Merge pull request #314 from wsexport/metadata-entities

Don't encode HTML entities before use
wikimedia · Jan 15, 2021 · bacc93b · bacc93b
2 parents f5f0628 + ddc095c
commit bacc93b
Show file tree

Hide file tree

Showing 3 changed files with 30 additions and 30 deletions.
diff --git a/src/BookProvider.php b/src/BookProvider.php
@@ -58,7 +58,14 @@ public function getMulti( array $titles, $isMetadata = false ) {
 		return $pages;
 	}
 
-	public function getMetadata( $title, $isMetadata, DOMDocument $doc ) {
+	/**
+	 * Get metadata etc. from a XHTML document.
+	 * @param string $title The book's titlepage's page name.
+	 * @param bool $isMetadata Whether this the book's content, chapters, credits, and pictures should also be extracted from the document.
+	 * @param DOMDocument $doc The document to read metadata from.
+	 * @return Book
+	 */
+	public function getMetadata( string $title, bool $isMetadata, DOMDocument $doc ): Book {
 		$pageList = [ $title ];
 		$parser = new PageParser( $doc );
 		$book = new Book();
@@ -76,18 +83,18 @@ public function getMetadata( $title, $isMetadata, DOMDocument $doc ) {
 		}
 
 		$book->type = $metadataParser->getMetadata( 'ws-type' );
-		$book->name = htmlspecialchars( $metadataParser->getMetadata( 'ws-title' ) );
+		$book->name = $metadataParser->getMetadata( 'ws-title' );
 		if ( $book->name == '' ) {
 			$book->name = $this->removeNamespacesFromTitle( str_replace( '_', ' ', $metadataSrc ) );
 		}
-		$book->periodical = htmlspecialchars( $metadataParser->getMetadata( 'ws-periodical' ) );
-		$book->author = htmlspecialchars( $metadataParser->getMetadata( 'ws-author' ) );
-		$book->translator = htmlspecialchars( $metadataParser->getMetadata( 'ws-translator' ) );
-		$book->illustrator = htmlspecialchars( $metadataParser->getMetadata( 'ws-illustrator' ) );
-		$book->school = htmlspecialchars( $metadataParser->getMetadata( 'ws-school' ) );
-		$book->publisher = htmlspecialchars( $metadataParser->getMetadata( 'ws-publisher' ) );
-		$book->year = htmlspecialchars( $metadataParser->getMetadata( 'ws-year' ) );
-		$book->place = htmlspecialchars( $metadataParser->getMetadata( 'ws-place' ) );
+		$book->periodical = $metadataParser->getMetadata( 'ws-periodical' );
+		$book->author = $metadataParser->getMetadata( 'ws-author' );
+		$book->translator = $metadataParser->getMetadata( 'ws-translator' );
+		$book->illustrator = $metadataParser->getMetadata( 'ws-illustrator' );
+		$book->school = $metadataParser->getMetadata( 'ws-school' );
+		$book->publisher = $metadataParser->getMetadata( 'ws-publisher' );
+		$book->year = $metadataParser->getMetadata( 'ws-year' );
+		$book->place = $metadataParser->getMetadata( 'ws-place' );
 		$book->key = $metadataParser->getMetadata( 'ws-key' );
 		$book->progress = $metadataParser->getMetadata( 'ws-progress' );
 		$book->volume = $metadataParser->getMetadata( 'ws-volume' );

diff --git a/src/Page.php b/src/Page.php
@@ -5,34 +5,19 @@
 use DOMDocument;
 
 /**
- * @author Thomas Pellissier Tanon
- * @copyright 2011 Thomas Pellissier Tanon
- * @license GPL-2.0-or-later
- */
-
-/**
- * container for a page of Wikisource
+ * Container for a page of Wikisource.
  */
 class Page {
 
-	/**
-	 * title of the book in Wikisource
-	 */
+	/** @var string Wiki page name of the page on Wikisource. */
 	public $title = '';
 
-	/**
-	 * name to display
-	 */
+	/** @var string The page's actual title, e.g. a page with title 'Foo/Bar' might have a name of 'Foo, Bar'. */
 	public $name = '';
 
-	/**
-	 * content of the page
-	 * @type DOMDocument
-	 */
+	/** @var DOMDocument Content of the page. */
 	public $content = null;
 
-	/**
-	 * list of the subpages as Page object
-	 */
+	/** @var Page[] List of the subpages. */
 	public $chapters = [];
 }
diff --git a/tests/Book/BookProviderTest.php b/tests/Book/BookProviderTest.php
@@ -82,4 +82,12 @@ private function parseDocument( $filename ) {
 		$this->assertTrue( $doc->loadHTMLFile( $filename ), 'parsing of "' . $filename . '"" failed' );
 		return $doc;
 	}
+
+	public function testTitleEntities() {
+		$doc = new DOMDocument();
+		$doc->loadHTML( '<body><p class="ws-title">Title &amp; name</p></body>' );
+		$book = $this->bookProvider->getMetadata( 'Title name', false, $doc );
+		$this->assertSame( 'Title name', $book->title );
+		$this->assertSame( 'Title & name', $book->name );
+	}
 }