infong
/
php-readability
mirror of https://github.com/j0k3r/php-readability
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #104 from jtojnar/html-shadowing

Browse Source 
Fix discarding `html[lang]`
pull/105/head 2.0.6
Jérémy Benoist 1 year ago committed by GitHub
parent a18cd0f2a9 efbbc86df9
commit 7413a38ff0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 89 additions and 27 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 43
      src/Readability.php
  2. 73
      tests/ReadabilityTest.php

43
src/Readability.php
Unescape View File

@ -1422,7 +1422,7 @@ class Readability implements LoggerAwareInterface
unset($tidy); unset($tidy);
} }
$this->html = '<meta charset="utf-8">' . (string) $this->html; $this->html = self::ensureMetaCharset((string) $this->html);
if ('html5lib' === $this->parser || 'html5' === $this->parser) { if ('html5lib' === $this->parser || 'html5' === $this->parser) {
$this->dom = (new HTML5())->loadHTML($this->html); $this->dom = (new HTML5())->loadHTML($this->html);
@ -1510,4 +1510,45 @@ class Readability implements LoggerAwareInterface
) )
&& !$node->hasAttribute('hidden'); && !$node->hasAttribute('hidden');
} }
/**
* Tries to insert `meta[charset]` tag into the proper place in the passed HTML document.
*
* `DOMDocument::loadHTML` will parse HTML documents as ISO-8859-1 if there is no `meta[charset]` tag.
* This means that UTF-8-encoded HTML fragments such as those coming from JSON-LD `articleBody` field would be parsed with incorrect encoding.
* Unfortunately, we cannot just put the tag at the start of the HTML fragment, since that would cause parser to auto-insert a `html` element, losing the attributes of the original `html` tag.
*
* @param string $html UTF-8 encoded document
*/
private static function ensureMetaCharset(string $html): string
{
$charsetTag = '<meta charset="utf-8">';
// Only look at first 1024 bytes since, according to HTML5 specification,
// that’s where <meta> elements declaring a character encoding must be located.
// https://developer.mozilla.org/en-US/docs/Web/HTML/Element/meta#charset
$start = substr($html, 0, 1000);
if (1 === preg_match('/<meta[^>]+charset/i', $start)) {
// <meta> tag is already present, no need for modification.
return $html;
}
if (1 === preg_match('/<head[^>]*>/i', $start)) {
// <head> tag was located, <meta> tags go there.
$html = preg_replace('/<head[^>]*>/i', '$0' . $charsetTag, $html, 1);
return $html;
}
if (1 === preg_match('/<html[^>]*>/i', $start)) {
// <html> tag was located, let’s put it inside and have parser create <head>.
$html = preg_replace('/<html[^>]*>/i', '$0' . $charsetTag, $html, 1);
return $html;
}
// Fallback – just plop the <meta> at the start of the fragment.
return $charsetTag . $html;
}
} }

73
tests/ReadabilityTest.php
Unescape View File

@ -24,7 +24,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$readability->init(); $readability->init();
$this->assertNull($readability->url); $this->assertNull($readability->url);
$this->assertInstanceOf('DomDocument', $readability->dom); $this->assertInstanceOf(\DOMDocument::class, $readability->dom);
} }
public function testConstructHtml5Parser(): void public function testConstructHtml5Parser(): void
@ -33,7 +33,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$readability->init(); $readability->init();
$this->assertSame('http://0.0.0.0', $readability->url); $this->assertSame('http://0.0.0.0', $readability->url);
$this->assertInstanceOf('DomDocument', $readability->dom); $this->assertInstanceOf(\DOMDocument::class, $readability->dom);
$this->assertSame('<html/>', $readability->original_html); $this->assertSame('<html/>', $readability->original_html);
} }
@ -46,7 +46,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$readability->init(); $readability->init();
$this->assertSame('http://0.0.0.0', $readability->url); $this->assertSame('http://0.0.0.0', $readability->url);
$this->assertInstanceOf('DomDocument', $readability->dom); $this->assertInstanceOf(\DOMDocument::class, $readability->dom);
$this->assertSame('<html/>', $readability->original_html); $this->assertSame('<html/>', $readability->original_html);
$this->assertTrue($readability->tidied); $this->assertTrue($readability->tidied);
} }
@ -60,7 +60,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$this->assertSame('', $readability->original_html); $this->assertSame('', $readability->original_html);
$this->assertFalse($readability->tidied); $this->assertFalse($readability->tidied);
$this->assertInstanceOf('DomDocument', $readability->dom); $this->assertInstanceOf(\DOMDocument::class, $readability->dom);
} }
public function testConstructSimpleWithoutTidy(): void public function testConstructSimpleWithoutTidy(): void
@ -69,7 +69,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$readability->init(); $readability->init();
$this->assertSame('http://0.0.0.0', $readability->url); $this->assertSame('http://0.0.0.0', $readability->url);
$this->assertInstanceOf('DomDocument', $readability->dom); $this->assertInstanceOf(\DOMDocument::class, $readability->dom);
$this->assertSame('<html/>', $readability->original_html); $this->assertSame('<html/>', $readability->original_html);
$this->assertFalse($readability->tidied); $this->assertFalse($readability->tidied);
} }
@ -115,7 +115,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testInitDiv(): void public function testInitDiv(): void
{ {
$readability = $this->getReadability('<div>' . str_repeat('This is the awesome content :)', 7) . '</div>', 'http://0.0.0.0'); $readability = $this->getReadability('<div>' . str_repeat('This is the awesome content :)', 7) . '</div>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -129,7 +128,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithFootnotes(): void public function testWithFootnotes(): void
{ {
$readability = $this->getReadability('<div>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '</div>', 'http://0.0.0.0'); $readability = $this->getReadability('<div>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '</div>', 'http://0.0.0.0');
$readability->debug = true;
$readability->convertLinksToFootnotes = true; $readability->convertLinksToFootnotes = true;
$res = $readability->init(); $res = $readability->init();
@ -146,7 +144,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testStandardClean(): void public function testStandardClean(): void
{ {
$readability = $this->getReadability('<div><h2>Title</h2>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<a href="#nofollow" rel="nofollow">will NOT be removed</a></div>', 'http://0.0.0.0'); $readability = $this->getReadability('<div><h2>Title</h2>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<a href="#nofollow" rel="nofollow">will NOT be removed</a></div>', 'http://0.0.0.0');
$readability->debug = true;
$readability->lightClean = false; $readability->lightClean = false;
$res = $readability->init(); $res = $readability->init();
@ -163,7 +160,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithIframe(): void public function testWithIframe(): void
{ {
$readability = $this->getReadability('<div><h2>Title</h2>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<p>This is an awesome text with some links, here there are <iframe src="http://youtube.com/test" href="#nofollow" rel="nofollow"></iframe><iframe>http://soundcloud.com/test</iframe></p></div>', 'http://0.0.0.0'); $readability = $this->getReadability('<div><h2>Title</h2>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<p>This is an awesome text with some links, here there are <iframe src="http://youtube.com/test" href="#nofollow" rel="nofollow"></iframe><iframe>http://soundcloud.com/test</iframe></p></div>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -178,7 +174,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithArticle(): void public function testWithArticle(): void
{ {
$readability = $this->getReadability('<article><p>' . str_repeat('This is an awesome text with some links, here there are: the awesome', 20) . '</p><p>This is an awesome text with some links, here there are <iframe src="http://youtube.com/test" href="#nofollow" rel="nofollow"></iframe></p></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<article><p>' . str_repeat('This is an awesome text with some links, here there are: the awesome', 20) . '</p><p>This is an awesome text with some links, here there are <iframe src="http://youtube.com/test" href="#nofollow" rel="nofollow"></iframe></p></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -193,7 +188,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithAside(): void public function testWithAside(): void
{ {
$readability = $this->getReadability('<article>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<footer><aside>' . str_repeat('<p>This is an awesome text with some links, here there are</p>', 8) . '</aside></footer></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<article>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<footer><aside>' . str_repeat('<p>This is an awesome text with some links, here there are</p>', 8) . '</aside></footer></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -208,7 +202,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithClasses(): void public function testWithClasses(): void
{ {
$readability = $this->getReadability('<article>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<div style="display:none">' . str_repeat('<p class="clock">This text should be removed</p>', 10) . '</div></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<article>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<div style="display:none">' . str_repeat('<p class="clock">This text should be removed</p>', 10) . '</div></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -223,7 +216,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithClassesWithoutLightClean(): void public function testWithClassesWithoutLightClean(): void
{ {
$readability = $this->getReadability('<article>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<div style="display:none">' . str_repeat('<p class="clock">This text should be removed</p>', 10) . '</div></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<article>' . str_repeat('<p>This is an awesome text with some links, here there are: <a href="http://0.0.0.0/test.html">the awesome</a></p>', 7) . '<div style="display:none">' . str_repeat('<p class="clock">This text should be removed</p>', 10) . '</div></article>', 'http://0.0.0.0');
$readability->debug = true;
$readability->lightClean = false; $readability->lightClean = false;
$res = $readability->init(); $res = $readability->init();
@ -239,7 +231,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithTd(): void public function testWithTd(): void
{ {
$readability = $this->getReadability('<table><tr>' . str_repeat('<td><p>This is an awesome text with some links, here there are the awesome</td>', 7) . '</tr></table>', 'http://0.0.0.0'); $readability = $this->getReadability('<table><tr>' . str_repeat('<td><p>This is an awesome text with some links, here there are the awesome</td>', 7) . '</tr></table>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -252,7 +243,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithSameClasses(): void public function testWithSameClasses(): void
{ {
$readability = $this->getReadability('<article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<div class="awesomecontent">This text is also an awesome text and you should know that !</div></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<div class="awesomecontent">This text is also an awesome text and you should know that !</div></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -266,7 +256,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testWithScript(): void public function testWithScript(): void
{ {
$readability = $this->getReadability('<article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p><script>This text is also an awesome text and you should know that !</script></p></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p><script>This text is also an awesome text and you should know that !</script></p></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -280,7 +269,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testTitle(): void public function testTitle(): void
{ {
$readability = $this->getReadability('<title>this is my title</title><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<title>this is my title</title><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -294,7 +282,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testTitleWithDash(): void public function testTitleWithDash(): void
{ {
$readability = $this->getReadability('<title> title2 - title3 </title><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<title> title2 - title3 </title><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -308,7 +295,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testTitleWithDoubleDot(): void public function testTitleWithDoubleDot(): void
{ {
$readability = $this->getReadability('<title> title2 : title3 </title><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<title> title2 : title3 </title><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -322,7 +308,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testTitleTooShortUseH1(): void public function testTitleTooShortUseH1(): void
{ {
$readability = $this->getReadability('<title>too short</title><h1>this is my h1 title !</h1><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0'); $readability = $this->getReadability('<title>too short</title><h1>this is my h1 title !</h1><article class="awesomecontent">' . str_repeat('<p>This is an awesome text with some links, here there are the awesome</p>', 7) . '<p></p></article>', 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -369,7 +354,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
</html>'; </html>';
$readability = $this->getReadability($data, 'http://iosgames.ru/?p=22030'); $readability = $this->getReadability($data, 'http://iosgames.ru/?p=22030');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
@ -437,7 +421,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
</html>'; </html>';
$readability = $this->getReadability($data, 'http://0.0.0.0'); $readability = $this->getReadability($data, 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
@ -474,7 +457,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$html = (string) file_get_contents('tests/fixtures/childNodeGoesNull.html'); $html = (string) file_get_contents('tests/fixtures/childNodeGoesNull.html');
$readability = $this->getReadability($html, 'http://0.0.0.0'); $readability = $this->getReadability($html, 'http://0.0.0.0');
$readability->debug = true;
$readability->convertLinksToFootnotes = true; $readability->convertLinksToFootnotes = true;
$res = $readability->init(); $res = $readability->init();
@ -487,7 +469,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$html = (string) file_get_contents('tests/fixtures/keepFootnotes.html'); $html = (string) file_get_contents('tests/fixtures/keepFootnotes.html');
$readability = $this->getReadability($html, 'http://0.0.0.0'); $readability = $this->getReadability($html, 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -501,7 +482,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
$html = (string) file_get_contents('tests/fixtures/wipedBody.html'); $html = (string) file_get_contents('tests/fixtures/wipedBody.html');
$readability = $this->getReadability($html, 'http://0.0.0.0', 'libxml', false); $readability = $this->getReadability($html, 'http://0.0.0.0', 'libxml', false);
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
$this->assertTrue($res); $this->assertTrue($res);
@ -540,7 +520,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
public function testVisibleNode(string $content, bool $shouldBeVisible): void public function testVisibleNode(string $content, bool $shouldBeVisible): void
{ {
$readability = $this->getReadability($content, 'http://0.0.0.0'); $readability = $this->getReadability($content, 'http://0.0.0.0');
$readability->debug = true;
$res = $readability->init(); $res = $readability->init();
if ($shouldBeVisible) { if ($shouldBeVisible) {
@ -550,6 +529,48 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
} }
} }
/**
* @return array<string, array{0: string, 1: string, 2?: bool}>
*/
public function dataForHtmlLang(): array
{
return [
'meta' => [
'<html lang="fr"><head><meta charset="utf-8"></head><body><article>' . str_repeat('<p>This is the awesome content :)</p>', 7) . '</article></body></html>',
'fr',
],
'head' => [
'<html lang="fr"><head><title>Foo</title></head><body><article>' . str_repeat('<p>This is the awesome content :)</p>', 7) . '</article></body></html>',
'fr',
],
'headless' => [
'<html lang="fr"><body><article>' . str_repeat('<p>This is the awesome content :)</p>', 7) . '</article></body></html>',
'fr',
// tidy would add <head> tag.
false,
],
'fragment' => [
'<article>' . str_repeat('<p>This is the awesome content :)</p>', 7) . '</article>',
'',
// tidy would add <html>.
false,
],
];
}
/**
* @dataProvider dataForHtmlLang
*/
public function testHtmlLang(string $html, string $lang, bool $useTidy = true): void
{
$readability = $this->getReadability($html, 'http://0.0.0.0', 'libxml', $useTidy);
$res = $readability->init();
$this->assertTrue($res);
$this->assertInstanceOf(\DOMDocument::class, $readability->dom);
$this->assertSame($lang, $readability->dom->documentElement->getAttribute('lang'));
}
private function getReadability(string $html, ?string $url = null, string $parser = 'libxml', bool $useTidy = true): Readability private function getReadability(string $html, ?string $url = null, string $parser = 'libxml', bool $useTidy = true): Readability
{ {
$readability = new Readability($html, $url, $parser, $useTidy); $readability = new Readability($html, $url, $parser, $useTidy);

Write Preview
Loading…
Cancel
Save