X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;ds=sidebyside;f=src%2Forg%2Fxwt%2Fjs%2FParser.java;h=1914cc7e3226acaa6f8f57fe9ae85373515ed1de;hb=16c24a73c1c1b2955db0bbbaf5a940215329bca1;hp=f70a13d7d8f99d5d37fc8cef2a887a503000aa32;hpb=11fb6d3f994008abe6c16b4402c3904a543e240e;p=org.ibex.core.git diff --git a/src/org/xwt/js/Parser.java b/src/org/xwt/js/Parser.java index f70a13d..1914cc7 100644 --- a/src/org/xwt/js/Parser.java +++ b/src/org/xwt/js/Parser.java @@ -1,29 +1,81 @@ -// Copyright 2002 Adam Megacz, see the COPYING file for licensing [GPL] +// Copyright 2004 Adam Megacz, see the COPYING file for licensing [GPL] package org.xwt.js; import org.xwt.util.*; import java.io.*; +/** + * Parses a stream of lexed tokens into a tree of JSFunction's. + * + * There are three kinds of things we parse: blocks, statements, and + * expressions. + * + * - Expressions are a special type of statement that evaluates to a + * value (for example, "break" is not an expression, * but "3+2" + * is). Some tokens sequences start expressions (for * example, + * literal numbers) and others continue an expression which * has + * already been begun (for example, '+'). Finally, some * + * expressions are valid targets for an assignment operation; after + * * each of these expressions, continueExprAfterAssignable() is + * called * to check for an assignment operation. + * + * - A statement ends with a semicolon and does not return a value. + * + * - A block is a single statement or a sequence of statements + * surrounded by curly braces. + * + * Each parsing method saves the parserLine before doing its actual + * work and restores it afterwards. This ensures that parsing a + * subexpression does not modify the line number until a token + * *after* the subexpression has been consumed by the parent + * expression. + * + * Technically it would be a better design for this class to build an + * intermediate parse tree and use that to emit bytecode. Here's the + * tradeoff: + * + * Advantages of building a parse tree: + * - easier to apply optimizations + * - would let us handle more sophisticated languages than JavaScript + * + * Advantages of leaving out the parse tree + * - faster compilation + * - less load on the garbage collector + * - much simpler code, easier to understand + * - less error-prone + * + * Fortunately JS is such a simple language that we can get away with + * the half-assed approach and still produce a working, complete + * compiler. + * + * The bytecode language emitted doesn't really cause any appreciable + * semantic loss, and is itself a parseable language very similar to + * Forth or a postfix variant of LISP. This means that the bytecode + * can be transformed into a parse tree, which can be manipulated. + * So if we ever want to add an optimizer, it could easily be done by + * producing a parse tree from the bytecode, optimizing that tree, + * and then re-emitting the bytecode. The parse tree node class + * would also be much simpler since the bytecode language has so few + * operators. + * + * Actually, the above paragraph is slightly inaccurate -- there are + * places where we push a value and then perform an arbitrary number + * of operations using it before popping it; this doesn't parse well. + * But these cases are clearly marked and easy to change if we do + * need to move to a parse tree format. + */ +class Parser extends Lexer implements ByteCodes { -/** parses a stream of lexed tokens into a tree of Expr's */ -public class Parser extends Lexer { // Constructors ////////////////////////////////////////////////////// - public Parser(Reader r, String sourceName, int line) throws IOException { - super(r); - this.sourceName = sourceName; - this.line = line; - } + public Parser(Reader r, String sourceName, int line) throws IOException { super(r, sourceName, line); } + /** for debugging */ public static void main(String[] s) throws Exception { - Parser p = new Parser(new InputStreamReader(System.in), "stdin", 0); - while(true) { - Expr block = p.parseBlock(false); - if (block == null) return; - System.out.println(block); - if (p.peekToken() == -1) return; - } + JS block = JS.fromReader("stdin", 0, new InputStreamReader(System.in)); + if (block == null) return; + System.out.println(block); } @@ -31,768 +83,868 @@ public class Parser extends Lexer { static byte[] precedence = new byte[MAX_TOKEN + 1]; static boolean[] isRightAssociative = new boolean[MAX_TOKEN + 1]; + // Use this as the precedence when we want anything up to the comma + private final static int NO_COMMA = 2; static { - precedence[ASSIGN] = 1; - isRightAssociative[ASSIGN] = true; - precedence[HOOK] = 2; - precedence[COMMA] = 3; - precedence[GT] = precedence[GE] = 4; - precedence[OR] = precedence[AND] = 5; - precedence[BITOR] = 6; - precedence[BITXOR] = 7; - precedence[BITAND] = 8; - precedence[EQ] = precedence[NE] = 9; - precedence[LT] = precedence[LE] = 10; - precedence[SHEQ] = precedence[SHNE] = 11; - precedence[LSH] = precedence[RSH] = precedence[URSH] = 12; - precedence[ADD] = precedence[SUB] = 13; - precedence[MUL] = precedence[DIV] = precedence[MOD] = 14; - precedence[BITNOT] = precedence[INSTANCEOF] = 15; - precedence[INC] = precedence[DEC] = 16; - precedence[LP] = 17; - precedence[LB] = 18; - precedence[DOT] = 19; + isRightAssociative[ASSIGN] = + isRightAssociative[ASSIGN_BITOR] = + isRightAssociative[ASSIGN_BITXOR] = + isRightAssociative[ASSIGN_BITAND] = + isRightAssociative[ASSIGN_LSH] = + isRightAssociative[ASSIGN_RSH] = + isRightAssociative[ASSIGN_URSH] = + isRightAssociative[ASSIGN_ADD] = + isRightAssociative[ASSIGN_SUB] = + isRightAssociative[ASSIGN_MUL] = + isRightAssociative[ASSIGN_DIV] = + isRightAssociative[ASSIGN_MOD] = + isRightAssociative[ADD_TRAP] = + isRightAssociative[DEL_TRAP] = + true; + + precedence[COMMA] = 1; + // 2 is intentionally left unassigned. we use minPrecedence==2 for comma separated lists + precedence[ASSIGN] = + precedence[ASSIGN_BITOR] = + precedence[ASSIGN_BITXOR] = + precedence[ASSIGN_BITAND] = + precedence[ASSIGN_LSH] = + precedence[ASSIGN_RSH] = + precedence[ASSIGN_URSH] = + precedence[ASSIGN_ADD] = + precedence[ASSIGN_SUB] = + precedence[ASSIGN_MUL] = + precedence[ASSIGN_DIV] = + precedence[ADD_TRAP] = + precedence[DEL_TRAP] = + precedence[ASSIGN_MOD] = 3; + precedence[HOOK] = 4; + precedence[OR] = 5; + precedence[AND] = 6; + precedence[BITOR] = 7; + precedence[BITXOR] = 8; + precedence[BITAND] = 9; + precedence[EQ] = precedence[NE] = precedence[SHEQ] = precedence[SHNE] = 10; + precedence[LT] = precedence[LE] = precedence[GT] = precedence[GE] = 11; + precedence[LSH] = precedence[RSH] = precedence[URSH] = 12; + precedence[ADD] = precedence[SUB] = 12; + precedence[MUL] = precedence[DIV] = precedence[MOD] = 13; + precedence[BITNOT] = precedence[BANG] = precedence[TYPEOF] = 14; + precedence[DOT] = precedence[LB] = precedence[LP] = precedence[INC] = precedence[DEC] = 15; } // Parsing Logic ///////////////////////////////////////////////////////// - /** a block is either a single statement or a list of statements surrounded by curly braces; all expressions are also statements */ - public Expr parseBlock(boolean requireBraces) throws IOException { - Expr ret = null; - int tok = peekToken(); - boolean braced = tok == LC; - if (requireBraces && !braced) throw new ParserException("expected {"); - if (braced) getToken(); - Expr head = null; - Expr tail = null; - int curLine = line; - OUTER: while(true) { - Expr smt; - - switch(tok = peekToken()) { - case -1: break OUTER; - case LC: smt = parseBlock(true); break; - case THROW: case RETURN: case ASSERT: - getToken(); - smt = new Expr(curLine, tok, parseMaximalExpr()); - if (getToken() != SEMI) throw new ParserException("expected ;"); - break; - - /* - FIXME - case NAME: { - getToken(); - String str = string; - if (getToken() != COLON) throw new ParserException("expected COLON after label"); - Expr labeledBlock = parseBlock(false); - if (labeledBlock.code != WHILE && labeledBlock.code != FOR && labeledBlock.code != SWITCH) - throw new ParserException("you can only label a WHILE, FOR, or SWITCH block"); - labeledBlock.string = str; - return labeledBlock; - } - */ - - case GOTO: case BREAK: case CONTINUE: { - getToken(); - int t = peekToken(); - if (t == NAME) { - getToken(); - smt = new Expr(curLine, tok, new Expr(curLine, string)); - } else if (t == GOTO) { - throw new ParserException("goto must be followed by a label"); - } else if (t == SEMI) { - getToken(); - smt = new Expr(curLine, tok); - } else { - throw new ParserException(codeToString[tok] + " followed by a " + codeToString[t] + "; expected NAME or SEMI"); - } - break; - } - - case RC: - if (braced) getToken(); - break OUTER; - - case SEMI: - getToken(); - if (!braced) break OUTER; - continue; - - default: - smt = parseMaximalExpr(); - if (smt == null) { - if (head == null) throw new ParserException("empty statement list; next token is " + codeToString[peekToken()]); - break OUTER; - } - break; - } - if (head == null) head = tail = smt; else tail = (tail.next = smt); - } - return new Expr(curLine, LC, head); + /** gets a token and throws an exception if it is not code */ + private void consume(int code) throws IOException { + if (getToken() != code) { + if(code == NAME) switch(op) { + case RETURN: case TYPEOF: case BREAK: case CONTINUE: case TRY: case THROW: + case ASSERT: case NULL: case TRUE: case FALSE: case IN: case IF: case ELSE: + case SWITCH: case CASE: case DEFAULT: case WHILE: case VAR: case WITH: + case CATCH: case FINALLY: + throw pe("Bad variable name; '" + codeToString[op].toLowerCase() + "' is a javascript keyword"); + } + throw pe("expected " + codeToString[code] + ", got " + (op == -1 ? "EOF" : codeToString[op])); + } } - /** throws an error if the next token is not code */ - public void expect(int code) throws IOException { - int got = peekToken(); - if (got != code) - throw new ParserException("expected " + codeToString[code] + ", got " + (got == -1 ? "EOL" : codeToString[got])); + /** + * Parse the largest possible expression containing no operators + * of precedence below minPrecedence and append the + * bytecodes for that expression to appendTo; the + * appended bytecodes MUST grow the stack by exactly one element. + */ + private void startExpr(JSFunction appendTo, int minPrecedence) throws IOException { + int saveParserLine = parserLine; + _startExpr(appendTo, minPrecedence); + parserLine = saveParserLine; + } + private void _startExpr(JSFunction appendTo, int minPrecedence) throws IOException { + int tok = getToken(); + JSFunction b = appendTo; + + switch (tok) { + case -1: throw pe("expected expression"); + + // all of these simply push values onto the stack + case NUMBER: b.add(parserLine, LITERAL, number); break; + case STRING: b.add(parserLine, LITERAL, string); break; + case NULL: b.add(parserLine, LITERAL, null); break; + case TRUE: case FALSE: b.add(parserLine, LITERAL, JS.B(tok == TRUE)); break; + + // (.foo) syntax + case DOT: { + consume(NAME); + b.add(parserLine, TOPSCOPE); + b.add(parserLine, LITERAL, ""); + b.add(parserLine, GET); + b.add(parserLine, LITERAL, string); + b.add(parserLine, GET); + continueExpr(b, minPrecedence); + break; + } + + case LB: { + b.add(parserLine, ARRAY, JS.ZERO); // push an array onto the stack + int size0 = b.size; + int i = 0; + if (peekToken() != RB) + while(true) { // iterate over the initialization values + int size = b.size; + b.add(parserLine, LITERAL, JS.N(i++)); // push the index in the array to place it into + if (peekToken() == COMMA || peekToken() == RB) + b.add(parserLine, LITERAL, null); // for stuff like [1,,2,] + else + startExpr(b, NO_COMMA); // push the value onto the stack + b.add(parserLine, PUT); // put it into the array + b.add(parserLine, POP); // discard the value remaining on the stack + if (peekToken() == RB) break; + consume(COMMA); + } + b.set(size0 - 1, JS.N(i)); // back at the ARRAY instruction, write the size of the array + consume(RB); + break; + } + case SUB: { // negative literal (like "3 * -1") + consume(NUMBER); + b.add(parserLine, LITERAL, JS.N(number.doubleValue() * -1)); + break; + } + case LP: { // grouping (not calling) + startExpr(b, -1); + consume(RP); + break; + } + case INC: case DEC: { // prefix (not postfix) + startExpr(b, precedence[tok]); + int prev = b.size - 1; + if (b.get(prev) == GET && b.getArg(prev) != null) + b.set(prev, LITERAL, b.getArg(prev)); + else if(b.get(prev) == GET) + b.pop(); + else + throw pe("prefixed increment/decrement can only be performed on a valid assignment target"); + b.add(parserLine, GET_PRESERVE, Boolean.TRUE); + b.add(parserLine, LITERAL, JS.N(1)); + b.add(parserLine, tok == INC ? ADD : SUB, JS.N(2)); + b.add(parserLine, PUT, null); + b.add(parserLine, SWAP, null); + b.add(parserLine, POP, null); + break; + } + case BANG: case BITNOT: case TYPEOF: { + startExpr(b, precedence[tok]); + b.add(parserLine, tok); + break; + } + case LC: { // object constructor + b.add(parserLine, OBJECT, null); // put an object on the stack + if (peekToken() != RC) + while(true) { + if (peekToken() != NAME && peekToken() != STRING) + throw pe("expected NAME or STRING"); + getToken(); + b.add(parserLine, LITERAL, string); // grab the key + consume(COLON); + startExpr(b, NO_COMMA); // grab the value + b.add(parserLine, PUT); // put the value into the object + b.add(parserLine, POP); // discard the remaining value + if (peekToken() == RC) break; + consume(COMMA); + if (peekToken() == RC) break; // we permit {,,} -- I'm not sure if ECMA does + } + consume(RC); + break; + } + case NAME: { + b.add(parserLine, TOPSCOPE); + b.add(parserLine, LITERAL, string); + continueExprAfterAssignable(b,minPrecedence); + break; + } + case FUNCTION: { + consume(LP); + int numArgs = 0; + JSFunction b2 = new JSFunction(sourceName, parserLine, null); + b.add(parserLine, NEWFUNCTION, b2); + + // function prelude; arguments array is already on the stack + b2.add(parserLine, TOPSCOPE); + b2.add(parserLine, SWAP); + b2.add(parserLine, DECLARE, "arguments"); // declare arguments (equivalent to 'var arguments;') + b2.add(parserLine, SWAP); // set this.arguments and leave the value on the stack + b2.add(parserLine, PUT); + + while(peekToken() != RP) { // run through the list of argument names + numArgs++; + if (peekToken() == NAME) { + consume(NAME); // a named argument + String varName = string; + + b2.add(parserLine, DUP); // dup the args array + b2.add(parserLine, GET, JS.N(numArgs - 1)); // retrieve it from the arguments array + b2.add(parserLine, TOPSCOPE); + b2.add(parserLine, SWAP); + b2.add(parserLine, DECLARE, varName); // declare the name + b2.add(parserLine, SWAP); + b2.add(parserLine, PUT); + b2.add(parserLine, POP); // pop the value + b2.add(parserLine, POP); // pop the scope + } + if (peekToken() == RP) break; + consume(COMMA); + } + consume(RP); + + b2.numFormalArgs = numArgs; + b2.add(parserLine, POP); // pop off the arguments array + b2.add(parserLine, POP); // pop off TOPSCOPE + + if(peekToken() != LC) + throw pe("JSFunctions must have a block surrounded by curly brackets"); + + parseBlock(b2, null); // the function body + + b2.add(parserLine, LITERAL, null); // in case we "fall out the bottom", return NULL + b2.add(parserLine, RETURN); + + break; + } + default: throw pe("expected expression, found " + codeToString[tok] + ", which cannot start an expression"); + } + + // attempt to continue the expression + continueExpr(b, minPrecedence); } - /** parses the largest possible expression */ - public Expr parseMaximalExpr() throws IOException { return parseMaximalExpr(null, -1); } - public Expr parseMaximalExpr(Expr prefix, int minPrecedence) throws IOException { - while(true) { - if (peekToken() == -1) break; - Expr save = prefix; - prefix = parseSingleExpr(prefix, minPrecedence); - if (save == prefix) break; - if (prefix == null) throw new ParserException("parseSingleExpr() returned null"); - } - return prefix; + private Grammar parseGrammar(Grammar g) throws IOException { + int tok = getToken(); + if (g != null) + switch(tok) { + case BITOR: return new Grammar.Alternative(g, parseGrammar(null)); + case ADD: return parseGrammar(new Grammar.Repetition(g, 1, Integer.MAX_VALUE)); + case MUL: return parseGrammar(new Grammar.Repetition(g, 0, Integer.MAX_VALUE)); + case HOOK: return parseGrammar(new Grammar.Repetition(g, 0, 1)); + } + Grammar g0 = null; + switch(tok) { + //case NUMBER: g0 = new Grammar.Literal(number); break; + case NAME: g0 = new Grammar.Reference(string); break; + case STRING: + g0 = new Grammar.Literal(string); + if (peekToken() == DOT) { + String old = string; + consume(DOT); + consume(DOT); + consume(STRING); + if (old.length() != 1 || string.length() != 1) throw pe("literal ranges must be single-char strings"); + g0 = new Grammar.Range(old.charAt(0), string.charAt(0)); + } + break; + case LP: g0 = parseGrammar(null); consume(RP); break; + default: pushBackToken(); return g; + } + if (g == null) return parseGrammar(g0); + return parseGrammar(new Grammar.Juxtaposition(g, g0)); } - public Expr parseSingleExpr(Expr prefix, int minPrecedence) throws IOException { - Expr e1 = null, e2 = null, e3 = null, head = null, tail = null, ret = null; - - int tok = peekToken(); - if (minPrecedence > 0 && - tok < precedence.length && - precedence[tok] != 0 && - (isRightAssociative[tok] ? - (precedence[tok] < minPrecedence) : - (precedence[tok] <= minPrecedence))) { - return prefix; - } - getToken(); - int curLine = line; - - // these case arms match the precedence of operators; each arm is a precedence level. - switch (tok) { - - case WITH: throw new ParserException("XWT does not allow the WITH keyword"); - case VOID: case RESERVED: throw new ParserException("reserved word that you shouldn't be using"); - case NAME: if (prefix != null) { pushBackToken(); return prefix; } else return parseMaximalExpr(new Expr(curLine, NAME, string), minPrecedence); - case STRING: if (prefix != null) { pushBackToken(); return prefix; } else return new Expr(curLine, string); - case NUMBER: if (prefix != null) { pushBackToken(); return prefix; } else return new Expr(curLine, number); - case NULL: case TRUE: case FALSE: case NOP: if (prefix != null) { pushBackToken(); return prefix; } else return new Expr(curLine, tok); - - case ASSIGN_BITOR: case ASSIGN_BITXOR: case ASSIGN_BITAND: case ASSIGN_LSH: - case ASSIGN_RSH: case ASSIGN_URSH: case ASSIGN_ADD: case ASSIGN_SUB: case ASSIGN_MUL: case ASSIGN_DIV: case ASSIGN_MOD: - return new Expr(curLine, ASSIGN, prefix, new Expr(curLine, tok - 1, prefix, parseMaximalExpr(null, precedence[ASSIGN]))); - - case COMMA: pushBackToken();return prefix; - - case THIS: - if (prefix != null) { pushBackToken(); return prefix; } - return new Expr(curLine, THIS); - - case SUB: - if (prefix == null && peekToken() == NUMBER) { - getToken(); - return new Expr(curLine, new Double(number.doubleValue() * -1)); - } // else fall through - case BITOR: case BITXOR: case BITAND: case SHEQ: case SHNE: case LSH: - case RSH: case URSH: case ADD: case MUL: case DIV: case MOD: - case ASSIGN: case GT: case GE: case OR: case AND: - case EQ: case NE: case LT: case LE: - if (prefix == null) throw new ParserException("the " + codeToString[tok] + " token cannot start an expression"); - return new Expr(curLine, tok, prefix, parseMaximalExpr(null, precedence[tok])); - - case DOT: { - //e1 = parseMaximalExpr(null, precedence[DOT]); - //if (e1.code == NAME) e1.code = STRING; - //else throw new ParserException("argument to DOT must be a NAME; got a " + codeToString[e1.code]); - //return new Expr(curLine, DOT, prefix, e1); - tok = getToken(); - if (tok != NAME) throw new ParserException("DOT must be followed by a NAME"); - return new Expr(curLine, DOT, prefix, new Expr(curLine, STRING, string)); - } - - case BANG: case BITNOT: case INSTANCEOF: case TYPEOF: - if (prefix != null) { pushBackToken(); return prefix; } - return new Expr(curLine, tok, parseMaximalExpr(null, precedence[tok])); - - case INC: case DEC: - if (prefix == null) { - // prefix - e1 = parseMaximalExpr(null, precedence[tok]); - return new Expr(curLine, ASSIGN, e1, new Expr(curLine, tok == INC ? ADD : SUB, e1, new Expr(curLine, new Integer(1)))); - } else { - // postfix - return new Expr(curLine, tok, prefix); - } - - case LP: - if (prefix == null) { // grouping - Expr r = parseMaximalExpr(); - expect(RP); getToken(); - return r; - - } else { // invocation - while(peekToken() != RP) { - Expr e = parseMaximalExpr(); - if (head == null) head = tail = e; else tail = tail.next = e; - tok = getToken(); - if (tok == RP) { pushBackToken(); break; } - if (tok != COMMA) throw new ParserException("expected comma or right paren, got " + codeToString[tok]); - } - getToken(); - return new Expr(curLine, LP, prefix, head); - } - - case LB: - if (prefix != null) { - // subscripting - e1 = parseMaximalExpr(); - if (getToken() != RB) throw new ParserException("expected a right brace"); - return new Expr(curLine, DOT, prefix, e1); - } else { - // array ctor - while(true) { - if (peekToken() == RB) { getToken(); return new Expr(curLine, LB, prefix, head); } - Expr eee = parseMaximalExpr(); - if (eee != null) { - if (head == null) head = tail = eee; - else tail.next = tail = eee; - } - tok = getToken(); - if (tok == RB) return new Expr(curLine, LB, prefix, head); - if (tok != COMMA) throw new ParserException("expected right bracket or comma"); - } - } - - case LC: - if (prefix != null) { pushBackToken(); return prefix; } - tok = getToken(); - if (tok == RC) return new Expr(curLine, RC, head); - while(true) { - if (tok != NAME && tok != STRING) throw new ParserException("expecting name, got " + codeToString[tok]); - Expr name = new Expr(curLine, NAME, string); - if (getToken() != COLON) throw new ParserException("expecting colon"); - e1 = new Expr(curLine, COLON, name, parseMaximalExpr()); - if (head == null) head = tail = e1; else tail = tail.next = e1; - tok = getToken(); - if (tok != COMMA && tok != RC) throw new ParserException("expected right curly or comma, got " + codeToString[tok]); - if (tok == RC) return new Expr(curLine, RC, head); - tok = getToken(); - if (tok == RC) return new Expr(curLine, RC, head); - } - - case HOOK: - e2 = parseMaximalExpr(); - if (getToken() != COLON) throw new ParserException("expected colon to close ?: expression"); - e3 = parseMaximalExpr(); - return new Expr(curLine, HOOK, prefix, new Expr(curLine, ELSE, e2, e3)); - - case SWITCH: { - if (prefix != null) { pushBackToken(); return prefix; } - if (getToken() != LP) throw new ParserException("expected left paren"); - Expr switchExpr = parseMaximalExpr(); - if (getToken() != RP) throw new ParserException("expected left paren"); - if (getToken() != LC) throw new ParserException("expected left brace"); - Expr firstExpr = null; - Expr lastExpr = null; - while(true) { - tok = getToken(); - Expr caseExpr; - if (tok == DEFAULT) { - caseExpr = null; - } else if (tok == CASE) { - // FIXME: we don't support non-brace-enclosed CASE blocks - caseExpr = parseMaximalExpr(); - } else { - throw new ParserException("expected CASE"); - } - expect(COLON); getToken(); - head = tail = null; - while(peekToken() != CASE && peekToken() != DEFAULT && peekToken() != RC) { - e1 = parseBlock(false); - if (e1 == null) break; - if (head == null) head = tail = e1; else tail = tail.next = e1; - } - e1 = new Expr(curLine, tok, caseExpr, new Expr(curLine, LC, head)); - if (lastExpr == null) firstExpr = e1; else lastExpr.next = e1; - lastExpr = e1; - if (peekToken() == RC) {getToken(); return new Expr(curLine, SWITCH, switchExpr, firstExpr); } - } - } - - case FUNCTION: { - if (prefix != null) { pushBackToken(); return prefix; } - if (getToken() != LP) throw new ParserException("function keyword must be followed by a left paren"); - Expr formalArgs = null, cur = null; - tok = getToken(); - while(tok != RP) { - if (tok != NAME) throw new ParserException("expected a variable name"); - if (cur == null) { formalArgs = cur = new Expr(curLine, string); } - else { cur.next = new Expr(curLine, NAME, string); cur = cur.next; } - tok = getToken(); - if (tok == RP) break; - if (tok != COMMA) throw new ParserException("function argument list must consist of alternating NAMEs and COMMAs"); - tok = getToken(); - } - return new Expr(curLine, FUNCTION, formalArgs, parseBlock(true)); - } - - case VAR: - if (prefix != null) { pushBackToken(); return prefix; } - while(true) { - if (getToken() != NAME) throw new ParserException("variable declarations must start with a variable name"); - Expr name = new Expr(curLine, NAME, string); - Expr initVal = null; - tok = peekToken(); - Expr e = null; - if (tok == ASSIGN) { - getToken(); - initVal = parseMaximalExpr(); - tok = peekToken(); - e = new Expr(curLine, ASSIGN, name, initVal); - } else { - e = new Expr(curLine, NAME, name); - } - if (head == null) head = tail = e; else tail = tail.next = e; - if (tok != COMMA) break; - getToken(); - } - return new Expr(curLine, VAR, head); - - case TRY: { - // We deliberately allow you to omit braces in catch{}/finally{} if they are single statements... - if (prefix != null) { pushBackToken(); return prefix; } - Expr tryBlock = parseBlock(true); - - tok = peekToken(); - if (tok == CATCH) { - getToken(); - if (getToken() != LP) throw new ParserException("expected ("); - if (getToken() != NAME) throw new ParserException("expected name"); - Expr name = new Expr(curLine, NAME, string); - if (getToken() != RP) throw new ParserException("expected )"); - head = tail = new Expr(curLine, CATCH, name, parseBlock(false)); - tok = peekToken(); - } - if (tok == FINALLY) { - getToken(); - e1 = new Expr(curLine, FINALLY, parseBlock(false)); - if (head == null) head = tail = e1; else tail = tail.next = e1; - } - - if (head == null) throw new ParserException("try without catch or finally"); - return new Expr(curLine, TRY, tryBlock, head); - } - - case IF: case WHILE: { - if (prefix != null) { pushBackToken(); return prefix; } - if (getToken() != LP) throw new ParserException("expected left paren"); - Expr parenExpr = parseMaximalExpr(); - int t; - if ((t = getToken()) != RP) throw new ParserException("expected right paren, but got " + codeToString[t]); - Expr firstBlock = parseBlock(false); - if (tok == IF && peekToken() == ELSE) { - getToken(); - return new Expr(curLine, tok, parenExpr, new Expr(curLine, ELSE, firstBlock, parseBlock(false))); - } else if (tok == IF) { - return new Expr(curLine, tok, parenExpr, new Expr(curLine, ELSE, firstBlock, new Expr(curLine, LC))); - } else if (tok == WHILE) { - return new Expr(curLine, tok, parenExpr, firstBlock); - } - } - - case IN: pushBackToken(); return prefix; - - case FOR: - if (prefix != null) { pushBackToken(); return prefix; } - if (getToken() != LP) throw new ParserException("expected left paren"); - e1 = parseMaximalExpr(null, -1); - if (peekToken() == IN) { - getToken(); - e2 = parseMaximalExpr(null, -1); - if (getToken() != RP) throw new ParserException("expected right paren"); - return new Expr(curLine, FOR, new Expr(curLine, IN, e1, e2), parseBlock(false)); - - } else { - if (getToken() != SEMI) throw new ParserException("expected ;"); - e2 = parseMaximalExpr(); - if (getToken() != SEMI) throw new ParserException("expected ;"); - e3 = parseMaximalExpr(); - if (getToken() != RP) throw new ParserException("expected right paren"); - return new Expr(curLine, LC, e1, new Expr(curLine, WHILE, e2, new Expr(curLine, LC, parseBlock(false), e3))); - } - - case DO: { - if (prefix != null) { pushBackToken(); return prefix; } - Expr firstBlock = parseBlock(false); - if (getToken() != WHILE) throw new ParserException("expecting WHILE"); - if (getToken() != LP) throw new ParserException("expected left paren"); - Expr whileExpr = parseMaximalExpr(); - if (getToken() != RP) throw new ParserException("expected right paren"); - if (getToken() != SEMI) throw new ParserException("semicolon"); - return new Expr(curLine, DO, firstBlock, whileExpr); - } - - default: - pushBackToken(); - return prefix; - } + /** + * Assuming that a complete assignable (lvalue) has just been + * parsed and the object and key are on the stack, + * continueExprAfterAssignable will attempt to parse an + * expression that modifies the assignable. This method always + * decreases the stack depth by exactly one element. + */ + private void continueExprAfterAssignable(JSFunction b,int minPrecedence) throws IOException { + int saveParserLine = parserLine; + _continueExprAfterAssignable(b,minPrecedence); + parserLine = saveParserLine; + } + private void _continueExprAfterAssignable(JSFunction b,int minPrecedence) throws IOException { + if (b == null) throw new Error("got null b; this should never happen"); + int tok = getToken(); + if (minPrecedence != -1 && (precedence[tok] < minPrecedence || (precedence[tok] == minPrecedence && !isRightAssociative[tok]))) + // force the default case + tok = -1; + switch(tok) { + + case GRAMMAR: { + b.add(parserLine, GET_PRESERVE); + Grammar g = parseGrammar(null); + if (peekToken() == LC) { + g.action = new JSFunction(sourceName, parserLine, null); + parseBlock((JSFunction)g.action); + ((JSFunction)g.action).add(parserLine, LITERAL, null); // in case we "fall out the bottom", return NULL + ((JSFunction)g.action).add(parserLine, RETURN); + } + b.add(parserLine, MAKE_GRAMMAR, g); + b.add(parserLine, PUT); + break; + } + + case ASSIGN_BITOR: case ASSIGN_BITXOR: case ASSIGN_BITAND: case ASSIGN_LSH: case ASSIGN_RSH: case ASSIGN_URSH: + case ASSIGN_MUL: case ASSIGN_DIV: case ASSIGN_MOD: case ASSIGN_ADD: case ASSIGN_SUB: case ADD_TRAP: case DEL_TRAP: { + if (tok != ADD_TRAP && tok != DEL_TRAP) b.add(parserLine, GET_PRESERVE); + + startExpr(b, precedence[tok]); + + int size = b.size; + + if (tok != ADD_TRAP && tok != DEL_TRAP) { + // tok-1 is always s/^ASSIGN_// (0 is BITOR, 1 is ASSIGN_BITOR, etc) + b.add(parserLine, tok - 1, tok-1==ADD ? JS.N(2) : null); + b.add(parserLine, PUT); + b.add(parserLine, SWAP); + b.add(parserLine, POP); + } else { + b.add(parserLine, tok); + } + break; + } + case INC: case DEC: { // postfix + b.add(parserLine, GET_PRESERVE, Boolean.TRUE); + b.add(parserLine, LITERAL, JS.N(1)); + b.add(parserLine, tok == INC ? ADD : SUB, JS.N(2)); + b.add(parserLine, PUT, null); + b.add(parserLine, SWAP, null); + b.add(parserLine, POP, null); + b.add(parserLine, LITERAL, JS.N(1)); + b.add(parserLine, tok == INC ? SUB : ADD, null); // undo what we just did, since this is postfix + break; + } + case ASSIGN: { + startExpr(b, precedence[tok]); + b.add(parserLine, PUT); + b.add(parserLine, SWAP); + b.add(parserLine, POP); + break; + } + case LP: { + + // Method calls are implemented by doing a GET_PRESERVE + // first. If the object supports method calls, it will + // return JS.METHOD + int n = parseArgs(b, 2); + b.add(parserLine, GET_PRESERVE); + b.add(parserLine, CALLMETHOD, JS.N(n)); + break; + } + default: { + pushBackToken(); + if(b.get(b.size-1) == LITERAL && b.getArg(b.size-1) != null) + b.set(b.size-1,GET,b.getArg(b.size-1)); + else + b.add(parserLine, GET); + return; + } + } } - - // Expr ////////////////////////////////////////////////////////////////////// - /** sorta like gcc trees */ - class Expr { - int code = -1; - - final Expr left; - final Expr right; - Expr next = null; // if this expr is part of a list - int line = -1; - String sourceName = "unknown"; - - String string = null; - Number number = null; + + /** + * Assuming that a complete expression has just been parsed, + * continueExpr will attempt to extend this expression by + * parsing additional tokens and appending additional bytecodes. + * + * No operators with precedence less than minPrecedence + * will be parsed. + * + * If any bytecodes are appended, they will not alter the stack + * depth. + */ + private void continueExpr(JSFunction b, int minPrecedence) throws IOException { + int saveParserLine = parserLine; + _continueExpr(b, minPrecedence); + parserLine = saveParserLine; + } + private void _continueExpr(JSFunction b, int minPrecedence) throws IOException { + if (b == null) throw new Error("got null b; this should never happen"); + int tok = getToken(); + if (tok == -1) return; + if (minPrecedence != -1 && (precedence[tok] < minPrecedence || (precedence[tok] == minPrecedence && !isRightAssociative[tok]))) { + pushBackToken(); + return; + } + + switch (tok) { + case LP: { // invocation (not grouping) + int n = parseArgs(b, 1); + b.add(parserLine, CALL, JS.N(n)); + break; + } + case BITOR: case BITXOR: case BITAND: case SHEQ: case SHNE: case LSH: + case RSH: case URSH: case MUL: case DIV: case MOD: + case GT: case GE: case EQ: case NE: case LT: case LE: case SUB: { + startExpr(b, precedence[tok]); + b.add(parserLine, tok); + break; + } + case ADD: { + int count=1; + int nextTok; + do { + startExpr(b,precedence[tok]); + count++; + nextTok = getToken(); + } while(nextTok == tok); + pushBackToken(); + b.add(parserLine, tok, JS.N(count)); + break; + } + case OR: case AND: { + b.add(parserLine, tok == AND ? b.JF : b.JT, JS.ZERO); // test to see if we can short-circuit + int size = b.size; + startExpr(b, precedence[tok]); // otherwise check the second value + b.add(parserLine, JMP, JS.N(2)); // leave the second value on the stack and jump to the end + b.add(parserLine, LITERAL, tok == AND ? + JS.B(false) : JS.B(true)); // target of the short-circuit jump is here + b.set(size - 1, JS.N(b.size - size)); // write the target of the short-circuit jump + break; + } + case DOT: { + // support foo..bar syntax for foo[""].bar + if (peekToken() == DOT) { + string = ""; + } else { + consume(NAME); + } + b.add(parserLine, LITERAL, string); + continueExprAfterAssignable(b,minPrecedence); + break; + } + case LB: { // subscripting (not array constructor) + startExpr(b, -1); + consume(RB); + continueExprAfterAssignable(b,minPrecedence); + break; + } + case HOOK: { + b.add(parserLine, JF, JS.ZERO); // jump to the if-false expression + int size = b.size; + startExpr(b, minPrecedence); // write the if-true expression + b.add(parserLine, JMP, JS.ZERO); // if true, jump *over* the if-false expression + b.set(size - 1, JS.N(b.size - size + 1)); // now we know where the target of the jump is + consume(COLON); + size = b.size; + startExpr(b, minPrecedence); // write the if-false expression + b.set(size - 1, JS.N(b.size - size + 1)); // this is the end; jump to here + break; + } + case COMMA: { + // pop the result of the previous expression, it is ignored + b.add(parserLine,POP); + startExpr(b,-1); + break; + } + default: { + pushBackToken(); + return; + } + } + + continueExpr(b, minPrecedence); // try to continue the expression + } - public String toString() { return toString(0); } - public String toString(int indent) { - String ret = ""; - for(int i=0; i> toLong(right.eval(s))); - case Lexer.URSH: return new Long(toLong(left.eval(s)) >>> toLong(right.eval(s))); - - case Lexer.LT: return toDouble(left.eval(s)) < toDouble(right.eval(s)) ? Boolean.TRUE : Boolean.FALSE; - case Lexer.LE: return toDouble(left.eval(s)) <= toDouble(right.eval(s)) ? Boolean.TRUE : Boolean.FALSE; - case Lexer.GT: return toDouble(left.eval(s)) > toDouble(right.eval(s)) ? Boolean.TRUE : Boolean.FALSE; - case Lexer.GE: return toDouble(left.eval(s)) >= toDouble(right.eval(s)) ? Boolean.TRUE : Boolean.FALSE; - - case Lexer.OR: return new Boolean(toBoolean(left.eval(s)) || toBoolean(right.eval(s))); - case Lexer.AND: return new Boolean(toBoolean(left.eval(s)) && toBoolean(right.eval(s))); - case Lexer.BANG: return new Boolean(!toBoolean(left.eval(s))); - - case Lexer.EQ: - case Lexer.NE: { - // FIXME: should use Javascript coercion-equality rules - Object l = left.eval(s); - Object r = right.eval(s); - boolean ret = (l == null && r == null) || (l != null && l.equals(r)); - return new Boolean(code == Lexer.EQ ? ret : !ret); - } - - case Lexer.INC: - case Lexer.DEC: - case Lexer.ASSIGN: { - Object v = (code == Lexer.ASSIGN) ? right.eval(s) : new Double(toDouble(left.eval(s)) + (code == Lexer.INC ? 1 : -1)); - if (left.code == Lexer.DOT) { - Object o = left.left.eval(s); - if (o instanceof String) { - throw new EvaluatorException("can't set properties on a String"); - } else if (o instanceof Number) { - throw new EvaluatorException("can't set properties on a Number"); - } else if (o instanceof Boolean) { - throw new EvaluatorException("can't set properties on a Boolean"); - } else { - JS target = (JS)left.left.eval(s); - if (target == null) throw new JS.Exn(new EvaluatorException("attempted to put to the null value")); - target.put(left.right.eval(s), v); - return v; - } - } else { - s.put(left.string, v); - return v; - } - } - - case Lexer.TYPEOF: { - Object o = left.eval(s); - if (o == null) return "null"; - if (o.getClass() == String.class) return "string"; - if (o.getClass() == Boolean.class) return "boolean"; - if (o instanceof Number) return "number"; - if (o instanceof JS.Array) return "array"; - if (o instanceof JS) return "object"; - throw new EvaluatorException("typeof " + o.getClass().getName() + " unknown"); - } - - case Lexer.NUMBER: return number; - case Lexer.STRING: return string; - - case Lexer.NULL: return null; - case Lexer.FALSE: return Boolean.FALSE; - case Lexer.TRUE: return Boolean.TRUE; - case Lexer.ASSERT: if (!toBoolean(left.eval(s))) throw new EvaluatorException("assertion failed"); - case Lexer.THROW: throw new JS.Exn(left.eval(s)); - case Lexer.NAME: return s.get(string); - case Lexer.THIS: return s.isTransparent() ? s : this.eval(s.getParentScope()); - - case Lexer.DOT: { - final Object o = left.eval(s); - Object v = ((right.code == Lexer.NAME) || (right.code == Lexer.STRING)) ? right.string : right.eval(s); - if (o instanceof String) { - if (v.equals("length")) return new Integer(((String)o).length()); - else if (v.equals("substring")) return new JS.Function() { - public Object _call(JS.Array args) { - if (args.length() == 1) return ((String)o).substring(toNumber(args.elementAt(0)).intValue()); - else if (args.length() == 2) return ((String)o).substring(toNumber(args.elementAt(0)).intValue(), - toNumber(args.elementAt(1)).intValue()); - else throw new EvaluatorException("String.substring() can only take one or two arguments"); - } - }; - else if (v.equals("indexOf")) return new JS.Function() { - public Object _call(JS.Array args) { - if (args.length() != 1) return null; - return new Integer(((String)o).indexOf(args.elementAt(0).toString())); - } - }; - throw new EvaluatorException("Not Implemented: properties on String objects"); - } else if (o instanceof Boolean) { - throw new EvaluatorException("Not Implemented: properties on Boolean objects"); - } else if (o instanceof Number) { - throw new EvaluatorException("Not Implemented: properties on Number objects"); - } else if (o instanceof JS) { - return ((JS)o).get(v); - } - } - - case Lexer.TRY: { - boolean safeToExit = false; - try { - Object ret = left.eval(s); - safeToExit = true; - return ret; - } catch (JS.Exn e) { - Expr c = right; - if (c.code == Lexer.CATCH) { - JS.Scope scope = new JS.Scope(s); - s.put(c.left.string, e); - c.right.eval(scope); - c = c.next; - } - if (c.code == Lexer.FINALLY) { - JS.Scope scope = new JS.Scope(s); - c.left.eval(scope); - } - } finally { - if (!safeToExit) Log.log(this, "WARNING: Java exception penetrated a JavaScript try{} block"); - } - return null; - } - - case Lexer.LP: - JS.Function f = (JS.Function)left.eval(s); - JS.Array arguments = new JS.Array(); - for(Expr e = right; e != null; e = e.next) arguments.addElement(e.eval(s)); - if (f == null) throw new JS.Exn(new EvaluatorException("attempted to call null")); - return f.call(arguments); - - case Lexer.FUNCTION: - return new JS.Function() { - public String toString() { return right.sourceName + ":" + right.line; } - public String getSourceName() throws JS.Exn { return right.sourceName; } - public Object _call(final JS.Array args) throws JS.Exn { - Function save = JS.getCurrentFunction(); - JS.currentFunction.put(Thread.currentThread(), this); - JS.Scope scope = new JS.Scope(s) { - // FIXME - public String getSourceName() { return sourceName; } - public boolean isTransparent() { return true; } - public Object get(Object key) throws JS.Exn { - if (key.equals("trapee")) return org.xwt.Trap.currentTrapee(); - else if (key.equals("cascade")) return org.xwt.Trap.cascadeFunction; - else if (key.equals("arguments")) return args; - return super.get(key); - } - }; - int i = 0; - for(Expr e = left; e != null; e = e.next) scope.put(e.string, args.get(new Integer(i++))); - try { - return right.eval(scope); - } catch (ReturnException r) { - return r.retval; - } catch (ControlTransferException c) { - throw new EvaluatorException("error, ControlTransferException tried to leave a function: " - + c); - } finally { - if (save == null) JS.currentFunction.remove(Thread.currentThread()); - else JS.currentFunction.put(Thread.currentThread(), save); - } - } - }; - - case Lexer.FOR: - Object[] keys = ((JS)left.right.eval(s)).keys(); - for(int i=0; i