Merge pull request #111 from j0k3r/fix/tidy-php85

Fix bad encoding for `tidy_repair_string`

.github/workflows/continuous-integration.yml

@@ -25,6 +25,7 @@ jobs:
           - "8.2"
           - "8.3"
           - "8.4"
+          - "8.5"
 
     steps:
       - name: "Checkout"

.gitignore

@@ -4,3 +4,4 @@ composer.lock
 .php_cs.cache
 .php-cs-fixer.cache
 .phpunit.result.cache
+phpstan.neon

composer.json

@@ -32,10 +32,10 @@
     "require-dev": {
         "friendsofphp/php-cs-fixer": "^3.0",
         "monolog/monolog": "^1.24|^2.1",
-        "symfony/phpunit-bridge": "^4.4|^5.3|^6.0",
+        "symfony/phpunit-bridge": "^4.4|^5.3|^6.0|^7.0",
-        "phpstan/phpstan": "^1.3",
+        "phpstan/phpstan": "^2.0",
-        "phpstan/phpstan-phpunit": "^1.0",
+        "phpstan/phpstan-phpunit": "^2.0",
-        "rector/rector": "^0.15.0"
+        "rector/rector": "^2.0.0"
     },
     "suggest": {
         "ext-tidy": "Used to clean up given HTML and to avoid problems with bad HTML structure."

phpstan.dist.neon

@@ -8,8 +8,6 @@ parameters:
     bootstrapFiles:
         - vendor/bin/.phpunit/phpunit/vendor/autoload.php
 
-    checkMissingIterableValueType: false
-
 includes:
     - vendor/phpstan/phpstan-phpunit/extension.neon
     - vendor/phpstan/phpstan-phpunit/rules.neon

rector.php

@@ -3,29 +3,16 @@
 declare(strict_types=1);
 
 use Rector\Config\RectorConfig;
-use Rector\Core\ValueObject\PhpVersion;
 use Rector\Set\ValueObject\LevelSetList;
 
-return static function (RectorConfig $rectorConfig): void {
+return RectorConfig::configure()
-    // paths to refactor; solid alternative to CLI arguments
+    ->withPaths([
-    $rectorConfig->paths([
         __DIR__ . '/src',
         __DIR__ . '/tests',
-    ]);
+    ])
-
+    ->withBootstrapFiles([
-    // Path to phpstan with extensions, that PHPSTan in Rector uses to determine types
-    $rectorConfig->phpstanConfig(__DIR__ . '/phpstan.neon');
-
-    $rectorConfig->bootstrapFiles([
         __DIR__ . '/vendor/bin/.phpunit/phpunit/vendor/autoload.php',
         __DIR__ . '/vendor/autoload.php',
-    ]);
+    ])
-
+    ->withSets([LevelSetList::UP_TO_PHP_74])
-    // Define what rule sets will be applied
+;
-    $rectorConfig->sets([
-        LevelSetList::UP_TO_PHP_74,
-    ]);
-
-    // is your PHP version different from the one your refactor to?
-    $rectorConfig->phpVersion(PhpVersion::PHP_74);
-};

src/Readability.php

@@ -88,7 +88,7 @@ class Readability implements LoggerAwareInterface
         'enclose-text' => true,
         'merge-divs' => true,
         // 'merge-spans' => true,
-        'input-encoding' => '????',
+        'input-encoding' => 'utf8',
         'output-encoding' => 'utf8',
         'hide-comments' => true,
     ];
@@ -1118,7 +1118,7 @@ class Readability implements LoggerAwareInterface
 
         $topCandidates = array_filter(
             $topCandidates,
-            fn ($v, $idx) => 0 === $idx || null !== $v,
+            static fn ($v, $idx) => 0 === $idx || null !== $v,
             \ARRAY_FILTER_USE_BOTH
         );
         $topCandidate = $topCandidates[0];
@@ -1383,7 +1383,10 @@ class Readability implements LoggerAwareInterface
         $this->logger->debug('Parsing URL: ' . $this->url);
 
         if ($this->url) {
-            $this->domainRegExp = '/' . strtr((string) preg_replace('/www\d*\./', '', (string) parse_url($this->url, \PHP_URL_HOST)), ['.' => '\.']) . '/';
+            $host = parse_url($this->url, \PHP_URL_HOST);
+            if (null !== $host) {
+                $this->domainRegExp = '/' . strtr((string) preg_replace('/www\d*\./', '', $host), ['.' => '\.']) . '/';
+            }
         }
 
         mb_internal_encoding('UTF-8');
@@ -1419,7 +1422,7 @@ class Readability implements LoggerAwareInterface
             unset($tidy);
         }
 
-        $this->html = '<meta charset="utf-8">' . (string) $this->html;
+        $this->html = self::entitizeNonAscii((string) $this->html);
 
         if ('html5lib' === $this->parser || 'html5' === $this->parser) {
             $this->dom = (new HTML5())->loadHTML($this->html);
@@ -1478,7 +1481,7 @@ class Readability implements LoggerAwareInterface
     private function hasSingleTagInsideElement(\DOMElement $node, string $tag): bool
     {
         $childNodes = iterator_to_array($node->childNodes);
-        $children = array_filter($childNodes, fn ($childNode) => $childNode instanceof \DOMElement);
+        $children = array_filter($childNodes, static fn ($childNode) => $childNode instanceof \DOMElement);
 
         // There should be exactly 1 element child with given tag
         if (1 !== \count($children) || $children[0]->nodeName !== $tag) {
@@ -1507,4 +1510,21 @@ class Readability implements LoggerAwareInterface
         )
         && !$node->hasAttribute('hidden');
     }
+
+    /**
+     * Converts non-ASCII UTF-8 characters to numeric HTML entities.
+     *
+     * `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.
+     *
+     * @param string $html UTF-8 encoded document
+     */
+    private static function entitizeNonAscii(string $html): string
+    {
+        $convmap = [
+            0x80, 0x1FFFFF, 0, 0x10FFFF,
+        ];
+
+        return mb_encode_numericentity($html, $convmap, 'utf8', true);
+    }
 }

tests/ReadabilityTest.php

@@ -21,10 +21,11 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     public function testConstructDefault(): void
     {
         $readability = $this->getReadability('');
+        $this->assertSame('utf8', $readability->tidy_config['input-encoding']);
         $readability->init();
 
         $this->assertNull($readability->url);
-        $this->assertInstanceOf('DomDocument', $readability->dom);
+        $this->assertInstanceOf(\DOMDocument::class, $readability->dom);
     }
 
     public function testConstructHtml5Parser(): void
@@ -33,7 +34,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $readability->init();
 
         $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);
     }
 
@@ -46,7 +47,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $readability->init();
 
         $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);
         $this->assertTrue($readability->tidied);
     }
@@ -60,7 +61,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $this->assertSame('', $readability->original_html);
         $this->assertFalse($readability->tidied);
 
-        $this->assertInstanceOf('DomDocument', $readability->dom);
+        $this->assertInstanceOf(\DOMDocument::class, $readability->dom);
     }
 
     public function testConstructSimpleWithoutTidy(): void
@@ -69,7 +70,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $readability->init();
 
         $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);
         $this->assertFalse($readability->tidied);
     }
@@ -115,7 +116,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     public function testInitDiv(): void
     {
         $readability = $this->getReadability('<div>' . str_repeat('This is the awesome content :)', 7) . '</div>', 'http://0.0.0.0');
-        $readability->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -129,7 +129,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $readability->convertLinksToFootnotes = true;
         $res = $readability->init();
 
@@ -146,7 +145,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $readability->lightClean = false;
         $res = $readability->init();
 
@@ -163,7 +161,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -178,7 +175,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -193,7 +189,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -208,7 +203,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -223,7 +217,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $readability->lightClean = false;
         $res = $readability->init();
 
@@ -239,7 +232,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -252,7 +244,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -266,7 +257,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -280,7 +270,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -294,7 +283,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -308,7 +296,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -322,7 +309,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     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->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -338,7 +324,7 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $oldErrorReporting = error_reporting(\E_ALL);
         $oldDisplayErrors = ini_set('display_errors', '1');
         // dummy function to be used to the next test
-        set_error_handler(function (int $errno, string $errstr, string $errfile, int $errline) {
+        set_error_handler(static function (int $errno, string $errstr, string $errfile, int $errline) {
             throw new \Exception($errstr, $errno);
         });
 
@@ -369,7 +355,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
                 </html>';
 
             $readability = $this->getReadability($data, 'http://iosgames.ru/?p=22030');
-            $readability->debug = true;
 
             $res = $readability->init();
 
@@ -437,7 +422,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
             </html>';
 
         $readability = $this->getReadability($data, 'http://0.0.0.0');
-        $readability->debug = true;
 
         $res = $readability->init();
 
@@ -474,7 +458,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $html = (string) file_get_contents('tests/fixtures/childNodeGoesNull.html');
 
         $readability = $this->getReadability($html, 'http://0.0.0.0');
-        $readability->debug = true;
         $readability->convertLinksToFootnotes = true;
         $res = $readability->init();
 
@@ -487,7 +470,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $html = (string) file_get_contents('tests/fixtures/keepFootnotes.html');
 
         $readability = $this->getReadability($html, 'http://0.0.0.0');
-        $readability->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -501,7 +483,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         $html = (string) file_get_contents('tests/fixtures/wipedBody.html');
 
         $readability = $this->getReadability($html, 'http://0.0.0.0', 'libxml', false);
-        $readability->debug = true;
         $res = $readability->init();
 
         $this->assertTrue($res);
@@ -540,7 +521,6 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
     public function testVisibleNode(string $content, bool $shouldBeVisible): void
     {
         $readability = $this->getReadability($content, 'http://0.0.0.0');
-        $readability->debug = true;
         $res = $readability->init();
 
         if ($shouldBeVisible) {
@@ -550,6 +530,64 @@ class ReadabilityTest extends \PHPUnit\Framework\TestCase
         }
     }
 
+    // https://github.com/wallabag/wallabag/issues/8158
+    public function testCharsetAfterTitle(): void
+    {
+        $readability = $this->getReadability('<!DOCTYPE html><html lang="et"><head><title>Tõde ja õigus I</title> <meta charset="utf-8"></head><body><p>See oli läinud aastasaja kolmanda veerandi lõpul. Päike lähenes silmapiirile, seistes sedavõrd madalas, et enam ei ulatunud valgustama ei mäkke ronivat hobust, kes puutelgedega vankrit vedas, ei vankril istuvat noort naist ega ka ligi kolmekümnelist meest, kes kõndis vankri kõrval.</p></body></html>', 'https://et.wikisource.org/wiki/T%C3%B5de_ja_%C3%B5igus_I/I');
+        $readability->convertLinksToFootnotes = true;
+        $res = $readability->init();
+
+        $this->assertTrue($res);
+        $this->assertInstanceOf('Readability\JSLikeHTMLElement', $readability->getContent());
+        $this->assertInstanceOf('Readability\JSLikeHTMLElement', $readability->getTitle());
+        $this->assertSame('Tõde ja õigus I', $readability->getTitle()->getInnerHtml());
+        $this->assertStringContainsString('Päike lähenes', $readability->getContent()->getInnerHtml());
+    }
+
+    /**
+     * @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>Tous les êtres humains naissent libres et égaux en dignité et en droits. Ils sont doués de raison et de conscience et doivent agir les uns envers les autres dans un esprit de fraternité.</p>', 7) . '</article></body></html>',
+                'fr',
+            ],
+            'head' => [
+                '<html lang="fr"><head><title>Foo</title></head><body><article>' . str_repeat('<p>Tous les êtres humains naissent libres et égaux en dignité et en droits. Ils sont doués de raison et de conscience et doivent agir les uns envers les autres dans un esprit de fraternité.</p>', 7) . '</article></body></html>',
+                'fr',
+            ],
+            'headless' => [
+                '<html lang="fr"><body><article>' . str_repeat('<p>Tous les êtres humains naissent libres et égaux en dignité et en droits. Ils sont doués de raison et de conscience et doivent agir les uns envers les autres dans un esprit de fraternité.</p>', 7) . '</article></body></html>',
+                'fr',
+                // tidy would add <head> tag.
+                false,
+            ],
+            'fragment' => [
+                '<article>' . str_repeat('<p>Tous les êtres humains naissent libres et égaux en dignité et en droits. Ils sont doués de raison et de conscience et doivent agir les uns envers les autres dans un esprit de fraternité.</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'));
+        $this->assertInstanceOf('Readability\JSLikeHTMLElement', $readability->getContent());
+        $this->assertStringContainsString('êtres', $readability->getContent()->getInnerHtml());
+    }
+
     private function getReadability(string $html, ?string $url = null, string $parser = 'libxml', bool $useTidy = true): Readability
     {
         $readability = new Readability($html, $url, $parser, $useTidy);