Start to move wikilink expansion to a separate async token transformer.

https://mediawiki.org/wiki/Special:Code/MediaWiki/112685

modules/parser/ext.core.AttributeExpander.js

@@ -65,6 +65,7 @@ AttributeExpander.prototype.onToken = function ( token, frame, cb ) {
 			return { async: true };
 		}
 	} else {
+		token.rank = this.rank;
 		return { token: token };
 	}
 };
@@ -80,6 +81,7 @@ AttributeExpander.prototype._returnAttributes = function ( expandData,
 	// Remove the target from the attributes
 	expandData.token.attribs = attributes;
 	if ( expandData.async ) {
+		expandData.token.rank = this.rank;
 		expandData.cb( [expandData.token], false );
 	}
 };

modules/parser/ext.core.LinkHandler.js

@@ -0,0 +1,48 @@
+/**
+ * Simple link handler. Registers after template expansions, as an
+ * asynchronous transform.
+ *
+ * @author Gabriel Wicke <gwicke@wikimedia.org>
+ * 
+ * * Collect description/parameter tokens between 'a' tags
+ * * Extract image options and add image html if target is media/image
+ *   namespace
+ * * 
+ *
+ *
+ * TODO: keep round-trip information in meta tag or the like
+ *
+ *
+ *
+ * Pro/Contra of single token vs. tags and tokens
+ * - Need to collect description tokens between a and /a
+ * + noinclude etc handled automatically by having all tokens on content level
+ */
+
+function WikiLinkHandler( manager, isInclude ) {
+	this.manager = manager;
+	this.manager.addTransform( this.onWikiLink.bind( this ), this.rank, 'tag', 'wikilink' );
+}
+
+WikiLinkHandler.prototype.rank = 1.15; // after AttributeExpander
+
+WikiLinkHandler.prototype.onWikiLink = function ( token, manager, cb ) {
+	// Split off and normalize namespace
+	// Compare with image/media namespaces
+	//	handle image
+	//	handle 
+	// Check if page exists
+	// 
+	var obj = new TagTk( 'a', [ this.manager.env.lookupKV( token.attribs, 'href' ) ] );
+    obj.attribs.push( new KV('data-mw-type', 'internal') );
+	var out = [obj].concat( this.manager.env.lookupKV( token.attribs, 'content' ).v, 
+							new EndTagTk( 'a' ) );
+	//console.warn( JSON.stringify( out, null, 2 ) );
+	return { tokens: out };
+};
+
+
+
+if (typeof module == "object") {
+	module.exports.WikiLinkHandler = WikiLinkHandler;
+}

modules/parser/mediawiki.parser.js

@@ -26,6 +26,7 @@ var fs = require('fs'),
 	Sanitizer                   = require('./ext.core.Sanitizer.js').Sanitizer,
 	TemplateHandler             = require('./ext.core.TemplateHandler.js').TemplateHandler,
 	AttributeExpander            = require('./ext.core.AttributeExpander.js').AttributeExpander,
+	WikiLinkHandler             = require('./ext.core.LinkHandler.js').WikiLinkHandler,
 	Cite                        = require('./ext.Cite.js').Cite,
 	FauxHTML5                   = require('./mediawiki.HTML5TreeBuilder.node.js').FauxHTML5,
 	DOMPostProcessor            = require('./mediawiki.DOMPostProcessor.js').DOMPostProcessor,
@@ -165,7 +166,8 @@ ParserPipeline.prototype._transformers = {
 				// Expand attributes after templates to avoid expanding unused branches
 				// XXX: Should we support further processing after attribute
 				// expansion?
-				AttributeExpander
+				AttributeExpander,
+				WikiLinkHandler
 				/* ExtensionHandler1, */
 				/* ExtensionHandler2, */
 			],

modules/parser/pegTokenizer.pegjs.txt

@@ -746,12 +746,11 @@ wikilink
     // class. Can we work out a static negative class instead?
     // XXX: Exclude uppercase chars from non-latin languages too!
     trail:(! [A-Z \t(),.:-] tc:text_char { return tc })* {
-      var obj = new TagTk( 'a', 
-                [
-                    new KV('data-mw-type', 'internal')
-                ] ),
+      var obj = new SelfclosingTagTk( 'wikilink' ),
           textTokens = [];
       obj.attribs.push( new KV('href', target) );
+
+      // Deal with content. XXX: Properly support pipe-trick etc
       if (lcontent && lcontent.length) {
           textTokens = lcontent;
           if (trail) {
@@ -765,8 +764,10 @@ wikilink
               textTokens = $.extend(true, [], target);
           }
       }
+
+      obj.attribs.push( new KV( 'content', flatten( textTokens ) ) );
       //console.warn( "XXX:" + pp([obj].concat(textTokens, [new EndTagTk( 'a' )])) );
-      return [obj].concat(textTokens, [new EndTagTk( 'a' )]);
+      return [obj];
   }
 
 link_text