X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=src%2Forg%2Fxwt%2Fjs%2FParser.java;h=4abbdea9867e24f0911046f6b20b7c922a427a15;hb=bebd488df5327f1ffe212c50236d6198f313ff3c;hp=67cb0a1e6b1e644c2cf661c76168137f8d254e07;hpb=3ab44d1c726c06f8843dee1601b0142fe94c3478;p=org.ibex.core.git diff --git a/src/org/xwt/js/Parser.java b/src/org/xwt/js/Parser.java index 67cb0a1..4abbdea 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 2003 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 ForthBlock's */ -public class Parser extends Lexer implements OpCodes { +/** + * Parses a stream of lexed tokens into a tree of CompiledFunctionImpl'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 { + // 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) { - ForthBlock block = p.parseStatement(false); - if (block == null) return; - System.out.println(block); - if (p.peekToken() == -1) return; - } + CompiledFunctionImpl block = new JS.CompiledFunction("stdin", 0, new InputStreamReader(System.in), null); + if (block == null) return; + System.out.println(block); } @@ -31,616 +83,705 @@ public class Parser extends Lexer implements OpCodes { 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[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; - 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] = 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] = 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 ///////////////////////////////////////////////////////// - public void consume(int code) throws IOException { - if (getToken() != code) - throw new ParserException("expected " + codeToString[code] + ", got " + (op == -1 ? "EOL" : codeToString[op])); + /** gets a token and throws an exception if it is not code */ + private void consume(int code) throws IOException { + if (getToken() != code) throw pe("expected " + codeToString[code] + ", got " + (op == -1 ? "EOF" : codeToString[op])); + } + + /** + * 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(CompiledFunctionImpl appendTo, int minPrecedence) throws IOException { + int saveParserLine = parserLine; + _startExpr(appendTo, minPrecedence); + parserLine = saveParserLine; } + private void _startExpr(CompiledFunctionImpl appendTo, int minPrecedence) throws IOException { + int tok = getToken(); + CompiledFunctionImpl 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 THIS: b.add(parserLine, TOPSCOPE, null); 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(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(); + b.add(parserLine, LITERAL, new Integer(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, 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(parserLine, LITERAL, new Double(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, tok, Boolean.TRUE); + 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; + CompiledFunctionImpl b2 = new JS.CompiledFunction(sourceName, parserLine, null, 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 + if (peekToken() == NAME) { + consume(NAME); // a named argument + String varName = string; + + b2.add(parserLine, DUP); // dup the args array + b2.add(parserLine, GET, new Integer(numArgs)); // 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(parserLine, POP); // pop off the arguments array + b2.add(parserLine, POP); // pop off TOPSCOPE + + if(peekToken() != LC) + throw pe("Functions must have a block surrounded by curly brackets"); + + parseBlock(b2, null); // the function body + + if(b2.get(b2.size()-1) != RETURN) { + 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 ForthBlock parseMaximalForthBlock() throws IOException { return parseMaximalForthBlock(null, -1); } - public ForthBlock parseMaximalForthBlock(ForthBlock prefix, int minPrecedence) throws IOException { - while(true) { - if (peekToken() == -1) break; - ForthBlock save = prefix; - prefix = parseSingleForthBlock(prefix, minPrecedence); - if (save == prefix) break; - if (prefix == null) throw new ParserException("parseSingleForthBlock() returned null"); - } - 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(CompiledFunctionImpl b,int minPrecedence) throws IOException { + int saveParserLine = parserLine; + _continueExprAfterAssignable(b,minPrecedence); + parserLine = saveParserLine; } + private void _continueExprAfterAssignable(CompiledFunctionImpl 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_ADD: case ASSIGN_SUB: case ASSIGN_MUL: case ASSIGN_DIV: case ASSIGN_MOD: { + b.add(parserLine, GET_PRESERVE); + startExpr(b, precedence[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); + break; + } + case INC: case DEC: { // postfix + b.add(parserLine, tok, Boolean.FALSE); + 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); + b.add(parserLine,CALLMETHOD,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; + } + } + } + - public ForthBlock parseSingleForthBlock(ForthBlock prefix, int minPrecedence) throws IOException { - ForthBlock 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; - - switch (tok) { - - case VAR: { - if (prefix != null) { pushBackToken(); return prefix; } - ForthBlock b = new ForthBlock(curLine, sourceName); - b.add(OpCodes.THIS); - while(true) { - consume(NAME); - String name = string; - b.add(b.LITERAL, name); - b.add(b.DECLARE); - if (peekToken() == ASSIGN) { - b.add(b.LITERAL, name); - consume(ASSIGN); - b.add(b.EXPR, parseMaximalForthBlock()); - b.add(b.PUT); - b.add(b.POP); - } - if (peekToken() != COMMA) break; - consume(COMMA); - } - return parseSingleForthBlock(b, minPrecedence); - } - - case IN: pushBackToken(); return prefix; - - case IF: { - if (prefix != null) { pushBackToken(); return prefix; } - ForthBlock b = new ForthBlock(curLine, sourceName); - consume(LP); - b.add(b.EXPR, parseMaximalForthBlock()); - consume(RP); - b.add(b.JF, new Integer(3)); - b.add(b.EXPR, parseStatement(false)); - b.add(b.JMP, new Integer(2)); - if (peekToken() != ELSE) return parseSingleForthBlock(b.add(b.LITERAL, null), minPrecedence); - consume(ELSE); - b.add(b.EXPR, parseStatement(false)); - return parseSingleForthBlock(b, minPrecedence); - } - - // FIXME: ugly hack!! - case ASSIGN_BITOR: if (tok == ASSIGN_BITOR) tok = BITOR; - case ASSIGN_BITXOR: if (tok == ASSIGN_BITXOR) tok = BITXOR; - case ASSIGN_BITAND: if (tok == ASSIGN_BITAND) tok = BITAND; - case ASSIGN_LSH: if (tok == ASSIGN_LSH) tok = LSH; - case ASSIGN_RSH: if (tok == ASSIGN_RSH) tok = RSH; - case ASSIGN_URSH: if (tok == ASSIGN_URSH) tok = URSH; - case ASSIGN_ADD: if (tok == ASSIGN_ADD) tok = ADD; - case ASSIGN_SUB: if (tok == ASSIGN_SUB) tok = SUB; - case ASSIGN_MUL: if (tok == ASSIGN_MUL) tok = MUL; - case ASSIGN_DIV: if (tok == ASSIGN_DIV) tok = DIV; - case ASSIGN_MOD: if (tok == ASSIGN_MOD) tok = MOD; - { - ForthBlock b = (ForthBlock)prefix; - b.set(b.size() - 1, b.GET_PRESERVE, new Boolean(true)); - b.add(b.EXPR, parseMaximalForthBlock(null, precedence[tok])); - b.add(tok); - b.add(b.PUT); - b.add(b.SWAP); - b.add(b.POP); - return parseSingleForthBlock(b, minPrecedence); - } - - case INC: case DEC: - if (prefix == null) { - // prefix - ForthBlock b = (ForthBlock)parseMaximalForthBlock(null, precedence[tok]); - b.set(b.size() - 1, tok, new Boolean(true)); - return parseSingleForthBlock(b, minPrecedence); - } else { - // postfix - ForthBlock b = (ForthBlock)prefix; - b.set(b.size() - 1, tok, new Boolean(false)); - return parseSingleForthBlock(b, minPrecedence); - } - - case LP: - if (prefix == null) { // grouping - ForthBlock b = new ForthBlock(curLine, sourceName, ForthBlock.EXPR, parseMaximalForthBlock()); - consume(RP); - return parseSingleForthBlock(b, minPrecedence); - - } else { // invocation - ForthBlock b = new ForthBlock(curLine, sourceName); - int i = 0; - b.add(b.EXPR, prefix); - while(peekToken() != RP) { - b.add(b.EXPR, parseMaximalForthBlock()); - i++; - if (peekToken() == RP) break; - consume(COMMA); - } - consume(RP); - b.add(b.CALL, new Integer(i)); - return parseSingleForthBlock(b, minPrecedence); - } - - case BANG: case BITNOT: case INSTANCEOF: case TYPEOF: { - if (prefix != null) { pushBackToken(); return prefix; } - ForthBlock b = new ForthBlock(curLine, sourceName); - b.add(b.EXPR, parseMaximalForthBlock(null, precedence[tok])); - b.add(tok); - return parseSingleForthBlock(b, minPrecedence); - } - - case SUB: - if (prefix == null && peekToken() == NUMBER) { - getToken(); - return parseSingleForthBlock(new ForthBlock(curLine, sourceName, ForthBlock.LITERAL, new Double(number.doubleValue() * -1)), minPrecedence); - } // else fall through + /** + * 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(CompiledFunctionImpl b, int minPrecedence) throws IOException { + int saveParserLine = parserLine; + _continueExpr(b, minPrecedence); + parserLine = saveParserLine; + } + private void _continueExpr(CompiledFunctionImpl 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); + 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 GT: case GE: case EQ: case NE: case LT: case LE: { - if (prefix == null) throw new ParserException("the " + codeToString[tok] + " token cannot start an expression"); - ForthBlock b = new ForthBlock(curLine, sourceName); - b.add(b.EXPR, prefix); - b.add(b.EXPR, parseMaximalForthBlock(null, precedence[tok])); - b.add(tok); - return parseSingleForthBlock(b, minPrecedence); - } - - // includes short-circuit logic - case OR: case AND: { - if (prefix == null) throw new ParserException("the " + codeToString[tok] + " token cannot start an expression"); - ForthBlock b = new ForthBlock(curLine, sourceName); - b.add(b.LITERAL, tok == AND ? new Boolean(false) : new Boolean(true)); - b.add(b.EXPR, prefix); - b.add(tok == AND ? b.JF : b.JT, new Integer(3)); - b.add(b.POP); - b.add(b.EXPR, parseMaximalForthBlock(null, precedence[tok])); - return parseSingleForthBlock(b, minPrecedence); - } - - 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 NUMBER: - if (prefix != null) { pushBackToken(); return prefix; } - return parseSingleForthBlock(new ForthBlock(curLine, sourceName, ForthBlock.LITERAL, number), minPrecedence); - - case STRING: - if (prefix != null) { pushBackToken(); return prefix; } - return parseSingleForthBlock(new ForthBlock(curLine, sourceName, ForthBlock.LITERAL, string), minPrecedence); - - case NULL: case TRUE: case FALSE: case NOP: - if (prefix != null) { pushBackToken(); return prefix; } - return parseSingleForthBlock(new ForthBlock(curLine, sourceName, ForthBlock.LITERAL, (tok == NULL || tok == NOP) ? null : new Boolean(tok == TRUE)), minPrecedence); - - case COMMA: pushBackToken(); return prefix; - - case Tokens.THIS: - if (prefix != null) { pushBackToken(); return prefix; } - return parseSingleForthBlock(new ForthBlock(curLine, sourceName, OpCodes.THIS, null), minPrecedence); - - case NAME: { - if (prefix != null) { pushBackToken(); return prefix; } - String name = string; - ForthBlock b = new ForthBlock(curLine, sourceName); - if (peekToken() == ASSIGN) { - consume(ASSIGN); - b.add(OpCodes.THIS); - b.add(ForthBlock.LITERAL, name); - b.add(ForthBlock.EXPR, parseMaximalForthBlock(null, minPrecedence)); - b.add(ForthBlock.PUT); - b.add(ForthBlock.SWAP); - b.add(ForthBlock.POP); - return parseSingleForthBlock(b, minPrecedence); - } else { - b.add(OpCodes.THIS); - b.add(ForthBlock.LITERAL, name); - b.add(ForthBlock.GET); - return parseSingleForthBlock(parseMaximalForthBlock(b, minPrecedence), minPrecedence); - } - } - - case DOT: { - consume(NAME); - String target = string; - ForthBlock b = new ForthBlock(curLine, sourceName); - b.add(b.EXPR, prefix); - if (peekToken() == ASSIGN) { - consume(ASSIGN); - ForthBlock val = parseMaximalForthBlock(); - b.add(b.LITERAL, target); - b.add(b.EXPR, val); - b.add(b.PUT); - b.add(b.SWAP); - b.add(b.POP); - } else { - b.add(b.LITERAL, target); - b.add(b.GET); - } - return parseSingleForthBlock(b, minPrecedence); - } - - case LB: { - ForthBlock b = new ForthBlock(curLine, sourceName); - if (prefix == null) { - b.add(b.ARRAY, new Integer(0)); - int i = 0; - while(true) { - ForthBlock e = parseMaximalForthBlock(); - if (e == null && peekToken() == RB) { consume(RB); return parseSingleForthBlock(b, minPrecedence); } - b.add(b.LITERAL, new Integer(i++)); - if (e == null) b.add(b.LITERAL, null); - else b.add(b.EXPR, e); - b.add(b.PUT); - b.add(b.POP); - if (peekToken() == RB) { consume(RB); return parseSingleForthBlock(b, minPrecedence); } - consume(COMMA); - } - } else { - b.add(b.EXPR, prefix); - b.add(b.EXPR, parseMaximalForthBlock()); - consume(RB); - if (peekToken() == ASSIGN) { - consume(ASSIGN); - b.add(b.EXPR, parseMaximalForthBlock()); - b.add(b.PUT); - b.add(b.SWAP); - b.add(b.POP); - } else { - b.add(b.GET); - } - return parseSingleForthBlock(b, minPrecedence); - } - } - - case LC: { - if (prefix != null) { pushBackToken(); return prefix; } - ForthBlock b = new ForthBlock(curLine, sourceName); - b.add(b.OBJECT, null); - if (peekToken() == RC) { consume(RC); return parseSingleForthBlock(b, minPrecedence); } - while(true) { - if (peekToken() != NAME && peekToken() != STRING) throw new Error("expected NAME or STRING"); - getToken(); - b.add(b.LITERAL, string); - consume(COLON); - b.add(b.EXPR, parseMaximalForthBlock()); - b.add(b.PUT); - b.add(b.POP); - if (peekToken() == RC) { consume(RC); return parseSingleForthBlock(b, minPrecedence); } - consume(COMMA); - if (peekToken() == RC) { consume(RC); return parseSingleForthBlock(b, minPrecedence); } - } - } - - case HOOK: { - ForthBlock b = new ForthBlock(curLine, sourceName); - b.add(b.EXPR, prefix); - b.add(b.JF, new Integer(3)); - b.add(b.EXPR, parseMaximalForthBlock()); - b.add(b.JMP, new Integer(2)); - consume(COLON); - b.add(b.EXPR, parseMaximalForthBlock()); - return parseSingleForthBlock(b, minPrecedence); - } - - case Tokens.FUNCTION: { - if (prefix != null) { pushBackToken(); return prefix; } - consume(LP); - ForthBlock b = new ForthBlock(curLine, sourceName); - int numArgs = 0; - b.add(OpCodes.THIS); - b.add(b.SWAP); - b.add(b.LITERAL, "arguments"); - b.add(b.LITERAL, "arguments"); - b.add(b.DECLARE); - b.add(b.SWAP); - b.add(b.PUT); - b.add(b.SWAP); - b.add(b.POP); - if (peekToken() == RP) consume(RP); - else while(true) { - if (peekToken() == COMMA) { - consume(COMMA); - } else { - consume(NAME); - - // declare the name - b.add(b.LITERAL, string); - b.add(b.DECLARE); - - // retrieve it from the arguments array - b.add(b.LITERAL, new Integer(numArgs)); - b.add(b.GET_PRESERVE); - b.add(b.SWAP); - b.add(b.POP); - - // put it to the current scope - b.add(OpCodes.THIS); - b.add(b.SWAP); - b.add(b.LITERAL, string); - b.add(b.SWAP); - b.add(b.PUT); - - // clean the stack - b.add(b.POP); - b.add(b.POP); - - if (peekToken() == RP) { consume(RP); break; } - consume(COMMA); - } - numArgs++; - } - // pop off the arguments array - b.add(b.POP); - parseStatement(true, b); - return parseSingleForthBlock(new ForthBlock(curLine, sourceName, OpCodes.FUNCTION, b), minPrecedence); - } - - case WHILE: { - if (prefix != null) { pushBackToken(); return prefix; } - consume(LP); - ForthBlock r = new ForthBlock(curLine, sourceName); - ForthBlock loop = new ForthBlock(curLine, sourceName); - r.add(loop.LOOP, loop); - r.add(r.LITERAL, null); - - loop.add(loop.EXPR, parseMaximalForthBlock()); - loop.add(loop.JT, new Integer(2)); - loop.add(Lexer.BREAK); - consume(RP); - parseStatement(false, loop); - - // if we fall out of the end, definately continue - loop.add(CONTINUE); - return parseSingleForthBlock(r, minPrecedence); - } - - case SWITCH: { - if (prefix != null) { pushBackToken(); return prefix; } - consume(LP); - ForthBlock r = new ForthBlock(curLine, sourceName); - ForthBlock loop = new ForthBlock(curLine, sourceName); - r.add(loop.LOOP, loop); - r.add(r.LITERAL, null); - loop.add(loop.EXPR, parseMaximalForthBlock()); - consume(RP); - consume(LC); - while(true) { - ForthBlock caseForthBlock; - tok = getToken(); - if (tok == CASE) { - loop.add(loop.DUP); - loop.add(loop.EXPR, parseMaximalForthBlock()); - loop.add(EQ); - loop.add(loop.JF, new Integer(2)); - } else if (tok != DEFAULT) throw new ParserException("expected CASE or DEFAULT"); - consume(COLON); - ForthBlock b = new ForthBlock(curLine, sourceName); - while(peekToken() != CASE && peekToken() != DEFAULT && peekToken() != RC) { - if ((e1 = parseStatement(false)) == null) break; - b.add(b.EXPR, e1); - } - loop.add(loop.EXPR, b); - if (peekToken() == RC) { - consume(RC); - r.add(BREAK); - return parseSingleForthBlock(r, minPrecedence); - } - } - } - - case DO: { - if (prefix != null) { pushBackToken(); return prefix; } - ForthBlock r = new ForthBlock(curLine, sourceName); - ForthBlock loop = new ForthBlock(curLine, sourceName); - r.add(loop.LOOP, loop); - r.add(r.LITERAL, null); - - parseStatement(false, loop); - consume(WHILE); - consume(LP); - loop.add(loop.EXPR, parseMaximalForthBlock()); - loop.add(loop.JT, new Integer(2)); - loop.add(Lexer.BREAK); - loop.add(Lexer.CONTINUE); - consume(RP); - consume(SEMI); - return parseSingleForthBlock(r, minPrecedence); - } - - case TRY: { - // FIXME: don't just ignore this! - // We deliberately allow you to omit braces in catch{}/finally{} if they are single statements... - if (prefix != null) { pushBackToken(); return prefix; } - ForthBlock tryBlock = parseStatement(true); - - tok = peekToken(); - if (tok == CATCH) { - getToken(); - if (getToken() != LP) throw new ParserException("expected ("); - if (getToken() != NAME) throw new ParserException("expected name"); - if (getToken() != RP) throw new ParserException("expected )"); - tok = peekToken(); - } - if (tok == FINALLY) getToken(); - - return parseSingleForthBlock(tryBlock, minPrecedence); - } - - case FOR: { - if (prefix != null) { pushBackToken(); return prefix; } - if (getToken() != LP) throw new ParserException("expected left paren"); - - tok = getToken(); - if (tok == VAR) tok = getToken(); - String varName = string; - boolean forIn = peekToken() == IN; - pushBackToken(tok, varName); - - ForthBlock b = new ForthBlock(curLine, sourceName); - if (forIn) { - consume(NAME); - consume(IN); - b.add(b.EXPR, parseMaximalForthBlock()); - b.add(b.PUSHKEYS); - b.add(b.LITERAL, "length"); - b.add(b.GET); - consume(RP); - ForthBlock b2 = new ForthBlock(curLine, sourceName); - b.add(b.SCOPE, b2); - b2.add(b.LITERAL, new Integer(1)); - b2.add(SUB); - b2.add(b.DUP); - b2.add(b.LITERAL, new Integer(0)); - b2.add(LT); - b2.add(b.JT, new Integer(7)); - b2.add(b.GET_PRESERVE); - b2.add(b.LITERAL, varName); - b2.add(b.LITERAL, varName); - b2.add(b.DECLARE); - b2.add(b.PUT); - b2.add(b.EXPR, parseStatement(false)); - b2.add(b.LITERAL, null); - return parseSingleForthBlock(b, minPrecedence); - - } else { - ForthBlock b2 = new ForthBlock(curLine, sourceName); - b.add(b.SCOPE, b2); - - e1 = parseMaximalForthBlock(); - if (e1 == null) e1 = new ForthBlock(curLine, sourceName, b.LITERAL, null); - - b2.add(b.EXPR, e1); - b2.add(b.POP); - consume(SEMI); - e2 = parseMaximalForthBlock(); - consume(SEMI); - e3 = parseMaximalForthBlock(); - consume(RP); - - if (e2 == null) e2 = new ForthBlock(curLine, sourceName, b.LITERAL, null); - if (e3 == null) e3 = new ForthBlock(curLine, sourceName, b.LITERAL, null); - - ForthBlock b3 = new ForthBlock(curLine, sourceName); - b2.add(b.LOOP, b3); - b2.add(b.LITERAL, null); - b3.add(b.JT, new Integer(3)); - b3.add(b.EXPR, e3); - b3.add(b.POP); - b3.add(b.EXPR, e2); - b3.add(b.JT, new Integer(2)); - b3.add(BREAK); - parseStatement(false, b3); - return parseSingleForthBlock(b, minPrecedence); - } - } - - default: - pushBackToken(); - return prefix; - } + 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, new Integer(count)); + break; + } + case OR: case AND: { + 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: { + 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, 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, 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: { + pushBackToken(); + return; + } + } + + continueExpr(b, minPrecedence); // try to continue the expression } - /** a block is either a single statement or a list of statements surrounded by curly braces; all expressions are also statements */ - public ForthBlock parseStatement(boolean requireBraces) throws IOException { return parseStatement(requireBraces, null); } - public ForthBlock parseStatement(boolean requireBraces, ForthBlock b) throws IOException { - ForthBlock smt = null; - int tok = peekToken(); - if (tok == -1) return null; - boolean braced = tok == LC; - if (requireBraces && !braced) throw new ParserException("expected {, got " + codeToString[tok]); - if (braced) consume(LC); - int curLine = line; - ForthBlock ret = new ForthBlock(curLine, sourceName); - ForthBlock block = b == null ? new ForthBlock(curLine, sourceName) : b; - block.add(ret.LITERAL, Boolean.TRUE); - ret.add(block.SCOPE, block); - while(true) { - switch(tok = peekToken()) { - - case LC: smt = parseStatement(true); break; - case GOTO: throw new ParserException("goto not supported"); - - case THROW: case RETURN: case ASSERT: { - getToken(); - ForthBlock r = new ForthBlock(curLine, sourceName); - if (tok == RETURN && peekToken() == SEMI) r.add(b.LITERAL, null); - else r.add(b.EXPR, parseMaximalForthBlock()); - consume(SEMI); - r.add(tok); - smt = r; - break; - } - - case BREAK: case CONTINUE: { - getToken(); - if (peekToken() == NAME) consume(NAME); - smt = new ForthBlock(curLine, sourceName, tok, string); - consume(SEMI); - break; - } - - case RC: - if (braced) consume(RC); - return block.size() == 0 ? null : ret; - - case SEMI: - consume(SEMI); - if (!braced) return block.size() == 0 ? null : ret; - continue; - - case NAME: { - String name = string; - consume(NAME); - if (peekToken() == COLON) { - consume(COLON); - smt = new ForthBlock(curLine, sourceName, ForthBlock.LABEL, string); - break; - } else { - pushBackToken(NAME, name); - // fall through to default case - } - } - - case -1: - default: - smt = parseMaximalForthBlock(); - if (smt == null) return block.size() == 0 ? null : ret; - if (peekToken() == SEMI) getToken(); - break; - } - - if (!braced) return smt; - block.add(block.EXPR, smt); - block.add(block.POP); - } + // parse a set of comma separated function arguments, assume LP has already been consumed + private int parseArgs(CompiledFunctionImpl b) throws IOException { + int i = 0; + while(peekToken() != RP) { + i++; + if (peekToken() != COMMA) { + startExpr(b, NO_COMMA); + if (peekToken() == RP) break; + } + consume(COMMA); + } + consume(RP); + return i; } - class ParserException extends RuntimeException { - public ParserException(String s) { super(sourceName + ":" + line + " " + s); } + /** Parse a block of statements which must be surrounded by LC..RC. */ + void parseBlock(CompiledFunctionImpl b) throws IOException { parseBlock(b, null); } + void parseBlock(CompiledFunctionImpl b, String label) throws IOException { + int saveParserLine = parserLine; + _parseBlock(b, label); + parserLine = saveParserLine; + } + void _parseBlock(CompiledFunctionImpl b, String label) throws IOException { + if (peekToken() == -1) return; + else if (peekToken() != LC) parseStatement(b, null); + else { + consume(LC); + while(peekToken() != RC && peekToken() != -1) parseStatement(b, null); + consume(RC); + } + } + + /** Parse a single statement, consuming the RC or SEMI which terminates it. */ + void parseStatement(CompiledFunctionImpl b, String label) throws IOException { + int saveParserLine = parserLine; + _parseStatement(b, label); + parserLine = saveParserLine; } + void _parseStatement(CompiledFunctionImpl 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(parserLine, LITERAL, null); + else + startExpr(b, -1); + b.add(parserLine, tok); + consume(SEMI); + break; + } + case BREAK: case CONTINUE: { + if (peekToken() == NAME) consume(NAME); + b.add(parserLine, tok, string); + consume(SEMI); + break; + } + case VAR: { + b.add(parserLine, TOPSCOPE); // push the current scope + while(true) { + consume(NAME); + b.add(parserLine, DECLARE, string); // declare it + if (peekToken() == ASSIGN) { // if there is an '=' after the variable name + consume(ASSIGN); + 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(parserLine, POP); // pop off the topscope + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1 && mostRecentlyReadToken != SEMI) consume(SEMI); + break; + } + case IF: { + consume(LP); + startExpr(b, -1); + consume(RP); + + 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(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 + break; + } + case WHILE: { + consume(LP); + 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(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(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(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 + consume(CASE); + b.add(parserLine, DUP); // duplicate the switch() value; we'll consume one copy + startExpr(b, -1); + consume(COLON); + 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)); + } 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(parserLine, BREAK); // break out of the loop if we 'fall through' + break; + } else { + throw pe("expected CASE, DEFAULT, or RC; got " + codeToString[peekToken()]); + } + b.set(size0 - 1, new Integer(b.size() - size0 + 1)); // end of the loop + break; + } + + case DO: { + 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(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 + break; + } + + case TRY: { + 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); + 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(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; + } + + case FOR: { + consume(LP); + + tok = getToken(); + 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 + pushBackToken(tok, varName); + + if (forIn) { + 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(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(parserLine, PUSHKEYS); // push the keys as an array; check the length + b.add(parserLine, LITERAL, "length"); + b.add(parserLine, GET); + consume(RP); + + 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(parserLine, NEWSCOPE); // grab a fresh scope + + parseStatement(b, null); // initializer + CompiledFunctionImpl e2 = // we need to put the incrementor before the test + new JS.CompiledFunction(sourceName, parserLine, null, null); // so we save the test here + if (peekToken() != SEMI) + startExpr(e2, -1); + else + e2.add(parserLine, b.LITERAL, Boolean.TRUE); // handle the for(foo;;foo) case + consume(SEMI); + if (label != null) b.add(parserLine, LABEL, label); + b.add(parserLine, LOOP); + int size2 = b.size(); + + 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(parserLine, POP); + } + 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(parserLine, JT, new Integer(2)); // break out if we don't meet the test + b.add(parserLine, BREAK); + parseStatement(b, null); + 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(parserLine, OLDSCOPE); // get our scope back + } + break; + } + + case NAME: { // either a label or an identifier; this is the one place we're not LL(1) + String possiblyTheLabel = string; + if (peekToken() == COLON) { // label + consume(COLON); + parseStatement(b, possiblyTheLabel); + break; + } else { // expression + pushBackToken(NAME, possiblyTheLabel); + startExpr(b, -1); + 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 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(parserLine, POP); + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1 && mostRecentlyReadToken != SEMI) consume(SEMI); + break; + } + } + } + + + // ParserException ////////////////////////////////////////////////////////////////////// + private IOException pe(String s) { return new IOException(sourceName + ":" + parserLine + " " + s); } }