X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=src%2Forg%2Fxwt%2Fjs%2FParser.java;h=f5ddd423052f82f8e8c113ef2308ef91fe499f59;hb=d5f0e6ca91a3359b09352d3df367d6b6bd22f096;hp=ff59d3bc5bc3de3458e5afa49bb4e07991c08f8c;hpb=1bd9bc0c2100ba4a1dc2a5f4ca25ec9e0d1b2265;p=org.ibex.core.git diff --git a/src/org/xwt/js/Parser.java b/src/org/xwt/js/Parser.java index ff59d3b..f5ddd42 100644 --- a/src/org/xwt/js/Parser.java +++ b/src/org/xwt/js/Parser.java @@ -1,31 +1,69 @@ // Copyright 2003 Adam Megacz, see the COPYING file for licensing [GPL] package org.xwt.js; +// FIXME: line number accuracy + import org.xwt.util.*; import java.io.*; -/** Parses a stream of lexed tokens into a tree of ByteCodeBlock's */ +/** + * Parses a stream of lexed tokens into a tree of CompiledFunction'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. + + * + * 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) { - ByteCodeBlock block = new ByteCodeBlock(0, "stdin"); - p.parseStatement(false, block); - if (block == null) return; - System.out.println(block); - if (p.peekToken() == -1) return; - } + CompiledFunction block = new CompiledFunction("stdin", 0, new InputStreamReader(System.in), null); + if (block == null) return; + System.out.println(block); } @@ -50,7 +88,7 @@ class Parser extends Lexer implements ByteCodes { 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[BITNOT] = 15; precedence[INC] = precedence[DEC] = 16; precedence[LP] = 17; precedence[LB] = 18; @@ -61,572 +99,586 @@ class Parser extends Lexer implements ByteCodes { // Parsing Logic ///////////////////////////////////////////////////////// /** gets a token and throws an exception if it is not code */ - public void consume(int code) throws IOException { - if (getToken() != code) - throw new ParserException("expected " + codeToString[code] + ", got " + (op == -1 ? "EOL" : codeToString[op])); + private void consume(int code) throws IOException { + if (getToken() != code) throw new ParserException("expected " + codeToString[code] + ", got " + (op == -1 ? "EOF" : codeToString[op])); } - /** append the largest expression beginning with prefix containing no operators of precedence below minPrecedence */ - public ByteCodeBlock startExpr() throws IOException { return startExpr(-1); } - public void startExpr(ByteCodeBlock block) throws IOException { startExpr(-1, block); } - public ByteCodeBlock startExpr(int minPrecedence) throws IOException { - ByteCodeBlock ret = new ByteCodeBlock(line, sourceName); - startExpr(minPrecedence, ret); - return ret.size() == 0 ? null : ret; - } - - public void startExpr(int minPrecedence, ByteCodeBlock appendTo) throws IOException { + /** + * 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(CompiledFunction appendTo, int minPrecedence) throws IOException { int tok = getToken(); - int curLine = line; - if (tok == -1) return; - if (minPrecedence > 0 && precedence[tok] != 0) - if (precedence[tok] < minPrecedence || (precedence[tok] == minPrecedence && !isRightAssociative[tok])) - { pushBackToken(); return; } - - ByteCodeBlock b = appendTo; + CompiledFunction b = appendTo; switch (tok) { - case NUMBER: continueExpr(b.add(ByteCodeBlock.LITERAL, number), minPrecedence); return; - case STRING: continueExpr(b.add(ByteCodeBlock.LITERAL, string), minPrecedence); return; - case THIS: continueExpr(b.add(TOPSCOPE, null), minPrecedence); return; - case NULL: continueExpr(b.add(ByteCodeBlock.LITERAL, null), minPrecedence); return; - case TRUE: case FALSE: continueExpr(b.add(ByteCodeBlock.LITERAL, new Boolean(tok == TRUE)), minPrecedence); return; + case -1: throw new ParserException("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 LB: { - b.add(ARRAY, new Integer(0)); + b.add(line, ARRAY, new Integer(0)); // push an array onto the stack + int size0 = b.size(); int i = 0; - while(true) { - int size = b.size(); - startExpr(b); - if (size == b.size()) - if (peekToken() == RB) { consume(RB); continueExpr(b, minPrecedence); return; } - b.add(LITERAL, new Integer(i++)); - if (size == b.size()) b.add(LITERAL, null); - b.add(PUT); - b.add(POP); - if (peekToken() == RB) { consume(RB); continueExpr(b, minPrecedence); return; } - consume(COMMA); - } + if (peekToken() != RB) + while(true) { // iterate over the initialization values + int size = b.size(); + if (peekToken() == COMMA || peekToken() == RB) + b.add(line, 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 + 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: { + case SUB: { // negative literal (like "3 * -1") consume(NUMBER); - continueExpr(b.add(ByteCodeBlock.LITERAL, new Double(number.doubleValue() * -1)), minPrecedence); - return; + b.add(line, LITERAL, new Double(number.doubleValue() * -1)); + break; } - case LP: { - startExpr(b); + case LP: { // grouping (not calling) + startExpr(b, -1); consume(RP); - continueExpr(b, minPrecedence); - return; + break; } - case INC: case DEC: { - // prefix - startExpr(precedence[tok], b); - b.set(b.size() - 1, tok, new Boolean(true)); - continueExpr(b, minPrecedence); - return; + case INC: case DEC: { // prefix (not postfix) + startExpr(b, precedence[tok]); + b.set(b.size() - 1, tok, new Boolean(true)); // FIXME, ugly; need startAssignTarget + break; } - case BANG: case BITNOT: case INSTANCEOF: case TYPEOF: { - startExpr(precedence[tok], b); - b.add(tok); - continueExpr(b, minPrecedence); - return; + case BANG: case BITNOT: case TYPEOF: { + startExpr(b, precedence[tok]); + b.add(line, tok); + break; } - case LC: { - b.add(OBJECT, null); - if (peekToken() == RC) { consume(RC); continueExpr(b, minPrecedence); return; } - while(true) { - if (peekToken() != NAME && peekToken() != STRING) throw new Error("expected NAME or STRING"); - getToken(); - b.add(LITERAL, string); - consume(COLON); - startExpr(b); - b.add(PUT); - b.add(POP); - if (peekToken() == RC) { consume(RC); continueExpr(b, minPrecedence); return; } - consume(COMMA); - if (peekToken() == RC) { consume(RC); continueExpr(b, minPrecedence); return; } - } + case LC: { // object constructor + b.add(line, 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"); + getToken(); + b.add(line, 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 + if (peekToken() == RC) break; + consume(COMMA); + if (peekToken() == RC) break; // we permit {,,} -- I'm not sure if ECMA does + } + consume(RC); + break; } - case NAME: { + case NAME: { // FIXME; this is an lvalue String name = string; if (peekToken() == ASSIGN) { consume(ASSIGN); - b.add(TOPSCOPE); - b.add(ByteCodeBlock.LITERAL, name); - startExpr(minPrecedence, b); - b.add(ByteCodeBlock.PUT); - b.add(ByteCodeBlock.SWAP); - b.add(ByteCodeBlock.POP); + 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(TOPSCOPE); - b.add(ByteCodeBlock.LITERAL, name); - b.add(ByteCodeBlock.GET); + b.add(line, TOPSCOPE); + b.add(line, LITERAL, name); + b.add(line, GET); } - continueExpr(b, minPrecedence); - return; + break; } case FUNCTION: { consume(LP); int numArgs = 0; - ByteCodeBlock b2 = new ByteCodeBlock(curLine, sourceName); - b2.add(TOPSCOPE); - b2.add(SWAP); - b2.add(LITERAL, "arguments"); - b2.add(LITERAL, "arguments"); - b2.add(DECLARE); - b2.add(SWAP); - b2.add(PUT); - b2.add(SWAP); - b2.add(POP); - if (peekToken() == RP) consume(RP); - else while(true) { - if (peekToken() == COMMA) { - consume(COMMA); - - } else { - consume(NAME); - - // declare the name - b2.add(LITERAL, string); - b2.add(DECLARE); - - // retrieve it from the arguments array - b2.add(LITERAL, new Integer(numArgs)); - b2.add(GET_PRESERVE); - b2.add(SWAP); - b2.add(POP); - - // put it to the current scope - b2.add(TOPSCOPE); - b2.add(SWAP); - b2.add(LITERAL, string); - b2.add(SWAP); - b2.add(PUT); - - // clean the stack - b2.add(POP); - b2.add(POP); - - if (peekToken() == RP) { consume(RP); break; } - consume(COMMA); + CompiledFunction b2 = new CompiledFunction(sourceName, line, null); + b.add(line, 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); + + while(peekToken() != RP) { // run through the list of argument names + 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); + + 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); } + if (peekToken() == RP) break; + consume(COMMA); numArgs++; } - // pop off the arguments array - b2.add(POP); - parseStatement(true, b2); - b2.add(LITERAL, null); - b2.add(RETURN); - continueExpr(b.add(NEWFUNCTION, b2), minPrecedence); - return; + consume(RP); + + b2.add(line, POP); // pop off the arguments array + + parseStatement(b2, null); // the function body + + b2.add(line, LITERAL, null); // in case we "fall out the bottom", return NULL + b2.add(line, RETURN); + + break; } - default: pushBackToken(); return; + default: throw new ParserException("expected expression, found " + codeToString[tok] + ", which cannot start an expression"); } + + // attempt to continue the expression + continueExpr(b, minPrecedence); } - public void continueExpr(ByteCodeBlock prefix, int minPrecedence) throws IOException { + /** + * 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(CompiledFunction b, int minPrecedence) throws IOException { + if (b == null) throw new Error("got null b; this should never happen"); int tok = getToken(); - int curLine = line; if (tok == -1) return; - if (minPrecedence > 0 && precedence[tok] != 0) - if (precedence[tok] < minPrecedence || (precedence[tok] == minPrecedence && !isRightAssociative[tok])) - { pushBackToken(); return; } - - if (prefix == null) throw new Error("got null prefix"); - ByteCodeBlock b = prefix; + if (minPrecedence != -1 && (precedence[tok] < minPrecedence || (precedence[tok] == minPrecedence && !isRightAssociative[tok]))) { + pushBackToken(); + return; + } 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: { - prefix.set(prefix.size() - 1, b.GET_PRESERVE, new Boolean(true)); - startExpr(precedence[tok - 1], b); - prefix.add(tok - 1); - prefix.add(PUT); - prefix.add(SWAP); - prefix.add(POP); - continueExpr(b, minPrecedence); - return; + 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)); - continueExpr(b, minPrecedence); - return; + case INC: case DEC: { // postfix + b.set(b.size() - 1, tok, new Boolean(false)); // FIXME use assignmenttarget + break; } - - case LP: { - // invocation + case LP: { // invocation (not grouping) int i = 0; while(peekToken() != RP) { - startExpr(b); i++; - if (peekToken() == RP) break; + if (peekToken() != COMMA) { + startExpr(b, -1); + if (peekToken() == RP) break; + } consume(COMMA); } consume(RP); - b.add(CALL, new Integer(i)); - continueExpr(b, minPrecedence); - return; + b.add(line, CALL, new Integer(i)); + 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: case SUB: { - startExpr(precedence[tok], b); - b.add(tok); - continueExpr(b, minPrecedence); - return; + startExpr(b, precedence[tok]); + b.add(line, tok); + break; } - case OR: case AND: { - b.add(tok == AND ? b.JF : b.JT, new Integer(0)); + 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(precedence[tok], b); - b.arg[size - 1] = new Integer(b.size() - size + 2); - b.add(JMP, new Integer(2)); - b.add(LITERAL, tok == AND ? new Boolean(false) : new Boolean(true)); - continueExpr(b, minPrecedence); - return; + 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 + break; } - - case DOT: { + case DOT: { // FIXME, assigntarget consume(NAME); String target = string; if (peekToken() == ASSIGN) { consume(ASSIGN); - b.add(LITERAL, target); - startExpr(b); - b.add(PUT); - b.add(SWAP); - b.add(POP); + b.add(line, LITERAL, target); + startExpr(b, -1); + b.add(line, PUT); + b.add(line, SWAP); + b.add(line, POP); } else { - b.add(LITERAL, target); - b.add(GET); + b.add(line, LITERAL, target); + b.add(line, GET); } - continueExpr(b, minPrecedence); - return; + break; } - - case LB: { - startExpr(b); + case LB: { // subscripting (not array constructor) + startExpr(b, -1); consume(RB); - if (peekToken() == ASSIGN) { + if (peekToken() == ASSIGN) { // FIXME: assigntarget consume(ASSIGN); - startExpr(b); - b.add(PUT); - b.add(SWAP); - b.add(POP); + startExpr(b, -1); + b.add(line, PUT); + b.add(line, SWAP); + b.add(line, POP); } else { - b.add(GET); + b.add(line, GET); } - continueExpr(b, minPrecedence); - return; + break; } - case HOOK: { - b.add(JF, new Integer(0)); + b.add(line, JF, new Integer(0)); // jump to the if-false expression int size = b.size(); - startExpr(b); - b.arg[size - 1] = new Integer(b.size() - size + 2); - b.add(JMP, new Integer(0)); + 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 consume(COLON); size = b.size(); - startExpr(b); - b.arg[size - 1] = new Integer(b.size() - size + 1); - continueExpr(b, minPrecedence); + 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 + break; + } + default: { + pushBackToken(); return; } - - 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 ByteCodeBlock parseStatement() throws IOException { - ByteCodeBlock ret = new ByteCodeBlock(line, sourceName); - ret.add(ret.LITERAL, null); - parseStatement(false, ret); - if (ret.size() == 1) return null; - return ret; + /** 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 { + if (peekToken() == -1) return; + else if (peekToken() != LC) parseStatement(b, null); + else { + consume(LC); + while(peekToken() != RC && peekToken() != -1) parseStatement(b, null); + consume(RC); + } } - public void parseStatement(boolean requireBraces) throws IOException { - parseStatement(requireBraces, new ByteCodeBlock(line, sourceName)); - } - public void parseStatement(boolean requireBraces, ByteCodeBlock b) throws IOException { + /** Parse a single statement, consuming the RC or SEMI which terminates it. */ + void parseStatement(CompiledFunction b, String label) throws IOException { int tok = peekToken(); if (tok == -1) return; - boolean braced = tok == LC; - if (requireBraces && !braced) throw new ParserException("expected {, got " + codeToString[tok]); - if (braced) consume(LC); - int curLine = line; - while(true) { - switch(tok = peekToken()) { - - case THROW: case RETURN: case ASSERT: { - getToken(); - if (tok == RETURN && peekToken() == SEMI) b.add(LITERAL, null); - else startExpr(b); - consume(SEMI); - b.add(tok); - break; - } - - case BREAK: case CONTINUE: { - getToken(); - if (peekToken() == NAME) consume(NAME); - b.add(tok, string); - consume(SEMI); - break; - } - - case SEMI: - consume(SEMI); - if (!braced) return; - break; - - case VAR: { - consume(VAR); - b.add(TOPSCOPE); // push the current scope - while(true) { - consume(NAME); - String name = string; - b.add(LITERAL, name); // push the name to be declared - b.add(DECLARE); // declare it - if (peekToken() == ASSIGN) { // if there is an '=' after the variable name - b.add(LITERAL, name); // put the var name back on the stack - consume(ASSIGN); - startExpr(b); - b.add(PUT); - b.add(POP); - } - if (peekToken() != COMMA) break; - consume(COMMA); - } - b.add(POP); - if (peekToken() == SEMI) consume(SEMI); - break; - } - - case IF: { - consume(IF); - consume(LP); - startExpr(b); - consume(RP); - - b.add(JF, new Integer(0)); - int size = b.size(); - parseStatement(false, b); - - if (peekToken() == ELSE) { - consume(ELSE); - b.arg[size - 1] = new Integer(2 + b.size() - size); - b.add(JMP, new Integer(0)); - size = b.size(); - parseStatement(false, b); + switch(tok = getToken()) { + + case THROW: case ASSERT: case RETURN: { + if (tok == RETURN && peekToken() == SEMI) b.add(line, LITERAL, null); + else startExpr(b, -1); + b.add(line, tok); + consume(SEMI); + break; + } + + case BREAK: case CONTINUE: { + if (peekToken() == NAME) consume(NAME); + b.add(line, tok, string); + consume(SEMI); + break; + } + + case VAR: { + b.add(line, 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 + consume(ASSIGN); + startExpr(b, -1); + b.add(line, PUT); + b.add(line, POP); } - b.arg[size - 1] = new Integer(1 + b.size() - size); - break; + if (peekToken() != COMMA) break; + consume(COMMA); } - - case WHILE: { - consume(WHILE); - consume(LP); - ByteCodeBlock loop = new ByteCodeBlock(curLine, sourceName); - b.add(LOOP, loop); - - loop.add(POP); - startExpr(loop); - loop.add(JT, new Integer(2)); - loop.add(BREAK); - consume(RP); - parseStatement(false, loop); - - // if we fall out of the end, definately continue - loop.add(CONTINUE); - break; + b.add(line, POP); + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1) consume(SEMI); + break; + } + + case IF: { + consume(LP); + startExpr(b, -1); + consume(RP); + + b.add(line, JF, new Integer(0)); + int size = b.size(); + parseStatement(b, null); + + if (peekToken() == ELSE) { + consume(ELSE); + b.set(size - 1, new Integer(2 + b.size() - size)); + b.add(line, JMP, new Integer(0)); + size = b.size(); + parseStatement(b, null); } - - case SWITCH: { - consume(SWITCH); - consume(LP); - ByteCodeBlock loop = new ByteCodeBlock(curLine, sourceName); - b.add(LOOP, loop); - startExpr(loop); - consume(RP); - consume(LC); - while(true) - if (peekToken() == CASE) { - consume(CASE); - loop.add(DUP); - startExpr(loop); - consume(COLON); - loop.add(EQ); - loop.add(JF, new Integer(0)); - int size = loop.size(); - while(peekToken() != CASE && peekToken() != DEFAULT && peekToken() != RC) { - int size2 = loop.size(); - parseStatement(false, loop); - if (size2 == loop.size()) break; - } - loop.arg[size - 1] = new Integer(1 + loop.size() - size); - } else if (peekToken() == DEFAULT) { - consume(DEFAULT); - consume(COLON); - while(peekToken() != CASE && peekToken() != DEFAULT && peekToken() != RC) { - int size2 = loop.size(); - parseStatement(false, loop); - if (size2 == loop.size()) break; - } - } else if (peekToken() == RC) { - consume(RC); - loop.add(BREAK); - break; - } else { - throw new ParserException("expected CASE, DEFAULT, or RC; got " + codeToString[peekToken()]); + b.set(size - 1, new Integer(1 + b.size() - size)); + 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); + startExpr(b, -1); + b.add(line, JT, new Integer(2)); + b.add(line, 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 + break; + } + + case SWITCH: { + consume(LP); + if (label != null) b.add(line, LABEL, label); + b.add(line, LOOP); + int size0 = b.size(); + startExpr(b, -1); + consume(RP); + consume(LC); + while(true) + if (peekToken() == CASE) { + consume(CASE); + b.add(line, DUP); + startExpr(b, -1); + consume(COLON); + b.add(line, EQ); + b.add(line, JF, new Integer(0)); + int size = b.size(); + while(peekToken() != CASE && peekToken() != DEFAULT && peekToken() != RC) { + int size2 = b.size(); + parseStatement(b, null); + if (size2 == b.size()) break; } - break; - } - - case DO: { - consume(DO); - ByteCodeBlock loop = new ByteCodeBlock(curLine, sourceName); - b.add(LOOP, loop); - - parseStatement(false, loop); - consume(WHILE); + 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) { + int size2 = b.size(); + parseStatement(b, null); + if (size2 == b.size()) break; + } + } else if (peekToken() == RC) { + consume(RC); + b.add(line, BREAK); + break; + } else { + throw new ParserException("expected CASE, DEFAULT, or RC; got " + codeToString[peekToken()]); + } + b.add(line, BREAK); + 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(); + parseStatement(b, null); + consume(WHILE); + consume(LP); + startExpr(b, -1); + b.add(line, JT, new Integer(2)); + b.add(line, BREAK); + b.add(line, CONTINUE); + consume(RP); + consume(SEMI); + b.set(size - 1, new Integer(b.size() - size + 1)); // end of the loop + break; + } + + case TRY: { + // We deliberately allow you to omit braces in catch{}/finally{} if they are single statements... + b.add(line, TRY); + int size = b.size(); + parseBlock(b, null); + 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 + + if (peekToken() == CATCH) { + getToken(); consume(LP); - startExpr(loop); - loop.add(JT, new Integer(2)); - loop.add(BREAK); - loop.add(CONTINUE); + consume(NAME); consume(RP); - consume(SEMI); - break; + // FIXME, we need an extra scope here + b.add(line, TOPSCOPE); // the exception is on top of the stack; put it to the variable + b.add(line, SWAP); + b.add(line, LITERAL); + b.add(line, SWAP); + b.add(line, PUT); + b.add(line, POP); + b.add(line, POP); + parseStatement(b, null); } - - case TRY: { - // We deliberately allow you to omit braces in catch{}/finally{} if they are single statements... - consume(TRY); - parseStatement(true, b); - - if (peekToken() == CATCH) { - getToken(); - consume(LP); - consume(NAME); - consume(RP); - parseStatement(); // just discard the catchblock - } - - if (peekToken() == FINALLY) { - consume(FINALLY); - parseStatement(false, b); - } - break; + + b.set(size2 - 1, new Integer(b.size() - size2 + 1)); // jump here if no exception was thrown + + // FIXME: not implemented correctly + if (peekToken() == FINALLY) { + consume(FINALLY); + parseStatement(b, null); } - + break; + } + case FOR: { - consume(FOR); consume(LP); - + tok = getToken(); - if (tok == VAR) tok = getToken(); + boolean hadVar = false; + if (tok == VAR) { hadVar = true; tok = getToken(); } String varName = string; boolean forIn = peekToken() == IN; pushBackToken(tok, varName); - + if (forIn) { + // FIXME: break needs to work in here consume(NAME); consume(IN); - startExpr(b); - b.add(PUSHKEYS); - b.add(LITERAL, "length"); - b.add(GET); + startExpr(b, -1); + b.add(line, PUSHKEYS); + b.add(line, LITERAL, "length"); + b.add(line, GET); consume(RP); - ByteCodeBlock b2 = new ByteCodeBlock(curLine, sourceName); - b.add(SCOPE, b2); - b2.add(LITERAL, new Integer(1)); - b2.add(SUB); - b2.add(DUP); - b2.add(LITERAL, new Integer(0)); - b2.add(LT); - b2.add(JT, new Integer(7)); - b2.add(GET_PRESERVE); - b2.add(LITERAL, varName); - b2.add(LITERAL, varName); - b2.add(DECLARE); - b2.add(PUT); - parseStatement(false, b2); + CompiledFunction b2 = new CompiledFunction(sourceName, line, null); + + b.add(line, NEWSCOPE); + + b.add(line, LITERAL, new Integer(1)); + b.add(line, SUB); + b.add(line, DUP); + b.add(line, LITERAL, new Integer(0)); + b.add(line, LT); + b.add(line, JT, new Integer(7)); + b.add(line, GET_PRESERVE); + b.add(line, LITERAL, varName); + b.add(line, LITERAL, varName); + b.add(line, DECLARE); + b.add(line, PUT); + parseStatement(b, null); + + b.add(line, OLDSCOPE); + break; - + } else { - ByteCodeBlock b2 = new ByteCodeBlock(curLine, sourceName); - b.add(SCOPE, b2); - b.add(POP); - - int size = b2.size(); - startExpr(b2); - if (b2.size() - size > 0) b2.add(POP); - consume(SEMI); - ByteCodeBlock e2 = startExpr(); + if (hadVar) pushBackToken(VAR, null); + b.add(line, NEWSCOPE); + + parseStatement(b, null); + CompiledFunction e2 = new CompiledFunction(sourceName, line, null); + if (peekToken() != SEMI) { + startExpr(e2, -1); + } else { + e2.add(line, b.LITERAL, Boolean.TRUE); + } consume(SEMI); - if (e2 == null) e2 = new ByteCodeBlock(curLine, sourceName, b.LITERAL, null); - - ByteCodeBlock b3 = new ByteCodeBlock(curLine, sourceName); - b2.add(LOOP, b3); - b2.add(LITERAL, null); - - b3.add(JT, new Integer(0)); - size = b3.size(); - startExpr(b3); + if (label != null) b.add(line, LABEL, label); + b.add(line, LOOP); + int size2 = b.size(); + + b.add(line, JT, new Integer(0)); + int size = b.size(); + if (peekToken() != RP) { + startExpr(b, -1); + b.add(line, POP); + } + b.set(size - 1, new Integer(b.size() - size + 1)); consume(RP); - if (b3.size() - size > 0) b3.add(POP); - b3.arg[size - 1] = new Integer(b3.size() - size + 1); - - b3.paste(e2); - b3.add(JT, new Integer(2)); - b3.add(BREAK); - parseStatement(false, b3); - b3.add(BREAK); + + b.paste(e2); + b.add(line, JT, new Integer(2)); + b.add(line, BREAK); + parseStatement(b, null); + b.add(line, CONTINUE); + b.set(size2 - 1, new Integer(b.size() - size2 + 1)); // end of the loop + + b.add(line, OLDSCOPE); break; } } - - case NAME: { - consume(NAME); - String name = string; - if (peekToken() == COLON) { - consume(COLON); - b.add(ByteCodeBlock.LABEL, string); - break; - } else { - pushBackToken(NAME, name); - // fall through to default case - } - } - // fall through - case RC: - if (tok == RC && braced) { consume(RC); return; } - // fall through - default: { - int size = b.size(); - startExpr(b); - if (size == b.size()) return; - b.add(POP); - if (peekToken() == SEMI) consume(SEMI); + + case NAME: { + String possiblyTheLabel = string; + if (peekToken() == COLON) { + consume(COLON); + label = possiblyTheLabel; + parseStatement(b, label); + break; + } else { + pushBackToken(NAME, possiblyTheLabel); + startExpr(b, -1); + b.add(line, POP); + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1) consume(SEMI); break; } - } - - if (!braced) return; + } + + case SEMI: return; + case LC: { + pushBackToken(); + parseBlock(b, label); + break; + } + + default: { + pushBackToken(); + startExpr(b, -1); + b.add(line, POP); + if ((mostRecentlyReadToken != RC || peekToken() == SEMI) && peekToken() != -1) consume(SEMI); + break; + } } } + + + // ParserException ////////////////////////////////////////////////////////////////////// - class ParserException extends RuntimeException { - public ParserException(String s) { super(sourceName + ":" + line + " " + s); } - } + private class ParserException extends IOException { public ParserException(String s) { super(sourceName + ":" + line + " " + s); } } }