SemanticScuttle/includes/js/dojo/_base/query.js

1191 lines
35 KiB
JavaScript

if(!dojo._hasResource["dojo._base.query"]){ //_hasResource checks added by build. Do not use _hasResource directly in your code.
dojo._hasResource["dojo._base.query"] = true;
dojo.provide("dojo._base.query");
dojo.require("dojo._base.NodeList");
/*
dojo.query() architectural overview:
dojo.query is a relatively full-featured CSS3 query library. It is
designed to take any valid CSS3 selector and return the nodes matching
the selector. To do this quickly, it processes queries in several
steps, applying caching where profitable.
The steps (roughly in reverse order of the way they appear in the code):
1.) check to see if we already have a "query dispatcher"
- if so, use that with the given parameterization. Skip to step 4.
2.) attempt to determine which branch to dispatch the query to:
- JS (optimized DOM iteration)
- xpath (for browsers that support it and where it's fast)
- native (not available in any browser yet)
3.) tokenize and convert to executable "query dispatcher"
- this is where the lion's share of the complexity in the
system lies. In the DOM version, the query dispatcher is
assembled as a chain of "yes/no" test functions pertaining to
a section of a simple query statement (".blah:nth-child(odd)"
but not "div div", which is 2 simple statements). Individual
statement dispatchers are cached (to prevent re-definition)
as are entire dispatch chains (to make re-execution of the
same query fast)
- in the xpath path, tokenization yeilds a concatenation of
parameterized xpath selectors. As with the DOM version, both
simple selector blocks and overall evaluators are cached to
prevent re-defintion
4.) the resulting query dispatcher is called in the passed scope (by default the top-level document)
- for DOM queries, this results in a recursive, top-down
evaluation of nodes based on each simple query section
- xpath queries can, thankfully, be executed in one shot
5.) matched nodes are pruned to ensure they are unique
*/
;(function(){
// define everything in a closure for compressability reasons. "d" is an
// alias to "dojo" since it's so frequently used. This seems a
// transformation that the build system could perform on a per-file basis.
////////////////////////////////////////////////////////////////////////
// Utility code
////////////////////////////////////////////////////////////////////////
var d = dojo;
var childNodesName = dojo.isIE ? "children" : "childNodes";
var caseSensitive = false;
var getQueryParts = function(query){
// summary: state machine for query tokenization
if(">~+".indexOf(query.charAt(query.length-1)) >= 0){
query += " *"
}
query += " "; // ensure that we terminate the state machine
var ts = function(s, e){
return d.trim(query.slice(s, e));
}
// the overall data graph of the full query, as represented by queryPart objects
var qparts = [];
// state keeping vars
var inBrackets = -1;
var inParens = -1;
var inMatchFor = -1;
var inPseudo = -1;
var inClass = -1;
var inId = -1;
var inTag = -1;
var lc = ""; // the last character
var cc = ""; // the current character
var pStart;
// iteration vars
var x = 0; // index in the query
var ql = query.length;
var currentPart = null; // data structure representing the entire clause
var _cp = null; // the current pseudo or attr matcher
var endTag = function(){
if(inTag >= 0){
var tv = (inTag == x) ? null : ts(inTag, x).toLowerCase();
currentPart[ (">~+".indexOf(tv) < 0) ? "tag" : "oper" ] = tv;
inTag = -1;
}
}
var endId = function(){
if(inId >= 0){
currentPart.id = ts(inId, x).replace(/\\/g, "");
inId = -1;
}
}
var endClass = function(){
if(inClass >= 0){
currentPart.classes.push(ts(inClass+1, x).replace(/\\/g, ""));
inClass = -1;
}
}
var endAll = function(){
endId(); endTag(); endClass();
}
for(; lc=cc, cc=query.charAt(x),x<ql; x++){
if(lc == "\\"){ continue; }
if(!currentPart){
// NOTE: I hate all this alloc, but it's shorter than writing tons of if's
pStart = x;
currentPart = {
query: null,
pseudos: [],
attrs: [],
classes: [],
tag: null,
oper: null,
id: null
};
inTag = x;
}
if(inBrackets >= 0){
// look for a the close first
if(cc == "]"){
if(!_cp.attr){
_cp.attr = ts(inBrackets+1, x);
}else{
_cp.matchFor = ts((inMatchFor||inBrackets+1), x);
}
var cmf = _cp.matchFor;
if(cmf){
if( (cmf.charAt(0) == '"') || (cmf.charAt(0) == "'") ){
_cp.matchFor = cmf.substring(1, cmf.length-1);
}
}
currentPart.attrs.push(_cp);
_cp = null; // necessaray?
inBrackets = inMatchFor = -1;
}else if(cc == "="){
var addToCc = ("|~^$*".indexOf(lc) >=0 ) ? lc : "";
_cp.type = addToCc+cc;
_cp.attr = ts(inBrackets+1, x-addToCc.length);
inMatchFor = x+1;
}
// now look for other clause parts
}else if(inParens >= 0){
if(cc == ")"){
if(inPseudo >= 0){
_cp.value = ts(inParens+1, x);
}
inPseudo = inParens = -1;
}
}else if(cc == "#"){
endAll();
inId = x+1;
}else if(cc == "."){
endAll();
inClass = x;
}else if(cc == ":"){
endAll();
inPseudo = x;
}else if(cc == "["){
endAll();
inBrackets = x;
_cp = {
/*=====
attr: null, type: null, matchFor: null
=====*/
};
}else if(cc == "("){
if(inPseudo >= 0){
_cp = {
name: ts(inPseudo+1, x),
value: null
}
currentPart.pseudos.push(_cp);
}
inParens = x;
}else if(cc == " " && lc != cc){
// note that we expect the string to be " " terminated
endAll();
if(inPseudo >= 0){
currentPart.pseudos.push({ name: ts(inPseudo+1, x) });
}
currentPart.hasLoops = (
currentPart.pseudos.length ||
currentPart.attrs.length ||
currentPart.classes.length );
currentPart.query = ts(pStart, x);
currentPart.tag = (currentPart["oper"]) ? null : (currentPart.tag || "*");
qparts.push(currentPart);
currentPart = null;
}
}
return qparts;
};
////////////////////////////////////////////////////////////////////////
// XPath query code
////////////////////////////////////////////////////////////////////////
// this array is a lookup used to generate an attribute matching function.
// There is a similar lookup/generator list for the DOM branch with similar
// calling semantics.
var xPathAttrs = {
"*=": function(attr, value){
return "[contains(@"+attr+", '"+ value +"')]";
},
"^=": function(attr, value){
return "[starts-with(@"+attr+", '"+ value +"')]";
},
"$=": function(attr, value){
return "[substring(@"+attr+", string-length(@"+attr+")-"+(value.length-1)+")='"+value+"']";
},
"~=": function(attr, value){
return "[contains(concat(' ',@"+attr+",' '), ' "+ value +" ')]";
},
"|=": function(attr, value){
return "[contains(concat(' ',@"+attr+",' '), ' "+ value +"-')]";
},
"=": function(attr, value){
return "[@"+attr+"='"+ value +"']";
}
};
// takes a list of attribute searches, the overall query, a function to
// generate a default matcher, and a closure-bound method for providing a
// matching function that generates whatever type of yes/no distinguisher
// the query method needs. The method is a bit tortured and hard to read
// because it needs to be used in both the XPath and DOM branches.
var handleAttrs = function( attrList,
query,
getDefault,
handleMatch){
d.forEach(query.attrs, function(attr){
var matcher;
// type, attr, matchFor
if(attr.type && attrList[attr.type]){
matcher = attrList[attr.type](attr.attr, attr.matchFor);
}else if(attr.attr.length){
matcher = getDefault(attr.attr);
}
if(matcher){ handleMatch(matcher); }
});
}
var buildPath = function(query){
var xpath = ".";
var qparts = getQueryParts(d.trim(query));
while(qparts.length){
var tqp = qparts.shift();
var prefix;
var postfix = "";
if(tqp.oper == ">"){
prefix = "/";
// prefix = "/child::*";
tqp = qparts.shift();
}else if(tqp.oper == "~"){
prefix = "/following-sibling::"; // get element following siblings
tqp = qparts.shift();
}else if(tqp.oper == "+"){
// FIXME:
// fails when selecting subsequent siblings by node type
// because the position() checks the position in the list
// of matching elements and not the localized siblings
prefix = "/following-sibling::";
postfix = "[position()=1]";
tqp = qparts.shift();
}else{
prefix = "//";
// prefix = "/descendant::*"
}
// get the tag name (if any)
xpath += prefix + tqp.tag + postfix;
// check to see if it's got an id. Needs to come first in xpath.
if(tqp.id){
xpath += "[@id='"+tqp.id+"'][1]";
}
d.forEach(tqp.classes, function(cn){
var cnl = cn.length;
var padding = " ";
if(cn.charAt(cnl-1) == "*"){
padding = ""; cn = cn.substr(0, cnl-1);
}
xpath +=
"[contains(concat(' ',@class,' '), ' "+
cn + padding + "')]";
});
handleAttrs(xPathAttrs, tqp,
function(condition){
return "[@"+condition+"]";
},
function(matcher){
xpath += matcher;
}
);
// FIXME: need to implement pseudo-class checks!!
};
return xpath;
};
var _xpathFuncCache = {};
var getXPathFunc = function(path){
if(_xpathFuncCache[path]){
return _xpathFuncCache[path];
}
var doc = d.doc;
// don't need to memoize. The closure scope handles it for us.
var xpath = buildPath(path);
var tf = function(parent){
// XPath query strings are memoized.
var ret = [];
var xpathResult;
try{
xpathResult = doc.evaluate(xpath, parent, null,
// XPathResult.UNORDERED_NODE_ITERATOR_TYPE, null);
XPathResult.ANY_TYPE, null);
}catch(e){
console.debug("failure in exprssion:", xpath, "under:", parent);
console.debug(e);
}
var result = xpathResult.iterateNext();
while(result){
ret.push(result);
result = xpathResult.iterateNext();
}
return ret;
}
return _xpathFuncCache[path] = tf;
};
/*
d.xPathMatch = function(query){
// XPath based DOM query system. Handles a small subset of CSS
// selectors, subset is identical to the non-XPath version of this
// function.
return getXPathFunc(query)();
}
*/
////////////////////////////////////////////////////////////////////////
// DOM query code
////////////////////////////////////////////////////////////////////////
var _filtersCache = {};
var _simpleFiltersCache = {};
// the basic building block of the yes/no chaining system. agree(f1, f2)
// generates a new function which returns the boolean results of both of
// the passed functions to a single logical-anded result.
var agree = function(first, second){
if(!first){ return second; }
if(!second){ return first; }
return function(){
return first.apply(window, arguments) && second.apply(window, arguments);
}
}
var _childElements = function(root){
var ret = [];
var te, x=0, tret = root[childNodesName];
while(te=tret[x++]){
if(te.nodeType == 1){ ret.push(te); }
}
return ret;
}
var _nextSiblings = function(root, single){
var ret = [];
var te = root;
while(te = te.nextSibling){
if(te.nodeType == 1){
ret.push(te);
if(single){ break; }
}
}
return ret;
}
var _filterDown = function(element, queryParts, matchArr, idx){
// NOTE:
// in the fast path! this function is called recursively and for
// every run of a query.
var nidx = idx+1;
var isFinal = (queryParts.length == nidx);
var tqp = queryParts[idx];
// see if we can constrain our next level to direct children
if(tqp.oper){
var ecn = (tqp.oper == ">") ?
_childElements(element) :
_nextSiblings(element, (tqp.oper == "+"));
if(!ecn || !ecn.length){
return;
}
nidx++;
isFinal = (queryParts.length == nidx);
// kinda janky, too much array alloc
var tf = getFilterFunc(queryParts[idx+1]);
// for(var x=ecn.length-1, te; x>=0, te=ecn[x]; x--){
for(var x=0, ecnl=ecn.length, te; x<ecnl, te=ecn[x]; x++){
if(tf(te)){
if(isFinal){
matchArr.push(te);
}else{
_filterDown(te, queryParts, matchArr, nidx);
}
}
/*
if(x==0){
break;
}
*/
}
}
// otherwise, keep going down, unless we'er at the end
var candidates = getElementsFunc(tqp)(element);
if(isFinal){
while(candidates.length){
matchArr.push(candidates.shift());
}
/*
candidates.unshift(0, matchArr.length-1);
matchArr.splice.apply(matchArr, candidates);
*/
}else{
// if we're not yet at the bottom, keep going!
while(candidates.length){
_filterDown(candidates.shift(), queryParts, matchArr, nidx);
}
}
}
var filterDown = function(elements, queryParts){
var ret = [];
// for every root, get the elements that match the descendant selector
// for(var x=elements.length-1, te; x>=0, te=elements[x]; x--){
var x = elements.length - 1, te;
while(te = elements[x--]){
_filterDown(te, queryParts, ret, 0);
}
return ret;
}
var getFilterFunc = function(q){
// note: query can't have spaces!
if(_filtersCache[q.query]){
return _filtersCache[q.query];
}
var ff = null;
// does it have a tagName component?
if(q.tag){
if(q.tag == "*"){
ff = agree(ff,
function(elem){
return (elem.nodeType == 1);
}
);
}else{
// tag name match
ff = agree(ff,
function(elem){
return (
(elem.nodeType == 1) &&
(q.tag == elem.tagName.toLowerCase())
);
// return isTn;
}
);
}
}
// does the node have an ID?
if(q.id){
ff = agree(ff,
function(elem){
return (
(elem.nodeType == 1) &&
(elem.id == q.id)
);
}
);
}
if(q.hasLoops){
// if we have other query param parts, make sure we add them to the
// filter chain
ff = agree(ff, getSimpleFilterFunc(q));
}
return _filtersCache[q.query] = ff;
}
var getNodeIndex = function(node){
// NOTE:
// we could have a more accurate caching mechanism by invalidating
// caches after the query has finished, but I think that'd lead to
// significantly more cache churn than the cache would provide
// value for in the common case. Generally, we're more
// conservative (and therefore, more accurate) than jQuery and
// DomQuery WRT node node indexes, but there may be corner cases
// in which we fall down. How much we care about them is TBD.
var pn = node.parentNode;
var pnc = pn.childNodes;
// check to see if we can trust the cache. If not, re-key the whole
// thing and return our node match from that.
var nidx = -1;
var child = pn.firstChild;
if(!child){
return nidx;
}
var ci = node["__cachedIndex"];
var cl = pn["__cachedLength"];
// only handle cache building if we've gone out of sync
if(((typeof cl == "number")&&(cl != pnc.length))||(typeof ci != "number")){
// rip though the whole set, building cache indexes as we go
pn["__cachedLength"] = pnc.length;
var idx = 1;
do{
// we only assign indexes for nodes with nodeType == 1, as per:
// http://www.w3.org/TR/css3-selectors/#nth-child-pseudo
// only elements are counted in the search order, and they
// begin at 1 for the first child's index
if(child === node){
nidx = idx;
}
if(child.nodeType == 1){
child["__cachedIndex"] = idx;
idx++;
}
child = child.nextSibling;
}while(child);
}else{
// NOTE:
// could be incorrect in some cases (node swaps involving the
// passed node, etc.), but we ignore those due to the relative
// unlikelihood of that occuring
nidx = ci;
}
return nidx;
}
var firedCount = 0;
var blank = "";
var _getAttr = function(elem, attr){
if(attr == "class"){
return elem.className || blank;
}
if(attr == "for"){
return elem.htmlFor || blank;
}
return elem.getAttribute(attr, 2) || blank;
}
var attrs = {
"*=": function(attr, value){
return function(elem){
// E[foo*="bar"]
// an E element whose "foo" attribute value contains
// the substring "bar"
return (_getAttr(elem, attr).indexOf(value)>=0);
}
},
"^=": function(attr, value){
// E[foo^="bar"]
// an E element whose "foo" attribute value begins exactly
// with the string "bar"
return function(elem){
return (_getAttr(elem, attr).indexOf(value)==0);
}
},
"$=": function(attr, value){
// E[foo$="bar"]
// an E element whose "foo" attribute value ends exactly
// with the string "bar"
var tval = " "+value;
return function(elem){
var ea = " "+_getAttr(elem, attr);
return (ea.lastIndexOf(value)==(ea.length-value.length));
}
},
"~=": function(attr, value){
// E[foo~="bar"]
// an E element whose "foo" attribute value is a list of
// space-separated values, one of which is exactly equal
// to "bar"
// return "[contains(concat(' ',@"+attr+",' '), ' "+ value +" ')]";
var tval = " "+value+" ";
return function(elem){
var ea = " "+_getAttr(elem, attr)+" ";
return (ea.indexOf(tval)>=0);
}
},
"|=": function(attr, value){
// E[hreflang|="en"]
// an E element whose "hreflang" attribute has a
// hyphen-separated list of values beginning (from the
// left) with "en"
var valueDash = " "+value+"-";
return function(elem){
var ea = " "+(elem.getAttribute(attr, 2) || "");
return (
(ea == value) ||
(ea.indexOf(valueDash)==0)
);
}
},
"=": function(attr, value){
return function(elem){
return (_getAttr(elem, attr) == value);
}
}
};
var pseudos = {
"first-child": function(name, condition){
return function(elem){
if(elem.nodeType != 1){ return false; }
// check to see if any of the previous siblings are elements
var fc = elem.previousSibling;
while(fc && (fc.nodeType != 1)){
fc = fc.previousSibling;
}
return (!fc);
}
},
"last-child": function(name, condition){
return function(elem){
if(elem.nodeType != 1){ return false; }
// check to see if any of the next siblings are elements
var nc = elem.nextSibling;
while(nc && (nc.nodeType != 1)){
nc = nc.nextSibling;
}
return (!nc);
}
},
"empty": function(name, condition){
return function(elem){
// DomQuery and jQuery get this wrong, oddly enough.
// The CSS 3 selectors spec is pretty explicit about
// it, too.
var cn = elem.childNodes;
var cnl = elem.childNodes.length;
// if(!cnl){ return true; }
for(var x=cnl-1; x >= 0; x--){
var nt = cn[x].nodeType;
if((nt == 1)||(nt == 3)){ return false; }
}
return true;
}
},
"contains": function(name, condition){
return function(elem){
// FIXME: I dislike this version of "contains", as
// whimsical attribute could set it off. An inner-text
// based version might be more accurate, but since
// jQuery and DomQuery also potentially get this wrong,
// I'm leaving it for now.
return (elem.innerHTML.indexOf(condition) >= 0);
}
},
"not": function(name, condition){
var ntf = getFilterFunc(getQueryParts(condition)[0]);
return function(elem){
return (!ntf(elem));
}
},
"nth-child": function(name, condition){
var pi = parseInt;
if(condition == "odd"){
return function(elem){
return (
((getNodeIndex(elem)) % 2) == 1
);
}
}else if((condition == "2n")||
(condition == "even")){
return function(elem){
return ((getNodeIndex(elem) % 2) == 0);
}
}else if(condition.indexOf("0n+") == 0){
var ncount = pi(condition.substr(3));
return function(elem){
return (elem.parentNode[childNodesName][ncount-1] === elem);
}
}else if( (condition.indexOf("n+") > 0) &&
(condition.length > 3) ){
var tparts = condition.split("n+", 2);
var pred = pi(tparts[0]);
var idx = pi(tparts[1]);
return function(elem){
return ((getNodeIndex(elem) % pred) == idx);
}
}else if(condition.indexOf("n") == -1){
var ncount = pi(condition);
return function(elem){
return (getNodeIndex(elem) == ncount);
}
}
}
};
var defaultGetter = (d.isIE) ? function(cond){
var clc = cond.toLowerCase();
return function(elem){
return elem[cond]||elem[clc];
}
} : function(cond){
return function(elem){
return (elem && elem.getAttribute && elem.hasAttribute(cond));
}
};
var getSimpleFilterFunc = function(query){
var fcHit = (_simpleFiltersCache[query.query]||_filtersCache[query.query]);
if(fcHit){ return fcHit; }
var ff = null;
// the only case where we'll need the tag name is if we came from an ID query
if(query.id){ // do we have an ID component?
if(query.tag != "*"){
ff = agree(ff, function(elem){
return (elem.tagName.toLowerCase() == query.tag);
});
}
}
// if there's a class in our query, generate a match function for it
d.forEach(query.classes, function(cname, idx, arr){
// get the class name
var isWildcard = cname.charAt(cname.length-1) == "*";
if(isWildcard){
cname = cname.substr(0, cname.length-1);
}
// I dislike the regex thing, even if memozied in a cache, but it's VERY short
var re = new RegExp("(?:^|\\s)" + cname + (isWildcard ? ".*" : "") + "(?:\\s|$)");
ff = agree(ff, function(elem){
return re.test(elem.className);
});
ff.count = idx;
});
d.forEach(query.pseudos, function(pseudo){
if(pseudos[pseudo.name]){
ff = agree(ff, pseudos[pseudo.name](pseudo.name, pseudo.value));
}
});
handleAttrs(attrs, query, defaultGetter,
function(tmatcher){ ff = agree(ff, tmatcher); }
);
if(!ff){
ff = function(){ return true; };
}
return _simpleFiltersCache[query.query] = ff;
}
var _getElementsFuncCache = { };
var getElementsFunc = function(query, root){
var fHit = _getElementsFuncCache[query.query];
if(fHit){ return fHit; }
// NOTE: this function is in the fast path! not memoized!!!
// the query doesn't contain any spaces, so there's only so many
// things it could be
if(query.id && !query.hasLoops && !query.tag){
// ID-only query. Easy.
return _getElementsFuncCache[query.query] = function(root){
// FIXME: if root != document, check for parenting!
return [ d.byId(query.id) ];
}
}
var filterFunc = getSimpleFilterFunc(query);
var retFunc;
if(query.tag && query.id && !query.hasLoops){
// we got a filtered ID search (e.g., "h4#thinger")
retFunc = function(root){
var te = d.byId(query.id);
if(filterFunc(te)){
return [ te ];
}
}
}else{
var tret;
if(!query.hasLoops){
// it's just a plain-ol elements-by-tag-name query from the root
retFunc = function(root){
var ret = [];
var te, x=0, tret = root.getElementsByTagName(query.tag);
while(te=tret[x++]){
ret.push(te);
}
return ret;
}
}else{
retFunc = function(root){
var ret = [];
var te, x=0, tret = root.getElementsByTagName(query.tag);
while(te=tret[x++]){
if(filterFunc(te)){
ret.push(te);
}
}
return ret;
}
}
}
return _getElementsFuncCache[query.query] = retFunc;
}
var _partsCache = {};
////////////////////////////////////////////////////////////////////////
// the query runner
////////////////////////////////////////////////////////////////////////
// this is the second level of spliting, from full-length queries (e.g.,
// "div.foo .bar") into simple query expressions (e.g., ["div.foo",
// ".bar"])
var _queryFuncCache = {
"*": d.isIE ?
function(root){
return root.all;
} :
function(root){
return root.getElementsByTagName("*");
},
"~": _nextSiblings,
"+": function(root){ return _nextSiblings(root, true); },
">": _childElements
};
var getStepQueryFunc = function(query){
// if it's trivial, get a fast-path dispatcher
var qparts = getQueryParts(d.trim(query));
// if(query[query.length-1] == ">"){ query += " *"; }
if(qparts.length == 1){
var tt = getElementsFunc(qparts[0]);
tt.nozip = true;
return tt;
}
// otherwise, break it up and return a runner that iterates over the parts recursively
var sqf = function(root){
var localQueryParts = qparts.slice(0); // clone the src arr
var candidates;
if(localQueryParts[0].oper == ">"){ // FIXME: what if it's + or ~?
candidates = [ root ];
// root = document;
}else{
candidates = getElementsFunc(localQueryParts.shift())(root);
}
return filterDown(candidates, localQueryParts);
}
return sqf;
}
// a specialized method that implements our primoridal "query optimizer".
// This allows us to dispatch queries to the fastest subsystem we can get.
var _getQueryFunc = (
// NOTE:
// XPath on the Webkit nighlies is slower than it's DOM iteration
// for most test cases
// FIXME:
// we should try to capture some runtime speed data for each query
// function to determine on the fly if we should stick w/ the
// potentially optimized variant or if we should try something
// new.
(document["evaluate"] && !d.isSafari) ?
function(query){
// has xpath support that's faster than DOM
var qparts = query.split(" ");
// can we handle it?
if( (document["evaluate"])&&
(query.indexOf(":") == -1)&&
(query.indexOf("+") == -1) // skip direct sibling matches. See line ~344
){
// dojo.debug(query);
// should we handle it?
// kind of a lame heuristic, but it works
if(
// a "div div div" style query
((qparts.length > 2)&&(query.indexOf(">") == -1))||
// or something else with moderate complexity. kinda janky
(qparts.length > 3)||
(query.indexOf("[")>=0)||
// or if it's a ".thinger" query
((1 == qparts.length)&&(0 <= query.indexOf(".")))
){
// use get and cache a xpath runner for this selector
return getXPathFunc(query);
}
}
// fallthrough
return getStepQueryFunc(query);
} : getStepQueryFunc
);
// uncomment to disable XPath for testing and tuning the DOM path
// _getQueryFunc = getStepQueryFunc;
// FIXME: we've got problems w/ the NodeList query()/filter() functions if we go XPath for everything
// uncomment to disable DOM queries for testing and tuning XPath
// _getQueryFunc = getXPathFunc;
// this is the primary caching for full-query results. The query dispatcher
// functions are generated here and then pickled for hash lookup in the
// future
var getQueryFunc = function(query){
// return a cached version if one is available
var qcz = query.charAt(0);
if(d.doc["querySelectorAll"] &&
( (!d.isSafari) || (d.isSafari > 3.1) ) && // see #5832
// as per CSS 3, we can't currently start w/ combinator:
// http://www.w3.org/TR/css3-selectors/#w3cselgrammar
(">+~".indexOf(qcz) == -1)
){
return function(root){
var r = root.querySelectorAll(query);
r.nozip = true; // skip expensive duplication checks and just wrap in a NodeList
return r;
};
}
if(_queryFuncCache[query]){ return _queryFuncCache[query]; }
if(0 > query.indexOf(",")){
// if it's not a compound query (e.g., ".foo, .bar"), cache and return a dispatcher
return _queryFuncCache[query] = _getQueryFunc(query);
}else{
// if it's a complex query, break it up into it's constituent parts
// and return a dispatcher that will merge the parts when run
// var parts = query.split(", ");
var parts = query.split(/\s*,\s*/);
var tf = function(root){
var pindex = 0; // avoid array alloc for every invocation
var ret = [];
var tp;
while(tp = parts[pindex++]){
ret = ret.concat(_getQueryFunc(tp, tp.indexOf(" "))(root));
}
return ret;
}
// ...cache and return
return _queryFuncCache[query] = tf;
}
}
// FIXME:
// Dean's Base2 uses a system whereby queries themselves note if
// they'll need duplicate filtering. We need to get on that plan!!
// attempt to efficiently determine if an item in a list is a dupe,
// returning a list of "uniques", hopefully in doucment order
var _zipIdx = 0;
var _zip = function(arr){
if(arr && arr.nozip){ return d.NodeList._wrap(arr); }
var ret = new d.NodeList();
if(!arr){ return ret; }
if(arr[0]){
ret.push(arr[0]);
}
if(arr.length < 2){ return ret; }
_zipIdx++;
arr[0]["_zipIdx"] = _zipIdx;
for(var x=1, te; te = arr[x]; x++){
if(arr[x]["_zipIdx"] != _zipIdx){
ret.push(te);
}
te["_zipIdx"] = _zipIdx;
}
// FIXME: should we consider stripping these properties?
return ret;
}
// the main executor
d.query = function(/*String*/ query, /*String|DOMNode?*/ root){
// summary:
// Returns nodes which match the given CSS3 selector, searching the
// entire document by default but optionally taking a node to scope
// the search by. Returns an instance of dojo.NodeList.
// description:
// dojo.query() is the swiss army knife of DOM node manipulation in
// Dojo. Much like Prototype's "$$" (bling-bling) function or JQuery's
// "$" function, dojo.query provides robust, high-performance
// CSS-based node selector support with the option of scoping searches
// to a particular sub-tree of a document.
//
// Supported Selectors:
// --------------------
//
// dojo.query() supports a rich set of CSS3 selectors, including:
//
// * class selectors (e.g., `.foo`)
// * node type selectors like `span`
// * ` ` descendant selectors
// * `>` child element selectors
// * `#foo` style ID selectors
// * `*` universal selector
// * `~`, the immediately preceeded-by sibling selector
// * `+`, the preceeded-by sibling selector
// * attribute queries:
// | * `[foo]` attribute presence selector
// | * `[foo='bar']` attribute value exact match
// | * `[foo~='bar']` attribute value list item match
// | * `[foo^='bar']` attribute start match
// | * `[foo$='bar']` attribute end match
// | * `[foo*='bar']` attribute substring match
// * `:first-child`, `:last-child` positional selectors
// * `:empty` content emtpy selector
// * `:empty` content emtpy selector
// * `:nth-child(n)`, `:nth-child(2n+1)` style positional calculations
// * `:nth-child(even)`, `:nth-child(odd)` positional selectors
// * `:not(...)` negation pseudo selectors
//
// Any legal combination of these selectors will work with
// `dojo.query()`, including compound selectors ("," delimited).
// Very complex and useful searches can be constructed with this
// palette of selectors and when combined with functions for
// maniplation presented by dojo.NodeList, many types of DOM
// manipulation operations become very straightforward.
//
// Unsupported Selectors:
// ----------------------
//
// While dojo.query handles many CSS3 selectors, some fall outside of
// what's resaonable for a programmatic node querying engine to
// handle. Currently unsupported selectors include:
//
// * namespace-differentiated selectors of any form
// * all `::` pseduo-element selectors
// * certain pseduo-selectors which don't get a lot of day-to-day use:
// | * `:root`, `:lang()`, `:target`, `:focus`
// * all visual and state selectors:
// | * `:root`, `:active`, `:hover`, `:visisted`, `:link`,
// `:enabled`, `:disabled`, `:checked`
// * `:*-of-type` pseudo selectors
//
// dojo.query and XML Documents:
// -----------------------------
//
// `dojo.query` currently only supports searching XML documents
// whose tags and attributes are 100% lower-case. This is a known
// limitation and will [be addressed soon](http://trac.dojotoolkit.org/ticket/3866)
// Non-selector Queries:
// ---------------------
//
// If something other than a String is passed for the query,
// `dojo.query` will return a new `dojo.NodeList` constructed from
// that parameter alone and all further processing will stop. This
// means that if you have a reference to a node or NodeList, you
// can quickly construct a new NodeList from the original by
// calling `dojo.query(node)` or `dojo.query(list)`.
//
// query:
// The CSS3 expression to match against. For details on the syntax of
// CSS3 selectors, see <http://www.w3.org/TR/css3-selectors/#selectors>
// root:
// A DOMNode (or node id) to scope the search from. Optional.
// returns: dojo.NodeList
// An instance of `dojo.NodeList`. Many methods are available on
// NodeLists for searching, iterating, manipulating, and handling
// events on the matched nodes in the returned list.
// example:
// search the entire document for elements with the class "foo":
// | dojo.query(".foo");
// these elements will match:
// | <span class="foo"></span>
// | <span class="foo bar"></span>
// | <p class="thud foo"></p>
// example:
// search the entire document for elements with the classes "foo" *and* "bar":
// | dojo.query(".foo.bar");
// these elements will match:
// | <span class="foo bar"></span>
// while these will not:
// | <span class="foo"></span>
// | <p class="thud foo"></p>
// example:
// find `<span>` elements which are descendants of paragraphs and
// which have a "highlighted" class:
// | dojo.query("p span.highlighted");
// the innermost span in this fragment matches:
// | <p class="foo">
// | <span>...
// | <span class="highlighted foo bar">...</span>
// | </span>
// | </p>
// example:
// set an "odd" class on all odd table rows inside of the table
// `#tabular_data`, using the `>` (direct child) selector to avoid
// affecting any nested tables:
// | dojo.query("#tabular_data > tbody > tr:nth-child(odd)").addClass("odd");
// example:
// remove all elements with the class "error" from the document
// and store them in a list:
// | var errors = dojo.query(".error").orphan();
// example:
// add an onclick handler to every submit button in the document
// which causes the form to be sent via Ajax instead:
// | dojo.query("input[type='submit']").onclick(function(e){
// | dojo.stopEvent(e); // prevent sending the form
// | var btn = e.target;
// | dojo.xhrPost({
// | form: btn.form,
// | load: function(data){
// | // replace the form with the response
// | var div = dojo.doc.createElement("div");
// | dojo.place(div, btn.form, "after");
// | div.innerHTML = data;
// | dojo.style(btn.form, "display", "none");
// | }
// | });
// | });
// NOTE: elementsById is not currently supported
// NOTE: ignores xpath-ish queries for now
if(query.constructor == d.NodeList){
return query;
}
if(!d.isString(query)){
return new d.NodeList(query); // dojo.NodeList
}
if(d.isString(root)){
root = d.byId(root);
}
return _zip(getQueryFunc(query)(root||d.doc)); // dojo.NodeList
}
/*
// exposing this was a mistake
d.query.attrs = attrs;
*/
// exposing this because new pseudo matches are only executed through the
// DOM query path (never through the xpath optimizing branch)
d.query.pseudos = pseudos;
// one-off function for filtering a NodeList based on a simple selector
d._filterQueryResult = function(nodeList, simpleFilter){
var tnl = new d.NodeList();
var ff = (simpleFilter) ? getFilterFunc(getQueryParts(simpleFilter)[0]) : function(){ return true; };
for(var x=0, te; te = nodeList[x]; x++){
if(ff(te)){ tnl.push(te); }
}
return tnl;
}
})();
}