in reply to Re: Simple Switch statement
in thread Simple Switch statement

I normally do just as dragonchild suggests, which is indeed right out of holy writ. However, don't be too quick to give up the old if/then/else. While maybe not as clean looking or as cool, the venerable standby can be about 2x as fast, though, admittedly, usually not noticeable unless you are testing thousands of conditions :-)

Switch/case is one of the RARE features I miss from traditionally frowned upon javascript.

Replies are listed 'Best First'.
Re: Re: Re: Simple Switch statement
by Roger (Parson) on Sep 13, 2003 at 01:55 UTC
    I agree with what bradcathey said about if/then/else. Also on testing of simple strings/values. The mundane
    if ($a eq "simple value")
{
  ...
}

    [download]
    runs faster than
    if ($a =~ /^simple value$/)
{
  ...
}

    [download]
    The point is to avoid using regular expressions on simple string testing. Any bit of speed increase in perl is good, especially when it has to be evaluated countless times.

[reply]
[d/l]
[select]

      Ah, but why not:

      for($switch_me) {
    /bob/   && do { print "Got 'bob'!\n"; break };
    /frnak/ && do { print "Got 'frnak'!\n"; break };
    $_ == 1 && do { print "Got 1!\n"; break };

    print "Got nuttin'!\n";
}

      [download]

      You can have your cake AND eat it too.

      Alakaboo

[reply]
[d/l]
        I have constructed the following test case to compare the performance of simple regular expression and simple eq test:

        use Benchmark;
@switch_me = qw / bob frnak 1 roger james /;

timethese (1000000, {
    'test_regex' => ' &test_regex; ',
    'test_eq' => ' &test_eq; '
});

sub test_regex()
{
    for(@switch_me) {
        /bob/   && do { break };
        /frnak/ && do { break };
        $_ == 1 && do { break };
    }
}

sub test_eq()
{
    for(@switch_me) {
        $_ eq "bob" && do { break };
        $_ eq "frnak" && do { break };
        $_ == 1 && do { break };
    }
}

        [download]
        The two subroutines are identical except for the string testing bit. The following are the test results:
        Benchmark: timing 1000000 iterations of test_eq, test_regex...
   test_eq:  8 wallclock secs ( 8.45 usr +  0.01 sys =  8.46 CPU)
             @ 118161.41/s (n=1000000)
test_regex: 10 wallclock secs ( 9.22 usr +  0.01 sys =  9.23 CPU)
            @ 108307.16/s (n=1000000)

        [download]
        It's about (118161.41-108307.16)/108307.16 = 10% speed increase in using eq than using regular expression. It's not a great increase, but it's nevertheless faster to use eq than regular expression.
[reply]
[d/l]
[select]

In Section Seekers of Perl Wisdom