Which is faster? switch vs. if

La_Parka

switch ($foo)
{
case 1:
$bar = "%";
break;

default:
$bar = "1";
break;
}

if ($foo == 1)
{
$bar = "%";
}
else
{
$bar = "1";
}

I'm probably splitting hairs, but which of the above is more efficient?

laserlight

I'm probably splitting hairs, but which of the above is more efficient?

You are definitely trying micro-optimisation that does not matter. My emprical tests show that the if-elseif-else version is slightly faster than a switch-case version. However, the difference is so slight as to be negligible in real use.

In your example, the switch does not make sense and obfuscates the flow control. If you really did need to compare multiple possible values to a single value, use a switch, or consider an array with in_array() or array_search().

NogDog

In a simple case like that, there's probably not much difference either way. A switch may gain you time though in situations where you have elseif's in there, as each elseif has to re-evaluate the conditional expression, whereas it is only evaluated once in a switch.

laserlight

A switch may gain you time though in situations where you have elseif's in there, as each elseif has to re-evaluate the conditional expression, whereas it is only evaluated once in a switch.

Surprisingly, comparing:

if ($foo == 1) {
} elseif ($foo == 2) {
} elseif ($foo == 3) {
} else {
}

and

switch ($foo)
{
case 1:
    break;
case 2:
    break;
case 3:
    break;
default:
}

a million times gave this result in seconds:

$foo  if      switch
1     0.205   0.249
2     0.287   0.329
3     0.382   0.422
4     0.395   0.395

I guess that the kind of optimisations possible for a switch for say, a C++ compiler is not so possible in normal PHP execution.

NogDog

If those are the values you're getting for a million iterations, I'm not going to choose which to use based on performance. 🙂

MarkR

It doesn't matter.
Use whichever enables you to create more maintainable code
Developer time is more expensive than CPU time, by a factor of several zillion
If you really, really, really do care, benchmark it.
If you benchmark it, do so in a production-spec environment

Mark

freebird2

Maybe it all depends on . In different situations the speed is different.