Re^4: matching the words

[BrowserUk]

Rather than loading the real words as an array, I'd load them as a single whitespace delimited string.

It is a couple of hundred times faster to invoke the regex engine, once, to search for one word in a string containing hundreds or thousands of words; than to invoke it hundreds or thousands of times to match against one word at a time:

#! perl -slw
use strict;
use Benchmark qw[ cmpthese ];

our $words = do{ local( @ARGV, $/ ) = 'words.txt'; <> };
our @words = split ' ', $words;

our $P //= 0.01;
our @toLookFor = map {
    rand() > $P ? ()
        :  do {
            my $w = $_;
            my $p = int( rand length()-1 );
            $w =~ s[.{$p}\K.][.];
            $w;
        };
} @words;

printf "Looking for %d terms amongst %d words\n",
    scalar @toLookFor, scalar @words;

cmpthese 1, {
    a => q[
        for my $re ( @toLookFor ) {
             m[^$re$]
                 #and print "a:$re :: $_"
                 for @words;
         }
    ],
    b => q[
        $words =~ m[\b($_)\b]
            #and print "b:$_ :: $1"
            for @toLookFor;
    ],
}

__END__
C:\test>junk42
Looking for 1846 terms amongst 178691 words
  s/iter     a     b
a   85.2    --  -95%
b   3.94 2065%    --

C:\test>junk42 -P=0.02
Looking for 3564 terms amongst 178691 words
  s/iter     a     b
a    166    --  -95%
b   7.83 2022%    --

cmpthese