/*! Copyright (c) 2011, Lloyd Hilaiel, ISC License */
/*
 * This is the JSONSelect reference implementation, in javascript.  This
 * code is designed to run under node.js or in a browser.  In the former
 * case, the "public API" is exposed as properties on the `export` object,
 * in the latter, as properties on `window.JSONSelect`.  That API is thus:
 *
 * .match(selector, object)
 *
 *   Parses and "compiles" the selector, then matches it against the object
 *   argument.  Matches are returned in an array.  Throws an error when
 *   there's a problem parsing the selector.
 *
 * .forEach(selector, object, callback)
 *
 *   Like match, but rather than returning an array, invokes the provided
 *   callback once per match as the matches are discovered. 
 * 
 * .compile(selector) 
 *
 *   Parses the selector and compiles it to an internal form, and returns
 *   an object which contains the compiled selector and has two properties:
 *   `match` and `forEach`.  These two functions work identically to the
 *   above, except they do not take a selector as an argument and instead
 *   use the compiled selector.
 *
 *   For cases where a complex selector is repeatedly used, this method
 *   should be faster as it will avoid recompiling the selector each time. 
 */
(function(exports) {

    var // localize references
    toString = Object.prototype.toString;

    function jsonParse(str) {
      try {
          if(JSON && JSON.parse){
              return JSON.parse(str);
          }
          return (new Function("return " + str))();
      } catch(e) {
        te("ijs", e.message);
      }
    }

    // emitted error codes.
    var errorCodes = {
        "bop":  "binary operator expected",
        "ee":   "expression expected",
        "epex": "closing paren expected ')'",
        "ijs":  "invalid json string",
        "mcp":  "missing closing paren",
        "mepf": "malformed expression in pseudo-function",
        "mexp": "multiple expressions not allowed",
        "mpc":  "multiple pseudo classes (:xxx) not allowed",
        "nmi":  "multiple ids not allowed",
        "pex":  "opening paren expected '('",
        "se":   "selector expected",
        "sex":  "string expected",
        "sra":  "string required after '.'",
        "uc":   "unrecognized char",
        "ucp":  "unexpected closing paren",
        "ujs":  "unclosed json string",
        "upc":  "unrecognized pseudo class"
    };

    // throw an error message
    function te(ec, context) {
      throw new Error(errorCodes[ec] + ( context && " in '" + context + "'"));
    }

    // THE LEXER
    var toks = {
        psc: 1, // pseudo class
        psf: 2, // pseudo class function
        typ: 3, // type
        str: 4, // string
        ide: 5  // identifiers (or "classes", stuff after a dot)
    };

    // The primary lexing regular expression in jsonselect
    var pat = new RegExp(
        "^(?:" +
        // (1) whitespace
        "([\\r\\n\\t\\ ]+)|" +
        // (2) one-char ops
        "([~*,>\\)\\(])|" +
        // (3) types names
        "(string|boolean|null|array|object|number)|" +
        // (4) pseudo classes
        "(:(?:root|first-child|last-child|only-child))|" +
        // (5) pseudo functions
        "(:(?:nth-child|nth-last-child|has|expr|val|contains))|" +
        // (6) bogusly named pseudo something or others
        "(:\\w+)|" +
        // (7 & 8) identifiers and JSON strings
        "(?:(\\.)?(\\\"(?:[^\\\\\\\"]|\\\\[^\\\"])*\\\"))|" +
        // (8) bogus JSON strings missing a trailing quote
        "(\\\")|" +
        // (9) identifiers (unquoted)
        "\\.((?:[_a-zA-Z]|[^\\0-\\0177]|\\\\[^\\r\\n\\f0-9a-fA-F])(?:[_a-zA-Z0-9\\-]|[^\\u0000-\\u0177]|(?:\\\\[^\\r\\n\\f0-9a-fA-F]))*)" +
        ")"
    );

    // A regular expression for matching "nth expressions" (see grammar, what :nth-child() eats)
    var nthPat = /^\s*\(\s*(?:([+\-]?)([0-9]*)n\s*(?:([+\-])\s*([0-9]))?|(odd|even)|([+\-]?[0-9]+))\s*\)/;
    function lex(str, off) {
        if (!off) off = 0;
        var m = pat.exec(str.substr(off));
        if (!m) return undefined;
        off+=m[0].length;
        var a;
        if (m[1]) a = [off, " "];
        else if (m[2]) a = [off, m[0]];
        else if (m[3]) a = [off, toks.typ, m[0]];
        else if (m[4]) a = [off, toks.psc, m[0]];
        else if (m[5]) a = [off, toks.psf, m[0]];
        else if (m[6]) te("upc", str);
        else if (m[8]) a = [off, m[7] ? toks.ide : toks.str, jsonParse(m[8])];
        else if (m[9]) te("ujs", str);
        else if (m[10]) a = [off, toks.ide, m[10].replace(/\\([^\r\n\f0-9a-fA-F])/g,"$1")];
        return a;
    }

    // THE EXPRESSION SUBSYSTEM

    var exprPat = new RegExp(
            // skip and don't capture leading whitespace
            "^\\s*(?:" +
            // (1) simple vals
            "(true|false|null)|" + 
            // (2) numbers
            "(-?\\d+(?:\\.\\d*)?(?:[eE][+\\-]?\\d+)?)|" +
            // (3) strings
            "(\"(?:[^\\]|\\[^\"])*\")|" +
            // (4) the 'x' value placeholder
            "(x)|" +
            // (5) binops
            "(&&|\\|\\||[\\$\\^<>!\\*]=|[=+\\-*/%<>])|" +
            // (6) parens
            "([\\(\\)])" +
            ")"
    );

    function is(o, t) { return typeof o === t; }
    var operators = {
        '*':  [ 9, function(lhs, rhs) { return lhs * rhs; } ],
        '/':  [ 9, function(lhs, rhs) { return lhs / rhs; } ],
        '%':  [ 9, function(lhs, rhs) { return lhs % rhs; } ],
        '+':  [ 7, function(lhs, rhs) { return lhs + rhs; } ],
        '-':  [ 7, function(lhs, rhs) { return lhs - rhs; } ],
        '<=': [ 5, function(lhs, rhs) { return is(lhs, 'number') && is(rhs, 'number') && lhs <= rhs; } ],
        '>=': [ 5, function(lhs, rhs) { return is(lhs, 'number') && is(rhs, 'number') && lhs >= rhs; } ],
        '$=': [ 5, function(lhs, rhs) { return is(lhs, 'string') && is(rhs, 'string') && lhs.lastIndexOf(rhs) === lhs.length - rhs.length; } ],
        '^=': [ 5, function(lhs, rhs) { return is(lhs, 'string') && is(rhs, 'string') && lhs.indexOf(rhs) === 0; } ],
        '*=': [ 5, function(lhs, rhs) { return is(lhs, 'string') && is(rhs, 'string') && lhs.indexOf(rhs) !== -1; } ],
        '>':  [ 5, function(lhs, rhs) { return is(lhs, 'number') && is(rhs, 'number') && lhs > rhs; } ],
        '<':  [ 5, function(lhs, rhs) { return is(lhs, 'number') && is(rhs, 'number') && lhs < rhs; } ],
        '=':  [ 3, function(lhs, rhs) { return lhs === rhs; } ],
        '!=': [ 3, function(lhs, rhs) { return lhs !== rhs; } ],
        '&&': [ 2, function(lhs, rhs) { return lhs && rhs; } ],
        '||': [ 1, function(lhs, rhs) { return lhs || rhs; } ]
    };

    function exprLex(str, off) {
        var v, m = exprPat.exec(str.substr(off));
        if (m) {
            off += m[0].length;
            v = m[1] || m[2] || m[3] || m[5] || m[6];
            if (m[1] || m[2] || m[3]) return [off, 0, jsonParse(v)];
            else if (m[4]) return [off, 0, undefined];
            return [off, v];
        }
    }

    function exprParse2(str, off) {
        if (!off) off = 0;
        // first we expect a value or a '('
        var l = exprLex(str, off),
            lhs;
        if (l && l[1] === '(') {
            lhs = exprParse2(str, l[0]);
            var p = exprLex(str, lhs[0]);
            if (!p || p[1] !== ')') te('epex', str);
            off = p[0];
            lhs = [ '(', lhs[1] ];
        } else if (!l || (l[1] && l[1] != 'x')) {
            te("ee", str + " - " + ( l[1] && l[1] ));
        } else {
            lhs = ((l[1] === 'x') ? undefined : l[2]);
            off = l[0];
        }

        // now we expect a binary operator or a ')'
        var op = exprLex(str, off);
        if (!op || op[1] == ')') return [off, lhs];
        else if (op[1] == 'x' || !op[1]) {
            te('bop', str + " - " + ( op[1] && op[1] ));
        }

        // tail recursion to fetch the rhs expression
        var rhs = exprParse2(str, op[0]);
        off = rhs[0];
        rhs = rhs[1];

        // and now precedence!  how shall we put everything together?
        var v;
        if (typeof rhs !== 'object' || rhs[0] === '(' || operators[op[1]][0] < operators[rhs[1]][0] ) {
            v = [lhs, op[1], rhs];
        }
        else {
            v = rhs;
            while (typeof rhs[0] === 'object' && rhs[0][0] != '(' && operators[op[1]][0] >= operators[rhs[0][1]][0]) {
                rhs = rhs[0];
            }
            rhs[0] = [lhs, op[1], rhs[0]];
        }
        return [off, v];
    }

    function exprParse(str, off) {
        function deparen(v) {
            if (typeof v !== 'object' || v === null) return v;
            else if (v[0] === '(') return deparen(v[1]);
            else return [deparen(v[0]), v[1], deparen(v[2])];
        }
        var e = exprParse2(str, off ? off : 0);
        return [e[0], deparen(e[1])];
    }

    function exprEval(expr, x) {
        if (expr === undefined) return x;
        else if (expr === null || typeof expr !== 'object') {
            return expr;
        }
        var lhs = exprEval(expr[0], x),
            rhs = exprEval(expr[2], x);
        return operators[expr[1]][1](lhs, rhs);
    }

    // THE PARSER

    function parse(str, off, nested, hints) {
        if (!nested) hints = {};

        var a = [], am, readParen;
        if (!off) off = 0; 

        while (true) {
            var s = parse_selector(str, off, hints);
            a.push(s[1]);
            s = lex(str, off = s[0]);
            if (s && s[1] === " ") s = lex(str, off = s[0]);
            if (!s) break;
            // now we've parsed a selector, and have something else...
            if (s[1] === ">" || s[1] === "~") {
                if (s[1] === "~") hints.usesSiblingOp = true;
                a.push(s[1]);
                off = s[0];
            } else if (s[1] === ",") {
                if (am === undefined) am = [ ",", a ];
                else am.push(a);
                a = [];
                off = s[0];
            } else if (s[1] === ")") {
                if (!nested) te("ucp", s[1]);
                readParen = 1;
                off = s[0];
                break;
            }
        }
        if (nested && !readParen) te("mcp", str);
        if (am) am.push(a);
        var rv;
        if (!nested && hints.usesSiblingOp) {
            rv = normalize(am ? am : a);
        } else {
            rv = am ? am : a;
        }
        return [off, rv];
    }

    function normalizeOne(sel) {
        var sels = [], s;
        for (var i = 0; i < sel.length; i++) {
            if (sel[i] === '~') {
                // `A ~ B` maps to `:has(:root > A) > B`
                // `Z A ~ B` maps to `Z :has(:root > A) > B, Z:has(:root > A) > B`
                // This first clause, takes care of the first case, and the first half of the latter case.
                if (i < 2 || sel[i-2] != '>') {
                    s = sel.slice(0,i-1);
                    s = s.concat([{has:[[{pc: ":root"}, ">", sel[i-1]]]}, ">"]);
                    s = s.concat(sel.slice(i+1));
                    sels.push(s);
                }
                // here we take care of the second half of above:
                // (`Z A ~ B` maps to `Z :has(:root > A) > B, Z :has(:root > A) > B`)
                // and a new case:
                // Z > A ~ B maps to Z:has(:root > A) > B
                if (i > 1) {
                    var at = sel[i-2] === '>' ? i-3 : i-2;
                    s = sel.slice(0,at);
                    var z = {};
                    for (var k in sel[at]) if (sel[at].hasOwnProperty(k)) z[k] = sel[at][k];
                    if (!z.has) z.has = [];
                    z.has.push([{pc: ":root"}, ">", sel[i-1]]);
                    s = s.concat(z, '>', sel.slice(i+1));
                    sels.push(s);
                }
                break;
            }
        }
        if (i == sel.length) return sel;
        return sels.length > 1 ? [','].concat(sels) : sels[0];
    }

    function normalize(sels) {
        if (sels[0] === ',') {
            var r = [","];
            for (var i = i; i < sels.length; i++) {
                var s = normalizeOne(s[i]);
                r = r.concat(s[0] === "," ? s.slice(1) : s);
            }
            return r;
        } else {
            return normalizeOne(sels);
        }
    }

    function parse_selector(str, off, hints) {
        var soff = off;
        var s = { };
        var l = lex(str, off);
        // skip space
        if (l && l[1] === " ") { soff = off = l[0]; l = lex(str, off); }
        if (l && l[1] === toks.typ) {
            s.type = l[2];
            l = lex(str, (off = l[0]));
        } else if (l && l[1] === "*") {
            // don't bother representing the universal sel, '*' in the
            // parse tree, cause it's the default
            l = lex(str, (off = l[0]));
        }

        // now support either an id or a pc
        while (true) {
            if (l === undefined) {
                break;
            } else if (l[1] === toks.ide) {
                if (s.id) te("nmi", l[1]);
                s.id = l[2];
            } else if (l[1] === toks.psc) {
                if (s.pc || s.pf) te("mpc", l[1]);
                // collapse first-child and last-child into nth-child expressions
                if (l[2] === ":first-child") {
                    s.pf = ":nth-child";
                    s.a = 0;
                    s.b = 1;
                } else if (l[2] === ":last-child") {
                    s.pf = ":nth-last-child";
                    s.a = 0;
                    s.b = 1;
                } else {
                    s.pc = l[2];
                }
            } else if (l[1] === toks.psf) {
                if (l[2] === ":val" || l[2] === ":contains") {
                    s.expr = [ undefined, l[2] === ":val" ? "=" : "*=", undefined];
                    // any amount of whitespace, followed by paren, string, paren
                    l = lex(str, (off = l[0]));
                    if (l && l[1] === " ") l = lex(str, off = l[0]);
                    if (!l || l[1] !== "(") te("pex", str);
                    l = lex(str, (off = l[0]));
                    if (l && l[1] === " ") l = lex(str, off = l[0]);
                    if (!l || l[1] !== toks.str) te("sex", str);
                    s.expr[2] = l[2];
                    l = lex(str, (off = l[0]));
                    if (l && l[1] === " ") l = lex(str, off = l[0]);
                    if (!l || l[1] !== ")") te("epex", str);
                } else if (l[2] === ":has") {
                    // any amount of whitespace, followed by paren
                    l = lex(str, (off = l[0]));
                    if (l && l[1] === " ") l = lex(str, off = l[0]);
                    if (!l || l[1] !== "(") te("pex", str);
                    var h = parse(str, l[0], true);
                    l[0] = h[0];
                    if (!s.has) s.has = [];
                    s.has.push(h[1]);
                } else if (l[2] === ":expr") {
                    if (s.expr) te("mexp", str);
                    var e = exprParse(str, l[0]);
                    l[0] = e[0];
                    s.expr = e[1];
                } else {
                    if (s.pc || s.pf ) te("mpc", str);
                    s.pf = l[2];
                    var m = nthPat.exec(str.substr(l[0]));
                    if (!m) te("mepf", str);
                    if (m[5]) {
                        s.a = 2;
                        s.b = (m[5] === "odd") ? 1 : 0;
                    } else if (m[6]) {
                        s.a = 0;
                        s.b = parseInt(m[6], 10);
                    } else {
                        s.a = parseInt((m[1] ? m[1] : "+") + (m[2] ? m[2] : "1"),10);
                        s.b = m[3] ? parseInt(m[3] + m[4],10) : 0;
                    }
                    l[0] += m[0].length;
                }
            } else {
                break;
            }
            l = lex(str, (off = l[0]));
        }

        // now if we didn't actually parse anything it's an error
        if (soff === off) te("se", str);

        return [off, s];
    }

    // THE EVALUATOR

    function isArray(o) {
        return Array.isArray ? Array.isArray(o) : 
          toString.call(o) === "[object Array]";
    }

    function mytypeof(o) {
        if (o === null) return "null";
        var to = typeof o;
        if (to === "object" && isArray(o)) to = "array";
        return to;
    }

    function mn(node, sel, id, num, tot) {
        var sels = [];
        var cs = (sel[0] === ">") ? sel[1] : sel[0];
        var m = true, mod;
        if (cs.type) m = m && (cs.type === mytypeof(node));
        if (cs.id)   m = m && (cs.id === id);
        if (m && cs.pf) {
            if (cs.pf === ":nth-last-child") num = tot - num;
            else num++;
            if (cs.a === 0) {
                m = cs.b === num;
            } else {
                mod = ((num - cs.b) % cs.a);

                m = (!mod && ((num*cs.a + cs.b) >= 0));
            }
        }
        if (m && cs.has) {
            // perhaps we should augment forEach to handle a return value
            // that indicates "client cancels traversal"?
            var bail = function() { throw 42; };
            for (var i = 0; i < cs.has.length; i++) {
                try {
                    forEach(cs.has[i], node, bail);
                } catch (e) {
                    if (e === 42) continue;
                }
                m = false;
                break;
            }
        }
        if (m && cs.expr) {
            m = exprEval(cs.expr, node);
        }
        // should we repeat this selector for descendants?
        if (sel[0] !== ">" && sel[0].pc !== ":root") sels.push(sel);

        if (m) {
            // is there a fragment that we should pass down?
            if (sel[0] === ">") { if (sel.length > 2) { m = false; sels.push(sel.slice(2)); } }
            else if (sel.length > 1) { m = false; sels.push(sel.slice(1)); }
        }

        return [m, sels];
    }

    function forEach(sel, obj, fun, id, num, tot) {
        var a = (sel[0] === ",") ? sel.slice(1) : [sel],
        a0 = [],
        call = false,
        i = 0, j = 0, k, x;
        for (i = 0; i < a.length; i++) {
            x = mn(obj, a[i], id, num, tot);
            if (x[0]) {
                call = true;
            }
            for (j = 0; j < x[1].length; j++) {
                a0.push(x[1][j]);
            }
        }
        if (a0.length && typeof obj === "object") {
            if (a0.length >= 1) {
                a0.unshift(",");
            }
            if (isArray(obj)) {
                for (i = 0; i < obj.length; i++) {
                    forEach(a0, obj[i], fun, undefined, i, obj.length);
                }
            } else {
                for (k in obj) {
                    if (obj.hasOwnProperty(k)) {
                        forEach(a0, obj[k], fun, k);
                    }
                }
            }
        }
        if (call && fun) {
            fun(obj);
        }
    }

    function match(sel, obj) {
        var a = [];
        forEach(sel, obj, function(x) {
            a.push(x);
        });
        return a;
    }

    function compile(sel) {
        return {
            sel: parse(sel)[1],
            match: function(obj){
                return match(this.sel, obj);
            },
            forEach: function(obj, fun) {
                return forEach(this.sel, obj, fun);
            }
        };
    }

    exports._lex = lex;
    exports._parse = parse;
    exports.match = function (sel, obj) {
        return compile(sel).match(obj);
    };
    exports.forEach = function(sel, obj, fun) {
        return compile(sel).forEach(obj, fun);
    };
    exports.compile = compile;
})(typeof exports === "undefined" ? (window.JSONSelect = {}) : exports);