CaRMtl/org/mozilla/javascript/Parser.java
2018-09-04 22:51:42 -04:00

2508 lines
83 KiB
Java

/* -*- Mode: java; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
*
* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is Rhino code, released
* May 6, 1999.
*
* The Initial Developer of the Original Code is
* Netscape Communications Corporation.
* Portions created by the Initial Developer are Copyright (C) 1997-1999
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
* Mike Ang
* Igor Bukanov
* Yuh-Ruey Chen
* Ethan Hugg
* Bob Jervis
* Terry Lucas
* Mike McCabe
* Milen Nankov
* Norris Boyd
*
* Alternatively, the contents of this file may be used under the terms of
* the GNU General Public License Version 2 or later (the "GPL"), in which
* case the provisions of the GPL are applicable instead of those above. If
* you wish to allow use of your version of this file only under the terms of
* the GPL and not to allow others to use your version of this file under the
* MPL, indicate your decision by deleting the provisions above and replacing
* them with the notice and other provisions required by the GPL. If you do
* not delete the provisions above, a recipient may use your version of this
* file under either the MPL or the GPL.
*
* ***** END LICENSE BLOCK ***** */
package org.mozilla.javascript;
import java.io.Reader;
import java.io.IOException;
import java.util.HashMap;
import java.util.Map;
/**
* This class implements the JavaScript parser.
*
* It is based on the C source files jsparse.c and jsparse.h
* in the jsref package.
*
* @see TokenStream
*
* @author Mike McCabe
* @author Brendan Eich
*/
public class Parser
{
// TokenInformation flags : currentFlaggedToken stores them together
// with token type
final static int
CLEAR_TI_MASK = 0xFFFF, // mask to clear token information bits
TI_AFTER_EOL = 1 << 16, // first token of the source line
TI_CHECK_LABEL = 1 << 17; // indicates to check for label
CompilerEnvirons compilerEnv;
private ErrorReporter errorReporter;
private String sourceURI;
boolean calledByCompileFunction;
private TokenStream ts;
private int currentFlaggedToken;
private int syntaxErrorCount;
private IRFactory nf;
private int nestingOfFunction;
private Decompiler decompiler;
private String encodedSource;
// The following are per function variables and should be saved/restored
// during function parsing.
// XXX Move to separated class?
ScriptOrFnNode currentScriptOrFn;
Node.Scope currentScope;
private int nestingOfWith;
private Map<String,Node> labelSet; // map of label names into nodes
private ObjArray loopSet;
private ObjArray loopAndSwitchSet;
private int endFlags;
// end of per function variables
public int getCurrentLineNumber() {
return ts.getLineno();
}
// Exception to unwind
private static class ParserException extends RuntimeException
{
static final long serialVersionUID = 5882582646773765630L;
}
public Parser(CompilerEnvirons compilerEnv, ErrorReporter errorReporter)
{
this.compilerEnv = compilerEnv;
this.errorReporter = errorReporter;
}
protected Decompiler createDecompiler(CompilerEnvirons compilerEnv)
{
return new Decompiler();
}
void addStrictWarning(String messageId, String messageArg)
{
if (compilerEnv.isStrictMode())
addWarning(messageId, messageArg);
}
void addWarning(String messageId, String messageArg)
{
String message = ScriptRuntime.getMessage1(messageId, messageArg);
if (compilerEnv.reportWarningAsError()) {
++syntaxErrorCount;
errorReporter.error(message, sourceURI, ts.getLineno(),
ts.getLine(), ts.getOffset());
} else
errorReporter.warning(message, sourceURI, ts.getLineno(),
ts.getLine(), ts.getOffset());
}
void addError(String messageId)
{
++syntaxErrorCount;
String message = ScriptRuntime.getMessage0(messageId);
errorReporter.error(message, sourceURI, ts.getLineno(),
ts.getLine(), ts.getOffset());
}
void addError(String messageId, String messageArg)
{
++syntaxErrorCount;
String message = ScriptRuntime.getMessage1(messageId, messageArg);
errorReporter.error(message, sourceURI, ts.getLineno(),
ts.getLine(), ts.getOffset());
}
RuntimeException reportError(String messageId)
{
addError(messageId);
// Throw a ParserException exception to unwind the recursive descent
// parse.
throw new ParserException();
}
private int peekToken()
throws IOException
{
int tt = currentFlaggedToken;
if (tt == Token.EOF) {
tt = ts.getToken();
if (tt == Token.EOL) {
do {
tt = ts.getToken();
} while (tt == Token.EOL);
tt |= TI_AFTER_EOL;
}
currentFlaggedToken = tt;
}
return tt & CLEAR_TI_MASK;
}
private int peekFlaggedToken()
throws IOException
{
peekToken();
return currentFlaggedToken;
}
private void consumeToken()
{
currentFlaggedToken = Token.EOF;
}
private int nextToken()
throws IOException
{
int tt = peekToken();
consumeToken();
return tt;
}
private int nextFlaggedToken()
throws IOException
{
peekToken();
int ttFlagged = currentFlaggedToken;
consumeToken();
return ttFlagged;
}
private boolean matchToken(int toMatch)
throws IOException
{
int tt = peekToken();
if (tt != toMatch) {
return false;
}
consumeToken();
return true;
}
private int peekTokenOrEOL()
throws IOException
{
int tt = peekToken();
// Check for last peeked token flags
if ((currentFlaggedToken & TI_AFTER_EOL) != 0) {
tt = Token.EOL;
}
return tt;
}
private void setCheckForLabel()
{
if ((currentFlaggedToken & CLEAR_TI_MASK) != Token.NAME)
throw Kit.codeBug();
currentFlaggedToken |= TI_CHECK_LABEL;
}
private void mustMatchToken(int toMatch, String messageId)
throws IOException, ParserException
{
if (!matchToken(toMatch)) {
reportError(messageId);
}
}
private void mustHaveXML()
{
if (!compilerEnv.isXmlAvailable()) {
reportError("msg.XML.not.available");
}
}
public String getEncodedSource()
{
return encodedSource;
}
public boolean eof()
{
return ts.eof();
}
boolean insideFunction()
{
return nestingOfFunction != 0;
}
void pushScope(Node node) {
Node.Scope scopeNode = (Node.Scope) node;
if (scopeNode.getParentScope() != null) throw Kit.codeBug();
scopeNode.setParent(currentScope);
currentScope = scopeNode;
}
void popScope() {
currentScope = currentScope.getParentScope();
}
private Node enterLoop(Node loopLabel, boolean doPushScope)
{
Node loop = nf.createLoopNode(loopLabel, ts.getLineno());
if (loopSet == null) {
loopSet = new ObjArray();
if (loopAndSwitchSet == null) {
loopAndSwitchSet = new ObjArray();
}
}
loopSet.push(loop);
loopAndSwitchSet.push(loop);
if (doPushScope) {
pushScope(loop);
}
return loop;
}
private void exitLoop(boolean doPopScope)
{
loopSet.pop();
loopAndSwitchSet.pop();
if (doPopScope) {
popScope();
}
}
private Node enterSwitch(Node switchSelector, int lineno)
{
Node switchNode = nf.createSwitch(switchSelector, lineno);
if (loopAndSwitchSet == null) {
loopAndSwitchSet = new ObjArray();
}
loopAndSwitchSet.push(switchNode);
return switchNode;
}
private void exitSwitch()
{
loopAndSwitchSet.pop();
}
/*
* Build a parse tree from the given sourceString.
*
* @return an Object representing the parsed
* program. If the parse fails, null will be returned. (The
* parse failure will result in a call to the ErrorReporter from
* CompilerEnvirons.)
*/
public ScriptOrFnNode parse(String sourceString,
String sourceURI, int lineno)
{
this.sourceURI = sourceURI;
this.ts = new TokenStream(this, null, sourceString, lineno);
try {
return parse();
} catch (IOException ex) {
// Should never happen
throw new IllegalStateException();
}
}
/*
* Build a parse tree from the given sourceString.
*
* @return an Object representing the parsed
* program. If the parse fails, null will be returned. (The
* parse failure will result in a call to the ErrorReporter from
* CompilerEnvirons.)
*/
public ScriptOrFnNode parse(Reader sourceReader,
String sourceURI, int lineno)
throws IOException
{
this.sourceURI = sourceURI;
this.ts = new TokenStream(this, sourceReader, null, lineno);
return parse();
}
private ScriptOrFnNode parse()
throws IOException
{
this.decompiler = createDecompiler(compilerEnv);
this.nf = new IRFactory(this);
currentScriptOrFn = nf.createScript();
currentScope = currentScriptOrFn;
int sourceStartOffset = decompiler.getCurrentOffset();
this.encodedSource = null;
decompiler.addToken(Token.SCRIPT);
this.currentFlaggedToken = Token.EOF;
this.syntaxErrorCount = 0;
int baseLineno = ts.getLineno(); // line number where source starts
/* so we have something to add nodes to until
* we've collected all the source */
Node pn = nf.createLeaf(Token.BLOCK);
try {
for (;;) {
int tt = peekToken();
if (tt <= Token.EOF) {
break;
}
Node n;
if (tt == Token.FUNCTION) {
consumeToken();
try {
n = function(calledByCompileFunction
? FunctionNode.FUNCTION_EXPRESSION
: FunctionNode.FUNCTION_STATEMENT);
} catch (ParserException e) {
break;
}
} else {
n = statement();
}
nf.addChildToBack(pn, n);
}
} catch (StackOverflowError ex) {
String msg = ScriptRuntime.getMessage0(
"msg.too.deep.parser.recursion");
throw Context.reportRuntimeError(msg, sourceURI,
ts.getLineno(), null, 0);
}
if (this.syntaxErrorCount != 0) {
String msg = String.valueOf(this.syntaxErrorCount);
msg = ScriptRuntime.getMessage1("msg.got.syntax.errors", msg);
throw errorReporter.runtimeError(msg, sourceURI, baseLineno,
null, 0);
}
currentScriptOrFn.setSourceName(sourceURI);
currentScriptOrFn.setBaseLineno(baseLineno);
currentScriptOrFn.setEndLineno(ts.getLineno());
int sourceEndOffset = decompiler.getCurrentOffset();
currentScriptOrFn.setEncodedSourceBounds(sourceStartOffset,
sourceEndOffset);
nf.initScript(currentScriptOrFn, pn);
if (compilerEnv.isGeneratingSource()) {
encodedSource = decompiler.getEncodedSource();
}
this.decompiler = null; // It helps GC
return currentScriptOrFn;
}
/*
* The C version of this function takes an argument list,
* which doesn't seem to be needed for tree generation...
* it'd only be useful for checking argument hiding, which
* I'm not doing anyway...
*/
private Node parseFunctionBody()
throws IOException
{
++nestingOfFunction;
Node pn = nf.createBlock(ts.getLineno());
try {
bodyLoop: for (;;) {
Node n;
int tt = peekToken();
switch (tt) {
case Token.ERROR:
case Token.EOF:
case Token.RC:
break bodyLoop;
case Token.FUNCTION:
consumeToken();
n = function(FunctionNode.FUNCTION_STATEMENT);
break;
default:
n = statement();
break;
}
nf.addChildToBack(pn, n);
}
} catch (ParserException e) {
// Ignore it
} finally {
--nestingOfFunction;
}
return pn;
}
private Node function(int functionType)
throws IOException, ParserException
{
int syntheticType = functionType;
int baseLineno = ts.getLineno(); // line number where source starts
int functionSourceStart = decompiler.markFunctionStart(functionType);
String name;
Node memberExprNode = null;
if (matchToken(Token.NAME)) {
name = ts.getString();
decompiler.addName(name);
if (!matchToken(Token.LP)) {
if (compilerEnv.isAllowMemberExprAsFunctionName()) {
// Extension to ECMA: if 'function <name>' does not follow
// by '(', assume <name> starts memberExpr
Node memberExprHead = nf.createName(name);
name = "";
memberExprNode = memberExprTail(false, memberExprHead);
}
mustMatchToken(Token.LP, "msg.no.paren.parms");
}
} else if (matchToken(Token.LP)) {
// Anonymous function
name = "";
} else {
name = "";
if (compilerEnv.isAllowMemberExprAsFunctionName()) {
// Note that memberExpr can not start with '(' like
// in function (1+2).toString(), because 'function (' already
// processed as anonymous function
memberExprNode = memberExpr(false);
}
mustMatchToken(Token.LP, "msg.no.paren.parms");
}
if (memberExprNode != null) {
syntheticType = FunctionNode.FUNCTION_EXPRESSION;
}
if (syntheticType != FunctionNode.FUNCTION_EXPRESSION &&
name.length() > 0)
{
// Function statements define a symbol in the enclosing scope
defineSymbol(Token.FUNCTION, false, name);
}
boolean nested = insideFunction();
FunctionNode fnNode = nf.createFunction(name);
if (nested || nestingOfWith > 0) {
// 1. Nested functions are not affected by the dynamic scope flag
// as dynamic scope is already a parent of their scope.
// 2. Functions defined under the with statement also immune to
// this setup, in which case dynamic scope is ignored in favor
// of with object.
fnNode.itsIgnoreDynamicScope = true;
}
int functionIndex = currentScriptOrFn.addFunction(fnNode);
int functionSourceEnd;
ScriptOrFnNode savedScriptOrFn = currentScriptOrFn;
currentScriptOrFn = fnNode;
Node.Scope savedCurrentScope = currentScope;
currentScope = fnNode;
int savedNestingOfWith = nestingOfWith;
nestingOfWith = 0;
Map<String,Node> savedLabelSet = labelSet;
labelSet = null;
ObjArray savedLoopSet = loopSet;
loopSet = null;
ObjArray savedLoopAndSwitchSet = loopAndSwitchSet;
loopAndSwitchSet = null;
int savedFunctionEndFlags = endFlags;
endFlags = 0;
Node destructuring = null;
Node body;
try {
decompiler.addToken(Token.LP);
if (!matchToken(Token.RP)) {
boolean first = true;
do {
if (!first)
decompiler.addToken(Token.COMMA);
first = false;
int tt = peekToken();
if (tt == Token.LB || tt == Token.LC) {
// Destructuring assignment for parameters: add a
// dummy parameter name, and add a statement to the
// body to initialize variables from the destructuring
// assignment
if (destructuring == null) {
destructuring = new Node(Token.COMMA);
}
String parmName = currentScriptOrFn.getNextTempName();
defineSymbol(Token.LP, false, parmName);
destructuring.addChildToBack(
nf.createDestructuringAssignment(Token.VAR,
primaryExpr(), nf.createName(parmName)));
} else {
mustMatchToken(Token.NAME, "msg.no.parm");
String s = ts.getString();
defineSymbol(Token.LP, false, s);
decompiler.addName(s);
}
} while (matchToken(Token.COMMA));
mustMatchToken(Token.RP, "msg.no.paren.after.parms");
}
decompiler.addToken(Token.RP);
mustMatchToken(Token.LC, "msg.no.brace.body");
decompiler.addEOL(Token.LC);
body = parseFunctionBody();
if (destructuring != null) {
body.addChildToFront(
new Node(Token.EXPR_VOID, destructuring, ts.getLineno()));
}
mustMatchToken(Token.RC, "msg.no.brace.after.body");
if (compilerEnv.isStrictMode() && !body.hasConsistentReturnUsage())
{
String msg = name.length() > 0 ? "msg.no.return.value"
: "msg.anon.no.return.value";
addStrictWarning(msg, name);
}
if (syntheticType == FunctionNode.FUNCTION_EXPRESSION &&
name.length() > 0 && currentScope.getSymbol(name) == null)
{
// Function expressions define a name only in the body of the
// function, and only if not hidden by a parameter name
defineSymbol(Token.FUNCTION, false, name);
}
decompiler.addToken(Token.RC);
functionSourceEnd = decompiler.markFunctionEnd(functionSourceStart);
if (functionType != FunctionNode.FUNCTION_EXPRESSION) {
// Add EOL only if function is not part of expression
// since it gets SEMI + EOL from Statement in that case
decompiler.addToken(Token.EOL);
}
}
finally {
endFlags = savedFunctionEndFlags;
loopAndSwitchSet = savedLoopAndSwitchSet;
loopSet = savedLoopSet;
labelSet = savedLabelSet;
nestingOfWith = savedNestingOfWith;
currentScriptOrFn = savedScriptOrFn;
currentScope = savedCurrentScope;
}
fnNode.setEncodedSourceBounds(functionSourceStart, functionSourceEnd);
fnNode.setSourceName(sourceURI);
fnNode.setBaseLineno(baseLineno);
fnNode.setEndLineno(ts.getLineno());
Node pn = nf.initFunction(fnNode, functionIndex, body, syntheticType);
if (memberExprNode != null) {
pn = nf.createAssignment(Token.ASSIGN, memberExprNode, pn);
if (functionType != FunctionNode.FUNCTION_EXPRESSION) {
// XXX check JScript behavior: should it be createExprStatement?
pn = nf.createExprStatementNoReturn(pn, baseLineno);
}
}
return pn;
}
private Node statements(Node scope)
throws IOException
{
Node pn = scope != null ? scope : nf.createBlock(ts.getLineno());
int tt;
while ((tt = peekToken()) > Token.EOF && tt != Token.RC) {
nf.addChildToBack(pn, statement());
}
return pn;
}
private Node condition()
throws IOException, ParserException
{
mustMatchToken(Token.LP, "msg.no.paren.cond");
decompiler.addToken(Token.LP);
Node pn = expr(false);
mustMatchToken(Token.RP, "msg.no.paren.after.cond");
decompiler.addToken(Token.RP);
// Report strict warning on code like "if (a = 7) ...". Suppress the
// warning if the condition is parenthesized, like "if ((a = 7)) ...".
if (pn.getProp(Node.PARENTHESIZED_PROP) == null &&
(pn.getType() == Token.SETNAME || pn.getType() == Token.SETPROP ||
pn.getType() == Token.SETELEM))
{
addStrictWarning("msg.equal.as.assign", "");
}
return pn;
}
// match a NAME; return null if no match.
private Node matchJumpLabelName()
throws IOException, ParserException
{
Node label = null;
int tt = peekTokenOrEOL();
if (tt == Token.NAME) {
consumeToken();
String name = ts.getString();
decompiler.addName(name);
if (labelSet != null) {
label = labelSet.get(name);
}
if (label == null) {
reportError("msg.undef.label");
}
}
return label;
}
private Node statement()
throws IOException
{
try {
Node pn = statementHelper(null);
if (pn != null) {
if (compilerEnv.isStrictMode() && !pn.hasSideEffects())
addStrictWarning("msg.no.side.effects", "");
return pn;
}
} catch (ParserException e) { }
// skip to end of statement
int lineno = ts.getLineno();
guessingStatementEnd: for (;;) {
int tt = peekTokenOrEOL();
consumeToken();
switch (tt) {
case Token.ERROR:
case Token.EOF:
case Token.EOL:
case Token.SEMI:
break guessingStatementEnd;
}
}
return nf.createExprStatement(nf.createName("error"), lineno);
}
private Node statementHelper(Node statementLabel)
throws IOException, ParserException
{
Node pn = null;
int tt = peekToken();
switch (tt) {
case Token.IF: {
consumeToken();
decompiler.addToken(Token.IF);
int lineno = ts.getLineno();
Node cond = condition();
decompiler.addEOL(Token.LC);
Node ifTrue = statement();
Node ifFalse = null;
if (matchToken(Token.ELSE)) {
decompiler.addToken(Token.RC);
decompiler.addToken(Token.ELSE);
decompiler.addEOL(Token.LC);
ifFalse = statement();
}
decompiler.addEOL(Token.RC);
pn = nf.createIf(cond, ifTrue, ifFalse, lineno);
return pn;
}
case Token.SWITCH: {
consumeToken();
decompiler.addToken(Token.SWITCH);
int lineno = ts.getLineno();
mustMatchToken(Token.LP, "msg.no.paren.switch");
decompiler.addToken(Token.LP);
pn = enterSwitch(expr(false), lineno);
try {
mustMatchToken(Token.RP, "msg.no.paren.after.switch");
decompiler.addToken(Token.RP);
mustMatchToken(Token.LC, "msg.no.brace.switch");
decompiler.addEOL(Token.LC);
boolean hasDefault = false;
switchLoop: for (;;) {
tt = nextToken();
Node caseExpression;
switch (tt) {
case Token.RC:
break switchLoop;
case Token.CASE:
decompiler.addToken(Token.CASE);
caseExpression = expr(false);
mustMatchToken(Token.COLON, "msg.no.colon.case");
decompiler.addEOL(Token.COLON);
break;
case Token.DEFAULT:
if (hasDefault) {
reportError("msg.double.switch.default");
}
decompiler.addToken(Token.DEFAULT);
hasDefault = true;
caseExpression = null;
mustMatchToken(Token.COLON, "msg.no.colon.case");
decompiler.addEOL(Token.COLON);
break;
default:
reportError("msg.bad.switch");
break switchLoop;
}
Node block = nf.createLeaf(Token.BLOCK);
while ((tt = peekToken()) != Token.RC
&& tt != Token.CASE
&& tt != Token.DEFAULT
&& tt != Token.EOF)
{
nf.addChildToBack(block, statement());
}
// caseExpression == null => add default label
nf.addSwitchCase(pn, caseExpression, block);
}
decompiler.addEOL(Token.RC);
nf.closeSwitch(pn);
} finally {
exitSwitch();
}
return pn;
}
case Token.WHILE: {
consumeToken();
decompiler.addToken(Token.WHILE);
Node loop = enterLoop(statementLabel, true);
try {
Node cond = condition();
decompiler.addEOL(Token.LC);
Node body = statement();
decompiler.addEOL(Token.RC);
pn = nf.createWhile(loop, cond, body);
} finally {
exitLoop(true);
}
return pn;
}
case Token.DO: {
consumeToken();
decompiler.addToken(Token.DO);
decompiler.addEOL(Token.LC);
Node loop = enterLoop(statementLabel, true);
try {
Node body = statement();
decompiler.addToken(Token.RC);
mustMatchToken(Token.WHILE, "msg.no.while.do");
decompiler.addToken(Token.WHILE);
Node cond = condition();
pn = nf.createDoWhile(loop, body, cond);
} finally {
exitLoop(true);
}
// Always auto-insert semicolon to follow SpiderMonkey:
// It is required by ECMAScript but is ignored by the rest of
// world, see bug 238945
matchToken(Token.SEMI);
decompiler.addEOL(Token.SEMI);
return pn;
}
case Token.FOR: {
consumeToken();
boolean isForEach = false;
decompiler.addToken(Token.FOR);
Node loop = enterLoop(statementLabel, true);
try {
Node init; // Node init is also foo in 'foo in object'
Node cond; // Node cond is also object in 'foo in object'
Node incr = null;
Node body;
int declType = -1;
// See if this is a for each () instead of just a for ()
if (matchToken(Token.NAME)) {
decompiler.addName(ts.getString());
if (ts.getString().equals("each")) {
isForEach = true;
} else {
reportError("msg.no.paren.for");
}
}
mustMatchToken(Token.LP, "msg.no.paren.for");
decompiler.addToken(Token.LP);
tt = peekToken();
if (tt == Token.SEMI) {
init = nf.createLeaf(Token.EMPTY);
} else {
if (tt == Token.VAR || tt == Token.LET) {
// set init to a var list or initial
consumeToken(); // consume the token
decompiler.addToken(tt);
init = variables(true, tt);
declType = tt;
}
else {
init = expr(true);
}
}
if (matchToken(Token.IN)) {
decompiler.addToken(Token.IN);
// 'cond' is the object over which we're iterating
cond = expr(false);
} else { // ordinary for loop
mustMatchToken(Token.SEMI, "msg.no.semi.for");
decompiler.addToken(Token.SEMI);
if (peekToken() == Token.SEMI) {
// no loop condition
cond = nf.createLeaf(Token.EMPTY);
} else {
cond = expr(false);
}
mustMatchToken(Token.SEMI, "msg.no.semi.for.cond");
decompiler.addToken(Token.SEMI);
if (peekToken() == Token.RP) {
incr = nf.createLeaf(Token.EMPTY);
} else {
incr = expr(false);
}
}
mustMatchToken(Token.RP, "msg.no.paren.for.ctrl");
decompiler.addToken(Token.RP);
decompiler.addEOL(Token.LC);
body = statement();
decompiler.addEOL(Token.RC);
if (incr == null) {
// cond could be null if 'in obj' got eaten
// by the init node.
pn = nf.createForIn(declType, loop, init, cond, body,
isForEach);
} else {
pn = nf.createFor(loop, init, cond, incr, body);
}
} finally {
exitLoop(true);
}
return pn;
}
case Token.TRY: {
consumeToken();
int lineno = ts.getLineno();
Node tryblock;
Node catchblocks = null;
Node finallyblock = null;
decompiler.addToken(Token.TRY);
if (peekToken() != Token.LC) {
reportError("msg.no.brace.try");
}
decompiler.addEOL(Token.LC);
tryblock = statement();
decompiler.addEOL(Token.RC);
catchblocks = nf.createLeaf(Token.BLOCK);
boolean sawDefaultCatch = false;
int peek = peekToken();
if (peek == Token.CATCH) {
while (matchToken(Token.CATCH)) {
if (sawDefaultCatch) {
reportError("msg.catch.unreachable");
}
decompiler.addToken(Token.CATCH);
mustMatchToken(Token.LP, "msg.no.paren.catch");
decompiler.addToken(Token.LP);
mustMatchToken(Token.NAME, "msg.bad.catchcond");
String varName = ts.getString();
decompiler.addName(varName);
Node catchCond = null;
if (matchToken(Token.IF)) {
decompiler.addToken(Token.IF);
catchCond = expr(false);
} else {
sawDefaultCatch = true;
}
mustMatchToken(Token.RP, "msg.bad.catchcond");
decompiler.addToken(Token.RP);
mustMatchToken(Token.LC, "msg.no.brace.catchblock");
decompiler.addEOL(Token.LC);
nf.addChildToBack(catchblocks,
nf.createCatch(varName, catchCond,
statements(null),
ts.getLineno()));
mustMatchToken(Token.RC, "msg.no.brace.after.body");
decompiler.addEOL(Token.RC);
}
} else if (peek != Token.FINALLY) {
mustMatchToken(Token.FINALLY, "msg.try.no.catchfinally");
}
if (matchToken(Token.FINALLY)) {
decompiler.addToken(Token.FINALLY);
decompiler.addEOL(Token.LC);
finallyblock = statement();
decompiler.addEOL(Token.RC);
}
pn = nf.createTryCatchFinally(tryblock, catchblocks,
finallyblock, lineno);
return pn;
}
case Token.THROW: {
consumeToken();
if (peekTokenOrEOL() == Token.EOL) {
// ECMAScript does not allow new lines before throw expression,
// see bug 256617
reportError("msg.bad.throw.eol");
}
int lineno = ts.getLineno();
decompiler.addToken(Token.THROW);
pn = nf.createThrow(expr(false), lineno);
break;
}
case Token.BREAK: {
consumeToken();
int lineno = ts.getLineno();
decompiler.addToken(Token.BREAK);
// matchJumpLabelName only matches if there is one
Node breakStatement = matchJumpLabelName();
if (breakStatement == null) {
if (loopAndSwitchSet == null || loopAndSwitchSet.size() == 0) {
reportError("msg.bad.break");
return null;
}
breakStatement = (Node)loopAndSwitchSet.peek();
}
pn = nf.createBreak(breakStatement, lineno);
break;
}
case Token.CONTINUE: {
consumeToken();
int lineno = ts.getLineno();
decompiler.addToken(Token.CONTINUE);
Node loop;
// matchJumpLabelName only matches if there is one
Node label = matchJumpLabelName();
if (label == null) {
if (loopSet == null || loopSet.size() == 0) {
reportError("msg.continue.outside");
return null;
}
loop = (Node)loopSet.peek();
} else {
loop = nf.getLabelLoop(label);
if (loop == null) {
reportError("msg.continue.nonloop");
return null;
}
}
pn = nf.createContinue(loop, lineno);
break;
}
case Token.WITH: {
consumeToken();
decompiler.addToken(Token.WITH);
int lineno = ts.getLineno();
mustMatchToken(Token.LP, "msg.no.paren.with");
decompiler.addToken(Token.LP);
Node obj = expr(false);
mustMatchToken(Token.RP, "msg.no.paren.after.with");
decompiler.addToken(Token.RP);
decompiler.addEOL(Token.LC);
++nestingOfWith;
Node body;
try {
body = statement();
} finally {
--nestingOfWith;
}
decompiler.addEOL(Token.RC);
pn = nf.createWith(obj, body, lineno);
return pn;
}
case Token.CONST:
case Token.VAR: {
consumeToken();
decompiler.addToken(tt);
pn = variables(false, tt);
break;
}
case Token.LET: {
consumeToken();
decompiler.addToken(Token.LET);
if (peekToken() == Token.LP) {
return let(true);
} else {
pn = variables(false, tt);
if (peekToken() == Token.SEMI)
break;
return pn;
}
}
case Token.RETURN:
case Token.YIELD: {
pn = returnOrYield(tt, false);
break;
}
case Token.DEBUGGER:
consumeToken();
decompiler.addToken(Token.DEBUGGER);
pn = nf.createDebugger(ts.getLineno());
break;
case Token.LC:
consumeToken();
if (statementLabel != null) {
decompiler.addToken(Token.LC);
}
Node scope = nf.createScopeNode(Token.BLOCK, ts.getLineno());
pushScope(scope);
try {
statements(scope);
mustMatchToken(Token.RC, "msg.no.brace.block");
if (statementLabel != null) {
decompiler.addEOL(Token.RC);
}
return scope;
} finally {
popScope();
}
case Token.ERROR:
// Fall thru, to have a node for error recovery to work on
case Token.SEMI:
consumeToken();
pn = nf.createLeaf(Token.EMPTY);
return pn;
case Token.FUNCTION: {
consumeToken();
pn = function(FunctionNode.FUNCTION_EXPRESSION_STATEMENT);
return pn;
}
case Token.DEFAULT :
consumeToken();
mustHaveXML();
decompiler.addToken(Token.DEFAULT);
int nsLine = ts.getLineno();
if (!(matchToken(Token.NAME)
&& ts.getString().equals("xml")))
{
reportError("msg.bad.namespace");
}
decompiler.addName(" xml");
if (!(matchToken(Token.NAME)
&& ts.getString().equals("namespace")))
{
reportError("msg.bad.namespace");
}
decompiler.addName(" namespace");
if (!matchToken(Token.ASSIGN)) {
reportError("msg.bad.namespace");
}
decompiler.addToken(Token.ASSIGN);
Node expr = expr(false);
pn = nf.createDefaultNamespace(expr, nsLine);
break;
case Token.NAME: {
int lineno = ts.getLineno();
String name = ts.getString();
setCheckForLabel();
pn = expr(false);
if (pn.getType() != Token.LABEL) {
pn = nf.createExprStatement(pn, lineno);
} else {
// Parsed the label: push back token should be
// colon that primaryExpr left untouched.
if (peekToken() != Token.COLON) Kit.codeBug();
consumeToken();
// depend on decompiling lookahead to guess that that
// last name was a label.
decompiler.addName(name);
decompiler.addEOL(Token.COLON);
if (labelSet == null) {
labelSet = new HashMap<String,Node>();
} else if (labelSet.containsKey(name)) {
reportError("msg.dup.label");
}
boolean firstLabel;
if (statementLabel == null) {
firstLabel = true;
statementLabel = pn;
} else {
// Discard multiple label nodes and use only
// the first: it allows to simplify IRFactory
firstLabel = false;
}
labelSet.put(name, statementLabel);
try {
pn = statementHelper(statementLabel);
} finally {
labelSet.remove(name);
}
if (firstLabel) {
pn = nf.createLabeledStatement(statementLabel, pn);
}
return pn;
}
break;
}
default: {
int lineno = ts.getLineno();
pn = expr(false);
pn = nf.createExprStatement(pn, lineno);
break;
}
}
int ttFlagged = peekFlaggedToken();
switch (ttFlagged & CLEAR_TI_MASK) {
case Token.SEMI:
// Consume ';' as a part of expression
consumeToken();
break;
case Token.ERROR:
case Token.EOF:
case Token.RC:
// Autoinsert ;
break;
default:
if ((ttFlagged & TI_AFTER_EOL) == 0) {
// Report error if no EOL or autoinsert ; otherwise
reportError("msg.no.semi.stmt");
}
break;
}
decompiler.addEOL(Token.SEMI);
return pn;
}
/**
* Returns whether or not the bits in the mask have changed to all set.
* @param before bits before change
* @param after bits after change
* @param mask mask for bits
* @return true if all the bits in the mask are set in "after" but not
* "before"
*/
private static final boolean nowAllSet(int before, int after, int mask)
{
return ((before & mask) != mask) && ((after & mask) == mask);
}
private Node returnOrYield(int tt, boolean exprContext)
throws IOException, ParserException
{
if (!insideFunction()) {
reportError(tt == Token.RETURN ? "msg.bad.return"
: "msg.bad.yield");
}
consumeToken();
decompiler.addToken(tt);
int lineno = ts.getLineno();
Node e;
/* This is ugly, but we don't want to require a semicolon. */
switch (peekTokenOrEOL()) {
case Token.SEMI:
case Token.RC:
case Token.EOF:
case Token.EOL:
case Token.ERROR:
case Token.RB:
case Token.RP:
case Token.YIELD:
e = null;
break;
default:
e = expr(false);
break;
}
int before = endFlags;
Node ret;
if (tt == Token.RETURN) {
if (e == null ) {
endFlags |= Node.END_RETURNS;
} else {
endFlags |= Node.END_RETURNS_VALUE;
}
ret = nf.createReturn(e, lineno);
// see if we need a strict mode warning
if (nowAllSet(before, endFlags,
Node.END_RETURNS|Node.END_RETURNS_VALUE))
{
addStrictWarning("msg.return.inconsistent", "");
}
} else {
endFlags |= Node.END_YIELDS;
ret = nf.createYield(e, lineno);
if (!exprContext)
ret = new Node(Token.EXPR_VOID, ret, lineno);
}
// see if we are mixing yields and value returns.
if (nowAllSet(before, endFlags,
Node.END_YIELDS|Node.END_RETURNS_VALUE))
{
String name = ((FunctionNode)currentScriptOrFn).getFunctionName();
if (name.length() == 0)
addError("msg.anon.generator.returns", "");
else
addError("msg.generator.returns", name);
}
return ret;
}
/**
* Parse a 'var' or 'const' statement, or a 'var' init list in a for
* statement.
* @param inFor true if we are currently in the midst of the init
* clause of a for.
* @param declType A token value: either VAR, CONST, or LET depending on
* context.
* @return The parsed statement
* @throws IOException
* @throws ParserException
*/
private Node variables(boolean inFor, int declType)
throws IOException, ParserException
{
Node result = nf.createVariables(declType, ts.getLineno());
boolean first = true;
for (;;) {
Node destructuring = null;
String s = null;
int tt = peekToken();
if (tt == Token.LB || tt == Token.LC) {
// Destructuring assignment, e.g., var [a,b] = ...
destructuring = primaryExpr();
} else {
// Simple variable name
mustMatchToken(Token.NAME, "msg.bad.var");
s = ts.getString();
if (!first)
decompiler.addToken(Token.COMMA);
first = false;
decompiler.addName(s);
defineSymbol(declType, inFor, s);
}
Node init = null;
if (matchToken(Token.ASSIGN)) {
decompiler.addToken(Token.ASSIGN);
init = assignExpr(inFor);
}
if (destructuring != null) {
if (init == null) {
if (!inFor)
reportError("msg.destruct.assign.no.init");
nf.addChildToBack(result, destructuring);
} else {
nf.addChildToBack(result,
nf.createDestructuringAssignment(declType,
destructuring, init));
}
} else {
Node name = nf.createName(s);
if (init != null)
nf.addChildToBack(name, init);
nf.addChildToBack(result, name);
}
if (!matchToken(Token.COMMA))
break;
}
return result;
}
private Node let(boolean isStatement)
throws IOException, ParserException
{
mustMatchToken(Token.LP, "msg.no.paren.after.let");
decompiler.addToken(Token.LP);
Node result = nf.createScopeNode(Token.LET, ts.getLineno());
pushScope(result);
try {
Node vars = variables(false, Token.LET);
nf.addChildToBack(result, vars);
mustMatchToken(Token.RP, "msg.no.paren.let");
decompiler.addToken(Token.RP);
if (isStatement && peekToken() == Token.LC) {
// let statement
consumeToken();
decompiler.addEOL(Token.LC);
nf.addChildToBack(result, statements(null));
mustMatchToken(Token.RC, "msg.no.curly.let");
decompiler.addToken(Token.RC);
} else {
// let expression
result.setType(Token.LETEXPR);
nf.addChildToBack(result, expr(false));
if (isStatement) {
// let expression in statement context
result = nf.createExprStatement(result, ts.getLineno());
}
}
} finally {
popScope();
}
return result;
}
void defineSymbol(int declType, boolean ignoreNotInBlock, String name) {
Node.Scope definingScope = currentScope.getDefiningScope(name);
Node.Scope.Symbol symbol = definingScope != null
? definingScope.getSymbol(name)
: null;
boolean error = false;
if (symbol != null && (symbol.declType == Token.CONST ||
declType == Token.CONST))
{
error = true;
} else {
switch (declType) {
case Token.LET:
if (symbol != null && definingScope == currentScope) {
error = symbol.declType == Token.LET;
}
int currentScopeType = currentScope.getType();
if (!ignoreNotInBlock &&
((currentScopeType == Token.LOOP) ||
(currentScopeType == Token.IF)))
{
addError("msg.let.decl.not.in.block");
}
currentScope.putSymbol(name,
new Node.Scope.Symbol(declType, name));
break;
case Token.VAR:
case Token.CONST:
case Token.FUNCTION:
if (symbol != null) {
if (symbol.declType == Token.VAR)
addStrictWarning("msg.var.redecl", name);
else if (symbol.declType == Token.LP) {
addStrictWarning("msg.var.hides.arg", name);
}
} else {
currentScriptOrFn.putSymbol(name,
new Node.Scope.Symbol(declType, name));
}
break;
case Token.LP:
if (symbol != null) {
// must be duplicate parameter. Second parameter hides the
// first, so go ahead and add the second pararameter
addWarning("msg.dup.parms", name);
}
currentScriptOrFn.putSymbol(name,
new Node.Scope.Symbol(declType, name));
break;
default:
throw Kit.codeBug();
}
}
if (error) {
addError(symbol.declType == Token.CONST ? "msg.const.redecl" :
symbol.declType == Token.LET ? "msg.let.redecl" :
symbol.declType == Token.VAR ? "msg.var.redecl" :
symbol.declType == Token.FUNCTION ? "msg.fn.redecl" :
"msg.parm.redecl", name);
}
}
private Node expr(boolean inForInit)
throws IOException, ParserException
{
Node pn = assignExpr(inForInit);
while (matchToken(Token.COMMA)) {
decompiler.addToken(Token.COMMA);
if (compilerEnv.isStrictMode() && !pn.hasSideEffects())
addStrictWarning("msg.no.side.effects", "");
if (peekToken() == Token.YIELD) {
reportError("msg.yield.parenthesized");
}
pn = nf.createBinary(Token.COMMA, pn, assignExpr(inForInit));
}
return pn;
}
private Node assignExpr(boolean inForInit)
throws IOException, ParserException
{
int tt = peekToken();
if (tt == Token.YIELD) {
consumeToken();
return returnOrYield(tt, true);
}
Node pn = condExpr(inForInit);
tt = peekToken();
if (Token.FIRST_ASSIGN <= tt && tt <= Token.LAST_ASSIGN) {
consumeToken();
decompiler.addToken(tt);
pn = nf.createAssignment(tt, pn, assignExpr(inForInit));
}
return pn;
}
private Node condExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = orExpr(inForInit);
if (matchToken(Token.HOOK)) {
decompiler.addToken(Token.HOOK);
Node ifTrue = assignExpr(false);
mustMatchToken(Token.COLON, "msg.no.colon.cond");
decompiler.addToken(Token.COLON);
Node ifFalse = assignExpr(inForInit);
return nf.createCondExpr(pn, ifTrue, ifFalse);
}
return pn;
}
private Node orExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = andExpr(inForInit);
if (matchToken(Token.OR)) {
decompiler.addToken(Token.OR);
pn = nf.createBinary(Token.OR, pn, orExpr(inForInit));
}
return pn;
}
private Node andExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = bitOrExpr(inForInit);
if (matchToken(Token.AND)) {
decompiler.addToken(Token.AND);
pn = nf.createBinary(Token.AND, pn, andExpr(inForInit));
}
return pn;
}
private Node bitOrExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = bitXorExpr(inForInit);
while (matchToken(Token.BITOR)) {
decompiler.addToken(Token.BITOR);
pn = nf.createBinary(Token.BITOR, pn, bitXorExpr(inForInit));
}
return pn;
}
private Node bitXorExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = bitAndExpr(inForInit);
while (matchToken(Token.BITXOR)) {
decompiler.addToken(Token.BITXOR);
pn = nf.createBinary(Token.BITXOR, pn, bitAndExpr(inForInit));
}
return pn;
}
private Node bitAndExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = eqExpr(inForInit);
while (matchToken(Token.BITAND)) {
decompiler.addToken(Token.BITAND);
pn = nf.createBinary(Token.BITAND, pn, eqExpr(inForInit));
}
return pn;
}
private Node eqExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = relExpr(inForInit);
for (;;) {
int tt = peekToken();
switch (tt) {
case Token.EQ:
case Token.NE:
case Token.SHEQ:
case Token.SHNE:
consumeToken();
int decompilerToken = tt;
int parseToken = tt;
if (compilerEnv.getLanguageVersion() == Context.VERSION_1_2) {
// JavaScript 1.2 uses shallow equality for == and != .
// In addition, convert === and !== for decompiler into
// == and != since the decompiler is supposed to show
// canonical source and in 1.2 ===, !== are allowed
// only as an alias to ==, !=.
switch (tt) {
case Token.EQ:
parseToken = Token.SHEQ;
break;
case Token.NE:
parseToken = Token.SHNE;
break;
case Token.SHEQ:
decompilerToken = Token.EQ;
break;
case Token.SHNE:
decompilerToken = Token.NE;
break;
}
}
decompiler.addToken(decompilerToken);
pn = nf.createBinary(parseToken, pn, relExpr(inForInit));
continue;
}
break;
}
return pn;
}
private Node relExpr(boolean inForInit)
throws IOException, ParserException
{
Node pn = shiftExpr();
for (;;) {
int tt = peekToken();
switch (tt) {
case Token.IN:
if (inForInit)
break;
// fall through
case Token.INSTANCEOF:
case Token.LE:
case Token.LT:
case Token.GE:
case Token.GT:
consumeToken();
decompiler.addToken(tt);
pn = nf.createBinary(tt, pn, shiftExpr());
continue;
}
break;
}
return pn;
}
private Node shiftExpr()
throws IOException, ParserException
{
Node pn = addExpr();
for (;;) {
int tt = peekToken();
switch (tt) {
case Token.LSH:
case Token.URSH:
case Token.RSH:
consumeToken();
decompiler.addToken(tt);
pn = nf.createBinary(tt, pn, addExpr());
continue;
}
break;
}
return pn;
}
private Node addExpr()
throws IOException, ParserException
{
Node pn = mulExpr();
for (;;) {
int tt = peekToken();
if (tt == Token.ADD || tt == Token.SUB) {
consumeToken();
decompiler.addToken(tt);
// flushNewLines
pn = nf.createBinary(tt, pn, mulExpr());
continue;
}
break;
}
return pn;
}
private Node mulExpr()
throws IOException, ParserException
{
Node pn = unaryExpr();
for (;;) {
int tt = peekToken();
switch (tt) {
case Token.MUL:
case Token.DIV:
case Token.MOD:
consumeToken();
decompiler.addToken(tt);
pn = nf.createBinary(tt, pn, unaryExpr());
continue;
}
break;
}
return pn;
}
private Node unaryExpr()
throws IOException, ParserException
{
int tt;
tt = peekToken();
switch(tt) {
case Token.VOID:
case Token.NOT:
case Token.BITNOT:
case Token.TYPEOF:
consumeToken();
decompiler.addToken(tt);
return nf.createUnary(tt, unaryExpr());
case Token.ADD:
consumeToken();
// Convert to special POS token in decompiler and parse tree
decompiler.addToken(Token.POS);
return nf.createUnary(Token.POS, unaryExpr());
case Token.SUB:
consumeToken();
// Convert to special NEG token in decompiler and parse tree
decompiler.addToken(Token.NEG);
return nf.createUnary(Token.NEG, unaryExpr());
case Token.INC:
case Token.DEC:
consumeToken();
decompiler.addToken(tt);
return nf.createIncDec(tt, false, memberExpr(true));
case Token.DELPROP:
consumeToken();
decompiler.addToken(Token.DELPROP);
return nf.createUnary(Token.DELPROP, unaryExpr());
case Token.ERROR:
consumeToken();
break;
// XML stream encountered in expression.
case Token.LT:
if (compilerEnv.isXmlAvailable()) {
consumeToken();
Node pn = xmlInitializer();
return memberExprTail(true, pn);
}
// Fall thru to the default handling of RELOP
default:
Node pn = memberExpr(true);
// Don't look across a newline boundary for a postfix incop.
tt = peekTokenOrEOL();
if (tt == Token.INC || tt == Token.DEC) {
consumeToken();
decompiler.addToken(tt);
return nf.createIncDec(tt, true, pn);
}
return pn;
}
return nf.createName("error"); // Only reached on error.Try to continue.
}
private Node xmlInitializer() throws IOException
{
int tt = ts.getFirstXMLToken();
if (tt != Token.XML && tt != Token.XMLEND) {
reportError("msg.syntax");
return null;
}
/* Make a NEW node to append to. */
Node pnXML = nf.createLeaf(Token.NEW);
String xml = ts.getString();
boolean fAnonymous = xml.trim().startsWith("<>");
Node pn = nf.createName(fAnonymous ? "XMLList" : "XML");
nf.addChildToBack(pnXML, pn);
pn = null;
Node expr;
for (;;tt = ts.getNextXMLToken()) {
switch (tt) {
case Token.XML:
xml = ts.getString();
decompiler.addName(xml);
mustMatchToken(Token.LC, "msg.syntax");
decompiler.addToken(Token.LC);
expr = (peekToken() == Token.RC)
? nf.createString("")
: expr(false);
mustMatchToken(Token.RC, "msg.syntax");
decompiler.addToken(Token.RC);
if (pn == null) {
pn = nf.createString(xml);
} else {
pn = nf.createBinary(Token.ADD, pn, nf.createString(xml));
}
if (ts.isXMLAttribute()) {
/* Need to put the result in double quotes */
expr = nf.createUnary(Token.ESCXMLATTR, expr);
Node prepend = nf.createBinary(Token.ADD,
nf.createString("\""),
expr);
expr = nf.createBinary(Token.ADD,
prepend,
nf.createString("\""));
} else {
expr = nf.createUnary(Token.ESCXMLTEXT, expr);
}
pn = nf.createBinary(Token.ADD, pn, expr);
break;
case Token.XMLEND:
xml = ts.getString();
decompiler.addName(xml);
if (pn == null) {
pn = nf.createString(xml);
} else {
pn = nf.createBinary(Token.ADD, pn, nf.createString(xml));
}
nf.addChildToBack(pnXML, pn);
return pnXML;
default:
reportError("msg.syntax");
return null;
}
}
}
private void argumentList(Node listNode)
throws IOException, ParserException
{
boolean matched;
matched = matchToken(Token.RP);
if (!matched) {
boolean first = true;
do {
if (!first)
decompiler.addToken(Token.COMMA);
first = false;
if (peekToken() == Token.YIELD) {
reportError("msg.yield.parenthesized");
}
nf.addChildToBack(listNode, assignExpr(false));
} while (matchToken(Token.COMMA));
mustMatchToken(Token.RP, "msg.no.paren.arg");
}
decompiler.addToken(Token.RP);
}
private Node memberExpr(boolean allowCallSyntax)
throws IOException, ParserException
{
int tt;
Node pn;
/* Check for new expressions. */
tt = peekToken();
if (tt == Token.NEW) {
/* Eat the NEW token. */
consumeToken();
decompiler.addToken(Token.NEW);
/* Make a NEW node to append to. */
pn = nf.createCallOrNew(Token.NEW, memberExpr(false));
if (matchToken(Token.LP)) {
decompiler.addToken(Token.LP);
/* Add the arguments to pn, if any are supplied. */
argumentList(pn);
}
/* XXX there's a check in the C source against
* "too many constructor arguments" - how many
* do we claim to support?
*/
/* Experimental syntax: allow an object literal to follow a new expression,
* which will mean a kind of anonymous class built with the JavaAdapter.
* the object literal will be passed as an additional argument to the constructor.
*/
tt = peekToken();
if (tt == Token.LC) {
nf.addChildToBack(pn, primaryExpr());
}
} else {
pn = primaryExpr();
}
return memberExprTail(allowCallSyntax, pn);
}
private Node memberExprTail(boolean allowCallSyntax, Node pn)
throws IOException, ParserException
{
tailLoop:
for (;;) {
int tt = peekToken();
switch (tt) {
case Token.DOT:
case Token.DOTDOT:
{
int memberTypeFlags;
String s;
consumeToken();
decompiler.addToken(tt);
memberTypeFlags = 0;
if (tt == Token.DOTDOT) {
mustHaveXML();
memberTypeFlags = Node.DESCENDANTS_FLAG;
}
if (!compilerEnv.isXmlAvailable()) {
mustMatchToken(Token.NAME, "msg.no.name.after.dot");
s = ts.getString();
decompiler.addName(s);
pn = nf.createPropertyGet(pn, null, s, memberTypeFlags);
break;
}
tt = nextToken();
switch (tt) {
// needed for generator.throw();
case Token.THROW:
decompiler.addName("throw");
pn = propertyName(pn, "throw", memberTypeFlags);
break;
// handles: name, ns::name, ns::*, ns::[expr]
case Token.NAME:
s = ts.getString();
decompiler.addName(s);
pn = propertyName(pn, s, memberTypeFlags);
break;
// handles: *, *::name, *::*, *::[expr]
case Token.MUL:
decompiler.addName("*");
pn = propertyName(pn, "*", memberTypeFlags);
break;
// handles: '@attr', '@ns::attr', '@ns::*', '@ns::*',
// '@::attr', '@::*', '@*', '@*::attr', '@*::*'
case Token.XMLATTR:
decompiler.addToken(Token.XMLATTR);
pn = attributeAccess(pn, memberTypeFlags);
break;
default:
reportError("msg.no.name.after.dot");
}
}
break;
case Token.DOTQUERY:
consumeToken();
mustHaveXML();
decompiler.addToken(Token.DOTQUERY);
pn = nf.createDotQuery(pn, expr(false), ts.getLineno());
mustMatchToken(Token.RP, "msg.no.paren");
decompiler.addToken(Token.RP);
break;
case Token.LB:
consumeToken();
decompiler.addToken(Token.LB);
pn = nf.createElementGet(pn, null, expr(false), 0);
mustMatchToken(Token.RB, "msg.no.bracket.index");
decompiler.addToken(Token.RB);
break;
case Token.LP:
if (!allowCallSyntax) {
break tailLoop;
}
consumeToken();
decompiler.addToken(Token.LP);
pn = nf.createCallOrNew(Token.CALL, pn);
/* Add the arguments to pn, if any are supplied. */
argumentList(pn);
break;
default:
break tailLoop;
}
}
return pn;
}
/*
* Xml attribute expression:
* '@attr', '@ns::attr', '@ns::*', '@ns::*', '@*', '@*::attr', '@*::*'
*/
private Node attributeAccess(Node pn, int memberTypeFlags)
throws IOException
{
memberTypeFlags |= Node.ATTRIBUTE_FLAG;
int tt = nextToken();
switch (tt) {
// handles: @name, @ns::name, @ns::*, @ns::[expr]
case Token.NAME:
{
String s = ts.getString();
decompiler.addName(s);
pn = propertyName(pn, s, memberTypeFlags);
}
break;
// handles: @*, @*::name, @*::*, @*::[expr]
case Token.MUL:
decompiler.addName("*");
pn = propertyName(pn, "*", memberTypeFlags);
break;
// handles @[expr]
case Token.LB:
decompiler.addToken(Token.LB);
pn = nf.createElementGet(pn, null, expr(false), memberTypeFlags);
mustMatchToken(Token.RB, "msg.no.bracket.index");
decompiler.addToken(Token.RB);
break;
default:
reportError("msg.no.name.after.xmlAttr");
pn = nf.createPropertyGet(pn, null, "?", memberTypeFlags);
break;
}
return pn;
}
/**
* Check if :: follows name in which case it becomes qualified name
*/
private Node propertyName(Node pn, String name, int memberTypeFlags)
throws IOException, ParserException
{
String namespace = null;
if (matchToken(Token.COLONCOLON)) {
decompiler.addToken(Token.COLONCOLON);
namespace = name;
int tt = nextToken();
switch (tt) {
// handles name::name
case Token.NAME:
name = ts.getString();
decompiler.addName(name);
break;
// handles name::*
case Token.MUL:
decompiler.addName("*");
name = "*";
break;
// handles name::[expr]
case Token.LB:
decompiler.addToken(Token.LB);
pn = nf.createElementGet(pn, namespace, expr(false),
memberTypeFlags);
mustMatchToken(Token.RB, "msg.no.bracket.index");
decompiler.addToken(Token.RB);
return pn;
default:
reportError("msg.no.name.after.coloncolon");
name = "?";
}
}
pn = nf.createPropertyGet(pn, namespace, name, memberTypeFlags);
return pn;
}
private Node arrayComprehension(String arrayName, Node expr)
throws IOException, ParserException
{
if (nextToken() != Token.FOR)
throw Kit.codeBug(); // shouldn't be here if next token isn't 'for'
decompiler.addName(" "); // space after array literal expr
decompiler.addToken(Token.FOR);
boolean isForEach = false;
if (matchToken(Token.NAME)) {
decompiler.addName(ts.getString());
if (ts.getString().equals("each")) {
isForEach = true;
} else {
reportError("msg.no.paren.for");
}
}
mustMatchToken(Token.LP, "msg.no.paren.for");
decompiler.addToken(Token.LP);
String name;
int tt = peekToken();
if (tt == Token.LB || tt == Token.LC) {
// handle destructuring assignment
name = currentScriptOrFn.getNextTempName();
defineSymbol(Token.LP, false, name);
expr = nf.createBinary(Token.COMMA,
nf.createAssignment(Token.ASSIGN, primaryExpr(),
nf.createName(name)),
expr);
} else if (tt == Token.NAME) {
consumeToken();
name = ts.getString();
decompiler.addName(name);
} else {
reportError("msg.bad.var");
return nf.createNumber(0);
}
Node init = nf.createName(name);
// Define as a let since we want the scope of the variable to
// be restricted to the array comprehension
defineSymbol(Token.LET, false, name);
mustMatchToken(Token.IN, "msg.in.after.for.name");
decompiler.addToken(Token.IN);
Node iterator = expr(false);
mustMatchToken(Token.RP, "msg.no.paren.for.ctrl");
decompiler.addToken(Token.RP);
Node body;
tt = peekToken();
if (tt == Token.FOR) {
body = arrayComprehension(arrayName, expr);
} else {
Node call = nf.createCallOrNew(Token.CALL,
nf.createPropertyGet(nf.createName(arrayName), null,
"push", 0));
call.addChildToBack(expr);
body = new Node(Token.EXPR_VOID, call, ts.getLineno());
if (tt == Token.IF) {
consumeToken();
decompiler.addToken(Token.IF);
int lineno = ts.getLineno();
Node cond = condition();
body = nf.createIf(cond, body, null, lineno);
}
mustMatchToken(Token.RB, "msg.no.bracket.arg");
decompiler.addToken(Token.RB);
}
Node loop = enterLoop(null, true);
try {
return nf.createForIn(Token.LET, loop, init, iterator, body,
isForEach);
} finally {
exitLoop(false);
}
}
private Node primaryExpr()
throws IOException, ParserException
{
Node pn;
int ttFlagged = nextFlaggedToken();
int tt = ttFlagged & CLEAR_TI_MASK;
switch(tt) {
case Token.FUNCTION:
return function(FunctionNode.FUNCTION_EXPRESSION);
case Token.LB: {
ObjArray elems = new ObjArray();
int skipCount = 0;
int destructuringLen = 0;
decompiler.addToken(Token.LB);
boolean after_lb_or_comma = true;
for (;;) {
tt = peekToken();
if (tt == Token.COMMA) {
consumeToken();
decompiler.addToken(Token.COMMA);
if (!after_lb_or_comma) {
after_lb_or_comma = true;
} else {
elems.add(null);
++skipCount;
}
} else if (tt == Token.RB) {
consumeToken();
decompiler.addToken(Token.RB);
// for ([a,] in obj) is legal, but for ([a] in obj) is
// not since we have both key and value supplied. The
// trick is that [a,] and [a] are equivalent in other
// array literal contexts. So we calculate a special
// length value just for destructuring assignment.
destructuringLen = elems.size() +
(after_lb_or_comma ? 1 : 0);
break;
} else if (skipCount == 0 && elems.size() == 1 &&
tt == Token.FOR)
{
Node scopeNode = nf.createScopeNode(Token.ARRAYCOMP,
ts.getLineno());
String tempName = currentScriptOrFn.getNextTempName();
pushScope(scopeNode);
try {
defineSymbol(Token.LET, false, tempName);
Node expr = (Node) elems.get(0);
Node block = nf.createBlock(ts.getLineno());
Node init = new Node(Token.EXPR_VOID,
nf.createAssignment(Token.ASSIGN,
nf.createName(tempName),
nf.createCallOrNew(Token.NEW,
nf.createName("Array"))), ts.getLineno());
block.addChildToBack(init);
block.addChildToBack(arrayComprehension(tempName,
expr));
scopeNode.addChildToBack(block);
scopeNode.addChildToBack(nf.createName(tempName));
return scopeNode;
} finally {
popScope();
}
} else {
if (!after_lb_or_comma) {
reportError("msg.no.bracket.arg");
}
elems.add(assignExpr(false));
after_lb_or_comma = false;
}
}
return nf.createArrayLiteral(elems, skipCount, destructuringLen);
}
case Token.LC: {
ObjArray elems = new ObjArray();
decompiler.addToken(Token.LC);
if (!matchToken(Token.RC)) {
boolean first = true;
commaloop:
do {
Object property;
if (!first)
decompiler.addToken(Token.COMMA);
else
first = false;
tt = peekToken();
switch(tt) {
case Token.NAME:
case Token.STRING:
consumeToken();
// map NAMEs to STRINGs in object literal context
// but tell the decompiler the proper type
String s = ts.getString();
if (tt == Token.NAME) {
if (s.equals("get") &&
peekToken() == Token.NAME) {
decompiler.addToken(Token.GET);
consumeToken();
s = ts.getString();
decompiler.addName(s);
property = ScriptRuntime.getIndexObject(s);
if (!getterSetterProperty(elems, property,
true))
break commaloop;
break;
} else if (s.equals("set") &&
peekToken() == Token.NAME) {
decompiler.addToken(Token.SET);
consumeToken();
s = ts.getString();
decompiler.addName(s);
property = ScriptRuntime.getIndexObject(s);
if (!getterSetterProperty(elems, property,
false))
break commaloop;
break;
}
decompiler.addName(s);
} else {
decompiler.addString(s);
}
property = ScriptRuntime.getIndexObject(s);
plainProperty(elems, property);
break;
case Token.NUMBER:
consumeToken();
double n = ts.getNumber();
decompiler.addNumber(n);
property = ScriptRuntime.getIndexObject(n);
plainProperty(elems, property);
break;
case Token.RC:
// trailing comma is OK.
break commaloop;
default:
reportError("msg.bad.prop");
break commaloop;
}
} while (matchToken(Token.COMMA));
mustMatchToken(Token.RC, "msg.no.brace.prop");
}
decompiler.addToken(Token.RC);
return nf.createObjectLiteral(elems);
}
case Token.LET:
decompiler.addToken(Token.LET);
return let(false);
case Token.LP:
/* Brendan's IR-jsparse.c makes a new node tagged with
* TOK_LP here... I'm not sure I understand why. Isn't
* the grouping already implicit in the structure of the
* parse tree? also TOK_LP is already overloaded (I
* think) in the C IR as 'function call.' */
decompiler.addToken(Token.LP);
pn = expr(false);
pn.putProp(Node.PARENTHESIZED_PROP, Boolean.TRUE);
decompiler.addToken(Token.RP);
mustMatchToken(Token.RP, "msg.no.paren");
return pn;
case Token.XMLATTR:
mustHaveXML();
decompiler.addToken(Token.XMLATTR);
pn = attributeAccess(null, 0);
return pn;
case Token.NAME: {
String name = ts.getString();
if ((ttFlagged & TI_CHECK_LABEL) != 0) {
if (peekToken() == Token.COLON) {
// Do not consume colon, it is used as unwind indicator
// to return to statementHelper.
// XXX Better way?
return nf.createLabel(ts.getLineno());
}
}
decompiler.addName(name);
if (compilerEnv.isXmlAvailable()) {
pn = propertyName(null, name, 0);
} else {
pn = nf.createName(name);
}
return pn;
}
case Token.NUMBER: {
double n = ts.getNumber();
decompiler.addNumber(n);
return nf.createNumber(n);
}
case Token.STRING: {
String s = ts.getString();
decompiler.addString(s);
return nf.createString(s);
}
case Token.DIV:
case Token.ASSIGN_DIV: {
// Got / or /= which should be treated as regexp in fact
ts.readRegExp(tt);
String flags = ts.regExpFlags;
ts.regExpFlags = null;
String re = ts.getString();
decompiler.addRegexp(re, flags);
int index = currentScriptOrFn.addRegexp(re, flags);
return nf.createRegExp(index);
}
case Token.NULL:
case Token.THIS:
case Token.FALSE:
case Token.TRUE:
decompiler.addToken(tt);
return nf.createLeaf(tt);
case Token.RESERVED:
reportError("msg.reserved.id");
break;
case Token.ERROR:
/* the scanner or one of its subroutines reported the error. */
break;
case Token.EOF:
reportError("msg.unexpected.eof");
break;
default:
reportError("msg.syntax");
break;
}
return null; // should never reach here
}
private void plainProperty(ObjArray elems, Object property)
throws IOException {
mustMatchToken(Token.COLON, "msg.no.colon.prop");
// OBJLIT is used as ':' in object literal for
// decompilation to solve spacing ambiguity.
decompiler.addToken(Token.OBJECTLIT);
elems.add(property);
elems.add(assignExpr(false));
}
private boolean getterSetterProperty(ObjArray elems, Object property,
boolean isGetter) throws IOException {
Node f = function(FunctionNode.FUNCTION_EXPRESSION);
if (f.getType() != Token.FUNCTION) {
reportError("msg.bad.prop");
return false;
}
int fnIndex = f.getExistingIntProp(Node.FUNCTION_PROP);
FunctionNode fn = currentScriptOrFn.getFunctionNode(fnIndex);
if (fn.getFunctionName().length() != 0) {
reportError("msg.bad.prop");
return false;
}
elems.add(property);
if (isGetter) {
elems.add(nf.createUnary(Token.GET, f));
} else {
elems.add(nf.createUnary(Token.SET, f));
}
return true;
}
}