// A framework for simple tokenizers. Takes care of newlines and
// white-space, and of getting the text from the source stream into
// the token object. A state is a function of two arguments -- a
// string stream and a setState function. The second can be used to
// change the tokenizer's state, and can be ignored for stateless
// tokenizers. This function should advance the stream over a token
// and return a string or object containing information about the next
// token, or null to pass and have the (new) state be called to finish
// the token. When a string is given, it is wrapped in a {style, type}
// object. In the resulting object, the characters consumed are stored
// under the content property. Any whitespace following them is also
// automatically consumed, and added to the value property. (Thus,
// content is the actual meaningful part of the token, while value
// contains all the text it spans.)

function tokenizer(source, state) {
  // Newlines are always a separate token.
  function isWhiteSpace(ch) {
    // The messy regexp is because IE's regexp matcher is of the
    // opinion that non-breaking spaces are no whitespace.
    return ch != "\n" && /^[\s\u00a0]*$/.test(ch);
  }

  var tokenizer = {
    state: state,

    take: function(type) {
      if (typeof(type) == "string")
        type = {style: type, type: type};

      type.content = (type.content || "") + source.get();
      if (!/\n$/.test(type.content))
        source.nextWhile(isWhiteSpace);
      type.value = type.content + source.get();
      return type;
    },

    next: function () {
      if (!source.more()) throw StopIteration;

      var type;
      if (source.equals("\n")) {
        source.next();
        return this.take("whitespace");
      }

      if (source.applies(isWhiteSpace))
        type = "whitespace";
      else
        while (!type)
          type = this.state(source, function(s) {tokenizer.state = s;});

      return this.take(type);
    }
  };
  return tokenizer;
}