Ext的DomQuery学习笔记

通过各种途径，得知Ext的选择器很不简单，最大的特点就是利用eval即时生成查询函数，让它在一些选择器类型中速度爆快。因此我觉得非常有必要学习一下Ext的这个模块了。

从最后一行得知，Ext.query方法是Ext.DomQuery.select的别名，那我们就顺着它的思路看呗。

select方法，我管它为入口函数。

select : function(path, root, type){
    if(!root || root == document){
        root = document;
    }
    if(typeof root == "string"){
        root = document.getElementById(root);
    }
    var paths = path.split(","),//把选择器按并联选择器分解
        results = [];
    for(var i = 0, len = paths.length; i < len; i++){
        var p = paths[i].replace(trimRe, "");//移除左右两边的空白节点
        if(!cache[p]){
            cache[p] = Ext.DomQuery.compile(p);//把刚编译出来的查询函数放进缓存体中
            if(!cache[p]){
                throw p + " is not a valid selector";
            }
        }
        var result = cache[p](root);//把文档对象传进去，获取目标元素
        if(result && result != document){//如果能获取元素或并不返回我们原来传入的那个文档对象，
            results = results.concat(result);//就把它并入最终结果中
        }
    }
    if(paths.length > 1){//去除重复元素
        return nodup(results);
    }
    return results;
},

compile方法，它用eval动态生成查询函数的做法确实让人一亮。

compile : function(path, type){//
    type = type || "select";
   //用于编译的代码
    var fn = ["var f = function(root){\n var mode; ++batch; var n = root || document;\n"],
        q = path, //选择器
        mode,
        lq,
        tk = Ext.DomQuery.matchers,
        tklen = tk.length,
        mm,
        //取出关系选择器的自符
        //modeRe = /^(\s?[\/>+~]\s?|\s|$)/,
        lmode = q.match(modeRe);
        //如 alert("> .aaa".match(/^(\s?[\/>+~]\s?|\s|$)/))
        //弹出 > ,>
    if(lmode && lmode[1]){//如果存在 / >  + ~ 这四个选择器，我们将对编译代码与选择器进行些操作
        //编译代码将增加，如 mode=">"的字段
        fn[fn.length] = 'mode="'+lmode[1].replace(trimRe, "")+'";';
        //选择器  > .aaa 将变成 .aaa
        q = q.replace(lmode[1], "");
    }
    //如果选择器被消耗到以断句符“/”开头，那么移除它，把第二行代入path
    //如 "\
    //     h1[title]"
    //的情形
    while(path.substr(0, 1)=="/"){
        path = path.substr(1);
    }
 
    while(q && lq != q){//如果选择器q不等于undefined或null
        lq = q;  //tagTokenRe = /^(#)?([\w-\*]+)/,
        var tm = q.match(tagTokenRe);// 判定其是ID选择器，标签选择器亦或通配符选择器
        if(type == "select"){
            if(tm){
                if(tm[1] == "#"){//如果是ID选择器，
                    fn[fn.length] = 'n = quickId(n, mode, root, "'+tm[2]+'");';
                }else{           //如果是标签选择器
                    fn[fn.length] = 'n = getNodes(n, mode, "'+tm[2]+'");';
                }
                q = q.replace(tm[0], "");
            }else if(q.substr(0, 1) != '@'){
                fn[fn.length] = 'n = getNodes(n, mode, "*");';
            }
        }else{
            if(tm){
                if(tm[1] == "#"){
                    fn[fn.length] = 'n = byId(n, null, "'+tm[2]+'");';
                }else{
                    fn[fn.length] = 'n = byTag(n, "'+tm[2]+'");';
                }
                q = q.replace(tm[0], "");
            }
        }
        while(!(mm = q.match(modeRe))){
            var matched = false;
            for(var j = 0; j < tklen; j++){
                var t = tk[j];
                var m = q.match(t.re);//用matchers里面的正则依次匹配选择器，
                if(m){ //如果通过则把matchers.select里面的{1},{2}这些东西替换为相应的字符
                    fn[fn.length] = t.select.replace(tplRe, function(x, i){
                                            return m[i];
                                        });
                    q = q.replace(m[0], "");//移除选择器相应的部分
                    matched = true;//中止循环
                    break;
                }
            }
            // prevent infinite loop on bad selector
            if(!matched){
                throw 'Error parsing selector, parsing failed at "' + q + '"';
            }
        }
        if(mm[1]){//添加编译代码，如 mode="~"的字段
            fn[fn.length] = 'mode="'+mm[1].replace(trimRe, "")+'";';
            q = q.replace(mm[1], "");//移除选择器相应的部分
        }
    }
    fn[fn.length] = "return nodup(n);\n}";//添加移除重复元素的编译代码
    eval(fn.join(""));//连结所有要编译的代码，用eval进行编译，于是当前作用域使增加一个叫f的函数
    return f;//返回f查询函数
},

f查询函数的生成依赖于一个叫做matchers的数组对象：

matchers : [{
        re: /^\.([\w-]+)/,
        select: 'n = byClassName(n, null, " {1} ");'
    }, {
        re: /^\:([\w-]+)(?:\(((?:[^\s>\/]*|.*?))\))?/,
        select: 'n = byPseudo(n, "{1}", "{2}");'
    },{
        re: /^(?:([\[\{])(?:@)?([\w-]+)\s?(?:(=|.=)\s?['"]?(.*?)["']?)?[\]\}])/,
        select: 'n = byAttribute(n, "{2}", "{4}", "{3}", "{1}");'
    }, {
        re: /^#([\w-]+)/,
        select: 'n = byId(n, null, "{1}");'
    },{
        re: /^@([\w-]+)/,
        select: 'return {firstChild:{nodeValue:attrValue(n, "{1}")}};'
    }
],

在处理ID选择器时，分为两个函数，分别为查找模式（type="select"）与过滤模式。个人觉得它好像不可能处理IE中的getElementById的bug。过滤模式用于反选选择器。

function quickId(ns, mode, root, id){
    if(ns == root){
       var d = root.ownerDocument || root;
       return d.getElementById(id);//IE下有bug
    }
    ns = getNodes(ns, mode, "*");
    return byId(ns, null, id);
}
function byId(cs, attr, id){
    if(cs.tagName || cs == document){
        cs = [cs];
    }
    if(!id){
        return cs;
    }
    var r = [], ri = -1;
    for(var i = 0,ci; ci = cs[i]; i++){
        if(ci && ci.id == id){//这里存在问题，因为IE下表单元素的id值不能为"id"，见我的博文《IE6的getElementById bug》
            r[++ri] = ci;
            return r;
        }
    }
    return r;
};

处理类选择器

function byClassName(c, a, v){//c为元素，v为className
      if(!v){
          return c;
      }
      var r = [], ri = -1, cn;
      for(var i = 0, ci; ci = c[i]; i++){
          if((' '+ci.className+' ').indexOf(v) != -1){
              r[++ri] = ci;
          }
      }
      return r;
  };

根据属性选择器筛选元素，不过在精确获取属性时，对于一些特殊属性无法辨识，具体可参见我的选择器query的属性转换列表。

function byAttribute(cs, attr, value, op, custom){
     var r = [], 
         ri = -1, 
         st = custom=="{",
         f = Ext.DomQuery.operators[op];
     for(var i = 0, ci; ci = cs[i]; i++){
         if(ci.nodeType != 1){
             continue;
         }
         var a;
         if(st){
             a = Ext.DomQuery.getStyle(ci, attr);
         }
         else if(attr == "class" || attr == "className"){
             a = ci.className;
         }else if(attr == "for"){
             a = ci.htmlFor;
         }else if(attr == "href"){
             a = ci.getAttribute("href", 2);
         }else{
             a = ci.getAttribute(attr);
         }
         if((f && f(a, value)) || (!f && a)){
             r[++ri] = ci;
         }
     }
     return r;
 };

getStyle方法不说了，它是调用style模块的。看看处理属性选择器的操作符，基本与jQuery的处理方式一下，不过Ext出现比较早，应该是抄它的。以前jQuery是用特慢的xpath模拟。

operators : {
    "=" : function(a, v){
        return a == v;
    },
    "!=" : function(a, v){
        return a != v;
    },
    "^=" : function(a, v){
        return a && a.substr(0, v.length) == v;
    },
    "$=" : function(a, v){
        return a && a.substr(a.length-v.length) == v;
    },
    "*=" : function(a, v){
        return a && a.indexOf(v) !== -1;
    },
    "%=" : function(a, v){
        return (a % v) == 0;
    },
    "|=" : function(a, v){
        return a && (a == v || a.substr(0, v.length+1) == v+'-');
    },
    "~=" : function(a, v){
        return a && (' '+a+' ').indexOf(' '+v+' ') != -1;
    }
},

看byPseudo方法，它只不过是个适配器，根据伪类的类型返回真正的处理函数。

function byPseudo(cs, name, value){
      return Ext.DomQuery.pseudos[name](cs, value);
  };

pseudos你可以管它做适配器对象，也可以称之为switch Object。嘛，叫什么都一样，它可以帮我们从无限的if...else if....else if 语句中解放出来。Ext运用的设计模式挺多的，这正是企业应用的特征之一，为以后添加新模块留下后路。

pseudos : {
    "first-child" : function(c){
        var r = [], ri = -1, n;
        for(var i = 0, ci; ci = n = c[i]; i++){//要求前面不能再有元素节点
            while((n = n.previousSibling) && n.nodeType != 1);
            if(!n){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "last-child" : function(c){
        var r = [], ri = -1, n;
        for(var i = 0, ci; ci = n = c[i]; i++){//要求其后不能再有元素节点
            while((n = n.nextSibling) && n.nodeType != 1);
            if(!n){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "nth-child" : function(c, a) {
        var r = [], ri = -1,
            m = nthRe.exec(a == "even" && "2n" || a == "odd" && "2n+1" || !nthRe2.test(a) && "n+" + a || a),
            f = (m[1] || 1) - 0, l = m[2] - 0;//和jQuery解析表达式的做法如出一辙
        for(var i = 0, n; n = c[i]; i++){
            var pn = n.parentNode;
            if (batch != pn._batch) {//在父节点上添加一个私有属性_batch，
                var j = 0;
                for(var cn = pn.firstChild; cn; cn = cn.nextSibling){
                    if(cn.nodeType == 1){
                       cn.nodeIndex = ++j;
                    }
                }
                pn._batch = batch;
            }
            if (f == 1) {//f就是an+b中的a，如果f为1时，那么只取出nodeIndex为b的元素节点即可
                if (l == 0 || n.nodeIndex == l){
                    r[++ri] = n;
                }
                //否则使用以下公式取元素(见实验2)
            } else if ((n.nodeIndex + l) % f == 0){
                r[++ri] = n;
            }
        }
 
        return r;
    },
 
    "only-child" : function(c){
        var r = [], ri = -1;;
        for(var i = 0, ci; ci = c[i]; i++){
            if(!prev(ci) && !next(ci)){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "empty" : function(c){
        var r = [], ri = -1;
        for(var i = 0, ci; ci = c[i]; i++){
            var cns = ci.childNodes, j = 0, cn, empty = true;
            while(cn = cns[j]){
                ++j;
                if(cn.nodeType == 1 || cn.nodeType == 3){
                    empty = false;
                    break;
                }
            }
            if(empty){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "contains" : function(c, v){
        var r = [], ri = -1;
        for(var i = 0, ci; ci = c[i]; i++){
            if((ci.textContent||ci.innerText||'').indexOf(v) != -1){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "nodeValue" : function(c, v){
        var r = [], ri = -1;
        for(var i = 0, ci; ci = c[i]; i++){
            if(ci.firstChild && ci.firstChild.nodeValue == v){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "checked" : function(c){
        var r = [], ri = -1;
        for(var i = 0, ci; ci = c[i]; i++){
            if(ci.checked == true){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "not" : function(c, ss){
        return Ext.DomQuery.filter(c, ss, true);
    },
 
    "any" : function(c, selectors){
        var ss = selectors.split('|'),
            r = [], ri = -1, s;
        for(var i = 0, ci; ci = c[i]; i++){
            for(var j = 0; s = ss[j]; j++){
                if(Ext.DomQuery.is(ci, s)){
                    r[++ri] = ci;
                    break;
                }
            }
        }
        return r;
    },
 
    "odd" : function(c){
        return this["nth-child"](c, "odd");
    },
 
    "even" : function(c){
        return this["nth-child"](c, "even");
    },
 
    "nth" : function(c, a){
        return c[a-1] || [];
    },
 
    "first" : function(c){
        return c[0] || [];
    },
 
    "last" : function(c){
        return c[c.length-1] || [];
    },
 
    "has" : function(c, ss){
        var s = Ext.DomQuery.select,
            r = [], ri = -1;
        for(var i = 0, ci; ci = c[i]; i++){
            if(s(ss, ci).length > 0){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "next" : function(c, ss){
        var is = Ext.DomQuery.is,
            r = [], ri = -1;
        for(var i = 0, ci; ci = c[i]; i++){
            var n = next(ci);
            if(n && is(n, ss)){
                r[++ri] = ci;
            }
        }
        return r;
    },
 
    "prev" : function(c, ss){
        var is = Ext.DomQuery.is,
            r = [], ri = -1;
        for(var i = 0, ci; ci = c[i]; i++){
            var n = prev(ci);
            if(n && is(n, ss)){
                r[++ri] = ci;
            }
        }
        return r;
    }
}

我们看一下"nth-child"模块，里面用到一个batch变量，它也动态生成的，还为元素添加两个私有属性_batch与nodeIndex。batch是从30803开始，这数字有什么深意吗？难道是Jack Slocum的银行卡密码？！看下面两个实验：

<!doctype html>
<html dir="ltr" lang="zh-CN">
  <head>
    <meta charset="utf-8"/>
    <title>Ext.query讲解 by 司徒正美</title>
    <style type="text/css">
    </style>
    <%= javascript_include_tag "ext-base"  %>
    <%= javascript_include_tag "ext-all"  %>
    <script type="text/javascript" charset="utf-8">
      window.onload = function(){
        alert(Ext.query("p span:nth-child(2n)"))
        var p = document.getElementsByTagName("p");
        alert(p[0]._batch)//30804
        alert(p[1]._batch)//30804
        alert(p[2]._batch)//30804
        alert(p[3]._batch)//30804
        alert(Ext.query("strong span:nth-child(1)"))
        var strong = document.getElementsByTagName("strong");
        alert(strong[0]._batch)//30805
      }
    </script>
  </head>
  <body>
    <p class="bbb">一<span class="aaa">二</span><span class="ccc">二</span><span class="bbb">二</span></p>
    <p class="aaa">一<span title="aaa">二</span></p>
    <p title="aaa">一<span title="aaa">二</span></p>
    <p id="na">一<span class="ccc">二</span><strong>Strong<span class="aaa">二</span></strong></p>
  </body>
</html>

<!doctype html>
<html dir="ltr" lang="zh-CN">
  <head>
    <meta charset="utf-8"/>
    <title>Ext.query讲解 by 司徒正美</title>
    <style type="text/css">
    </style>
    <%= javascript_include_tag "ext-base"  %>
    <%= javascript_include_tag "ext-all"  %>
    <script type="text/javascript" charset="utf-8">
      window.onload = function(){
        var r = Ext.query("p span:nth-child(2n)")
        alert(r[0].innerHTML)//span2
        alert(r[1].innerHTML)//span4
        alert(r[0].nodeIndex)//2
        alert(r[1].nodeIndex)//4
      }
    </script>
  </head>
  <body>
    <p class="bbb"><span class="aaa">span1</span>
      <span class="aaa">span2</span>
      <span class="ccc">span3</span>
      <span class="bbb">span4</span>
    </p>
    <p class="aaa">一<span title="aaa">二</span></p>
    <p title="aaa">一<span title="aaa">二</span></p>
    <p id="na">一<span class="ccc">二</span><strong>Strong<span class="aaa">二</span></strong></p>
  </body>
</html>

看反选选择器

"not" : function(c, ss){//c为上次搜索的结果集，ss为:not(***)中括号里面的内容
  return Ext.DomQuery.filter(c, ss, true);
},
filter : function(els, ss, nonMatches){
  ss = ss.replace(trimRe, "");//移除左右两边的空白节点
  if(!simpleCache[ss]){//如果缓存体不存在此选择器（Ext的缓存体蛮多的）
    simpleCache[ss] = Ext.DomQuery.compile(ss, "simple");//动态生成一个查询函数
  }
  var result = simpleCache[ss](els);//求取结果
  return nonMatches ? quickDiff(result, els) : result;//如果为true则调用quickDiff方法，否则直接返回结果集
},

看它如何进行取反，我以前也是用这种技术，就是利用了数学上全集与子集与补集的关系。quickDiff的第一个参数为子集，第二个参数为全集，既然是取反，当然取其补集。过程是在子集的元素节点中设置一个私有属性_diff，然后在全集范围的元素节点内找那些没有被标记，或标记不相同的元素，放进结果集。

function quickDiff(c1, c2){//子集，全集
    var len1 = c1.length,
        d = ++key,
        r = [];
    if(!len1){
        return c2;
    }
    if(isIE && typeof c1[0].selectSingleNode != "undefined"){
        return quickDiffIEXml(c1, c2);
    }        
    for(var i = 0; i < len1; i++){
        c1[i]._qdiff = d;//往子集元素设置一个私有属性_qdiff，起始数为30803
    }        
    for(var i = 0, len = c2.length; i < len; i++){
        if(c2[i]._qdiff != d){//然后在全集范围内找那些没有被标记，或标记不相同的元素，放进结果集
            r[r.length] = c2[i];
        }
    }
    return r;
}

不过对于IE的XML则利用setAttribute来标记私有属性，还要筛选后去除这私有属性。

function quickDiffIEXml(c1, c2){
    var d = ++key,
        r = [];
    for(var i = 0, len = c1.length; i < len; i++){
        c1[i].setAttribute("_qdiff", d);
    }        
    for(var i = 0, len = c2.length; i < len; i++){
        if(c2[i].getAttribute("_qdiff") != d){
            r[r.length] = c2[i];
        }
    }
    for(var i = 0, len = c1.length; i < len; i++){
       c1[i].removeAttribute("_qdiff");
    }
    return r;
}