Merge pull request #165 from Zegnat/fix-e-html-output

Stop adding whitespace in html that is not in the source document

Author: aaronpk

Mf2/Parser.php

@@ -897,10 +897,17 @@ public function parseE(\DOMElement $e) {
 		// TODO: as it is this is not relative to only children, make this .// and rerun tests
 		$this->resolveChildUrls($e);
 
-		$html = '';
-		foreach ($e->childNodes as $node) {
-			$html .= $node->ownerDocument->saveHTML($node);
-		}
+		// Temporarily move all descendants into a separate DocumentFragment.
+		// This way we can DOMDocument::saveHTML on the entire collection at once.
+		// Running DOMDocument::saveHTML per node may add whitespace that isn't in source.
+		// See https://stackoverflow.com/q/38317903
+		$innerNodes = $e->ownerDocument->createDocumentFragment();
+		while ($e->hasChildNodes()) {
+			$innerNodes->appendChild($e->firstChild);
+		}
+		$html = $e->ownerDocument->saveHtml($innerNodes);
+		// Put the nodes back in place.
+		$e->appendChild($innerNodes);
 
 		$return = array(
 			'html' => unicodeTrim($html),

tests/Mf2/ParserTest.php

@@ -696,6 +696,13 @@ public function testMultiLevelRecursion() {
     $this->assertEquals('Comment D', $three['children'][1]['properties']['name'][0]);
     $this->assertEquals('Four', $output['items'][0]['children'][3]['properties']['name'][0]);
   }
+  
+  public function testNoErrantWhitespaceOnEHtml()
+  {
+    $input = '<div class="h-entry"><div class="e-content"><p>1</p><p>2</p></div></div>';
+    $output = Mf2\parse($input);
+    $this->assertEquals('<p>1</p><p>2</p>', $output['items'][0]['properties']['content'][0]['html']);
+  }
 
 	/**
 	 * @see https://github.com/indieweb/php-mf2/issues/158