|
|
Posted by Rik on 07/31/07 01:53
On Tue, 31 Jul 2007 03:20:15 +0200, Jerry Stuckle =
<jstucklex@attglobal.net> wrote:
> Rik wrote:
>> On Tue, 31 Jul 2007 02:38:04 +0200, Jerry Stuckle =
>> <jstucklex@attglobal.net> wrote:
>>> I take this back - the result of the operation is unpredictable. =
>>> That's because:
>>>
>>> $rank[$i] =3D trim($rank[$i++]);
>>>
>>> $rank[$i++]
>>>
>>> is evaluated as $rank[$i]. However, when
>>>
>>> $rank[$i]
>>>
>>> is evaluated, does it use the old or the new version of $i?
>> This is indeed dubious. Which of the '[' gets evaluated first? =
>> Allthough, left-associtivity might indicate the first one should be =
>> evaluated first (PHP4 behaviour).
>> However, for:
>> $array[$i] =3D $i++;
>> Precedence as defined should kick in clearly. The '[' isn't in the =
>> table just for show.
>>
>>> The bottom line: don't change a value and use it in the same =
>>> statement. Results are unpredictable.
>> And makes for more readable code indeed. However, if I read the =
>> documentation I fully expect the PHP4 result from my example, not the=
=
>> PHP5 one...
>> --Rik Wasmus
He, still trouble with my sig-seperator? :P
> Rik,
>
> The key here is, precedence defines the order of OPERATOR evaluation -=
=
> but not the order of OPERAND evaluation. A subtle, but important =
> difference in cases like this.
>
> For instance, it defines that '++' will be evaluated before '[]',
Well, the other way around.
> it doesn't define whether [$i] will use the old or new value of $i++. =
=
> So even though [$i] is evaluated before [$i++], $i++ is evaluated firs=
t =
> and the value of $i is unpredictable.
Hmm, you've got a point there. However, I still wonder what warrented th=
e =
change from PHP4 to PHP5. Maybe it's got something to do with the fact =
references are more integrated, but that's just a wild guess.
What IS clear however, is that this is a nice example why one-liners are=
=
overrated. Nothing wrong with a few lines if it just clarifies things =
more, especially if it takes close to no extra CPU time and/or memory.
-- =
Rik Wasmus
[Back to original message]
|