X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=src%2Forg%2Fxwt%2Fjs%2FParser.java;h=ffb12d86bb4d57537351ea03d81f62ff7ef42e18;hb=67eeff476179a91ae930ea89cbecde22132ca532;hp=e82b408d7841cc8a04ae75e103085ecf3948d8b4;hpb=466cd55c399b5b7a6ae922432f38dee204dc5001;p=org.ibex.core.git diff --git a/src/org/xwt/js/Parser.java b/src/org/xwt/js/Parser.java index e82b408..ffb12d8 100644 --- a/src/org/xwt/js/Parser.java +++ b/src/org/xwt/js/Parser.java @@ -1,23 +1,36 @@ // Copyright 2003 Adam Megacz, see the COPYING file for licensing [GPL] package org.xwt.js; -// FIXME: line number accuracy -// FIXME: scope management - import org.xwt.util.*; import java.io.*; +// FEATURE intern Integers/Numbers + /** - * Parses a stream of lexed tokens into a tree of CompiledFunction's. + * Parses a stream of lexed tokens into a tree of JSFunction's. * - - * There are three kinds of things we parse: blocks, statements, - * expressions. Expressions are a special type of statement that - * evaluates to a value (for example, "break" is not an expression, - * but "3+2" is). AssignmentTargets are a special kind of expression - * that can be 'put' to (for example, "foo()" is not an - * assignmentTarget, but "foo[7]" is). FIXME. - + * 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 @@ -62,7 +75,7 @@ class Parser extends Lexer implements ByteCodes { /** for debugging */ public static void main(String[] s) throws Exception { - CompiledFunction block = new CompiledFunction("stdin", 0, new InputStreamReader(System.in), null); + JSFunction block = new JSFunction("stdin", 0, new InputStreamReader(System.in), null); if (block == null) return; System.out.println(block); } @@ -72,28 +85,49 @@ class Parser extends Lexer implements ByteCodes { 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 { - isRightAssociative[ASSIGN] = true; - - precedence[ASSIGN] = 1; - precedence[HOOK] = 2; - precedence[COMMA] = 3; - precedence[OR] = precedence[AND] = 4; - precedence[GT] = precedence[GE] = 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; + 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] = 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[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] = 13; - precedence[MUL] = precedence[DIV] = precedence[MOD] = 14; - precedence[BITNOT] = 15; - precedence[INC] = precedence[DEC] = 16; - precedence[LP] = 17; - precedence[LB] = 18; - precedence[DOT] = 19; + 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; } @@ -101,7 +135,7 @@ class Parser extends Lexer implements ByteCodes { /** gets a token and throws an exception if it is not code */ private void consume(int code) throws IOException { - if (getToken() != code) throw new ParserException("expected " + codeToString[code] + ", got " + (op == -1 ? "EOF" : codeToString[op])); + if (getToken() != code) throw pe("expected " + codeToString[code] + ", got " + (op == -1 ? "EOF" : codeToString[op])); } /** @@ -109,45 +143,49 @@ class Parser extends Lexer implements ByteCodes { * 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(CompiledFunction appendTo, int minPrecedence) throws IOException { + */ + 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(); - CompiledFunction b = appendTo; + JSFunction b = appendTo; switch (tok) { - case -1: throw new ParserException("expected expression"); + case -1: throw pe("expected expression"); // all of these simply push values onto the stack - case NUMBER: b.add(line, LITERAL, number); break; - case STRING: b.add(line, LITERAL, string); break; - case THIS: b.add(line, TOPSCOPE, null); break; - case NULL: b.add(line, LITERAL, null); break; - case TRUE: case FALSE: b.add(line, LITERAL, new Boolean(tok == TRUE)); break; + 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, new Boolean(tok == TRUE)); break; case LB: { - b.add(line, ARRAY, new Integer(0)); // push an array onto the stack - int size0 = b.size(); + b.add(parserLine, ARRAY, new Integer(0)); // 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(); + while(true) { // iterate over the initialization values + int size = b.size; + b.add(parserLine, LITERAL, new Integer(i++)); // push the index in the array to place it into if (peekToken() == COMMA || peekToken() == RB) - b.add(line, LITERAL, null); // for stuff like [1,,2,] + b.add(parserLine, LITERAL, null); // for stuff like [1,,2,] else - startExpr(b, -1); // push the value onto the stack - b.add(line, LITERAL, new Integer(i++)); // push the index in the array to place it into - b.add(line, PUT); // put it into the array - b.add(line, POP); // discard the value remaining on the stack + 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, new Integer(i)); // back at the ARRAY instruction, write the size of the array + b.set(size0 - 1, new Integer(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(line, LITERAL, new Double(number.doubleValue() * -1)); + b.add(parserLine, LITERAL, new Double(number.doubleValue() * -1)); break; } case LP: { // grouping (not calling) @@ -157,26 +195,38 @@ class Parser extends Lexer implements ByteCodes { } case INC: case DEC: { // prefix (not postfix) startExpr(b, precedence[tok]); - b.set(b.size() - 1, tok, new Boolean(true)); // FIXME, ugly; need startAssignTarget + 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, new Integer(1)); + b.add(parserLine, tok == INC ? ADD : SUB, null); + 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(line, tok); + b.add(parserLine, tok); break; } case LC: { // object constructor - b.add(line, OBJECT, null); // put an object on the stack + b.add(parserLine, OBJECT, null); // put an object on the stack if (peekToken() != RC) while(true) { if (peekToken() != NAME && peekToken() != STRING) - throw new ParserException("expected NAME or STRING"); + throw pe("expected NAME or STRING"); getToken(); - b.add(line, LITERAL, string); // grab the key + b.add(parserLine, LITERAL, string); // grab the key consume(COLON); - startExpr(b, -1); // grab the value - b.add(line, PUT); // put the value into the object - b.add(line, POP); // discard the remaining value + 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 @@ -184,85 +234,142 @@ class Parser extends Lexer implements ByteCodes { consume(RC); break; } - case NAME: { // FIXME; this is an lvalue - String name = string; - if (peekToken() == ASSIGN) { - consume(ASSIGN); - b.add(line, TOPSCOPE); - b.add(line, LITERAL, name); - startExpr(b, minPrecedence); - b.add(line, PUT); - b.add(line, SWAP); - b.add(line, POP); - } else { - b.add(line, TOPSCOPE); - b.add(line, LITERAL, name); - b.add(line, GET); - } + case NAME: { + b.add(parserLine, TOPSCOPE); + b.add(parserLine, LITERAL, string); + continueExprAfterAssignable(b,minPrecedence); break; } case FUNCTION: { consume(LP); int numArgs = 0; - CompiledFunction b2 = new CompiledFunction(sourceName, line, null); - b.add(line, NEWFUNCTION, b2); + JSFunction b2 = new JSFunction(sourceName, parserLine, null, null); + b.add(parserLine, NEWFUNCTION, b2); // function prelude; arguments array is already on the stack - b2.add(line, TOPSCOPE); // push the scope onto the stack - b2.add(line, SWAP); // swap 'this' and 'arguments' - - b2.add(line, LITERAL, "arguments"); // declare arguments (equivalent to 'var arguments;') - b2.add(line, DECLARE); - - b2.add(line, LITERAL, "arguments"); // set this.arguments and leave the value on the stack - b2.add(line, SWAP); - b2.add(line, PUT); - b2.add(line, SWAP); - b2.add(line, POP); + 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 + while(peekToken() != RP) { // run through the list of argument names + numArgs++; if (peekToken() == NAME) { - consume(NAME); // a named argument - - b2.add(line, LITERAL, string); // declare the name - b2.add(line, DECLARE); - - b2.add(line, LITERAL, new Integer(numArgs)); // retrieve it from the arguments array - b2.add(line, GET_PRESERVE); - b2.add(line, SWAP); - b2.add(line, POP); + consume(NAME); // a named argument + String varName = string; - b2.add(line, TOPSCOPE); // put it to the current scope - b2.add(line, SWAP); - b2.add(line, LITERAL, string); - b2.add(line, SWAP); - b2.add(line, PUT); - - b2.add(line, POP); // clean the stack - b2.add(line, POP); + b2.add(parserLine, DUP); // dup the args array + b2.add(parserLine, GET, new Integer(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); - numArgs++; } consume(RP); - b2.add(line, POP); // pop off the arguments array - - parseStatement(b2, null); // the function body + 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(line, LITERAL, null); // in case we "fall out the bottom", return NULL - b2.add(line, RETURN); + b2.add(parserLine, LITERAL, null); // in case we "fall out the bottom", return NULL + b2.add(parserLine, RETURN); break; } - default: throw new ParserException("expected expression, found " + codeToString[tok] + ", which cannot start an expression"); + default: throw pe("expected expression, found " + codeToString[tok] + ", which cannot start an expression"); } // attempt to continue the expression continueExpr(b, minPrecedence); } + + /** + * 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 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: { + b.add(parserLine, GET_PRESERVE); + startExpr(b, precedence[tok]); + int size = b.size; + if (tok == ASSIGN_ADD || tok == ASSIGN_SUB) { + b.add(parserLine, tok); + } + // tok-1 is always s/^ASSIGN_// (0 is BITOR, 1 is ASSIGN_BITOR, etc) + b.add(parserLine, tok - 1, tok-1==ADD ? new Integer(2) : null); + b.add(parserLine, PUT); + b.add(parserLine, SWAP); + b.add(parserLine, POP); + if (tok == ASSIGN_ADD || tok == ASSIGN_SUB) b.set(size, tok, new Integer(b.size - size)); + break; + } + case INC: case DEC: { // postfix + b.add(parserLine, GET_PRESERVE, Boolean.TRUE); + b.add(parserLine, LITERAL, new Integer(1)); + b.add(parserLine, tok == INC ? ADD : SUB, null); + b.add(parserLine, PUT, null); + b.add(parserLine, SWAP, null); + b.add(parserLine, POP, null); + b.add(parserLine, LITERAL, new Integer(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: { + int n = parseArgs(b, false); + + // if the object supports GETCALL, we use this, and jump over the following two instructions + b.add(parserLine,CALLMETHOD,new Integer(n)); + b.add(parserLine,GET); + b.add(parserLine,CALL,new Integer(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; + } + } + } + + /** * Assuming that a complete expression has just been parsed, * continueExpr will attempt to extend this expression by @@ -274,7 +381,12 @@ class Parser extends Lexer implements ByteCodes { * If any bytecodes are appended, they will not alter the stack * depth. */ - private void continueExpr(CompiledFunction b, int minPrecedence) throws IOException { + 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; @@ -284,90 +396,73 @@ class Parser extends Lexer implements ByteCodes { } switch (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: { - b.set(b.size() - 1, b.GET_PRESERVE, new Boolean(true)); // FIXME should use AssignTarget - startExpr(b, precedence[tok - 1]); - b.add(line, tok - 1); - b.add(line, PUT); - b.add(line, SWAP); - b.add(line, POP); - break; - } - case INC: case DEC: { // postfix - b.set(b.size() - 1, tok, new Boolean(false)); // FIXME use assignmenttarget - break; - } case LP: { // invocation (not grouping) - int i = 0; - while(peekToken() != RP) { - i++; - if (peekToken() != COMMA) { - startExpr(b, -1); - if (peekToken() == RP) break; - } - consume(COMMA); - } - consume(RP); - b.add(line, CALL, new Integer(i)); + int n = parseArgs(b, true); + b.add(parserLine, CALL, new Integer(n)); break; } 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 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(line, 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, new Integer(count)); break; } case OR: case AND: { - b.add(line, tok == AND ? b.JF : b.JT, new Integer(0)); // test to see if we can short-circuit - int size = b.size(); - startExpr(b, precedence[tok]); // otherwise check the second value - b.add(line, JMP, new Integer(2)); // leave the second value on the stack and jump to the end - b.add(line, LITERAL, tok == AND ? new Boolean(false) : new Boolean(true)); // target of the short-circuit jump is here - b.set(size - 1, new Integer(b.size() - size)); // write the target of the short-circuit jump + b.add(parserLine, tok == AND ? b.JF : b.JT, new Integer(0)); // 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, new Integer(2)); // leave the second value on the stack and jump to the end + b.add(parserLine, LITERAL, tok == AND ? + new Boolean(false) : new Boolean(true)); // target of the short-circuit jump is here + b.set(size - 1, new Integer(b.size - size)); // write the target of the short-circuit jump break; } - case DOT: { // FIXME, assigntarget - consume(NAME); - String target = string; - if (peekToken() == ASSIGN) { - consume(ASSIGN); - b.add(line, LITERAL, target); - startExpr(b, -1); - b.add(line, PUT); - b.add(line, SWAP); - b.add(line, POP); + case DOT: { + // support foo..bar syntax for foo[""].bar + if (peekToken() == DOT) { + string = ""; } else { - b.add(line, LITERAL, target); - b.add(line, GET); + consume(NAME); } + b.add(parserLine, LITERAL, string); + continueExprAfterAssignable(b,minPrecedence); break; } case LB: { // subscripting (not array constructor) startExpr(b, -1); consume(RB); - if (peekToken() == ASSIGN) { // FIXME: assigntarget - consume(ASSIGN); - startExpr(b, -1); - b.add(line, PUT); - b.add(line, SWAP); - b.add(line, POP); - } else { - b.add(line, GET); - } + continueExprAfterAssignable(b,minPrecedence); break; } case HOOK: { - b.add(line, JF, new Integer(0)); // jump to the if-false expression - int size = b.size(); - startExpr(b, -1); // write the if-true expression - b.add(line, JMP, new Integer(0)); // if true, jump *over* the if-false expression - b.set(size - 1, new Integer(b.size() - size + 1)); // now we know where the target of the jump is + b.add(parserLine, JF, new Integer(0)); // jump to the if-false expression + int size = b.size; + startExpr(b, minPrecedence); // write the if-true expression + b.add(parserLine, JMP, new Integer(0)); // if true, jump *over* the if-false expression + b.set(size - 1, new Integer(b.size - size + 1)); // now we know where the target of the jump is consume(COLON); - size = b.size(); - startExpr(b, -1); // write the if-false expression - b.set(size - 1, new Integer(b.size() - size + 1)); // this is the end; jump to here + size = b.size; + startExpr(b, minPrecedence); // write the if-false expression + b.set(size - 1, new Integer(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: { @@ -379,9 +474,31 @@ class Parser extends Lexer implements ByteCodes { continueExpr(b, minPrecedence); // try to continue the expression } + // parse a set of comma separated function arguments, assume LP has already been consumed + // if swap is true, (because the function is already on the stack) we will SWAP after each argument to keep it on top + private int parseArgs(JSFunction b, boolean swap) throws IOException { + int i = 0; + while(peekToken() != RP) { + i++; + if (peekToken() != COMMA) { + startExpr(b, NO_COMMA); + if (swap) b.add(parserLine, SWAP); + if (peekToken() == RP) break; + } + consume(COMMA); + } + consume(RP); + return i; + } + /** Parse a block of statements which must be surrounded by LC..RC. */ - void parseBlock(CompiledFunction b) throws IOException { parseBlock(b, null); } - void parseBlock(CompiledFunction b, String label) throws IOException { + void parseBlock(JSFunction b) throws IOException { parseBlock(b, null); } + void parseBlock(JSFunction b, String label) throws IOException { + int saveParserLine = parserLine; + _parseBlock(b, label); + parserLine = saveParserLine; + } + void _parseBlock(JSFunction b, String label) throws IOException { if (peekToken() == -1) return; else if (peekToken() != LC) parseStatement(b, null); else { @@ -392,45 +509,49 @@ class Parser extends Lexer implements ByteCodes { } /** Parse a single statement, consuming the RC or SEMI which terminates it. */ - void parseStatement(CompiledFunction b, String label) throws IOException { + void parseStatement(JSFunction b, String label) throws IOException { + int saveParserLine = parserLine; + _parseStatement(b, label); + parserLine = saveParserLine; + } + void _parseStatement(JSFunction b, String label) throws IOException { int tok = peekToken(); if (tok == -1) return; switch(tok = getToken()) { case THROW: case ASSERT: case RETURN: { if (tok == RETURN && peekToken() == SEMI) - b.add(line, LITERAL, null); + b.add(parserLine, LITERAL, null); else startExpr(b, -1); - b.add(line, tok); + b.add(parserLine, tok); consume(SEMI); break; } case BREAK: case CONTINUE: { if (peekToken() == NAME) consume(NAME); - b.add(line, tok, string); + b.add(parserLine, tok, string); consume(SEMI); break; } case VAR: { - b.add(line, TOPSCOPE); // push the current scope + b.add(parserLine, TOPSCOPE); // push the current scope while(true) { consume(NAME); - String name = string; - b.add(line, LITERAL, name); // push the name to be declared - b.add(line, DECLARE); // declare it - if (peekToken() == ASSIGN) { // if there is an '=' after the variable name - b.add(line, LITERAL, name); // put the var name back on the stack + b.add(parserLine, DECLARE, string); // declare it + if (peekToken() == ASSIGN) { // if there is an '=' after the variable name consume(ASSIGN); - startExpr(b, -1); - b.add(line, PUT); // assign it - b.add(line, POP); // clean the stack - } + startExpr(b, NO_COMMA); + b.add(parserLine, PUT); // assign it + b.add(parserLine, POP); // clean the stack + } else { + b.add(parserLine, POP); // pop the string pushed by declare + } if (peekToken() != COMMA) break; consume(COMMA); } - b.add(line, POP); // pop off the topscope - if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1) consume(SEMI); + b.add(parserLine, POP); // pop off the topscope + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1 && mostRecentlyReadToken != SEMI) consume(SEMI); break; } case IF: { @@ -438,119 +559,137 @@ class Parser extends Lexer implements ByteCodes { startExpr(b, -1); consume(RP); - b.add(line, JF, new Integer(0)); // if false, jump to the else-block - int size = b.size(); + b.add(parserLine, JF, new Integer(0)); // if false, jump to the else-block + int size = b.size; parseStatement(b, null); if (peekToken() == ELSE) { consume(ELSE); - b.add(line, JMP, new Integer(0)); // if we took the true-block, jump over the else-block - b.set(size - 1, new Integer(b.size() - size + 1)); - size = b.size(); + b.add(parserLine, JMP, new Integer(0)); // if we took the true-block, jump over the else-block + b.set(size - 1, new Integer(b.size - size + 1)); + size = b.size; parseStatement(b, null); } - b.set(size - 1, new Integer(b.size() - size + 1)); // regardless of which branch we took, b[size] needs to point here + b.set(size - 1, new Integer(b.size - size + 1)); // regardless of which branch we took, b[size] needs to point here break; } case WHILE: { consume(LP); - if (label != null) b.add(line, LABEL, label); - b.add(line, LOOP); - int size = b.size(); - b.add(line, POP); // discard the first-iteration indicator + if (label != null) b.add(parserLine, LABEL, label); + b.add(parserLine, LOOP); + int size = b.size; + b.add(parserLine, POP); // discard the first-iteration indicator startExpr(b, -1); - b.add(line, JT, new Integer(2)); // if the while() clause is true, jump over the BREAK - b.add(line, BREAK); + b.add(parserLine, JT, new Integer(2)); // if the while() clause is true, jump over the BREAK + b.add(parserLine, BREAK); consume(RP); parseStatement(b, null); - b.add(line, CONTINUE); // if we fall out of the end, definately continue - b.set(size - 1, new Integer(b.size() - size + 1)); // end of the loop + b.add(parserLine, CONTINUE); // if we fall out of the end, definately continue + b.set(size - 1, new Integer(b.size - size + 1)); // end of the loop break; } case SWITCH: { consume(LP); - if (label != null) b.add(line, LABEL, label); - b.add(line, LOOP); - int size0 = b.size(); + if (label != null) b.add(parserLine, LABEL, label); + b.add(parserLine, LOOP); + int size0 = b.size; startExpr(b, -1); consume(RP); consume(LC); while(true) - if (peekToken() == CASE) { // we compile CASE statements like a bunch of if..else's + if (peekToken() == CASE) { // we compile CASE statements like a bunch of if..else's consume(CASE); - b.add(line, DUP); // duplicate the switch() value; we'll consume one copy + b.add(parserLine, DUP); // duplicate the switch() value; we'll consume one copy startExpr(b, -1); consume(COLON); - b.add(line, EQ); // check if we should do this case-block - b.add(line, JF, new Integer(0)); // if not, jump to the next one - int size = b.size(); + b.add(parserLine, EQ); // check if we should do this case-block + b.add(parserLine, JF, new Integer(0)); // if not, jump to the next one + int size = b.size; while(peekToken() != CASE && peekToken() != DEFAULT && peekToken() != RC) parseStatement(b, null); - b.set(size - 1, new Integer(1 + b.size() - size)); + b.set(size - 1, new Integer(1 + b.size - size)); } else if (peekToken() == DEFAULT) { consume(DEFAULT); consume(COLON); while(peekToken() != CASE && peekToken() != DEFAULT && peekToken() != RC) parseStatement(b, null); } else if (peekToken() == RC) { consume(RC); - b.add(line, BREAK); // break out of the loop if we 'fall through' + b.add(parserLine, BREAK); // break out of the loop if we 'fall through' break; } else { - throw new ParserException("expected CASE, DEFAULT, or RC; got " + codeToString[peekToken()]); + throw pe("expected CASE, DEFAULT, or RC; got " + codeToString[peekToken()]); } - b.set(size0 - 1, new Integer(b.size() - size0 + 1)); // end of the loop + b.set(size0 - 1, new Integer(b.size - size0 + 1)); // end of the loop break; } case DO: { - if (label != null) b.add(line, LABEL, label); - b.add(line, LOOP); - int size = b.size(); + if (label != null) b.add(parserLine, LABEL, label); + b.add(parserLine, LOOP); + int size = b.size; parseStatement(b, null); consume(WHILE); consume(LP); startExpr(b, -1); - b.add(line, JT, new Integer(2)); // check the while() clause; jump over the BREAK if true - b.add(line, BREAK); - b.add(line, CONTINUE); + b.add(parserLine, JT, new Integer(2)); // check the while() clause; jump over the BREAK if true + b.add(parserLine, BREAK); + b.add(parserLine, CONTINUE); consume(RP); consume(SEMI); - b.set(size - 1, new Integer(b.size() - size + 1)); // end of the loop; write this location to the LOOP instruction + b.set(size - 1, new Integer(b.size - size + 1)); // end of the loop; write this location to the LOOP instruction break; } case TRY: { - b.add(line, TRY); - int size = b.size(); - parseStatement(b, null); // parse the expression to be TRYed - b.add(line, POP); // pop the TryMarker - b.add(line, JMP); // jump forward to the end of the catch block - int size2 = b.size(); - b.set(size - 1, new Integer(b.size() - size + 1)); // the TRY argument points at the start of the CATCH block + b.add(parserLine, TRY); // try bytecode causes a TryMarker to be pushed + int tryInsn = b.size - 1; + // parse the expression to be TRYed + parseStatement(b, null); + // pop the try marker. this is pushed when the TRY bytecode is executed + b.add(parserLine, POP); + // jump forward to the end of the catch block, start of the finally block + b.add(parserLine, JMP); + int successJMPInsn = b.size - 1; + if (peekToken() != CATCH && peekToken() != FINALLY) + throw pe("try without catch or finally"); + + int catchJMPDistance = -1; if (peekToken() == CATCH) { + catchJMPDistance = b.size - tryInsn; + String exceptionVar; getToken(); consume(LP); consume(NAME); + exceptionVar = string; consume(RP); - // FIXME, we need an extra scope here - b.add(line, TOPSCOPE); // the exception is on top of the stack; put it to the chosen name - b.add(line, SWAP); - b.add(line, LITERAL); - b.add(line, SWAP); - b.add(line, PUT); - b.add(line, POP); - b.add(line, POP); + b.add(parserLine, TOPSCOPE); // the exception is on top of the stack; put it to the chosen name + b.add(parserLine, SWAP); + b.add(parserLine, LITERAL,exceptionVar); + b.add(parserLine, SWAP); + b.add(parserLine, PUT); + b.add(parserLine, POP); + b.add(parserLine, POP); parseStatement(b, null); + // pop the try and catch markers + b.add(parserLine,POP); + b.add(parserLine,POP); } // jump here if no exception was thrown - b.set(size2 - 1, new Integer(b.size() - size2 + 1)); - - // FIXME: not implemented correctly + b.set(successJMPInsn, new Integer(b.size - successJMPInsn)); + + int finallyJMPDistance = -1; if (peekToken() == FINALLY) { + b.add(parserLine, LITERAL, null); // null FinallyData + finallyJMPDistance = b.size - tryInsn; consume(FINALLY); parseStatement(b, null); + b.add(parserLine,FINALLY_DONE); } + + // setup the TRY arguments + b.set(tryInsn, new int[] { catchJMPDistance, finallyJMPDistance }); + break; } @@ -558,77 +697,77 @@ class Parser extends Lexer implements ByteCodes { consume(LP); tok = getToken(); - boolean hadVar = false; // if it's a for..in, we ignore the VAR + boolean hadVar = false; // if it's a for..in, we ignore the VAR if (tok == VAR) { hadVar = true; tok = getToken(); } String varName = string; - boolean forIn = peekToken() == IN; // determine if this is a for..in loop or not + boolean forIn = peekToken() == IN; // determine if this is a for..in loop or not pushBackToken(tok, varName); if (forIn) { - b.add(line, NEWSCOPE); - b.add(line, LITERAL, varName); // declare the new variable - b.add(line, DECLARE); + b.add(parserLine, NEWSCOPE); // for-loops always create new scopes + b.add(parserLine, LITERAL, varName); // declare the new variable + b.add(parserLine, DECLARE); - b.add(line, LOOP); // we actually only add this to ensure that BREAK works - b.add(line, POP); // discard the first-iteration indicator - int size = b.size(); + b.add(parserLine, LOOP); // we actually only add this to ensure that BREAK works + b.add(parserLine, POP); // discard the first-iteration indicator + int size = b.size; consume(NAME); consume(IN); startExpr(b, -1); - b.add(line, PUSHKEYS); // push the keys as an array; check the length - b.add(line, LITERAL, "length"); - b.add(line, GET); + b.add(parserLine, PUSHKEYS); // push the keys as an array; check the length + b.add(parserLine, LITERAL, "length"); + b.add(parserLine, GET); consume(RP); - b.add(line, LITERAL, new Integer(1)); // decrement the length - b.add(line, SUB); - b.add(line, DUP); - b.add(line, LITERAL, new Integer(0)); // see if we've exhausted all the elements - b.add(line, LT); - b.add(line, JF, new Integer(2)); - b.add(line, BREAK); // if we have, then BREAK - b.add(line, GET_PRESERVE); // get the key out of the keys array - b.add(line, LITERAL, varName); - b.add(line, PUT); // write it to this[varName] - parseStatement(b, null); // do some stuff - b.add(line, CONTINUE); // continue if we fall out the bottom - - b.set(size - 1, new Integer(b.size() - size + 1)); // BREAK to here - b.add(line, OLDSCOPE); // restore the scope + b.add(parserLine, LITERAL, new Integer(1)); // decrement the length + b.add(parserLine, SUB); + b.add(parserLine, DUP); + b.add(parserLine, LITERAL, new Integer(0)); // see if we've exhausted all the elements + b.add(parserLine, LT); + b.add(parserLine, JF, new Integer(2)); + b.add(parserLine, BREAK); // if we have, then BREAK + b.add(parserLine, GET_PRESERVE); // get the key out of the keys array + b.add(parserLine, LITERAL, varName); + b.add(parserLine, PUT); // write it to this[varName] + parseStatement(b, null); // do some stuff + b.add(parserLine, CONTINUE); // continue if we fall out the bottom + + b.set(size - 1, new Integer(b.size - size + 1)); // BREAK to here + b.add(parserLine, OLDSCOPE); // restore the scope } else { - if (hadVar) pushBackToken(VAR, null); // yeah, this actually matters - b.add(line, NEWSCOPE); // grab a fresh scope + if (hadVar) pushBackToken(VAR, null); // yeah, this actually matters + b.add(parserLine, NEWSCOPE); // grab a fresh scope - parseStatement(b, null); // initializer - CompiledFunction e2 = // we need to put the incrementor before the test - new CompiledFunction(sourceName, line, null); // so we save the test here + parseStatement(b, null); // initializer + JSFunction e2 = // we need to put the incrementor before the test + new JSFunction(sourceName, parserLine, null, null); // so we save the test here if (peekToken() != SEMI) startExpr(e2, -1); else - e2.add(line, b.LITERAL, Boolean.TRUE); // handle the for(foo;;foo) case + e2.add(parserLine, b.LITERAL, Boolean.TRUE); // handle the for(foo;;foo) case consume(SEMI); - if (label != null) b.add(line, LABEL, label); - b.add(line, LOOP); - int size2 = b.size(); + if (label != null) b.add(parserLine, LABEL, label); + b.add(parserLine, LOOP); + int size2 = b.size; - b.add(line, JT, new Integer(0)); // if we're on the first iteration, jump over the incrementor - int size = b.size(); - if (peekToken() != RP) { // do the increment thing + b.add(parserLine, JT, new Integer(0)); // if we're on the first iteration, jump over the incrementor + int size = b.size; + if (peekToken() != RP) { // do the increment thing startExpr(b, -1); - b.add(line, POP); + b.add(parserLine, POP); } - b.set(size - 1, new Integer(b.size() - size + 1)); + b.set(size - 1, new Integer(b.size - size + 1)); consume(RP); - b.paste(e2); // ok, *now* test if we're done yet - b.add(line, JT, new Integer(2)); // break out if we don't meet the test - b.add(line, BREAK); + b.paste(e2); // ok, *now* test if we're done yet + b.add(parserLine, JT, new Integer(2)); // break out if we don't meet the test + b.add(parserLine, BREAK); parseStatement(b, null); - b.add(line, CONTINUE); // if we fall out the bottom, CONTINUE - b.set(size2 - 1, new Integer(b.size() - size2 + 1)); // end of the loop + b.add(parserLine, CONTINUE); // if we fall out the bottom, CONTINUE + b.set(size2 - 1, new Integer(b.size - size2 + 1)); // end of the loop - b.add(line, OLDSCOPE); // get our scope back + b.add(parserLine, OLDSCOPE); // get our scope back } break; } @@ -642,25 +781,27 @@ class Parser extends Lexer implements ByteCodes { } else { // expression pushBackToken(NAME, possiblyTheLabel); startExpr(b, -1); - b.add(line, POP); - if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1) consume(SEMI); + b.add(parserLine, POP); + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1 && mostRecentlyReadToken != SEMI) consume(SEMI); break; } } - case SEMI: return; // yep, the null statement is valid + case SEMI: return; // yep, the null statement is valid case LC: { // blocks are statements too pushBackToken(); + b.add(parserLine, NEWSCOPE); parseBlock(b, label); + b.add(parserLine, OLDSCOPE); break; } default: { // hope that it's an expression pushBackToken(); startExpr(b, -1); - b.add(line, POP); - if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1) consume(SEMI); + b.add(parserLine, POP); + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1 && mostRecentlyReadToken != SEMI) consume(SEMI); break; } } @@ -668,8 +809,7 @@ class Parser extends Lexer implements ByteCodes { // ParserException ////////////////////////////////////////////////////////////////////// - - private class ParserException extends IOException { public ParserException(String s) { super(sourceName + ":" + line + " " + s); } } + private IOException pe(String s) { return new IOException(sourceName + ":" + parserLine + " " + s); } }