
To: "Sara Golemon" <pollita@php.net>,internals@lists.php.net
In-Reply-To: <20040729235555.37129.qmail@pb1.pair.com>
References: <20040729021337.73894.qmail@pb1.pair.com>
 <1091136355.490.4.camel@localhost>
 <1091139076.29517.12.camel@blobule.suds>
 <410991E1.6070000@hristov.com>
 <1091140927.29516.18.camel@blobule.suds>
 <41099925.4020700@hristov.com>
 <20040729225322.94418.qmail@pb1.pair.com>
 <41099C4A.4080206@hristov.com>
 <75910961.20040729190842@ionzoft.com>
 <41099EB2.9090403@hristov.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Subject: Re: [PHP-DEV] GOTO operator
From: andi@zend.com (Andi Gutmans)

At 04:55 PM 7/29/2004 -0700, Sara Golemon wrote:
>Ack! Gods no.... Andi is already winding up to rip me a new one for
>prematurely committing the patch earlier today, don't push him over the edge
>by suggesting das uberevil ini options, even in jest! ;)
>
>-Sara
>
>P.S. - Andi: Ya know I love ya more than my luggage... :)

Just got home from OSCON and about to go to sleep.
Sara: I have gotten over my initial heart attack from this morning so don't 
worry about it :)

Let's for a second skip discussion of the patch itself, because although it 
was an interesting exercise the patch is not production ready.
However, I think before talking about implementation, we need to decide if 
goto is really good for PHP.
We talked about it quite a bit at OSCON, and I must say that I personally 
think that adding goto to PHP will do more harm than good.
It is true that for some advanced programmers, goto can come in handy, but 
the fact of the matter is that the PHP community has always taken a lot of 
pride in providing a language which keeps away from these kind of shortcuts.

I think there are only two cases where goto is really interesting:
a) Error handling.
b) Auto-generating code or compiler compilers (Sterling mentioned these two).

I think that the cases where you need it for error handling in PHP code 
aren't that great. Especially as the do..while(0) mechanism is much more 
useful in PHP than in other languages due to the ability to do "break n;". 
This ability in my opinion, is a good argument for why it's not as relevant 
in PHP as in other languages. And unlike someone mentioned in this thread, 
it is actually pretty fast and also takes care of all of the ugly stuff 
such as memory leaks from opcodes (not that it couldn't be implemented for 
goto but I'm just giving some more info and why it's a good thing).

I think that although slightly harder, code generators can probably also 
use similar constructs and survive without goto. I wouldn't want to add 
such a construct just for code generation. Sterling suggested calling the 
language construct something like "goto_only_use_for_autogenerated_code" 
(you get the point). His reasoning was that he doesn't really think it's 
needed by the average developer but it would be useful for automated tools. 
Not something I can completely disagree with but I think the downside 
outweighs the benefits.

I think that one can always find examples of how goto is useful, but it's 
hard to think ahead of time of how badly it will be abused. From my 
experience, although not quite as clear, it'll definitely be used in 
spaghetti code and will overall encourage bad coding practices, especially 
for less advanced programmers than people on this list.
Yes, any programmer can shoot himself in the foot with other PHP features, 
but we do have a responsibility to minimize that. One example is our 
introduction of E_STRICT. We don't warn about enough things but we do try. 
Why give our developers yet another way of shooting themselves in the foot? 
Or should I say, blow off their leg? :)

Andi