fuzzy match implementation

phanan-forks · Jul 7, 2017 · fdd9189 · fdd9189
1 parent 1687c93
commit fdd9189
Show file tree

Hide file tree

Showing 4 changed files with 1,775 additions and 12 deletions.
diff --git a/helper/helpers.php b/helper/helpers.php
@@ -9,18 +9,17 @@ function stringEndsWith($haystack, $needle)
 }
 
 if (!function_exists('fuzzyMatch')) {
-    function fuzzyMatch($pattern, $str)
+    function fuzzyMatch($pattern, $items)
     {
-        $j             = 0;
-        $patternLength = strlen($pattern);
-        $strLength     = strlen($str);
-
-        for ($i = 0; $i < $strLength && $j < $patternLength; $i++) {
-            if ($pattern[$j] == $str[$i]) {
-                $j++;
-            }
-        }
+        $fm = new TeamTNT\TNTSearch\TNTFuzzyMatch;
+        return $fm->fuzzyMatch($pattern, $items);
+    }
+}
 
-        return ($j == $patternLength);
+if (!function_exists('fuzzyMatchFromFile')) {
+    function fuzzyMatchFromFile($pattern, $path)
+    {
+        $fm = new TeamTNT\TNTSearch\TNTFuzzyMatch;
+        return $fm->fuzzyMatchFromFile($pattern, $path);
     }
-}
+}
diff --git a/src/TNTFuzzyMatch.php b/src/TNTFuzzyMatch.php
@@ -0,0 +1,144 @@
+<?php
+
+namespace TeamTNT\TNTSearch;
+
+class TNTFuzzyMatch
+{
+    public function norm($vec)
+    {
+        $norm       = 0;
+        $components = count($vec);
+
+        for ($i = 0; $i < $components; $i++) {
+            $norm += $vec[$i] * $vec[$i];
+        }
+
+        return sqrt($norm);
+    }
+
+    public function dot($vec1, $vec2)
+    {
+        $prod       = 0;
+        $components = count($vec1);
+
+        for ($i = 0; $i < $components; $i++) {
+            $prod += ($vec1[$i] * $vec2[$i]);
+        }
+
+        return $prod;
+    }
+
+    public function wordToVector($word)
+    {
+        $alphabet = "aAbBcCčČćĆdDđĐeEfFgGhHiIjJkKlLmMnNoOpPqQrRsSšŠtTvVuUwWxXyYzZžŽ1234567890'+ /";
+
+        $result = [];
+        foreach (str_split($word) as $w) {
+            $result[] = strpos($alphabet, $w) + 1000000;
+        }
+        return $result;
+    }
+
+    public function angleBetweenVectors($a, $b)
+    {
+        $denominator = ($this->norm($a) * $this->norm($b));
+
+        if ($denominator == 0) {
+            return 0;
+        }
+
+        return $this->dot($a, $b) / $denominator;
+    }
+
+    public function hasCommonSubsequence($pattern, $str)
+    {
+        $j             = 0;
+        $patternLength = strlen($pattern);
+        $strLength     = strlen($str);
+
+        for ($i = 0; $i < $strLength && $j < $patternLength; $i++) {
+            if ($pattern[$j] == $str[$i]) {
+                $j++;
+            }
+        }
+
+        return ($j == $patternLength);
+    }
+
+    public function makeVectorSameLength($str, $pattern)
+    {
+        $j   = 0;
+        $max = max(count($pattern), count($str));
+        $a   = [];
+        $b   = [];
+
+        for ($i = 0; $i < $max && $j < $max; $i++) {
+            if (isset($pattern[$j]) && isset($str[$i]) && $pattern[$j] == $str[$i]) {
+                $j++;
+                $b[] = $str[$i];
+            } else {
+                $b[] = 0;
+            }
+        }
+
+        return $b;
+    }
+
+    public function fuzzyMatchFromFile($pattern, $path)
+    {
+        $res   = [];
+        $lines = fopen($path, "r");
+        if ($lines) {
+            while (!feof($lines)) {
+                $line = fgets($lines, 4096);
+                if ($this->hasCommonSubsequence($pattern, $line)) {
+                    $res[] = $line;
+                }
+            }
+            fclose($lines);
+        }
+
+        $paternVector = $this->wordToVector($pattern);
+
+        $sorted = [];
+        foreach ($res as $word) {
+            $word                   = trim($word);
+            $wordVector             = $this->wordToVector($word);
+            $normalizedPaternVector = $this->makeVectorSameLength($wordVector, $paternVector);
+
+            $angle         = $this->angleBetweenVectors($wordVector, $normalizedPaternVector);
+            $sorted[$word] = $angle;
+        }
+
+        arsort($sorted);
+
+        return $sorted;
+    }
+
+    public function fuzzyMatch($pattern, $items)
+    {
+        $res = [];
+
+        foreach ($items as $item) {
+            if ($this->hasCommonSubsequence($pattern, $item)) {
+                $res[] = $item;
+            }
+        }
+
+        $paternVector = $this->wordToVector($pattern);
+
+        $sorted = [];
+        foreach ($res as $word) {
+            $word                   = trim($word);
+            $wordVector             = $this->wordToVector($word);
+            $normalizedPaternVector = $this->makeVectorSameLength($wordVector, $paternVector);
+
+            $angle         = $this->angleBetweenVectors($wordVector, $normalizedPaternVector);
+            $sorted[$word] = $angle;
+        }
+
+        arsort($sorted);
+
+        return $sorted;
+    }
+}
diff --git a/tests/TNTFuzzyMatchTest.php b/tests/TNTFuzzyMatchTest.php
@@ -0,0 +1,110 @@
+<?php
+
+use TeamTNT\TNTSearch\TNTFuzzyMatch;
+
+class TNTGeoSearchTest extends PHPUnit_Framework_TestCase
+{
+    public function __construct()
+    {
+        $this->fm = new TNTFuzzyMatch;
+    }
+
+    public function testNorm()
+    {
+        $vector     = [3, 4];
+        $normalized = $this->fm->norm($vector);
+        $this->assertEquals(5, $normalized);
+
+        $vector     = [1, 2, 3, 4, 5];
+        $normalized = $this->fm->norm($vector);
+        $this->assertEquals(7.416198487095663, $normalized);
+    }
+
+    public function testDot()
+    {
+        $vector1 = [1, 2, -5];
+        $vector2 = [4, 8, 1];
+
+        $product = $this->fm->dot($vector1, $vector2);
+
+        $this->assertEquals(15, $product);
+    }
+
+    public function testWordToVector()
+    {
+        $word   = "TNT";
+        $vector = $this->fm->wordToVector($word);
+        $this->assertEquals($vector, [1000055, 1000039, 1000055]);
+    }
+
+    public function testAngleBetweenVectors()
+    {
+        $vector1 = [1, 2, 3];
+        $vector2 = [4, 5, 6];
+
+        $angle = $this->fm->angleBetweenVectors($vector1, $vector2);
+
+        $this->assertEquals(0.97463184619707621, $angle);
+    }
+
+    public function testHasCommonSubsequence()
+    {
+        $pattern1 = "tnsarh";
+        $pattern2 = "ntnsearch";
+
+        $res1 = $this->fm->hasCommonSubsequence($pattern1, 'tntsearch');
+        $res2 = $this->fm->hasCommonSubsequence($pattern2, 'tntsearch');
+
+        $this->assertEquals($res1, true);
+        $this->assertEquals($res2, false);
+    }
+
+    public function testMakeVectorSameLength()
+    {
+        $wordVector    = $this->fm->wordToVector("tntsearch");
+        $patternVector = $this->fm->wordToVector("tnth");
+
+        $res = $this->fm->makeVectorSameLength($wordVector, $patternVector);
+        $this->assertEquals([1000054, 1000038, 1000054, 0, 0, 0, 0, 0, 1000026], $res);
+    }
+
+    public function testFuzzyMatchFromFile()
+    {
+        $res = $this->fm->fuzzyMatchFromFile('search', __DIR__.'/_files/english_wordlist_2k.txt');
+
+        $this->assertEquals([
+            'search'   => 1,
+            'research' => 0.86602345529065
+        ], $res);
+    }
+
+    public function testFuzzyMatchFromFileFunction()
+    {
+        $res = fuzzyMatchFromFile('search', __DIR__.'/_files/english_wordlist_2k.txt');
+
+        $this->assertEquals([
+            'search'   => 1,
+            'research' => 0.86602345529065
+        ], $res);
+    }
+
+    public function testFuzzyMatch()
+    {
+        $res = $this->fm->fuzzyMatch('search', ['search', 'research', 'something']);
+
+        $this->assertEquals([
+            'search'   => 1,
+            'research' => 0.86602345529065
+        ], $res);
+    }
+
+    public function testFuzzyMatchFunction()
+    {
+        $res = fuzzyMatch('search', ['search', 'research', 'something']);
+
+        $this->assertEquals([
+            'search'   => 1,
+            'research' => 0.86602345529065
+        ], $res);
+    }
+}