Rewritten highlighting algorithm. IrregularWordsStemmerInterface is r…

…emoved as it is not used anymore.
parpalak · Aug 24, 2024 · c93a642 · c93a642
1 parent 251778d
commit c93a642
Show file tree

Hide file tree

Showing 18 changed files with 434 additions and 363 deletions.
diff --git a/src/S2/Rose/Entity/HighlightIntervals.php b/src/S2/Rose/Entity/HighlightIntervals.php
@@ -0,0 +1,36 @@
+<?php
+/**
+ * @copyright 2024 Roman Parpalak
+ * @license   https://opensource.org/license/mit MIT
+ */
+
+declare(strict_types=1);
+
+namespace S2\Rose\Entity;
+
+class HighlightIntervals
+{
+    protected array $highlightIntervals = [];
+    protected bool $hasPreviousInterval = false;
+
+    public function addInterval(int $start, int $end): void
+    {
+        if (!$this->hasPreviousInterval) {
+            $this->highlightIntervals[] = [$start, $end];
+        } else {
+            $this->highlightIntervals[\count($this->highlightIntervals) - 1][1] = $end;
+        }
+
+        $this->hasPreviousInterval = true;
+    }
+
+    public function skipInterval(): void
+    {
+        $this->hasPreviousInterval = false;
+    }
+
+    public function toArray(): array
+    {
+        return $this->highlightIntervals;
+    }
+}
diff --git a/src/S2/Rose/Entity/ResultItem.php b/src/S2/Rose/Entity/ResultItem.php
@@ -1,16 +1,17 @@
-<?php declare(strict_types=1);
+<?php
 /**
- * @copyright 2016-2023 Roman Parpalak
+ * @copyright 2016-2024 Roman Parpalak
  * @license   MIT
  */
 
+declare(strict_types=1);
+
 namespace S2\Rose\Entity;
 
 use S2\Rose\Entity\Metadata\ImgCollection;
 use S2\Rose\Entity\Metadata\SnippetSource;
 use S2\Rose\Exception\InvalidArgumentException;
 use S2\Rose\Exception\RuntimeException;
-use S2\Rose\Snippet\WordsByStemsExtractor;
 use S2\Rose\Stemmer\StemmerInterface;
 
 class ResultItem
@@ -113,7 +114,7 @@ public function getSnippet(): string
             return $this->description;
         }
 
-        $snippet = $this->snippet->toString(0.3);
+        $snippet = $this->snippet->toString();
         if ($snippet) {
             return $snippet;
         }
@@ -127,7 +128,7 @@ public function getFormattedSnippet(): string
             return $this->description;
         }
 
-        $snippet = $this->snippet->toString(0.3, true);
+        $snippet = $this->snippet->toString(true);
         if ($snippet) {
             return $snippet;
         }
@@ -137,8 +138,6 @@ public function getFormattedSnippet(): string
 
     /**
      * @param string[] $words
-     *
-     * @return $this
      */
     public function setFoundWords(array $words): self
     {
@@ -148,14 +147,7 @@ public function setFoundWords(array $words): self
     }
 
     /**
-     * TODO Refactor the highlight logic to a separate class.
-     *
-     * @param StemmerInterface $stemmer
-     *
-     * @return string
-     *
      * @throws RuntimeException
-     * @see \S2\Rose\Snippet\SnippetBuilder::buildSnippet for dublicated logic
      */
     public function getHighlightedTitle(StemmerInterface $stemmer): string
     {
@@ -165,16 +157,12 @@ public function getHighlightedTitle(StemmerInterface $stemmer): string
             throw new InvalidArgumentException('Highlight template must contain "%s" substring for sprintf() function.');
         }
 
-        $extractor = new WordsByStemsExtractor($stemmer, $this->foundWords);
-
-        [$foundWords,] = $extractor->extract($this->title);
-
-
         $snippetLine = new SnippetLine(
             $this->title,
             SnippetSource::FORMAT_PLAIN_TEXT,
-            array_keys($foundWords),
-            \count($foundWords)
+            $stemmer,
+            $this->foundWords,
+            0
         );
 
         return $snippetLine->getHighlighted($template, false);

diff --git a/src/S2/Rose/Entity/Snippet.php b/src/S2/Rose/Entity/Snippet.php
@@ -1,13 +1,12 @@
-<?php declare(strict_types=1);
+<?php
 /**
- * @copyright 2016-2023 Roman Parpalak
+ * @copyright 2016-2024 Roman Parpalak
  * @license   MIT
  */
 
-namespace S2\Rose\Entity;
+declare(strict_types=1);
 
-use S2\Rose\Entity\Metadata\SnippetSource;
-use S2\Rose\Helper\StringHelper;
+namespace S2\Rose\Entity;
 
 class Snippet
 {
@@ -19,23 +18,19 @@ class Snippet
      * @var SnippetLine[]
      */
     protected array $snippetLines = [];
-    protected array $snippetLineWeights = [];
 
     /**
      * @var SnippetLine[]
      */
     protected array $introductionSnippetLines;
     protected string $textIntroduction = '';
 
-    protected int $foundWordCount = 0;
-
     protected string $highlightTemplate;
-    private array $snippetMinWordPositions = [];
-    private array $snippetMaxWordPositions = [];
+    protected array $snippetMinWordPositions = [];
+    protected array $snippetMaxWordPositions = [];
 
-    public function __construct(int $foundWordNum, string $highlightTemplate, SnippetLine ...$introductionSnippetLines)
+    public function __construct(string $highlightTemplate, SnippetLine ...$introductionSnippetLines)
     {
-        $this->foundWordCount           = $foundWordNum;
         $this->highlightTemplate        = $highlightTemplate;
         $this->introductionSnippetLines = $introductionSnippetLines;
     }
@@ -50,7 +45,6 @@ public function setLineSeparator(string $lineSeparator): self
     public function attachSnippetLine(int $minWordPosition, int $maxWordPosition, SnippetLine $snippetLine): self
     {
         $this->snippetLines[]            = $snippetLine;
-        $this->snippetLineWeights[]      = $snippetLine->getRelevance();
         $this->snippetMinWordPositions[] = $minWordPosition;
         $this->snippetMaxWordPositions[] = $maxWordPosition;
 
@@ -67,7 +61,7 @@ public function getTextIntroduction(bool $includeFormatting = false): string
         return implode(' ', $result);
     }
 
-    public function toString(float $acceptableRelevance = 0.6, bool $includeFormatting = false): ?string
+    public function toString(bool $includeFormatting = false): ?string
     {
         $stat = [];
         foreach ($this->snippetLines as $index => $snippetLine) {
@@ -99,17 +93,11 @@ public function toString(float $acceptableRelevance = 0.6, bool $includeFormatti
             $resultSnippetLines[$idx] = $this->snippetLines[$idx];
         }
 
-        if ($this->calcLinesRelevance($resultSnippetLines) < $acceptableRelevance) {
-            return null;
-        }
-
         return $this->implodeLines($resultSnippetLines, $includeFormatting);
     }
 
     /**
      * @param array|SnippetLine[] $snippetLines
-     *
-     * @return string
      */
     private function implodeLines(array $snippetLines, bool $includeFormatting): string
     {
@@ -149,25 +137,4 @@ private function implodeLines(array $snippetLines, bool $includeFormatting): str
 
         return $result;
     }
-
-    /**
-     * @param array|SnippetLine[] $snippetLines
-     *
-     * @return float|int
-     */
-    private function calcLinesRelevance(array $snippetLines)
-    {
-        if (!($this->foundWordCount > 0)) {
-            return 0;
-        }
-
-        $foundWords = [];
-        foreach ($snippetLines as $snippetLine) {
-            foreach ($snippetLine->getFoundWords() as $word) {
-                $foundWords[$word] = 1;
-            }
-        }
-
-        return \count($foundWords) * 1.0 / $this->foundWordCount;
-    }
 }