in reply to Pattern searching allowing for mis-matches...

My favorite trick for this sort of fuzzy matching is to combine bit wise xor and the transliteration operator (used in this context for counting). Consider:

use strict;
use warnings;

my $target = 'TGATTGAATCAAGGTGTTTT';
my $match = 'TGAT';
my $quality = 0.75;
my $matchLen = length $match;
my $matchNum = int ($matchLen * $quality);

for my $offset (0 .. length ($target) - $matchLen) {
    my $test = substr $target, $offset, $matchLen;
    my $matched = ($test ^ $match) =~ tr/\x00//;

    next if $matched < $matchNum;
    print "Found <$test> at offset $offset which matches in $matched p
+laces\n";
}

[download]

Prints:

Found <TGAT> at offset 0 which matches in 4 places
Found <TGAA> at offset 4 which matches in 3 places
Found <TGTT> at offset 14 which matches in 3 places

[download]
True laziness is hard work

Replies are listed 'Best First'.
Re^2: Pattern searching allowing for mis-matches...
by MaroonBalloon (Acolyte) on Dec 15, 2009 at 07:53 UTC
    The only problem with that is the threshold is not exact, for example, if the threshold is instead .5, and the length of the query is three characters, I still come up with all the matches that have at least (1/3). I have tried fixing this to no avail, here is my attempt: 
    foreach my $key (keys %o) {
my $keydisplay=1;
my $target = $o{$key};
#my $threshold = 0.79;
my $searchlength = length $search;
my $truethreshold = int($searchlength * $threshold*100) /100; #(page46
+)

    if ($searchlength - $truethreshold >= .005) {
$truethreshold += .01;
}
$searchlength = $truethreshold;


for my $position (0 .. (length ($target) - $searchlength)) {
    my $test = substr $target, $position, $searchlength;
    my $matched = ($test ^ $search) =~ tr/\x00//;

next if $matched < $truethreshold;
    print "\t\tFound <$test> at position $position which matches in $m
+atched of $searchlength places\n";
  }
}

    [download]
[reply]
[d/l]

      Characters are discrete entities. What outcome do you expect from matching 1/2 a character?

      True laziness is hard work
[reply]
        Right, maybe that was a poor example..... A threshold for .6 on a three character query still searches for 1/3, 2/3 or 3/3 identity. Similarly, a threshold of .7 on a 5 character search yields 3/5, 4/5 and 5/5 match. I would say I prefer to round it UP, rather than down, as by definition somethin shouldn't be below a threshold. I am just looking for a way to tighten it up.
[reply]

In Section Seekers of Perl Wisdom