13 Responses to “Stripping whitespace from both ends of a string…”

1. chackoc says:

   March 1, 2010 at 7:37 pm

   $str =~ s/(^\s+|\s+$)//g;

   Reply
2. brunov says:

   March 1, 2010 at 7:37 pm

   For the epitome of clear:

   use perl5i::latest;

   $str->trim;

   Reply
3. Adam Kennedy says:

   March 1, 2010 at 8:19 pm

   $str =~ s/(?:^\s+|\s+$)//g;

   Reply
4. zloyrusskiy says:

   March 1, 2010 at 8:31 pm

   I’m using:

   s/(?:^\s+|\s+$)//g

   it’s one regular expression + has good speed

   Reply
5. Mark Grimes says:

   March 1, 2010 at 8:32 pm

   What about:

   $str =~ s/^\s*|\s*$//g;

   I can’t test or benchmark it right now, but I know I’ve used something along these lines before.

   Reply
6. Yanick says:

   March 1, 2010 at 8:49 pm

   What about

   $str =~ s/^\s+|\s+$//g

   ?

   Reply
7. Yanick says:

   March 1, 2010 at 8:50 pm

   What about

   $str =~ s/^\s+|\s+$//g

   ?

   It’s faster, and not too obtuse.

   Reply
8. Dave Rolsky says:

   March 1, 2010 at 9:05 pm

   Yes, it’s slower, but look at the absolute speed. The “slow” version can still execute 440,000 or so per minute!

   Write code as clearly as possible, _then_ optimize the slow parts. This is what profilers are for.

   This sort of premature micro-optimization makes for messy code that may actually be no faster than clean code.

   Also, I think the simplest way to write it is …

   $str =~ s/^\s+|\s+$//g;

   Reply
9. Hercynium says:

   March 1, 2010 at 10:34 pm

   What about:

   $str =~ s/^\s*|\s*$//g

   Seems clearer and will probably be faster.

   Reply
10. mattp says:

    March 2, 2010 at 7:40 am

    its slower due to the backtracking required from the combination of two greedy quantifiers with the one non greedy.
    to quote knuth .. premature optimization is the root of all evil

    Reply
11. dami says:

    March 2, 2010 at 12:58 pm

    Jeffrey Friedl recommends using 2 regexes. But you can do it in one line :

    s/^\s+//, s/\s+$// for $str;

    Reply
12. John Wiersba says:

    March 5, 2010 at 1:49 pm

    I think the hands down winner is $str =~ s/^\s+//; $str =~ s/\s+\z//;

    It says exactly what you want and it’s the fastest. No need to “optimize” with a single regex. Although dami’s postfix-for form is 7 characters shorter, the need for a comma instead of semicolon is a blemish, plus it’s about 25% slower.

    #!/usr/bin/perl -w
    use strict;
    use Benchmark;
    use vars qw($str1);
    $str1 = ” asf asd fasd fasd f asdf asd asdf as fsda fasdf “;
    timethese( -5, {
    test1_0 => ‘ $str = $str1; $str =~ s/^\s+|\s+\z//g ‘,
    test1_1 => ‘ $str = $str1; $str =~ s/^\s*|\s*\z//g ‘,
    test1_2 => ‘ $str = $str1; $str =~ s/^\s+|\s+$//g ‘,
    test1_3 => ‘ $str = $str1; $str =~ s/^\s*|\s*$//g ‘,
    test2_0 => ‘ $str = $str1; $str =~ s/^\s+//; $str =~ s/\s+\z// ‘,
    test2_1 => ‘ $str = $str1; $str =~ s/^\s+//; $str =~ s/\s+$// ‘,
    test2_2 => ‘ $str = $str1; $str =~ s/^\s*//; $str =~ s/\s*\z// ‘,
    test2_3 => ‘ $str = $str1; $str =~ s/^\s*//; $str =~ s/\s*$// ‘,
    test3 => ‘ $str = $str1; s/^\s+//, s/\s+\z// for $str ‘,
    });

    Benchmark: running test1_0, test1_1, test1_2, test1_3, test2_0, test2_1, test2_2, test2_3, test3 for at least 5 CPU seconds…
    test1_0: 6 wallclock secs ( 5.29 usr + 0.00 sys = 5.29 CPU) @ 185981.10/s (n=983840)
    test1_1: 5 wallclock secs ( 5.22 usr + 0.00 sys = 5.22 CPU) @ 154480.84/s (n=806390)
    test1_2: 6 wallclock secs ( 5.22 usr + 0.00 sys = 5.22 CPU) @ 188475.10/s (n=983840)
    test1_3: 5 wallclock secs ( 5.25 usr + 0.00 sys = 5.25 CPU) @ 154564.00/s (n=811461)
    test2_0: 5 wallclock secs ( 5.26 usr + 0.01 sys = 5.27 CPU) @ 672072.87/s (n=3541824)
    test2_1: 6 wallclock secs ( 5.44 usr + 0.00 sys = 5.44 CPU) @ 651070.77/s (n=3541825)
    test2_2: 5 wallclock secs ( 5.17 usr + 0.00 sys = 5.17 CPU) @ 248851.26/s (n=1286561)
    test2_3: 5 wallclock secs ( 5.23 usr + 0.00 sys = 5.23 CPU) @ 241523.71/s (n=1263169)
    test3: 4 wallclock secs ( 5.29 usr + 0.00 sys = 5.29 CPU) @ 526948.02/s (n=2787555)

    Reply
13. Benchmarking Versus Profiling | House Absolute(ly) Pointless says:

    November 22, 2012 at 12:43 pm

    […] noticed a couple blog entries in the Planet Perl Iron Man feed discussing which way of stripping whitespace from […]

    Reply