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

Improve regex to extract HTML metadata (title, description, etc.)
This commit is contained in:
ArthurHoaro 2020-10-13 12:26:55 +02:00 committed by GitHub
commit 458b6b9918
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 93 additions and 2 deletions

View file

@ -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

View file

@ -94,8 +94,78 @@ class LinkUtilsTest extends TestCase
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 @@ class LinkUtilsTest extends TestCase
{
$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));
}
/**