1 // Copyright 2012 the V8 project authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4 
5 #include "src/v8.h"
6 
7 #include "src/api.h"
8 #include "src/ast.h"
9 #include "src/bailout-reason.h"
10 #include "src/base/platform/platform.h"
11 #include "src/bootstrapper.h"
12 #include "src/char-predicates-inl.h"
13 #include "src/codegen.h"
14 #include "src/compiler.h"
15 #include "src/messages.h"
16 #include "src/parser.h"
17 #include "src/preparser.h"
18 #include "src/runtime.h"
19 #include "src/scanner-character-streams.h"
20 #include "src/scopeinfo.h"
21 #include "src/string-stream.h"
22 
23 namespace v8 {
24 namespace internal {
25 
RegExpBuilder(Zone * zone)26 RegExpBuilder::RegExpBuilder(Zone* zone)
27     : zone_(zone),
28       pending_empty_(false),
29       characters_(NULL),
30       terms_(),
31       alternatives_()
32 #ifdef DEBUG
33     , last_added_(ADD_NONE)
34 #endif
35   {}
36 
37 
FlushCharacters()38 void RegExpBuilder::FlushCharacters() {
39   pending_empty_ = false;
40   if (characters_ != NULL) {
41     RegExpTree* atom = new(zone()) RegExpAtom(characters_->ToConstVector());
42     characters_ = NULL;
43     text_.Add(atom, zone());
44     LAST(ADD_ATOM);
45   }
46 }
47 
48 
FlushText()49 void RegExpBuilder::FlushText() {
50   FlushCharacters();
51   int num_text = text_.length();
52   if (num_text == 0) {
53     return;
54   } else if (num_text == 1) {
55     terms_.Add(text_.last(), zone());
56   } else {
57     RegExpText* text = new(zone()) RegExpText(zone());
58     for (int i = 0; i < num_text; i++)
59       text_.Get(i)->AppendToText(text, zone());
60     terms_.Add(text, zone());
61   }
62   text_.Clear();
63 }
64 
65 
AddCharacter(uc16 c)66 void RegExpBuilder::AddCharacter(uc16 c) {
67   pending_empty_ = false;
68   if (characters_ == NULL) {
69     characters_ = new(zone()) ZoneList<uc16>(4, zone());
70   }
71   characters_->Add(c, zone());
72   LAST(ADD_CHAR);
73 }
74 
75 
AddEmpty()76 void RegExpBuilder::AddEmpty() {
77   pending_empty_ = true;
78 }
79 
80 
AddAtom(RegExpTree * term)81 void RegExpBuilder::AddAtom(RegExpTree* term) {
82   if (term->IsEmpty()) {
83     AddEmpty();
84     return;
85   }
86   if (term->IsTextElement()) {
87     FlushCharacters();
88     text_.Add(term, zone());
89   } else {
90     FlushText();
91     terms_.Add(term, zone());
92   }
93   LAST(ADD_ATOM);
94 }
95 
96 
AddAssertion(RegExpTree * assert)97 void RegExpBuilder::AddAssertion(RegExpTree* assert) {
98   FlushText();
99   terms_.Add(assert, zone());
100   LAST(ADD_ASSERT);
101 }
102 
103 
NewAlternative()104 void RegExpBuilder::NewAlternative() {
105   FlushTerms();
106 }
107 
108 
FlushTerms()109 void RegExpBuilder::FlushTerms() {
110   FlushText();
111   int num_terms = terms_.length();
112   RegExpTree* alternative;
113   if (num_terms == 0) {
114     alternative = RegExpEmpty::GetInstance();
115   } else if (num_terms == 1) {
116     alternative = terms_.last();
117   } else {
118     alternative = new(zone()) RegExpAlternative(terms_.GetList(zone()));
119   }
120   alternatives_.Add(alternative, zone());
121   terms_.Clear();
122   LAST(ADD_NONE);
123 }
124 
125 
ToRegExp()126 RegExpTree* RegExpBuilder::ToRegExp() {
127   FlushTerms();
128   int num_alternatives = alternatives_.length();
129   if (num_alternatives == 0) {
130     return RegExpEmpty::GetInstance();
131   }
132   if (num_alternatives == 1) {
133     return alternatives_.last();
134   }
135   return new(zone()) RegExpDisjunction(alternatives_.GetList(zone()));
136 }
137 
138 
AddQuantifierToAtom(int min,int max,RegExpQuantifier::QuantifierType quantifier_type)139 void RegExpBuilder::AddQuantifierToAtom(
140     int min, int max, RegExpQuantifier::QuantifierType quantifier_type) {
141   if (pending_empty_) {
142     pending_empty_ = false;
143     return;
144   }
145   RegExpTree* atom;
146   if (characters_ != NULL) {
147     DCHECK(last_added_ == ADD_CHAR);
148     // Last atom was character.
149     Vector<const uc16> char_vector = characters_->ToConstVector();
150     int num_chars = char_vector.length();
151     if (num_chars > 1) {
152       Vector<const uc16> prefix = char_vector.SubVector(0, num_chars - 1);
153       text_.Add(new(zone()) RegExpAtom(prefix), zone());
154       char_vector = char_vector.SubVector(num_chars - 1, num_chars);
155     }
156     characters_ = NULL;
157     atom = new(zone()) RegExpAtom(char_vector);
158     FlushText();
159   } else if (text_.length() > 0) {
160     DCHECK(last_added_ == ADD_ATOM);
161     atom = text_.RemoveLast();
162     FlushText();
163   } else if (terms_.length() > 0) {
164     DCHECK(last_added_ == ADD_ATOM);
165     atom = terms_.RemoveLast();
166     if (atom->max_match() == 0) {
167       // Guaranteed to only match an empty string.
168       LAST(ADD_TERM);
169       if (min == 0) {
170         return;
171       }
172       terms_.Add(atom, zone());
173       return;
174     }
175   } else {
176     // Only call immediately after adding an atom or character!
177     UNREACHABLE();
178     return;
179   }
180   terms_.Add(
181       new(zone()) RegExpQuantifier(min, max, quantifier_type, atom), zone());
182   LAST(ADD_TERM);
183 }
184 
185 
GetFunctionEntry(int start)186 FunctionEntry ParseData::GetFunctionEntry(int start) {
187   // The current pre-data entry must be a FunctionEntry with the given
188   // start position.
189   if ((function_index_ + FunctionEntry::kSize <= Length()) &&
190       (static_cast<int>(Data()[function_index_]) == start)) {
191     int index = function_index_;
192     function_index_ += FunctionEntry::kSize;
193     Vector<unsigned> subvector(&(Data()[index]), FunctionEntry::kSize);
194     return FunctionEntry(subvector);
195   }
196   return FunctionEntry();
197 }
198 
199 
FunctionCount()200 int ParseData::FunctionCount() {
201   int functions_size = FunctionsSize();
202   if (functions_size < 0) return 0;
203   if (functions_size % FunctionEntry::kSize != 0) return 0;
204   return functions_size / FunctionEntry::kSize;
205 }
206 
207 
IsSane()208 bool ParseData::IsSane() {
209   // Check that the header data is valid and doesn't specify
210   // point to positions outside the store.
211   int data_length = Length();
212   if (data_length < PreparseDataConstants::kHeaderSize) return false;
213   if (Magic() != PreparseDataConstants::kMagicNumber) return false;
214   if (Version() != PreparseDataConstants::kCurrentVersion) return false;
215   if (HasError()) return false;
216   // Check that the space allocated for function entries is sane.
217   int functions_size = FunctionsSize();
218   if (functions_size < 0) return false;
219   if (functions_size % FunctionEntry::kSize != 0) return false;
220   // Check that the total size has room for header and function entries.
221   int minimum_size =
222       PreparseDataConstants::kHeaderSize + functions_size;
223   if (data_length < minimum_size) return false;
224   return true;
225 }
226 
227 
Initialize()228 void ParseData::Initialize() {
229   // Prepares state for use.
230   int data_length = Length();
231   if (data_length >= PreparseDataConstants::kHeaderSize) {
232     function_index_ = PreparseDataConstants::kHeaderSize;
233   }
234 }
235 
236 
HasError()237 bool ParseData::HasError() {
238   return Data()[PreparseDataConstants::kHasErrorOffset];
239 }
240 
241 
Magic()242 unsigned ParseData::Magic() {
243   return Data()[PreparseDataConstants::kMagicOffset];
244 }
245 
246 
Version()247 unsigned ParseData::Version() {
248   return Data()[PreparseDataConstants::kVersionOffset];
249 }
250 
251 
FunctionsSize()252 int ParseData::FunctionsSize() {
253   return static_cast<int>(Data()[PreparseDataConstants::kFunctionsSizeOffset]);
254 }
255 
256 
SetCachedData()257 void Parser::SetCachedData() {
258   if (compile_options() == ScriptCompiler::kNoCompileOptions) {
259     cached_parse_data_ = NULL;
260   } else {
261     DCHECK(info_->cached_data() != NULL);
262     if (compile_options() == ScriptCompiler::kConsumeParserCache) {
263       cached_parse_data_ = new ParseData(*info_->cached_data());
264     }
265   }
266 }
267 
268 
NewScope(Scope * parent,ScopeType scope_type)269 Scope* Parser::NewScope(Scope* parent, ScopeType scope_type) {
270   DCHECK(ast_value_factory());
271   Scope* result =
272       new (zone()) Scope(parent, scope_type, ast_value_factory(), zone());
273   result->Initialize();
274   return result;
275 }
276 
277 
278 // ----------------------------------------------------------------------------
279 // Target is a support class to facilitate manipulation of the
280 // Parser's target_stack_ (the stack of potential 'break' and
281 // 'continue' statement targets). Upon construction, a new target is
282 // added; it is removed upon destruction.
283 
284 class Target BASE_EMBEDDED {
285  public:
Target(Target ** variable,AstNode * node)286   Target(Target** variable, AstNode* node)
287       : variable_(variable), node_(node), previous_(*variable) {
288     *variable = this;
289   }
290 
~Target()291   ~Target() {
292     *variable_ = previous_;
293   }
294 
previous()295   Target* previous() { return previous_; }
node()296   AstNode* node() { return node_; }
297 
298  private:
299   Target** variable_;
300   AstNode* node_;
301   Target* previous_;
302 };
303 
304 
305 class TargetScope BASE_EMBEDDED {
306  public:
TargetScope(Target ** variable)307   explicit TargetScope(Target** variable)
308       : variable_(variable), previous_(*variable) {
309     *variable = NULL;
310   }
311 
~TargetScope()312   ~TargetScope() {
313     *variable_ = previous_;
314   }
315 
316  private:
317   Target** variable_;
318   Target* previous_;
319 };
320 
321 
322 // ----------------------------------------------------------------------------
323 // The CHECK_OK macro is a convenient macro to enforce error
324 // handling for functions that may fail (by returning !*ok).
325 //
326 // CAUTION: This macro appends extra statements after a call,
327 // thus it must never be used where only a single statement
328 // is correct (e.g. an if statement branch w/o braces)!
329 
330 #define CHECK_OK  ok);   \
331   if (!*ok) return NULL; \
332   ((void)0
333 #define DUMMY )  // to make indentation work
334 #undef DUMMY
335 
336 #define CHECK_FAILED  /**/);   \
337   if (failed_) return NULL; \
338   ((void)0
339 #define DUMMY )  // to make indentation work
340 #undef DUMMY
341 
342 // ----------------------------------------------------------------------------
343 // Implementation of Parser
344 
345 class ParserTraits::Checkpoint
346     : public ParserBase<ParserTraits>::CheckpointBase {
347  public:
Checkpoint(ParserBase<ParserTraits> * parser)348   explicit Checkpoint(ParserBase<ParserTraits>* parser)
349       : CheckpointBase(parser), parser_(parser) {
350     saved_ast_node_id_gen_ = *parser_->ast_node_id_gen_;
351   }
352 
Restore()353   void Restore() {
354     CheckpointBase::Restore();
355     *parser_->ast_node_id_gen_ = saved_ast_node_id_gen_;
356   }
357 
358  private:
359   ParserBase<ParserTraits>* parser_;
360   AstNode::IdGen saved_ast_node_id_gen_;
361 };
362 
363 
IsEvalOrArguments(const AstRawString * identifier) const364 bool ParserTraits::IsEvalOrArguments(const AstRawString* identifier) const {
365   return identifier == parser_->ast_value_factory()->eval_string() ||
366          identifier == parser_->ast_value_factory()->arguments_string();
367 }
368 
369 
IsPrototype(const AstRawString * identifier) const370 bool ParserTraits::IsPrototype(const AstRawString* identifier) const {
371   return identifier == parser_->ast_value_factory()->prototype_string();
372 }
373 
374 
IsConstructor(const AstRawString * identifier) const375 bool ParserTraits::IsConstructor(const AstRawString* identifier) const {
376   return identifier == parser_->ast_value_factory()->constructor_string();
377 }
378 
379 
IsThisProperty(Expression * expression)380 bool ParserTraits::IsThisProperty(Expression* expression) {
381   DCHECK(expression != NULL);
382   Property* property = expression->AsProperty();
383   return property != NULL &&
384       property->obj()->AsVariableProxy() != NULL &&
385       property->obj()->AsVariableProxy()->is_this();
386 }
387 
388 
IsIdentifier(Expression * expression)389 bool ParserTraits::IsIdentifier(Expression* expression) {
390   VariableProxy* operand = expression->AsVariableProxy();
391   return operand != NULL && !operand->is_this();
392 }
393 
394 
PushPropertyName(FuncNameInferrer * fni,Expression * expression)395 void ParserTraits::PushPropertyName(FuncNameInferrer* fni,
396                                     Expression* expression) {
397   if (expression->IsPropertyName()) {
398     fni->PushLiteralName(expression->AsLiteral()->AsRawPropertyName());
399   } else {
400     fni->PushLiteralName(
401         parser_->ast_value_factory()->anonymous_function_string());
402   }
403 }
404 
405 
CheckAssigningFunctionLiteralToProperty(Expression * left,Expression * right)406 void ParserTraits::CheckAssigningFunctionLiteralToProperty(Expression* left,
407                                                            Expression* right) {
408   DCHECK(left != NULL);
409   if (left->AsProperty() != NULL &&
410       right->AsFunctionLiteral() != NULL) {
411     right->AsFunctionLiteral()->set_pretenure();
412   }
413 }
414 
415 
CheckPossibleEvalCall(Expression * expression,Scope * scope)416 void ParserTraits::CheckPossibleEvalCall(Expression* expression,
417                                          Scope* scope) {
418   VariableProxy* callee = expression->AsVariableProxy();
419   if (callee != NULL &&
420       callee->raw_name() == parser_->ast_value_factory()->eval_string()) {
421     scope->DeclarationScope()->RecordEvalCall();
422   }
423 }
424 
425 
MarkExpressionAsAssigned(Expression * expression)426 Expression* ParserTraits::MarkExpressionAsAssigned(Expression* expression) {
427   VariableProxy* proxy =
428       expression != NULL ? expression->AsVariableProxy() : NULL;
429   if (proxy != NULL) proxy->set_is_assigned();
430   return expression;
431 }
432 
433 
ShortcutNumericLiteralBinaryExpression(Expression ** x,Expression * y,Token::Value op,int pos,AstNodeFactory<AstConstructionVisitor> * factory)434 bool ParserTraits::ShortcutNumericLiteralBinaryExpression(
435     Expression** x, Expression* y, Token::Value op, int pos,
436     AstNodeFactory<AstConstructionVisitor>* factory) {
437   if ((*x)->AsLiteral() && (*x)->AsLiteral()->raw_value()->IsNumber() &&
438       y->AsLiteral() && y->AsLiteral()->raw_value()->IsNumber()) {
439     double x_val = (*x)->AsLiteral()->raw_value()->AsNumber();
440     double y_val = y->AsLiteral()->raw_value()->AsNumber();
441     switch (op) {
442       case Token::ADD:
443         *x = factory->NewNumberLiteral(x_val + y_val, pos);
444         return true;
445       case Token::SUB:
446         *x = factory->NewNumberLiteral(x_val - y_val, pos);
447         return true;
448       case Token::MUL:
449         *x = factory->NewNumberLiteral(x_val * y_val, pos);
450         return true;
451       case Token::DIV:
452         *x = factory->NewNumberLiteral(x_val / y_val, pos);
453         return true;
454       case Token::BIT_OR: {
455         int value = DoubleToInt32(x_val) | DoubleToInt32(y_val);
456         *x = factory->NewNumberLiteral(value, pos);
457         return true;
458       }
459       case Token::BIT_AND: {
460         int value = DoubleToInt32(x_val) & DoubleToInt32(y_val);
461         *x = factory->NewNumberLiteral(value, pos);
462         return true;
463       }
464       case Token::BIT_XOR: {
465         int value = DoubleToInt32(x_val) ^ DoubleToInt32(y_val);
466         *x = factory->NewNumberLiteral(value, pos);
467         return true;
468       }
469       case Token::SHL: {
470         int value = DoubleToInt32(x_val) << (DoubleToInt32(y_val) & 0x1f);
471         *x = factory->NewNumberLiteral(value, pos);
472         return true;
473       }
474       case Token::SHR: {
475         uint32_t shift = DoubleToInt32(y_val) & 0x1f;
476         uint32_t value = DoubleToUint32(x_val) >> shift;
477         *x = factory->NewNumberLiteral(value, pos);
478         return true;
479       }
480       case Token::SAR: {
481         uint32_t shift = DoubleToInt32(y_val) & 0x1f;
482         int value = ArithmeticShiftRight(DoubleToInt32(x_val), shift);
483         *x = factory->NewNumberLiteral(value, pos);
484         return true;
485       }
486       default:
487         break;
488     }
489   }
490   return false;
491 }
492 
493 
BuildUnaryExpression(Expression * expression,Token::Value op,int pos,AstNodeFactory<AstConstructionVisitor> * factory)494 Expression* ParserTraits::BuildUnaryExpression(
495     Expression* expression, Token::Value op, int pos,
496     AstNodeFactory<AstConstructionVisitor>* factory) {
497   DCHECK(expression != NULL);
498   if (expression->IsLiteral()) {
499     const AstValue* literal = expression->AsLiteral()->raw_value();
500     if (op == Token::NOT) {
501       // Convert the literal to a boolean condition and negate it.
502       bool condition = literal->BooleanValue();
503       return factory->NewBooleanLiteral(!condition, pos);
504     } else if (literal->IsNumber()) {
505       // Compute some expressions involving only number literals.
506       double value = literal->AsNumber();
507       switch (op) {
508         case Token::ADD:
509           return expression;
510         case Token::SUB:
511           return factory->NewNumberLiteral(-value, pos);
512         case Token::BIT_NOT:
513           return factory->NewNumberLiteral(~DoubleToInt32(value), pos);
514         default:
515           break;
516       }
517     }
518   }
519   // Desugar '+foo' => 'foo*1'
520   if (op == Token::ADD) {
521     return factory->NewBinaryOperation(
522         Token::MUL, expression, factory->NewNumberLiteral(1, pos), pos);
523   }
524   // The same idea for '-foo' => 'foo*(-1)'.
525   if (op == Token::SUB) {
526     return factory->NewBinaryOperation(
527         Token::MUL, expression, factory->NewNumberLiteral(-1, pos), pos);
528   }
529   // ...and one more time for '~foo' => 'foo^(~0)'.
530   if (op == Token::BIT_NOT) {
531     return factory->NewBinaryOperation(
532         Token::BIT_XOR, expression, factory->NewNumberLiteral(~0, pos), pos);
533   }
534   return factory->NewUnaryOperation(op, expression, pos);
535 }
536 
537 
NewThrowReferenceError(const char * message,int pos)538 Expression* ParserTraits::NewThrowReferenceError(const char* message, int pos) {
539   return NewThrowError(
540       parser_->ast_value_factory()->make_reference_error_string(), message,
541       NULL, pos);
542 }
543 
544 
NewThrowSyntaxError(const char * message,const AstRawString * arg,int pos)545 Expression* ParserTraits::NewThrowSyntaxError(
546     const char* message, const AstRawString* arg, int pos) {
547   return NewThrowError(parser_->ast_value_factory()->make_syntax_error_string(),
548                        message, arg, pos);
549 }
550 
551 
NewThrowTypeError(const char * message,const AstRawString * arg,int pos)552 Expression* ParserTraits::NewThrowTypeError(
553     const char* message, const AstRawString* arg, int pos) {
554   return NewThrowError(parser_->ast_value_factory()->make_type_error_string(),
555                        message, arg, pos);
556 }
557 
558 
NewThrowError(const AstRawString * constructor,const char * message,const AstRawString * arg,int pos)559 Expression* ParserTraits::NewThrowError(
560     const AstRawString* constructor, const char* message,
561     const AstRawString* arg, int pos) {
562   Zone* zone = parser_->zone();
563   int argc = arg != NULL ? 1 : 0;
564   const AstRawString* type =
565       parser_->ast_value_factory()->GetOneByteString(message);
566   ZoneList<const AstRawString*>* array =
567       new (zone) ZoneList<const AstRawString*>(argc, zone);
568   if (arg != NULL) {
569     array->Add(arg, zone);
570   }
571   ZoneList<Expression*>* args = new (zone) ZoneList<Expression*>(2, zone);
572   args->Add(parser_->factory()->NewStringLiteral(type, pos), zone);
573   args->Add(parser_->factory()->NewStringListLiteral(array, pos), zone);
574   CallRuntime* call_constructor =
575       parser_->factory()->NewCallRuntime(constructor, NULL, args, pos);
576   return parser_->factory()->NewThrow(call_constructor, pos);
577 }
578 
579 
ReportMessageAt(Scanner::Location source_location,const char * message,const char * arg,bool is_reference_error)580 void ParserTraits::ReportMessageAt(Scanner::Location source_location,
581                                    const char* message,
582                                    const char* arg,
583                                    bool is_reference_error) {
584   if (parser_->stack_overflow()) {
585     // Suppress the error message (syntax error or such) in the presence of a
586     // stack overflow. The isolate allows only one pending exception at at time
587     // and we want to report the stack overflow later.
588     return;
589   }
590   parser_->has_pending_error_ = true;
591   parser_->pending_error_location_ = source_location;
592   parser_->pending_error_message_ = message;
593   parser_->pending_error_char_arg_ = arg;
594   parser_->pending_error_arg_ = NULL;
595   parser_->pending_error_is_reference_error_ = is_reference_error;
596 }
597 
598 
ReportMessage(const char * message,const char * arg,bool is_reference_error)599 void ParserTraits::ReportMessage(const char* message,
600                                  const char* arg,
601                                  bool is_reference_error) {
602   Scanner::Location source_location = parser_->scanner()->location();
603   ReportMessageAt(source_location, message, arg, is_reference_error);
604 }
605 
606 
ReportMessage(const char * message,const AstRawString * arg,bool is_reference_error)607 void ParserTraits::ReportMessage(const char* message,
608                                  const AstRawString* arg,
609                                  bool is_reference_error) {
610   Scanner::Location source_location = parser_->scanner()->location();
611   ReportMessageAt(source_location, message, arg, is_reference_error);
612 }
613 
614 
ReportMessageAt(Scanner::Location source_location,const char * message,const AstRawString * arg,bool is_reference_error)615 void ParserTraits::ReportMessageAt(Scanner::Location source_location,
616                                    const char* message,
617                                    const AstRawString* arg,
618                                    bool is_reference_error) {
619   if (parser_->stack_overflow()) {
620     // Suppress the error message (syntax error or such) in the presence of a
621     // stack overflow. The isolate allows only one pending exception at at time
622     // and we want to report the stack overflow later.
623     return;
624   }
625   parser_->has_pending_error_ = true;
626   parser_->pending_error_location_ = source_location;
627   parser_->pending_error_message_ = message;
628   parser_->pending_error_char_arg_ = NULL;
629   parser_->pending_error_arg_ = arg;
630   parser_->pending_error_is_reference_error_ = is_reference_error;
631 }
632 
633 
GetSymbol(Scanner * scanner)634 const AstRawString* ParserTraits::GetSymbol(Scanner* scanner) {
635   const AstRawString* result =
636       parser_->scanner()->CurrentSymbol(parser_->ast_value_factory());
637   DCHECK(result != NULL);
638   return result;
639 }
640 
641 
GetNumberAsSymbol(Scanner * scanner)642 const AstRawString* ParserTraits::GetNumberAsSymbol(Scanner* scanner) {
643   double double_value = parser_->scanner()->DoubleValue();
644   char array[100];
645   const char* string =
646       DoubleToCString(double_value, Vector<char>(array, arraysize(array)));
647   return ast_value_factory()->GetOneByteString(string);
648 }
649 
650 
GetNextSymbol(Scanner * scanner)651 const AstRawString* ParserTraits::GetNextSymbol(Scanner* scanner) {
652   return parser_->scanner()->NextSymbol(parser_->ast_value_factory());
653 }
654 
655 
ThisExpression(Scope * scope,AstNodeFactory<AstConstructionVisitor> * factory,int pos)656 Expression* ParserTraits::ThisExpression(
657     Scope* scope, AstNodeFactory<AstConstructionVisitor>* factory, int pos) {
658   return factory->NewVariableProxy(scope->receiver(), pos);
659 }
660 
SuperReference(Scope * scope,AstNodeFactory<AstConstructionVisitor> * factory,int pos)661 Expression* ParserTraits::SuperReference(
662     Scope* scope, AstNodeFactory<AstConstructionVisitor>* factory, int pos) {
663   return factory->NewSuperReference(
664       ThisExpression(scope, factory, pos)->AsVariableProxy(),
665       pos);
666 }
667 
ClassLiteral(const AstRawString * name,Expression * extends,Expression * constructor,ZoneList<ObjectLiteral::Property * > * properties,int pos,AstNodeFactory<AstConstructionVisitor> * factory)668 Expression* ParserTraits::ClassLiteral(
669     const AstRawString* name, Expression* extends, Expression* constructor,
670     ZoneList<ObjectLiteral::Property*>* properties, int pos,
671     AstNodeFactory<AstConstructionVisitor>* factory) {
672   return factory->NewClassLiteral(name, extends, constructor, properties, pos);
673 }
674 
ExpressionFromLiteral(Token::Value token,int pos,Scanner * scanner,AstNodeFactory<AstConstructionVisitor> * factory)675 Literal* ParserTraits::ExpressionFromLiteral(
676     Token::Value token, int pos,
677     Scanner* scanner,
678     AstNodeFactory<AstConstructionVisitor>* factory) {
679   switch (token) {
680     case Token::NULL_LITERAL:
681       return factory->NewNullLiteral(pos);
682     case Token::TRUE_LITERAL:
683       return factory->NewBooleanLiteral(true, pos);
684     case Token::FALSE_LITERAL:
685       return factory->NewBooleanLiteral(false, pos);
686     case Token::NUMBER: {
687       double value = scanner->DoubleValue();
688       return factory->NewNumberLiteral(value, pos);
689     }
690     default:
691       DCHECK(false);
692   }
693   return NULL;
694 }
695 
696 
ExpressionFromIdentifier(const AstRawString * name,int pos,Scope * scope,AstNodeFactory<AstConstructionVisitor> * factory)697 Expression* ParserTraits::ExpressionFromIdentifier(
698     const AstRawString* name, int pos, Scope* scope,
699     AstNodeFactory<AstConstructionVisitor>* factory) {
700   if (parser_->fni_ != NULL) parser_->fni_->PushVariableName(name);
701   // The name may refer to a module instance object, so its type is unknown.
702 #ifdef DEBUG
703   if (FLAG_print_interface_details)
704     PrintF("# Variable %.*s ", name->length(), name->raw_data());
705 #endif
706   Interface* interface = Interface::NewUnknown(parser_->zone());
707   return scope->NewUnresolved(factory, name, interface, pos);
708 }
709 
710 
ExpressionFromString(int pos,Scanner * scanner,AstNodeFactory<AstConstructionVisitor> * factory)711 Expression* ParserTraits::ExpressionFromString(
712     int pos, Scanner* scanner,
713     AstNodeFactory<AstConstructionVisitor>* factory) {
714   const AstRawString* symbol = GetSymbol(scanner);
715   if (parser_->fni_ != NULL) parser_->fni_->PushLiteralName(symbol);
716   return factory->NewStringLiteral(symbol, pos);
717 }
718 
719 
GetIterator(Expression * iterable,AstNodeFactory<AstConstructionVisitor> * factory)720 Expression* ParserTraits::GetIterator(
721     Expression* iterable, AstNodeFactory<AstConstructionVisitor>* factory) {
722   Expression* iterator_symbol_literal =
723       factory->NewSymbolLiteral("symbolIterator", RelocInfo::kNoPosition);
724   int pos = iterable->position();
725   Expression* prop =
726       factory->NewProperty(iterable, iterator_symbol_literal, pos);
727   Zone* zone = parser_->zone();
728   ZoneList<Expression*>* args = new (zone) ZoneList<Expression*>(0, zone);
729   return factory->NewCall(prop, args, pos);
730 }
731 
732 
GetLiteralTheHole(int position,AstNodeFactory<AstConstructionVisitor> * factory)733 Literal* ParserTraits::GetLiteralTheHole(
734     int position, AstNodeFactory<AstConstructionVisitor>* factory) {
735   return factory->NewTheHoleLiteral(RelocInfo::kNoPosition);
736 }
737 
738 
ParseV8Intrinsic(bool * ok)739 Expression* ParserTraits::ParseV8Intrinsic(bool* ok) {
740   return parser_->ParseV8Intrinsic(ok);
741 }
742 
743 
ParseFunctionLiteral(const AstRawString * name,Scanner::Location function_name_location,bool name_is_strict_reserved,FunctionKind kind,int function_token_position,FunctionLiteral::FunctionType type,FunctionLiteral::ArityRestriction arity_restriction,bool * ok)744 FunctionLiteral* ParserTraits::ParseFunctionLiteral(
745     const AstRawString* name, Scanner::Location function_name_location,
746     bool name_is_strict_reserved, FunctionKind kind,
747     int function_token_position, FunctionLiteral::FunctionType type,
748     FunctionLiteral::ArityRestriction arity_restriction, bool* ok) {
749   return parser_->ParseFunctionLiteral(
750       name, function_name_location, name_is_strict_reserved, kind,
751       function_token_position, type, arity_restriction, ok);
752 }
753 
754 
Parser(CompilationInfo * info,ParseInfo * parse_info)755 Parser::Parser(CompilationInfo* info, ParseInfo* parse_info)
756     : ParserBase<ParserTraits>(&scanner_, parse_info->stack_limit,
757                                info->extension(), NULL, info->zone(),
758                                info->ast_node_id_gen(), this),
759       scanner_(parse_info->unicode_cache),
760       reusable_preparser_(NULL),
761       original_scope_(NULL),
762       target_stack_(NULL),
763       cached_parse_data_(NULL),
764       info_(info),
765       has_pending_error_(false),
766       pending_error_message_(NULL),
767       pending_error_arg_(NULL),
768       pending_error_char_arg_(NULL),
769       total_preparse_skipped_(0),
770       pre_parse_timer_(NULL) {
771   DCHECK(!script().is_null() || info->source_stream() != NULL);
772   set_allow_harmony_scoping(!info->is_native() && FLAG_harmony_scoping);
773   set_allow_modules(!info->is_native() && FLAG_harmony_modules);
774   set_allow_natives_syntax(FLAG_allow_natives_syntax || info->is_native());
775   set_allow_lazy(false);  // Must be explicitly enabled.
776   set_allow_arrow_functions(FLAG_harmony_arrow_functions);
777   set_allow_harmony_numeric_literals(FLAG_harmony_numeric_literals);
778   set_allow_classes(FLAG_harmony_classes);
779   set_allow_harmony_object_literals(FLAG_harmony_object_literals);
780   for (int feature = 0; feature < v8::Isolate::kUseCounterFeatureCount;
781        ++feature) {
782     use_counts_[feature] = 0;
783   }
784   if (info->ast_value_factory() == NULL) {
785     // info takes ownership of AstValueFactory.
786     info->SetAstValueFactory(
787         new AstValueFactory(zone(), parse_info->hash_seed));
788   }
789 }
790 
791 
ParseProgram()792 FunctionLiteral* Parser::ParseProgram() {
793   // TODO(bmeurer): We temporarily need to pass allow_nesting = true here,
794   // see comment for HistogramTimerScope class.
795 
796   // It's OK to use the counters here, since this function is only called in
797   // the main thread.
798   HistogramTimerScope timer_scope(isolate()->counters()->parse(), true);
799   Handle<String> source(String::cast(script()->source()));
800   isolate()->counters()->total_parse_size()->Increment(source->length());
801   base::ElapsedTimer timer;
802   if (FLAG_trace_parse) {
803     timer.Start();
804   }
805   fni_ = new (zone()) FuncNameInferrer(ast_value_factory(), zone());
806 
807   // Initialize parser state.
808   CompleteParserRecorder recorder;
809 
810   if (compile_options() == ScriptCompiler::kProduceParserCache) {
811     log_ = &recorder;
812   } else if (compile_options() == ScriptCompiler::kConsumeParserCache) {
813     cached_parse_data_->Initialize();
814   }
815 
816   source = String::Flatten(source);
817   FunctionLiteral* result;
818 
819   Scope* top_scope = NULL;
820   Scope* eval_scope = NULL;
821   if (source->IsExternalTwoByteString()) {
822     // Notice that the stream is destroyed at the end of the branch block.
823     // The last line of the blocks can't be moved outside, even though they're
824     // identical calls.
825     ExternalTwoByteStringUtf16CharacterStream stream(
826         Handle<ExternalTwoByteString>::cast(source), 0, source->length());
827     scanner_.Initialize(&stream);
828     result = DoParseProgram(info(), &top_scope, &eval_scope);
829   } else {
830     GenericStringUtf16CharacterStream stream(source, 0, source->length());
831     scanner_.Initialize(&stream);
832     result = DoParseProgram(info(), &top_scope, &eval_scope);
833   }
834   top_scope->set_end_position(source->length());
835   if (eval_scope != NULL) {
836     eval_scope->set_end_position(source->length());
837   }
838   HandleSourceURLComments();
839 
840   if (FLAG_trace_parse && result != NULL) {
841     double ms = timer.Elapsed().InMillisecondsF();
842     if (info()->is_eval()) {
843       PrintF("[parsing eval");
844     } else if (info()->script()->name()->IsString()) {
845       String* name = String::cast(info()->script()->name());
846       SmartArrayPointer<char> name_chars = name->ToCString();
847       PrintF("[parsing script: %s", name_chars.get());
848     } else {
849       PrintF("[parsing script");
850     }
851     PrintF(" - took %0.3f ms]\n", ms);
852   }
853   if (compile_options() == ScriptCompiler::kProduceParserCache) {
854     if (result != NULL) *info_->cached_data() = recorder.GetScriptData();
855     log_ = NULL;
856   }
857   return result;
858 }
859 
860 
DoParseProgram(CompilationInfo * info,Scope ** scope,Scope ** eval_scope)861 FunctionLiteral* Parser::DoParseProgram(CompilationInfo* info, Scope** scope,
862                                         Scope** eval_scope) {
863   DCHECK(scope_ == NULL);
864   DCHECK(target_stack_ == NULL);
865 
866   FunctionLiteral* result = NULL;
867   {
868     *scope = NewScope(scope_, GLOBAL_SCOPE);
869     info->SetGlobalScope(*scope);
870     if (!info->context().is_null() && !info->context()->IsNativeContext()) {
871       *scope = Scope::DeserializeScopeChain(*info->context(), *scope, zone());
872       // The Scope is backed up by ScopeInfo (which is in the V8 heap); this
873       // means the Parser cannot operate independent of the V8 heap. Tell the
874       // string table to internalize strings and values right after they're
875       // created.
876       ast_value_factory()->Internalize(isolate());
877     }
878     original_scope_ = *scope;
879     if (info->is_eval()) {
880       if (!(*scope)->is_global_scope() || info->strict_mode() == STRICT) {
881         *scope = NewScope(*scope, EVAL_SCOPE);
882       }
883     } else if (info->is_global()) {
884       *scope = NewScope(*scope, GLOBAL_SCOPE);
885     }
886     (*scope)->set_start_position(0);
887     // End position will be set by the caller.
888 
889     // Compute the parsing mode.
890     Mode mode = (FLAG_lazy && allow_lazy()) ? PARSE_LAZILY : PARSE_EAGERLY;
891     if (allow_natives_syntax() || extension_ != NULL ||
892         (*scope)->is_eval_scope()) {
893       mode = PARSE_EAGERLY;
894     }
895     ParsingModeScope parsing_mode(this, mode);
896 
897     // Enters 'scope'.
898     FunctionState function_state(&function_state_, &scope_, *scope, zone(),
899                                  ast_value_factory(), info->ast_node_id_gen());
900 
901     scope_->SetStrictMode(info->strict_mode());
902     ZoneList<Statement*>* body = new(zone()) ZoneList<Statement*>(16, zone());
903     bool ok = true;
904     int beg_pos = scanner()->location().beg_pos;
905     ParseSourceElements(body, Token::EOS, info->is_eval(), true, eval_scope,
906                         &ok);
907 
908     if (ok && strict_mode() == STRICT) {
909       CheckOctalLiteral(beg_pos, scanner()->location().end_pos, &ok);
910     }
911 
912     if (ok && allow_harmony_scoping() && strict_mode() == STRICT) {
913       CheckConflictingVarDeclarations(scope_, &ok);
914     }
915 
916     if (ok && info->parse_restriction() == ONLY_SINGLE_FUNCTION_LITERAL) {
917       if (body->length() != 1 ||
918           !body->at(0)->IsExpressionStatement() ||
919           !body->at(0)->AsExpressionStatement()->
920               expression()->IsFunctionLiteral()) {
921         ReportMessage("single_function_literal");
922         ok = false;
923       }
924     }
925 
926     if (ok) {
927       result = factory()->NewFunctionLiteral(
928           ast_value_factory()->empty_string(), ast_value_factory(), scope_,
929           body, function_state.materialized_literal_count(),
930           function_state.expected_property_count(),
931           function_state.handler_count(), 0,
932           FunctionLiteral::kNoDuplicateParameters,
933           FunctionLiteral::ANONYMOUS_EXPRESSION, FunctionLiteral::kGlobalOrEval,
934           FunctionLiteral::kNotParenthesized, FunctionKind::kNormalFunction, 0);
935       result->set_ast_properties(factory()->visitor()->ast_properties());
936       result->set_dont_optimize_reason(
937           factory()->visitor()->dont_optimize_reason());
938     }
939   }
940 
941   // Make sure the target stack is empty.
942   DCHECK(target_stack_ == NULL);
943 
944   return result;
945 }
946 
947 
ParseLazy()948 FunctionLiteral* Parser::ParseLazy() {
949   // It's OK to use the counters here, since this function is only called in
950   // the main thread.
951   HistogramTimerScope timer_scope(isolate()->counters()->parse_lazy());
952   Handle<String> source(String::cast(script()->source()));
953   isolate()->counters()->total_parse_size()->Increment(source->length());
954   base::ElapsedTimer timer;
955   if (FLAG_trace_parse) {
956     timer.Start();
957   }
958   Handle<SharedFunctionInfo> shared_info = info()->shared_info();
959 
960   // Initialize parser state.
961   source = String::Flatten(source);
962   FunctionLiteral* result;
963   if (source->IsExternalTwoByteString()) {
964     ExternalTwoByteStringUtf16CharacterStream stream(
965         Handle<ExternalTwoByteString>::cast(source),
966         shared_info->start_position(),
967         shared_info->end_position());
968     result = ParseLazy(&stream);
969   } else {
970     GenericStringUtf16CharacterStream stream(source,
971                                              shared_info->start_position(),
972                                              shared_info->end_position());
973     result = ParseLazy(&stream);
974   }
975 
976   if (FLAG_trace_parse && result != NULL) {
977     double ms = timer.Elapsed().InMillisecondsF();
978     SmartArrayPointer<char> name_chars = result->debug_name()->ToCString();
979     PrintF("[parsing function: %s - took %0.3f ms]\n", name_chars.get(), ms);
980   }
981   return result;
982 }
983 
984 
ParseLazy(Utf16CharacterStream * source)985 FunctionLiteral* Parser::ParseLazy(Utf16CharacterStream* source) {
986   Handle<SharedFunctionInfo> shared_info = info()->shared_info();
987   scanner_.Initialize(source);
988   DCHECK(scope_ == NULL);
989   DCHECK(target_stack_ == NULL);
990 
991   Handle<String> name(String::cast(shared_info->name()));
992   DCHECK(ast_value_factory());
993   fni_ = new (zone()) FuncNameInferrer(ast_value_factory(), zone());
994   const AstRawString* raw_name = ast_value_factory()->GetString(name);
995   fni_->PushEnclosingName(raw_name);
996 
997   ParsingModeScope parsing_mode(this, PARSE_EAGERLY);
998 
999   // Place holder for the result.
1000   FunctionLiteral* result = NULL;
1001 
1002   {
1003     // Parse the function literal.
1004     Scope* scope = NewScope(scope_, GLOBAL_SCOPE);
1005     info()->SetGlobalScope(scope);
1006     if (!info()->closure().is_null()) {
1007       scope = Scope::DeserializeScopeChain(info()->closure()->context(), scope,
1008                                            zone());
1009     }
1010     original_scope_ = scope;
1011     FunctionState function_state(&function_state_, &scope_, scope, zone(),
1012                                  ast_value_factory(),
1013                                  info()->ast_node_id_gen());
1014     DCHECK(scope->strict_mode() == SLOPPY || info()->strict_mode() == STRICT);
1015     DCHECK(info()->strict_mode() == shared_info->strict_mode());
1016     scope->SetStrictMode(shared_info->strict_mode());
1017     FunctionLiteral::FunctionType function_type = shared_info->is_expression()
1018         ? (shared_info->is_anonymous()
1019               ? FunctionLiteral::ANONYMOUS_EXPRESSION
1020               : FunctionLiteral::NAMED_EXPRESSION)
1021         : FunctionLiteral::DECLARATION;
1022     bool ok = true;
1023 
1024     if (shared_info->is_arrow()) {
1025       Expression* expression = ParseExpression(false, &ok);
1026       DCHECK(expression->IsFunctionLiteral());
1027       result = expression->AsFunctionLiteral();
1028     } else {
1029       result = ParseFunctionLiteral(raw_name, Scanner::Location::invalid(),
1030                                     false,  // Strict mode name already checked.
1031                                     shared_info->kind(), RelocInfo::kNoPosition,
1032                                     function_type,
1033                                     FunctionLiteral::NORMAL_ARITY, &ok);
1034     }
1035     // Make sure the results agree.
1036     DCHECK(ok == (result != NULL));
1037   }
1038 
1039   // Make sure the target stack is empty.
1040   DCHECK(target_stack_ == NULL);
1041 
1042   if (result != NULL) {
1043     Handle<String> inferred_name(shared_info->inferred_name());
1044     result->set_inferred_name(inferred_name);
1045   }
1046   return result;
1047 }
1048 
1049 
ParseSourceElements(ZoneList<Statement * > * processor,int end_token,bool is_eval,bool is_global,Scope ** eval_scope,bool * ok)1050 void* Parser::ParseSourceElements(ZoneList<Statement*>* processor,
1051                                   int end_token, bool is_eval, bool is_global,
1052                                   Scope** eval_scope, bool* ok) {
1053   // SourceElements ::
1054   //   (ModuleElement)* <end_token>
1055 
1056   // Allocate a target stack to use for this set of source
1057   // elements. This way, all scripts and functions get their own
1058   // target stack thus avoiding illegal breaks and continues across
1059   // functions.
1060   TargetScope scope(&this->target_stack_);
1061 
1062   DCHECK(processor != NULL);
1063   bool directive_prologue = true;     // Parsing directive prologue.
1064 
1065   while (peek() != end_token) {
1066     if (directive_prologue && peek() != Token::STRING) {
1067       directive_prologue = false;
1068     }
1069 
1070     Scanner::Location token_loc = scanner()->peek_location();
1071     Statement* stat;
1072     if (is_global && !is_eval) {
1073       stat = ParseModuleElement(NULL, CHECK_OK);
1074     } else {
1075       stat = ParseBlockElement(NULL, CHECK_OK);
1076     }
1077     if (stat == NULL || stat->IsEmpty()) {
1078       directive_prologue = false;   // End of directive prologue.
1079       continue;
1080     }
1081 
1082     if (directive_prologue) {
1083       // A shot at a directive.
1084       ExpressionStatement* e_stat;
1085       Literal* literal;
1086       // Still processing directive prologue?
1087       if ((e_stat = stat->AsExpressionStatement()) != NULL &&
1088           (literal = e_stat->expression()->AsLiteral()) != NULL &&
1089           literal->raw_value()->IsString()) {
1090         // Check "use strict" directive (ES5 14.1) and "use asm" directive. Only
1091         // one can be present.
1092         if (strict_mode() == SLOPPY &&
1093             literal->raw_value()->AsString() ==
1094                 ast_value_factory()->use_strict_string() &&
1095             token_loc.end_pos - token_loc.beg_pos ==
1096                 ast_value_factory()->use_strict_string()->length() + 2) {
1097           // TODO(mstarzinger): Global strict eval calls, need their own scope
1098           // as specified in ES5 10.4.2(3). The correct fix would be to always
1099           // add this scope in DoParseProgram(), but that requires adaptations
1100           // all over the code base, so we go with a quick-fix for now.
1101           // In the same manner, we have to patch the parsing mode.
1102           if (is_eval && !scope_->is_eval_scope()) {
1103             DCHECK(scope_->is_global_scope());
1104             Scope* scope = NewScope(scope_, EVAL_SCOPE);
1105             scope->set_start_position(scope_->start_position());
1106             scope->set_end_position(scope_->end_position());
1107             scope_ = scope;
1108             if (eval_scope != NULL) {
1109               // Caller will correct the positions of the ad hoc eval scope.
1110               *eval_scope = scope;
1111             }
1112             mode_ = PARSE_EAGERLY;
1113           }
1114           scope_->SetStrictMode(STRICT);
1115           // "use strict" is the only directive for now.
1116           directive_prologue = false;
1117         } else if (literal->raw_value()->AsString() ==
1118                        ast_value_factory()->use_asm_string() &&
1119                    token_loc.end_pos - token_loc.beg_pos ==
1120                        ast_value_factory()->use_asm_string()->length() + 2) {
1121           // Store the usage count; The actual use counter on the isolate is
1122           // incremented after parsing is done.
1123           ++use_counts_[v8::Isolate::kUseAsm];
1124           scope_->SetAsmModule();
1125         }
1126       } else {
1127         // End of the directive prologue.
1128         directive_prologue = false;
1129       }
1130     }
1131 
1132     processor->Add(stat, zone());
1133   }
1134 
1135   return 0;
1136 }
1137 
1138 
ParseModuleElement(ZoneList<const AstRawString * > * labels,bool * ok)1139 Statement* Parser::ParseModuleElement(ZoneList<const AstRawString*>* labels,
1140                                       bool* ok) {
1141   // (Ecma 262 5th Edition, clause 14):
1142   // SourceElement:
1143   //    Statement
1144   //    FunctionDeclaration
1145   //
1146   // In harmony mode we allow additionally the following productions
1147   // ModuleElement:
1148   //    LetDeclaration
1149   //    ConstDeclaration
1150   //    ModuleDeclaration
1151   //    ImportDeclaration
1152   //    ExportDeclaration
1153   //    GeneratorDeclaration
1154 
1155   switch (peek()) {
1156     case Token::FUNCTION:
1157       return ParseFunctionDeclaration(NULL, ok);
1158     case Token::CLASS:
1159       return ParseClassDeclaration(NULL, ok);
1160     case Token::IMPORT:
1161       return ParseImportDeclaration(ok);
1162     case Token::EXPORT:
1163       return ParseExportDeclaration(ok);
1164     case Token::CONST:
1165       return ParseVariableStatement(kModuleElement, NULL, ok);
1166     case Token::LET:
1167       DCHECK(allow_harmony_scoping());
1168       if (strict_mode() == STRICT) {
1169         return ParseVariableStatement(kModuleElement, NULL, ok);
1170       }
1171       // Fall through.
1172     default: {
1173       Statement* stmt = ParseStatement(labels, CHECK_OK);
1174       // Handle 'module' as a context-sensitive keyword.
1175       if (FLAG_harmony_modules &&
1176           peek() == Token::IDENTIFIER &&
1177           !scanner()->HasAnyLineTerminatorBeforeNext() &&
1178           stmt != NULL) {
1179         ExpressionStatement* estmt = stmt->AsExpressionStatement();
1180         if (estmt != NULL && estmt->expression()->AsVariableProxy() != NULL &&
1181             estmt->expression()->AsVariableProxy()->raw_name() ==
1182                 ast_value_factory()->module_string() &&
1183             !scanner()->literal_contains_escapes()) {
1184           return ParseModuleDeclaration(NULL, ok);
1185         }
1186       }
1187       return stmt;
1188     }
1189   }
1190 }
1191 
1192 
ParseModuleDeclaration(ZoneList<const AstRawString * > * names,bool * ok)1193 Statement* Parser::ParseModuleDeclaration(ZoneList<const AstRawString*>* names,
1194                                           bool* ok) {
1195   // ModuleDeclaration:
1196   //    'module' Identifier Module
1197 
1198   int pos = peek_position();
1199   const AstRawString* name =
1200       ParseIdentifier(kDontAllowEvalOrArguments, CHECK_OK);
1201 
1202 #ifdef DEBUG
1203   if (FLAG_print_interface_details)
1204     PrintF("# Module %.*s ", name->length(), name->raw_data());
1205 #endif
1206 
1207   Module* module = ParseModule(CHECK_OK);
1208   VariableProxy* proxy = NewUnresolved(name, MODULE, module->interface());
1209   Declaration* declaration =
1210       factory()->NewModuleDeclaration(proxy, module, scope_, pos);
1211   Declare(declaration, true, CHECK_OK);
1212 
1213 #ifdef DEBUG
1214   if (FLAG_print_interface_details)
1215     PrintF("# Module %.*s ", name->length(), name->raw_data());
1216   if (FLAG_print_interfaces) {
1217     PrintF("module %.*s: ", name->length(), name->raw_data());
1218     module->interface()->Print();
1219   }
1220 #endif
1221 
1222   if (names) names->Add(name, zone());
1223   if (module->body() == NULL)
1224     return factory()->NewEmptyStatement(pos);
1225   else
1226     return factory()->NewModuleStatement(proxy, module->body(), pos);
1227 }
1228 
1229 
ParseModule(bool * ok)1230 Module* Parser::ParseModule(bool* ok) {
1231   // Module:
1232   //    '{' ModuleElement '}'
1233   //    '=' ModulePath ';'
1234   //    'at' String ';'
1235 
1236   switch (peek()) {
1237     case Token::LBRACE:
1238       return ParseModuleLiteral(ok);
1239 
1240     case Token::ASSIGN: {
1241       Expect(Token::ASSIGN, CHECK_OK);
1242       Module* result = ParseModulePath(CHECK_OK);
1243       ExpectSemicolon(CHECK_OK);
1244       return result;
1245     }
1246 
1247     default: {
1248       ExpectContextualKeyword(CStrVector("at"), CHECK_OK);
1249       Module* result = ParseModuleUrl(CHECK_OK);
1250       ExpectSemicolon(CHECK_OK);
1251       return result;
1252     }
1253   }
1254 }
1255 
1256 
ParseModuleLiteral(bool * ok)1257 Module* Parser::ParseModuleLiteral(bool* ok) {
1258   // Module:
1259   //    '{' ModuleElement '}'
1260 
1261   int pos = peek_position();
1262   // Construct block expecting 16 statements.
1263   Block* body = factory()->NewBlock(NULL, 16, false, RelocInfo::kNoPosition);
1264 #ifdef DEBUG
1265   if (FLAG_print_interface_details) PrintF("# Literal ");
1266 #endif
1267   Scope* scope = NewScope(scope_, MODULE_SCOPE);
1268 
1269   Expect(Token::LBRACE, CHECK_OK);
1270   scope->set_start_position(scanner()->location().beg_pos);
1271   scope->SetStrictMode(STRICT);
1272 
1273   {
1274     BlockState block_state(&scope_, scope);
1275     TargetCollector collector(zone());
1276     Target target(&this->target_stack_, &collector);
1277     Target target_body(&this->target_stack_, body);
1278 
1279     while (peek() != Token::RBRACE) {
1280       Statement* stat = ParseModuleElement(NULL, CHECK_OK);
1281       if (stat && !stat->IsEmpty()) {
1282         body->AddStatement(stat, zone());
1283       }
1284     }
1285   }
1286 
1287   Expect(Token::RBRACE, CHECK_OK);
1288   scope->set_end_position(scanner()->location().end_pos);
1289   body->set_scope(scope);
1290 
1291   // Check that all exports are bound.
1292   Interface* interface = scope->interface();
1293   for (Interface::Iterator it = interface->iterator();
1294        !it.done(); it.Advance()) {
1295     if (scope->LookupLocal(it.name()) == NULL) {
1296       ParserTraits::ReportMessage("module_export_undefined", it.name());
1297       *ok = false;
1298       return NULL;
1299     }
1300   }
1301 
1302   interface->MakeModule(ok);
1303   DCHECK(*ok);
1304   interface->Freeze(ok);
1305   DCHECK(*ok);
1306   return factory()->NewModuleLiteral(body, interface, pos);
1307 }
1308 
1309 
ParseModulePath(bool * ok)1310 Module* Parser::ParseModulePath(bool* ok) {
1311   // ModulePath:
1312   //    Identifier
1313   //    ModulePath '.' Identifier
1314 
1315   int pos = peek_position();
1316   Module* result = ParseModuleVariable(CHECK_OK);
1317   while (Check(Token::PERIOD)) {
1318     const AstRawString* name = ParseIdentifierName(CHECK_OK);
1319 #ifdef DEBUG
1320     if (FLAG_print_interface_details)
1321       PrintF("# Path .%.*s ", name->length(), name->raw_data());
1322 #endif
1323     Module* member = factory()->NewModulePath(result, name, pos);
1324     result->interface()->Add(name, member->interface(), zone(), ok);
1325     if (!*ok) {
1326 #ifdef DEBUG
1327       if (FLAG_print_interfaces) {
1328         PrintF("PATH TYPE ERROR at '%.*s'\n", name->length(), name->raw_data());
1329         PrintF("result: ");
1330         result->interface()->Print();
1331         PrintF("member: ");
1332         member->interface()->Print();
1333       }
1334 #endif
1335       ParserTraits::ReportMessage("invalid_module_path", name);
1336       return NULL;
1337     }
1338     result = member;
1339   }
1340 
1341   return result;
1342 }
1343 
1344 
ParseModuleVariable(bool * ok)1345 Module* Parser::ParseModuleVariable(bool* ok) {
1346   // ModulePath:
1347   //    Identifier
1348 
1349   int pos = peek_position();
1350   const AstRawString* name =
1351       ParseIdentifier(kDontAllowEvalOrArguments, CHECK_OK);
1352 #ifdef DEBUG
1353   if (FLAG_print_interface_details)
1354     PrintF("# Module variable %.*s ", name->length(), name->raw_data());
1355 #endif
1356   VariableProxy* proxy = scope_->NewUnresolved(
1357       factory(), name, Interface::NewModule(zone()),
1358       scanner()->location().beg_pos);
1359 
1360   return factory()->NewModuleVariable(proxy, pos);
1361 }
1362 
1363 
ParseModuleUrl(bool * ok)1364 Module* Parser::ParseModuleUrl(bool* ok) {
1365   // Module:
1366   //    String
1367 
1368   int pos = peek_position();
1369   Expect(Token::STRING, CHECK_OK);
1370   const AstRawString* symbol = GetSymbol(scanner());
1371 
1372   // TODO(ES6): Request JS resource from environment...
1373 
1374 #ifdef DEBUG
1375   if (FLAG_print_interface_details) PrintF("# Url ");
1376 #endif
1377 
1378   // Create an empty literal as long as the feature isn't finished.
1379   USE(symbol);
1380   Scope* scope = NewScope(scope_, MODULE_SCOPE);
1381   Block* body = factory()->NewBlock(NULL, 1, false, RelocInfo::kNoPosition);
1382   body->set_scope(scope);
1383   Interface* interface = scope->interface();
1384   Module* result = factory()->NewModuleLiteral(body, interface, pos);
1385   interface->Freeze(ok);
1386   DCHECK(*ok);
1387   interface->Unify(scope->interface(), zone(), ok);
1388   DCHECK(*ok);
1389   return result;
1390 }
1391 
1392 
ParseModuleSpecifier(bool * ok)1393 Module* Parser::ParseModuleSpecifier(bool* ok) {
1394   // ModuleSpecifier:
1395   //    String
1396   //    ModulePath
1397 
1398   if (peek() == Token::STRING) {
1399     return ParseModuleUrl(ok);
1400   } else {
1401     return ParseModulePath(ok);
1402   }
1403 }
1404 
1405 
ParseImportDeclaration(bool * ok)1406 Block* Parser::ParseImportDeclaration(bool* ok) {
1407   // ImportDeclaration:
1408   //    'import' IdentifierName (',' IdentifierName)* 'from' ModuleSpecifier ';'
1409   //
1410   // TODO(ES6): implement destructuring ImportSpecifiers
1411 
1412   int pos = peek_position();
1413   Expect(Token::IMPORT, CHECK_OK);
1414   ZoneList<const AstRawString*> names(1, zone());
1415 
1416   const AstRawString* name = ParseIdentifierName(CHECK_OK);
1417   names.Add(name, zone());
1418   while (peek() == Token::COMMA) {
1419     Consume(Token::COMMA);
1420     name = ParseIdentifierName(CHECK_OK);
1421     names.Add(name, zone());
1422   }
1423 
1424   ExpectContextualKeyword(CStrVector("from"), CHECK_OK);
1425   Module* module = ParseModuleSpecifier(CHECK_OK);
1426   ExpectSemicolon(CHECK_OK);
1427 
1428   // Generate a separate declaration for each identifier.
1429   // TODO(ES6): once we implement destructuring, make that one declaration.
1430   Block* block = factory()->NewBlock(NULL, 1, true, RelocInfo::kNoPosition);
1431   for (int i = 0; i < names.length(); ++i) {
1432 #ifdef DEBUG
1433     if (FLAG_print_interface_details)
1434       PrintF("# Import %.*s ", name->length(), name->raw_data());
1435 #endif
1436     Interface* interface = Interface::NewUnknown(zone());
1437     module->interface()->Add(names[i], interface, zone(), ok);
1438     if (!*ok) {
1439 #ifdef DEBUG
1440       if (FLAG_print_interfaces) {
1441         PrintF("IMPORT TYPE ERROR at '%.*s'\n", name->length(),
1442                name->raw_data());
1443         PrintF("module: ");
1444         module->interface()->Print();
1445       }
1446 #endif
1447       ParserTraits::ReportMessage("invalid_module_path", name);
1448       return NULL;
1449     }
1450     VariableProxy* proxy = NewUnresolved(names[i], LET, interface);
1451     Declaration* declaration =
1452         factory()->NewImportDeclaration(proxy, module, scope_, pos);
1453     Declare(declaration, true, CHECK_OK);
1454   }
1455 
1456   return block;
1457 }
1458 
1459 
ParseExportDeclaration(bool * ok)1460 Statement* Parser::ParseExportDeclaration(bool* ok) {
1461   // ExportDeclaration:
1462   //    'export' Identifier (',' Identifier)* ';'
1463   //    'export' VariableDeclaration
1464   //    'export' FunctionDeclaration
1465   //    'export' GeneratorDeclaration
1466   //    'export' ModuleDeclaration
1467   //
1468   // TODO(ES6): implement structuring ExportSpecifiers
1469 
1470   Expect(Token::EXPORT, CHECK_OK);
1471 
1472   Statement* result = NULL;
1473   ZoneList<const AstRawString*> names(1, zone());
1474   switch (peek()) {
1475     case Token::IDENTIFIER: {
1476       int pos = position();
1477       const AstRawString* name =
1478           ParseIdentifier(kDontAllowEvalOrArguments, CHECK_OK);
1479       // Handle 'module' as a context-sensitive keyword.
1480       if (name != ast_value_factory()->module_string()) {
1481         names.Add(name, zone());
1482         while (peek() == Token::COMMA) {
1483           Consume(Token::COMMA);
1484           name = ParseIdentifier(kDontAllowEvalOrArguments, CHECK_OK);
1485           names.Add(name, zone());
1486         }
1487         ExpectSemicolon(CHECK_OK);
1488         result = factory()->NewEmptyStatement(pos);
1489       } else {
1490         result = ParseModuleDeclaration(&names, CHECK_OK);
1491       }
1492       break;
1493     }
1494 
1495     case Token::FUNCTION:
1496       result = ParseFunctionDeclaration(&names, CHECK_OK);
1497       break;
1498 
1499     case Token::CLASS:
1500       result = ParseClassDeclaration(&names, CHECK_OK);
1501       break;
1502 
1503     case Token::VAR:
1504     case Token::LET:
1505     case Token::CONST:
1506       result = ParseVariableStatement(kModuleElement, &names, CHECK_OK);
1507       break;
1508 
1509     default:
1510       *ok = false;
1511       ReportUnexpectedToken(scanner()->current_token());
1512       return NULL;
1513   }
1514 
1515   // Every export of a module may be assigned.
1516   for (int i = 0; i < names.length(); ++i) {
1517     Variable* var = scope_->Lookup(names[i]);
1518     if (var == NULL) {
1519       // TODO(sigurds) This is an export that has no definition yet,
1520       // not clear what to do in this case.
1521       continue;
1522     }
1523     if (!IsImmutableVariableMode(var->mode())) {
1524       var->set_maybe_assigned();
1525     }
1526   }
1527 
1528   // Extract declared names into export declarations and interface.
1529   Interface* interface = scope_->interface();
1530   for (int i = 0; i < names.length(); ++i) {
1531 #ifdef DEBUG
1532     if (FLAG_print_interface_details)
1533       PrintF("# Export %.*s ", names[i]->length(), names[i]->raw_data());
1534 #endif
1535     Interface* inner = Interface::NewUnknown(zone());
1536     interface->Add(names[i], inner, zone(), CHECK_OK);
1537     if (!*ok)
1538       return NULL;
1539     VariableProxy* proxy = NewUnresolved(names[i], LET, inner);
1540     USE(proxy);
1541     // TODO(rossberg): Rethink whether we actually need to store export
1542     // declarations (for compilation?).
1543     // ExportDeclaration* declaration =
1544     //     factory()->NewExportDeclaration(proxy, scope_, position);
1545     // scope_->AddDeclaration(declaration);
1546   }
1547 
1548   DCHECK(result != NULL);
1549   return result;
1550 }
1551 
1552 
ParseBlockElement(ZoneList<const AstRawString * > * labels,bool * ok)1553 Statement* Parser::ParseBlockElement(ZoneList<const AstRawString*>* labels,
1554                                      bool* ok) {
1555   // (Ecma 262 5th Edition, clause 14):
1556   // SourceElement:
1557   //    Statement
1558   //    FunctionDeclaration
1559   //
1560   // In harmony mode we allow additionally the following productions
1561   // BlockElement (aka SourceElement):
1562   //    LetDeclaration
1563   //    ConstDeclaration
1564   //    GeneratorDeclaration
1565   //    ClassDeclaration
1566 
1567   switch (peek()) {
1568     case Token::FUNCTION:
1569       return ParseFunctionDeclaration(NULL, ok);
1570     case Token::CLASS:
1571       return ParseClassDeclaration(NULL, ok);
1572     case Token::CONST:
1573       return ParseVariableStatement(kModuleElement, NULL, ok);
1574     case Token::LET:
1575       DCHECK(allow_harmony_scoping());
1576       if (strict_mode() == STRICT) {
1577         return ParseVariableStatement(kModuleElement, NULL, ok);
1578       }
1579       // Fall through.
1580     default:
1581       return ParseStatement(labels, ok);
1582   }
1583 }
1584 
1585 
ParseStatement(ZoneList<const AstRawString * > * labels,bool * ok)1586 Statement* Parser::ParseStatement(ZoneList<const AstRawString*>* labels,
1587                                   bool* ok) {
1588   // Statement ::
1589   //   Block
1590   //   VariableStatement
1591   //   EmptyStatement
1592   //   ExpressionStatement
1593   //   IfStatement
1594   //   IterationStatement
1595   //   ContinueStatement
1596   //   BreakStatement
1597   //   ReturnStatement
1598   //   WithStatement
1599   //   LabelledStatement
1600   //   SwitchStatement
1601   //   ThrowStatement
1602   //   TryStatement
1603   //   DebuggerStatement
1604 
1605   // Note: Since labels can only be used by 'break' and 'continue'
1606   // statements, which themselves are only valid within blocks,
1607   // iterations or 'switch' statements (i.e., BreakableStatements),
1608   // labels can be simply ignored in all other cases; except for
1609   // trivial labeled break statements 'label: break label' which is
1610   // parsed into an empty statement.
1611   switch (peek()) {
1612     case Token::LBRACE:
1613       return ParseBlock(labels, ok);
1614 
1615     case Token::SEMICOLON:
1616       Next();
1617       return factory()->NewEmptyStatement(RelocInfo::kNoPosition);
1618 
1619     case Token::IF:
1620       return ParseIfStatement(labels, ok);
1621 
1622     case Token::DO:
1623       return ParseDoWhileStatement(labels, ok);
1624 
1625     case Token::WHILE:
1626       return ParseWhileStatement(labels, ok);
1627 
1628     case Token::FOR:
1629       return ParseForStatement(labels, ok);
1630 
1631     case Token::CONTINUE:
1632       return ParseContinueStatement(ok);
1633 
1634     case Token::BREAK:
1635       return ParseBreakStatement(labels, ok);
1636 
1637     case Token::RETURN:
1638       return ParseReturnStatement(ok);
1639 
1640     case Token::WITH:
1641       return ParseWithStatement(labels, ok);
1642 
1643     case Token::SWITCH:
1644       return ParseSwitchStatement(labels, ok);
1645 
1646     case Token::THROW:
1647       return ParseThrowStatement(ok);
1648 
1649     case Token::TRY: {
1650       // NOTE: It is somewhat complicated to have labels on
1651       // try-statements. When breaking out of a try-finally statement,
1652       // one must take great care not to treat it as a
1653       // fall-through. It is much easier just to wrap the entire
1654       // try-statement in a statement block and put the labels there
1655       Block* result =
1656           factory()->NewBlock(labels, 1, false, RelocInfo::kNoPosition);
1657       Target target(&this->target_stack_, result);
1658       TryStatement* statement = ParseTryStatement(CHECK_OK);
1659       if (result) result->AddStatement(statement, zone());
1660       return result;
1661     }
1662 
1663     case Token::FUNCTION: {
1664       // FunctionDeclaration is only allowed in the context of SourceElements
1665       // (Ecma 262 5th Edition, clause 14):
1666       // SourceElement:
1667       //    Statement
1668       //    FunctionDeclaration
1669       // Common language extension is to allow function declaration in place
1670       // of any statement. This language extension is disabled in strict mode.
1671       //
1672       // In Harmony mode, this case also handles the extension:
1673       // Statement:
1674       //    GeneratorDeclaration
1675       if (strict_mode() == STRICT) {
1676         ReportMessageAt(scanner()->peek_location(), "strict_function");
1677         *ok = false;
1678         return NULL;
1679       }
1680       return ParseFunctionDeclaration(NULL, ok);
1681     }
1682 
1683     case Token::CLASS:
1684       return ParseClassDeclaration(NULL, ok);
1685 
1686     case Token::DEBUGGER:
1687       return ParseDebuggerStatement(ok);
1688 
1689     case Token::VAR:
1690     case Token::CONST:
1691       return ParseVariableStatement(kStatement, NULL, ok);
1692 
1693     case Token::LET:
1694       DCHECK(allow_harmony_scoping());
1695       if (strict_mode() == STRICT) {
1696         return ParseVariableStatement(kStatement, NULL, ok);
1697       }
1698       // Fall through.
1699     default:
1700       return ParseExpressionOrLabelledStatement(labels, ok);
1701   }
1702 }
1703 
1704 
NewUnresolved(const AstRawString * name,VariableMode mode,Interface * interface)1705 VariableProxy* Parser::NewUnresolved(const AstRawString* name,
1706                                      VariableMode mode, Interface* interface) {
1707   // If we are inside a function, a declaration of a var/const variable is a
1708   // truly local variable, and the scope of the variable is always the function
1709   // scope.
1710   // Let/const variables in harmony mode are always added to the immediately
1711   // enclosing scope.
1712   return DeclarationScope(mode)->NewUnresolved(
1713       factory(), name, interface, position());
1714 }
1715 
1716 
Declare(Declaration * declaration,bool resolve,bool * ok)1717 void Parser::Declare(Declaration* declaration, bool resolve, bool* ok) {
1718   VariableProxy* proxy = declaration->proxy();
1719   DCHECK(proxy->raw_name() != NULL);
1720   const AstRawString* name = proxy->raw_name();
1721   VariableMode mode = declaration->mode();
1722   Scope* declaration_scope = DeclarationScope(mode);
1723   Variable* var = NULL;
1724 
1725   // If a suitable scope exists, then we can statically declare this
1726   // variable and also set its mode. In any case, a Declaration node
1727   // will be added to the scope so that the declaration can be added
1728   // to the corresponding activation frame at runtime if necessary.
1729   // For instance declarations inside an eval scope need to be added
1730   // to the calling function context.
1731   // Similarly, strict mode eval scope does not leak variable declarations to
1732   // the caller's scope so we declare all locals, too.
1733   if (declaration_scope->is_function_scope() ||
1734       declaration_scope->is_strict_eval_scope() ||
1735       declaration_scope->is_block_scope() ||
1736       declaration_scope->is_module_scope() ||
1737       declaration_scope->is_global_scope()) {
1738     // Declare the variable in the declaration scope.
1739     // For the global scope, we have to check for collisions with earlier
1740     // (i.e., enclosing) global scopes, to maintain the illusion of a single
1741     // global scope.
1742     var = declaration_scope->is_global_scope()
1743         ? declaration_scope->Lookup(name)
1744         : declaration_scope->LookupLocal(name);
1745     if (var == NULL) {
1746       // Declare the name.
1747       var = declaration_scope->DeclareLocal(name, mode,
1748                                             declaration->initialization(),
1749                                             kNotAssigned, proxy->interface());
1750     } else if (IsLexicalVariableMode(mode) || IsLexicalVariableMode(var->mode())
1751                || ((mode == CONST_LEGACY || var->mode() == CONST_LEGACY) &&
1752                    !declaration_scope->is_global_scope())) {
1753       // The name was declared in this scope before; check for conflicting
1754       // re-declarations. We have a conflict if either of the declarations is
1755       // not a var (in the global scope, we also have to ignore legacy const for
1756       // compatibility). There is similar code in runtime.cc in the Declare
1757       // functions. The function CheckConflictingVarDeclarations checks for
1758       // var and let bindings from different scopes whereas this is a check for
1759       // conflicting declarations within the same scope. This check also covers
1760       // the special case
1761       //
1762       // function () { let x; { var x; } }
1763       //
1764       // because the var declaration is hoisted to the function scope where 'x'
1765       // is already bound.
1766       DCHECK(IsDeclaredVariableMode(var->mode()));
1767       if (allow_harmony_scoping() && strict_mode() == STRICT) {
1768         // In harmony we treat re-declarations as early errors. See
1769         // ES5 16 for a definition of early errors.
1770         ParserTraits::ReportMessage("var_redeclaration", name);
1771         *ok = false;
1772         return;
1773       }
1774       Expression* expression = NewThrowTypeError(
1775           "var_redeclaration", name, declaration->position());
1776       declaration_scope->SetIllegalRedeclaration(expression);
1777     } else if (mode == VAR) {
1778       var->set_maybe_assigned();
1779     }
1780   }
1781 
1782   // We add a declaration node for every declaration. The compiler
1783   // will only generate code if necessary. In particular, declarations
1784   // for inner local variables that do not represent functions won't
1785   // result in any generated code.
1786   //
1787   // Note that we always add an unresolved proxy even if it's not
1788   // used, simply because we don't know in this method (w/o extra
1789   // parameters) if the proxy is needed or not. The proxy will be
1790   // bound during variable resolution time unless it was pre-bound
1791   // below.
1792   //
1793   // WARNING: This will lead to multiple declaration nodes for the
1794   // same variable if it is declared several times. This is not a
1795   // semantic issue as long as we keep the source order, but it may be
1796   // a performance issue since it may lead to repeated
1797   // RuntimeHidden_DeclareLookupSlot calls.
1798   declaration_scope->AddDeclaration(declaration);
1799 
1800   if (mode == CONST_LEGACY && declaration_scope->is_global_scope()) {
1801     // For global const variables we bind the proxy to a variable.
1802     DCHECK(resolve);  // should be set by all callers
1803     Variable::Kind kind = Variable::NORMAL;
1804     var = new (zone())
1805         Variable(declaration_scope, name, mode, true, kind,
1806                  kNeedsInitialization, kNotAssigned, proxy->interface());
1807   } else if (declaration_scope->is_eval_scope() &&
1808              declaration_scope->strict_mode() == SLOPPY) {
1809     // For variable declarations in a sloppy eval scope the proxy is bound
1810     // to a lookup variable to force a dynamic declaration using the
1811     // DeclareLookupSlot runtime function.
1812     Variable::Kind kind = Variable::NORMAL;
1813     // TODO(sigurds) figure out if kNotAssigned is OK here
1814     var = new (zone()) Variable(declaration_scope, name, mode, true, kind,
1815                                 declaration->initialization(), kNotAssigned,
1816                                 proxy->interface());
1817     var->AllocateTo(Variable::LOOKUP, -1);
1818     resolve = true;
1819   }
1820 
1821   // If requested and we have a local variable, bind the proxy to the variable
1822   // at parse-time. This is used for functions (and consts) declared inside
1823   // statements: the corresponding function (or const) variable must be in the
1824   // function scope and not a statement-local scope, e.g. as provided with a
1825   // 'with' statement:
1826   //
1827   //   with (obj) {
1828   //     function f() {}
1829   //   }
1830   //
1831   // which is translated into:
1832   //
1833   //   with (obj) {
1834   //     // in this case this is not: 'var f; f = function () {};'
1835   //     var f = function () {};
1836   //   }
1837   //
1838   // Note that if 'f' is accessed from inside the 'with' statement, it
1839   // will be allocated in the context (because we must be able to look
1840   // it up dynamically) but it will also be accessed statically, i.e.,
1841   // with a context slot index and a context chain length for this
1842   // initialization code. Thus, inside the 'with' statement, we need
1843   // both access to the static and the dynamic context chain; the
1844   // runtime needs to provide both.
1845   if (resolve && var != NULL) {
1846     proxy->BindTo(var);
1847 
1848     if (FLAG_harmony_modules) {
1849       bool ok;
1850 #ifdef DEBUG
1851       if (FLAG_print_interface_details) {
1852         PrintF("# Declare %.*s ", var->raw_name()->length(),
1853                var->raw_name()->raw_data());
1854       }
1855 #endif
1856       proxy->interface()->Unify(var->interface(), zone(), &ok);
1857       if (!ok) {
1858 #ifdef DEBUG
1859         if (FLAG_print_interfaces) {
1860           PrintF("DECLARE TYPE ERROR\n");
1861           PrintF("proxy: ");
1862           proxy->interface()->Print();
1863           PrintF("var: ");
1864           var->interface()->Print();
1865         }
1866 #endif
1867         ParserTraits::ReportMessage("module_type_error", name);
1868       }
1869     }
1870   }
1871 }
1872 
1873 
1874 // Language extension which is only enabled for source files loaded
1875 // through the API's extension mechanism.  A native function
1876 // declaration is resolved by looking up the function through a
1877 // callback provided by the extension.
ParseNativeDeclaration(bool * ok)1878 Statement* Parser::ParseNativeDeclaration(bool* ok) {
1879   int pos = peek_position();
1880   Expect(Token::FUNCTION, CHECK_OK);
1881   // Allow "eval" or "arguments" for backward compatibility.
1882   const AstRawString* name = ParseIdentifier(kAllowEvalOrArguments, CHECK_OK);
1883   Expect(Token::LPAREN, CHECK_OK);
1884   bool done = (peek() == Token::RPAREN);
1885   while (!done) {
1886     ParseIdentifier(kAllowEvalOrArguments, CHECK_OK);
1887     done = (peek() == Token::RPAREN);
1888     if (!done) {
1889       Expect(Token::COMMA, CHECK_OK);
1890     }
1891   }
1892   Expect(Token::RPAREN, CHECK_OK);
1893   Expect(Token::SEMICOLON, CHECK_OK);
1894 
1895   // Make sure that the function containing the native declaration
1896   // isn't lazily compiled. The extension structures are only
1897   // accessible while parsing the first time not when reparsing
1898   // because of lazy compilation.
1899   DeclarationScope(VAR)->ForceEagerCompilation();
1900 
1901   // TODO(1240846): It's weird that native function declarations are
1902   // introduced dynamically when we meet their declarations, whereas
1903   // other functions are set up when entering the surrounding scope.
1904   VariableProxy* proxy = NewUnresolved(name, VAR, Interface::NewValue());
1905   Declaration* declaration =
1906       factory()->NewVariableDeclaration(proxy, VAR, scope_, pos);
1907   Declare(declaration, true, CHECK_OK);
1908   NativeFunctionLiteral* lit = factory()->NewNativeFunctionLiteral(
1909       name, extension_, RelocInfo::kNoPosition);
1910   return factory()->NewExpressionStatement(
1911       factory()->NewAssignment(
1912           Token::INIT_VAR, proxy, lit, RelocInfo::kNoPosition),
1913       pos);
1914 }
1915 
1916 
ParseFunctionDeclaration(ZoneList<const AstRawString * > * names,bool * ok)1917 Statement* Parser::ParseFunctionDeclaration(
1918     ZoneList<const AstRawString*>* names, bool* ok) {
1919   // FunctionDeclaration ::
1920   //   'function' Identifier '(' FormalParameterListopt ')' '{' FunctionBody '}'
1921   // GeneratorDeclaration ::
1922   //   'function' '*' Identifier '(' FormalParameterListopt ')'
1923   //      '{' FunctionBody '}'
1924   Expect(Token::FUNCTION, CHECK_OK);
1925   int pos = position();
1926   bool is_generator = Check(Token::MUL);
1927   bool is_strict_reserved = false;
1928   const AstRawString* name = ParseIdentifierOrStrictReservedWord(
1929       &is_strict_reserved, CHECK_OK);
1930   FunctionLiteral* fun =
1931       ParseFunctionLiteral(name, scanner()->location(), is_strict_reserved,
1932                            is_generator ? FunctionKind::kGeneratorFunction
1933                                         : FunctionKind::kNormalFunction,
1934                            pos, FunctionLiteral::DECLARATION,
1935                            FunctionLiteral::NORMAL_ARITY, CHECK_OK);
1936   // Even if we're not at the top-level of the global or a function
1937   // scope, we treat it as such and introduce the function with its
1938   // initial value upon entering the corresponding scope.
1939   // In ES6, a function behaves as a lexical binding, except in the
1940   // global scope, or the initial scope of eval or another function.
1941   VariableMode mode =
1942       allow_harmony_scoping() && strict_mode() == STRICT &&
1943       !(scope_->is_global_scope() || scope_->is_eval_scope() ||
1944           scope_->is_function_scope()) ? LET : VAR;
1945   VariableProxy* proxy = NewUnresolved(name, mode, Interface::NewValue());
1946   Declaration* declaration =
1947       factory()->NewFunctionDeclaration(proxy, mode, fun, scope_, pos);
1948   Declare(declaration, true, CHECK_OK);
1949   if (names) names->Add(name, zone());
1950   return factory()->NewEmptyStatement(RelocInfo::kNoPosition);
1951 }
1952 
1953 
ParseClassDeclaration(ZoneList<const AstRawString * > * names,bool * ok)1954 Statement* Parser::ParseClassDeclaration(ZoneList<const AstRawString*>* names,
1955                                          bool* ok) {
1956   // ClassDeclaration ::
1957   //   'class' Identifier ('extends' LeftHandExpression)? '{' ClassBody '}'
1958   //
1959   // A ClassDeclaration
1960   //
1961   //   class C { ... }
1962   //
1963   // has the same semantics as:
1964   //
1965   //   let C = class C { ... };
1966   //
1967   // so rewrite it as such.
1968 
1969   Expect(Token::CLASS, CHECK_OK);
1970   int pos = position();
1971   bool is_strict_reserved = false;
1972   const AstRawString* name =
1973       ParseIdentifierOrStrictReservedWord(&is_strict_reserved, CHECK_OK);
1974   Expression* value = ParseClassLiteral(name, scanner()->location(),
1975                                         is_strict_reserved, pos, CHECK_OK);
1976 
1977   Block* block = factory()->NewBlock(NULL, 1, true, pos);
1978   VariableMode mode = LET;
1979   VariableProxy* proxy = NewUnresolved(name, mode, Interface::NewValue());
1980   Declaration* declaration =
1981       factory()->NewVariableDeclaration(proxy, mode, scope_, pos);
1982   Declare(declaration, true, CHECK_OK);
1983 
1984   Token::Value init_op = Token::INIT_LET;
1985   Assignment* assignment = factory()->NewAssignment(init_op, proxy, value, pos);
1986   block->AddStatement(
1987       factory()->NewExpressionStatement(assignment, RelocInfo::kNoPosition),
1988       zone());
1989 
1990   if (names) names->Add(name, zone());
1991   return block;
1992 }
1993 
1994 
ParseBlock(ZoneList<const AstRawString * > * labels,bool * ok)1995 Block* Parser::ParseBlock(ZoneList<const AstRawString*>* labels, bool* ok) {
1996   if (allow_harmony_scoping() && strict_mode() == STRICT) {
1997     return ParseScopedBlock(labels, ok);
1998   }
1999 
2000   // Block ::
2001   //   '{' Statement* '}'
2002 
2003   // Note that a Block does not introduce a new execution scope!
2004   // (ECMA-262, 3rd, 12.2)
2005   //
2006   // Construct block expecting 16 statements.
2007   Block* result =
2008       factory()->NewBlock(labels, 16, false, RelocInfo::kNoPosition);
2009   Target target(&this->target_stack_, result);
2010   Expect(Token::LBRACE, CHECK_OK);
2011   while (peek() != Token::RBRACE) {
2012     Statement* stat = ParseStatement(NULL, CHECK_OK);
2013     if (stat && !stat->IsEmpty()) {
2014       result->AddStatement(stat, zone());
2015     }
2016   }
2017   Expect(Token::RBRACE, CHECK_OK);
2018   return result;
2019 }
2020 
2021 
ParseScopedBlock(ZoneList<const AstRawString * > * labels,bool * ok)2022 Block* Parser::ParseScopedBlock(ZoneList<const AstRawString*>* labels,
2023                                 bool* ok) {
2024   // The harmony mode uses block elements instead of statements.
2025   //
2026   // Block ::
2027   //   '{' BlockElement* '}'
2028 
2029   // Construct block expecting 16 statements.
2030   Block* body =
2031       factory()->NewBlock(labels, 16, false, RelocInfo::kNoPosition);
2032   Scope* block_scope = NewScope(scope_, BLOCK_SCOPE);
2033 
2034   // Parse the statements and collect escaping labels.
2035   Expect(Token::LBRACE, CHECK_OK);
2036   block_scope->set_start_position(scanner()->location().beg_pos);
2037   { BlockState block_state(&scope_, block_scope);
2038     TargetCollector collector(zone());
2039     Target target(&this->target_stack_, &collector);
2040     Target target_body(&this->target_stack_, body);
2041 
2042     while (peek() != Token::RBRACE) {
2043       Statement* stat = ParseBlockElement(NULL, CHECK_OK);
2044       if (stat && !stat->IsEmpty()) {
2045         body->AddStatement(stat, zone());
2046       }
2047     }
2048   }
2049   Expect(Token::RBRACE, CHECK_OK);
2050   block_scope->set_end_position(scanner()->location().end_pos);
2051   block_scope = block_scope->FinalizeBlockScope();
2052   body->set_scope(block_scope);
2053   return body;
2054 }
2055 
2056 
ParseVariableStatement(VariableDeclarationContext var_context,ZoneList<const AstRawString * > * names,bool * ok)2057 Block* Parser::ParseVariableStatement(VariableDeclarationContext var_context,
2058                                       ZoneList<const AstRawString*>* names,
2059                                       bool* ok) {
2060   // VariableStatement ::
2061   //   VariableDeclarations ';'
2062 
2063   const AstRawString* ignore;
2064   Block* result =
2065       ParseVariableDeclarations(var_context, NULL, names, &ignore, CHECK_OK);
2066   ExpectSemicolon(CHECK_OK);
2067   return result;
2068 }
2069 
2070 
2071 // If the variable declaration declares exactly one non-const
2072 // variable, then *out is set to that variable. In all other cases,
2073 // *out is untouched; in particular, it is the caller's responsibility
2074 // to initialize it properly. This mechanism is used for the parsing
2075 // of 'for-in' loops.
ParseVariableDeclarations(VariableDeclarationContext var_context,VariableDeclarationProperties * decl_props,ZoneList<const AstRawString * > * names,const AstRawString ** out,bool * ok)2076 Block* Parser::ParseVariableDeclarations(
2077     VariableDeclarationContext var_context,
2078     VariableDeclarationProperties* decl_props,
2079     ZoneList<const AstRawString*>* names,
2080     const AstRawString** out,
2081     bool* ok) {
2082   // VariableDeclarations ::
2083   //   ('var' | 'const' | 'let') (Identifier ('=' AssignmentExpression)?)+[',']
2084   //
2085   // The ES6 Draft Rev3 specifies the following grammar for const declarations
2086   //
2087   // ConstDeclaration ::
2088   //   const ConstBinding (',' ConstBinding)* ';'
2089   // ConstBinding ::
2090   //   Identifier '=' AssignmentExpression
2091   //
2092   // TODO(ES6):
2093   // ConstBinding ::
2094   //   BindingPattern '=' AssignmentExpression
2095 
2096   int pos = peek_position();
2097   VariableMode mode = VAR;
2098   // True if the binding needs initialization. 'let' and 'const' declared
2099   // bindings are created uninitialized by their declaration nodes and
2100   // need initialization. 'var' declared bindings are always initialized
2101   // immediately by their declaration nodes.
2102   bool needs_init = false;
2103   bool is_const = false;
2104   Token::Value init_op = Token::INIT_VAR;
2105   if (peek() == Token::VAR) {
2106     Consume(Token::VAR);
2107   } else if (peek() == Token::CONST) {
2108     // TODO(ES6): The ES6 Draft Rev4 section 12.2.2 reads:
2109     //
2110     // ConstDeclaration : const ConstBinding (',' ConstBinding)* ';'
2111     //
2112     // * It is a Syntax Error if the code that matches this production is not
2113     //   contained in extended code.
2114     //
2115     // However disallowing const in sloppy mode will break compatibility with
2116     // existing pages. Therefore we keep allowing const with the old
2117     // non-harmony semantics in sloppy mode.
2118     Consume(Token::CONST);
2119     switch (strict_mode()) {
2120       case SLOPPY:
2121         mode = CONST_LEGACY;
2122         init_op = Token::INIT_CONST_LEGACY;
2123         break;
2124       case STRICT:
2125         if (allow_harmony_scoping()) {
2126           if (var_context == kStatement) {
2127             // In strict mode 'const' declarations are only allowed in source
2128             // element positions.
2129             ReportMessage("unprotected_const");
2130             *ok = false;
2131             return NULL;
2132           }
2133           mode = CONST;
2134           init_op = Token::INIT_CONST;
2135         } else {
2136           ReportMessage("strict_const");
2137           *ok = false;
2138           return NULL;
2139         }
2140     }
2141     is_const = true;
2142     needs_init = true;
2143   } else if (peek() == Token::LET && strict_mode() == STRICT) {
2144     DCHECK(allow_harmony_scoping());
2145     Consume(Token::LET);
2146     if (var_context == kStatement) {
2147       // Let declarations are only allowed in source element positions.
2148       ReportMessage("unprotected_let");
2149       *ok = false;
2150       return NULL;
2151     }
2152     mode = LET;
2153     needs_init = true;
2154     init_op = Token::INIT_LET;
2155   } else {
2156     UNREACHABLE();  // by current callers
2157   }
2158 
2159   Scope* declaration_scope = DeclarationScope(mode);
2160 
2161   // The scope of a var/const declared variable anywhere inside a function
2162   // is the entire function (ECMA-262, 3rd, 10.1.3, and 12.2). Thus we can
2163   // transform a source-level var/const declaration into a (Function)
2164   // Scope declaration, and rewrite the source-level initialization into an
2165   // assignment statement. We use a block to collect multiple assignments.
2166   //
2167   // We mark the block as initializer block because we don't want the
2168   // rewriter to add a '.result' assignment to such a block (to get compliant
2169   // behavior for code such as print(eval('var x = 7')), and for cosmetic
2170   // reasons when pretty-printing. Also, unless an assignment (initialization)
2171   // is inside an initializer block, it is ignored.
2172   //
2173   // Create new block with one expected declaration.
2174   Block* block = factory()->NewBlock(NULL, 1, true, pos);
2175   int nvars = 0;  // the number of variables declared
2176   const AstRawString* name = NULL;
2177   do {
2178     if (fni_ != NULL) fni_->Enter();
2179 
2180     // Parse variable name.
2181     if (nvars > 0) Consume(Token::COMMA);
2182     name = ParseIdentifier(kDontAllowEvalOrArguments, CHECK_OK);
2183     if (fni_ != NULL) fni_->PushVariableName(name);
2184 
2185     // Declare variable.
2186     // Note that we *always* must treat the initial value via a separate init
2187     // assignment for variables and constants because the value must be assigned
2188     // when the variable is encountered in the source. But the variable/constant
2189     // is declared (and set to 'undefined') upon entering the function within
2190     // which the variable or constant is declared. Only function variables have
2191     // an initial value in the declaration (because they are initialized upon
2192     // entering the function).
2193     //
2194     // If we have a const declaration, in an inner scope, the proxy is always
2195     // bound to the declared variable (independent of possibly surrounding with
2196     // statements).
2197     // For let/const declarations in harmony mode, we can also immediately
2198     // pre-resolve the proxy because it resides in the same scope as the
2199     // declaration.
2200     Interface* interface =
2201         is_const ? Interface::NewConst() : Interface::NewValue();
2202     VariableProxy* proxy = NewUnresolved(name, mode, interface);
2203     Declaration* declaration =
2204         factory()->NewVariableDeclaration(proxy, mode, scope_, pos);
2205     Declare(declaration, mode != VAR, CHECK_OK);
2206     nvars++;
2207     if (declaration_scope->num_var_or_const() > kMaxNumFunctionLocals) {
2208       ReportMessage("too_many_variables");
2209       *ok = false;
2210       return NULL;
2211     }
2212     if (names) names->Add(name, zone());
2213 
2214     // Parse initialization expression if present and/or needed. A
2215     // declaration of the form:
2216     //
2217     //    var v = x;
2218     //
2219     // is syntactic sugar for:
2220     //
2221     //    var v; v = x;
2222     //
2223     // In particular, we need to re-lookup 'v' (in scope_, not
2224     // declaration_scope) as it may be a different 'v' than the 'v' in the
2225     // declaration (e.g., if we are inside a 'with' statement or 'catch'
2226     // block).
2227     //
2228     // However, note that const declarations are different! A const
2229     // declaration of the form:
2230     //
2231     //   const c = x;
2232     //
2233     // is *not* syntactic sugar for:
2234     //
2235     //   const c; c = x;
2236     //
2237     // The "variable" c initialized to x is the same as the declared
2238     // one - there is no re-lookup (see the last parameter of the
2239     // Declare() call above).
2240 
2241     Scope* initialization_scope = is_const ? declaration_scope : scope_;
2242     Expression* value = NULL;
2243     int pos = -1;
2244     // Harmony consts have non-optional initializers.
2245     if (peek() == Token::ASSIGN || mode == CONST) {
2246       Expect(Token::ASSIGN, CHECK_OK);
2247       pos = position();
2248       value = ParseAssignmentExpression(var_context != kForStatement, CHECK_OK);
2249       // Don't infer if it is "a = function(){...}();"-like expression.
2250       if (fni_ != NULL &&
2251           value->AsCall() == NULL &&
2252           value->AsCallNew() == NULL) {
2253         fni_->Infer();
2254       } else {
2255         fni_->RemoveLastFunction();
2256       }
2257       if (decl_props != NULL) *decl_props = kHasInitializers;
2258     }
2259 
2260     // Record the end position of the initializer.
2261     if (proxy->var() != NULL) {
2262       proxy->var()->set_initializer_position(position());
2263     }
2264 
2265     // Make sure that 'const x' and 'let x' initialize 'x' to undefined.
2266     if (value == NULL && needs_init) {
2267       value = GetLiteralUndefined(position());
2268     }
2269 
2270     // Global variable declarations must be compiled in a specific
2271     // way. When the script containing the global variable declaration
2272     // is entered, the global variable must be declared, so that if it
2273     // doesn't exist (on the global object itself, see ES5 errata) it
2274     // gets created with an initial undefined value. This is handled
2275     // by the declarations part of the function representing the
2276     // top-level global code; see Runtime::DeclareGlobalVariable. If
2277     // it already exists (in the object or in a prototype), it is
2278     // *not* touched until the variable declaration statement is
2279     // executed.
2280     //
2281     // Executing the variable declaration statement will always
2282     // guarantee to give the global object an own property.
2283     // This way, global variable declarations can shadow
2284     // properties in the prototype chain, but only after the variable
2285     // declaration statement has been executed. This is important in
2286     // browsers where the global object (window) has lots of
2287     // properties defined in prototype objects.
2288     if (initialization_scope->is_global_scope() &&
2289         !IsLexicalVariableMode(mode)) {
2290       // Compute the arguments for the runtime call.
2291       ZoneList<Expression*>* arguments =
2292           new(zone()) ZoneList<Expression*>(3, zone());
2293       // We have at least 1 parameter.
2294       arguments->Add(factory()->NewStringLiteral(name, pos), zone());
2295       CallRuntime* initialize;
2296 
2297       if (is_const) {
2298         arguments->Add(value, zone());
2299         value = NULL;  // zap the value to avoid the unnecessary assignment
2300 
2301         // Construct the call to Runtime_InitializeConstGlobal
2302         // and add it to the initialization statement block.
2303         // Note that the function does different things depending on
2304         // the number of arguments (1 or 2).
2305         initialize = factory()->NewCallRuntime(
2306             ast_value_factory()->initialize_const_global_string(),
2307             Runtime::FunctionForId(Runtime::kInitializeConstGlobal), arguments,
2308             pos);
2309       } else {
2310         // Add strict mode.
2311         // We may want to pass singleton to avoid Literal allocations.
2312         StrictMode strict_mode = initialization_scope->strict_mode();
2313         arguments->Add(factory()->NewNumberLiteral(strict_mode, pos), zone());
2314 
2315         // Be careful not to assign a value to the global variable if
2316         // we're in a with. The initialization value should not
2317         // necessarily be stored in the global object in that case,
2318         // which is why we need to generate a separate assignment node.
2319         if (value != NULL && !inside_with()) {
2320           arguments->Add(value, zone());
2321           value = NULL;  // zap the value to avoid the unnecessary assignment
2322           // Construct the call to Runtime_InitializeVarGlobal
2323           // and add it to the initialization statement block.
2324           initialize = factory()->NewCallRuntime(
2325               ast_value_factory()->initialize_var_global_string(),
2326               Runtime::FunctionForId(Runtime::kInitializeVarGlobal), arguments,
2327               pos);
2328         } else {
2329           initialize = NULL;
2330         }
2331       }
2332 
2333       if (initialize != NULL) {
2334         block->AddStatement(factory()->NewExpressionStatement(
2335                                 initialize, RelocInfo::kNoPosition),
2336                             zone());
2337       }
2338     } else if (needs_init) {
2339       // Constant initializations always assign to the declared constant which
2340       // is always at the function scope level. This is only relevant for
2341       // dynamically looked-up variables and constants (the start context for
2342       // constant lookups is always the function context, while it is the top
2343       // context for var declared variables). Sigh...
2344       // For 'let' and 'const' declared variables in harmony mode the
2345       // initialization also always assigns to the declared variable.
2346       DCHECK(proxy != NULL);
2347       DCHECK(proxy->var() != NULL);
2348       DCHECK(value != NULL);
2349       Assignment* assignment =
2350           factory()->NewAssignment(init_op, proxy, value, pos);
2351       block->AddStatement(
2352           factory()->NewExpressionStatement(assignment, RelocInfo::kNoPosition),
2353           zone());
2354       value = NULL;
2355     }
2356 
2357     // Add an assignment node to the initialization statement block if we still
2358     // have a pending initialization value.
2359     if (value != NULL) {
2360       DCHECK(mode == VAR);
2361       // 'var' initializations are simply assignments (with all the consequences
2362       // if they are inside a 'with' statement - they may change a 'with' object
2363       // property).
2364       VariableProxy* proxy =
2365           initialization_scope->NewUnresolved(factory(), name, interface);
2366       Assignment* assignment =
2367           factory()->NewAssignment(init_op, proxy, value, pos);
2368       block->AddStatement(
2369           factory()->NewExpressionStatement(assignment, RelocInfo::kNoPosition),
2370           zone());
2371     }
2372 
2373     if (fni_ != NULL) fni_->Leave();
2374   } while (peek() == Token::COMMA);
2375 
2376   // If there was a single non-const declaration, return it in the output
2377   // parameter for possible use by for/in.
2378   if (nvars == 1 && !is_const) {
2379     *out = name;
2380   }
2381 
2382   return block;
2383 }
2384 
2385 
ContainsLabel(ZoneList<const AstRawString * > * labels,const AstRawString * label)2386 static bool ContainsLabel(ZoneList<const AstRawString*>* labels,
2387                           const AstRawString* label) {
2388   DCHECK(label != NULL);
2389   if (labels != NULL) {
2390     for (int i = labels->length(); i-- > 0; ) {
2391       if (labels->at(i) == label) {
2392         return true;
2393       }
2394     }
2395   }
2396   return false;
2397 }
2398 
2399 
ParseExpressionOrLabelledStatement(ZoneList<const AstRawString * > * labels,bool * ok)2400 Statement* Parser::ParseExpressionOrLabelledStatement(
2401     ZoneList<const AstRawString*>* labels, bool* ok) {
2402   // ExpressionStatement | LabelledStatement ::
2403   //   Expression ';'
2404   //   Identifier ':' Statement
2405   int pos = peek_position();
2406   bool starts_with_idenfifier = peek_any_identifier();
2407   Expression* expr = ParseExpression(true, CHECK_OK);
2408   if (peek() == Token::COLON && starts_with_idenfifier && expr != NULL &&
2409       expr->AsVariableProxy() != NULL &&
2410       !expr->AsVariableProxy()->is_this()) {
2411     // Expression is a single identifier, and not, e.g., a parenthesized
2412     // identifier.
2413     VariableProxy* var = expr->AsVariableProxy();
2414     const AstRawString* label = var->raw_name();
2415     // TODO(1240780): We don't check for redeclaration of labels
2416     // during preparsing since keeping track of the set of active
2417     // labels requires nontrivial changes to the way scopes are
2418     // structured.  However, these are probably changes we want to
2419     // make later anyway so we should go back and fix this then.
2420     if (ContainsLabel(labels, label) || TargetStackContainsLabel(label)) {
2421       ParserTraits::ReportMessage("label_redeclaration", label);
2422       *ok = false;
2423       return NULL;
2424     }
2425     if (labels == NULL) {
2426       labels = new(zone()) ZoneList<const AstRawString*>(4, zone());
2427     }
2428     labels->Add(label, zone());
2429     // Remove the "ghost" variable that turned out to be a label
2430     // from the top scope. This way, we don't try to resolve it
2431     // during the scope processing.
2432     scope_->RemoveUnresolved(var);
2433     Expect(Token::COLON, CHECK_OK);
2434     return ParseStatement(labels, ok);
2435   }
2436 
2437   // If we have an extension, we allow a native function declaration.
2438   // A native function declaration starts with "native function" with
2439   // no line-terminator between the two words.
2440   if (extension_ != NULL && peek() == Token::FUNCTION &&
2441       !scanner()->HasAnyLineTerminatorBeforeNext() && expr != NULL &&
2442       expr->AsVariableProxy() != NULL &&
2443       expr->AsVariableProxy()->raw_name() ==
2444           ast_value_factory()->native_string() &&
2445       !scanner()->literal_contains_escapes()) {
2446     return ParseNativeDeclaration(ok);
2447   }
2448 
2449   // Parsed expression statement, or the context-sensitive 'module' keyword.
2450   // Only expect semicolon in the former case.
2451   if (!FLAG_harmony_modules || peek() != Token::IDENTIFIER ||
2452       scanner()->HasAnyLineTerminatorBeforeNext() ||
2453       expr->AsVariableProxy() == NULL ||
2454       expr->AsVariableProxy()->raw_name() !=
2455           ast_value_factory()->module_string() ||
2456       scanner()->literal_contains_escapes()) {
2457     ExpectSemicolon(CHECK_OK);
2458   }
2459   return factory()->NewExpressionStatement(expr, pos);
2460 }
2461 
2462 
ParseIfStatement(ZoneList<const AstRawString * > * labels,bool * ok)2463 IfStatement* Parser::ParseIfStatement(ZoneList<const AstRawString*>* labels,
2464                                       bool* ok) {
2465   // IfStatement ::
2466   //   'if' '(' Expression ')' Statement ('else' Statement)?
2467 
2468   int pos = peek_position();
2469   Expect(Token::IF, CHECK_OK);
2470   Expect(Token::LPAREN, CHECK_OK);
2471   Expression* condition = ParseExpression(true, CHECK_OK);
2472   Expect(Token::RPAREN, CHECK_OK);
2473   Statement* then_statement = ParseStatement(labels, CHECK_OK);
2474   Statement* else_statement = NULL;
2475   if (peek() == Token::ELSE) {
2476     Next();
2477     else_statement = ParseStatement(labels, CHECK_OK);
2478   } else {
2479     else_statement = factory()->NewEmptyStatement(RelocInfo::kNoPosition);
2480   }
2481   return factory()->NewIfStatement(
2482       condition, then_statement, else_statement, pos);
2483 }
2484 
2485 
ParseContinueStatement(bool * ok)2486 Statement* Parser::ParseContinueStatement(bool* ok) {
2487   // ContinueStatement ::
2488   //   'continue' Identifier? ';'
2489 
2490   int pos = peek_position();
2491   Expect(Token::CONTINUE, CHECK_OK);
2492   const AstRawString* label = NULL;
2493   Token::Value tok = peek();
2494   if (!scanner()->HasAnyLineTerminatorBeforeNext() &&
2495       tok != Token::SEMICOLON && tok != Token::RBRACE && tok != Token::EOS) {
2496     // ECMA allows "eval" or "arguments" as labels even in strict mode.
2497     label = ParseIdentifier(kAllowEvalOrArguments, CHECK_OK);
2498   }
2499   IterationStatement* target = LookupContinueTarget(label, CHECK_OK);
2500   if (target == NULL) {
2501     // Illegal continue statement.
2502     const char* message = "illegal_continue";
2503     if (label != NULL) {
2504       message = "unknown_label";
2505     }
2506     ParserTraits::ReportMessage(message, label);
2507     *ok = false;
2508     return NULL;
2509   }
2510   ExpectSemicolon(CHECK_OK);
2511   return factory()->NewContinueStatement(target, pos);
2512 }
2513 
2514 
ParseBreakStatement(ZoneList<const AstRawString * > * labels,bool * ok)2515 Statement* Parser::ParseBreakStatement(ZoneList<const AstRawString*>* labels,
2516                                        bool* ok) {
2517   // BreakStatement ::
2518   //   'break' Identifier? ';'
2519 
2520   int pos = peek_position();
2521   Expect(Token::BREAK, CHECK_OK);
2522   const AstRawString* label = NULL;
2523   Token::Value tok = peek();
2524   if (!scanner()->HasAnyLineTerminatorBeforeNext() &&
2525       tok != Token::SEMICOLON && tok != Token::RBRACE && tok != Token::EOS) {
2526     // ECMA allows "eval" or "arguments" as labels even in strict mode.
2527     label = ParseIdentifier(kAllowEvalOrArguments, CHECK_OK);
2528   }
2529   // Parse labeled break statements that target themselves into
2530   // empty statements, e.g. 'l1: l2: l3: break l2;'
2531   if (label != NULL && ContainsLabel(labels, label)) {
2532     ExpectSemicolon(CHECK_OK);
2533     return factory()->NewEmptyStatement(pos);
2534   }
2535   BreakableStatement* target = NULL;
2536   target = LookupBreakTarget(label, CHECK_OK);
2537   if (target == NULL) {
2538     // Illegal break statement.
2539     const char* message = "illegal_break";
2540     if (label != NULL) {
2541       message = "unknown_label";
2542     }
2543     ParserTraits::ReportMessage(message, label);
2544     *ok = false;
2545     return NULL;
2546   }
2547   ExpectSemicolon(CHECK_OK);
2548   return factory()->NewBreakStatement(target, pos);
2549 }
2550 
2551 
ParseReturnStatement(bool * ok)2552 Statement* Parser::ParseReturnStatement(bool* ok) {
2553   // ReturnStatement ::
2554   //   'return' Expression? ';'
2555 
2556   // Consume the return token. It is necessary to do that before
2557   // reporting any errors on it, because of the way errors are
2558   // reported (underlining).
2559   Expect(Token::RETURN, CHECK_OK);
2560   Scanner::Location loc = scanner()->location();
2561 
2562   Token::Value tok = peek();
2563   Statement* result;
2564   Expression* return_value;
2565   if (scanner()->HasAnyLineTerminatorBeforeNext() ||
2566       tok == Token::SEMICOLON ||
2567       tok == Token::RBRACE ||
2568       tok == Token::EOS) {
2569     return_value = GetLiteralUndefined(position());
2570   } else {
2571     return_value = ParseExpression(true, CHECK_OK);
2572   }
2573   ExpectSemicolon(CHECK_OK);
2574   if (is_generator()) {
2575     Expression* generator = factory()->NewVariableProxy(
2576         function_state_->generator_object_variable());
2577     Expression* yield = factory()->NewYield(
2578         generator, return_value, Yield::kFinal, loc.beg_pos);
2579     result = factory()->NewExpressionStatement(yield, loc.beg_pos);
2580   } else {
2581     result = factory()->NewReturnStatement(return_value, loc.beg_pos);
2582   }
2583 
2584   Scope* decl_scope = scope_->DeclarationScope();
2585   if (decl_scope->is_global_scope() || decl_scope->is_eval_scope()) {
2586     ReportMessageAt(loc, "illegal_return");
2587     *ok = false;
2588     return NULL;
2589   }
2590   return result;
2591 }
2592 
2593 
ParseWithStatement(ZoneList<const AstRawString * > * labels,bool * ok)2594 Statement* Parser::ParseWithStatement(ZoneList<const AstRawString*>* labels,
2595                                       bool* ok) {
2596   // WithStatement ::
2597   //   'with' '(' Expression ')' Statement
2598 
2599   Expect(Token::WITH, CHECK_OK);
2600   int pos = position();
2601 
2602   if (strict_mode() == STRICT) {
2603     ReportMessage("strict_mode_with");
2604     *ok = false;
2605     return NULL;
2606   }
2607 
2608   Expect(Token::LPAREN, CHECK_OK);
2609   Expression* expr = ParseExpression(true, CHECK_OK);
2610   Expect(Token::RPAREN, CHECK_OK);
2611 
2612   scope_->DeclarationScope()->RecordWithStatement();
2613   Scope* with_scope = NewScope(scope_, WITH_SCOPE);
2614   Statement* stmt;
2615   { BlockState block_state(&scope_, with_scope);
2616     with_scope->set_start_position(scanner()->peek_location().beg_pos);
2617     stmt = ParseStatement(labels, CHECK_OK);
2618     with_scope->set_end_position(scanner()->location().end_pos);
2619   }
2620   return factory()->NewWithStatement(with_scope, expr, stmt, pos);
2621 }
2622 
2623 
ParseCaseClause(bool * default_seen_ptr,bool * ok)2624 CaseClause* Parser::ParseCaseClause(bool* default_seen_ptr, bool* ok) {
2625   // CaseClause ::
2626   //   'case' Expression ':' Statement*
2627   //   'default' ':' Statement*
2628 
2629   Expression* label = NULL;  // NULL expression indicates default case
2630   if (peek() == Token::CASE) {
2631     Expect(Token::CASE, CHECK_OK);
2632     label = ParseExpression(true, CHECK_OK);
2633   } else {
2634     Expect(Token::DEFAULT, CHECK_OK);
2635     if (*default_seen_ptr) {
2636       ReportMessage("multiple_defaults_in_switch");
2637       *ok = false;
2638       return NULL;
2639     }
2640     *default_seen_ptr = true;
2641   }
2642   Expect(Token::COLON, CHECK_OK);
2643   int pos = position();
2644   ZoneList<Statement*>* statements =
2645       new(zone()) ZoneList<Statement*>(5, zone());
2646   while (peek() != Token::CASE &&
2647          peek() != Token::DEFAULT &&
2648          peek() != Token::RBRACE) {
2649     Statement* stat = ParseStatement(NULL, CHECK_OK);
2650     statements->Add(stat, zone());
2651   }
2652 
2653   return factory()->NewCaseClause(label, statements, pos);
2654 }
2655 
2656 
ParseSwitchStatement(ZoneList<const AstRawString * > * labels,bool * ok)2657 SwitchStatement* Parser::ParseSwitchStatement(
2658     ZoneList<const AstRawString*>* labels, bool* ok) {
2659   // SwitchStatement ::
2660   //   'switch' '(' Expression ')' '{' CaseClause* '}'
2661 
2662   SwitchStatement* statement =
2663       factory()->NewSwitchStatement(labels, peek_position());
2664   Target target(&this->target_stack_, statement);
2665 
2666   Expect(Token::SWITCH, CHECK_OK);
2667   Expect(Token::LPAREN, CHECK_OK);
2668   Expression* tag = ParseExpression(true, CHECK_OK);
2669   Expect(Token::RPAREN, CHECK_OK);
2670 
2671   bool default_seen = false;
2672   ZoneList<CaseClause*>* cases = new(zone()) ZoneList<CaseClause*>(4, zone());
2673   Expect(Token::LBRACE, CHECK_OK);
2674   while (peek() != Token::RBRACE) {
2675     CaseClause* clause = ParseCaseClause(&default_seen, CHECK_OK);
2676     cases->Add(clause, zone());
2677   }
2678   Expect(Token::RBRACE, CHECK_OK);
2679 
2680   if (statement) statement->Initialize(tag, cases);
2681   return statement;
2682 }
2683 
2684 
ParseThrowStatement(bool * ok)2685 Statement* Parser::ParseThrowStatement(bool* ok) {
2686   // ThrowStatement ::
2687   //   'throw' Expression ';'
2688 
2689   Expect(Token::THROW, CHECK_OK);
2690   int pos = position();
2691   if (scanner()->HasAnyLineTerminatorBeforeNext()) {
2692     ReportMessage("newline_after_throw");
2693     *ok = false;
2694     return NULL;
2695   }
2696   Expression* exception = ParseExpression(true, CHECK_OK);
2697   ExpectSemicolon(CHECK_OK);
2698 
2699   return factory()->NewExpressionStatement(
2700       factory()->NewThrow(exception, pos), pos);
2701 }
2702 
2703 
ParseTryStatement(bool * ok)2704 TryStatement* Parser::ParseTryStatement(bool* ok) {
2705   // TryStatement ::
2706   //   'try' Block Catch
2707   //   'try' Block Finally
2708   //   'try' Block Catch Finally
2709   //
2710   // Catch ::
2711   //   'catch' '(' Identifier ')' Block
2712   //
2713   // Finally ::
2714   //   'finally' Block
2715 
2716   Expect(Token::TRY, CHECK_OK);
2717   int pos = position();
2718 
2719   TargetCollector try_collector(zone());
2720   Block* try_block;
2721 
2722   { Target target(&this->target_stack_, &try_collector);
2723     try_block = ParseBlock(NULL, CHECK_OK);
2724   }
2725 
2726   Token::Value tok = peek();
2727   if (tok != Token::CATCH && tok != Token::FINALLY) {
2728     ReportMessage("no_catch_or_finally");
2729     *ok = false;
2730     return NULL;
2731   }
2732 
2733   // If we can break out from the catch block and there is a finally block,
2734   // then we will need to collect escaping targets from the catch
2735   // block. Since we don't know yet if there will be a finally block, we
2736   // always collect the targets.
2737   TargetCollector catch_collector(zone());
2738   Scope* catch_scope = NULL;
2739   Variable* catch_variable = NULL;
2740   Block* catch_block = NULL;
2741   const AstRawString* name = NULL;
2742   if (tok == Token::CATCH) {
2743     Consume(Token::CATCH);
2744 
2745     Expect(Token::LPAREN, CHECK_OK);
2746     catch_scope = NewScope(scope_, CATCH_SCOPE);
2747     catch_scope->set_start_position(scanner()->location().beg_pos);
2748     name = ParseIdentifier(kDontAllowEvalOrArguments, CHECK_OK);
2749 
2750     Expect(Token::RPAREN, CHECK_OK);
2751 
2752     Target target(&this->target_stack_, &catch_collector);
2753     VariableMode mode =
2754         allow_harmony_scoping() && strict_mode() == STRICT ? LET : VAR;
2755     catch_variable = catch_scope->DeclareLocal(name, mode, kCreatedInitialized);
2756     BlockState block_state(&scope_, catch_scope);
2757     catch_block = ParseBlock(NULL, CHECK_OK);
2758 
2759     catch_scope->set_end_position(scanner()->location().end_pos);
2760     tok = peek();
2761   }
2762 
2763   Block* finally_block = NULL;
2764   DCHECK(tok == Token::FINALLY || catch_block != NULL);
2765   if (tok == Token::FINALLY) {
2766     Consume(Token::FINALLY);
2767     finally_block = ParseBlock(NULL, CHECK_OK);
2768   }
2769 
2770   // Simplify the AST nodes by converting:
2771   //   'try B0 catch B1 finally B2'
2772   // to:
2773   //   'try { try B0 catch B1 } finally B2'
2774 
2775   if (catch_block != NULL && finally_block != NULL) {
2776     // If we have both, create an inner try/catch.
2777     DCHECK(catch_scope != NULL && catch_variable != NULL);
2778     int index = function_state_->NextHandlerIndex();
2779     TryCatchStatement* statement = factory()->NewTryCatchStatement(
2780         index, try_block, catch_scope, catch_variable, catch_block,
2781         RelocInfo::kNoPosition);
2782     statement->set_escaping_targets(try_collector.targets());
2783     try_block = factory()->NewBlock(NULL, 1, false, RelocInfo::kNoPosition);
2784     try_block->AddStatement(statement, zone());
2785     catch_block = NULL;  // Clear to indicate it's been handled.
2786   }
2787 
2788   TryStatement* result = NULL;
2789   if (catch_block != NULL) {
2790     DCHECK(finally_block == NULL);
2791     DCHECK(catch_scope != NULL && catch_variable != NULL);
2792     int index = function_state_->NextHandlerIndex();
2793     result = factory()->NewTryCatchStatement(
2794         index, try_block, catch_scope, catch_variable, catch_block, pos);
2795   } else {
2796     DCHECK(finally_block != NULL);
2797     int index = function_state_->NextHandlerIndex();
2798     result = factory()->NewTryFinallyStatement(
2799         index, try_block, finally_block, pos);
2800     // Combine the jump targets of the try block and the possible catch block.
2801     try_collector.targets()->AddAll(*catch_collector.targets(), zone());
2802   }
2803 
2804   result->set_escaping_targets(try_collector.targets());
2805   return result;
2806 }
2807 
2808 
ParseDoWhileStatement(ZoneList<const AstRawString * > * labels,bool * ok)2809 DoWhileStatement* Parser::ParseDoWhileStatement(
2810     ZoneList<const AstRawString*>* labels, bool* ok) {
2811   // DoStatement ::
2812   //   'do' Statement 'while' '(' Expression ')' ';'
2813 
2814   DoWhileStatement* loop =
2815       factory()->NewDoWhileStatement(labels, peek_position());
2816   Target target(&this->target_stack_, loop);
2817 
2818   Expect(Token::DO, CHECK_OK);
2819   Statement* body = ParseStatement(NULL, CHECK_OK);
2820   Expect(Token::WHILE, CHECK_OK);
2821   Expect(Token::LPAREN, CHECK_OK);
2822 
2823   Expression* cond = ParseExpression(true, CHECK_OK);
2824   Expect(Token::RPAREN, CHECK_OK);
2825 
2826   // Allow do-statements to be terminated with and without
2827   // semi-colons. This allows code such as 'do;while(0)return' to
2828   // parse, which would not be the case if we had used the
2829   // ExpectSemicolon() functionality here.
2830   if (peek() == Token::SEMICOLON) Consume(Token::SEMICOLON);
2831 
2832   if (loop != NULL) loop->Initialize(cond, body);
2833   return loop;
2834 }
2835 
2836 
ParseWhileStatement(ZoneList<const AstRawString * > * labels,bool * ok)2837 WhileStatement* Parser::ParseWhileStatement(
2838     ZoneList<const AstRawString*>* labels, bool* ok) {
2839   // WhileStatement ::
2840   //   'while' '(' Expression ')' Statement
2841 
2842   WhileStatement* loop = factory()->NewWhileStatement(labels, peek_position());
2843   Target target(&this->target_stack_, loop);
2844 
2845   Expect(Token::WHILE, CHECK_OK);
2846   Expect(Token::LPAREN, CHECK_OK);
2847   Expression* cond = ParseExpression(true, CHECK_OK);
2848   Expect(Token::RPAREN, CHECK_OK);
2849   Statement* body = ParseStatement(NULL, CHECK_OK);
2850 
2851   if (loop != NULL) loop->Initialize(cond, body);
2852   return loop;
2853 }
2854 
2855 
CheckInOrOf(bool accept_OF,ForEachStatement::VisitMode * visit_mode)2856 bool Parser::CheckInOrOf(bool accept_OF,
2857                          ForEachStatement::VisitMode* visit_mode) {
2858   if (Check(Token::IN)) {
2859     *visit_mode = ForEachStatement::ENUMERATE;
2860     return true;
2861   } else if (accept_OF && CheckContextualKeyword(CStrVector("of"))) {
2862     *visit_mode = ForEachStatement::ITERATE;
2863     return true;
2864   }
2865   return false;
2866 }
2867 
2868 
InitializeForEachStatement(ForEachStatement * stmt,Expression * each,Expression * subject,Statement * body)2869 void Parser::InitializeForEachStatement(ForEachStatement* stmt,
2870                                         Expression* each,
2871                                         Expression* subject,
2872                                         Statement* body) {
2873   ForOfStatement* for_of = stmt->AsForOfStatement();
2874 
2875   if (for_of != NULL) {
2876     Variable* iterator = scope_->DeclarationScope()->NewTemporary(
2877         ast_value_factory()->dot_iterator_string());
2878     Variable* result = scope_->DeclarationScope()->NewTemporary(
2879         ast_value_factory()->dot_result_string());
2880 
2881     Expression* assign_iterator;
2882     Expression* next_result;
2883     Expression* result_done;
2884     Expression* assign_each;
2885 
2886     // var iterator = subject[Symbol.iterator]();
2887     assign_iterator = factory()->NewAssignment(
2888         Token::ASSIGN, factory()->NewVariableProxy(iterator),
2889         GetIterator(subject, factory()), RelocInfo::kNoPosition);
2890 
2891     // var result = iterator.next();
2892     {
2893       Expression* iterator_proxy = factory()->NewVariableProxy(iterator);
2894       Expression* next_literal = factory()->NewStringLiteral(
2895           ast_value_factory()->next_string(), RelocInfo::kNoPosition);
2896       Expression* next_property = factory()->NewProperty(
2897           iterator_proxy, next_literal, RelocInfo::kNoPosition);
2898       ZoneList<Expression*>* next_arguments =
2899           new(zone()) ZoneList<Expression*>(0, zone());
2900       Expression* next_call = factory()->NewCall(
2901           next_property, next_arguments, RelocInfo::kNoPosition);
2902       Expression* result_proxy = factory()->NewVariableProxy(result);
2903       next_result = factory()->NewAssignment(
2904           Token::ASSIGN, result_proxy, next_call, RelocInfo::kNoPosition);
2905     }
2906 
2907     // result.done
2908     {
2909       Expression* done_literal = factory()->NewStringLiteral(
2910           ast_value_factory()->done_string(), RelocInfo::kNoPosition);
2911       Expression* result_proxy = factory()->NewVariableProxy(result);
2912       result_done = factory()->NewProperty(
2913           result_proxy, done_literal, RelocInfo::kNoPosition);
2914     }
2915 
2916     // each = result.value
2917     {
2918       Expression* value_literal = factory()->NewStringLiteral(
2919           ast_value_factory()->value_string(), RelocInfo::kNoPosition);
2920       Expression* result_proxy = factory()->NewVariableProxy(result);
2921       Expression* result_value = factory()->NewProperty(
2922           result_proxy, value_literal, RelocInfo::kNoPosition);
2923       assign_each = factory()->NewAssignment(
2924           Token::ASSIGN, each, result_value, RelocInfo::kNoPosition);
2925     }
2926 
2927     for_of->Initialize(each, subject, body,
2928                        assign_iterator,
2929                        next_result,
2930                        result_done,
2931                        assign_each);
2932   } else {
2933     stmt->Initialize(each, subject, body);
2934   }
2935 }
2936 
2937 
DesugarLetBindingsInForStatement(Scope * inner_scope,ZoneList<const AstRawString * > * names,ForStatement * loop,Statement * init,Expression * cond,Statement * next,Statement * body,bool * ok)2938 Statement* Parser::DesugarLetBindingsInForStatement(
2939     Scope* inner_scope, ZoneList<const AstRawString*>* names,
2940     ForStatement* loop, Statement* init, Expression* cond, Statement* next,
2941     Statement* body, bool* ok) {
2942   // ES6 13.6.3.4 specifies that on each loop iteration the let variables are
2943   // copied into a new environment. After copying, the "next" statement of the
2944   // loop is executed to update the loop variables. The loop condition is
2945   // checked and the loop body is executed.
2946   //
2947   // We rewrite a for statement of the form
2948   //
2949   //  for (let x = i; cond; next) body
2950   //
2951   // into
2952   //
2953   //  {
2954   //     let x = i;
2955   //     temp_x = x;
2956   //     flag = 1;
2957   //     for (;;) {
2958   //        let x = temp_x;
2959   //        if (flag == 1) {
2960   //          flag = 0;
2961   //        } else {
2962   //          next;
2963   //        }
2964   //        if (cond) {
2965   //          <empty>
2966   //        } else {
2967   //          break;
2968   //        }
2969   //        b
2970   //        temp_x = x;
2971   //     }
2972   //  }
2973 
2974   DCHECK(names->length() > 0);
2975   Scope* for_scope = scope_;
2976   ZoneList<Variable*> temps(names->length(), zone());
2977 
2978   Block* outer_block = factory()->NewBlock(NULL, names->length() + 3, false,
2979                                            RelocInfo::kNoPosition);
2980   outer_block->AddStatement(init, zone());
2981 
2982   const AstRawString* temp_name = ast_value_factory()->dot_for_string();
2983 
2984   // For each let variable x:
2985   //   make statement: temp_x = x.
2986   for (int i = 0; i < names->length(); i++) {
2987     VariableProxy* proxy =
2988         NewUnresolved(names->at(i), LET, Interface::NewValue());
2989     Variable* temp = scope_->DeclarationScope()->NewTemporary(temp_name);
2990     VariableProxy* temp_proxy = factory()->NewVariableProxy(temp);
2991     Assignment* assignment = factory()->NewAssignment(
2992         Token::ASSIGN, temp_proxy, proxy, RelocInfo::kNoPosition);
2993     Statement* assignment_statement = factory()->NewExpressionStatement(
2994         assignment, RelocInfo::kNoPosition);
2995     outer_block->AddStatement(assignment_statement, zone());
2996     temps.Add(temp, zone());
2997   }
2998 
2999   Variable* flag = scope_->DeclarationScope()->NewTemporary(temp_name);
3000   // Make statement: flag = 1.
3001   {
3002     VariableProxy* flag_proxy = factory()->NewVariableProxy(flag);
3003     Expression* const1 = factory()->NewSmiLiteral(1, RelocInfo::kNoPosition);
3004     Assignment* assignment = factory()->NewAssignment(
3005         Token::ASSIGN, flag_proxy, const1, RelocInfo::kNoPosition);
3006     Statement* assignment_statement = factory()->NewExpressionStatement(
3007         assignment, RelocInfo::kNoPosition);
3008     outer_block->AddStatement(assignment_statement, zone());
3009   }
3010 
3011   outer_block->AddStatement(loop, zone());
3012   outer_block->set_scope(for_scope);
3013   scope_ = inner_scope;
3014 
3015   Block* inner_block = factory()->NewBlock(NULL, 2 * names->length() + 3,
3016                                            false, RelocInfo::kNoPosition);
3017   int pos = scanner()->location().beg_pos;
3018   ZoneList<Variable*> inner_vars(names->length(), zone());
3019 
3020   // For each let variable x:
3021   //    make statement: let x = temp_x.
3022   for (int i = 0; i < names->length(); i++) {
3023     VariableProxy* proxy =
3024         NewUnresolved(names->at(i), LET, Interface::NewValue());
3025     Declaration* declaration =
3026         factory()->NewVariableDeclaration(proxy, LET, scope_, pos);
3027     Declare(declaration, true, CHECK_OK);
3028     inner_vars.Add(declaration->proxy()->var(), zone());
3029     VariableProxy* temp_proxy = factory()->NewVariableProxy(temps.at(i));
3030     Assignment* assignment = factory()->NewAssignment(
3031         Token::INIT_LET, proxy, temp_proxy, pos);
3032     Statement* assignment_statement = factory()->NewExpressionStatement(
3033         assignment, pos);
3034     proxy->var()->set_initializer_position(pos);
3035     inner_block->AddStatement(assignment_statement, zone());
3036   }
3037 
3038   // Make statement: if (flag == 1) { flag = 0; } else { next; }.
3039   if (next) {
3040     Expression* compare = NULL;
3041     // Make compare expresion: flag == 1.
3042     {
3043       Expression* const1 = factory()->NewSmiLiteral(1, RelocInfo::kNoPosition);
3044       VariableProxy* flag_proxy = factory()->NewVariableProxy(flag);
3045       compare = factory()->NewCompareOperation(
3046           Token::EQ, flag_proxy, const1, pos);
3047     }
3048     Statement* clear_flag = NULL;
3049     // Make statement: flag = 0.
3050     {
3051       VariableProxy* flag_proxy = factory()->NewVariableProxy(flag);
3052       Expression* const0 = factory()->NewSmiLiteral(0, RelocInfo::kNoPosition);
3053       Assignment* assignment = factory()->NewAssignment(
3054           Token::ASSIGN, flag_proxy, const0, RelocInfo::kNoPosition);
3055       clear_flag = factory()->NewExpressionStatement(assignment, pos);
3056     }
3057     Statement* clear_flag_or_next = factory()->NewIfStatement(
3058         compare, clear_flag, next, RelocInfo::kNoPosition);
3059     inner_block->AddStatement(clear_flag_or_next, zone());
3060   }
3061 
3062 
3063   // Make statement: if (cond) { } else { break; }.
3064   if (cond) {
3065     Statement* empty = factory()->NewEmptyStatement(RelocInfo::kNoPosition);
3066     BreakableStatement* t = LookupBreakTarget(NULL, CHECK_OK);
3067     Statement* stop = factory()->NewBreakStatement(t, RelocInfo::kNoPosition);
3068     Statement* if_not_cond_break = factory()->NewIfStatement(
3069         cond, empty, stop, cond->position());
3070     inner_block->AddStatement(if_not_cond_break, zone());
3071   }
3072 
3073   inner_block->AddStatement(body, zone());
3074 
3075   // For each let variable x:
3076   //   make statement: temp_x = x;
3077   for (int i = 0; i < names->length(); i++) {
3078     VariableProxy* temp_proxy = factory()->NewVariableProxy(temps.at(i));
3079     int pos = scanner()->location().end_pos;
3080     VariableProxy* proxy = factory()->NewVariableProxy(inner_vars.at(i), pos);
3081     Assignment* assignment = factory()->NewAssignment(
3082         Token::ASSIGN, temp_proxy, proxy, RelocInfo::kNoPosition);
3083     Statement* assignment_statement = factory()->NewExpressionStatement(
3084         assignment, RelocInfo::kNoPosition);
3085     inner_block->AddStatement(assignment_statement, zone());
3086   }
3087 
3088   inner_scope->set_end_position(scanner()->location().end_pos);
3089   inner_block->set_scope(inner_scope);
3090   scope_ = for_scope;
3091 
3092   loop->Initialize(NULL, NULL, NULL, inner_block);
3093   return outer_block;
3094 }
3095 
3096 
ParseForStatement(ZoneList<const AstRawString * > * labels,bool * ok)3097 Statement* Parser::ParseForStatement(ZoneList<const AstRawString*>* labels,
3098                                      bool* ok) {
3099   // ForStatement ::
3100   //   'for' '(' Expression? ';' Expression? ';' Expression? ')' Statement
3101 
3102   int pos = peek_position();
3103   Statement* init = NULL;
3104   ZoneList<const AstRawString*> let_bindings(1, zone());
3105 
3106   // Create an in-between scope for let-bound iteration variables.
3107   Scope* saved_scope = scope_;
3108   Scope* for_scope = NewScope(scope_, BLOCK_SCOPE);
3109   scope_ = for_scope;
3110 
3111   Expect(Token::FOR, CHECK_OK);
3112   Expect(Token::LPAREN, CHECK_OK);
3113   for_scope->set_start_position(scanner()->location().beg_pos);
3114   if (peek() != Token::SEMICOLON) {
3115     if (peek() == Token::VAR || peek() == Token::CONST) {
3116       bool is_const = peek() == Token::CONST;
3117       const AstRawString* name = NULL;
3118       VariableDeclarationProperties decl_props = kHasNoInitializers;
3119       Block* variable_statement =
3120           ParseVariableDeclarations(kForStatement, &decl_props, NULL, &name,
3121                                     CHECK_OK);
3122       bool accept_OF = decl_props == kHasNoInitializers;
3123       ForEachStatement::VisitMode mode;
3124 
3125       if (name != NULL && CheckInOrOf(accept_OF, &mode)) {
3126         Interface* interface =
3127             is_const ? Interface::NewConst() : Interface::NewValue();
3128         ForEachStatement* loop =
3129             factory()->NewForEachStatement(mode, labels, pos);
3130         Target target(&this->target_stack_, loop);
3131 
3132         Expression* enumerable = ParseExpression(true, CHECK_OK);
3133         Expect(Token::RPAREN, CHECK_OK);
3134 
3135         VariableProxy* each =
3136             scope_->NewUnresolved(factory(), name, interface);
3137         Statement* body = ParseStatement(NULL, CHECK_OK);
3138         InitializeForEachStatement(loop, each, enumerable, body);
3139         Block* result =
3140             factory()->NewBlock(NULL, 2, false, RelocInfo::kNoPosition);
3141         result->AddStatement(variable_statement, zone());
3142         result->AddStatement(loop, zone());
3143         scope_ = saved_scope;
3144         for_scope->set_end_position(scanner()->location().end_pos);
3145         for_scope = for_scope->FinalizeBlockScope();
3146         DCHECK(for_scope == NULL);
3147         // Parsed for-in loop w/ variable/const declaration.
3148         return result;
3149       } else {
3150         init = variable_statement;
3151       }
3152     } else if (peek() == Token::LET && strict_mode() == STRICT) {
3153       DCHECK(allow_harmony_scoping());
3154       const AstRawString* name = NULL;
3155       VariableDeclarationProperties decl_props = kHasNoInitializers;
3156       Block* variable_statement =
3157           ParseVariableDeclarations(kForStatement, &decl_props, &let_bindings,
3158                                     &name, CHECK_OK);
3159       bool accept_IN = name != NULL && decl_props != kHasInitializers;
3160       bool accept_OF = decl_props == kHasNoInitializers;
3161       ForEachStatement::VisitMode mode;
3162 
3163       if (accept_IN && CheckInOrOf(accept_OF, &mode)) {
3164         // Rewrite a for-in statement of the form
3165         //
3166         //   for (let x in e) b
3167         //
3168         // into
3169         //
3170         //   <let x' be a temporary variable>
3171         //   for (x' in e) {
3172         //     let x;
3173         //     x = x';
3174         //     b;
3175         //   }
3176 
3177         // TODO(keuchel): Move the temporary variable to the block scope, after
3178         // implementing stack allocated block scoped variables.
3179         Variable* temp = scope_->DeclarationScope()->NewTemporary(
3180             ast_value_factory()->dot_for_string());
3181         VariableProxy* temp_proxy = factory()->NewVariableProxy(temp);
3182         ForEachStatement* loop =
3183             factory()->NewForEachStatement(mode, labels, pos);
3184         Target target(&this->target_stack_, loop);
3185 
3186         // The expression does not see the loop variable.
3187         scope_ = saved_scope;
3188         Expression* enumerable = ParseExpression(true, CHECK_OK);
3189         scope_ = for_scope;
3190         Expect(Token::RPAREN, CHECK_OK);
3191 
3192         VariableProxy* each =
3193             scope_->NewUnresolved(factory(), name, Interface::NewValue());
3194         Statement* body = ParseStatement(NULL, CHECK_OK);
3195         Block* body_block =
3196             factory()->NewBlock(NULL, 3, false, RelocInfo::kNoPosition);
3197         Assignment* assignment = factory()->NewAssignment(
3198             Token::ASSIGN, each, temp_proxy, RelocInfo::kNoPosition);
3199         Statement* assignment_statement = factory()->NewExpressionStatement(
3200             assignment, RelocInfo::kNoPosition);
3201         body_block->AddStatement(variable_statement, zone());
3202         body_block->AddStatement(assignment_statement, zone());
3203         body_block->AddStatement(body, zone());
3204         InitializeForEachStatement(loop, temp_proxy, enumerable, body_block);
3205         scope_ = saved_scope;
3206         for_scope->set_end_position(scanner()->location().end_pos);
3207         for_scope = for_scope->FinalizeBlockScope();
3208         body_block->set_scope(for_scope);
3209         // Parsed for-in loop w/ let declaration.
3210         return loop;
3211 
3212       } else {
3213         init = variable_statement;
3214       }
3215     } else {
3216       Scanner::Location lhs_location = scanner()->peek_location();
3217       Expression* expression = ParseExpression(false, CHECK_OK);
3218       ForEachStatement::VisitMode mode;
3219       bool accept_OF = expression->AsVariableProxy();
3220 
3221       if (CheckInOrOf(accept_OF, &mode)) {
3222         expression = this->CheckAndRewriteReferenceExpression(
3223             expression, lhs_location, "invalid_lhs_in_for", CHECK_OK);
3224 
3225         ForEachStatement* loop =
3226             factory()->NewForEachStatement(mode, labels, pos);
3227         Target target(&this->target_stack_, loop);
3228 
3229         Expression* enumerable = ParseExpression(true, CHECK_OK);
3230         Expect(Token::RPAREN, CHECK_OK);
3231 
3232         Statement* body = ParseStatement(NULL, CHECK_OK);
3233         InitializeForEachStatement(loop, expression, enumerable, body);
3234         scope_ = saved_scope;
3235         for_scope->set_end_position(scanner()->location().end_pos);
3236         for_scope = for_scope->FinalizeBlockScope();
3237         DCHECK(for_scope == NULL);
3238         // Parsed for-in loop.
3239         return loop;
3240 
3241       } else {
3242         init = factory()->NewExpressionStatement(
3243             expression, RelocInfo::kNoPosition);
3244       }
3245     }
3246   }
3247 
3248   // Standard 'for' loop
3249   ForStatement* loop = factory()->NewForStatement(labels, pos);
3250   Target target(&this->target_stack_, loop);
3251 
3252   // Parsed initializer at this point.
3253   Expect(Token::SEMICOLON, CHECK_OK);
3254 
3255   // If there are let bindings, then condition and the next statement of the
3256   // for loop must be parsed in a new scope.
3257   Scope* inner_scope = NULL;
3258   if (let_bindings.length() > 0) {
3259     inner_scope = NewScope(for_scope, BLOCK_SCOPE);
3260     inner_scope->set_start_position(scanner()->location().beg_pos);
3261     scope_ = inner_scope;
3262   }
3263 
3264   Expression* cond = NULL;
3265   if (peek() != Token::SEMICOLON) {
3266     cond = ParseExpression(true, CHECK_OK);
3267   }
3268   Expect(Token::SEMICOLON, CHECK_OK);
3269 
3270   Statement* next = NULL;
3271   if (peek() != Token::RPAREN) {
3272     Expression* exp = ParseExpression(true, CHECK_OK);
3273     next = factory()->NewExpressionStatement(exp, RelocInfo::kNoPosition);
3274   }
3275   Expect(Token::RPAREN, CHECK_OK);
3276 
3277   Statement* body = ParseStatement(NULL, CHECK_OK);
3278 
3279   Statement* result = NULL;
3280   if (let_bindings.length() > 0) {
3281     scope_ = for_scope;
3282     result = DesugarLetBindingsInForStatement(inner_scope, &let_bindings, loop,
3283                                               init, cond, next, body, CHECK_OK);
3284     scope_ = saved_scope;
3285     for_scope->set_end_position(scanner()->location().end_pos);
3286   } else {
3287     scope_ = saved_scope;
3288     for_scope->set_end_position(scanner()->location().end_pos);
3289     for_scope = for_scope->FinalizeBlockScope();
3290     if (for_scope) {
3291       // Rewrite a for statement of the form
3292       //   for (const x = i; c; n) b
3293       //
3294       // into
3295       //
3296       //   {
3297       //     const x = i;
3298       //     for (; c; n) b
3299       //   }
3300       DCHECK(init != NULL);
3301       Block* block =
3302           factory()->NewBlock(NULL, 2, false, RelocInfo::kNoPosition);
3303       block->AddStatement(init, zone());
3304       block->AddStatement(loop, zone());
3305       block->set_scope(for_scope);
3306       loop->Initialize(NULL, cond, next, body);
3307       result = block;
3308     } else {
3309       loop->Initialize(init, cond, next, body);
3310       result = loop;
3311     }
3312   }
3313   return result;
3314 }
3315 
3316 
ParseDebuggerStatement(bool * ok)3317 DebuggerStatement* Parser::ParseDebuggerStatement(bool* ok) {
3318   // In ECMA-262 'debugger' is defined as a reserved keyword. In some browser
3319   // contexts this is used as a statement which invokes the debugger as i a
3320   // break point is present.
3321   // DebuggerStatement ::
3322   //   'debugger' ';'
3323 
3324   int pos = peek_position();
3325   Expect(Token::DEBUGGER, CHECK_OK);
3326   ExpectSemicolon(CHECK_OK);
3327   return factory()->NewDebuggerStatement(pos);
3328 }
3329 
3330 
IsCompileTimeValue(Expression * expression)3331 bool CompileTimeValue::IsCompileTimeValue(Expression* expression) {
3332   if (expression->IsLiteral()) return true;
3333   MaterializedLiteral* lit = expression->AsMaterializedLiteral();
3334   return lit != NULL && lit->is_simple();
3335 }
3336 
3337 
GetValue(Isolate * isolate,Expression * expression)3338 Handle<FixedArray> CompileTimeValue::GetValue(Isolate* isolate,
3339                                               Expression* expression) {
3340   Factory* factory = isolate->factory();
3341   DCHECK(IsCompileTimeValue(expression));
3342   Handle<FixedArray> result = factory->NewFixedArray(2, TENURED);
3343   ObjectLiteral* object_literal = expression->AsObjectLiteral();
3344   if (object_literal != NULL) {
3345     DCHECK(object_literal->is_simple());
3346     if (object_literal->fast_elements()) {
3347       result->set(kLiteralTypeSlot, Smi::FromInt(OBJECT_LITERAL_FAST_ELEMENTS));
3348     } else {
3349       result->set(kLiteralTypeSlot, Smi::FromInt(OBJECT_LITERAL_SLOW_ELEMENTS));
3350     }
3351     result->set(kElementsSlot, *object_literal->constant_properties());
3352   } else {
3353     ArrayLiteral* array_literal = expression->AsArrayLiteral();
3354     DCHECK(array_literal != NULL && array_literal->is_simple());
3355     result->set(kLiteralTypeSlot, Smi::FromInt(ARRAY_LITERAL));
3356     result->set(kElementsSlot, *array_literal->constant_elements());
3357   }
3358   return result;
3359 }
3360 
3361 
GetLiteralType(Handle<FixedArray> value)3362 CompileTimeValue::LiteralType CompileTimeValue::GetLiteralType(
3363     Handle<FixedArray> value) {
3364   Smi* literal_type = Smi::cast(value->get(kLiteralTypeSlot));
3365   return static_cast<LiteralType>(literal_type->value());
3366 }
3367 
3368 
GetElements(Handle<FixedArray> value)3369 Handle<FixedArray> CompileTimeValue::GetElements(Handle<FixedArray> value) {
3370   return Handle<FixedArray>(FixedArray::cast(value->get(kElementsSlot)));
3371 }
3372 
3373 
CheckAndDeclareArrowParameter(ParserTraits * traits,Expression * expression,Scope * scope,int * num_params,Scanner::Location * dupe_loc)3374 bool CheckAndDeclareArrowParameter(ParserTraits* traits, Expression* expression,
3375                                    Scope* scope, int* num_params,
3376                                    Scanner::Location* dupe_loc) {
3377   // Case for empty parameter lists:
3378   //   () => ...
3379   if (expression == NULL) return true;
3380 
3381   // Too many parentheses around expression:
3382   //   (( ... )) => ...
3383   if (expression->parenthesization_level() > 1) return false;
3384 
3385   // Case for a single parameter:
3386   //   (foo) => ...
3387   //   foo => ...
3388   if (expression->IsVariableProxy()) {
3389     if (expression->AsVariableProxy()->is_this()) return false;
3390 
3391     const AstRawString* raw_name = expression->AsVariableProxy()->raw_name();
3392     if (traits->IsEvalOrArguments(raw_name) ||
3393         traits->IsFutureStrictReserved(raw_name))
3394       return false;
3395 
3396     if (scope->IsDeclared(raw_name)) {
3397       *dupe_loc = Scanner::Location(
3398           expression->position(), expression->position() + raw_name->length());
3399       return false;
3400     }
3401 
3402     scope->DeclareParameter(raw_name, VAR);
3403     ++(*num_params);
3404     return true;
3405   }
3406 
3407   // Case for more than one parameter:
3408   //   (foo, bar [, ...]) => ...
3409   if (expression->IsBinaryOperation()) {
3410     BinaryOperation* binop = expression->AsBinaryOperation();
3411     if (binop->op() != Token::COMMA || binop->left()->is_parenthesized() ||
3412         binop->right()->is_parenthesized())
3413       return false;
3414 
3415     return CheckAndDeclareArrowParameter(traits, binop->left(), scope,
3416                                          num_params, dupe_loc) &&
3417            CheckAndDeclareArrowParameter(traits, binop->right(), scope,
3418                                          num_params, dupe_loc);
3419   }
3420 
3421   // Any other kind of expression is not a valid parameter list.
3422   return false;
3423 }
3424 
3425 
DeclareArrowParametersFromExpression(Expression * expression,Scope * scope,Scanner::Location * dupe_loc,bool * ok)3426 int ParserTraits::DeclareArrowParametersFromExpression(
3427     Expression* expression, Scope* scope, Scanner::Location* dupe_loc,
3428     bool* ok) {
3429   int num_params = 0;
3430   *ok = CheckAndDeclareArrowParameter(this, expression, scope, &num_params,
3431                                       dupe_loc);
3432   return num_params;
3433 }
3434 
3435 
ParseFunctionLiteral(const AstRawString * function_name,Scanner::Location function_name_location,bool name_is_strict_reserved,FunctionKind kind,int function_token_pos,FunctionLiteral::FunctionType function_type,FunctionLiteral::ArityRestriction arity_restriction,bool * ok)3436 FunctionLiteral* Parser::ParseFunctionLiteral(
3437     const AstRawString* function_name, Scanner::Location function_name_location,
3438     bool name_is_strict_reserved, FunctionKind kind, int function_token_pos,
3439     FunctionLiteral::FunctionType function_type,
3440     FunctionLiteral::ArityRestriction arity_restriction, bool* ok) {
3441   // Function ::
3442   //   '(' FormalParameterList? ')' '{' FunctionBody '}'
3443   //
3444   // Getter ::
3445   //   '(' ')' '{' FunctionBody '}'
3446   //
3447   // Setter ::
3448   //   '(' PropertySetParameterList ')' '{' FunctionBody '}'
3449 
3450   int pos = function_token_pos == RelocInfo::kNoPosition
3451       ? peek_position() : function_token_pos;
3452 
3453   bool is_generator = IsGeneratorFunction(kind);
3454 
3455   // Anonymous functions were passed either the empty symbol or a null
3456   // handle as the function name.  Remember if we were passed a non-empty
3457   // handle to decide whether to invoke function name inference.
3458   bool should_infer_name = function_name == NULL;
3459 
3460   // We want a non-null handle as the function name.
3461   if (should_infer_name) {
3462     function_name = ast_value_factory()->empty_string();
3463   }
3464 
3465   int num_parameters = 0;
3466   // Function declarations are function scoped in normal mode, so they are
3467   // hoisted. In harmony block scoping mode they are block scoped, so they
3468   // are not hoisted.
3469   //
3470   // One tricky case are function declarations in a local sloppy-mode eval:
3471   // their declaration is hoisted, but they still see the local scope. E.g.,
3472   //
3473   // function() {
3474   //   var x = 0
3475   //   try { throw 1 } catch (x) { eval("function g() { return x }") }
3476   //   return g()
3477   // }
3478   //
3479   // needs to return 1. To distinguish such cases, we need to detect
3480   // (1) whether a function stems from a sloppy eval, and
3481   // (2) whether it actually hoists across the eval.
3482   // Unfortunately, we do not represent sloppy eval scopes, so we do not have
3483   // either information available directly, especially not when lazily compiling
3484   // a function like 'g'. We hence rely on the following invariants:
3485   // - (1) is the case iff the innermost scope of the deserialized scope chain
3486   //   under which we compile is _not_ a declaration scope. This holds because
3487   //   in all normal cases, function declarations are fully hoisted to a
3488   //   declaration scope and compiled relative to that.
3489   // - (2) is the case iff the current declaration scope is still the original
3490   //   one relative to the deserialized scope chain. Otherwise we must be
3491   //   compiling a function in an inner declaration scope in the eval, e.g. a
3492   //   nested function, and hoisting works normally relative to that.
3493   Scope* declaration_scope = scope_->DeclarationScope();
3494   Scope* original_declaration_scope = original_scope_->DeclarationScope();
3495   Scope* scope =
3496       function_type == FunctionLiteral::DECLARATION &&
3497       (!allow_harmony_scoping() || strict_mode() == SLOPPY) &&
3498       (original_scope_ == original_declaration_scope ||
3499        declaration_scope != original_declaration_scope)
3500           ? NewScope(declaration_scope, FUNCTION_SCOPE)
3501           : NewScope(scope_, FUNCTION_SCOPE);
3502   ZoneList<Statement*>* body = NULL;
3503   int materialized_literal_count = -1;
3504   int expected_property_count = -1;
3505   int handler_count = 0;
3506   FunctionLiteral::ParameterFlag duplicate_parameters =
3507       FunctionLiteral::kNoDuplicateParameters;
3508   FunctionLiteral::IsParenthesizedFlag parenthesized = parenthesized_function_
3509       ? FunctionLiteral::kIsParenthesized
3510       : FunctionLiteral::kNotParenthesized;
3511   AstProperties ast_properties;
3512   BailoutReason dont_optimize_reason = kNoReason;
3513   // Parse function body.
3514   {
3515     FunctionState function_state(&function_state_, &scope_, scope, zone(),
3516                                  ast_value_factory(),
3517                                  info()->ast_node_id_gen());
3518     scope_->SetScopeName(function_name);
3519 
3520     if (is_generator) {
3521       // For generators, allocating variables in contexts is currently a win
3522       // because it minimizes the work needed to suspend and resume an
3523       // activation.
3524       scope_->ForceContextAllocation();
3525 
3526       // Calling a generator returns a generator object.  That object is stored
3527       // in a temporary variable, a definition that is used by "yield"
3528       // expressions. This also marks the FunctionState as a generator.
3529       Variable* temp = scope_->DeclarationScope()->NewTemporary(
3530           ast_value_factory()->dot_generator_object_string());
3531       function_state.set_generator_object_variable(temp);
3532     }
3533 
3534     //  FormalParameterList ::
3535     //    '(' (Identifier)*[','] ')'
3536     Expect(Token::LPAREN, CHECK_OK);
3537     scope->set_start_position(scanner()->location().beg_pos);
3538 
3539     // We don't yet know if the function will be strict, so we cannot yet
3540     // produce errors for parameter names or duplicates. However, we remember
3541     // the locations of these errors if they occur and produce the errors later.
3542     Scanner::Location eval_args_error_log = Scanner::Location::invalid();
3543     Scanner::Location dupe_error_loc = Scanner::Location::invalid();
3544     Scanner::Location reserved_loc = Scanner::Location::invalid();
3545 
3546     bool done = arity_restriction == FunctionLiteral::GETTER_ARITY ||
3547         (peek() == Token::RPAREN &&
3548          arity_restriction != FunctionLiteral::SETTER_ARITY);
3549     while (!done) {
3550       bool is_strict_reserved = false;
3551       const AstRawString* param_name =
3552           ParseIdentifierOrStrictReservedWord(&is_strict_reserved, CHECK_OK);
3553 
3554       // Store locations for possible future error reports.
3555       if (!eval_args_error_log.IsValid() && IsEvalOrArguments(param_name)) {
3556         eval_args_error_log = scanner()->location();
3557       }
3558       if (!reserved_loc.IsValid() && is_strict_reserved) {
3559         reserved_loc = scanner()->location();
3560       }
3561       if (!dupe_error_loc.IsValid() && scope_->IsDeclared(param_name)) {
3562         duplicate_parameters = FunctionLiteral::kHasDuplicateParameters;
3563         dupe_error_loc = scanner()->location();
3564       }
3565 
3566       Variable* var = scope_->DeclareParameter(param_name, VAR);
3567       if (scope->strict_mode() == SLOPPY) {
3568         // TODO(sigurds) Mark every parameter as maybe assigned. This is a
3569         // conservative approximation necessary to account for parameters
3570         // that are assigned via the arguments array.
3571         var->set_maybe_assigned();
3572       }
3573 
3574       num_parameters++;
3575       if (num_parameters > Code::kMaxArguments) {
3576         ReportMessage("too_many_parameters");
3577         *ok = false;
3578         return NULL;
3579       }
3580       if (arity_restriction == FunctionLiteral::SETTER_ARITY) break;
3581       done = (peek() == Token::RPAREN);
3582       if (!done) Expect(Token::COMMA, CHECK_OK);
3583     }
3584     Expect(Token::RPAREN, CHECK_OK);
3585 
3586     Expect(Token::LBRACE, CHECK_OK);
3587 
3588     // If we have a named function expression, we add a local variable
3589     // declaration to the body of the function with the name of the
3590     // function and let it refer to the function itself (closure).
3591     // NOTE: We create a proxy and resolve it here so that in the
3592     // future we can change the AST to only refer to VariableProxies
3593     // instead of Variables and Proxis as is the case now.
3594     Variable* fvar = NULL;
3595     Token::Value fvar_init_op = Token::INIT_CONST_LEGACY;
3596     if (function_type == FunctionLiteral::NAMED_EXPRESSION) {
3597       if (allow_harmony_scoping() && strict_mode() == STRICT) {
3598         fvar_init_op = Token::INIT_CONST;
3599       }
3600       VariableMode fvar_mode =
3601           allow_harmony_scoping() && strict_mode() == STRICT
3602               ? CONST : CONST_LEGACY;
3603       DCHECK(function_name != NULL);
3604       fvar = new (zone())
3605           Variable(scope_, function_name, fvar_mode, true /* is valid LHS */,
3606                    Variable::NORMAL, kCreatedInitialized, kNotAssigned,
3607                    Interface::NewConst());
3608       VariableProxy* proxy = factory()->NewVariableProxy(fvar);
3609       VariableDeclaration* fvar_declaration = factory()->NewVariableDeclaration(
3610           proxy, fvar_mode, scope_, RelocInfo::kNoPosition);
3611       scope_->DeclareFunctionVar(fvar_declaration);
3612     }
3613 
3614     // Determine if the function can be parsed lazily. Lazy parsing is different
3615     // from lazy compilation; we need to parse more eagerly than we compile.
3616 
3617     // We can only parse lazily if we also compile lazily. The heuristics for
3618     // lazy compilation are:
3619     // - It must not have been prohibited by the caller to Parse (some callers
3620     //   need a full AST).
3621     // - The outer scope must allow lazy compilation of inner functions.
3622     // - The function mustn't be a function expression with an open parenthesis
3623     //   before; we consider that a hint that the function will be called
3624     //   immediately, and it would be a waste of time to make it lazily
3625     //   compiled.
3626     // These are all things we can know at this point, without looking at the
3627     // function itself.
3628 
3629     // In addition, we need to distinguish between these cases:
3630     // (function foo() {
3631     //   bar = function() { return 1; }
3632     //  })();
3633     // and
3634     // (function foo() {
3635     //   var a = 1;
3636     //   bar = function() { return a; }
3637     //  })();
3638 
3639     // Now foo will be parsed eagerly and compiled eagerly (optimization: assume
3640     // parenthesis before the function means that it will be called
3641     // immediately). The inner function *must* be parsed eagerly to resolve the
3642     // possible reference to the variable in foo's scope. However, it's possible
3643     // that it will be compiled lazily.
3644 
3645     // To make this additional case work, both Parser and PreParser implement a
3646     // logic where only top-level functions will be parsed lazily.
3647     bool is_lazily_parsed = (mode() == PARSE_LAZILY &&
3648                              scope_->AllowsLazyCompilation() &&
3649                              !parenthesized_function_);
3650     parenthesized_function_ = false;  // The bit was set for this function only.
3651 
3652     if (is_lazily_parsed) {
3653       SkipLazyFunctionBody(function_name, &materialized_literal_count,
3654                            &expected_property_count, CHECK_OK);
3655     } else {
3656       body = ParseEagerFunctionBody(function_name, pos, fvar, fvar_init_op,
3657                                     is_generator, CHECK_OK);
3658       materialized_literal_count = function_state.materialized_literal_count();
3659       expected_property_count = function_state.expected_property_count();
3660       handler_count = function_state.handler_count();
3661     }
3662 
3663     // Validate strict mode.
3664     // Concise methods use StrictFormalParameters.
3665     if (strict_mode() == STRICT || IsConciseMethod(kind)) {
3666       CheckStrictFunctionNameAndParameters(function_name,
3667                                            name_is_strict_reserved,
3668                                            function_name_location,
3669                                            eval_args_error_log,
3670                                            dupe_error_loc,
3671                                            reserved_loc,
3672                                            CHECK_OK);
3673     }
3674     if (strict_mode() == STRICT) {
3675       CheckOctalLiteral(scope->start_position(),
3676                         scope->end_position(),
3677                         CHECK_OK);
3678     }
3679     ast_properties = *factory()->visitor()->ast_properties();
3680     dont_optimize_reason = factory()->visitor()->dont_optimize_reason();
3681 
3682     if (allow_harmony_scoping() && strict_mode() == STRICT) {
3683       CheckConflictingVarDeclarations(scope, CHECK_OK);
3684     }
3685   }
3686 
3687   FunctionLiteral* function_literal = factory()->NewFunctionLiteral(
3688       function_name, ast_value_factory(), scope, body,
3689       materialized_literal_count, expected_property_count, handler_count,
3690       num_parameters, duplicate_parameters, function_type,
3691       FunctionLiteral::kIsFunction, parenthesized, kind, pos);
3692   function_literal->set_function_token_position(function_token_pos);
3693   function_literal->set_ast_properties(&ast_properties);
3694   function_literal->set_dont_optimize_reason(dont_optimize_reason);
3695 
3696   if (fni_ != NULL && should_infer_name) fni_->AddFunction(function_literal);
3697   return function_literal;
3698 }
3699 
3700 
SkipLazyFunctionBody(const AstRawString * function_name,int * materialized_literal_count,int * expected_property_count,bool * ok)3701 void Parser::SkipLazyFunctionBody(const AstRawString* function_name,
3702                                   int* materialized_literal_count,
3703                                   int* expected_property_count,
3704                                   bool* ok) {
3705   int function_block_pos = position();
3706   if (compile_options() == ScriptCompiler::kConsumeParserCache) {
3707     // If we have cached data, we use it to skip parsing the function body. The
3708     // data contains the information we need to construct the lazy function.
3709     FunctionEntry entry =
3710         cached_parse_data_->GetFunctionEntry(function_block_pos);
3711     // Check that cached data is valid.
3712     CHECK(entry.is_valid());
3713     // End position greater than end of stream is safe, and hard to check.
3714     CHECK(entry.end_pos() > function_block_pos);
3715     scanner()->SeekForward(entry.end_pos() - 1);
3716 
3717     scope_->set_end_position(entry.end_pos());
3718     Expect(Token::RBRACE, ok);
3719     if (!*ok) {
3720       return;
3721     }
3722     total_preparse_skipped_ += scope_->end_position() - function_block_pos;
3723     *materialized_literal_count = entry.literal_count();
3724     *expected_property_count = entry.property_count();
3725     scope_->SetStrictMode(entry.strict_mode());
3726   } else {
3727     // With no cached data, we partially parse the function, without building an
3728     // AST. This gathers the data needed to build a lazy function.
3729     SingletonLogger logger;
3730     PreParser::PreParseResult result =
3731         ParseLazyFunctionBodyWithPreParser(&logger);
3732     if (result == PreParser::kPreParseStackOverflow) {
3733       // Propagate stack overflow.
3734       set_stack_overflow();
3735       *ok = false;
3736       return;
3737     }
3738     if (logger.has_error()) {
3739       ParserTraits::ReportMessageAt(
3740           Scanner::Location(logger.start(), logger.end()),
3741           logger.message(), logger.argument_opt(), logger.is_reference_error());
3742       *ok = false;
3743       return;
3744     }
3745     scope_->set_end_position(logger.end());
3746     Expect(Token::RBRACE, ok);
3747     if (!*ok) {
3748       return;
3749     }
3750     total_preparse_skipped_ += scope_->end_position() - function_block_pos;
3751     *materialized_literal_count = logger.literals();
3752     *expected_property_count = logger.properties();
3753     scope_->SetStrictMode(logger.strict_mode());
3754     if (compile_options() == ScriptCompiler::kProduceParserCache) {
3755       DCHECK(log_);
3756       // Position right after terminal '}'.
3757       int body_end = scanner()->location().end_pos;
3758       log_->LogFunction(function_block_pos, body_end,
3759                         *materialized_literal_count,
3760                         *expected_property_count,
3761                         scope_->strict_mode());
3762     }
3763   }
3764 }
3765 
3766 
ParseEagerFunctionBody(const AstRawString * function_name,int pos,Variable * fvar,Token::Value fvar_init_op,bool is_generator,bool * ok)3767 ZoneList<Statement*>* Parser::ParseEagerFunctionBody(
3768     const AstRawString* function_name, int pos, Variable* fvar,
3769     Token::Value fvar_init_op, bool is_generator, bool* ok) {
3770   // Everything inside an eagerly parsed function will be parsed eagerly
3771   // (see comment above).
3772   ParsingModeScope parsing_mode(this, PARSE_EAGERLY);
3773   ZoneList<Statement*>* body = new(zone()) ZoneList<Statement*>(8, zone());
3774   if (fvar != NULL) {
3775     VariableProxy* fproxy = scope_->NewUnresolved(
3776         factory(), function_name, Interface::NewConst());
3777     fproxy->BindTo(fvar);
3778     body->Add(factory()->NewExpressionStatement(
3779         factory()->NewAssignment(fvar_init_op,
3780                                  fproxy,
3781                                  factory()->NewThisFunction(pos),
3782                                  RelocInfo::kNoPosition),
3783         RelocInfo::kNoPosition), zone());
3784   }
3785 
3786   // For generators, allocate and yield an iterator on function entry.
3787   if (is_generator) {
3788     ZoneList<Expression*>* arguments =
3789         new(zone()) ZoneList<Expression*>(0, zone());
3790     CallRuntime* allocation = factory()->NewCallRuntime(
3791         ast_value_factory()->empty_string(),
3792         Runtime::FunctionForId(Runtime::kCreateJSGeneratorObject), arguments,
3793         pos);
3794     VariableProxy* init_proxy = factory()->NewVariableProxy(
3795         function_state_->generator_object_variable());
3796     Assignment* assignment = factory()->NewAssignment(
3797         Token::INIT_VAR, init_proxy, allocation, RelocInfo::kNoPosition);
3798     VariableProxy* get_proxy = factory()->NewVariableProxy(
3799         function_state_->generator_object_variable());
3800     Yield* yield = factory()->NewYield(
3801         get_proxy, assignment, Yield::kInitial, RelocInfo::kNoPosition);
3802     body->Add(factory()->NewExpressionStatement(
3803         yield, RelocInfo::kNoPosition), zone());
3804   }
3805 
3806   ParseSourceElements(body, Token::RBRACE, false, false, NULL, CHECK_OK);
3807 
3808   if (is_generator) {
3809     VariableProxy* get_proxy = factory()->NewVariableProxy(
3810         function_state_->generator_object_variable());
3811     Expression* undefined =
3812         factory()->NewUndefinedLiteral(RelocInfo::kNoPosition);
3813     Yield* yield = factory()->NewYield(get_proxy, undefined, Yield::kFinal,
3814                                        RelocInfo::kNoPosition);
3815     body->Add(factory()->NewExpressionStatement(
3816         yield, RelocInfo::kNoPosition), zone());
3817   }
3818 
3819   Expect(Token::RBRACE, CHECK_OK);
3820   scope_->set_end_position(scanner()->location().end_pos);
3821 
3822   return body;
3823 }
3824 
3825 
ParseLazyFunctionBodyWithPreParser(SingletonLogger * logger)3826 PreParser::PreParseResult Parser::ParseLazyFunctionBodyWithPreParser(
3827     SingletonLogger* logger) {
3828   // This function may be called on a background thread too; record only the
3829   // main thread preparse times.
3830   if (pre_parse_timer_ != NULL) {
3831     pre_parse_timer_->Start();
3832   }
3833   DCHECK_EQ(Token::LBRACE, scanner()->current_token());
3834 
3835   if (reusable_preparser_ == NULL) {
3836     reusable_preparser_ = new PreParser(&scanner_, NULL, stack_limit_);
3837     reusable_preparser_->set_allow_harmony_scoping(allow_harmony_scoping());
3838     reusable_preparser_->set_allow_modules(allow_modules());
3839     reusable_preparser_->set_allow_natives_syntax(allow_natives_syntax());
3840     reusable_preparser_->set_allow_lazy(true);
3841     reusable_preparser_->set_allow_arrow_functions(allow_arrow_functions());
3842     reusable_preparser_->set_allow_harmony_numeric_literals(
3843         allow_harmony_numeric_literals());
3844     reusable_preparser_->set_allow_classes(allow_classes());
3845     reusable_preparser_->set_allow_harmony_object_literals(
3846         allow_harmony_object_literals());
3847   }
3848   PreParser::PreParseResult result =
3849       reusable_preparser_->PreParseLazyFunction(strict_mode(),
3850                                                 is_generator(),
3851                                                 logger);
3852   if (pre_parse_timer_ != NULL) {
3853     pre_parse_timer_->Stop();
3854   }
3855   return result;
3856 }
3857 
3858 
ParseV8Intrinsic(bool * ok)3859 Expression* Parser::ParseV8Intrinsic(bool* ok) {
3860   // CallRuntime ::
3861   //   '%' Identifier Arguments
3862 
3863   int pos = peek_position();
3864   Expect(Token::MOD, CHECK_OK);
3865   // Allow "eval" or "arguments" for backward compatibility.
3866   const AstRawString* name = ParseIdentifier(kAllowEvalOrArguments, CHECK_OK);
3867   ZoneList<Expression*>* args = ParseArguments(CHECK_OK);
3868 
3869   if (extension_ != NULL) {
3870     // The extension structures are only accessible while parsing the
3871     // very first time not when reparsing because of lazy compilation.
3872     scope_->DeclarationScope()->ForceEagerCompilation();
3873   }
3874 
3875   const Runtime::Function* function = Runtime::FunctionForName(name->string());
3876 
3877   // Check for built-in IS_VAR macro.
3878   if (function != NULL &&
3879       function->intrinsic_type == Runtime::RUNTIME &&
3880       function->function_id == Runtime::kIS_VAR) {
3881     // %IS_VAR(x) evaluates to x if x is a variable,
3882     // leads to a parse error otherwise.  Could be implemented as an
3883     // inline function %_IS_VAR(x) to eliminate this special case.
3884     if (args->length() == 1 && args->at(0)->AsVariableProxy() != NULL) {
3885       return args->at(0);
3886     } else {
3887       ReportMessage("not_isvar");
3888       *ok = false;
3889       return NULL;
3890     }
3891   }
3892 
3893   // Check that the expected number of arguments are being passed.
3894   if (function != NULL &&
3895       function->nargs != -1 &&
3896       function->nargs != args->length()) {
3897     ReportMessage("illegal_access");
3898     *ok = false;
3899     return NULL;
3900   }
3901 
3902   // Check that the function is defined if it's an inline runtime call.
3903   if (function == NULL && name->FirstCharacter() == '_') {
3904     ParserTraits::ReportMessage("not_defined", name);
3905     *ok = false;
3906     return NULL;
3907   }
3908 
3909   // We have a valid intrinsics call or a call to a builtin.
3910   return factory()->NewCallRuntime(name, function, args, pos);
3911 }
3912 
3913 
GetLiteralUndefined(int position)3914 Literal* Parser::GetLiteralUndefined(int position) {
3915   return factory()->NewUndefinedLiteral(position);
3916 }
3917 
3918 
CheckConflictingVarDeclarations(Scope * scope,bool * ok)3919 void Parser::CheckConflictingVarDeclarations(Scope* scope, bool* ok) {
3920   Declaration* decl = scope->CheckConflictingVarDeclarations();
3921   if (decl != NULL) {
3922     // In harmony mode we treat conflicting variable bindinds as early
3923     // errors. See ES5 16 for a definition of early errors.
3924     const AstRawString* name = decl->proxy()->raw_name();
3925     int position = decl->proxy()->position();
3926     Scanner::Location location = position == RelocInfo::kNoPosition
3927         ? Scanner::Location::invalid()
3928         : Scanner::Location(position, position + 1);
3929     ParserTraits::ReportMessageAt(location, "var_redeclaration", name);
3930     *ok = false;
3931   }
3932 }
3933 
3934 
3935 // ----------------------------------------------------------------------------
3936 // Parser support
3937 
3938 
TargetStackContainsLabel(const AstRawString * label)3939 bool Parser::TargetStackContainsLabel(const AstRawString* label) {
3940   for (Target* t = target_stack_; t != NULL; t = t->previous()) {
3941     BreakableStatement* stat = t->node()->AsBreakableStatement();
3942     if (stat != NULL && ContainsLabel(stat->labels(), label))
3943       return true;
3944   }
3945   return false;
3946 }
3947 
3948 
LookupBreakTarget(const AstRawString * label,bool * ok)3949 BreakableStatement* Parser::LookupBreakTarget(const AstRawString* label,
3950                                               bool* ok) {
3951   bool anonymous = label == NULL;
3952   for (Target* t = target_stack_; t != NULL; t = t->previous()) {
3953     BreakableStatement* stat = t->node()->AsBreakableStatement();
3954     if (stat == NULL) continue;
3955     if ((anonymous && stat->is_target_for_anonymous()) ||
3956         (!anonymous && ContainsLabel(stat->labels(), label))) {
3957       RegisterTargetUse(stat->break_target(), t->previous());
3958       return stat;
3959     }
3960   }
3961   return NULL;
3962 }
3963 
3964 
LookupContinueTarget(const AstRawString * label,bool * ok)3965 IterationStatement* Parser::LookupContinueTarget(const AstRawString* label,
3966                                                  bool* ok) {
3967   bool anonymous = label == NULL;
3968   for (Target* t = target_stack_; t != NULL; t = t->previous()) {
3969     IterationStatement* stat = t->node()->AsIterationStatement();
3970     if (stat == NULL) continue;
3971 
3972     DCHECK(stat->is_target_for_anonymous());
3973     if (anonymous || ContainsLabel(stat->labels(), label)) {
3974       RegisterTargetUse(stat->continue_target(), t->previous());
3975       return stat;
3976     }
3977   }
3978   return NULL;
3979 }
3980 
3981 
RegisterTargetUse(Label * target,Target * stop)3982 void Parser::RegisterTargetUse(Label* target, Target* stop) {
3983   // Register that a break target found at the given stop in the
3984   // target stack has been used from the top of the target stack. Add
3985   // the break target to any TargetCollectors passed on the stack.
3986   for (Target* t = target_stack_; t != stop; t = t->previous()) {
3987     TargetCollector* collector = t->node()->AsTargetCollector();
3988     if (collector != NULL) collector->AddTarget(target, zone());
3989   }
3990 }
3991 
3992 
HandleSourceURLComments()3993 void Parser::HandleSourceURLComments() {
3994   if (scanner_.source_url()->length() > 0) {
3995     Handle<String> source_url = scanner_.source_url()->Internalize(isolate());
3996     info_->script()->set_source_url(*source_url);
3997   }
3998   if (scanner_.source_mapping_url()->length() > 0) {
3999     Handle<String> source_mapping_url =
4000         scanner_.source_mapping_url()->Internalize(isolate());
4001     info_->script()->set_source_mapping_url(*source_mapping_url);
4002   }
4003 }
4004 
4005 
ThrowPendingError()4006 void Parser::ThrowPendingError() {
4007   DCHECK(ast_value_factory()->IsInternalized());
4008   if (has_pending_error_) {
4009     MessageLocation location(script(), pending_error_location_.beg_pos,
4010                              pending_error_location_.end_pos);
4011     Factory* factory = isolate()->factory();
4012     bool has_arg =
4013         pending_error_arg_ != NULL || pending_error_char_arg_ != NULL;
4014     Handle<FixedArray> elements = factory->NewFixedArray(has_arg ? 1 : 0);
4015     if (pending_error_arg_ != NULL) {
4016       Handle<String> arg_string = pending_error_arg_->string();
4017       elements->set(0, *arg_string);
4018     } else if (pending_error_char_arg_ != NULL) {
4019       Handle<String> arg_string =
4020           factory->NewStringFromUtf8(CStrVector(pending_error_char_arg_))
4021           .ToHandleChecked();
4022       elements->set(0, *arg_string);
4023     }
4024     isolate()->debug()->OnCompileError(script());
4025 
4026     Handle<JSArray> array = factory->NewJSArrayWithElements(elements);
4027     Handle<Object> error;
4028     MaybeHandle<Object> maybe_error =
4029         pending_error_is_reference_error_
4030             ? factory->NewReferenceError(pending_error_message_, array)
4031             : factory->NewSyntaxError(pending_error_message_, array);
4032     if (maybe_error.ToHandle(&error)) isolate()->Throw(*error, &location);
4033   }
4034 }
4035 
4036 
Internalize()4037 void Parser::Internalize() {
4038   // Internalize strings.
4039   ast_value_factory()->Internalize(isolate());
4040 
4041   // Error processing.
4042   if (info()->function() == NULL) {
4043     if (stack_overflow()) {
4044       isolate()->StackOverflow();
4045     } else {
4046       ThrowPendingError();
4047     }
4048   }
4049 
4050   // Move statistics to Isolate.
4051   for (int feature = 0; feature < v8::Isolate::kUseCounterFeatureCount;
4052        ++feature) {
4053     for (int i = 0; i < use_counts_[feature]; ++i) {
4054       isolate()->CountUsage(v8::Isolate::UseCounterFeature(feature));
4055     }
4056   }
4057   isolate()->counters()->total_preparse_skipped()->Increment(
4058       total_preparse_skipped_);
4059 }
4060 
4061 
4062 // ----------------------------------------------------------------------------
4063 // Regular expressions
4064 
4065 
RegExpParser(FlatStringReader * in,Handle<String> * error,bool multiline,Zone * zone)4066 RegExpParser::RegExpParser(FlatStringReader* in,
4067                            Handle<String>* error,
4068                            bool multiline,
4069                            Zone* zone)
4070     : isolate_(zone->isolate()),
4071       zone_(zone),
4072       error_(error),
4073       captures_(NULL),
4074       in_(in),
4075       current_(kEndMarker),
4076       next_pos_(0),
4077       capture_count_(0),
4078       has_more_(true),
4079       multiline_(multiline),
4080       simple_(false),
4081       contains_anchor_(false),
4082       is_scanned_for_captures_(false),
4083       failed_(false) {
4084   Advance();
4085 }
4086 
4087 
Next()4088 uc32 RegExpParser::Next() {
4089   if (has_next()) {
4090     return in()->Get(next_pos_);
4091   } else {
4092     return kEndMarker;
4093   }
4094 }
4095 
4096 
Advance()4097 void RegExpParser::Advance() {
4098   if (next_pos_ < in()->length()) {
4099     StackLimitCheck check(isolate());
4100     if (check.HasOverflowed()) {
4101       ReportError(CStrVector(Isolate::kStackOverflowMessage));
4102     } else if (zone()->excess_allocation()) {
4103       ReportError(CStrVector("Regular expression too large"));
4104     } else {
4105       current_ = in()->Get(next_pos_);
4106       next_pos_++;
4107     }
4108   } else {
4109     current_ = kEndMarker;
4110     has_more_ = false;
4111   }
4112 }
4113 
4114 
Reset(int pos)4115 void RegExpParser::Reset(int pos) {
4116   next_pos_ = pos;
4117   has_more_ = (pos < in()->length());
4118   Advance();
4119 }
4120 
4121 
Advance(int dist)4122 void RegExpParser::Advance(int dist) {
4123   next_pos_ += dist - 1;
4124   Advance();
4125 }
4126 
4127 
simple()4128 bool RegExpParser::simple() {
4129   return simple_;
4130 }
4131 
4132 
ReportError(Vector<const char> message)4133 RegExpTree* RegExpParser::ReportError(Vector<const char> message) {
4134   failed_ = true;
4135   *error_ = isolate()->factory()->NewStringFromAscii(message).ToHandleChecked();
4136   // Zip to the end to make sure the no more input is read.
4137   current_ = kEndMarker;
4138   next_pos_ = in()->length();
4139   return NULL;
4140 }
4141 
4142 
4143 // Pattern ::
4144 //   Disjunction
ParsePattern()4145 RegExpTree* RegExpParser::ParsePattern() {
4146   RegExpTree* result = ParseDisjunction(CHECK_FAILED);
4147   DCHECK(!has_more());
4148   // If the result of parsing is a literal string atom, and it has the
4149   // same length as the input, then the atom is identical to the input.
4150   if (result->IsAtom() && result->AsAtom()->length() == in()->length()) {
4151     simple_ = true;
4152   }
4153   return result;
4154 }
4155 
4156 
4157 // Disjunction ::
4158 //   Alternative
4159 //   Alternative | Disjunction
4160 // Alternative ::
4161 //   [empty]
4162 //   Term Alternative
4163 // Term ::
4164 //   Assertion
4165 //   Atom
4166 //   Atom Quantifier
ParseDisjunction()4167 RegExpTree* RegExpParser::ParseDisjunction() {
4168   // Used to store current state while parsing subexpressions.
4169   RegExpParserState initial_state(NULL, INITIAL, 0, zone());
4170   RegExpParserState* stored_state = &initial_state;
4171   // Cache the builder in a local variable for quick access.
4172   RegExpBuilder* builder = initial_state.builder();
4173   while (true) {
4174     switch (current()) {
4175     case kEndMarker:
4176       if (stored_state->IsSubexpression()) {
4177         // Inside a parenthesized group when hitting end of input.
4178         ReportError(CStrVector("Unterminated group") CHECK_FAILED);
4179       }
4180       DCHECK_EQ(INITIAL, stored_state->group_type());
4181       // Parsing completed successfully.
4182       return builder->ToRegExp();
4183     case ')': {
4184       if (!stored_state->IsSubexpression()) {
4185         ReportError(CStrVector("Unmatched ')'") CHECK_FAILED);
4186       }
4187       DCHECK_NE(INITIAL, stored_state->group_type());
4188 
4189       Advance();
4190       // End disjunction parsing and convert builder content to new single
4191       // regexp atom.
4192       RegExpTree* body = builder->ToRegExp();
4193 
4194       int end_capture_index = captures_started();
4195 
4196       int capture_index = stored_state->capture_index();
4197       SubexpressionType group_type = stored_state->group_type();
4198 
4199       // Restore previous state.
4200       stored_state = stored_state->previous_state();
4201       builder = stored_state->builder();
4202 
4203       // Build result of subexpression.
4204       if (group_type == CAPTURE) {
4205         RegExpCapture* capture = new(zone()) RegExpCapture(body, capture_index);
4206         captures_->at(capture_index - 1) = capture;
4207         body = capture;
4208       } else if (group_type != GROUPING) {
4209         DCHECK(group_type == POSITIVE_LOOKAHEAD ||
4210                group_type == NEGATIVE_LOOKAHEAD);
4211         bool is_positive = (group_type == POSITIVE_LOOKAHEAD);
4212         body = new(zone()) RegExpLookahead(body,
4213                                    is_positive,
4214                                    end_capture_index - capture_index,
4215                                    capture_index);
4216       }
4217       builder->AddAtom(body);
4218       // For compatability with JSC and ES3, we allow quantifiers after
4219       // lookaheads, and break in all cases.
4220       break;
4221     }
4222     case '|': {
4223       Advance();
4224       builder->NewAlternative();
4225       continue;
4226     }
4227     case '*':
4228     case '+':
4229     case '?':
4230       return ReportError(CStrVector("Nothing to repeat"));
4231     case '^': {
4232       Advance();
4233       if (multiline_) {
4234         builder->AddAssertion(
4235             new(zone()) RegExpAssertion(RegExpAssertion::START_OF_LINE));
4236       } else {
4237         builder->AddAssertion(
4238             new(zone()) RegExpAssertion(RegExpAssertion::START_OF_INPUT));
4239         set_contains_anchor();
4240       }
4241       continue;
4242     }
4243     case '$': {
4244       Advance();
4245       RegExpAssertion::AssertionType assertion_type =
4246           multiline_ ? RegExpAssertion::END_OF_LINE :
4247                        RegExpAssertion::END_OF_INPUT;
4248       builder->AddAssertion(new(zone()) RegExpAssertion(assertion_type));
4249       continue;
4250     }
4251     case '.': {
4252       Advance();
4253       // everything except \x0a, \x0d, \u2028 and \u2029
4254       ZoneList<CharacterRange>* ranges =
4255           new(zone()) ZoneList<CharacterRange>(2, zone());
4256       CharacterRange::AddClassEscape('.', ranges, zone());
4257       RegExpTree* atom = new(zone()) RegExpCharacterClass(ranges, false);
4258       builder->AddAtom(atom);
4259       break;
4260     }
4261     case '(': {
4262       SubexpressionType subexpr_type = CAPTURE;
4263       Advance();
4264       if (current() == '?') {
4265         switch (Next()) {
4266           case ':':
4267             subexpr_type = GROUPING;
4268             break;
4269           case '=':
4270             subexpr_type = POSITIVE_LOOKAHEAD;
4271             break;
4272           case '!':
4273             subexpr_type = NEGATIVE_LOOKAHEAD;
4274             break;
4275           default:
4276             ReportError(CStrVector("Invalid group") CHECK_FAILED);
4277             break;
4278         }
4279         Advance(2);
4280       } else {
4281         if (captures_ == NULL) {
4282           captures_ = new(zone()) ZoneList<RegExpCapture*>(2, zone());
4283         }
4284         if (captures_started() >= kMaxCaptures) {
4285           ReportError(CStrVector("Too many captures") CHECK_FAILED);
4286         }
4287         captures_->Add(NULL, zone());
4288       }
4289       // Store current state and begin new disjunction parsing.
4290       stored_state = new(zone()) RegExpParserState(stored_state, subexpr_type,
4291                                                    captures_started(), zone());
4292       builder = stored_state->builder();
4293       continue;
4294     }
4295     case '[': {
4296       RegExpTree* atom = ParseCharacterClass(CHECK_FAILED);
4297       builder->AddAtom(atom);
4298       break;
4299     }
4300     // Atom ::
4301     //   \ AtomEscape
4302     case '\\':
4303       switch (Next()) {
4304       case kEndMarker:
4305         return ReportError(CStrVector("\\ at end of pattern"));
4306       case 'b':
4307         Advance(2);
4308         builder->AddAssertion(
4309             new(zone()) RegExpAssertion(RegExpAssertion::BOUNDARY));
4310         continue;
4311       case 'B':
4312         Advance(2);
4313         builder->AddAssertion(
4314             new(zone()) RegExpAssertion(RegExpAssertion::NON_BOUNDARY));
4315         continue;
4316       // AtomEscape ::
4317       //   CharacterClassEscape
4318       //
4319       // CharacterClassEscape :: one of
4320       //   d D s S w W
4321       case 'd': case 'D': case 's': case 'S': case 'w': case 'W': {
4322         uc32 c = Next();
4323         Advance(2);
4324         ZoneList<CharacterRange>* ranges =
4325             new(zone()) ZoneList<CharacterRange>(2, zone());
4326         CharacterRange::AddClassEscape(c, ranges, zone());
4327         RegExpTree* atom = new(zone()) RegExpCharacterClass(ranges, false);
4328         builder->AddAtom(atom);
4329         break;
4330       }
4331       case '1': case '2': case '3': case '4': case '5': case '6':
4332       case '7': case '8': case '9': {
4333         int index = 0;
4334         if (ParseBackReferenceIndex(&index)) {
4335           RegExpCapture* capture = NULL;
4336           if (captures_ != NULL && index <= captures_->length()) {
4337             capture = captures_->at(index - 1);
4338           }
4339           if (capture == NULL) {
4340             builder->AddEmpty();
4341             break;
4342           }
4343           RegExpTree* atom = new(zone()) RegExpBackReference(capture);
4344           builder->AddAtom(atom);
4345           break;
4346         }
4347         uc32 first_digit = Next();
4348         if (first_digit == '8' || first_digit == '9') {
4349           // Treat as identity escape
4350           builder->AddCharacter(first_digit);
4351           Advance(2);
4352           break;
4353         }
4354       }
4355       // FALLTHROUGH
4356       case '0': {
4357         Advance();
4358         uc32 octal = ParseOctalLiteral();
4359         builder->AddCharacter(octal);
4360         break;
4361       }
4362       // ControlEscape :: one of
4363       //   f n r t v
4364       case 'f':
4365         Advance(2);
4366         builder->AddCharacter('\f');
4367         break;
4368       case 'n':
4369         Advance(2);
4370         builder->AddCharacter('\n');
4371         break;
4372       case 'r':
4373         Advance(2);
4374         builder->AddCharacter('\r');
4375         break;
4376       case 't':
4377         Advance(2);
4378         builder->AddCharacter('\t');
4379         break;
4380       case 'v':
4381         Advance(2);
4382         builder->AddCharacter('\v');
4383         break;
4384       case 'c': {
4385         Advance();
4386         uc32 controlLetter = Next();
4387         // Special case if it is an ASCII letter.
4388         // Convert lower case letters to uppercase.
4389         uc32 letter = controlLetter & ~('a' ^ 'A');
4390         if (letter < 'A' || 'Z' < letter) {
4391           // controlLetter is not in range 'A'-'Z' or 'a'-'z'.
4392           // This is outside the specification. We match JSC in
4393           // reading the backslash as a literal character instead
4394           // of as starting an escape.
4395           builder->AddCharacter('\\');
4396         } else {
4397           Advance(2);
4398           builder->AddCharacter(controlLetter & 0x1f);
4399         }
4400         break;
4401       }
4402       case 'x': {
4403         Advance(2);
4404         uc32 value;
4405         if (ParseHexEscape(2, &value)) {
4406           builder->AddCharacter(value);
4407         } else {
4408           builder->AddCharacter('x');
4409         }
4410         break;
4411       }
4412       case 'u': {
4413         Advance(2);
4414         uc32 value;
4415         if (ParseHexEscape(4, &value)) {
4416           builder->AddCharacter(value);
4417         } else {
4418           builder->AddCharacter('u');
4419         }
4420         break;
4421       }
4422       default:
4423         // Identity escape.
4424         builder->AddCharacter(Next());
4425         Advance(2);
4426         break;
4427       }
4428       break;
4429     case '{': {
4430       int dummy;
4431       if (ParseIntervalQuantifier(&dummy, &dummy)) {
4432         ReportError(CStrVector("Nothing to repeat") CHECK_FAILED);
4433       }
4434       // fallthrough
4435     }
4436     default:
4437       builder->AddCharacter(current());
4438       Advance();
4439       break;
4440     }  // end switch(current())
4441 
4442     int min;
4443     int max;
4444     switch (current()) {
4445     // QuantifierPrefix ::
4446     //   *
4447     //   +
4448     //   ?
4449     //   {
4450     case '*':
4451       min = 0;
4452       max = RegExpTree::kInfinity;
4453       Advance();
4454       break;
4455     case '+':
4456       min = 1;
4457       max = RegExpTree::kInfinity;
4458       Advance();
4459       break;
4460     case '?':
4461       min = 0;
4462       max = 1;
4463       Advance();
4464       break;
4465     case '{':
4466       if (ParseIntervalQuantifier(&min, &max)) {
4467         if (max < min) {
4468           ReportError(CStrVector("numbers out of order in {} quantifier.")
4469                       CHECK_FAILED);
4470         }
4471         break;
4472       } else {
4473         continue;
4474       }
4475     default:
4476       continue;
4477     }
4478     RegExpQuantifier::QuantifierType quantifier_type = RegExpQuantifier::GREEDY;
4479     if (current() == '?') {
4480       quantifier_type = RegExpQuantifier::NON_GREEDY;
4481       Advance();
4482     } else if (FLAG_regexp_possessive_quantifier && current() == '+') {
4483       // FLAG_regexp_possessive_quantifier is a debug-only flag.
4484       quantifier_type = RegExpQuantifier::POSSESSIVE;
4485       Advance();
4486     }
4487     builder->AddQuantifierToAtom(min, max, quantifier_type);
4488   }
4489 }
4490 
4491 
4492 #ifdef DEBUG
4493 // Currently only used in an DCHECK.
IsSpecialClassEscape(uc32 c)4494 static bool IsSpecialClassEscape(uc32 c) {
4495   switch (c) {
4496     case 'd': case 'D':
4497     case 's': case 'S':
4498     case 'w': case 'W':
4499       return true;
4500     default:
4501       return false;
4502   }
4503 }
4504 #endif
4505 
4506 
4507 // In order to know whether an escape is a backreference or not we have to scan
4508 // the entire regexp and find the number of capturing parentheses.  However we
4509 // don't want to scan the regexp twice unless it is necessary.  This mini-parser
4510 // is called when needed.  It can see the difference between capturing and
4511 // noncapturing parentheses and can skip character classes and backslash-escaped
4512 // characters.
ScanForCaptures()4513 void RegExpParser::ScanForCaptures() {
4514   // Start with captures started previous to current position
4515   int capture_count = captures_started();
4516   // Add count of captures after this position.
4517   int n;
4518   while ((n = current()) != kEndMarker) {
4519     Advance();
4520     switch (n) {
4521       case '\\':
4522         Advance();
4523         break;
4524       case '[': {
4525         int c;
4526         while ((c = current()) != kEndMarker) {
4527           Advance();
4528           if (c == '\\') {
4529             Advance();
4530           } else {
4531             if (c == ']') break;
4532           }
4533         }
4534         break;
4535       }
4536       case '(':
4537         if (current() != '?') capture_count++;
4538         break;
4539     }
4540   }
4541   capture_count_ = capture_count;
4542   is_scanned_for_captures_ = true;
4543 }
4544 
4545 
ParseBackReferenceIndex(int * index_out)4546 bool RegExpParser::ParseBackReferenceIndex(int* index_out) {
4547   DCHECK_EQ('\\', current());
4548   DCHECK('1' <= Next() && Next() <= '9');
4549   // Try to parse a decimal literal that is no greater than the total number
4550   // of left capturing parentheses in the input.
4551   int start = position();
4552   int value = Next() - '0';
4553   Advance(2);
4554   while (true) {
4555     uc32 c = current();
4556     if (IsDecimalDigit(c)) {
4557       value = 10 * value + (c - '0');
4558       if (value > kMaxCaptures) {
4559         Reset(start);
4560         return false;
4561       }
4562       Advance();
4563     } else {
4564       break;
4565     }
4566   }
4567   if (value > captures_started()) {
4568     if (!is_scanned_for_captures_) {
4569       int saved_position = position();
4570       ScanForCaptures();
4571       Reset(saved_position);
4572     }
4573     if (value > capture_count_) {
4574       Reset(start);
4575       return false;
4576     }
4577   }
4578   *index_out = value;
4579   return true;
4580 }
4581 
4582 
4583 // QuantifierPrefix ::
4584 //   { DecimalDigits }
4585 //   { DecimalDigits , }
4586 //   { DecimalDigits , DecimalDigits }
4587 //
4588 // Returns true if parsing succeeds, and set the min_out and max_out
4589 // values. Values are truncated to RegExpTree::kInfinity if they overflow.
ParseIntervalQuantifier(int * min_out,int * max_out)4590 bool RegExpParser::ParseIntervalQuantifier(int* min_out, int* max_out) {
4591   DCHECK_EQ(current(), '{');
4592   int start = position();
4593   Advance();
4594   int min = 0;
4595   if (!IsDecimalDigit(current())) {
4596     Reset(start);
4597     return false;
4598   }
4599   while (IsDecimalDigit(current())) {
4600     int next = current() - '0';
4601     if (min > (RegExpTree::kInfinity - next) / 10) {
4602       // Overflow. Skip past remaining decimal digits and return -1.
4603       do {
4604         Advance();
4605       } while (IsDecimalDigit(current()));
4606       min = RegExpTree::kInfinity;
4607       break;
4608     }
4609     min = 10 * min + next;
4610     Advance();
4611   }
4612   int max = 0;
4613   if (current() == '}') {
4614     max = min;
4615     Advance();
4616   } else if (current() == ',') {
4617     Advance();
4618     if (current() == '}') {
4619       max = RegExpTree::kInfinity;
4620       Advance();
4621     } else {
4622       while (IsDecimalDigit(current())) {
4623         int next = current() - '0';
4624         if (max > (RegExpTree::kInfinity - next) / 10) {
4625           do {
4626             Advance();
4627           } while (IsDecimalDigit(current()));
4628           max = RegExpTree::kInfinity;
4629           break;
4630         }
4631         max = 10 * max + next;
4632         Advance();
4633       }
4634       if (current() != '}') {
4635         Reset(start);
4636         return false;
4637       }
4638       Advance();
4639     }
4640   } else {
4641     Reset(start);
4642     return false;
4643   }
4644   *min_out = min;
4645   *max_out = max;
4646   return true;
4647 }
4648 
4649 
ParseOctalLiteral()4650 uc32 RegExpParser::ParseOctalLiteral() {
4651   DCHECK(('0' <= current() && current() <= '7') || current() == kEndMarker);
4652   // For compatibility with some other browsers (not all), we parse
4653   // up to three octal digits with a value below 256.
4654   uc32 value = current() - '0';
4655   Advance();
4656   if ('0' <= current() && current() <= '7') {
4657     value = value * 8 + current() - '0';
4658     Advance();
4659     if (value < 32 && '0' <= current() && current() <= '7') {
4660       value = value * 8 + current() - '0';
4661       Advance();
4662     }
4663   }
4664   return value;
4665 }
4666 
4667 
ParseHexEscape(int length,uc32 * value)4668 bool RegExpParser::ParseHexEscape(int length, uc32 *value) {
4669   int start = position();
4670   uc32 val = 0;
4671   bool done = false;
4672   for (int i = 0; !done; i++) {
4673     uc32 c = current();
4674     int d = HexValue(c);
4675     if (d < 0) {
4676       Reset(start);
4677       return false;
4678     }
4679     val = val * 16 + d;
4680     Advance();
4681     if (i == length - 1) {
4682       done = true;
4683     }
4684   }
4685   *value = val;
4686   return true;
4687 }
4688 
4689 
ParseClassCharacterEscape()4690 uc32 RegExpParser::ParseClassCharacterEscape() {
4691   DCHECK(current() == '\\');
4692   DCHECK(has_next() && !IsSpecialClassEscape(Next()));
4693   Advance();
4694   switch (current()) {
4695     case 'b':
4696       Advance();
4697       return '\b';
4698     // ControlEscape :: one of
4699     //   f n r t v
4700     case 'f':
4701       Advance();
4702       return '\f';
4703     case 'n':
4704       Advance();
4705       return '\n';
4706     case 'r':
4707       Advance();
4708       return '\r';
4709     case 't':
4710       Advance();
4711       return '\t';
4712     case 'v':
4713       Advance();
4714       return '\v';
4715     case 'c': {
4716       uc32 controlLetter = Next();
4717       uc32 letter = controlLetter & ~('A' ^ 'a');
4718       // For compatibility with JSC, inside a character class
4719       // we also accept digits and underscore as control characters.
4720       if ((controlLetter >= '0' && controlLetter <= '9') ||
4721           controlLetter == '_' ||
4722           (letter >= 'A' && letter <= 'Z')) {
4723         Advance(2);
4724         // Control letters mapped to ASCII control characters in the range
4725         // 0x00-0x1f.
4726         return controlLetter & 0x1f;
4727       }
4728       // We match JSC in reading the backslash as a literal
4729       // character instead of as starting an escape.
4730       return '\\';
4731     }
4732     case '0': case '1': case '2': case '3': case '4': case '5':
4733     case '6': case '7':
4734       // For compatibility, we interpret a decimal escape that isn't
4735       // a back reference (and therefore either \0 or not valid according
4736       // to the specification) as a 1..3 digit octal character code.
4737       return ParseOctalLiteral();
4738     case 'x': {
4739       Advance();
4740       uc32 value;
4741       if (ParseHexEscape(2, &value)) {
4742         return value;
4743       }
4744       // If \x is not followed by a two-digit hexadecimal, treat it
4745       // as an identity escape.
4746       return 'x';
4747     }
4748     case 'u': {
4749       Advance();
4750       uc32 value;
4751       if (ParseHexEscape(4, &value)) {
4752         return value;
4753       }
4754       // If \u is not followed by a four-digit hexadecimal, treat it
4755       // as an identity escape.
4756       return 'u';
4757     }
4758     default: {
4759       // Extended identity escape. We accept any character that hasn't
4760       // been matched by a more specific case, not just the subset required
4761       // by the ECMAScript specification.
4762       uc32 result = current();
4763       Advance();
4764       return result;
4765     }
4766   }
4767   return 0;
4768 }
4769 
4770 
ParseClassAtom(uc16 * char_class)4771 CharacterRange RegExpParser::ParseClassAtom(uc16* char_class) {
4772   DCHECK_EQ(0, *char_class);
4773   uc32 first = current();
4774   if (first == '\\') {
4775     switch (Next()) {
4776       case 'w': case 'W': case 'd': case 'D': case 's': case 'S': {
4777         *char_class = Next();
4778         Advance(2);
4779         return CharacterRange::Singleton(0);  // Return dummy value.
4780       }
4781       case kEndMarker:
4782         return ReportError(CStrVector("\\ at end of pattern"));
4783       default:
4784         uc32 c = ParseClassCharacterEscape(CHECK_FAILED);
4785         return CharacterRange::Singleton(c);
4786     }
4787   } else {
4788     Advance();
4789     return CharacterRange::Singleton(first);
4790   }
4791 }
4792 
4793 
4794 static const uc16 kNoCharClass = 0;
4795 
4796 // Adds range or pre-defined character class to character ranges.
4797 // If char_class is not kInvalidClass, it's interpreted as a class
4798 // escape (i.e., 's' means whitespace, from '\s').
AddRangeOrEscape(ZoneList<CharacterRange> * ranges,uc16 char_class,CharacterRange range,Zone * zone)4799 static inline void AddRangeOrEscape(ZoneList<CharacterRange>* ranges,
4800                                     uc16 char_class,
4801                                     CharacterRange range,
4802                                     Zone* zone) {
4803   if (char_class != kNoCharClass) {
4804     CharacterRange::AddClassEscape(char_class, ranges, zone);
4805   } else {
4806     ranges->Add(range, zone);
4807   }
4808 }
4809 
4810 
ParseCharacterClass()4811 RegExpTree* RegExpParser::ParseCharacterClass() {
4812   static const char* kUnterminated = "Unterminated character class";
4813   static const char* kRangeOutOfOrder = "Range out of order in character class";
4814 
4815   DCHECK_EQ(current(), '[');
4816   Advance();
4817   bool is_negated = false;
4818   if (current() == '^') {
4819     is_negated = true;
4820     Advance();
4821   }
4822   ZoneList<CharacterRange>* ranges =
4823       new(zone()) ZoneList<CharacterRange>(2, zone());
4824   while (has_more() && current() != ']') {
4825     uc16 char_class = kNoCharClass;
4826     CharacterRange first = ParseClassAtom(&char_class CHECK_FAILED);
4827     if (current() == '-') {
4828       Advance();
4829       if (current() == kEndMarker) {
4830         // If we reach the end we break out of the loop and let the
4831         // following code report an error.
4832         break;
4833       } else if (current() == ']') {
4834         AddRangeOrEscape(ranges, char_class, first, zone());
4835         ranges->Add(CharacterRange::Singleton('-'), zone());
4836         break;
4837       }
4838       uc16 char_class_2 = kNoCharClass;
4839       CharacterRange next = ParseClassAtom(&char_class_2 CHECK_FAILED);
4840       if (char_class != kNoCharClass || char_class_2 != kNoCharClass) {
4841         // Either end is an escaped character class. Treat the '-' verbatim.
4842         AddRangeOrEscape(ranges, char_class, first, zone());
4843         ranges->Add(CharacterRange::Singleton('-'), zone());
4844         AddRangeOrEscape(ranges, char_class_2, next, zone());
4845         continue;
4846       }
4847       if (first.from() > next.to()) {
4848         return ReportError(CStrVector(kRangeOutOfOrder) CHECK_FAILED);
4849       }
4850       ranges->Add(CharacterRange::Range(first.from(), next.to()), zone());
4851     } else {
4852       AddRangeOrEscape(ranges, char_class, first, zone());
4853     }
4854   }
4855   if (!has_more()) {
4856     return ReportError(CStrVector(kUnterminated) CHECK_FAILED);
4857   }
4858   Advance();
4859   if (ranges->length() == 0) {
4860     ranges->Add(CharacterRange::Everything(), zone());
4861     is_negated = !is_negated;
4862   }
4863   return new(zone()) RegExpCharacterClass(ranges, is_negated);
4864 }
4865 
4866 
4867 // ----------------------------------------------------------------------------
4868 // The Parser interface.
4869 
ParseRegExp(FlatStringReader * input,bool multiline,RegExpCompileData * result,Zone * zone)4870 bool RegExpParser::ParseRegExp(FlatStringReader* input,
4871                                bool multiline,
4872                                RegExpCompileData* result,
4873                                Zone* zone) {
4874   DCHECK(result != NULL);
4875   RegExpParser parser(input, &result->error, multiline, zone);
4876   RegExpTree* tree = parser.ParsePattern();
4877   if (parser.failed()) {
4878     DCHECK(tree == NULL);
4879     DCHECK(!result->error.is_null());
4880   } else {
4881     DCHECK(tree != NULL);
4882     DCHECK(result->error.is_null());
4883     result->tree = tree;
4884     int capture_count = parser.captures_started();
4885     result->simple = tree->IsAtom() && parser.simple() && capture_count == 0;
4886     result->contains_anchor = parser.contains_anchor();
4887     result->capture_count = capture_count;
4888   }
4889   return !parser.failed();
4890 }
4891 
4892 
Parse()4893 bool Parser::Parse() {
4894   DCHECK(info()->function() == NULL);
4895   FunctionLiteral* result = NULL;
4896   pre_parse_timer_ = isolate()->counters()->pre_parse();
4897   if (FLAG_trace_parse || allow_natives_syntax() || extension_ != NULL) {
4898     // If intrinsics are allowed, the Parser cannot operate independent of the
4899     // V8 heap because of Runtime. Tell the string table to internalize strings
4900     // and values right after they're created.
4901     ast_value_factory()->Internalize(isolate());
4902   }
4903 
4904   if (info()->is_lazy()) {
4905     DCHECK(!info()->is_eval());
4906     if (info()->shared_info()->is_function()) {
4907       result = ParseLazy();
4908     } else {
4909       result = ParseProgram();
4910     }
4911   } else {
4912     SetCachedData();
4913     result = ParseProgram();
4914   }
4915   info()->SetFunction(result);
4916 
4917   Internalize();
4918   DCHECK(ast_value_factory()->IsInternalized());
4919   return (result != NULL);
4920 }
4921 
4922 
ParseOnBackground()4923 void Parser::ParseOnBackground() {
4924   DCHECK(info()->function() == NULL);
4925   FunctionLiteral* result = NULL;
4926   fni_ = new (zone()) FuncNameInferrer(ast_value_factory(), zone());
4927 
4928   CompleteParserRecorder recorder;
4929   if (compile_options() == ScriptCompiler::kProduceParserCache) {
4930     log_ = &recorder;
4931   }
4932 
4933   DCHECK(info()->source_stream() != NULL);
4934   ExternalStreamingStream stream(info()->source_stream(),
4935                                  info()->source_stream_encoding());
4936   scanner_.Initialize(&stream);
4937   DCHECK(info()->context().is_null() || info()->context()->IsNativeContext());
4938 
4939   // When streaming, we don't know the length of the source until we have parsed
4940   // it. The raw data can be UTF-8, so we wouldn't know the source length until
4941   // we have decoded it anyway even if we knew the raw data length (which we
4942   // don't). We work around this by storing all the scopes which need their end
4943   // position set at the end of the script (the top scope and possible eval
4944   // scopes) and set their end position after we know the script length.
4945   Scope* top_scope = NULL;
4946   Scope* eval_scope = NULL;
4947   result = DoParseProgram(info(), &top_scope, &eval_scope);
4948 
4949   top_scope->set_end_position(scanner()->location().end_pos);
4950   if (eval_scope != NULL) {
4951     eval_scope->set_end_position(scanner()->location().end_pos);
4952   }
4953 
4954   info()->SetFunction(result);
4955 
4956   // We cannot internalize on a background thread; a foreground task will take
4957   // care of calling Parser::Internalize just before compilation.
4958 
4959   if (compile_options() == ScriptCompiler::kProduceParserCache) {
4960     if (result != NULL) *info_->cached_data() = recorder.GetScriptData();
4961     log_ = NULL;
4962   }
4963 }
4964 } }  // namespace v8::internal
4965