1 /* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ 2 /* vim: set ts=8 sts=2 et sw=2 tw=80: */ 3 /* This Source Code Form is subject to the terms of the Mozilla Public 4 * License, v. 2.0. If a copy of the MPL was not distributed with this 5 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ 6 7 #ifndef Tokenizer_h__ 8 #define Tokenizer_h__ 9 10 #include <type_traits> 11 12 #include "nsString.h" 13 #include "mozilla/CheckedInt.h" 14 #include "mozilla/ScopeExit.h" 15 #include "mozilla/UniquePtr.h" 16 #include "nsTArray.h" 17 18 namespace mozilla { 19 20 template <typename TChar> 21 class TokenizerBase { 22 public: 23 typedef nsTSubstring<TChar> TAString; 24 typedef nsTString<TChar> TString; 25 typedef nsTDependentString<TChar> TDependentString; 26 typedef nsTDependentSubstring<TChar> TDependentSubstring; 27 28 static TChar const sWhitespaces[]; 29 30 /** 31 * The analyzer works with elements in the input cut to a sequence of token 32 * where each token has an elementary type 33 */ 34 enum TokenType : uint32_t { 35 TOKEN_UNKNOWN, 36 TOKEN_RAW, 37 TOKEN_ERROR, 38 TOKEN_INTEGER, 39 TOKEN_WORD, 40 TOKEN_CHAR, 41 TOKEN_WS, 42 TOKEN_EOL, 43 TOKEN_EOF, 44 TOKEN_CUSTOM0 = 1000 45 }; 46 47 enum ECaseSensitivity { CASE_SENSITIVE, CASE_INSENSITIVE }; 48 49 /** 50 * Class holding the type and the value of a token. It can be manually 51 * created to allow checks against it via methods of TTokenizer or are results 52 * of some of the TTokenizer's methods. 53 */ 54 class Token { 55 TokenType mType; 56 TDependentSubstring mWord; 57 TString mCustom; 58 TChar mChar; 59 uint64_t mInteger; 60 ECaseSensitivity mCustomCaseInsensitivity; 61 bool mCustomEnabled; 62 63 // If this token is a result of the parsing process, this member is 64 // referencing a sub-string in the input buffer. If this is externally 65 // created Token this member is left an empty string. 66 TDependentSubstring mFragment; 67 68 friend class TokenizerBase<TChar>; 69 void AssignFragment(typename TAString::const_char_iterator begin, 70 typename TAString::const_char_iterator end); 71 72 static Token Raw(); 73 74 public: 75 Token(); 76 Token(const Token& aOther); 77 Token& operator=(const Token& aOther); 78 79 // Static constructors of tokens by type and value 80 static Token Word(TAString const& aWord); 81 static Token Char(TChar const aChar); 82 static Token Number(uint64_t const aNumber); 83 static Token Whitespace(); 84 static Token NewLine(); 85 static Token EndOfFile(); 86 static Token Error(); 87 88 // Compares the two tokens, type must be identical and value 89 // of one of the tokens must be 'any' or equal. 90 bool Equals(const Token& aOther) const; 91 Type()92 TokenType Type() const { return mType; } 93 TChar AsChar() const; 94 TDependentSubstring AsString() const; 95 uint64_t AsInteger() const; 96 Fragment()97 TDependentSubstring Fragment() const { return mFragment; } 98 }; 99 100 /** 101 * Consumers may register a custom string that, when found in the input, is 102 * considered a token and returned by Next*() and accepted by Check*() 103 * methods. AddCustomToken() returns a reference to a token that can then be 104 * comapred using Token::Equals() againts the output from Next*() or be passed 105 * to Check*(). 106 */ 107 Token AddCustomToken(const TAString& aValue, 108 ECaseSensitivity aCaseInsensitivity, 109 bool aEnabled = true); 110 template <uint32_t N> 111 Token AddCustomToken(const TChar (&aValue)[N], 112 ECaseSensitivity aCaseInsensitivity, 113 bool aEnabled = true) { 114 return AddCustomToken(TDependentSubstring(aValue, N - 1), 115 aCaseInsensitivity, aEnabled); 116 } 117 void RemoveCustomToken(Token& aToken); 118 /** 119 * Only applies to a custom type of a Token (see AddCustomToken above.) 120 * This turns on and off token recognition. When a custom token is disabled, 121 * it's ignored as never added as a custom token. 122 */ 123 void EnableCustomToken(Token const& aToken, bool aEnable); 124 125 /** 126 * Mode of tokenization. 127 * FULL tokenization, the default, recognizes built-in tokens and any custom 128 * tokens, if added. CUSTOM_ONLY will only recognize custom tokens, the rest 129 * is seen as 'raw'. This mode can be understood as a 'binary' mode. 130 */ 131 enum class Mode { FULL, CUSTOM_ONLY }; 132 void SetTokenizingMode(Mode aMode); 133 134 /** 135 * Return false iff the last Check*() call has returned false or when we've 136 * read past the end of the input string. 137 */ 138 [[nodiscard]] bool HasFailed() const; 139 140 protected: 141 explicit TokenizerBase(const TChar* aWhitespaces = nullptr, 142 const TChar* aAdditionalWordChars = nullptr); 143 144 // false if we have already read the EOF token. 145 bool HasInput() const; 146 // Main parsing function, it doesn't shift the read cursor, just returns the 147 // next token position. 148 typename TAString::const_char_iterator Parse(Token& aToken) const; 149 // Is read cursor at the end? 150 bool IsEnd(const typename TAString::const_char_iterator& caret) const; 151 // True, when we are at the end of the input data, but it has not been marked 152 // as complete yet. In that case we cannot proceed with providing a 153 // multi-TChar token. 154 bool IsPending(const typename TAString::const_char_iterator& caret) const; 155 // Is read cursor on a character that is a word start? 156 bool IsWordFirst(const TChar aInput) const; 157 // Is read cursor on a character that is an in-word letter? 158 bool IsWord(const TChar aInput) const; 159 // Is read cursor on a character that is a valid number? 160 // TODO - support multiple radix 161 bool IsNumber(const TChar aInput) const; 162 // Is equal to the given custom token? 163 bool IsCustom(const typename TAString::const_char_iterator& caret, 164 const Token& aCustomToken, uint32_t* aLongest = nullptr) const; 165 166 // Friendly helper to assign a fragment on a Token 167 static void AssignFragment(Token& aToken, 168 typename TAString::const_char_iterator begin, 169 typename TAString::const_char_iterator end); 170 171 // true iff we have already read the EOF token 172 bool mPastEof; 173 // true iff the last Check*() call has returned false, reverts to true on 174 // Rollback() call 175 bool mHasFailed; 176 // true if the input string is final (finished), false when we expect more 177 // data yet to be fed to the tokenizer (see IncrementalTokenizer derived 178 // class). 179 bool mInputFinished; 180 // custom only vs full tokenizing mode, see the Parse() method 181 Mode mMode; 182 // minimal raw data chunked delivery during incremental feed 183 uint32_t mMinRawDelivery; 184 185 // Customizable list of whitespaces 186 const TChar* mWhitespaces; 187 // Additinal custom word characters 188 const TChar* mAdditionalWordChars; 189 190 // All these point to the original buffer passed to the constructor or to the 191 // incremental buffer after FeedInput. 192 typename TAString::const_char_iterator 193 mCursor; // Position of the current (actually next to read) token start 194 typename TAString::const_char_iterator mEnd; // End of the input position 195 196 // This is the list of tokens user has registered with AddCustomToken() 197 nsTArray<UniquePtr<Token>> mCustomTokens; 198 uint32_t mNextCustomTokenID; 199 200 private: 201 TokenizerBase() = delete; 202 TokenizerBase(const TokenizerBase&) = delete; 203 TokenizerBase(TokenizerBase&&) = delete; 204 TokenizerBase(const TokenizerBase&&) = delete; 205 TokenizerBase& operator=(const TokenizerBase&) = delete; 206 }; 207 208 /** 209 * This is a simple implementation of a lexical analyzer or maybe better 210 * called a tokenizer. 211 * 212 * Please use Tokenizer or Tokenizer16 classes, that are specializations 213 * of this template class. Tokenizer is for ASCII input, Tokenizer16 may 214 * handle char16_t input, but doesn't recognize whitespaces or numbers 215 * other than standard `char` specialized Tokenizer class. 216 */ 217 template <typename TChar> 218 class TTokenizer : public TokenizerBase<TChar> { 219 public: 220 typedef TokenizerBase<TChar> base; 221 222 /** 223 * @param aSource 224 * The string to parse. 225 * IMPORTANT NOTE: TTokenizer doesn't ensure the input string buffer 226 * lifetime. It's up to the consumer to make sure the string's buffer outlives 227 * the TTokenizer! 228 * @param aWhitespaces 229 * If non-null TTokenizer will use this custom set of whitespaces for 230 * CheckWhite() and SkipWhites() calls. By default the list consists of space 231 * and tab. 232 * @param aAdditionalWordChars 233 * If non-null it will be added to the list of characters that consist a 234 * word. This is useful when you want to accept e.g. '-' in HTTP headers. By 235 * default a word character is consider any character for which upper case 236 * is different from lower case. 237 * 238 * If there is an overlap between aWhitespaces and aAdditionalWordChars, the 239 * check for word characters is made first. 240 */ 241 explicit TTokenizer(const typename base::TAString& aSource, 242 const TChar* aWhitespaces = nullptr, 243 const TChar* aAdditionalWordChars = nullptr); 244 explicit TTokenizer(const TChar* aSource, const TChar* aWhitespaces = nullptr, 245 const TChar* aAdditionalWordChars = nullptr); 246 247 /** 248 * When there is still anything to read from the input, tokenize it, store the 249 * token type and value to aToken result and shift the cursor past this just 250 * parsed token. Each call to Next() reads another token from the input and 251 * shifts the cursor. Returns false if we have passed the end of the input. 252 */ 253 [[nodiscard]] bool Next(typename base::Token& aToken); 254 255 /** 256 * Parse the token on the input read cursor position, check its type is equal 257 * to aTokenType and if so, put it into aResult, shift the cursor and return 258 * true. Otherwise, leave the input read cursor position intact and return 259 * false. 260 */ 261 [[nodiscard]] bool Check(const typename base::TokenType aTokenType, 262 typename base::Token& aResult); 263 /** 264 * Same as above method, just compares both token type and token value passed 265 * in aToken. When both the type and the value equals, shift the cursor and 266 * return true. Otherwise return false. 267 */ 268 [[nodiscard]] bool Check(const typename base::Token& aToken); 269 270 /** 271 * SkipWhites method (below) may also skip new line characters automatically. 272 */ 273 enum WhiteSkipping { 274 /** 275 * SkipWhites will only skip what is defined as a white space (default). 276 */ 277 DONT_INCLUDE_NEW_LINE = 0, 278 /** 279 * SkipWhites will skip definited white spaces as well as new lines 280 * automatically. 281 */ 282 INCLUDE_NEW_LINE = 1 283 }; 284 285 /** 286 * Skips any occurence of whitespaces specified in mWhitespaces member, 287 * optionally skip also new lines. 288 */ 289 void SkipWhites(WhiteSkipping aIncludeNewLines = DONT_INCLUDE_NEW_LINE); 290 291 /** 292 * Skips all tokens until the given one is found or EOF is hit. The token 293 * or EOF are next to read. 294 */ 295 void SkipUntil(typename base::Token const& aToken); 296 297 // These are mostly shortcuts for the Check() methods above. 298 299 /** 300 * Check whitespace character is present. 301 */ CheckWhite()302 [[nodiscard]] bool CheckWhite() { return Check(base::Token::Whitespace()); } 303 /** 304 * Check there is a single character on the read cursor position. If so, 305 * shift the read cursor position and return true. Otherwise false. 306 */ CheckChar(const TChar aChar)307 [[nodiscard]] bool CheckChar(const TChar aChar) { 308 return Check(base::Token::Char(aChar)); 309 } 310 /** 311 * This is a customizable version of CheckChar. aClassifier is a function 312 * called with value of the character on the current input read position. If 313 * this user function returns true, read cursor is shifted and true returned. 314 * Otherwise false. The user classifiction function is not called when we are 315 * at or past the end and false is immediately returned. 316 */ 317 [[nodiscard]] bool CheckChar(bool (*aClassifier)(const TChar aChar)); 318 /** 319 * Check for a whole expected word. 320 */ CheckWord(const typename base::TAString & aWord)321 [[nodiscard]] bool CheckWord(const typename base::TAString& aWord) { 322 return Check(base::Token::Word(aWord)); 323 } 324 /** 325 * Shortcut for literal const word check with compile time length calculation. 326 */ 327 template <uint32_t N> CheckWord(const TChar (& aWord)[N])328 [[nodiscard]] bool CheckWord(const TChar (&aWord)[N]) { 329 return Check( 330 base::Token::Word(typename base::TDependentString(aWord, N - 1))); 331 } 332 /** 333 * Checks \r, \n or \r\n. 334 */ CheckEOL()335 [[nodiscard]] bool CheckEOL() { return Check(base::Token::NewLine()); } 336 /** 337 * Checks we are at the end of the input string reading. If so, shift past 338 * the end and returns true. Otherwise does nothing and returns false. 339 */ CheckEOF()340 [[nodiscard]] bool CheckEOF() { return Check(base::Token::EndOfFile()); } 341 342 /** 343 * These are shortcuts to obtain the value immediately when the token type 344 * matches. 345 */ 346 [[nodiscard]] bool ReadChar(TChar* aValue); 347 [[nodiscard]] bool ReadChar(bool (*aClassifier)(const TChar aChar), 348 TChar* aValue); 349 [[nodiscard]] bool ReadWord(typename base::TAString& aValue); 350 [[nodiscard]] bool ReadWord(typename base::TDependentSubstring& aValue); 351 352 /** 353 * This is an integer read helper. It returns false and doesn't move the read 354 * cursor when any of the following happens: 355 * - the token at the read cursor is not an integer 356 * - the final number doesn't fit the T type 357 * Otherwise true is returned, aValue is filled with the integral number 358 * and the cursor is moved forward. 359 */ 360 template <typename T> ReadInteger(T * aValue)361 [[nodiscard]] bool ReadInteger(T* aValue) { 362 MOZ_RELEASE_ASSERT(aValue); 363 364 typename base::TAString::const_char_iterator rollback = mRollback; 365 typename base::TAString::const_char_iterator cursor = base::mCursor; 366 typename base::Token t; 367 if (!Check(base::TOKEN_INTEGER, t)) { 368 return false; 369 } 370 371 mozilla::CheckedInt<T> checked(t.AsInteger()); 372 if (!checked.isValid()) { 373 // Move to a state as if Check() call has failed 374 mRollback = rollback; 375 base::mCursor = cursor; 376 base::mHasFailed = true; 377 return false; 378 } 379 380 *aValue = checked.value(); 381 return true; 382 } 383 384 /** 385 * Same as above, but accepts an integer with an optional minus sign. 386 */ 387 template <typename T, typename V = std::enable_if_t< 388 std::is_signed_v<std::remove_pointer_t<T>>, 389 std::remove_pointer_t<T>>> ReadSignedInteger(T * aValue)390 [[nodiscard]] bool ReadSignedInteger(T* aValue) { 391 MOZ_RELEASE_ASSERT(aValue); 392 393 typename base::TAString::const_char_iterator rollback = mRollback; 394 typename base::TAString::const_char_iterator cursor = base::mCursor; 395 auto revert = MakeScopeExit([&] { 396 // Move to a state as if Check() call has failed 397 mRollback = rollback; 398 base::mCursor = cursor; 399 base::mHasFailed = true; 400 }); 401 402 // Using functional raw access because '-' could be part of the word set 403 // making CheckChar('-') not work. 404 bool minus = CheckChar([](const TChar aChar) { return aChar == '-'; }); 405 406 typename base::Token t; 407 if (!Check(base::TOKEN_INTEGER, t)) { 408 return false; 409 } 410 411 mozilla::CheckedInt<T> checked(t.AsInteger()); 412 if (minus) { 413 checked *= -1; 414 } 415 416 if (!checked.isValid()) { 417 return false; 418 } 419 420 *aValue = checked.value(); 421 revert.release(); 422 return true; 423 } 424 425 /** 426 * Returns the read cursor position back as it was before the last call of any 427 * parsing method of TTokenizer (Next, Check*, Skip*, Read*) so that the last 428 * operation can be repeated. Rollback cannot be used multiple times, it only 429 * reverts the last successfull parse operation. It also cannot be used 430 * before any parsing operation has been called on the TTokenizer. 431 */ 432 void Rollback(); 433 434 /** 435 * Record() and Claim() are collecting the input as it is being parsed to 436 * obtain a substring between particular syntax bounderies defined by any 437 * recursive descent parser or simple parser the TTokenizer is used to read 438 * the input for. Inlucsion of a token that has just been parsed can be 439 * controlled using an arguemnt. 440 */ 441 enum ClaimInclusion { 442 /** 443 * Include resulting (or passed) token of the last lexical analyzer 444 * operation in the result. 445 */ 446 INCLUDE_LAST, 447 /** 448 * Do not include it. 449 */ 450 EXCLUDE_LAST 451 }; 452 453 /** 454 * Start the process of recording. Based on aInclude value the begining of 455 * the recorded sub-string is at the current position (EXCLUDE_LAST) or at the 456 * position before the last parsed token (INCLUDE_LAST). 457 */ 458 void Record(ClaimInclusion aInclude = EXCLUDE_LAST); 459 /** 460 * Claim result of the record started with Record() call before. Depending on 461 * aInclude the ending of the sub-string result includes or excludes the last 462 * parsed or checked token. 463 */ 464 void Claim(typename base::TAString& aResult, 465 ClaimInclusion aInclude = EXCLUDE_LAST); 466 void Claim(typename base::TDependentSubstring& aResult, 467 ClaimInclusion aInclude = EXCLUDE_LAST); 468 469 /** 470 * If aToken is found, aResult is set to the substring between the current 471 * position and the position of aToken, potentially including aToken depending 472 * on aInclude. 473 * If aToken isn't found aResult is set to the substring between the current 474 * position and the end of the string. 475 * If aToken is found, the method returns true. Otherwise it returns false. 476 * 477 * Calling Rollback() after ReadUntil() will return the read cursor to the 478 * position it had before ReadUntil was called. 479 */ 480 [[nodiscard]] bool ReadUntil(typename base::Token const& aToken, 481 typename base::TDependentSubstring& aResult, 482 ClaimInclusion aInclude = EXCLUDE_LAST); 483 [[nodiscard]] bool ReadUntil(typename base::Token const& aToken, 484 typename base::TAString& aResult, 485 ClaimInclusion aInclude = EXCLUDE_LAST); 486 487 protected: 488 // All these point to the original buffer passed to the TTokenizer's 489 // constructor 490 typename base::TAString::const_char_iterator 491 mRecord; // Position where the recorded sub-string for Claim() is 492 typename base::TAString::const_char_iterator 493 mRollback; // Position of the previous token start 494 495 private: 496 TTokenizer() = delete; 497 TTokenizer(const TTokenizer&) = delete; 498 TTokenizer(TTokenizer&&) = delete; 499 TTokenizer(const TTokenizer&&) = delete; 500 TTokenizer& operator=(const TTokenizer&) = delete; 501 }; 502 503 typedef TTokenizer<char> Tokenizer; 504 typedef TTokenizer<char16_t> Tokenizer16; 505 506 } // namespace mozilla 507 508 #endif // Tokenizer_h__ 509