Beefy Boxes and Bandwidth Generously Provided by pair Networks
There's more than one way to do things
 
PerlMonks  

regex alternation

by Anonymous Monk
on Sep 30, 2011 at 10:02 UTC ( #928732=perlquestion: print w/replies, xml ) Need Help??
Anonymous Monk has asked for the wisdom of the Perl Monks concerning the following question:

In the code below I want both strings in the test_array to match with the regex $test3.

This one works but I wonder if there is a more efficient way of doing it than the one I came up with? I ask because the regex will be run thousands of times and performance is a factor.

my $test3 = qr<^in directory '/directory/tmp(/\..*'|')$>;

my @test_array = (
    "in directory '/directory/tmp'",
    "in directory '/directory/tmp/.idBAA5KaaYWfb'"
);

for my $line (@test_array) {
    print "TESTING <$line>\n";
    if ($line =~ $test3) {
        print "MATCHED\n";
    }
    else {
        print "NO MATCH\n";
    }
}

[download]

Replies are listed 'Best First'.
Re: regex alternation
by moritz (Cardinal) on Sep 30, 2011 at 10:19 UTC

    You can try something along these lines to avoid invoking the regex engine: 

    my $prefix = "in directory '/directory/tmp";
my $prefix_len = length $prefix;
my $test3 = qr<^$prefix(/\..*'|')$>;

for my $line (@test_array) {
   if (substr($line, 0, $prefix_len) eq $prefix
            && $line =~ $test3) {
        print "MATCHED\n";
   }
  else {
        print "NO MATCH\n";
    }
}

    [download]

    But you need to Benchmark it to find out if it's actually faster with your data.

    Perl 6 - second systems done right
[reply]
[d/l]
Re: regex alternation
by keszler (Priest) on Sep 30, 2011 at 10:24 UTC

    Why capture the closing single-quote? If you don't need it the alternation is unnecessary overhead. qr<^in directory '/directory/tmp(/\..*)?'$> would be better in that case. Also, if you're not planning to do anything with $1 after the match then there's no need to capture at all: qr<^in directory '/directory/tmp(?:/\..*)?'$>

    For overall regex performance it's hard - for me anyway - to determine without knowing what the non-matching data looks like. In general you want the non-matching tests to fail as quickly as possible. (Your anchoring "in directory" to the start of the string is good for that.)

    update - moritz's prefix method is even better

[reply]
[d/l]
[select]
Re: regex alternation
by Anonymous Monk on Sep 30, 2011 at 10:50 UTC

    Hooray, make of this what you will, at your own peril

    And see also: Why does global match run faster than none global?, Multiple Regex evaluations or one big one? 

    #!/usr/bin/perl --
use strict;
use warnings;
use Benchmark qw[ cmpthese ];
print "##########\n";
print "perl $] \n";

my @test_array = (
    "in directory '/directory/tmp'",
    "in directory '/directory/tmp/.idBAA5KaaYWfb'",
    "in directory '/directory/not'",
    ( "in directory '/directory/notit'" ) x 2,
);


cmpthese -3, {
    a_ => sub {
        my $match = 0;
        my $test3 = qr<^in directory '/directory/tmp(/\..*'|')$>;
        for (@test_array) {
            $match++ if $_ =~ $test3;
        }
    },
    a => sub {
        my $match = 0;
        my $test3 = qr<^in directory '/directory/tmp(/\..*'|')$>;
        for my $line (@test_array) {
            $match++ if $line =~ $test3;
        }
    },
    b => sub {
        my $match = 0;
        my $test3 = qr<^in directory '/directory/tmp(?:/\..*'|')$>;
        for my $line (@test_array) {
            $match++ if $line =~ $test3;
        }
    },
    c => sub {
        my $match = 0;
        my $test3 = qr<^in directory '/directory/tmp(?:'|/\..*')$>;
        for my $line (@test_array) {
            $match++ if $line =~ $test3;
        }
    },
    f => sub {
        my $match = 0;
        my $test3 = qr<^in directory '/directory/tmp(?:/\..*)?'$>;
        for my $line (@test_array) {
            $match++ if $line =~ $test3;
        }
    },
    g => sub {
        my $match = 0;
        my $prefix = q<in directory '/directory/tmp>;
        my $prefix_len = length $prefix;
        my $test3 = qr<^$prefix(/\..*'|')$>;
        for my $line (@test_array) {
            $match++ if substr($line, 0, $prefix_len) eq $prefix
                        and $line =~ $test3
        }
    },
    d => sub {
        my $match = 0;
        my $test3 = qr<^in directory '/directory/tmp'$>;
        my $test4 = qr<^in directory '/directory/tmp/\..*'$>;
        for my $line (@test_array) {
            $match++ if $line =~ $test3 or $line =~ $test4;
        }
    },
    e => sub {
        my $match = 0;
        my $test3 = q<in directory '/directory/tmp'>;
        my $test4 = qr<^in directory '/directory/tmp/\..*'$>;
        for my $line (@test_array) {
            $match++ if $line eq $test3 or $line =~ $test4;
        }
    },
};

__END__
##########
perl 5.006001
Benchmark: running a, a_, b, c, d, e, f, g, each for at least 3 CPU se
+conds...
         a:  3 wallclock secs ( 3.03 usr +  0.00 sys =  3.03 CPU) @ 19
+0674.36/s (n=577934)
        a_:  4 wallclock secs ( 3.03 usr +  0.00 sys =  3.03 CPU) @ 19
+0674.36/s (n=577934)
         b:  4 wallclock secs ( 3.22 usr +  0.00 sys =  3.22 CPU) @ 20
+6861.45/s (n=665887)
         c:  3 wallclock secs ( 3.06 usr +  0.00 sys =  3.06 CPU) @ 21
+3192.95/s (n=653010)
         d:  3 wallclock secs ( 3.11 usr +  0.00 sys =  3.11 CPU) @ 15
+6451.74/s (n=486252)
         e:  4 wallclock secs ( 3.25 usr +  0.00 sys =  3.25 CPU) @ 21
+6385.04/s (n=703035)
         f:  4 wallclock secs ( 3.11 usr +  0.00 sys =  3.11 CPU) @ 19
+4816.72/s (n=605880)
         g:  3 wallclock secs ( 3.22 usr +  0.00 sys =  3.22 CPU) @ 14
+8028.89/s (n=476505)
       Rate    g    d   a_    a    f    b    c    e
g  148029/s   --  -5% -22% -22% -24% -28% -31% -32%
d  156452/s   6%   -- -18% -18% -20% -24% -27% -28%
a_ 190674/s  29%  22%   --  -0%  -2%  -8% -11% -12%
a  190674/s  29%  22%   0%   --  -2%  -8% -11% -12%
f  194817/s  32%  25%   2%   2%   --  -6%  -9% -10%
b  206861/s  40%  32%   8%   8%   6%   --  -3%  -4%
c  213193/s  44%  36%  12%  12%   9%   3%   --  -1%
e  216385/s  46%  38%  13%  13%  11%   5%   1%   --


##########
perl 5.014001
      Rate    d   a_    a    f    b    c    g    e
d  49781/s   -- -31% -31% -34% -35% -38% -38% -46%
a_ 71676/s  44%   --  -1%  -4%  -6% -10% -11% -22%
a  72090/s  45%   1%   --  -4%  -6% -10% -10% -21%
f  75019/s  51%   5%   4%   --  -2%  -6%  -6% -18%
b  76633/s  54%   7%   6%   2%   --  -4%  -4% -16%
c  79672/s  60%  11%  11%   6%   4%   --  -1% -13%
g  80135/s  61%  12%  11%   7%   5%   1%   -- -13%
e  91585/s  84%  28%  27%  22%  20%  15%  14%   --

    [download]
[reply]
[d/l]
      Interesting... Is that a debug build of 5.14? Because those rates are WAAAY slower than you got for 5.6.

      Mike
[reply]
[reply]
Re: regex alternation
by jethro (Monsignor) on Sep 30, 2011 at 10:18 UTC
    One obvious improvement would be to use the regex modifier o to tell perl that the regex won't change.
[reply]
[reply]
Re: regex alternation
by AlexTape (Monk) on Sep 30, 2011 at 10:32 UTC
    at first it will be a bit faster if you dont use an alias for $_. perhaps try map to increase the speed.. i´m intrested too in this case.. probably there is a solution via hashing.. u need exactly match||no match prints? or you "only" need the matched lines? <= if so it can be done faster 4 sure..
    for(@test_array){print "MATCHED\n" if $_ =~ $test3 && next;
print "NO MATCH\n";} #untested

    [download]
    futhermore check:
    How do I optimize a regular expression? and
    Perl regexp matching is slow??

    $perlig =~ s/pec/cep/g if 'errors expected';
[reply]
[d/l]

Back to Seekers of Perl Wisdom

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: perlquestion [id://928732]
Approved by Perlbotics
help
Chatterbox?
and all is quiet...

How do I use this? | Other CB clients
Other Users?
Others making s'mores by the fire in the courtyard of the Monastery: (8)
As of 2017-06-27 17:35 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?
    How many monitors do you use while coding?















    Results (611 votes). Check out past polls.