/* Parse function for JavaScript. Makes use of the tokenizer from
 * tokenizejavascript.js. Note that your parsers do not have to be
 * this complicated -- if you don't want to recognize local variables,
 * in many languages it is enough to just look for braces, semicolons,
 * parentheses, etc, and know when you are inside a string or comment.
 *
 * See manual.html for more info about the parser interface.
 */

var JSParser = Editor.Parser = (function() {
  // Token types that can be considered to be atoms.
  var atomicTypes = {"atom": true, "number": true, "variable": true, "string": true, "regexp": true};
  // Setting that can be used to have JSON data indent properly.
  var json = false;
  // Constructor for the lexical context objects.
  function JSLexical(indented, column, type, align, prev, info) {
    // indentation at start of this line
    this.indented = indented;
    // column at which this scope was opened
    this.column = column;
    // type of scope ('vardef', 'stat' (statement), 'form' (special form), '[', '{', or '(')
    this.type = type;
    // '[', '{', or '(' blocks that have any text after their opening
    // character are said to be 'aligned' -- any lines below are
    // indented all the way to the opening character.
    if (align != null)
      this.align = align;
    // Parent scope, if any.
    this.prev = prev;
    this.info = info;
  }

  // My favourite JavaScript indentation rules.
  function indentJS(lexical) {
    return function(firstChars) {
      var firstChar = firstChars && firstChars.charAt(0), type = lexical.type;
      var closing = firstChar == type;
      if (type == "vardef")
        return lexical.indented + 4;
      else if (type == "form" && firstChar == "{")
        return lexical.indented;
      else if (type == "stat" || type == "form")
        return lexical.indented + indentUnit;
      else if (lexical.info == "switch" && !closing)
        return lexical.indented + (/^(?:case|default)\b/.test(firstChars) ? indentUnit : 2 * indentUnit);
      else if (lexical.align)
        return lexical.column - (closing ? 1 : 0);
      else
        return lexical.indented + (closing ? 0 : indentUnit);
    };
  }

  // The parser-iterator-producing function itself.
  function parseJS(input, basecolumn) {
    // Wrap the input in a token stream
    var tokens = tokenizeJavaScript(input);
    // The parser state. cc is a stack of actions that have to be
    // performed to finish the current statement. For example we might
    // know that we still need to find a closing parenthesis and a
    // semicolon. Actions at the end of the stack go first. It is
    // initialized with an infinitely looping action that consumes
    // whole statements.
    var cc = [json ? expressions : statements];
    // Context contains information about the current local scope, the
    // variables defined in that, and the scopes above it.
    var context = null;
    // The lexical scope, used mostly for indentation.
    var lexical = new JSLexical((basecolumn || 0) - indentUnit, 0, "block", false);
    // Current column, and the indentation at the start of the current
    // line. Used to create lexical scope objects.
    var column = 0;
    var indented = 0;
    // Variables which are used by the mark, cont, and pass functions
    // below to communicate with the driver loop in the 'next'
    // function.
    var consume, marked;

    // The iterator object.
    var parser = {next: next, copy: copy};

    function next(){
      // Start by performing any 'lexical' actions (adjusting the
      // lexical variable), or the operations below will be working
      // with the wrong lexical state.
      while(cc[cc.length - 1].lex)
        cc.pop()();

      // Fetch a token.
      var token = tokens.next();

      // Adjust column and indented.
      if (token.type == "whitespace" && column == 0)
        indented = token.value.length;
      column += token.value.length;
      if (token.content == "\n"){
        indented = column = 0;
        // If the lexical scope's align property is still undefined at
        // the end of the line, it is an un-aligned scope.
        if (!("align" in lexical))
          lexical.align = false;
        // Newline tokens get an indentation function associated with
        // them.
        token.indentation = indentJS(lexical);
      }
      // No more processing for meaningless tokens.
      if (token.type == "whitespace" || token.type == "comment")
        return token;
      // When a meaningful token is found and the lexical scope's
      // align is undefined, it is an aligned scope.
      if (!("align" in lexical))
        lexical.align = true;

      // Execute actions until one 'consumes' the token and we can
      // return it.
      while(true) {
        consume = marked = false;
        // Take and execute the topmost action.
        cc.pop()(token.type, token.content);
        if (consume){
          // Marked is used to change the style of the current token.
          if (marked)
            token.style = marked;
          // Here we differentiate between local and global variables.
          else if (token.type == "variable" && inScope(token.content))
            token.style = "js-localvariable";
          return token;
        }
      }
    }

    // This makes a copy of the parser state. It stores all the
    // stateful variables in a closure, and returns a function that
    // will restore them when called with a new input stream. Note
    // that the cc array has to be copied, because it is contantly
    // being modified. Lexical objects are not mutated, and context
    // objects are not mutated in a harmful way, so they can be shared
    // between runs of the parser.
    function copy(){
      var _context = context, _lexical = lexical, _cc = cc.concat([]), _tokenState = tokens.state;

      return function copyParser(input){
        context = _context;
        lexical = _lexical;
        cc = _cc.concat([]); // copies the array
        column = indented = 0;
        tokens = tokenizeJavaScript(input, _tokenState);
        return parser;
      };
    }

    // Helper function for pushing a number of actions onto the cc
    // stack in reverse order.
    function push(fs){
      for (var i = fs.length - 1; i >= 0; i--)
        cc.push(fs[i]);
    }
    // cont and pass are used by the action functions to add other
    // actions to the stack. cont will cause the current token to be
    // consumed, pass will leave it for the next action.
    function cont(){
      push(arguments);
      consume = true;
    }
    function pass(){
      push(arguments);
      consume = false;
    }
    // Used to change the style of the current token.
    function mark(style){
      marked = style;
    }

    // Push a new scope. Will automatically link the current scope.
    function pushcontext(){
      context = {prev: context, vars: {"this": true, "arguments": true}};
    }
    // Pop off the current scope.
    function popcontext(){
      context = context.prev;
    }
    // Register a variable in the current scope.
    function register(varname){
      if (context){
        mark("js-variabledef");
        context.vars[varname] = true;
      }
    }
    // Check whether a variable is defined in the current scope.
    function inScope(varname){
      var cursor = context;
      while (cursor) {
        if (cursor.vars[varname])
          return true;
        cursor = cursor.prev;
      }
      return false;
    }

    // Push a new lexical context of the given type.
    function pushlex(type, info) {
      var result = function(){
        lexical = new JSLexical(indented, column, type, null, lexical, info)
      };
      result.lex = true;
      return result;
    }
    // Pop off the current lexical context.
    function poplex(){
      if (lexical.type == ")")
        indented = lexical.indented;
      lexical = lexical.prev;
    }
    poplex.lex = true;
    // The 'lex' flag on these actions is used by the 'next' function
    // to know they can (and have to) be ran before moving on to the
    // next token.

    // Creates an action that discards tokens until it finds one of
    // the given type.
    function expect(wanted){
      return function expecting(type){
        if (type == wanted) cont();
        else if (wanted == ";") pass();
        else cont(arguments.callee);
      };
    }

    // Looks for a statement, and then calls itself.
    function statements(type){
      return pass(statement, statements);
    }
    function expressions(type){
      return pass(expression, expressions);
    }
    // Dispatches various types of statements based on the type of the
    // current token.
    function statement(type){
      if (type == "var") cont(pushlex("vardef"), vardef1, expect(";"), poplex);
      else if (type == "keyword a") cont(pushlex("form"), expression, statement, poplex);
      else if (type == "keyword b") cont(pushlex("form"), statement, poplex);
      else if (type == "{") cont(pushlex("}"), block, poplex);
      else if (type == ";") cont();
      else if (type == "function") cont(functiondef);
      else if (type == "for") cont(pushlex("form"), expect("("), pushlex(")"), forspec1, expect(")"), poplex, statement, poplex);
      else if (type == "variable") cont(pushlex("stat"), maybelabel);
      else if (type == "switch") cont(pushlex("form"), expression, pushlex("}", "switch"), expect("{"), block, poplex, poplex);
      else if (type == "case") cont(expression, expect(":"));
      else if (type == "default") cont(expect(":"));
      else if (type == "catch") cont(pushlex("form"), pushcontext, expect("("), funarg, expect(")"), statement, poplex, popcontext);
      else pass(pushlex("stat"), expression, expect(";"), poplex);
    }
    // Dispatch expression types.
    function expression(type){
      if (atomicTypes.hasOwnProperty(type)) cont(maybeoperator);
      else if (type == "function") cont(functiondef);
      else if (type == "keyword c") cont(expression);
      else if (type == "(") cont(pushlex(")"), expression, expect(")"), poplex, maybeoperator);
      else if (type == "operator") cont(expression);
      else if (type == "[") cont(pushlex("]"), commasep(expression, "]"), poplex, maybeoperator);
      else if (type == "{") cont(pushlex("}"), commasep(objprop, "}"), poplex, maybeoperator);
      else cont();
    }
    // Called for places where operators, function calls, or
    // subscripts are valid. Will skip on to the next action if none
    // is found.
    function maybeoperator(type, value){
      if (type == "operator" && /\+\+|--/.test(value)) cont(maybeoperator);
      else if (type == "operator") cont(expression);
      else if (type == ";") pass();
      else if (type == "(") cont(pushlex(")"), commasep(expression, ")"), poplex, maybeoperator);
      else if (type == ".") cont(property, maybeoperator);
      else if (type == "[") cont(pushlex("]"), expression, expect("]"), poplex, maybeoperator);
    }
    // When a statement starts with a variable name, it might be a
    // label. If no colon follows, it's a regular statement.
    function maybelabel(type){
      if (type == ":") cont(poplex, statement);
      else pass(maybeoperator, expect(";"), poplex);
    }
    // Property names need to have their style adjusted -- the
    // tokenizer thinks they are variables.
    function property(type){
      if (type == "variable") {mark("js-property"); cont();}
    }
    // This parses a property and its value in an object literal.
    function objprop(type){
      if (type == "variable") mark("js-property");
      if (atomicTypes.hasOwnProperty(type)) cont(expect(":"), expression);
    }
    // Parses a comma-separated list of the things that are recognized
    // by the 'what' argument.
    function commasep(what, end){
      function proceed(type) {
        if (type == ",") cont(what, proceed);
        else if (type == end) cont();
        else cont(expect(end));
      }
      return function commaSeparated(type) {
        if (type == end) cont();
        else pass(what, proceed);
      };
    }
    // Look for statements until a closing brace is found.
    function block(type){
      if (type == "}") cont();
      else pass(statement, block);
    }
    // Variable definitions are split into two actions -- 1 looks for
    // a name or the end of the definition, 2 looks for an '=' sign or
    // a comma.
    function vardef1(type, value){
      if (type == "variable"){register(value); cont(vardef2);}
      else cont();
    }
    function vardef2(type, value){
      if (value == "=") cont(expression, vardef2);
      else if (type == ",") cont(vardef1);
    }
    // For loops.
    function forspec1(type){
      if (type == "var") cont(vardef1, forspec2);
      else if (type == ";") pass(forspec2);
      else if (type == "variable") cont(formaybein);
      else pass(forspec2);
    }
    function formaybein(type, value){
      if (value == "in") cont(expression);
      else cont(maybeoperator, forspec2);
    }
    function forspec2(type, value){
      if (type == ";") cont(forspec3);
      else if (value == "in") cont(expression);
      else cont(expression, expect(";"), forspec3);
    }
    function forspec3(type) {
      if (type == ")") pass();
      else cont(expression);
    }
    // A function definition creates a new context, and the variables
    // in its argument list have to be added to this context.
    function functiondef(type, value){
      if (type == "variable"){register(value); cont(functiondef);}
      else if (type == "(") cont(pushlex(")"), pushcontext, commasep(funarg, ")"), poplex, statement, popcontext);
    }
    function funarg(type, value){
      if (type == "variable"){register(value); cont();}
    }

    return parser;
  }

  return {
    make: parseJS,
    electricChars: "{}:",
    configure: function(obj) {
      if (obj.json != null) json = obj.json;
    }
  };
})();