blob: fec1567c0ec6d46cfb0182010f297b05a1570ad3 [file] [log] [blame]
// Copyright 2010 the V8 project authors. All rights reserved.
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following
// disclaimer in the documentation and/or other materials provided
// with the distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived
// from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include "../include/v8stdint.h"
#include "unicode.h"
#include "globals.h"
#include "checks.h"
#include "allocation.h"
#include "utils.h"
#include "list.h"
#include "scanner-base.h"
#include "preparse-data.h"
#include "preparser.h"
namespace v8 {
namespace preparser {
// Preparsing checks a JavaScript program and emits preparse-data that helps
// a later parsing to be faster.
// See preparser-data.h for the data.
// The PreParser checks that the syntax follows the grammar for JavaScript,
// and collects some information about the program along the way.
// The grammar check is only performed in order to understand the program
// sufficiently to deduce some information about it, that can be used
// to speed up later parsing. Finding errors is not the goal of pre-parsing,
// rather it is to speed up properly written and correct programs.
// That means that contextual checks (like a label being declared where
// it is used) are generally omitted.
namespace i = ::v8::internal;
#define CHECK_OK ok); \
if (!*ok) return -1; \
((void)0
#define DUMMY ) // to make indentation work
#undef DUMMY
void PreParser::ReportUnexpectedToken(i::Token::Value token) {
// We don't report stack overflows here, to avoid increasing the
// stack depth even further. Instead we report it after parsing is
// over, in ParseProgram.
if (token == i::Token::ILLEGAL && stack_overflow_) {
return;
}
i::JavaScriptScanner::Location source_location = scanner_->location();
// Four of the tokens are treated specially
switch (token) {
case i::Token::EOS:
return ReportMessageAt(source_location.beg_pos, source_location.end_pos,
"unexpected_eos", NULL);
case i::Token::NUMBER:
return ReportMessageAt(source_location.beg_pos, source_location.end_pos,
"unexpected_token_number", NULL);
case i::Token::STRING:
return ReportMessageAt(source_location.beg_pos, source_location.end_pos,
"unexpected_token_string", NULL);
case i::Token::IDENTIFIER:
case i::Token::FUTURE_RESERVED_WORD:
return ReportMessageAt(source_location.beg_pos, source_location.end_pos,
"unexpected_token_identifier", NULL);
default:
const char* name = i::Token::String(token);
ReportMessageAt(source_location.beg_pos, source_location.end_pos,
"unexpected_token", name);
}
}
PreParser::SourceElements PreParser::ParseSourceElements(int end_token,
bool* ok) {
// SourceElements ::
// (Statement)* <end_token>
while (peek() != end_token) {
ParseStatement(CHECK_OK);
}
return kUnknownSourceElements;
}
PreParser::Statement PreParser::ParseStatement(bool* ok) {
// Statement ::
// Block
// VariableStatement
// EmptyStatement
// ExpressionStatement
// IfStatement
// IterationStatement
// ContinueStatement
// BreakStatement
// ReturnStatement
// WithStatement
// LabelledStatement
// SwitchStatement
// ThrowStatement
// TryStatement
// DebuggerStatement
// Note: Since labels can only be used by 'break' and 'continue'
// statements, which themselves are only valid within blocks,
// iterations or 'switch' statements (i.e., BreakableStatements),
// labels can be simply ignored in all other cases; except for
// trivial labeled break statements 'label: break label' which is
// parsed into an empty statement.
// Keep the source position of the statement
switch (peek()) {
case i::Token::LBRACE:
return ParseBlock(ok);
case i::Token::CONST:
case i::Token::VAR:
return ParseVariableStatement(ok);
case i::Token::SEMICOLON:
Next();
return kUnknownStatement;
case i::Token::IF:
return ParseIfStatement(ok);
case i::Token::DO:
return ParseDoWhileStatement(ok);
case i::Token::WHILE:
return ParseWhileStatement(ok);
case i::Token::FOR:
return ParseForStatement(ok);
case i::Token::CONTINUE:
return ParseContinueStatement(ok);
case i::Token::BREAK:
return ParseBreakStatement(ok);
case i::Token::RETURN:
return ParseReturnStatement(ok);
case i::Token::WITH:
return ParseWithStatement(ok);
case i::Token::SWITCH:
return ParseSwitchStatement(ok);
case i::Token::THROW:
return ParseThrowStatement(ok);
case i::Token::TRY:
return ParseTryStatement(ok);
case i::Token::FUNCTION:
return ParseFunctionDeclaration(ok);
case i::Token::NATIVE:
return ParseNativeDeclaration(ok);
case i::Token::DEBUGGER:
return ParseDebuggerStatement(ok);
default:
return ParseExpressionOrLabelledStatement(ok);
}
}
PreParser::Statement PreParser::ParseFunctionDeclaration(bool* ok) {
// FunctionDeclaration ::
// 'function' Identifier '(' FormalParameterListopt ')' '{' FunctionBody '}'
Expect(i::Token::FUNCTION, CHECK_OK);
ParseIdentifier(CHECK_OK);
ParseFunctionLiteral(CHECK_OK);
return kUnknownStatement;
}
// Language extension which is only enabled for source files loaded
// through the API's extension mechanism. A native function
// declaration is resolved by looking up the function through a
// callback provided by the extension.
PreParser::Statement PreParser::ParseNativeDeclaration(bool* ok) {
Expect(i::Token::NATIVE, CHECK_OK);
Expect(i::Token::FUNCTION, CHECK_OK);
ParseIdentifier(CHECK_OK);
Expect(i::Token::LPAREN, CHECK_OK);
bool done = (peek() == i::Token::RPAREN);
while (!done) {
ParseIdentifier(CHECK_OK);
done = (peek() == i::Token::RPAREN);
if (!done) {
Expect(i::Token::COMMA, CHECK_OK);
}
}
Expect(i::Token::RPAREN, CHECK_OK);
Expect(i::Token::SEMICOLON, CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseBlock(bool* ok) {
// Block ::
// '{' Statement* '}'
// Note that a Block does not introduce a new execution scope!
// (ECMA-262, 3rd, 12.2)
//
Expect(i::Token::LBRACE, CHECK_OK);
while (peek() != i::Token::RBRACE) {
ParseStatement(CHECK_OK);
}
Expect(i::Token::RBRACE, CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseVariableStatement(bool* ok) {
// VariableStatement ::
// VariableDeclarations ';'
Statement result = ParseVariableDeclarations(true, NULL, CHECK_OK);
ExpectSemicolon(CHECK_OK);
return result;
}
// If the variable declaration declares exactly one non-const
// variable, then *var is set to that variable. In all other cases,
// *var is untouched; in particular, it is the caller's responsibility
// to initialize it properly. This mechanism is also used for the parsing
// of 'for-in' loops.
PreParser::Statement PreParser::ParseVariableDeclarations(bool accept_IN,
int* num_decl,
bool* ok) {
// VariableDeclarations ::
// ('var' | 'const') (Identifier ('=' AssignmentExpression)?)+[',']
if (peek() == i::Token::VAR) {
Consume(i::Token::VAR);
} else if (peek() == i::Token::CONST) {
Consume(i::Token::CONST);
} else {
*ok = false;
return 0;
}
// The scope of a variable/const declared anywhere inside a function
// is the entire function (ECMA-262, 3rd, 10.1.3, and 12.2). .
int nvars = 0; // the number of variables declared
do {
// Parse variable name.
if (nvars > 0) Consume(i::Token::COMMA);
ParseIdentifier(CHECK_OK);
nvars++;
if (peek() == i::Token::ASSIGN) {
Expect(i::Token::ASSIGN, CHECK_OK);
ParseAssignmentExpression(accept_IN, CHECK_OK);
}
} while (peek() == i::Token::COMMA);
if (num_decl != NULL) *num_decl = nvars;
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseExpressionOrLabelledStatement(
bool* ok) {
// ExpressionStatement | LabelledStatement ::
// Expression ';'
// Identifier ':' Statement
Expression expr = ParseExpression(true, CHECK_OK);
if (peek() == i::Token::COLON && expr == kIdentifierExpression) {
Consume(i::Token::COLON);
return ParseStatement(ok);
}
// Parsed expression statement.
ExpectSemicolon(CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseIfStatement(bool* ok) {
// IfStatement ::
// 'if' '(' Expression ')' Statement ('else' Statement)?
Expect(i::Token::IF, CHECK_OK);
Expect(i::Token::LPAREN, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
ParseStatement(CHECK_OK);
if (peek() == i::Token::ELSE) {
Next();
ParseStatement(CHECK_OK);
}
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseContinueStatement(bool* ok) {
// ContinueStatement ::
// 'continue' [no line terminator] Identifier? ';'
Expect(i::Token::CONTINUE, CHECK_OK);
i::Token::Value tok = peek();
if (!scanner_->has_line_terminator_before_next() &&
tok != i::Token::SEMICOLON &&
tok != i::Token::RBRACE &&
tok != i::Token::EOS) {
ParseIdentifier(CHECK_OK);
}
ExpectSemicolon(CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseBreakStatement(bool* ok) {
// BreakStatement ::
// 'break' [no line terminator] Identifier? ';'
Expect(i::Token::BREAK, CHECK_OK);
i::Token::Value tok = peek();
if (!scanner_->has_line_terminator_before_next() &&
tok != i::Token::SEMICOLON &&
tok != i::Token::RBRACE &&
tok != i::Token::EOS) {
ParseIdentifier(CHECK_OK);
}
ExpectSemicolon(CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseReturnStatement(bool* ok) {
// ReturnStatement ::
// 'return' [no line terminator] Expression? ';'
// Consume the return token. It is necessary to do the before
// reporting any errors on it, because of the way errors are
// reported (underlining).
Expect(i::Token::RETURN, CHECK_OK);
// An ECMAScript program is considered syntactically incorrect if it
// contains a return statement that is not within the body of a
// function. See ECMA-262, section 12.9, page 67.
// This is not handled during preparsing.
i::Token::Value tok = peek();
if (!scanner_->has_line_terminator_before_next() &&
tok != i::Token::SEMICOLON &&
tok != i::Token::RBRACE &&
tok != i::Token::EOS) {
ParseExpression(true, CHECK_OK);
}
ExpectSemicolon(CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseWithStatement(bool* ok) {
// WithStatement ::
// 'with' '(' Expression ')' Statement
Expect(i::Token::WITH, CHECK_OK);
Expect(i::Token::LPAREN, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
scope_->EnterWith();
ParseStatement(CHECK_OK);
scope_->LeaveWith();
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseSwitchStatement(bool* ok) {
// SwitchStatement ::
// 'switch' '(' Expression ')' '{' CaseClause* '}'
Expect(i::Token::SWITCH, CHECK_OK);
Expect(i::Token::LPAREN, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
Expect(i::Token::LBRACE, CHECK_OK);
i::Token::Value token = peek();
while (token != i::Token::RBRACE) {
if (token == i::Token::CASE) {
Expect(i::Token::CASE, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::COLON, CHECK_OK);
} else if (token == i::Token::DEFAULT) {
Expect(i::Token::DEFAULT, CHECK_OK);
Expect(i::Token::COLON, CHECK_OK);
} else {
ParseStatement(CHECK_OK);
}
token = peek();
}
Expect(i::Token::RBRACE, CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseDoWhileStatement(bool* ok) {
// DoStatement ::
// 'do' Statement 'while' '(' Expression ')' ';'
Expect(i::Token::DO, CHECK_OK);
ParseStatement(CHECK_OK);
Expect(i::Token::WHILE, CHECK_OK);
Expect(i::Token::LPAREN, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseWhileStatement(bool* ok) {
// WhileStatement ::
// 'while' '(' Expression ')' Statement
Expect(i::Token::WHILE, CHECK_OK);
Expect(i::Token::LPAREN, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
ParseStatement(CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseForStatement(bool* ok) {
// ForStatement ::
// 'for' '(' Expression? ';' Expression? ';' Expression? ')' Statement
Expect(i::Token::FOR, CHECK_OK);
Expect(i::Token::LPAREN, CHECK_OK);
if (peek() != i::Token::SEMICOLON) {
if (peek() == i::Token::VAR || peek() == i::Token::CONST) {
int decl_count;
ParseVariableDeclarations(false, &decl_count, CHECK_OK);
if (peek() == i::Token::IN && decl_count == 1) {
Expect(i::Token::IN, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
ParseStatement(CHECK_OK);
return kUnknownStatement;
}
} else {
ParseExpression(false, CHECK_OK);
if (peek() == i::Token::IN) {
Expect(i::Token::IN, CHECK_OK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
ParseStatement(CHECK_OK);
return kUnknownStatement;
}
}
}
// Parsed initializer at this point.
Expect(i::Token::SEMICOLON, CHECK_OK);
if (peek() != i::Token::SEMICOLON) {
ParseExpression(true, CHECK_OK);
}
Expect(i::Token::SEMICOLON, CHECK_OK);
if (peek() != i::Token::RPAREN) {
ParseExpression(true, CHECK_OK);
}
Expect(i::Token::RPAREN, CHECK_OK);
ParseStatement(CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseThrowStatement(bool* ok) {
// ThrowStatement ::
// 'throw' [no line terminator] Expression ';'
Expect(i::Token::THROW, CHECK_OK);
if (scanner_->has_line_terminator_before_next()) {
i::JavaScriptScanner::Location pos = scanner_->location();
ReportMessageAt(pos.beg_pos, pos.end_pos,
"newline_after_throw", NULL);
*ok = false;
return kUnknownStatement;
}
ParseExpression(true, CHECK_OK);
ExpectSemicolon(CHECK_OK);
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseTryStatement(bool* ok) {
// TryStatement ::
// 'try' Block Catch
// 'try' Block Finally
// 'try' Block Catch Finally
//
// Catch ::
// 'catch' '(' Identifier ')' Block
//
// Finally ::
// 'finally' Block
// In preparsing, allow any number of catch/finally blocks, including zero
// of both.
Expect(i::Token::TRY, CHECK_OK);
ParseBlock(CHECK_OK);
bool catch_or_finally_seen = false;
if (peek() == i::Token::CATCH) {
Consume(i::Token::CATCH);
Expect(i::Token::LPAREN, CHECK_OK);
ParseIdentifier(CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
scope_->EnterWith();
ParseBlock(ok);
scope_->LeaveWith();
if (!*ok) return kUnknownStatement;
catch_or_finally_seen = true;
}
if (peek() == i::Token::FINALLY) {
Consume(i::Token::FINALLY);
ParseBlock(CHECK_OK);
catch_or_finally_seen = true;
}
if (!catch_or_finally_seen) {
*ok = false;
}
return kUnknownStatement;
}
PreParser::Statement PreParser::ParseDebuggerStatement(bool* ok) {
// In ECMA-262 'debugger' is defined as a reserved keyword. In some browser
// contexts this is used as a statement which invokes the debugger as if a
// break point is present.
// DebuggerStatement ::
// 'debugger' ';'
Expect(i::Token::DEBUGGER, CHECK_OK);
ExpectSemicolon(CHECK_OK);
return kUnknownStatement;
}
// Precedence = 1
PreParser::Expression PreParser::ParseExpression(bool accept_IN, bool* ok) {
// Expression ::
// AssignmentExpression
// Expression ',' AssignmentExpression
Expression result = ParseAssignmentExpression(accept_IN, CHECK_OK);
while (peek() == i::Token::COMMA) {
Expect(i::Token::COMMA, CHECK_OK);
ParseAssignmentExpression(accept_IN, CHECK_OK);
result = kUnknownExpression;
}
return result;
}
// Precedence = 2
PreParser::Expression PreParser::ParseAssignmentExpression(bool accept_IN,
bool* ok) {
// AssignmentExpression ::
// ConditionalExpression
// LeftHandSideExpression AssignmentOperator AssignmentExpression
Expression expression = ParseConditionalExpression(accept_IN, CHECK_OK);
if (!i::Token::IsAssignmentOp(peek())) {
// Parsed conditional expression only (no assignment).
return expression;
}
i::Token::Value op = Next(); // Get assignment operator.
ParseAssignmentExpression(accept_IN, CHECK_OK);
if ((op == i::Token::ASSIGN) && (expression == kThisPropertyExpression)) {
scope_->AddProperty();
}
return kUnknownExpression;
}
// Precedence = 3
PreParser::Expression PreParser::ParseConditionalExpression(bool accept_IN,
bool* ok) {
// ConditionalExpression ::
// LogicalOrExpression
// LogicalOrExpression '?' AssignmentExpression ':' AssignmentExpression
// We start using the binary expression parser for prec >= 4 only!
Expression expression = ParseBinaryExpression(4, accept_IN, CHECK_OK);
if (peek() != i::Token::CONDITIONAL) return expression;
Consume(i::Token::CONDITIONAL);
// In parsing the first assignment expression in conditional
// expressions we always accept the 'in' keyword; see ECMA-262,
// section 11.12, page 58.
ParseAssignmentExpression(true, CHECK_OK);
Expect(i::Token::COLON, CHECK_OK);
ParseAssignmentExpression(accept_IN, CHECK_OK);
return kUnknownExpression;
}
int PreParser::Precedence(i::Token::Value tok, bool accept_IN) {
if (tok == i::Token::IN && !accept_IN)
return 0; // 0 precedence will terminate binary expression parsing
return i::Token::Precedence(tok);
}
// Precedence >= 4
PreParser::Expression PreParser::ParseBinaryExpression(int prec,
bool accept_IN,
bool* ok) {
Expression result = ParseUnaryExpression(CHECK_OK);
for (int prec1 = Precedence(peek(), accept_IN); prec1 >= prec; prec1--) {
// prec1 >= 4
while (Precedence(peek(), accept_IN) == prec1) {
Next();
ParseBinaryExpression(prec1 + 1, accept_IN, CHECK_OK);
result = kUnknownExpression;
}
}
return result;
}
PreParser::Expression PreParser::ParseUnaryExpression(bool* ok) {
// UnaryExpression ::
// PostfixExpression
// 'delete' UnaryExpression
// 'void' UnaryExpression
// 'typeof' UnaryExpression
// '++' UnaryExpression
// '--' UnaryExpression
// '+' UnaryExpression
// '-' UnaryExpression
// '~' UnaryExpression
// '!' UnaryExpression
i::Token::Value op = peek();
if (i::Token::IsUnaryOp(op) || i::Token::IsCountOp(op)) {
op = Next();
ParseUnaryExpression(ok);
return kUnknownExpression;
} else {
return ParsePostfixExpression(ok);
}
}
PreParser::Expression PreParser::ParsePostfixExpression(bool* ok) {
// PostfixExpression ::
// LeftHandSideExpression ('++' | '--')?
Expression expression = ParseLeftHandSideExpression(CHECK_OK);
if (!scanner_->has_line_terminator_before_next() &&
i::Token::IsCountOp(peek())) {
Next();
return kUnknownExpression;
}
return expression;
}
PreParser::Expression PreParser::ParseLeftHandSideExpression(bool* ok) {
// LeftHandSideExpression ::
// (NewExpression | MemberExpression) ...
Expression result;
if (peek() == i::Token::NEW) {
result = ParseNewExpression(CHECK_OK);
} else {
result = ParseMemberExpression(CHECK_OK);
}
while (true) {
switch (peek()) {
case i::Token::LBRACK: {
Consume(i::Token::LBRACK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RBRACK, CHECK_OK);
if (result == kThisExpression) {
result = kThisPropertyExpression;
} else {
result = kUnknownExpression;
}
break;
}
case i::Token::LPAREN: {
ParseArguments(CHECK_OK);
result = kUnknownExpression;
break;
}
case i::Token::PERIOD: {
Consume(i::Token::PERIOD);
ParseIdentifierName(CHECK_OK);
if (result == kThisExpression) {
result = kThisPropertyExpression;
} else {
result = kUnknownExpression;
}
break;
}
default:
return result;
}
}
}
PreParser::Expression PreParser::ParseNewExpression(bool* ok) {
// NewExpression ::
// ('new')+ MemberExpression
// The grammar for new expressions is pretty warped. The keyword
// 'new' can either be a part of the new expression (where it isn't
// followed by an argument list) or a part of the member expression,
// where it must be followed by an argument list. To accommodate
// this, we parse the 'new' keywords greedily and keep track of how
// many we have parsed. This information is then passed on to the
// member expression parser, which is only allowed to match argument
// lists as long as it has 'new' prefixes left
unsigned new_count = 0;
do {
Consume(i::Token::NEW);
new_count++;
} while (peek() == i::Token::NEW);
return ParseMemberWithNewPrefixesExpression(new_count, ok);
}
PreParser::Expression PreParser::ParseMemberExpression(bool* ok) {
return ParseMemberWithNewPrefixesExpression(0, ok);
}
PreParser::Expression PreParser::ParseMemberWithNewPrefixesExpression(
unsigned new_count, bool* ok) {
// MemberExpression ::
// (PrimaryExpression | FunctionLiteral)
// ('[' Expression ']' | '.' Identifier | Arguments)*
// Parse the initial primary or function expression.
Expression result = kUnknownExpression;
if (peek() == i::Token::FUNCTION) {
Consume(i::Token::FUNCTION);
if (peek_any_identifier()) {
ParseIdentifier(CHECK_OK);
}
result = ParseFunctionLiteral(CHECK_OK);
} else {
result = ParsePrimaryExpression(CHECK_OK);
}
while (true) {
switch (peek()) {
case i::Token::LBRACK: {
Consume(i::Token::LBRACK);
ParseExpression(true, CHECK_OK);
Expect(i::Token::RBRACK, CHECK_OK);
if (result == kThisExpression) {
result = kThisPropertyExpression;
} else {
result = kUnknownExpression;
}
break;
}
case i::Token::PERIOD: {
Consume(i::Token::PERIOD);
ParseIdentifierName(CHECK_OK);
if (result == kThisExpression) {
result = kThisPropertyExpression;
} else {
result = kUnknownExpression;
}
break;
}
case i::Token::LPAREN: {
if (new_count == 0) return result;
// Consume one of the new prefixes (already parsed).
ParseArguments(CHECK_OK);
new_count--;
result = kUnknownExpression;
break;
}
default:
return result;
}
}
}
PreParser::Expression PreParser::ParsePrimaryExpression(bool* ok) {
// PrimaryExpression ::
// 'this'
// 'null'
// 'true'
// 'false'
// Identifier
// Number
// String
// ArrayLiteral
// ObjectLiteral
// RegExpLiteral
// '(' Expression ')'
Expression result = kUnknownExpression;
switch (peek()) {
case i::Token::THIS: {
Next();
result = kThisExpression;
break;
}
case i::Token::IDENTIFIER:
case i::Token::FUTURE_RESERVED_WORD: {
ParseIdentifier(CHECK_OK);
result = kIdentifierExpression;
break;
}
case i::Token::NULL_LITERAL:
case i::Token::TRUE_LITERAL:
case i::Token::FALSE_LITERAL:
case i::Token::NUMBER: {
Next();
break;
}
case i::Token::STRING: {
Next();
result = GetStringSymbol();
break;
}
case i::Token::ASSIGN_DIV:
result = ParseRegExpLiteral(true, CHECK_OK);
break;
case i::Token::DIV:
result = ParseRegExpLiteral(false, CHECK_OK);
break;
case i::Token::LBRACK:
result = ParseArrayLiteral(CHECK_OK);
break;
case i::Token::LBRACE:
result = ParseObjectLiteral(CHECK_OK);
break;
case i::Token::LPAREN:
Consume(i::Token::LPAREN);
parenthesized_function_ = (peek() == i::Token::FUNCTION);
result = ParseExpression(true, CHECK_OK);
Expect(i::Token::RPAREN, CHECK_OK);
if (result == kIdentifierExpression) result = kUnknownExpression;
break;
case i::Token::MOD:
result = ParseV8Intrinsic(CHECK_OK);
break;
default: {
Next();
*ok = false;
return kUnknownExpression;
}
}
return result;
}
PreParser::Expression PreParser::ParseArrayLiteral(bool* ok) {
// ArrayLiteral ::
// '[' Expression? (',' Expression?)* ']'
Expect(i::Token::LBRACK, CHECK_OK);
while (peek() != i::Token::RBRACK) {
if (peek() != i::Token::COMMA) {
ParseAssignmentExpression(true, CHECK_OK);
}
if (peek() != i::Token::RBRACK) {
Expect(i::Token::COMMA, CHECK_OK);
}
}
Expect(i::Token::RBRACK, CHECK_OK);
scope_->NextMaterializedLiteralIndex();
return kUnknownExpression;
}
PreParser::Expression PreParser::ParseObjectLiteral(bool* ok) {
// ObjectLiteral ::
// '{' (
// ((IdentifierName | String | Number) ':' AssignmentExpression)
// | (('get' | 'set') (IdentifierName | String | Number) FunctionLiteral)
// )*[','] '}'
Expect(i::Token::LBRACE, CHECK_OK);
while (peek() != i::Token::RBRACE) {
i::Token::Value next = peek();
switch (next) {
case i::Token::IDENTIFIER:
case i::Token::FUTURE_RESERVED_WORD: {
bool is_getter = false;
bool is_setter = false;
ParseIdentifierOrGetOrSet(&is_getter, &is_setter, CHECK_OK);
if ((is_getter || is_setter) && peek() != i::Token::COLON) {
i::Token::Value name = Next();
bool is_keyword = i::Token::IsKeyword(name);
if (name != i::Token::IDENTIFIER &&
name != i::Token::FUTURE_RESERVED_WORD &&
name != i::Token::NUMBER &&
name != i::Token::STRING &&
!is_keyword) {
*ok = false;
return kUnknownExpression;
}
if (!is_keyword) {
LogSymbol();
}
ParseFunctionLiteral(CHECK_OK);
if (peek() != i::Token::RBRACE) {
Expect(i::Token::COMMA, CHECK_OK);
}
continue; // restart the while
}
break;
}
case i::Token::STRING:
Consume(next);
GetStringSymbol();
break;
case i::Token::NUMBER:
Consume(next);
break;
default:
if (i::Token::IsKeyword(next)) {
Consume(next);
} else {
// Unexpected token.
*ok = false;
return kUnknownExpression;
}
}
Expect(i::Token::COLON, CHECK_OK);
ParseAssignmentExpression(true, CHECK_OK);
// TODO(1240767): Consider allowing trailing comma.
if (peek() != i::Token::RBRACE) Expect(i::Token::COMMA, CHECK_OK);
}
Expect(i::Token::RBRACE, CHECK_OK);
scope_->NextMaterializedLiteralIndex();
return kUnknownExpression;
}
PreParser::Expression PreParser::ParseRegExpLiteral(bool seen_equal,
bool* ok) {
if (!scanner_->ScanRegExpPattern(seen_equal)) {
Next();
i::JavaScriptScanner::Location location = scanner_->location();
ReportMessageAt(location.beg_pos, location.end_pos,
"unterminated_regexp", NULL);
*ok = false;
return kUnknownExpression;
}
scope_->NextMaterializedLiteralIndex();
if (!scanner_->ScanRegExpFlags()) {
Next();
i::JavaScriptScanner::Location location = scanner_->location();
ReportMessageAt(location.beg_pos, location.end_pos,
"invalid_regexp_flags", NULL);
*ok = false;
return kUnknownExpression;
}
Next();
return kUnknownExpression;
}
PreParser::Arguments PreParser::ParseArguments(bool* ok) {
// Arguments ::
// '(' (AssignmentExpression)*[','] ')'
Expect(i::Token::LPAREN, CHECK_OK);
bool done = (peek() == i::Token::RPAREN);
int argc = 0;
while (!done) {
ParseAssignmentExpression(true, CHECK_OK);
argc++;
done = (peek() == i::Token::RPAREN);
if (!done) Expect(i::Token::COMMA, CHECK_OK);
}
Expect(i::Token::RPAREN, CHECK_OK);
return argc;
}
PreParser::Expression PreParser::ParseFunctionLiteral(bool* ok) {
// Function ::
// '(' FormalParameterList? ')' '{' FunctionBody '}'
// Parse function body.
ScopeType outer_scope_type = scope_->type();
bool inside_with = scope_->IsInsideWith();
Scope function_scope(&scope_, kFunctionScope);
// FormalParameterList ::
// '(' (Identifier)*[','] ')'
Expect(i::Token::LPAREN, CHECK_OK);
bool done = (peek() == i::Token::RPAREN);
while (!done) {
ParseIdentifier(CHECK_OK);
done = (peek() == i::Token::RPAREN);
if (!done) {
Expect(i::Token::COMMA, CHECK_OK);
}
}
Expect(i::Token::RPAREN, CHECK_OK);
Expect(i::Token::LBRACE, CHECK_OK);
int function_block_pos = scanner_->location().beg_pos;
// Determine if the function will be lazily compiled.
// Currently only happens to top-level functions.
// Optimistically assume that all top-level functions are lazily compiled.
bool is_lazily_compiled = (outer_scope_type == kTopLevelScope &&
!inside_with && allow_lazy_ &&
!parenthesized_function_);
parenthesized_function_ = false;
if (is_lazily_compiled) {
log_->PauseRecording();
ParseSourceElements(i::Token::RBRACE, ok);
log_->ResumeRecording();
if (!*ok) return kUnknownExpression;
Expect(i::Token::RBRACE, CHECK_OK);
// Position right after terminal '}'.
int end_pos = scanner_->location().end_pos;
log_->LogFunction(function_block_pos, end_pos,
function_scope.materialized_literal_count(),
function_scope.expected_properties());
} else {
ParseSourceElements(i::Token::RBRACE, CHECK_OK);
Expect(i::Token::RBRACE, CHECK_OK);
}
return kUnknownExpression;
}
PreParser::Expression PreParser::ParseV8Intrinsic(bool* ok) {
// CallRuntime ::
// '%' Identifier Arguments
Expect(i::Token::MOD, CHECK_OK);
ParseIdentifier(CHECK_OK);
ParseArguments(CHECK_OK);
return kUnknownExpression;
}
void PreParser::ExpectSemicolon(bool* ok) {
// Check for automatic semicolon insertion according to
// the rules given in ECMA-262, section 7.9, page 21.
i::Token::Value tok = peek();
if (tok == i::Token::SEMICOLON) {
Next();
return;
}
if (scanner_->has_line_terminator_before_next() ||
tok == i::Token::RBRACE ||
tok == i::Token::EOS) {
return;
}
Expect(i::Token::SEMICOLON, ok);
}
void PreParser::LogSymbol() {
int identifier_pos = scanner_->location().beg_pos;
if (scanner_->is_literal_ascii()) {
log_->LogAsciiSymbol(identifier_pos, scanner_->literal_ascii_string());
} else {
log_->LogUC16Symbol(identifier_pos, scanner_->literal_uc16_string());
}
}
PreParser::Identifier PreParser::GetIdentifierSymbol() {
LogSymbol();
return kUnknownIdentifier;
}
PreParser::Expression PreParser::GetStringSymbol() {
LogSymbol();
return kUnknownExpression;
}
PreParser::Identifier PreParser::ParseIdentifier(bool* ok) {
if (!Check(i::Token::FUTURE_RESERVED_WORD)) {
Expect(i::Token::IDENTIFIER, ok);
}
if (!*ok) return kUnknownIdentifier;
return GetIdentifierSymbol();
}
PreParser::Identifier PreParser::ParseIdentifierName(bool* ok) {
i::Token::Value next = Next();
if (i::Token::IsKeyword(next)) {
int pos = scanner_->location().beg_pos;
const char* keyword = i::Token::String(next);
log_->LogAsciiSymbol(pos, i::Vector<const char>(keyword,
i::StrLength(keyword)));
return kUnknownExpression;
}
if (next == i::Token::IDENTIFIER ||
next == i::Token::FUTURE_RESERVED_WORD) {
return GetIdentifierSymbol();
}
*ok = false;
return kUnknownIdentifier;
}
// This function reads an identifier and determines whether or not it
// is 'get' or 'set'.
PreParser::Identifier PreParser::ParseIdentifierOrGetOrSet(bool* is_get,
bool* is_set,
bool* ok) {
PreParser::Identifier result = ParseIdentifier(CHECK_OK);
if (scanner_->is_literal_ascii() && scanner_->literal_length() == 3) {
const char* token = scanner_->literal_ascii_string().start();
*is_get = strncmp(token, "get", 3) == 0;
*is_set = !*is_get && strncmp(token, "set", 3) == 0;
}
return result;
}
bool PreParser::peek_any_identifier() {
i::Token::Value next = peek();
return next == i::Token::IDENTIFIER ||
next == i::Token::FUTURE_RESERVED_WORD;
}
#undef CHECK_OK
} } // v8::preparser