Merge pull request #1540 from ArthurHoaro/fix/metadata-regexes

Improve regex to extract HTML metadata (title, description, etc.)
shaarli · Oct 13, 2020 · 458b6b9 · 458b6b9
2 parents 543b16b + 2cd0509
commit 458b6b9
Show file tree

Hide file tree

Showing 2 changed files with 93 additions and 2 deletions.
diff --git a/application/bookmark/LinkUtils.php b/application/bookmark/LinkUtils.php
@@ -66,11 +66,13 @@ function html_extract_tag($tag, $html)
 {
     $propertiesKey = ['property', 'name', 'itemprop'];
     $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 .')=["\']?(?:og:)?'. $tag .'["\'\s][^>]*content=["\']?(.*?)["\'/>]#';
+    $ogRegex = '#<meta[^>]+(?:'. $properties .')=(?:'. $orCondition .')[^>]*content=["\'](.*?)["\'].*?>#';
     // 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 .')=["\']?(?:og)?:'. $tag .'["\'\s/>]#';
+    $ogRegexReverse = '#<meta[^>]+content=["\'](.*?)["\'][^>]+(?:'. $properties .')=(?:'. $orCondition .').*?>#';
 
     if (preg_match($ogRegex, $html, $matches) > 0
         || preg_match($ogRegexReverse, $html, $matches) > 0

diff --git a/tests/bookmark/LinkUtilsTest.php b/tests/bookmark/LinkUtilsTest.php
@@ -94,8 +94,78 @@ public function testHtmlExtractNonExistentCharset()
     public function testHtmlExtractExistentNameTag()
     {
         $description = 'Bob and Alice share cookies.';
+
+        // Simple one line
         $html = '<html><meta>stuff2</meta><meta name="description" content="' . $description . '"/></html>';
         $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // Simple OpenGraph
+        $html = '<meta property="og:description" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // Simple reversed OpenGraph
+        $html = '<meta content="' . $description . '" property="og:description">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // ItemProp OpenGraph
+        $html = '<meta itemprop="og:description" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph without quotes
+        $html = '<meta property=og:description content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph reversed without quotes
+        $html = '<meta content="' . $description . '" property=og:description>';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph with noise
+        $html = '<meta tag1="content1" property="og:description" tag2="content2" content="' .
+            $description . '" tag3="content3">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph reversed with noise
+        $html = '<meta tag1="content1" content="' . $description . '" ' .
+            'tag3="content3" tag2="content2" property="og:description">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph multiple properties start
+        $html = '<meta property="unrelated og:description" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph multiple properties end
+        $html = '<meta property="og:description unrelated" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph multiple properties both end
+        $html = '<meta property="og:unrelated1 og:description og:unrelated2" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph multiple properties both end with noise
+        $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));
+
+        // OpenGraph reversed multiple properties start
+        $html = '<meta content="' . $description . '" property="unrelated og:description">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph reversed multiple properties end
+        $html = '<meta content="' . $description . '" property="og:description unrelated">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph reversed multiple properties both end
+        $html = '<meta content="' . $description . '" property="og:unrelated1 og:description og:unrelated2">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // OpenGraph reversed multiple properties both end with noise
+        $html = '<meta tag1="content1" content="' . $description . '" tag2="content2" '.
+            'property="og:unrelated1 og:description og:unrelated2" tag3="content3">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
+
+        // Suggestion from #1375
+        $html = '<meta property="og:description" name="description" content="' . $description . '">';
+        $this->assertEquals($description, html_extract_tag('description', $html));
     }
 
     /**
@@ -105,6 +175,25 @@ public function testHtmlExtractNonExistentNameTag()
     {
         $html = '<html><meta>stuff2</meta><meta name="image" content="img"/></html>';
         $this->assertFalse(html_extract_tag('description', $html));
+
+        // Partial meta tag
+        $html = '<meta content="Brief description">';
+        $this->assertFalse(html_extract_tag('description', $html));
+
+        $html = '<meta property="og:description">';
+        $this->assertFalse(html_extract_tag('description', $html));
+
+        $html = '<meta tag1="content1" property="og:description">';
+        $this->assertFalse(html_extract_tag('description', $html));
+
+        $html = '<meta property="og:description" tag1="content1">';
+        $this->assertFalse(html_extract_tag('description', $html));
+
+        $html = '<meta tag1="content1" content="Brief description">';
+        $this->assertFalse(html_extract_tag('description', $html));
+
+        $html = '<meta content="Brief description" tag1="content1">';
+        $this->assertFalse(html_extract_tag('description', $html));
     }
 
     /**