Thanks, Adrian. I knew it could be done as a one liner.
- wil
- wil
Nov 30, 2001, 1:27 AM
Veteran / Moderator (4108 posts)
Nov 30, 2001, 1:27 AM
Post #27 of 28
Views: 2580
OK, Paul. As you obviously don't believe me. Here goes...
I said:
$x = "12345123451234512345x456456456456y8989";
$x =~ s/\d+//g; # faster than s/\d//g
Can you see why?
And here's my reasoning:
If you use s/\d//g on that string, you will do 36 match-and-replace's.
If you use s/\d+//g, you do 3 match-and-replace's.
Running a benchmark...
cmpthese(-5, {
fast => sub {
my $x = "12345123451234512345x456456456456y8989";
$x =~ s/\d+//g;
},
slow => sub {
my $x = "12345123451234512345x456456456456y8989";
$x =~ s/\d//g;
}
});
__END__
Benchmark: running fast, slow for at least 5 CPU seconds...
fast: 48872.57/s (n=256581)
slow: 11055.86/s (n=60365)
Rate slow fast
slow 11056/s -- -77%
fast 48873/s 342% --
These results prove to you how much faster using \d+ is.
- wil
I said:
$x = "12345123451234512345x456456456456y8989";
$x =~ s/\d+//g; # faster than s/\d//g
Can you see why?
And here's my reasoning:
If you use s/\d//g on that string, you will do 36 match-and-replace's.
If you use s/\d+//g, you do 3 match-and-replace's.
Running a benchmark...
Code:
wil@localhost [12:20pm] ~ #130> bleadperl -MBenchmark=cmpthese cmpthese(-5, {
fast => sub {
my $x = "12345123451234512345x456456456456y8989";
$x =~ s/\d+//g;
},
slow => sub {
my $x = "12345123451234512345x456456456456y8989";
$x =~ s/\d//g;
}
});
__END__
Benchmark: running fast, slow for at least 5 CPU seconds...
fast: 48872.57/s (n=256581)
slow: 11055.86/s (n=60365)
Rate slow fast
slow 11056/s -- -77%
fast 48873/s 342% --
These results prove to you how much faster using \d+ is.
- wil
Dec 2, 2001, 8:37 AM
Veteran / Moderator (4108 posts)
Dec 2, 2001, 8:37 AM
Post #28 of 28
Views: 2599
Finally! I've found it. In "The Camel", 3rd edition - page 598 under the heading "Time Efficiency". Look at the second bullet point down:
If you're deleting characters, tr/abc//d is faster than s/[abc]//g.
End of argument. :-)
- wil