From 6289f3c48ca8dde511d227411e16b731d3c05328 Mon Sep 17 00:00:00 2001
From: Nik Everett <neverett@wikimedia.org>
Date: Fri, 18 Apr 2014 09:16:04 -0400
Subject: [PATCH] Make HtmlFormatter return removed elements

This shouldn't cause much overhead and is useful for getting the text of
the removed tags.

Change-Id: I97cf66014719244b8bb2b0509b419c82202bdb01
---
 includes/HtmlFormatter.php                   | 22 +++++++++++-------
 tests/phpunit/includes/HtmlFormatterTest.php | 24 ++++++++++++++++----
 2 files changed, 34 insertions(+), 12 deletions(-)

diff --git a/includes/HtmlFormatter.php b/includes/HtmlFormatter.php
index 7f590e5207..96ffe9ea5c 100644
--- a/includes/HtmlFormatter.php
+++ b/includes/HtmlFormatter.php
@@ -128,7 +128,9 @@ class HtmlFormatter {
 	}
 
 	/**
-	 * Removes content we've chosen to remove
+	 * Removes content we've chosen to remove.  The text of the removed elements can be
+	 * extracted with the getText method.
+	 * @return array of removed DOMElements
 	 */
 	public function filterContent() {
 		wfProfileIn( __METHOD__ );
@@ -156,8 +158,7 @@ class HtmlFormatter {
 				}
 			}
 		}
-
-		$this->removeElements( $domElemsToRemove );
+		$removed = $this->removeElements( $domElemsToRemove );
 
 		// Elements with named IDs
 		$domElemsToRemove = array();
@@ -167,7 +168,7 @@ class HtmlFormatter {
 				$domElemsToRemove[] = $itemToRemoveNode;
 			}
 		}
-		$this->removeElements( $domElemsToRemove );
+		$removed = array_merge( $removed, $this->removeElements( $domElemsToRemove ) );
 
 		// CSS Classes
 		$domElemsToRemove = array();
@@ -183,7 +184,7 @@ class HtmlFormatter {
 				}
 			}
 		}
-		$this->removeElements( $domElemsToRemove );
+		$removed = array_merge( $removed, $this->removeElements( $domElemsToRemove ) );
 
 		// Tags with CSS Classes
 		foreach ( $removals['TAG_CLASS'] as $classToRemove ) {
@@ -192,16 +193,17 @@ class HtmlFormatter {
 			$elements = $xpath->query(
 				'//' . $parts[0] . '[@class="' . $parts[1] . '"]'
 			);
-
-			$this->removeElements( $elements );
+			$removed = array_merge( $removed, $this->removeElements( $elements ) );
 		}
 
 		wfProfileOut( __METHOD__ );
+		return $removed;
 	}
 
 	/**
 	 * Removes a list of elelments from DOMDocument
 	 * @param array|DOMNodeList $elements
+	 * @return array of removed elements
 	 */
 	private function removeElements( $elements ) {
 		$list = $elements;
@@ -217,6 +219,7 @@ class HtmlFormatter {
 				$element->parentNode->removeChild( $element );
 			}
 		}
+		return $list;
 	}
 
 	/**
@@ -245,7 +248,10 @@ class HtmlFormatter {
 	}
 
 	/**
-	 * Performs final transformations and returns resulting HTML
+	 * Performs final transformations and returns resulting HTML.  Note that if you want to call this
+	 * both without an element and with an element you should call it without an element first.  If you
+	 * specify the $element in the method it'll change the underlying dom and you won't be able to get
+	 * it back.
 	 *
 	 * @param DOMElement|string|null $element ID of element to get HTML from or false to get it from the whole tree
 	 * @return string Processed HTML
diff --git a/tests/phpunit/includes/HtmlFormatterTest.php b/tests/phpunit/includes/HtmlFormatterTest.php
index 99a6efd659..98eff7b387 100644
--- a/tests/phpunit/includes/HtmlFormatterTest.php
+++ b/tests/phpunit/includes/HtmlFormatterTest.php
@@ -8,17 +8,23 @@ class HtmlFormatterTest extends MediaWikiTestCase {
 	 * @dataProvider getHtmlData
 	 * @covers HtmlFormatter::getText
 	 */
-	public function testTransform( $input, $expected, $callback = false ) {
+	public function testTransform( $input, $expectedText, $expectedRemoved = array(), $callback = false ) {
 		$input = self::normalize( $input );
 		$formatter = new HtmlFormatter( HtmlFormatter::wrapHTML( $input ) );
 		if ( $callback ) {
 			$callback( $formatter );
 		}
-		$formatter->filterContent();
+		$removedElements = $formatter->filterContent();
 		$html = $formatter->getText();
+		$removed = array();
+		foreach ( $removedElements as $removedElement ) {
+			$removed[] = self::normalize( $formatter->getText( $removedElement ) );
+		}
+		$expectedRemoved = array_map( 'self::normalize', $expectedRemoved );
 
 		$this->assertValidHtmlSnippet( $html );
-		$this->assertEquals( self::normalize( $expected ), self::normalize( $html ) );
+		$this->assertEquals( self::normalize( $expectedText ), self::normalize( $html ) );
+		$this->assertEquals( asort( $expectedRemoved ), asort( $removed ) );
 	}
 
 	private static function normalize( $s ) {
@@ -45,6 +51,7 @@ class HtmlFormatterTest extends MediaWikiTestCase {
 			array(
 				'<img src="/foo/bar.jpg" alt="Blah"/>',
 				'',
+				array( '<img src="/foo/bar.jpg" alt="Blah">' ),
 				$removeImages,
 			),
 			// basic tag removal
@@ -52,21 +59,30 @@ class HtmlFormatterTest extends MediaWikiTestCase {
 				'<table><tr><td>foo</td></tr></table><div class="foo">foo</div><div class="foo quux">foo</div><span id="bar">bar</span>
 <strong class="foo" id="bar">foobar</strong><div class="notfoo">test</div><div class="baz"/>
 <span class="baz">baz</span>',
-
 				'<div class="notfoo">test</div>
 <span class="baz">baz</span>',
+				array(
+					'<table><tr><td>foo</td></tr></table>',
+					'<div class="foo">foo</div>',
+					'<div class="foo quux">foo</div>',
+					'<span id="bar">bar</span>',
+					'<strong class="foo" id="bar">foobar</strong>',
+					'<div class="baz"/>',
+				),
 				$removeTags,
 			),
 			// don't flatten tags that start like chosen ones
 			array(
 				'<div><s>foo</s> <span>bar</span></div>',
 				'foo <span>bar</span>',
+				array(),
 				$flattenSomeStuff,
 			),
 			// total flattening
 			array(
 				'<div style="foo">bar<sup>2</sup></div>',
 				'bar2',
+				array(),
 				$flattenEverything,
 			),
 			// UTF-8 preservation and security
-- 
2.20.1