Updated search query cleaning algorithm to be closer to indexing clea…

…ning algorithm.
parpalak · Aug 21, 2024 · 251778d · 251778d
1 parent 468d5e7
commit 251778d
Show file tree

Hide file tree

Showing 4 changed files with 76 additions and 34 deletions.
diff --git a/src/S2/Rose/Entity/Metadata/SentenceCollection.php b/src/S2/Rose/Entity/Metadata/SentenceCollection.php
@@ -101,10 +101,10 @@ private function buildWordsInfo(): void
      */
     public static function breakIntoWords(string $content): array
     {
-        // Replace comma as decimal separator to dot
-        $content = preg_replace('/[\s()]\d+\K,(?=\d+(?:[\s()]|\.\s))/', '.', $content);
+        // Replace decimal separator: ',' -> '.'
+        $content = preg_replace('#(?:^|[\s()])-?\d+\K,(?=\d+(?:$|[\s()]|\.\s))#', '.', $content);
 
-        // We allow letters, digits and some punctuation: ".,-"
+        // We allow letters, digits and some punctuation: ".,-^_"
         $content = str_replace(',', ', ', $content);
         $content = preg_replace('#[^\\-.,0-9\\p{L}^_]+#u', ' ', $content);
         $content = mb_strtolower($content);

diff --git a/src/S2/Rose/Entity/Query.php b/src/S2/Rose/Entity/Query.php
@@ -115,46 +115,44 @@ public function valueToArray()
 
         // Normalize
         $content = str_replace(['«', '»', '“', '”', '‘', '’'], '"', $content);
-        $content = str_replace(['---', '--', '–', '−'], '—', $content);
+        $content = str_replace('−', '-', $content); // Replace minus sign to a hyphen
+        $content = str_replace(['---', '–', '−'], '—', $content); // Normalize dashes
         $content = preg_replace('#,\\s+,#u', ',,', $content);
         $content = preg_replace('#[^\\-\\p{L}0-9^_.,()";?!…:—]+#iu', ' ', $content);
-        $content = preg_replace('#\\n+#', ' ', $content);
-        $content = preg_replace('#\\s+#u', ' ', $content);
         $content = mb_strtolower($content);
 
-        $content = preg_replace('#(,+)#u', '\\1 ', $content);
+        // Replace decimal separators: ',' -> '.'
+        $content = preg_replace('#(?<=^|\\s)(\\-?\\d+),(\\d+)(?=\\s|$)#u', '\\1.\\2', $content);
 
-        $content = preg_replace('#[ |\\/]+#', ' ', $content);
-
-        $words = explode(' ', $content);
-        foreach ($words as $k => $v) {
-            // Separate special chars from the letter combination
-            if (strlen($v) > 1) {
-                foreach (['—', '^', '(', ')', '"', ':', '?', '!'] as $specialChar) {
-                    if (mb_substr($v, 0, 1) == $specialChar || mb_substr($v, -1) == $specialChar) {
-                        $words[$k] = str_replace($specialChar, '', $v);
-                        $words[]   = $specialChar;
-                    }
-                }
+        // Separate special chars at the beginning of the word
+        while (true) {
+            $content = preg_replace('#(?:^|\\s)\K([—^()"?:!])(?=[^\s])#u', '\\1 ', $content, -1, $count);
+            if ($count === 0) {
+                break;
             }
+        }
 
-            // Separate hyphen from the letter combination
-            if (strlen($v) > 1 && (substr($v, 0, 1) == '-' || substr($v, -1) == '-')) {
-                $words[$k] = str_replace('-', '', $v);
-                $words[]   = '-';
+        // Separate special chars at the end of the word
+        while (true) {
+            $content = preg_replace('#(?<=[^\s])([—^()"?:!])(?=\\s|$)#u', ' \\1', $content, -1, $count);
+            if ($count === 0) {
+                break;
             }
+        }
 
-            // Replace 'ё' inside words
-            if (false !== strpos($v, 'ё') && $v != 'ё') {
-                $words[$k] = str_replace('ё', 'е', $v);
-            }
+        // Separate groups of commas
+        $content = preg_replace('#(,+)#u', ' \\1 ', $content);
 
-            // Remove ','
-            if (preg_match('#^[^,]+,$#u', $v) || preg_match('#^,[^,]+$#u', $v)) {
-                $words[$k] = str_replace(',', '', $v);
-                $words[]   = ',';
+        $words = preg_split('#\\s+#', $content);
+        foreach ($words as $k => &$v) {
+            // Replace 'ё' inside words
+            if ($v !== 'ё' && false !== strpos($v, 'ё')) {
+                $v = str_replace('ё', 'е', $v);
             }
         }
+        unset($v);
+
+        $words = array_unique($words);
 
         StringHelper::removeLongWords($words);
 

diff --git a/src/S2/Rose/Snippet/WordsByStemsExtractor.php b/src/S2/Rose/Snippet/WordsByStemsExtractor.php
@@ -1,8 +1,8 @@
 <?php
 /**
  * @copyright 2024 Roman Parpalak
- * @license MIT
- * @package Rose
+ * @license   MIT
+ * @package   Rose
  */
 
 declare(strict_types=1);
@@ -85,6 +85,7 @@ public function extract(string $text): array
      * check each fragment for a match with the searched stem.
      *
      * @param string $text
+     *
      * @return string[]
      */
     private function getWords(string $text): array
@@ -93,6 +94,6 @@ private function getWords(string $text): array
             return [$text];
         }
 
-        return array_merge(explode('-', $text), [$text]);
+        return array_merge(array_filter(explode('-', $text), static fn(string $word) => $word !== ''), [$text]);
     }
 }
diff --git a/tests/unit/Rose/Entity/QueryTest.php b/tests/unit/Rose/Entity/QueryTest.php
@@ -11,17 +11,60 @@
 
 /**
  * @group entity
+ * @group query
  */
 class QueryTest extends Unit
 {
     public function testFilterInput(): void
     {
+        // Tests for splitting strings by special delimiters
         $this->assertEquals([1, 2], (new Query('1|||2'))->valueToArray());
         $this->assertEquals([1, 2], (new Query('1\\\\\\2'))->valueToArray());
         $this->assertEquals(['a', 'b'], (new Query('a/b'))->valueToArray());
         $this->assertEquals(['a', 'b'], (new Query(' a   b   '))->valueToArray());
         $this->assertEquals(['..'], (new Query('..'))->valueToArray());
         $this->assertEquals(['...'], (new Query('...'))->valueToArray());
         $this->assertEquals(['a..b'], (new Query('a..b'))->valueToArray());
+
+        // Tests for replacing numbers
+        $this->assertEquals(['1.2'], (new Query('1,2'))->valueToArray());
+        // $this->assertEquals(['-1.2'], (new Query('-1,2'))->valueToArray());
+        $this->assertEquals(['1.2'], (new Query('1.2'))->valueToArray());
+
+        // Tests for replacing typographic quotes
+        $this->assertEquals(['"', 'text'], (new Query('«text»'))->valueToArray());
+        $this->assertEquals(['"', 'text'], (new Query('“text”'))->valueToArray());
+
+        // Tests for replacing dashes
+        $this->assertEquals(['a--b'], (new Query('a--b'))->valueToArray());
+        $this->assertEquals(['a—b'], (new Query('a---b'))->valueToArray()); // --- to mdash
+        $this->assertEquals(['a—b'], (new Query('a–b'))->valueToArray()); // ndash to mdash
+        $this->assertEquals(['a-b'], (new Query('a−b'))->valueToArray()); // Minus to hyphen
+
+        // Test for replacing line breaks and extra spaces
+        $this->assertEquals(['a', 'b'], (new Query("a\n\nb"))->valueToArray());
+        $this->assertEquals(['a', 'b'], (new Query("a \t   b"))->valueToArray());
+
+        // Tests for separating special characters
+        $this->assertEquals(['a!b'], (new Query('a!b'))->valueToArray());
+        $this->assertEquals(['!', 'ab'], (new Query('!ab'))->valueToArray());
+        $this->assertEquals(['!', 'a!b'], (new Query('!a!b'))->valueToArray());
+        $this->assertEquals(['(', 'word', ')'], (new Query('(word)'))->valueToArray());
+        $this->assertEquals(['mysql', '--all-databases'], (new Query('mysql --all-databases'))->valueToArray());
+
+        // Test for replacing "ё" with "е"
+        $this->assertEquals(['ё', 'полет', 'field'], (new Query('ё полёт field'))->valueToArray());
+
+        // Tests for handling commas
+        $this->assertEquals(['a', ',', 'b'], (new Query('a,b'))->valueToArray());
+        $this->assertEquals(['a', ',,', 'b'], (new Query('a,,b'))->valueToArray());
+        $this->assertEquals(['a', ',,,', 'b'], (new Query('a,,,b'))->valueToArray());
+
+        // Tests for removing long words
+        $this->assertEquals(['a', 'c'], (new Query('a ' . str_repeat('b', 101) . ' c'))->valueToArray());
+
+        // Tests for compatibility of multiple rules
+        $this->assertEquals(['a—b', '"', 'text'], (new Query('a–b «text»'))->valueToArray());
+        $this->assertEquals(['a', ',', 'b'], (new Query(" a, \n   b "))->valueToArray());
     }
 }