Dojo 的css 2.0 lite 选择器源代码
Dojo 包含两个选择器 lite.js 和 acme.js, lite只是一个基础的选择器，包含id, className, tagName, 以及css2.0的属性选择器， 文件库相对于acme.js来说小很多，常用的选择方法都已经覆盖， 而acme.js 比较全面，支持全部的css3.0的要求， 但文件很大
Dojo中还支持 sizzle，可以通过 selectorEngine 来自定义选择器， 关于选择器的使用可以参考 http://dojotoolkit.org/reference-guide/1.10/dojo/query.html
以下是对 lite.js 源码注释
/*
选择器分为两部分:
拥有 querySelectorAll的浏览器
正则表达式判断出 #id, .class, input等，不是 css格式的选择符
分别调用 getElementById, getElementsByClassName, getElementsByTagName
对于没有getElementsByClassName的浏览器IE8, 则调用querySelectorAll
没有符合正则表达式， css 的选择器字符串 "div#id .a"
直接调用 querySelectorAll 函数
没有 querySelectorAll的浏览器 IE6, IE7
如果查询字符串中包含逗号分隔 "div, input"
调用combine方法, 分隔成单独的查询字符串, 在执行以下的步骤
正则表达式会匹配 div#clock .aa i[id='test'],div,clock,.aa i[id='test'],,,
先查找 #clock元素，然后设用 .aa i[id='test'], #clock, 变为 .aa i[id='test'],,,,.aa ,i,[id='test']
正则表达式匹配 .aa i[id='test']
先在root里找getElementsByTagName('i'), 构建新的selector， .aa [id='test']
调用 jsMatchesSelector, 先匹配getElementsByTagName中的元素 其属性是否为[id='test'], 然后他的父元素是否有 .aa类
其它
构造 match 类方法
调整 querySelectorAll 方法
*/
define(["../has", "../_base/kernel"], function(has, dojo){
"use strict";
//test
var testDiv = document.createElement("div");
var matchesSelector = testDiv.matches || testDiv.webkitMatchesSelector || testDiv.mozMatchesSelector || testDiv.msMatchesSelector || testDiv.oMatchesSelector; //IE9以下才支持，testDiv.matches('div') 如果每个元素跟css选择器相同，返回ture, 否则返回false
var querySelectorAll = testDiv.querySelectorAll;
var unionSplit = /([^\s,](?:"(?:\\.|[^"])+"|'(?:\\.|[^'])+'|[^,])*)/g;
has.add("dom-matches-selector", !!matchesSelector);
has.add("dom-qsa", !!querySelectorAll);
// this is a simple query engine. It has handles basic selectors, and for simple
// common selectors is extremely fast
var liteEngine = function(selector, root){
// summary:
// A small lightweight query selector engine that implements CSS2.1 selectors
// minus pseudo-classes and the sibling combinator, plus CSS3 attribute selectors
// 当 query('div, p") 以及 IE7及以下，才调用combine方法
if(combine && selector.indexOf(',') > -1){
return combine(selector, root);
}
// use the root's ownerDocument if provided, otherwise try to use dojo.doc. Note
// that we don't use dojo/_base/window's doc to reduce dependencies, and
// fallback to plain document if dojo.doc hasn't been defined (by dojo/_base/window).
// presumably we will have a better way to do this in 2.0
/*
如果浏览器提供了node.ownerDocument属性（IE6.0引进，其它未知), doc为拥有这个元素的document对像.
*/
var doc = root ? root.ownerDocument || root : dojo.doc || document,
/*
针对现在浏览器：可以匹配div#id, div.class, input, 但对于 span i等， match 返回空
div#id
返回["div.id", "div", "id"]
div.class
返回["div.class", "div", undefined", ".", "class", undefined]
query(".red")
返回 [".red", undefined, undefined, ".", "red", undefined]
query("div#id .red") 查找div#id 的 .red子元素
querySelectorAl ? null : div#clock1 .red,div,clock1,.red,,,
数组的格式是 ['匹配到的字符串', 'div', 'id', '.', 'className', 'element']
*/
match = (querySelectorAll ?
/^([\w]*)#([\w\-]+$)|^(\.)([\w\-\*]+$)|^(\w+$)/ : // 简单的查询， div#id, class， 元素input
/*
每一个匹配部分， 都会用于查询的手动过滤
^([\w]*)#([\w\-]+)(?:\s+(.*))?$ 用于匹配 "div#id .red" 或者 "div#id input"
(?:^|(>|.+\s+))([\w\-\*]+)(\S*$) div > ipnut > a:visit result: ["div > ipnut > a:visited", undefined, undefined, undefined, "div > ipnut > ", "a", ":visited"]
返回的数组格式是
['匹配到的字符串', 'div', 'id', '.class elementName', 'div>input>' ,'a', ":visited']
*/
/^([\w]*)#([\w\-]+)(?:\s+(.*))?$|(?:^|(>|.+\s+))([\w\-\*]+)(\S*$)/) // this one matches parts of the query that we can use to speed up manual filtering
.exec(selector);
root = root || doc;
if(match){
// fast path regardless of whether or not querySelectorAll exists
if(match[2]){
// an #id
// use dojo.byId if available as it fixes the id retrieval in IE, note that we can't use the dojo namespace in 2.0, but if there is a conditional module use, we will use that
// IE 的getElementById bug是， name 和 id不分
var found = dojo.byId ? dojo.byId(match[2], doc) : doc.getElementById(match[2]);
if(!found || (match[1] && match[1] != found.tagName.toLowerCase())){
// if there is a tag qualifer and it doesn't match, no matches
return [];
}
if(root != doc){
// there is a root element, make sure we are a child of it 如果指定了root， 那么确保root是这个元素的父元素
var parent = found;
while(parent != root){
parent = parent.parentNode;
if(!parent){
return [];
}
}
}
// 对于没有querySelectAll的浏览器， “div#id .red”， 现在只是找到了div#id元素， 还需要在次查找.red元素， 并在找查.red元素时，设置root为div#id;
return match[3] ?
liteEngine(match[3], found)
: [found];
}
if(match[3] && root.getElementsByClassName){
// a .class IE8, IE7 不支持
return root.getElementsByClassName(match[4]);
}
var found;
if(match[5]){
// a tag 现代化浏览器，直接根据tagName查找，并返回
found = root.getElementsByTagName(match[5]);
// IE7及以下的浏览器，选查找到元素，并构造新的选择器字符串
/*
#clock\'1 .aa i[id='test'],,,,#clock\'1 .aa ,i,[id='test']
#clock\'1 .aa [id='test']
*/
if(match[4] || match[6]){
selector = (match[4] || "") + match[6];
}else{
// that was the entirety of the query, return results
return found;
}
}
}
if(querySelectorAll){
// qSA works strangely on Element-rooted queries
// We can work around this by specifying an extra ID on the root
// and working up from there (Thanks to Andrew Dupont for the technique)
// IE 8 doesn't work on object elements 不支持在<object> 标签调用querySelectorAll <object width="400" height="400" data="helloworld.swf"></object>
if (root.nodeType === 1 && root.nodeName.toLowerCase() !== "object"){
return useRoot(root, selector, root.querySelectorAll);
}else{
// we can use the native qSA
// 应用于document
// IE8 不能应用此方法 object elements
return root.querySelectorAll(selector);
}
}else if(!found){ // 对于没有querySelectorAll的浏览器
// search all children and then filter
found = root.getElementsByTagName("*");
}
// now we filter the nodes that were found using the matchesSelector
var results = [];
for(var i = 0, l = found.length; i < l; i++){
var node = found[i];
if(node.nodeType == 1 && jsMatchesSelector(node, selector, root)){
// keep the nodes that match the selector
results.push(node);
}
}
return results;
};
var useRoot = function(context, query, method){
// this function creates a temporary id so we can do rooted qSA queries, this is taken from sizzle
var oldContext = context,
old = context.getAttribute("id"),
nid = old || "__dojo__",
hasParent = context.parentNode,
relativeHierarchySelector = /^\s*[+~]/.test(query); // query('+input', #form1); 相对于context的元素
/*
prev + next, 返回找到的next元素，next是prev的后一个兄弟元素, prev ~ siblings， 返回prev之后的兄弟元素
<label>
<input>
<span>
<input>
label + input 返回 第一个input
label ~ input 返回 两个input, span不返回
*/
if(relativeHierarchySelector && !hasParent){
//这个表达式应该没有用，因为传递给useRoot不可能为Document对像，只可能是Element， 而Element对像都是有parentNode
return [];
}
if(!old){
context.setAttribute("id", nid);
}else{
nid = nid.replace(/'/g, "\\$&"); // $&引用与 regexp 相匹配的子串， $` '左侧的文本 $' '右侧的文本
}
if(relativeHierarchySelector && hasParent){
context = context.parentNode;
}
//对query中的每个元素定义id， query('+ span, ~span", context);
var selectors = query.match(unionSplit);
for(var i = 0; i < selectors.length; i++){
selectors[i] = "[id='" + nid + "'] " + selectors[i];
}
/*
<div id="tes't"><span><i></i></span></div>
那么： query('span i', document.getElementById("test'1")) 会返回 [id='tes\'t'] span i
在div.parentNode上调用 querySelectorAll("[id='tes\'t'] span i") 查找对应的元素
如果你在元素上直接调用 div.querySelectorAll('body div span i") 这个也是会返回 i 元素，因为querySelectorAll是从整个文档查询 div span i， 然后在过滤出 div#test't内的 span i.
useRoot 会让 div.querySelectAll("body div span i") 返回 [], 相等于查询 [id='tes\'t'] body div span i
*/
query = selectors.join(",");
try{
return method.call(context, query);
}finally{
if(!old){
oldContext.removeAttribute("id");
}
}
};
if(!has("dom-matches-selector")){
var jsMatchesSelector = (function(){
// a JS implementation of CSS selector matching, first we start with the various handlers
var caseFix = testDiv.tagName == "div" ? "toLowerCase" : "toUpperCase"; //xml 或者 xhtml 会保留原始的大小写， 而html中，都是大小的
var selectorTypes = {
"": function(tagName){
tagName = tagName[caseFix]();
return function(node){
return node.tagName == tagName;
};
},
".": function(className){
var classNameSpaced = ' ' + className + ' '; //如果要查询，.aa, 而有一个元素的类名是 aaSuffix, 对过添加左右空格，就能排除这种情况
return function(node){
return node.className.indexOf(className) > -1 && (' ' + node.className + ' ').indexOf(classNameSpaced) > -1;
};
},
"#": function(id){
return function(node){
return node.id == id;
};
}
};
var attrComparators = {
"^=": function(attrValue, value){
return attrValue.indexOf(value) == 0;
},
"*=": function(attrValue, value){
return attrValue.indexOf(value) > -1;
},
"$=": function(attrValue, value){
return attrValue.substring(attrValue.length - value.length, attrValue.length) == value;
},
"~=": function(attrValue, value){
return (' ' + attrValue + ' ').indexOf(' ' + value + ' ') > -1;
},
"|=": function(attrValue, value){
return (attrValue + '-').indexOf(value + '-') == 0;
},
"=": function(attrValue, value){
return attrValue == value;
},
"": function(attrValue, value){
return true;
}
};
function attr(name, value, type){
var firstChar = value.charAt(0);
if(firstChar == '"' || firstChar == "'"){
// it is quoted, remove the quotes
value = value.slice(1, -1);
}
value = value.replace(/\\/g,'');
var comparator = attrComparators[type || ""];
return function(node){
var attrValue = node.getAttribute(name);
return attrValue && comparator(attrValue, value);
};
}
function ancestor(matcher){
return function(node, root){
while((node = node.parentNode) != root){
if(matcher(node, root)){
return true;
}
}
};
}
function parent(matcher){
return function(node, root){
node = node.parentNode;
return matcher ?
node != root && matcher(node, root)
: node == root;
};
}
var cache = {};
function and(matcher, next){
return matcher ?
function(node, root){
return next(node) && matcher(node, root);
}
: next;
}
return function(node, selector, root){
// this returns true or false based on if the node matches the selector (optionally within the given root)
var matcher = cache[selector]; // check to see if we have created a matcher function for the given selector
if(!matcher){
// create a matcher function for the given selector
// parse the selectors
/*
正则表达式翻译
1. 检测字符串中的 >或者空格，并记录到 &1
2. 检测字符串中是否有 #或者. ，并记录到 &2， 并记录它们后面的字符 &3 (转义字符，eg. id="first.last"， 查找时#first\\.last" , 其它可包含的字符为 [\w-]
3. 检测字符串中是否有 [], 并记录 attrName 到 &4, attrType(*=, |=, =, ^=)到&5, 对于 attrValue有三种类型 "aa", aa, aa ([^\]]) 检测非"]" 结束的字符
*/
/*
e.g: query("#clock\\'1 .aa i[id='test']")
match: #clock\'1 .aa i[id='test'],,,,#clock\'1 .aa ,i,[id='test']
selector: #clock\'1 .aa [id='test']
执行 selector.replace
1st matcher = function(node){
return node.id == id;
};
2st
matcher = ancestor(matcher) // 返回一个函数 function(node, root){}, 返回的这个函数会调用 1st中的函数
3st
matcher = function(className){
var classNameSpaced = ' ' + className + ' '; //如果要查询，.aa, 而有一个元素的类名是 aaSuffix, 对过添加左右空格，就能排除这种情况
return function(node){
return node.className.indexOf(className) > -1 && (' ' + node.className + ' ').indexOf(classNameSpaced) > -1;
};
} && matcher(node, root) // 调用第二步的matcher
4st 同2st
5st matcher = attr() && matcher （4st)
6 当调用 matcher时， 先调用执行 attr() -> matcher (4st） -> matcher(3st) -> matcher(2st) -> matcher(1st)
*/
if(selector.replace(/(?:\s*([> ])\s*)|(#|\.)?((?:\\.|[\w-])+)|\[\s*([\w-]+)\s*(.?=)?\s*("(?:\\.|[^"])+"|'(?:\\.|[^'])+'|(?:\\.|[^\]])*)\s*\]/g, function(t, combinator, type, value, attrName, attrType, attrValue){
if(value){
matcher = and(matcher, selectorTypes[type || ""](value.replace(/\\/g, '')));
}
else if(combinator){
matcher = (combinator == " " ? ancestor : parent)(matcher);
}
else if(attrName){
matcher = and(matcher, attr(attrName, attrValue, attrType));
}
return "";
})){
throw new Error("Syntax error in query");
}
if(!matcher){
return true;
}
cache[selector] = matcher;
}
// now run the matcher function on the node
return matcher(node, root);
};
})();
}
/*
IE7及以下的浏览器，没有定义querySelectorAll方法
lite 支持的css2 选择器用法 http://dojotoolkit.org/reference-guide/1.10/dojo/query.html
*/
if(!has("dom-qsa")){
var combine = function(selector, root){
// combined queries
//var unionSplit = /([^\s,](?:"(?:\\.|[^"])+"|'(?:\\.|[^'])+'|[^,])*)/g;
/*
1. "div, a" 或才 " div, a", [^\s,]以非空格，逗号开头, |[^,]* 当遇到逗号时，停止这次匹配
2. \\.|[^"] 用来匹配转义后的字符，如"aaa\\"bbbbccc" , 我要匹配整个字任串时，如果只用[^"]只能匹配到"aaa\\" ， 为了提升正则表达式的效率，可以改成以下方式
"(?: [^"\\]|\\.)+" 因为 正常字符的出现，要比转义出现的概率大， 所以把 [^"\]提前
3. /([^\s,]([^,])*)/g 但是必须要考虑双引号或者单引号之间有的逗号
4. 如果遇到双引号，就应当一次性的读取“”之间的内容, 内容的形式为 [^"] 和 \\. 转义字符, querySelectAll(".class:test") //错误
e.g:
query("div, a");
query("div[id='my\\:name']") <div id="my:name"></div>
query("[data-name='hello, world']")
*/
var selectors = selector.match(unionSplit);
var indexed = [];
// add all results and keep unique ones, this only runs in IE, so we take advantage
// of known IE features, particularly sourceIndex which is unique and allows us to
// order the results
for(var i = 0; i < selectors.length; i++){
selector = new String(selectors[i].replace(/\s*$/,''));
selector.indexOf = escape; // keep it from recursively entering combine
var results = liteEngine(selector, root);
for(var j = 0, l = results.length; j < l; j++){
var node = results[j];
indexed[node.sourceIndex] = node;
}
}
// now convert from a sparse array to a dense array 将一个稀疏数组变为稠密数组
var totalResults = [];
for(i in indexed){
totalResults.push(indexed[i]);
}
return totalResults;
};
}
liteEngine.match = matchesSelector ? function(node, selector, root){
if(root && root.nodeType != 9){
// doesn't support three args, use rooted id trick
return useRoot(root, selector, function(query){
return matchesSelector.call(node, query);
});
}
// we have a native matchesSelector, use that
return matchesSelector.call(node, selector);
} : jsMatchesSelector; // otherwise use the JS matches impl
return liteEngine;
});