Skip to content

Instantly share code, notes, and snippets.

@rohjay
Created May 11, 2023 15:37
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save rohjay/248ab08f219f390528805f4fa76899bc to your computer and use it in GitHub Desktop.
Save rohjay/248ab08f219f390528805f4fa76899bc to your computer and use it in GitHub Desktop.
PHP function to compare strings using the Ratcliff/Obershelp pattern-matching algorithm. Great for checking sameness of words / phrases.
<?php
// I PHP-ified Ben's work from here: https://github.com/ben-yocum/gestalt-pattern-matcher (thanks for sharing Ben!)
// made a few modifications and added comments to better understand how this works. Enjoy =]
function compare_strings(string $input1, string $input2, bool $case_insensitive = true): float {
if ( $input1 == '' || $input2 == '' ) {
return 0.0;
}
// Case insensitive is the preferred, but you do you
if ( $case_insensitive ) {
$input1 = strtolower($input1);
$input2 = strtolower($input2);
}
// Build the foundation of the stack with our comparison strings...
$stack = [
$input1,
$input2
];
// Set the "score" at zero.
$score = 0;
// I added this in here due to a lack of understanding how this worked. Keeping it 😜
$prevention_count = 1000;
while( count($stack) > 0 && $prevention_count > 0 ) {
// Pop the comparison strings off the top...
$string2 = array_pop($stack);
$string1 = array_pop($stack);
// Use a split array of the strings so I don't have to substr with offsets all over the place.
$string1_split = str_split($string1);
$string2_split = str_split($string2);
// Grab the string length once
$str1_len = strlen($string1);
$str2_len = strlen($string2);
// Set up the indicators for sameness
$sequence_length = 0;
$sequence_index1 = -1;
$sequence_index2 = -1;
// Loop the length of string 1
for($i = 0; $i < $str1_len; $i++) {
// Loop the length of string 2
for($j = 0; $j < $str2_len; $j++) {
// Our wildcard to track from our current position out to wherever the strings stop matching
$k = 0;
while(isset($string1_split[$i+$k]) && isset($string2_split[$j+$k]) && $string1_split[$i+$k] === $string2_split[$j+$k]) {
++$k;
}
// Now that we've seen how far out $k gets us... Let's see if this is streak-worthy ;)
if($k > $sequence_length) {
$sequence_length = $k; // Our wildcard that tracks how far out a same-streak goes
$sequence_index1 = $i; // Where the streak starts in string 1
$sequence_index2 = $j; // Where the streak starts in string 2
}
}
}
// We've looped through both strings independently and found ONE similarity, let's tally the score and add more
// things to look at around where this was.
if($sequence_length !== 0) {
// Tally the score based off the sequence length (remember our wildcard $k from before? This is the biggest
// we saw of that)
$score += $sequence_length * 2;
// If our matching sequence didn't start at the beginning of either string, there could be matches here yet.
if( $sequence_index1 !== 0 && $sequence_index2 !== 0 ) {
$stack[] = substr($string1, 0, $sequence_index1);
$stack[] = substr($string2, 0, $sequence_index2);
}
// If our matching sequence didn't butt up to the end of either string, there are still potential matches.
if(
$sequence_index1 + $sequence_length !== $str1_len
&& $sequence_index2 + $sequence_length !== $str2_len
) {
$stack[] = substr($string1, $sequence_index1 + $sequence_length, $str1_len);
$stack[] = substr($string2, $sequence_index2 + $sequence_length, $str2_len);
}
}
// Because I never trust this will stop. Never.
--$prevention_count;
}
// This gives us a relative score normalized by the length of the two strings.
$strlengths = strlen($input1) + strlen($input2);
$rating = $score / ($strlengths != 0 ? $strlengths : 1); // Habit. Never divide a goose egg.
// I multiply by 100 here and round to the 5th decimal to give me a Percentage figure. For instance:
// 12.34567% match is not a great match. 98.76543% match is pretty good! An exact match is 100.
return round($rating * 100, 5);
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment