Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1 | // Copyright 2011 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 | // Features shared by parsing and pre-parsing scanners. |
| 6 | |
| 7 | #ifndef V8_PARSING_SCANNER_H_ |
| 8 | #define V8_PARSING_SCANNER_H_ |
| 9 | |
| 10 | #include "src/allocation.h" |
| 11 | #include "src/base/logging.h" |
| 12 | #include "src/char-predicates.h" |
| 13 | #include "src/globals.h" |
| 14 | #include "src/hashmap.h" |
| 15 | #include "src/list.h" |
| 16 | #include "src/parsing/token.h" |
| 17 | #include "src/unicode.h" |
| 18 | #include "src/unicode-decoder.h" |
| 19 | #include "src/utils.h" |
| 20 | |
| 21 | namespace v8 { |
| 22 | namespace internal { |
| 23 | |
| 24 | |
| 25 | class AstRawString; |
| 26 | class AstValueFactory; |
| 27 | class ParserRecorder; |
| 28 | class UnicodeCache; |
| 29 | |
| 30 | |
| 31 | // --------------------------------------------------------------------- |
| 32 | // Buffered stream of UTF-16 code units, using an internal UTF-16 buffer. |
| 33 | // A code unit is a 16 bit value representing either a 16 bit code point |
| 34 | // or one part of a surrogate pair that make a single 21 bit code point. |
| 35 | |
| 36 | class Utf16CharacterStream { |
| 37 | public: |
| 38 | Utf16CharacterStream() : pos_(0) { } |
| 39 | virtual ~Utf16CharacterStream() { } |
| 40 | |
| 41 | // Returns and advances past the next UTF-16 code unit in the input |
| 42 | // stream. If there are no more code units, it returns a negative |
| 43 | // value. |
| 44 | inline uc32 Advance() { |
| 45 | if (buffer_cursor_ < buffer_end_ || ReadBlock()) { |
| 46 | pos_++; |
| 47 | return static_cast<uc32>(*(buffer_cursor_++)); |
| 48 | } |
| 49 | // Note: currently the following increment is necessary to avoid a |
| 50 | // parser problem! The scanner treats the final kEndOfInput as |
| 51 | // a code unit with a position, and does math relative to that |
| 52 | // position. |
| 53 | pos_++; |
| 54 | |
| 55 | return kEndOfInput; |
| 56 | } |
| 57 | |
| 58 | // Return the current position in the code unit stream. |
| 59 | // Starts at zero. |
| 60 | inline size_t pos() const { return pos_; } |
| 61 | |
| 62 | // Skips forward past the next code_unit_count UTF-16 code units |
| 63 | // in the input, or until the end of input if that comes sooner. |
| 64 | // Returns the number of code units actually skipped. If less |
| 65 | // than code_unit_count, |
| 66 | inline size_t SeekForward(size_t code_unit_count) { |
| 67 | size_t buffered_chars = buffer_end_ - buffer_cursor_; |
| 68 | if (code_unit_count <= buffered_chars) { |
| 69 | buffer_cursor_ += code_unit_count; |
| 70 | pos_ += code_unit_count; |
| 71 | return code_unit_count; |
| 72 | } |
| 73 | return SlowSeekForward(code_unit_count); |
| 74 | } |
| 75 | |
| 76 | // Pushes back the most recently read UTF-16 code unit (or negative |
| 77 | // value if at end of input), i.e., the value returned by the most recent |
| 78 | // call to Advance. |
| 79 | // Must not be used right after calling SeekForward. |
| 80 | virtual void PushBack(int32_t code_unit) = 0; |
| 81 | |
| 82 | virtual bool SetBookmark(); |
| 83 | virtual void ResetToBookmark(); |
| 84 | |
| 85 | protected: |
| 86 | static const uc32 kEndOfInput = -1; |
| 87 | |
| 88 | // Ensures that the buffer_cursor_ points to the code_unit at |
| 89 | // position pos_ of the input, if possible. If the position |
| 90 | // is at or after the end of the input, return false. If there |
| 91 | // are more code_units available, return true. |
| 92 | virtual bool ReadBlock() = 0; |
| 93 | virtual size_t SlowSeekForward(size_t code_unit_count) = 0; |
| 94 | |
| 95 | const uint16_t* buffer_cursor_; |
| 96 | const uint16_t* buffer_end_; |
| 97 | size_t pos_; |
| 98 | }; |
| 99 | |
| 100 | |
| 101 | // --------------------------------------------------------------------- |
| 102 | // DuplicateFinder discovers duplicate symbols. |
| 103 | |
| 104 | class DuplicateFinder { |
| 105 | public: |
| 106 | explicit DuplicateFinder(UnicodeCache* constants) |
| 107 | : unicode_constants_(constants), |
| 108 | backing_store_(16), |
| 109 | map_(&Match) { } |
| 110 | |
| 111 | int AddOneByteSymbol(Vector<const uint8_t> key, int value); |
| 112 | int AddTwoByteSymbol(Vector<const uint16_t> key, int value); |
| 113 | // Add a a number literal by converting it (if necessary) |
| 114 | // to the string that ToString(ToNumber(literal)) would generate. |
| 115 | // and then adding that string with AddOneByteSymbol. |
| 116 | // This string is the actual value used as key in an object literal, |
| 117 | // and the one that must be different from the other keys. |
| 118 | int AddNumber(Vector<const uint8_t> key, int value); |
| 119 | |
| 120 | private: |
| 121 | int AddSymbol(Vector<const uint8_t> key, bool is_one_byte, int value); |
| 122 | // Backs up the key and its length in the backing store. |
| 123 | // The backup is stored with a base 127 encoding of the |
| 124 | // length (plus a bit saying whether the string is one byte), |
| 125 | // followed by the bytes of the key. |
| 126 | uint8_t* BackupKey(Vector<const uint8_t> key, bool is_one_byte); |
| 127 | |
| 128 | // Compare two encoded keys (both pointing into the backing store) |
| 129 | // for having the same base-127 encoded lengths and representation. |
| 130 | // and then having the same 'length' bytes following. |
| 131 | static bool Match(void* first, void* second); |
| 132 | // Creates a hash from a sequence of bytes. |
| 133 | static uint32_t Hash(Vector<const uint8_t> key, bool is_one_byte); |
| 134 | // Checks whether a string containing a JS number is its canonical |
| 135 | // form. |
| 136 | static bool IsNumberCanonical(Vector<const uint8_t> key); |
| 137 | |
| 138 | // Size of buffer. Sufficient for using it to call DoubleToCString in |
| 139 | // from conversions.h. |
| 140 | static const int kBufferSize = 100; |
| 141 | |
| 142 | UnicodeCache* unicode_constants_; |
| 143 | // Backing store used to store strings used as hashmap keys. |
| 144 | SequenceCollector<unsigned char> backing_store_; |
| 145 | HashMap map_; |
| 146 | // Buffer used for string->number->canonical string conversions. |
| 147 | char number_buffer_[kBufferSize]; |
| 148 | }; |
| 149 | |
| 150 | |
| 151 | // ---------------------------------------------------------------------------- |
| 152 | // LiteralBuffer - Collector of chars of literals. |
| 153 | |
| 154 | class LiteralBuffer { |
| 155 | public: |
| 156 | LiteralBuffer() : is_one_byte_(true), position_(0), backing_store_() { } |
| 157 | |
| 158 | ~LiteralBuffer() { backing_store_.Dispose(); } |
| 159 | |
| 160 | INLINE(void AddChar(uint32_t code_unit)) { |
| 161 | if (position_ >= backing_store_.length()) ExpandBuffer(); |
| 162 | if (is_one_byte_) { |
| 163 | if (code_unit <= unibrow::Latin1::kMaxChar) { |
| 164 | backing_store_[position_] = static_cast<byte>(code_unit); |
| 165 | position_ += kOneByteSize; |
| 166 | return; |
| 167 | } |
| 168 | ConvertToTwoByte(); |
| 169 | } |
| 170 | if (code_unit <= unibrow::Utf16::kMaxNonSurrogateCharCode) { |
| 171 | *reinterpret_cast<uint16_t*>(&backing_store_[position_]) = code_unit; |
| 172 | position_ += kUC16Size; |
| 173 | } else { |
| 174 | *reinterpret_cast<uint16_t*>(&backing_store_[position_]) = |
| 175 | unibrow::Utf16::LeadSurrogate(code_unit); |
| 176 | position_ += kUC16Size; |
| 177 | if (position_ >= backing_store_.length()) ExpandBuffer(); |
| 178 | *reinterpret_cast<uint16_t*>(&backing_store_[position_]) = |
| 179 | unibrow::Utf16::TrailSurrogate(code_unit); |
| 180 | position_ += kUC16Size; |
| 181 | } |
| 182 | } |
| 183 | |
| 184 | bool is_one_byte() const { return is_one_byte_; } |
| 185 | |
| 186 | bool is_contextual_keyword(Vector<const char> keyword) const { |
| 187 | return is_one_byte() && keyword.length() == position_ && |
| 188 | (memcmp(keyword.start(), backing_store_.start(), position_) == 0); |
| 189 | } |
| 190 | |
| 191 | Vector<const uint16_t> two_byte_literal() const { |
| 192 | DCHECK(!is_one_byte_); |
| 193 | DCHECK((position_ & 0x1) == 0); |
| 194 | return Vector<const uint16_t>( |
| 195 | reinterpret_cast<const uint16_t*>(backing_store_.start()), |
| 196 | position_ >> 1); |
| 197 | } |
| 198 | |
| 199 | Vector<const uint8_t> one_byte_literal() const { |
| 200 | DCHECK(is_one_byte_); |
| 201 | return Vector<const uint8_t>( |
| 202 | reinterpret_cast<const uint8_t*>(backing_store_.start()), |
| 203 | position_); |
| 204 | } |
| 205 | |
| 206 | int length() const { |
| 207 | return is_one_byte_ ? position_ : (position_ >> 1); |
| 208 | } |
| 209 | |
| 210 | void ReduceLength(int delta) { |
| 211 | position_ -= delta * (is_one_byte_ ? kOneByteSize : kUC16Size); |
| 212 | } |
| 213 | |
| 214 | void Reset() { |
| 215 | position_ = 0; |
| 216 | is_one_byte_ = true; |
| 217 | } |
| 218 | |
| 219 | Handle<String> Internalize(Isolate* isolate) const; |
| 220 | |
| 221 | void CopyFrom(const LiteralBuffer* other) { |
| 222 | if (other == nullptr) { |
| 223 | Reset(); |
| 224 | } else { |
| 225 | is_one_byte_ = other->is_one_byte_; |
| 226 | position_ = other->position_; |
| 227 | backing_store_.Dispose(); |
| 228 | backing_store_ = other->backing_store_.Clone(); |
| 229 | } |
| 230 | } |
| 231 | |
| 232 | private: |
| 233 | static const int kInitialCapacity = 16; |
| 234 | static const int kGrowthFactory = 4; |
| 235 | static const int kMinConversionSlack = 256; |
| 236 | static const int kMaxGrowth = 1 * MB; |
| 237 | inline int NewCapacity(int min_capacity) { |
| 238 | int capacity = Max(min_capacity, backing_store_.length()); |
| 239 | int new_capacity = Min(capacity * kGrowthFactory, capacity + kMaxGrowth); |
| 240 | return new_capacity; |
| 241 | } |
| 242 | |
| 243 | void ExpandBuffer() { |
| 244 | Vector<byte> new_store = Vector<byte>::New(NewCapacity(kInitialCapacity)); |
| 245 | MemCopy(new_store.start(), backing_store_.start(), position_); |
| 246 | backing_store_.Dispose(); |
| 247 | backing_store_ = new_store; |
| 248 | } |
| 249 | |
| 250 | void ConvertToTwoByte() { |
| 251 | DCHECK(is_one_byte_); |
| 252 | Vector<byte> new_store; |
| 253 | int new_content_size = position_ * kUC16Size; |
| 254 | if (new_content_size >= backing_store_.length()) { |
| 255 | // Ensure room for all currently read code units as UC16 as well |
| 256 | // as the code unit about to be stored. |
| 257 | new_store = Vector<byte>::New(NewCapacity(new_content_size)); |
| 258 | } else { |
| 259 | new_store = backing_store_; |
| 260 | } |
| 261 | uint8_t* src = backing_store_.start(); |
| 262 | uint16_t* dst = reinterpret_cast<uint16_t*>(new_store.start()); |
| 263 | for (int i = position_ - 1; i >= 0; i--) { |
| 264 | dst[i] = src[i]; |
| 265 | } |
| 266 | if (new_store.start() != backing_store_.start()) { |
| 267 | backing_store_.Dispose(); |
| 268 | backing_store_ = new_store; |
| 269 | } |
| 270 | position_ = new_content_size; |
| 271 | is_one_byte_ = false; |
| 272 | } |
| 273 | |
| 274 | bool is_one_byte_; |
| 275 | int position_; |
| 276 | Vector<byte> backing_store_; |
| 277 | |
| 278 | DISALLOW_COPY_AND_ASSIGN(LiteralBuffer); |
| 279 | }; |
| 280 | |
| 281 | |
| 282 | // ---------------------------------------------------------------------------- |
| 283 | // JavaScript Scanner. |
| 284 | |
| 285 | class Scanner { |
| 286 | public: |
| 287 | // Scoped helper for literal recording. Automatically drops the literal |
| 288 | // if aborting the scanning before it's complete. |
| 289 | class LiteralScope { |
| 290 | public: |
| 291 | explicit LiteralScope(Scanner* self) : scanner_(self), complete_(false) { |
| 292 | scanner_->StartLiteral(); |
| 293 | } |
| 294 | ~LiteralScope() { |
| 295 | if (!complete_) scanner_->DropLiteral(); |
| 296 | } |
| 297 | void Complete() { |
| 298 | complete_ = true; |
| 299 | } |
| 300 | |
| 301 | private: |
| 302 | Scanner* scanner_; |
| 303 | bool complete_; |
| 304 | }; |
| 305 | |
| 306 | // Scoped helper for a re-settable bookmark. |
| 307 | class BookmarkScope { |
| 308 | public: |
| 309 | explicit BookmarkScope(Scanner* scanner) : scanner_(scanner) { |
| 310 | DCHECK_NOT_NULL(scanner_); |
| 311 | } |
| 312 | ~BookmarkScope() { scanner_->DropBookmark(); } |
| 313 | |
| 314 | bool Set() { return scanner_->SetBookmark(); } |
| 315 | void Reset() { scanner_->ResetToBookmark(); } |
| 316 | bool HasBeenSet() { return scanner_->BookmarkHasBeenSet(); } |
| 317 | bool HasBeenReset() { return scanner_->BookmarkHasBeenReset(); } |
| 318 | |
| 319 | private: |
| 320 | Scanner* scanner_; |
| 321 | |
| 322 | DISALLOW_COPY_AND_ASSIGN(BookmarkScope); |
| 323 | }; |
| 324 | |
| 325 | // Representation of an interval of source positions. |
| 326 | struct Location { |
| 327 | Location(int b, int e) : beg_pos(b), end_pos(e) { } |
| 328 | Location() : beg_pos(0), end_pos(0) { } |
| 329 | |
| 330 | bool IsValid() const { |
| 331 | return beg_pos >= 0 && end_pos >= beg_pos; |
| 332 | } |
| 333 | |
| 334 | static Location invalid() { return Location(-1, -1); } |
| 335 | |
| 336 | int beg_pos; |
| 337 | int end_pos; |
| 338 | }; |
| 339 | |
| 340 | // -1 is outside of the range of any real source code. |
| 341 | static const int kNoOctalLocation = -1; |
| 342 | |
| 343 | explicit Scanner(UnicodeCache* scanner_contants); |
| 344 | |
| 345 | void Initialize(Utf16CharacterStream* source); |
| 346 | |
| 347 | // Returns the next token and advances input. |
| 348 | Token::Value Next(); |
| 349 | // Returns the token following peek() |
| 350 | Token::Value PeekAhead(); |
| 351 | // Returns the current token again. |
| 352 | Token::Value current_token() { return current_.token; } |
| 353 | // Returns the location information for the current token |
| 354 | // (the token last returned by Next()). |
| 355 | Location location() const { return current_.location; } |
| 356 | |
| 357 | // Similar functions for the upcoming token. |
| 358 | |
| 359 | // One token look-ahead (past the token returned by Next()). |
| 360 | Token::Value peek() const { return next_.token; } |
| 361 | |
| 362 | Location peek_location() const { return next_.location; } |
| 363 | |
| 364 | bool literal_contains_escapes() const { |
| 365 | return LiteralContainsEscapes(current_); |
| 366 | } |
| 367 | bool next_literal_contains_escapes() const { |
| 368 | return LiteralContainsEscapes(next_); |
| 369 | } |
| 370 | bool is_literal_contextual_keyword(Vector<const char> keyword) { |
| 371 | DCHECK_NOT_NULL(current_.literal_chars); |
| 372 | return current_.literal_chars->is_contextual_keyword(keyword); |
| 373 | } |
| 374 | bool is_next_contextual_keyword(Vector<const char> keyword) { |
| 375 | DCHECK_NOT_NULL(next_.literal_chars); |
| 376 | return next_.literal_chars->is_contextual_keyword(keyword); |
| 377 | } |
| 378 | |
| 379 | const AstRawString* CurrentSymbol(AstValueFactory* ast_value_factory); |
| 380 | const AstRawString* NextSymbol(AstValueFactory* ast_value_factory); |
| 381 | const AstRawString* CurrentRawSymbol(AstValueFactory* ast_value_factory); |
| 382 | |
| 383 | double DoubleValue(); |
| 384 | bool ContainsDot(); |
| 385 | bool LiteralMatches(const char* data, int length, bool allow_escapes = true) { |
| 386 | if (is_literal_one_byte() && |
| 387 | literal_length() == length && |
| 388 | (allow_escapes || !literal_contains_escapes())) { |
| 389 | const char* token = |
| 390 | reinterpret_cast<const char*>(literal_one_byte_string().start()); |
| 391 | return !strncmp(token, data, length); |
| 392 | } |
| 393 | return false; |
| 394 | } |
| 395 | inline bool UnescapedLiteralMatches(const char* data, int length) { |
| 396 | return LiteralMatches(data, length, false); |
| 397 | } |
| 398 | |
| 399 | void IsGetOrSet(bool* is_get, bool* is_set) { |
| 400 | if (is_literal_one_byte() && |
| 401 | literal_length() == 3 && |
| 402 | !literal_contains_escapes()) { |
| 403 | const char* token = |
| 404 | reinterpret_cast<const char*>(literal_one_byte_string().start()); |
| 405 | *is_get = strncmp(token, "get", 3) == 0; |
| 406 | *is_set = !*is_get && strncmp(token, "set", 3) == 0; |
| 407 | } |
| 408 | } |
| 409 | |
| 410 | int FindSymbol(DuplicateFinder* finder, int value); |
| 411 | |
| 412 | UnicodeCache* unicode_cache() { return unicode_cache_; } |
| 413 | |
| 414 | // Returns the location of the last seen octal literal. |
| 415 | Location octal_position() const { return octal_pos_; } |
| 416 | void clear_octal_position() { octal_pos_ = Location::invalid(); } |
| 417 | |
| 418 | // Returns the value of the last smi that was scanned. |
| 419 | int smi_value() const { return current_.smi_value_; } |
| 420 | |
| 421 | // Seek forward to the given position. This operation does not |
| 422 | // work in general, for instance when there are pushed back |
| 423 | // characters, but works for seeking forward until simple delimiter |
| 424 | // tokens, which is what it is used for. |
| 425 | void SeekForward(int pos); |
| 426 | |
| 427 | // Returns true if there was a line terminator before the peek'ed token, |
| 428 | // possibly inside a multi-line comment. |
| 429 | bool HasAnyLineTerminatorBeforeNext() const { |
| 430 | return has_line_terminator_before_next_ || |
| 431 | has_multiline_comment_before_next_; |
| 432 | } |
| 433 | |
| 434 | // Scans the input as a regular expression pattern, previous |
| 435 | // character(s) must be /(=). Returns true if a pattern is scanned. |
| 436 | bool ScanRegExpPattern(bool seen_equal); |
| 437 | // Scans the input as regular expression flags. Returns the flags on success. |
| 438 | Maybe<RegExp::Flags> ScanRegExpFlags(); |
| 439 | |
| 440 | // Scans the input as a template literal |
| 441 | Token::Value ScanTemplateStart(); |
| 442 | Token::Value ScanTemplateContinuation(); |
| 443 | |
| 444 | const LiteralBuffer* source_url() const { return &source_url_; } |
| 445 | const LiteralBuffer* source_mapping_url() const { |
| 446 | return &source_mapping_url_; |
| 447 | } |
| 448 | |
| 449 | bool IdentifierIsFutureStrictReserved(const AstRawString* string) const; |
| 450 | |
Ben Murdoch | 097c5b2 | 2016-05-18 11:27:45 +0100 | [diff] [blame^] | 451 | bool FoundHtmlComment() const { return found_html_comment_; } |
| 452 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 453 | private: |
| 454 | // The current and look-ahead token. |
| 455 | struct TokenDesc { |
| 456 | Token::Value token; |
| 457 | Location location; |
| 458 | LiteralBuffer* literal_chars; |
| 459 | LiteralBuffer* raw_literal_chars; |
| 460 | int smi_value_; |
| 461 | }; |
| 462 | |
| 463 | static const int kCharacterLookaheadBufferSize = 1; |
| 464 | |
| 465 | // Scans octal escape sequence. Also accepts "\0" decimal escape sequence. |
| 466 | template <bool capture_raw> |
| 467 | uc32 ScanOctalEscape(uc32 c, int length); |
| 468 | |
| 469 | // Call this after setting source_ to the input. |
| 470 | void Init() { |
| 471 | // Set c0_ (one character ahead) |
| 472 | STATIC_ASSERT(kCharacterLookaheadBufferSize == 1); |
| 473 | Advance(); |
| 474 | // Initialize current_ to not refer to a literal. |
| 475 | current_.literal_chars = NULL; |
| 476 | current_.raw_literal_chars = NULL; |
| 477 | next_next_.token = Token::UNINITIALIZED; |
Ben Murdoch | 097c5b2 | 2016-05-18 11:27:45 +0100 | [diff] [blame^] | 478 | found_html_comment_ = false; |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 479 | } |
| 480 | |
| 481 | // Support BookmarkScope functionality. |
| 482 | bool SetBookmark(); |
| 483 | void ResetToBookmark(); |
| 484 | bool BookmarkHasBeenSet(); |
| 485 | bool BookmarkHasBeenReset(); |
| 486 | void DropBookmark(); |
| 487 | static void CopyTokenDesc(TokenDesc* to, TokenDesc* from); |
| 488 | |
| 489 | // Literal buffer support |
| 490 | inline void StartLiteral() { |
| 491 | LiteralBuffer* free_buffer = |
| 492 | (current_.literal_chars == &literal_buffer0_) |
| 493 | ? &literal_buffer1_ |
| 494 | : (current_.literal_chars == &literal_buffer1_) ? &literal_buffer2_ |
| 495 | : &literal_buffer0_; |
| 496 | free_buffer->Reset(); |
| 497 | next_.literal_chars = free_buffer; |
| 498 | } |
| 499 | |
| 500 | inline void StartRawLiteral() { |
| 501 | LiteralBuffer* free_buffer = |
| 502 | (current_.raw_literal_chars == &raw_literal_buffer0_) |
| 503 | ? &raw_literal_buffer1_ |
| 504 | : (current_.raw_literal_chars == &raw_literal_buffer1_) |
| 505 | ? &raw_literal_buffer2_ |
| 506 | : &raw_literal_buffer0_; |
| 507 | free_buffer->Reset(); |
| 508 | next_.raw_literal_chars = free_buffer; |
| 509 | } |
| 510 | |
| 511 | INLINE(void AddLiteralChar(uc32 c)) { |
| 512 | DCHECK_NOT_NULL(next_.literal_chars); |
| 513 | next_.literal_chars->AddChar(c); |
| 514 | } |
| 515 | |
| 516 | INLINE(void AddRawLiteralChar(uc32 c)) { |
| 517 | DCHECK_NOT_NULL(next_.raw_literal_chars); |
| 518 | next_.raw_literal_chars->AddChar(c); |
| 519 | } |
| 520 | |
| 521 | INLINE(void ReduceRawLiteralLength(int delta)) { |
| 522 | DCHECK_NOT_NULL(next_.raw_literal_chars); |
| 523 | next_.raw_literal_chars->ReduceLength(delta); |
| 524 | } |
| 525 | |
| 526 | // Stops scanning of a literal and drop the collected characters, |
| 527 | // e.g., due to an encountered error. |
| 528 | inline void DropLiteral() { |
| 529 | next_.literal_chars = NULL; |
| 530 | next_.raw_literal_chars = NULL; |
| 531 | } |
| 532 | |
| 533 | inline void AddLiteralCharAdvance() { |
| 534 | AddLiteralChar(c0_); |
| 535 | Advance(); |
| 536 | } |
| 537 | |
| 538 | // Low-level scanning support. |
| 539 | template <bool capture_raw = false, bool check_surrogate = true> |
| 540 | void Advance() { |
| 541 | if (capture_raw) { |
| 542 | AddRawLiteralChar(c0_); |
| 543 | } |
| 544 | c0_ = source_->Advance(); |
| 545 | if (check_surrogate) HandleLeadSurrogate(); |
| 546 | } |
| 547 | |
| 548 | void HandleLeadSurrogate() { |
| 549 | if (unibrow::Utf16::IsLeadSurrogate(c0_)) { |
| 550 | uc32 c1 = source_->Advance(); |
| 551 | if (!unibrow::Utf16::IsTrailSurrogate(c1)) { |
| 552 | source_->PushBack(c1); |
| 553 | } else { |
| 554 | c0_ = unibrow::Utf16::CombineSurrogatePair(c0_, c1); |
| 555 | } |
| 556 | } |
| 557 | } |
| 558 | |
| 559 | void PushBack(uc32 ch) { |
| 560 | if (ch > static_cast<uc32>(unibrow::Utf16::kMaxNonSurrogateCharCode)) { |
| 561 | source_->PushBack(unibrow::Utf16::TrailSurrogate(c0_)); |
| 562 | source_->PushBack(unibrow::Utf16::LeadSurrogate(c0_)); |
| 563 | } else { |
| 564 | source_->PushBack(c0_); |
| 565 | } |
| 566 | c0_ = ch; |
| 567 | } |
| 568 | |
| 569 | inline Token::Value Select(Token::Value tok) { |
| 570 | Advance(); |
| 571 | return tok; |
| 572 | } |
| 573 | |
| 574 | inline Token::Value Select(uc32 next, Token::Value then, Token::Value else_) { |
| 575 | Advance(); |
| 576 | if (c0_ == next) { |
| 577 | Advance(); |
| 578 | return then; |
| 579 | } else { |
| 580 | return else_; |
| 581 | } |
| 582 | } |
| 583 | |
| 584 | // Returns the literal string, if any, for the current token (the |
| 585 | // token last returned by Next()). The string is 0-terminated. |
| 586 | // Literal strings are collected for identifiers, strings, numbers as well |
| 587 | // as for template literals. For template literals we also collect the raw |
| 588 | // form. |
| 589 | // These functions only give the correct result if the literal was scanned |
| 590 | // when a LiteralScope object is alive. |
| 591 | Vector<const uint8_t> literal_one_byte_string() { |
| 592 | DCHECK_NOT_NULL(current_.literal_chars); |
| 593 | return current_.literal_chars->one_byte_literal(); |
| 594 | } |
| 595 | Vector<const uint16_t> literal_two_byte_string() { |
| 596 | DCHECK_NOT_NULL(current_.literal_chars); |
| 597 | return current_.literal_chars->two_byte_literal(); |
| 598 | } |
| 599 | bool is_literal_one_byte() { |
| 600 | DCHECK_NOT_NULL(current_.literal_chars); |
| 601 | return current_.literal_chars->is_one_byte(); |
| 602 | } |
| 603 | int literal_length() const { |
| 604 | DCHECK_NOT_NULL(current_.literal_chars); |
| 605 | return current_.literal_chars->length(); |
| 606 | } |
| 607 | // Returns the literal string for the next token (the token that |
| 608 | // would be returned if Next() were called). |
| 609 | Vector<const uint8_t> next_literal_one_byte_string() { |
| 610 | DCHECK_NOT_NULL(next_.literal_chars); |
| 611 | return next_.literal_chars->one_byte_literal(); |
| 612 | } |
| 613 | Vector<const uint16_t> next_literal_two_byte_string() { |
| 614 | DCHECK_NOT_NULL(next_.literal_chars); |
| 615 | return next_.literal_chars->two_byte_literal(); |
| 616 | } |
| 617 | bool is_next_literal_one_byte() { |
| 618 | DCHECK_NOT_NULL(next_.literal_chars); |
| 619 | return next_.literal_chars->is_one_byte(); |
| 620 | } |
| 621 | Vector<const uint8_t> raw_literal_one_byte_string() { |
| 622 | DCHECK_NOT_NULL(current_.raw_literal_chars); |
| 623 | return current_.raw_literal_chars->one_byte_literal(); |
| 624 | } |
| 625 | Vector<const uint16_t> raw_literal_two_byte_string() { |
| 626 | DCHECK_NOT_NULL(current_.raw_literal_chars); |
| 627 | return current_.raw_literal_chars->two_byte_literal(); |
| 628 | } |
| 629 | bool is_raw_literal_one_byte() { |
| 630 | DCHECK_NOT_NULL(current_.raw_literal_chars); |
| 631 | return current_.raw_literal_chars->is_one_byte(); |
| 632 | } |
| 633 | |
| 634 | template <bool capture_raw> |
| 635 | uc32 ScanHexNumber(int expected_length); |
| 636 | // Scan a number of any length but not bigger than max_value. For example, the |
| 637 | // number can be 000000001, so it's very long in characters but its value is |
| 638 | // small. |
| 639 | template <bool capture_raw> |
| 640 | uc32 ScanUnlimitedLengthHexNumber(int max_value); |
| 641 | |
| 642 | // Scans a single JavaScript token. |
| 643 | void Scan(); |
| 644 | |
| 645 | bool SkipWhiteSpace(); |
| 646 | Token::Value SkipSingleLineComment(); |
| 647 | Token::Value SkipSourceURLComment(); |
| 648 | void TryToParseSourceURLComment(); |
| 649 | Token::Value SkipMultiLineComment(); |
| 650 | // Scans a possible HTML comment -- begins with '<!'. |
| 651 | Token::Value ScanHtmlComment(); |
| 652 | |
| 653 | void ScanDecimalDigits(); |
| 654 | Token::Value ScanNumber(bool seen_period); |
| 655 | Token::Value ScanIdentifierOrKeyword(); |
| 656 | Token::Value ScanIdentifierSuffix(LiteralScope* literal, bool escaped); |
| 657 | |
| 658 | Token::Value ScanString(); |
| 659 | |
| 660 | // Scans an escape-sequence which is part of a string and adds the |
| 661 | // decoded character to the current literal. Returns true if a pattern |
| 662 | // is scanned. |
| 663 | template <bool capture_raw, bool in_template_literal> |
| 664 | bool ScanEscape(); |
| 665 | |
| 666 | // Decodes a Unicode escape-sequence which is part of an identifier. |
| 667 | // If the escape sequence cannot be decoded the result is kBadChar. |
| 668 | uc32 ScanIdentifierUnicodeEscape(); |
| 669 | // Helper for the above functions. |
| 670 | template <bool capture_raw> |
| 671 | uc32 ScanUnicodeEscape(); |
| 672 | |
| 673 | Token::Value ScanTemplateSpan(); |
| 674 | |
| 675 | // Return the current source position. |
| 676 | int source_pos() { |
| 677 | return static_cast<int>(source_->pos()) - kCharacterLookaheadBufferSize; |
| 678 | } |
| 679 | |
| 680 | static bool LiteralContainsEscapes(const TokenDesc& token) { |
| 681 | Location location = token.location; |
| 682 | int source_length = (location.end_pos - location.beg_pos); |
| 683 | if (token.token == Token::STRING) { |
| 684 | // Subtract delimiters. |
| 685 | source_length -= 2; |
| 686 | } |
| 687 | return token.literal_chars->length() != source_length; |
| 688 | } |
| 689 | |
| 690 | UnicodeCache* unicode_cache_; |
| 691 | |
| 692 | // Buffers collecting literal strings, numbers, etc. |
| 693 | LiteralBuffer literal_buffer0_; |
| 694 | LiteralBuffer literal_buffer1_; |
| 695 | LiteralBuffer literal_buffer2_; |
| 696 | |
| 697 | // Values parsed from magic comments. |
| 698 | LiteralBuffer source_url_; |
| 699 | LiteralBuffer source_mapping_url_; |
| 700 | |
| 701 | // Buffer to store raw string values |
| 702 | LiteralBuffer raw_literal_buffer0_; |
| 703 | LiteralBuffer raw_literal_buffer1_; |
| 704 | LiteralBuffer raw_literal_buffer2_; |
| 705 | |
| 706 | TokenDesc current_; // desc for current token (as returned by Next()) |
| 707 | TokenDesc next_; // desc for next token (one token look-ahead) |
| 708 | TokenDesc next_next_; // desc for the token after next (after PeakAhead()) |
| 709 | |
| 710 | // Variables for Scanner::BookmarkScope and the *Bookmark implementation. |
| 711 | // These variables contain the scanner state when a bookmark is set. |
| 712 | // |
| 713 | // We will use bookmark_c0_ as a 'control' variable, where: |
| 714 | // - bookmark_c0_ >= 0: A bookmark has been set and this contains c0_. |
| 715 | // - bookmark_c0_ == -1: No bookmark has been set. |
| 716 | // - bookmark_c0_ == -2: The bookmark has been applied (ResetToBookmark). |
| 717 | // |
| 718 | // Which state is being bookmarked? The parser state is distributed over |
| 719 | // several variables, roughly like this: |
| 720 | // ... 1234 + 5678 ..... [character stream] |
| 721 | // [current_] [next_] c0_ | [scanner state] |
| 722 | // So when the scanner is logically at the beginning of an expression |
| 723 | // like "1234 + 4567", then: |
| 724 | // - current_ contains "1234" |
| 725 | // - next_ contains "+" |
| 726 | // - c0_ contains ' ' (the space between "+" and "5678", |
| 727 | // - the source_ character stream points to the beginning of "5678". |
| 728 | // To be able to restore this state, we will keep copies of current_, next_, |
| 729 | // and c0_; we'll ask the stream to bookmark itself, and we'll copy the |
| 730 | // contents of current_'s and next_'s literal buffers to bookmark_*_literal_. |
| 731 | static const uc32 kNoBookmark = -1; |
| 732 | static const uc32 kBookmarkWasApplied = -2; |
| 733 | uc32 bookmark_c0_; |
| 734 | TokenDesc bookmark_current_; |
| 735 | TokenDesc bookmark_next_; |
| 736 | LiteralBuffer bookmark_current_literal_; |
| 737 | LiteralBuffer bookmark_current_raw_literal_; |
| 738 | LiteralBuffer bookmark_next_literal_; |
| 739 | LiteralBuffer bookmark_next_raw_literal_; |
| 740 | |
| 741 | // Input stream. Must be initialized to an Utf16CharacterStream. |
| 742 | Utf16CharacterStream* source_; |
| 743 | |
| 744 | |
| 745 | // Start position of the octal literal last scanned. |
| 746 | Location octal_pos_; |
| 747 | |
| 748 | // One Unicode character look-ahead; c0_ < 0 at the end of the input. |
| 749 | uc32 c0_; |
| 750 | |
| 751 | // Whether there is a line terminator whitespace character after |
| 752 | // the current token, and before the next. Does not count newlines |
| 753 | // inside multiline comments. |
| 754 | bool has_line_terminator_before_next_; |
| 755 | // Whether there is a multi-line comment that contains a |
| 756 | // line-terminator after the current token, and before the next. |
| 757 | bool has_multiline_comment_before_next_; |
Ben Murdoch | 097c5b2 | 2016-05-18 11:27:45 +0100 | [diff] [blame^] | 758 | |
| 759 | // Whether this scanner encountered an HTML comment. |
| 760 | bool found_html_comment_; |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 761 | }; |
| 762 | |
| 763 | } // namespace internal |
| 764 | } // namespace v8 |
| 765 | |
| 766 | #endif // V8_PARSING_SCANNER_H_ |