Use preg_replace_callback throughout; lazy checks.

The checks for whether a page is a potential link target or not (depending
on the absence of the __NOAUTOLINKTARGET__ magic word and if it is not a
redirect to the current page) have now been moved into the callback
functions, so that they are only performed if a page really is a candidate
for linking (i.e, its title occurs on the currently edited page).

The change resulted in a ~10-fold increase in speed.

LinkTitles.body.php

@@ -29,6 +29,17 @@
 	/// Central class of the extension. Sets up parser hooks.
 	/// This class contains only static functions; do not instantiate.
 	class LinkTitles {
+		/// A Title object for the page that is being parsed.
+		private static $mCurrentTitle;
+
+		/// A Title object for the target page currently being examined.
+		private static $mTargetTitle;
+
+		/// The content object for the currently processed target page.
+		/// This variable is necessary to be able to prevent loading the target 
+		/// content twice.
+		private static $mTargetContent;
+
 		/// Setup function, hooks the extension's functions to MediaWiki events.
 		public static function setup() {
 			global $wgLinkTitlesParseOnEdit;
@@ -74,6 +85,7 @@
 		/// @param $content Content object that holds the article content
 		/// @returns true
 		static function parseContent( &$article, &$content ) {
+			wfProfileIn( __METHOD__ );
 
 			// If the page contains the magic word '__NOAUTOLINKS__', do not parse
 			// the content.
@@ -93,17 +105,9 @@
 			global $wgLinkTitlesWordEndOnly;
 			global $wgLinkTitlesSmartMode;
 			global $wgCapitalLinks;
-			global $wgLinkTitlesEnableNoTargetMagicWord;
-			global $wgLinkTitlesCheckRedirect;
 
 			( $wgLinkTitlesWordStartOnly ) ? $wordStartDelim = '\b' : $wordStartDelim = '';
 			( $wgLinkTitlesWordEndOnly ) ? $wordEndDelim = '\b' : $wordEndDelim = '';
-			// ( $wgLinkTitlesIgnoreCase ) ? $regexModifier = 'i' : $regexModifier = '';
-
-			// To prevent adding self-references, we now
-			// extract the current page's title.
-			$myTitle = $article->getTitle();
-			$myTitleText = $myTitle->GetText();
 
 			( $wgLinkTitlesPreferShortTitles ) ? $sort_order = 'ASC' : $sort_order = 'DESC';
 			( $wgLinkTitlesFirstOnly ) ? $limit = 1 : $limit = -1;
@@ -115,6 +119,10 @@
 				$templatesDelimiter = '{{[^|]+?}}|{{.+\||';
 			};
 
+			LinkTitles::$mCurrentTitle = $article->getTitle();
+			$text = $content->getContentHandler()->serializeContent($content);
+			$newText = $text;
+
 			// Build a regular expression that will capture existing wiki links ("[[...]]"),
 			// wiki headings ("= ... =", "== ... ==" etc.),  
 			// urls ("http://example.com", "[http://example.com]", "[http://example.com Description]",
@@ -143,8 +151,7 @@
 			// targets. This includes the current page.
 			$black_list = str_replace( '_', ' ',
 				'("' . implode( '", "',$wgLinkTitlesBlackList ) . 
-				$myTitle->getDbKey() . '")' );
+				LinkTitles::$mCurrentTitle->getDbKey() . '")' );
-
 
 			// Build an SQL query and fetch all page titles ordered by length from 
 			// shortest to longest. Only titles from 'normal' pages (namespace uid 
@@ -177,59 +184,30 @@
 				);
 			}
 
-			$text = $content->getContentHandler()->serializeContent($content);
+			// Build an anonymous callback function to be used in simple mode.
+			$simpleModeCallback = function( $matches ) {
+				if ( LinkTitles::checkTargetPage() ) {
+					return '[[' . $matches[0] . ']]';
+				}
+				else
+				{
+					return $matches[0];
+				}
+			};
 
 			// Iterate through the page titles
+			wfProfileIn('LinkTitles::parseContent-row_iterator');
 			foreach( $res as $row ) {
-				// Obtain an instance of a Title class for the current database row.
+				LinkTitles::newTarget( $row->page_title );
-				$targetTitle = Title::makeTitle(NS_MAIN, $row->page_title);
 
-				if ( $wgLinkTitlesCheckRedirect || $wgLinkTitlesEnableNoTargetMagicWord ) {
-					// Obtain a page object for the current title, so we can check for 
-					// the presence of the __NOAUTOLINKTARGET__ magic keyword.
-					$targetPageContent = WikiPage::factory($targetTitle)->getContent();
-
-					// To prevent linking to pages that redirect to the current page,
-					// obtain the title that the target page redirects to. Will be null 
-					// if there is no redirect.
-					if ( $wgLinkTitlesCheckRedirect ) {
-						$redirectTitle = $targetPageContent->getUltimateRedirectTarget();
-						$redirectCheck = !( $redirectTitle && $redirectTitle->equals($myTitle) );
-					}
-					else
-					{
-						$redirectCheck = true;
-					};
-
-					if ( $wgLinkTitlesEnableNoTargetMagicWord ) {
-						$magicWordCheck = ! $targetPageContent->matchMagicWord(
-							MagicWord::get('MAG_LINKTITLES_NOTARGET') ); 
-					}
-					else 
-					{
-						$magicWordCheck = true;
-					};
-				}
-				else
-				{
-					$redirectCheck = true;
-					$magicWordCheck = true;
-				}
-
-				// Proceed only if the currently examined page does not redirect to 
-				// our page and does not contain the no-target magic word.
-				// If the corresponding configuration variables are set to false,
-				// both 'check' variables below will be set to true by the code 
-				// above.
-				if ( $redirectCheck && $magicWordCheck ) {
 				// split the page content by [[...]] groups
 				// credits to inhan @ StackOverflow for suggesting preg_split
 				// see http://stackoverflow.com/questions/10672286
-					$arr = preg_split( $delimiter, $text, -1, PREG_SPLIT_DELIM_CAPTURE );
+				$arr = preg_split( $delimiter, $newText, -1, PREG_SPLIT_DELIM_CAPTURE );
 
 				// Escape certain special characters in the page title to prevent
 				// regexp compilation errors
-					$targetTitleText = $targetTitle->getText();
+				$targetTitleText = LinkTitles::$mTargetTitle->getText();
 				$quotedTitle = preg_quote($targetTitleText, '/');
 
 				// Depending on the global configuration setting $wgCapitalLinks,
@@ -245,9 +223,9 @@
 
 				for ( $i = 0; $i < count( $arr ); $i+=2 ) {
 					// even indexes will point to text that is not enclosed by brackets
-						$arr[$i] = preg_replace( '/(?<![\:\.\@\/\?\&])' .
+					$arr[$i] = preg_replace_callback( '/(?<![\:\.\@\/\?\&])' .
 						$wordStartDelim . $searchTerm . $wordEndDelim . '/',
-							'[[$1]]', $arr[$i], $limit, $count );
+						$simpleModeCallback, $arr[$i], $limit, $count );
 					if (( $limit >= 0 ) && ( $count > 0  )) {
 						break; 
 					};
@@ -273,6 +251,7 @@
 						// we need to ignore the first letter of the page titles, as 
 						// it does not matter for linking.
 						$callback = function ($matches) use ($targetTitleText) {
+							if ( LinkTitles::checkTargetPage() ) {
 								if ( strcmp(substr($targetTitleText, 1), substr($matches[0], 1)) == 0 ) {
 									// Case-sensitive match: no need to bulid piped link.
 									return '[[' . $matches[0] . ']]';
@@ -280,6 +259,11 @@
 									// Case-insensitive match: build piped link.
 									return '[[' . $targetTitleText . '|' . $matches[0] . ']]';
 								}
+							}
+							else
+							{
+								return $matches[0];
+							}
 						};
 					}
 					else
@@ -287,6 +271,7 @@
 						// If $wgCapitalLinks is false, we can use the simple variant 
 						// of the callback function.
 						$callback = function ($matches) use ($targetTitleText) {
+							if ( LinkTitles::checkTargetPage() ) {
 								if ( strcmp($targetTitleText, $matches[0]) == 0 ) {
 									// Case-sensitive match: no need to bulid piped link.
 									return '[[' . $matches[0] . ']]';
@@ -294,6 +279,11 @@
 									// Case-insensitive match: build piped link.
 									return '[[' . $targetTitleText . '|' . $matches[0] . ']]';
 								}
+							}
+							else
+							{
+								return $matches[0];
+							}
 						};
 					}
 
@@ -309,12 +299,13 @@
 						};
 					};
 					$newText = implode( '', $arr );
+				} // $wgLinkTitlesSmartMode
+			}; // foreach $res as $row
+			wfProfileOut('LinkTitles::parseContent-row_iterator');
 			if ( $newText != $text ) {
 				$content = $content->getContentHandler()->unserializeContent( $newText );
 			}
-					} // $wgLinkTitlesSmartMode
+			wfProfileOut( __METHOD__ );
-				}
-			}; // foreach $res as $row
 			return true;
 		}
 		
@@ -354,6 +345,63 @@
 			$mwa->matchAndRemove( $text );
 			return true;
 		}
+
+		/// Sets member variables for the current target page.
+		private function newTarget( $titleString ) {
+			// @todo Make this wiki namespace aware.
+			LinkTitles::$mTargetTitle = Title::makeTitle( NS_MAIN, $titleString );
+			LinkTitles::$mTargetContent = null;
+		}
+
+		/// Returns the content of the current target page.
+		/// This function serves to be used in preg_replace_callback callback 
+		/// functions, in order to load the target page content from the 
+		/// database only when needed.
+		/// @note It is absolutely necessary that the newTarget() 
+		/// function is called for every new page.
+		private function getTargetContent() {
+			if ( ! isset( $mTargetContent ) ) {
+				LinkTitles::$mTargetContent = WikiPage::factory(
+					LinkTitles::$mTargetTitle)->getContent();
+			};
+			return LinkTitles::$mTargetContent;
+		}
+
+		/// Examines the current target page. Returns true if it may be linked; 
+		/// false if not. This depends on the settings 
+		/// $wgLinkTitlesCheckRedirect and $wgLinkTitlesEnableNoTargetMagicWord 
+		/// and whether the target page is a redirect or contains the 
+		/// __NOAUTOLINKTARGET__ magic word.
+		/// @returns boolean
+		private function checkTargetPage() {
+			wfProfileIn( __METHOD__ );
+			global $wgLinkTitlesEnableNoTargetMagicWord;
+			global $wgLinkTitlesCheckRedirect;
+
+			// If checking for redirects is enabled and the target page does 
+			// indeed redirect to the current page, return the page title as-is 
+			// (unlinked).
+			if ( $wgLinkTitlesCheckRedirect ) {
+				$redirectTitle = LinkTitles::getTargetContent()->getUltimateRedirectTarget();
+				if ( $redirectTitle && $redirectTitle->equals(LinkTitles::$mCurrentTitle) ) {
+					wfProfileOut( __METHOD__ );
+					return false;
+				}
+			};
+
+			// If the magic word __NOAUTOLINKTARGET__ is enabled and the target 
+			// page does indeed contain this magic word, return the page title 
+			// as-is (unlinked).
+			if ( $wgLinkTitlesEnableNoTargetMagicWord ) {
+				if ( LinkTitles::getTargetContent()->matchMagicWord(
+						MagicWord::get('MAG_LINKTITLES_NOTARGET') ) ) {
+					wfProfileOut( __METHOD__ );
+					return false;
+				}
+			};
+			wfProfileOut( __METHOD__ );
+			return true;
+		}
 	}
 
 // vim: ts=2:sw=2:noet:comments^=\:///