| 
							- /**
 -  * 
 -  * htmlParser改造自: https://github.com/blowsie/Pure-JavaScript-HTML5-Parser
 -  * 
 -  * author: Di (微信小程序开发工程师)
 -  * organization: WeAppDev(微信小程序开发论坛)(http://weappdev.com)
 -  *               垂直微信小程序开发交流社区
 -  * 
 -  * github地址: https://github.com/icindy/wxParse
 -  * 
 -  * for: 微信小程序富文本解析
 -  * detail : http://weappdev.com/t/wxparse-alpha0-1-html-markdown/184
 -  */
 - // Regular Expressions for parsing tags and attributes
 - var startTag = /^<([-A-Za-z0-9_]+)((?:\s+[a-zA-Z_:][-a-zA-Z0-9_:.]*(?:\s*=\s*(?:(?:"[^"]*")|(?:'[^']*')|[^>\s]+))?)*)\s*(\/?)>/,
 - 	endTag = /^<\/([-A-Za-z0-9_]+)[^>]*>/,
 - 	attr = /([a-zA-Z_:][-a-zA-Z0-9_:.]*)(?:\s*=\s*(?:(?:"((?:\\.|[^"])*)")|(?:'((?:\\.|[^'])*)')|([^>\s]+)))?/g;
 - 
 - // Empty Elements - HTML 5
 - var empty = makeMap("area,base,basefont,br,col,frame,hr,img,input,link,meta,param,embed,command,keygen,source,track,wbr");
 - 
 - // Block Elements - HTML 5
 - var block = makeMap("a,address,code,article,applet,aside,audio,blockquote,button,canvas,center,dd,del,dir,div,dl,dt,fieldset,figcaption,figure,footer,form,frameset,h1,h2,h3,h4,h5,h6,header,hgroup,hr,iframe,ins,isindex,li,map,menu,noframes,noscript,object,ol,output,p,pre,section,script,table,tbody,td,tfoot,th,thead,tr,ul,video");
 - 
 - // Inline Elements - HTML 5
 - var inline = makeMap("abbr,acronym,applet,b,basefont,bdo,big,br,button,cite,del,dfn,em,font,i,iframe,img,input,ins,kbd,label,map,object,q,s,samp,script,select,small,span,strike,strong,sub,sup,textarea,tt,u,var");
 - 
 - // Elements that you can, intentionally, leave open
 - // (and which close themselves)
 - var closeSelf = makeMap("colgroup,dd,dt,li,options,p,td,tfoot,th,thead,tr");
 - 
 - // Attributes that have their values filled in disabled="disabled"
 - var fillAttrs = makeMap("checked,compact,declare,defer,disabled,ismap,multiple,nohref,noresize,noshade,nowrap,readonly,selected");
 - 
 - // Special Elements (can contain anything)
 - var special = makeMap("wxxxcode-style,script,style,view,scroll-view,block");
 - 
 - function HTMLParser(html, handler) {
 - 	var index, chars, match, stack = [], last = html;
 - 	stack.last = function () {
 - 		return this[this.length - 1];
 - 	};
 - 
 - 	while (html) {
 - 		chars = true;
 - 
 - 		// Make sure we're not in a script or style element
 - 		if (!stack.last() || !special[stack.last()]) {
 - 
 - 			// Comment
 - 			if (html.indexOf("<!--") == 0) {
 - 				index = html.indexOf("-->");
 - 
 - 				if (index >= 0) {
 - 					if (handler.comment)
 - 						handler.comment(html.substring(4, index));
 - 					html = html.substring(index + 3);
 - 					chars = false;
 - 				}
 - 
 - 				// end tag
 - 			} else if (html.indexOf("</") == 0) {
 - 				match = html.match(endTag);
 - 
 - 				if (match) {
 - 					html = html.substring(match[0].length);
 - 					match[0].replace(endTag, parseEndTag);
 - 					chars = false;
 - 				}
 - 
 - 				// start tag
 - 			} else if (html.indexOf("<") == 0) {
 - 				match = html.match(startTag);
 - 
 - 				if (match) {
 - 					html = html.substring(match[0].length);
 - 					match[0].replace(startTag, parseStartTag);
 - 					chars = false;
 - 				}
 - 			}
 - 
 - 			if (chars) {
 - 				index = html.indexOf("<");
 - 				var text = ''
 - 				while (index === 0) {
 -                                   text += "<";
 -                                   html = html.substring(1);
 -                                   index = html.indexOf("<");
 - 				}
 - 				text += index < 0 ? html : html.substring(0, index);
 - 				html = index < 0 ? "" : html.substring(index);
 - 
 - 				if (handler.chars)
 - 					handler.chars(text);
 - 			}
 - 
 - 		} else {
 - 
 - 			html = html.replace(new RegExp("([\\s\\S]*?)<\/" + stack.last() + "[^>]*>"), function (all, text) {
 - 				text = text.replace(/<!--([\s\S]*?)-->|<!\[CDATA\[([\s\S]*?)]]>/g, "$1$2");
 - 				if (handler.chars)
 - 					handler.chars(text);
 - 
 - 				return "";
 - 			});
 - 
 - 
 - 			parseEndTag("", stack.last());
 - 		}
 - 
 - 		if (html == last)
 - 			throw "Parse Error: " + html;
 - 		last = html;
 - 	}
 - 
 - 	// Clean up any remaining tags
 - 	parseEndTag();
 - 
 - 	function parseStartTag(tag, tagName, rest, unary) {
 - 		tagName = tagName.toLowerCase();
 - 
 - 		if (block[tagName]) {
 - 			while (stack.last() && inline[stack.last()]) {
 - 				parseEndTag("", stack.last());
 - 			}
 - 		}
 - 
 - 		if (closeSelf[tagName] && stack.last() == tagName) {
 - 			parseEndTag("", tagName);
 - 		}
 - 
 - 		unary = empty[tagName] || !!unary;
 - 
 - 		if (!unary)
 - 			stack.push(tagName);
 - 
 - 		if (handler.start) {
 - 			var attrs = [];
 - 
 - 			rest.replace(attr, function (match, name) {
 - 				var value = arguments[2] ? arguments[2] :
 - 					arguments[3] ? arguments[3] :
 - 						arguments[4] ? arguments[4] :
 - 							fillAttrs[name] ? name : "";
 - 
 - 				attrs.push({
 - 					name: name,
 - 					value: value,
 - 					escaped: value.replace(/(^|[^\\])"/g, '$1\\\"') //"
 - 				});
 - 			});
 - 
 - 			if (handler.start) {
 - 				handler.start(tagName, attrs, unary);
 - 			}
 - 
 - 		}
 - 	}
 - 
 - 	function parseEndTag(tag, tagName) {
 - 		// If no tag name is provided, clean shop
 - 		if (!tagName)
 - 			var pos = 0;
 - 
 - 		// Find the closest opened tag of the same type
 - 		else {
 - 			tagName = tagName.toLowerCase();
 - 			for (var pos = stack.length - 1; pos >= 0; pos--)
 - 				if (stack[pos] == tagName)
 - 					break;
 - 		}
 - 		if (pos >= 0) {
 - 			// Close all the open elements, up the stack
 - 			for (var i = stack.length - 1; i >= pos; i--)
 - 				if (handler.end)
 - 					handler.end(stack[i]);
 - 
 - 			// Remove the open elements from the stack
 - 			stack.length = pos;
 - 		}
 - 	}
 - };
 - 
 - 
 - function makeMap(str) {
 - 	var obj = {}, items = str.split(",");
 - 	for (var i = 0; i < items.length; i++)
 - 		obj[items[i]] = true;
 - 	return obj;
 - }
 - 
 - module.exports = HTMLParser;
 
 
  |