1248 lines
41 KiB
JavaScript
1248 lines
41 KiB
JavaScript
|
/***********************************************************************
|
||
|
|
||
|
A JavaScript tokenizer / parser / beautifier / compressor.
|
||
|
|
||
|
This version is suitable for Node.js. With minimal changes (the
|
||
|
exports stuff) it should work on any JS platform.
|
||
|
|
||
|
This file contains the tokenizer/parser. It is a port to JavaScript
|
||
|
of parse-js [1], a JavaScript parser library written in Common Lisp
|
||
|
by Marijn Haverbeke. Thank you Marijn!
|
||
|
|
||
|
[1] http://marijn.haverbeke.nl/parse-js/
|
||
|
|
||
|
Exported functions:
|
||
|
|
||
|
- tokenizer(code) -- returns a function. Call the returned
|
||
|
function to fetch the next token.
|
||
|
|
||
|
- parse(code) -- returns an AST of the given JavaScript code.
|
||
|
|
||
|
-------------------------------- (C) ---------------------------------
|
||
|
|
||
|
Author: Mihai Bazon
|
||
|
<mihai.bazon@gmail.com>
|
||
|
http://mihai.bazon.net/blog
|
||
|
|
||
|
Distributed under the BSD license:
|
||
|
|
||
|
Copyright 2010 (c) Mihai Bazon <mihai.bazon@gmail.com>
|
||
|
Based on parse-js (http://marijn.haverbeke.nl/parse-js/).
|
||
|
|
||
|
Redistribution and use in source and binary forms, with or without
|
||
|
modification, are permitted provided that the following conditions
|
||
|
are met:
|
||
|
|
||
|
* Redistributions of source code must retain the above
|
||
|
copyright notice, this list of conditions and the following
|
||
|
disclaimer.
|
||
|
|
||
|
* Redistributions in binary form must reproduce the above
|
||
|
copyright notice, this list of conditions and the following
|
||
|
disclaimer in the documentation and/or other materials
|
||
|
provided with the distribution.
|
||
|
|
||
|
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER “AS IS” AND ANY
|
||
|
EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||
|
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||
|
PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER BE
|
||
|
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY,
|
||
|
OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||
|
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||
|
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
||
|
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
|
||
|
TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
|
||
|
THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
||
|
SUCH DAMAGE.
|
||
|
|
||
|
***********************************************************************/
|
||
|
|
||
|
/* -----[ Tokenizer (constants) ]----- */
|
||
|
|
||
|
var KEYWORDS = array_to_hash([
|
||
|
"break",
|
||
|
"case",
|
||
|
"catch",
|
||
|
"const",
|
||
|
"continue",
|
||
|
"default",
|
||
|
"delete",
|
||
|
"do",
|
||
|
"else",
|
||
|
"finally",
|
||
|
"for",
|
||
|
"function",
|
||
|
"if",
|
||
|
"in",
|
||
|
"instanceof",
|
||
|
"new",
|
||
|
"return",
|
||
|
"switch",
|
||
|
"throw",
|
||
|
"try",
|
||
|
"typeof",
|
||
|
"var",
|
||
|
"void",
|
||
|
"while",
|
||
|
"with"
|
||
|
]);
|
||
|
|
||
|
var RESERVED_WORDS = array_to_hash([
|
||
|
"abstract",
|
||
|
"boolean",
|
||
|
"byte",
|
||
|
"char",
|
||
|
"class",
|
||
|
"debugger",
|
||
|
"double",
|
||
|
"enum",
|
||
|
"export",
|
||
|
"extends",
|
||
|
"final",
|
||
|
"float",
|
||
|
"goto",
|
||
|
"implements",
|
||
|
"import",
|
||
|
"int",
|
||
|
"interface",
|
||
|
"long",
|
||
|
"native",
|
||
|
"package",
|
||
|
"private",
|
||
|
"protected",
|
||
|
"public",
|
||
|
"short",
|
||
|
"static",
|
||
|
"super",
|
||
|
"synchronized",
|
||
|
"throws",
|
||
|
"transient",
|
||
|
"volatile"
|
||
|
]);
|
||
|
|
||
|
var KEYWORDS_BEFORE_EXPRESSION = array_to_hash([
|
||
|
"return",
|
||
|
"new",
|
||
|
"delete",
|
||
|
"throw",
|
||
|
"else",
|
||
|
"case"
|
||
|
]);
|
||
|
|
||
|
var KEYWORDS_ATOM = array_to_hash([
|
||
|
"false",
|
||
|
"null",
|
||
|
"true",
|
||
|
"undefined"
|
||
|
]);
|
||
|
|
||
|
var OPERATOR_CHARS = array_to_hash(characters("+-*&%=<>!?|~^"));
|
||
|
|
||
|
var RE_HEX_NUMBER = /^0x[0-9a-f]+$/i;
|
||
|
var RE_OCT_NUMBER = /^0[0-7]+$/;
|
||
|
var RE_DEC_NUMBER = /^\d*\.?\d*(?:e[+-]?\d*(?:\d\.?|\.?\d)\d*)?$/i;
|
||
|
|
||
|
var OPERATORS = array_to_hash([
|
||
|
"in",
|
||
|
"instanceof",
|
||
|
"typeof",
|
||
|
"new",
|
||
|
"void",
|
||
|
"delete",
|
||
|
"++",
|
||
|
"--",
|
||
|
"+",
|
||
|
"-",
|
||
|
"!",
|
||
|
"~",
|
||
|
"&",
|
||
|
"|",
|
||
|
"^",
|
||
|
"*",
|
||
|
"/",
|
||
|
"%",
|
||
|
">>",
|
||
|
"<<",
|
||
|
">>>",
|
||
|
"<",
|
||
|
">",
|
||
|
"<=",
|
||
|
">=",
|
||
|
"==",
|
||
|
"===",
|
||
|
"!=",
|
||
|
"!==",
|
||
|
"?",
|
||
|
"=",
|
||
|
"+=",
|
||
|
"-=",
|
||
|
"/=",
|
||
|
"*=",
|
||
|
"%=",
|
||
|
">>=",
|
||
|
"<<=",
|
||
|
">>>=",
|
||
|
"%=",
|
||
|
"|=",
|
||
|
"^=",
|
||
|
"&=",
|
||
|
"&&",
|
||
|
"||"
|
||
|
]);
|
||
|
|
||
|
var WHITESPACE_CHARS = array_to_hash(characters(" \n\r\t"));
|
||
|
|
||
|
var PUNC_BEFORE_EXPRESSION = array_to_hash(characters("[{}(,.;:"));
|
||
|
|
||
|
var PUNC_CHARS = array_to_hash(characters("[]{}(),;:"));
|
||
|
|
||
|
var REGEXP_MODIFIERS = array_to_hash(characters("gmsiy"));
|
||
|
|
||
|
/* -----[ Tokenizer ]----- */
|
||
|
|
||
|
function is_alphanumeric_char(ch) {
|
||
|
ch = ch.charCodeAt(0);
|
||
|
return (ch >= 48 && ch <= 57) ||
|
||
|
(ch >= 65 && ch <= 90) ||
|
||
|
(ch >= 97 && ch <= 122);
|
||
|
};
|
||
|
|
||
|
function is_identifier_char(ch) {
|
||
|
return is_alphanumeric_char(ch) || ch == "$" || ch == "_";
|
||
|
};
|
||
|
|
||
|
function is_digit(ch) {
|
||
|
ch = ch.charCodeAt(0);
|
||
|
return ch >= 48 && ch <= 57;
|
||
|
};
|
||
|
|
||
|
function parse_js_number(num) {
|
||
|
if (RE_HEX_NUMBER.test(num)) {
|
||
|
return parseInt(num.substr(2), 16);
|
||
|
} else if (RE_OCT_NUMBER.test(num)) {
|
||
|
return parseInt(num.substr(1), 8);
|
||
|
} else if (RE_DEC_NUMBER.test(num)) {
|
||
|
return parseFloat(num);
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function JS_Parse_Error(message, line, col, pos) {
|
||
|
this.message = message;
|
||
|
this.line = line;
|
||
|
this.col = col;
|
||
|
this.pos = pos;
|
||
|
try {
|
||
|
({})();
|
||
|
} catch(ex) {
|
||
|
this.stack = ex.stack;
|
||
|
};
|
||
|
};
|
||
|
|
||
|
JS_Parse_Error.prototype.toString = function() {
|
||
|
return this.message + " (line: " + this.line + ", col: " + this.col + ", pos: " + this.pos + ")" + "\n\n" + this.stack;
|
||
|
};
|
||
|
|
||
|
function js_error(message, line, col, pos) {
|
||
|
throw new JS_Parse_Error(message, line, col, pos);
|
||
|
};
|
||
|
|
||
|
function is_token(token, type, val) {
|
||
|
return token.type == type && (val == null || token.value == val);
|
||
|
};
|
||
|
|
||
|
var EX_EOF = {};
|
||
|
|
||
|
function tokenizer($TEXT) {
|
||
|
|
||
|
var S = {
|
||
|
text : $TEXT.replace(/\r\n?|[\n\u2028\u2029]/g, "\n").replace(/^\uFEFF/, ''),
|
||
|
pos : 0,
|
||
|
tokpos : 0,
|
||
|
line : 0,
|
||
|
tokline : 0,
|
||
|
col : 0,
|
||
|
tokcol : 0,
|
||
|
newline_before : false,
|
||
|
regex_allowed : false,
|
||
|
comments_before : []
|
||
|
};
|
||
|
|
||
|
function peek() { return S.text.charAt(S.pos); };
|
||
|
|
||
|
function next(signal_eof) {
|
||
|
var ch = S.text.charAt(S.pos++);
|
||
|
if (signal_eof && !ch)
|
||
|
throw EX_EOF;
|
||
|
if (ch == "\n") {
|
||
|
S.newline_before = true;
|
||
|
++S.line;
|
||
|
S.col = 0;
|
||
|
} else {
|
||
|
++S.col;
|
||
|
}
|
||
|
return ch;
|
||
|
};
|
||
|
|
||
|
function eof() {
|
||
|
return !S.peek();
|
||
|
};
|
||
|
|
||
|
function find(what, signal_eof) {
|
||
|
var pos = S.text.indexOf(what, S.pos);
|
||
|
if (signal_eof && pos == -1) throw EX_EOF;
|
||
|
return pos;
|
||
|
};
|
||
|
|
||
|
function start_token() {
|
||
|
S.tokline = S.line;
|
||
|
S.tokcol = S.col;
|
||
|
S.tokpos = S.pos;
|
||
|
};
|
||
|
|
||
|
function token(type, value, is_comment) {
|
||
|
S.regex_allowed = ((type == "operator" && !HOP(UNARY_POSTFIX, value)) ||
|
||
|
(type == "keyword" && HOP(KEYWORDS_BEFORE_EXPRESSION, value)) ||
|
||
|
(type == "punc" && HOP(PUNC_BEFORE_EXPRESSION, value)));
|
||
|
var ret = {
|
||
|
type : type,
|
||
|
value : value,
|
||
|
line : S.tokline,
|
||
|
col : S.tokcol,
|
||
|
pos : S.tokpos,
|
||
|
nlb : S.newline_before
|
||
|
};
|
||
|
if (!is_comment) {
|
||
|
ret.comments_before = S.comments_before;
|
||
|
S.comments_before = [];
|
||
|
}
|
||
|
S.newline_before = false;
|
||
|
return ret;
|
||
|
};
|
||
|
|
||
|
function skip_whitespace() {
|
||
|
while (HOP(WHITESPACE_CHARS, peek()))
|
||
|
next();
|
||
|
};
|
||
|
|
||
|
function read_while(pred) {
|
||
|
var ret = "", ch = peek(), i = 0;
|
||
|
while (ch && pred(ch, i++)) {
|
||
|
ret += next();
|
||
|
ch = peek();
|
||
|
}
|
||
|
return ret;
|
||
|
};
|
||
|
|
||
|
function parse_error(err) {
|
||
|
js_error(err, S.tokline, S.tokcol, S.tokpos);
|
||
|
};
|
||
|
|
||
|
function read_num(prefix) {
|
||
|
var has_e = false, after_e = false, has_x = false, has_dot = prefix == ".";
|
||
|
var num = read_while(function(ch, i){
|
||
|
if (ch == "x" || ch == "X") {
|
||
|
if (has_x) return false;
|
||
|
return has_x = true;
|
||
|
}
|
||
|
if (!has_x && (ch == "E" || ch == "e")) {
|
||
|
if (has_e) return false;
|
||
|
return has_e = after_e = true;
|
||
|
}
|
||
|
if (ch == "-") {
|
||
|
if (after_e || (i == 0 && !prefix)) return true;
|
||
|
return false;
|
||
|
}
|
||
|
if (ch == "+") return after_e;
|
||
|
after_e = false;
|
||
|
if (ch == ".") {
|
||
|
if (!has_dot)
|
||
|
return has_dot = true;
|
||
|
return false;
|
||
|
}
|
||
|
return is_alphanumeric_char(ch);
|
||
|
});
|
||
|
if (prefix)
|
||
|
num = prefix + num;
|
||
|
var valid = parse_js_number(num);
|
||
|
if (!isNaN(valid)) {
|
||
|
return token("num", valid);
|
||
|
} else {
|
||
|
parse_error("Invalid syntax: " + num);
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function read_escaped_char() {
|
||
|
var ch = next(true);
|
||
|
switch (ch) {
|
||
|
case "n" : return "\n";
|
||
|
case "r" : return "\r";
|
||
|
case "t" : return "\t";
|
||
|
case "b" : return "\b";
|
||
|
case "v" : return "\v";
|
||
|
case "f" : return "\f";
|
||
|
case "0" : return "\0";
|
||
|
case "x" : return String.fromCharCode(hex_bytes(2));
|
||
|
case "u" : return String.fromCharCode(hex_bytes(4));
|
||
|
default : return ch;
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function hex_bytes(n) {
|
||
|
var num = 0;
|
||
|
for (; n > 0; --n) {
|
||
|
var digit = parseInt(next(true), 16);
|
||
|
if (isNaN(digit))
|
||
|
parse_error("Invalid hex-character pattern in string");
|
||
|
num = (num << 4) | digit;
|
||
|
}
|
||
|
return num;
|
||
|
};
|
||
|
|
||
|
function read_string() {
|
||
|
return with_eof_error("Unterminated string constant", function(){
|
||
|
var quote = next(), ret = "";
|
||
|
for (;;) {
|
||
|
var ch = next(true);
|
||
|
if (ch == "\\") ch = read_escaped_char();
|
||
|
else if (ch == quote) break;
|
||
|
ret += ch;
|
||
|
}
|
||
|
return token("string", ret);
|
||
|
});
|
||
|
};
|
||
|
|
||
|
function read_line_comment() {
|
||
|
next();
|
||
|
var i = find("\n"), ret;
|
||
|
if (i == -1) {
|
||
|
ret = S.text.substr(S.pos);
|
||
|
S.pos = S.text.length;
|
||
|
} else {
|
||
|
ret = S.text.substring(S.pos, i);
|
||
|
S.pos = i;
|
||
|
}
|
||
|
return token("comment1", ret, true);
|
||
|
};
|
||
|
|
||
|
function read_multiline_comment() {
|
||
|
next();
|
||
|
return with_eof_error("Unterminated multiline comment", function(){
|
||
|
var i = find("*/", true),
|
||
|
text = S.text.substring(S.pos, i),
|
||
|
tok = token("comment2", text, true);
|
||
|
S.pos = i + 2;
|
||
|
S.line += text.split("\n").length - 1;
|
||
|
S.newline_before = text.indexOf("\n") >= 0;
|
||
|
return tok;
|
||
|
});
|
||
|
};
|
||
|
|
||
|
function read_regexp() {
|
||
|
return with_eof_error("Unterminated regular expression", function(){
|
||
|
var prev_backslash = false, regexp = "", ch, in_class = false;
|
||
|
while ((ch = next(true))) if (prev_backslash) {
|
||
|
regexp += "\\" + ch;
|
||
|
prev_backslash = false;
|
||
|
} else if (ch == "[") {
|
||
|
in_class = true;
|
||
|
regexp += ch;
|
||
|
} else if (ch == "]" && in_class) {
|
||
|
in_class = false;
|
||
|
regexp += ch;
|
||
|
} else if (ch == "/" && !in_class) {
|
||
|
break;
|
||
|
} else if (ch == "\\") {
|
||
|
prev_backslash = true;
|
||
|
} else {
|
||
|
regexp += ch;
|
||
|
}
|
||
|
var mods = read_while(function(ch){
|
||
|
return HOP(REGEXP_MODIFIERS, ch);
|
||
|
});
|
||
|
return token("regexp", [ regexp, mods ]);
|
||
|
});
|
||
|
};
|
||
|
|
||
|
function read_operator(prefix) {
|
||
|
function grow(op) {
|
||
|
if (!peek()) return op;
|
||
|
var bigger = op + peek();
|
||
|
if (HOP(OPERATORS, bigger)) {
|
||
|
next();
|
||
|
return grow(bigger);
|
||
|
} else {
|
||
|
return op;
|
||
|
}
|
||
|
};
|
||
|
return token("operator", grow(prefix || next()));
|
||
|
};
|
||
|
|
||
|
function handle_slash() {
|
||
|
next();
|
||
|
var regex_allowed = S.regex_allowed;
|
||
|
switch (peek()) {
|
||
|
case "/":
|
||
|
S.comments_before.push(read_line_comment());
|
||
|
S.regex_allowed = regex_allowed;
|
||
|
return next_token();
|
||
|
case "*":
|
||
|
S.comments_before.push(read_multiline_comment());
|
||
|
S.regex_allowed = regex_allowed;
|
||
|
return next_token();
|
||
|
}
|
||
|
return S.regex_allowed ? read_regexp() : read_operator("/");
|
||
|
};
|
||
|
|
||
|
function handle_dot() {
|
||
|
next();
|
||
|
return is_digit(peek())
|
||
|
? read_num(".")
|
||
|
: token("punc", ".");
|
||
|
};
|
||
|
|
||
|
function read_word() {
|
||
|
var word = read_while(is_identifier_char);
|
||
|
return !HOP(KEYWORDS, word)
|
||
|
? token("name", word)
|
||
|
: HOP(OPERATORS, word)
|
||
|
? token("operator", word)
|
||
|
: HOP(KEYWORDS_ATOM, word)
|
||
|
? token("atom", word)
|
||
|
: token("keyword", word);
|
||
|
};
|
||
|
|
||
|
function with_eof_error(eof_error, cont) {
|
||
|
try {
|
||
|
return cont();
|
||
|
} catch(ex) {
|
||
|
if (ex === EX_EOF) parse_error(eof_error);
|
||
|
else throw ex;
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function next_token(force_regexp) {
|
||
|
if (force_regexp)
|
||
|
return read_regexp();
|
||
|
skip_whitespace();
|
||
|
start_token();
|
||
|
var ch = peek();
|
||
|
if (!ch) return token("eof");
|
||
|
if (is_digit(ch)) return read_num();
|
||
|
if (ch == '"' || ch == "'") return read_string();
|
||
|
if (HOP(PUNC_CHARS, ch)) return token("punc", next());
|
||
|
if (ch == ".") return handle_dot();
|
||
|
if (ch == "/") return handle_slash();
|
||
|
if (HOP(OPERATOR_CHARS, ch)) return read_operator();
|
||
|
if (is_identifier_char(ch)) return read_word();
|
||
|
parse_error("Unexpected character '" + ch + "'");
|
||
|
};
|
||
|
|
||
|
next_token.context = function(nc) {
|
||
|
if (nc) S = nc;
|
||
|
return S;
|
||
|
};
|
||
|
|
||
|
return next_token;
|
||
|
|
||
|
};
|
||
|
|
||
|
/* -----[ Parser (constants) ]----- */
|
||
|
|
||
|
var UNARY_PREFIX = array_to_hash([
|
||
|
"typeof",
|
||
|
"void",
|
||
|
"delete",
|
||
|
"--",
|
||
|
"++",
|
||
|
"!",
|
||
|
"~",
|
||
|
"-",
|
||
|
"+"
|
||
|
]);
|
||
|
|
||
|
var UNARY_POSTFIX = array_to_hash([ "--", "++" ]);
|
||
|
|
||
|
var ASSIGNMENT = (function(a, ret, i){
|
||
|
while (i < a.length) {
|
||
|
ret[a[i]] = a[i].substr(0, a[i].length - 1);
|
||
|
i++;
|
||
|
}
|
||
|
return ret;
|
||
|
})(
|
||
|
["+=", "-=", "/=", "*=", "%=", ">>=", "<<=", ">>>=", "|=", "^=", "&="],
|
||
|
{ "=": true },
|
||
|
0
|
||
|
);
|
||
|
|
||
|
var PRECEDENCE = (function(a, ret){
|
||
|
for (var i = 0, n = 1; i < a.length; ++i, ++n) {
|
||
|
var b = a[i];
|
||
|
for (var j = 0; j < b.length; ++j) {
|
||
|
ret[b[j]] = n;
|
||
|
}
|
||
|
}
|
||
|
return ret;
|
||
|
})(
|
||
|
[
|
||
|
["||"],
|
||
|
["&&"],
|
||
|
["|"],
|
||
|
["^"],
|
||
|
["&"],
|
||
|
["==", "===", "!=", "!=="],
|
||
|
["<", ">", "<=", ">=", "in", "instanceof"],
|
||
|
[">>", "<<", ">>>"],
|
||
|
["+", "-"],
|
||
|
["*", "/", "%"]
|
||
|
],
|
||
|
{}
|
||
|
);
|
||
|
|
||
|
var STATEMENTS_WITH_LABELS = array_to_hash([ "for", "do", "while", "switch" ]);
|
||
|
|
||
|
var ATOMIC_START_TOKEN = array_to_hash([ "atom", "num", "string", "regexp", "name" ]);
|
||
|
|
||
|
/* -----[ Parser ]----- */
|
||
|
|
||
|
function NodeWithToken(str, start, end) {
|
||
|
this.name = str;
|
||
|
this.start = start;
|
||
|
this.end = end;
|
||
|
};
|
||
|
|
||
|
NodeWithToken.prototype.toString = function() { return this.name; };
|
||
|
|
||
|
function parse($TEXT, strict_mode, embed_tokens) {
|
||
|
|
||
|
var S = {
|
||
|
input : typeof $TEXT == "string" ? tokenizer($TEXT, true) : $TEXT,
|
||
|
token : null,
|
||
|
prev : null,
|
||
|
peeked : null,
|
||
|
in_function : 0,
|
||
|
in_loop : 0,
|
||
|
labels : []
|
||
|
};
|
||
|
|
||
|
S.token = next();
|
||
|
|
||
|
function is(type, value) {
|
||
|
return is_token(S.token, type, value);
|
||
|
};
|
||
|
|
||
|
function peek() { return S.peeked || (S.peeked = S.input()); };
|
||
|
|
||
|
function next() {
|
||
|
S.prev = S.token;
|
||
|
if (S.peeked) {
|
||
|
S.token = S.peeked;
|
||
|
S.peeked = null;
|
||
|
} else {
|
||
|
S.token = S.input();
|
||
|
}
|
||
|
return S.token;
|
||
|
};
|
||
|
|
||
|
function prev() {
|
||
|
return S.prev;
|
||
|
};
|
||
|
|
||
|
function croak(msg, line, col, pos) {
|
||
|
var ctx = S.input.context();
|
||
|
js_error(msg,
|
||
|
line != null ? line : ctx.tokline,
|
||
|
col != null ? col : ctx.tokcol,
|
||
|
pos != null ? pos : ctx.tokpos);
|
||
|
};
|
||
|
|
||
|
function token_error(token, msg) {
|
||
|
croak(msg, token.line, token.col);
|
||
|
};
|
||
|
|
||
|
function unexpected(token) {
|
||
|
if (token == null)
|
||
|
token = S.token;
|
||
|
token_error(token, "Unexpected token: " + token.type + " (" + token.value + ")");
|
||
|
};
|
||
|
|
||
|
function expect_token(type, val) {
|
||
|
if (is(type, val)) {
|
||
|
return next();
|
||
|
}
|
||
|
token_error(S.token, "Unexpected token " + S.token.type + ", expected " + type);
|
||
|
};
|
||
|
|
||
|
function expect(punc) { return expect_token("punc", punc); };
|
||
|
|
||
|
function can_insert_semicolon() {
|
||
|
return !strict_mode && (
|
||
|
S.token.nlb || is("eof") || is("punc", "}")
|
||
|
);
|
||
|
};
|
||
|
|
||
|
function semicolon() {
|
||
|
if (is("punc", ";")) next();
|
||
|
else if (!can_insert_semicolon()) unexpected();
|
||
|
};
|
||
|
|
||
|
function as() {
|
||
|
return slice(arguments);
|
||
|
};
|
||
|
|
||
|
function parenthesised() {
|
||
|
expect("(");
|
||
|
var ex = expression();
|
||
|
expect(")");
|
||
|
return ex;
|
||
|
};
|
||
|
|
||
|
function add_tokens(str, start, end) {
|
||
|
return new NodeWithToken(str, start, end);
|
||
|
};
|
||
|
|
||
|
var statement = embed_tokens ? function() {
|
||
|
var start = S.token;
|
||
|
var stmt = $statement();
|
||
|
stmt[0] = add_tokens(stmt[0], start, prev());
|
||
|
return stmt;
|
||
|
} : $statement;
|
||
|
|
||
|
function $statement() {
|
||
|
if (is("operator", "/")) {
|
||
|
S.peeked = null;
|
||
|
S.token = S.input(true); // force regexp
|
||
|
}
|
||
|
switch (S.token.type) {
|
||
|
case "num":
|
||
|
case "string":
|
||
|
case "regexp":
|
||
|
case "operator":
|
||
|
case "atom":
|
||
|
return simple_statement();
|
||
|
|
||
|
case "name":
|
||
|
return is_token(peek(), "punc", ":")
|
||
|
? labeled_statement(prog1(S.token.value, next, next))
|
||
|
: simple_statement();
|
||
|
|
||
|
case "punc":
|
||
|
switch (S.token.value) {
|
||
|
case "{":
|
||
|
return as("block", block_());
|
||
|
case "[":
|
||
|
case "(":
|
||
|
return simple_statement();
|
||
|
case ";":
|
||
|
next();
|
||
|
return as("block");
|
||
|
default:
|
||
|
unexpected();
|
||
|
}
|
||
|
|
||
|
case "keyword":
|
||
|
switch (prog1(S.token.value, next)) {
|
||
|
case "break":
|
||
|
return break_cont("break");
|
||
|
|
||
|
case "continue":
|
||
|
return break_cont("continue");
|
||
|
|
||
|
case "debugger":
|
||
|
semicolon();
|
||
|
return as("debugger");
|
||
|
|
||
|
case "do":
|
||
|
return (function(body){
|
||
|
expect_token("keyword", "while");
|
||
|
return as("do", prog1(parenthesised, semicolon), body);
|
||
|
})(in_loop(statement));
|
||
|
|
||
|
case "for":
|
||
|
return for_();
|
||
|
|
||
|
case "function":
|
||
|
return function_(true);
|
||
|
|
||
|
case "if":
|
||
|
return if_();
|
||
|
|
||
|
case "return":
|
||
|
if (S.in_function == 0)
|
||
|
croak("'return' outside of function");
|
||
|
return as("return",
|
||
|
is("punc", ";")
|
||
|
? (next(), null)
|
||
|
: can_insert_semicolon()
|
||
|
? null
|
||
|
: prog1(expression, semicolon));
|
||
|
|
||
|
case "switch":
|
||
|
return as("switch", parenthesised(), switch_block_());
|
||
|
|
||
|
case "throw":
|
||
|
return as("throw", prog1(expression, semicolon));
|
||
|
|
||
|
case "try":
|
||
|
return try_();
|
||
|
|
||
|
case "var":
|
||
|
return prog1(var_, semicolon);
|
||
|
|
||
|
case "const":
|
||
|
return prog1(const_, semicolon);
|
||
|
|
||
|
case "while":
|
||
|
return as("while", parenthesised(), in_loop(statement));
|
||
|
|
||
|
case "with":
|
||
|
return as("with", parenthesised(), statement());
|
||
|
|
||
|
default:
|
||
|
unexpected();
|
||
|
}
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function labeled_statement(label) {
|
||
|
S.labels.push(label);
|
||
|
var start = S.token, stat = statement();
|
||
|
if (strict_mode && !HOP(STATEMENTS_WITH_LABELS, stat[0]))
|
||
|
unexpected(start);
|
||
|
S.labels.pop();
|
||
|
return as("label", label, stat);
|
||
|
};
|
||
|
|
||
|
function simple_statement() {
|
||
|
return as("stat", prog1(expression, semicolon));
|
||
|
};
|
||
|
|
||
|
function break_cont(type) {
|
||
|
var name = is("name") ? S.token.value : null;
|
||
|
if (name != null) {
|
||
|
next();
|
||
|
if (!member(name, S.labels))
|
||
|
croak("Label " + name + " without matching loop or statement");
|
||
|
}
|
||
|
else if (S.in_loop == 0)
|
||
|
croak(type + " not inside a loop or switch");
|
||
|
semicolon();
|
||
|
return as(type, name);
|
||
|
};
|
||
|
|
||
|
function for_() {
|
||
|
expect("(");
|
||
|
var has_var = is("keyword", "var");
|
||
|
if (has_var)
|
||
|
next();
|
||
|
if (is("name") && is_token(peek(), "operator", "in")) {
|
||
|
// for (i in foo)
|
||
|
var name = S.token.value;
|
||
|
next(); next();
|
||
|
var obj = expression();
|
||
|
expect(")");
|
||
|
return as("for-in", has_var, name, obj, in_loop(statement));
|
||
|
} else {
|
||
|
// classic for
|
||
|
var init = is("punc", ";") ? null : has_var ? var_() : expression();
|
||
|
expect(";");
|
||
|
var test = is("punc", ";") ? null : expression();
|
||
|
expect(";");
|
||
|
var step = is("punc", ")") ? null : expression();
|
||
|
expect(")");
|
||
|
return as("for", init, test, step, in_loop(statement));
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function function_(in_statement) {
|
||
|
var name = is("name") ? prog1(S.token.value, next) : null;
|
||
|
if (in_statement && !name)
|
||
|
unexpected();
|
||
|
expect("(");
|
||
|
return as(in_statement ? "defun" : "function",
|
||
|
name,
|
||
|
// arguments
|
||
|
(function(first, a){
|
||
|
while (!is("punc", ")")) {
|
||
|
if (first) first = false; else expect(",");
|
||
|
if (!is("name")) unexpected();
|
||
|
a.push(S.token.value);
|
||
|
next();
|
||
|
}
|
||
|
next();
|
||
|
return a;
|
||
|
})(true, []),
|
||
|
// body
|
||
|
(function(){
|
||
|
++S.in_function;
|
||
|
var loop = S.in_loop;
|
||
|
S.in_loop = 0;
|
||
|
var a = block_();
|
||
|
--S.in_function;
|
||
|
S.in_loop = loop;
|
||
|
return a;
|
||
|
})());
|
||
|
};
|
||
|
|
||
|
function if_() {
|
||
|
var cond = parenthesised(), body = statement(), belse;
|
||
|
if (is("keyword", "else")) {
|
||
|
next();
|
||
|
belse = statement();
|
||
|
}
|
||
|
return as("if", cond, body, belse);
|
||
|
};
|
||
|
|
||
|
function block_() {
|
||
|
expect("{");
|
||
|
var a = [];
|
||
|
while (!is("punc", "}")) {
|
||
|
if (is("eof")) unexpected();
|
||
|
a.push(statement());
|
||
|
}
|
||
|
next();
|
||
|
return a;
|
||
|
};
|
||
|
|
||
|
var switch_block_ = curry(in_loop, function(){
|
||
|
expect("{");
|
||
|
var a = [], cur = null;
|
||
|
while (!is("punc", "}")) {
|
||
|
if (is("eof")) unexpected();
|
||
|
if (is("keyword", "case")) {
|
||
|
next();
|
||
|
cur = [];
|
||
|
a.push([ expression(), cur ]);
|
||
|
expect(":");
|
||
|
}
|
||
|
else if (is("keyword", "default")) {
|
||
|
next();
|
||
|
expect(":");
|
||
|
cur = [];
|
||
|
a.push([ null, cur ]);
|
||
|
}
|
||
|
else {
|
||
|
if (!cur) unexpected();
|
||
|
cur.push(statement());
|
||
|
}
|
||
|
}
|
||
|
next();
|
||
|
return a;
|
||
|
});
|
||
|
|
||
|
function try_() {
|
||
|
var body = block_(), bcatch, bfinally;
|
||
|
if (is("keyword", "catch")) {
|
||
|
next();
|
||
|
expect("(");
|
||
|
if (!is("name"))
|
||
|
croak("Name expected");
|
||
|
var name = S.token.value;
|
||
|
next();
|
||
|
expect(")");
|
||
|
bcatch = [ name, block_() ];
|
||
|
}
|
||
|
if (is("keyword", "finally")) {
|
||
|
next();
|
||
|
bfinally = block_();
|
||
|
}
|
||
|
if (!bcatch && !bfinally)
|
||
|
croak("Missing catch/finally blocks");
|
||
|
return as("try", body, bcatch, bfinally);
|
||
|
};
|
||
|
|
||
|
function vardefs() {
|
||
|
var a = [];
|
||
|
for (;;) {
|
||
|
if (!is("name"))
|
||
|
unexpected();
|
||
|
var name = S.token.value;
|
||
|
next();
|
||
|
if (is("operator", "=")) {
|
||
|
next();
|
||
|
a.push([ name, expression(false) ]);
|
||
|
} else {
|
||
|
a.push([ name ]);
|
||
|
}
|
||
|
if (!is("punc", ","))
|
||
|
break;
|
||
|
next();
|
||
|
}
|
||
|
return a;
|
||
|
};
|
||
|
|
||
|
function var_() {
|
||
|
return as("var", vardefs());
|
||
|
};
|
||
|
|
||
|
function const_() {
|
||
|
return as("const", vardefs());
|
||
|
};
|
||
|
|
||
|
function new_() {
|
||
|
var newexp = expr_atom(false), args;
|
||
|
if (is("punc", "(")) {
|
||
|
next();
|
||
|
args = expr_list(")");
|
||
|
} else {
|
||
|
args = [];
|
||
|
}
|
||
|
return subscripts(as("new", newexp, args), true);
|
||
|
};
|
||
|
|
||
|
function expr_atom(allow_calls) {
|
||
|
if (is("operator", "new")) {
|
||
|
next();
|
||
|
return new_();
|
||
|
}
|
||
|
if (is("operator") && HOP(UNARY_PREFIX, S.token.value)) {
|
||
|
return make_unary("unary-prefix",
|
||
|
prog1(S.token.value, next),
|
||
|
expr_atom(allow_calls));
|
||
|
}
|
||
|
if (is("punc")) {
|
||
|
switch (S.token.value) {
|
||
|
case "(":
|
||
|
next();
|
||
|
return subscripts(prog1(expression, curry(expect, ")")), allow_calls);
|
||
|
case "[":
|
||
|
next();
|
||
|
return subscripts(array_(), allow_calls);
|
||
|
case "{":
|
||
|
next();
|
||
|
return subscripts(object_(), allow_calls);
|
||
|
}
|
||
|
unexpected();
|
||
|
}
|
||
|
if (is("keyword", "function")) {
|
||
|
next();
|
||
|
return subscripts(function_(false), allow_calls);
|
||
|
}
|
||
|
if (HOP(ATOMIC_START_TOKEN, S.token.type)) {
|
||
|
var atom = S.token.type == "regexp"
|
||
|
? as("regexp", S.token.value[0], S.token.value[1])
|
||
|
: as(S.token.type, S.token.value);
|
||
|
return subscripts(prog1(atom, next), allow_calls);
|
||
|
}
|
||
|
unexpected();
|
||
|
};
|
||
|
|
||
|
function expr_list(closing, allow_trailing_comma, allow_empty) {
|
||
|
var first = true, a = [];
|
||
|
while (!is("punc", closing)) {
|
||
|
if (first) first = false; else expect(",");
|
||
|
if (allow_trailing_comma && is("punc", closing)) break;
|
||
|
if (is("punc", ",") && allow_empty) {
|
||
|
a.push([ "atom", "undefined" ]);
|
||
|
} else {
|
||
|
a.push(expression(false));
|
||
|
}
|
||
|
}
|
||
|
next();
|
||
|
return a;
|
||
|
};
|
||
|
|
||
|
function array_() {
|
||
|
return as("array", expr_list("]", !strict_mode, true));
|
||
|
};
|
||
|
|
||
|
function object_() {
|
||
|
var first = true, a = [];
|
||
|
while (!is("punc", "}")) {
|
||
|
if (first) first = false; else expect(",");
|
||
|
if (!strict_mode && is("punc", "}"))
|
||
|
// allow trailing comma
|
||
|
break;
|
||
|
var type = S.token.type;
|
||
|
var name = as_property_name();
|
||
|
if (type == "name" && (name == "get" || name == "set") && !is("punc", ":")) {
|
||
|
a.push([ as_name(), function_(false), name ]);
|
||
|
} else {
|
||
|
expect(":");
|
||
|
a.push([ name, expression(false) ]);
|
||
|
}
|
||
|
}
|
||
|
next();
|
||
|
return as("object", a);
|
||
|
};
|
||
|
|
||
|
function as_property_name() {
|
||
|
switch (S.token.type) {
|
||
|
case "num":
|
||
|
case "string":
|
||
|
return prog1(S.token.value, next);
|
||
|
}
|
||
|
return as_name();
|
||
|
};
|
||
|
|
||
|
function as_name() {
|
||
|
switch (S.token.type) {
|
||
|
case "name":
|
||
|
case "operator":
|
||
|
case "keyword":
|
||
|
case "atom":
|
||
|
return prog1(S.token.value, next);
|
||
|
default:
|
||
|
unexpected();
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function subscripts(expr, allow_calls) {
|
||
|
if (is("punc", ".")) {
|
||
|
next();
|
||
|
return subscripts(as("dot", expr, as_name()), allow_calls);
|
||
|
}
|
||
|
if (is("punc", "[")) {
|
||
|
next();
|
||
|
return subscripts(as("sub", expr, prog1(expression, curry(expect, "]"))), allow_calls);
|
||
|
}
|
||
|
if (allow_calls && is("punc", "(")) {
|
||
|
next();
|
||
|
return subscripts(as("call", expr, expr_list(")")), true);
|
||
|
}
|
||
|
if (allow_calls && is("operator") && HOP(UNARY_POSTFIX, S.token.value)) {
|
||
|
return prog1(curry(make_unary, "unary-postfix", S.token.value, expr),
|
||
|
next);
|
||
|
}
|
||
|
return expr;
|
||
|
};
|
||
|
|
||
|
function make_unary(tag, op, expr) {
|
||
|
if ((op == "++" || op == "--") && !is_assignable(expr))
|
||
|
croak("Invalid use of " + op + " operator");
|
||
|
return as(tag, op, expr);
|
||
|
};
|
||
|
|
||
|
function expr_op(left, min_prec) {
|
||
|
var op = is("operator") ? S.token.value : null;
|
||
|
var prec = op != null ? PRECEDENCE[op] : null;
|
||
|
if (prec != null && prec > min_prec) {
|
||
|
next();
|
||
|
var right = expr_op(expr_atom(true), prec);
|
||
|
return expr_op(as("binary", op, left, right), min_prec);
|
||
|
}
|
||
|
return left;
|
||
|
};
|
||
|
|
||
|
function expr_ops() {
|
||
|
return expr_op(expr_atom(true), 0);
|
||
|
};
|
||
|
|
||
|
function maybe_conditional() {
|
||
|
var expr = expr_ops();
|
||
|
if (is("operator", "?")) {
|
||
|
next();
|
||
|
var yes = expression(false);
|
||
|
expect(":");
|
||
|
return as("conditional", expr, yes, expression(false));
|
||
|
}
|
||
|
return expr;
|
||
|
};
|
||
|
|
||
|
function is_assignable(expr) {
|
||
|
switch (expr[0]) {
|
||
|
case "dot":
|
||
|
case "sub":
|
||
|
return true;
|
||
|
case "name":
|
||
|
return expr[1] != "this";
|
||
|
}
|
||
|
};
|
||
|
|
||
|
function maybe_assign() {
|
||
|
var left = maybe_conditional(), val = S.token.value;
|
||
|
if (is("operator") && HOP(ASSIGNMENT, val)) {
|
||
|
if (is_assignable(left)) {
|
||
|
next();
|
||
|
return as("assign", ASSIGNMENT[val], left, maybe_assign());
|
||
|
}
|
||
|
croak("Invalid assignment");
|
||
|
}
|
||
|
return left;
|
||
|
};
|
||
|
|
||
|
function expression(commas) {
|
||
|
if (arguments.length == 0)
|
||
|
commas = true;
|
||
|
var expr = maybe_assign();
|
||
|
if (commas && is("punc", ",")) {
|
||
|
next();
|
||
|
return as("seq", expr, expression());
|
||
|
}
|
||
|
return expr;
|
||
|
};
|
||
|
|
||
|
function in_loop(cont) {
|
||
|
try {
|
||
|
++S.in_loop;
|
||
|
return cont();
|
||
|
} finally {
|
||
|
--S.in_loop;
|
||
|
}
|
||
|
};
|
||
|
|
||
|
return as("toplevel", (function(a){
|
||
|
while (!is("eof"))
|
||
|
a.push(statement());
|
||
|
return a;
|
||
|
})([]));
|
||
|
|
||
|
};
|
||
|
|
||
|
/* -----[ Utilities ]----- */
|
||
|
|
||
|
function curry(f) {
|
||
|
var args = slice(arguments, 1);
|
||
|
return function() { return f.apply(this, args.concat(slice(arguments))); };
|
||
|
};
|
||
|
|
||
|
function prog1(ret) {
|
||
|
if (ret instanceof Function)
|
||
|
ret = ret();
|
||
|
for (var i = 1, n = arguments.length; --n > 0; ++i)
|
||
|
arguments[i]();
|
||
|
return ret;
|
||
|
};
|
||
|
|
||
|
function array_to_hash(a) {
|
||
|
var ret = {};
|
||
|
for (var i = 0; i < a.length; ++i)
|
||
|
ret[a[i]] = true;
|
||
|
return ret;
|
||
|
};
|
||
|
|
||
|
function slice(a, start) {
|
||
|
return Array.prototype.slice.call(a, start == null ? 0 : start);
|
||
|
};
|
||
|
|
||
|
function characters(str) {
|
||
|
return str.split("");
|
||
|
};
|
||
|
|
||
|
function member(name, array) {
|
||
|
for (var i = array.length; --i >= 0;)
|
||
|
if (array[i] === name)
|
||
|
return true;
|
||
|
return false;
|
||
|
};
|
||
|
|
||
|
function HOP(obj, prop) {
|
||
|
return Object.prototype.hasOwnProperty.call(obj, prop);
|
||
|
};
|
||
|
|
||
|
/* -----[ Exports ]----- */
|
||
|
|
||
|
exports.tokenizer = tokenizer;
|
||
|
exports.parse = parse;
|
||
|
exports.slice = slice;
|
||
|
exports.curry = curry;
|
||
|
exports.member = member;
|
||
|
exports.array_to_hash = array_to_hash;
|
||
|
exports.PRECEDENCE = PRECEDENCE;
|
||
|
exports.KEYWORDS_ATOM = KEYWORDS_ATOM;
|
||
|
exports.RESERVED_WORDS = RESERVED_WORDS;
|
||
|
exports.KEYWORDS = KEYWORDS;
|
||
|
exports.ATOMIC_START_TOKEN = ATOMIC_START_TOKEN;
|
||
|
exports.OPERATORS = OPERATORS;
|
||
|
exports.is_alphanumeric_char = is_alphanumeric_char;
|
||
|
exports.is_identifier_char = is_identifier_char;
|