[DomCrawler] Give choice of used parser

victor-prdh · nicolas-grekas · commit 68800eb7dad4 · 2023-03-06T12:07:16.000+01:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -4,6 +4,7 @@ CHANGELOG
 6.3
 ---
 
+ * Add `$useHtml5Parser` argument to `Crawler`
  * Add `CrawlerSelectorCount` test constraint
  * Add argument `$normalizeWhitespace` to `Crawler::innerText()`
  * Make `Crawler::innerText()` return the first non-empty text
diff --git a/Crawler.php b/Crawler.php
@@ -58,16 +58,17 @@ class Crawler implements \Countable, \IteratorAggregate
      */
     private bool $isHtml = true;
 
-    private HTML5 $html5Parser;
+
+    private ?HTML5 $html5Parser = null;
 
     /**
      * @param \DOMNodeList|\DOMNode|\DOMNode[]|string|null $node A Node to use as the base for the crawling
      */
-    public function __construct(\DOMNodeList|\DOMNode|array|string $node = null, string $uri = null, string $baseHref = null)
+    public function __construct(\DOMNodeList|\DOMNode|array|string $node = null, string $uri = null, string $baseHref = null, bool $useHtml5Parser = true)
     {
         $this->uri = $uri;
         $this->baseHref = $baseHref ?: $uri;
-        $this->html5Parser = new HTML5(['disable_html_ns' => true]);
+        $this->html5Parser = $useHtml5Parser ? new HTML5(['disable_html_ns' => true]) : null;
         $this->cachedNamespaces = new \ArrayObject();
 
         $this->add($node);
@@ -621,7 +622,7 @@ public function html(string $default = null): string
         $node = $this->getNode(0);
         $owner = $node->ownerDocument;
 
-        if ('<!DOCTYPE html>' === $owner->saveXML($owner->childNodes[0])) {
+        if ($this->html5Parser && '<!DOCTYPE html>' === $owner->saveXML($owner->childNodes[0])) {
             $owner = $this->html5Parser;
         }
 
@@ -642,7 +643,7 @@ public function outerHtml(): string
         $node = $this->getNode(0);
         $owner = $node->ownerDocument;
 
-        if ('<!DOCTYPE html>' === $owner->saveXML($owner->childNodes[0])) {
+        if ($this->html5Parser && '<!DOCTYPE html>' === $owner->saveXML($owner->childNodes[0])) {
             $owner = $this->html5Parser;
         }
 
@@ -1215,6 +1216,10 @@ private function parseHtmlString(string $content, string $charset): \DOMDocument
 
     private function canParseHtml5String(string $content): bool
     {
+        if (!$this->html5Parser) {
+            return false;
+        }
+
         if (false === ($pos = stripos($content, '<!doctype html>'))) {
             return false;
         }
diff --git a/Tests/AbstractCrawlerTestCase.php b/Tests/AbstractCrawlerTestCase.php
@@ -21,9 +21,9 @@ abstract class AbstractCrawlerTestCase extends TestCase
 {
     abstract public static function getDoctype(): string;
 
-    protected function createCrawler($node = null, string $uri = null, string $baseHref = null)
+    protected function createCrawler($node = null, string $uri = null, string $baseHref = null, bool $useHtml5Parser = true)
     {
-        return new Crawler($node, $uri, $baseHref);
+        return new Crawler($node, $uri, $baseHref, $useHtml5Parser);
     }
 
     public function testConstructor()
diff --git a/Tests/Html5ParserCrawlerTest.php b/Tests/Html5ParserCrawlerTest.php
@@ -46,6 +46,20 @@ public function testHtml5ParserWithInvalidHeadedContent(string $content)
         self::assertEmpty($crawler->filterXPath('//h1')->text(), '->addHtmlContent failed as expected');
     }
 
+    public function testHtml5ParserNotSameAsNativeParserForSpecificHtml()
+    {
+        // Html who create a bug specific to the DOM extension (see https://github.com/symfony/symfony/issues/28596)
+        $html = $this->getDoctype().'<html><body><h1><p>Foo</p></h1></body></html>';
+
+        $html5Crawler = $this->createCrawler(null, null, null, true);
+        $html5Crawler->add($html);
+
+        $nativeCrawler = $this->createCrawler(null, null, null, false);
+        $nativeCrawler->add($html);
+
+        $this->assertNotEquals($nativeCrawler->filterXPath('//h1')->text(), $html5Crawler->filterXPath('//h1')->text(), 'Native parser and Html5 parser must be different');
+    }
+
     public static function validHtml5Provider(): iterable
     {
         $html = self::getDoctype().'<html><body><h1><p>Foo</p></h1></body></html>';

Original file line number	Diff line number	Diff line change
`@@ -21,9 +21,9 @@ abstract class AbstractCrawlerTestCase extends TestCase`
`21`	`21`	`{`
`22`	`22`	`abstract public static function getDoctype(): string;`
`23`	`23`
`24`		`- protected function createCrawler($node = null, string $uri = null, string $baseHref = null)`
	`24`	`+ protected function createCrawler($node = null, string $uri = null, string $baseHref = null, bool $useHtml5Parser = true)`
`25`	`25`	`{`
`26`		`- return new Crawler($node, $uri, $baseHref);`
	`26`	`+ return new Crawler($node, $uri, $baseHref, $useHtml5Parser);`
`27`	`27`	`}`
`28`	`28`
`29`	`29`	`public function testConstructor()`