Fix an issue truncating extracted metadata content

Previous regex forced the selection to stop at either the first single or double quote found, regardless of the opening quote. Using '\1', we're sure to wait for the proper quote before stopping the capture.

diff --git a/application/bookmark/LinkUtils.php b/application/bookmark/LinkUtils.php
index 17c379796cd39b239be09459d78e3cad22fdfb0b..a74fda5795c09e9f0fd6b3ed1e8d98b81226bc5d 100644 (file)
--- a/application/bookmark/LinkUtils.php
+++ b/application/bookmark/LinkUtils.php
@@ -68,16 +68,16 @@ function html_extract_tag($tag, $html)
     $properties = implode('|', $propertiesKey);
     // We need a OR here to accept either 'property=og:noquote' or 'property="og:unrelated og:my-tag"'
     $orCondition  = '["\']?(?:og:)?'. $tag .'["\']?|["\'][^\'"]*?(?:og:)?' . $tag . '[^\'"]*?[\'"]';
-    // Try to retrieve OpenGraph image.
-    $ogRegex = '#<meta[^>]+(?:'. $properties .')=(?:'. $orCondition .')[^>]*content=["\'](.*?)["\'].*?>#';
+    // Try to retrieve OpenGraph tag.
+    $ogRegex = '#<meta[^>]+(?:'. $properties .')=(?:'. $orCondition .')[^>]*content=(["\'])([^\1]*?)\1.*?>#';
     // If the attributes are not in the order property => content (e.g. Github)
     // New regex to keep this readable... more or less.
-    $ogRegexReverse = '#<meta[^>]+content=["\'](.*?)["\'][^>]+(?:'. $properties .')=(?:'. $orCondition .').*?>#';
+    $ogRegexReverse = '#<meta[^>]+content=(["\'])([^\1]*?)\1[^>]+(?:'. $properties .')=(?:'. $orCondition .').*?>#';
 
     if (preg_match($ogRegex, $html, $matches) > 0
         || preg_match($ogRegexReverse, $html, $matches) > 0
     ) {
-        return $matches[1];
+        return $matches[2];
     }
 
     return false;

diff --git a/tests/bookmark/LinkUtilsTest.php b/tests/bookmark/LinkUtilsTest.php
index 3321242fae07f018c91b98b2f40067fa2d6a9e22..9bddf84b3f02939ec9afb479948eaee368d4a1ff 100644 (file)
--- a/tests/bookmark/LinkUtilsTest.php
+++ b/tests/bookmark/LinkUtilsTest.php
@@ -168,6 +168,36 @@ class LinkUtilsTest extends TestCase
         $this->assertEquals($description, html_extract_tag('description', $html));
     }
 
+    /**
+     * Test html_extract_tag() with double quoted content containing single quote, and the opposite.
+     */
+    public function testHtmlExtractExistentNameTagWithMixedQuotes(): void
+    {
+        $description = 'Bob and Alice share M&M\'s.';
+
+        $html = '<meta property="og:description" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        $html = '<meta tag1="content1" property="og:unrelated1 og:description og:unrelated2" '.
+            'tag2="content2" content="' . $description . '" tag3="content3">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        $html = '<meta property="og:description" name="description" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        $description = 'Bob and Alice share "cookies".';
+
+        $html = '<meta property="og:description" content=\'' . $description . '\'>';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        $html = '<meta tag1="content1" property="og:unrelated1 og:description og:unrelated2" '.
+            'tag2="content2" content=\'' . $description . '\' tag3="content3">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        $html = '<meta property="og:description" name="description" content=\'' . $description . '\'>';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+    }
+
     /**
      * Test html_extract_tag() when the tag <meta name= is not found.
      */