StringRef.cpp revision 01d53ec176a6be4585df1f43af11151988ca4b35
1//===-- StringRef.cpp - Lightweight String References ---------------------===// 2// 3// The LLVM Compiler Infrastructure 4// 5// This file is distributed under the University of Illinois Open Source 6// License. See LICENSE.TXT for details. 7// 8//===----------------------------------------------------------------------===// 9 10#include "llvm/ADT/StringRef.h" 11#include "llvm/ADT/APInt.h" 12#include "llvm/ADT/OwningPtr.h" 13#include "llvm/ADT/edit_distance.h" 14#include <bitset> 15 16using namespace llvm; 17 18// MSVC emits references to this into the translation units which reference it. 19#ifndef _MSC_VER 20const size_t StringRef::npos; 21#endif 22 23static char ascii_tolower(char x) { 24 if (x >= 'A' && x <= 'Z') 25 return x - 'A' + 'a'; 26 return x; 27} 28 29static char ascii_toupper(char x) { 30 if (x >= 'a' && x <= 'z') 31 return x - 'a' + 'A'; 32 return x; 33} 34 35static bool ascii_isdigit(char x) { 36 return x >= '0' && x <= '9'; 37} 38 39/// compare_lower - Compare strings, ignoring case. 40int StringRef::compare_lower(StringRef RHS) const { 41 for (size_t I = 0, E = min(Length, RHS.Length); I != E; ++I) { 42 unsigned char LHC = ascii_tolower(Data[I]); 43 unsigned char RHC = ascii_tolower(RHS.Data[I]); 44 if (LHC != RHC) 45 return LHC < RHC ? -1 : 1; 46 } 47 48 if (Length == RHS.Length) 49 return 0; 50 return Length < RHS.Length ? -1 : 1; 51} 52 53/// compare_numeric - Compare strings, handle embedded numbers. 54int StringRef::compare_numeric(StringRef RHS) const { 55 for (size_t I = 0, E = min(Length, RHS.Length); I != E; ++I) { 56 // Check for sequences of digits. 57 if (ascii_isdigit(Data[I]) && ascii_isdigit(RHS.Data[I])) { 58 // The longer sequence of numbers is considered larger. 59 // This doesn't really handle prefixed zeros well. 60 size_t J; 61 for (J = I + 1; J != E + 1; ++J) { 62 bool ld = J < Length && ascii_isdigit(Data[J]); 63 bool rd = J < RHS.Length && ascii_isdigit(RHS.Data[J]); 64 if (ld != rd) 65 return rd ? -1 : 1; 66 if (!rd) 67 break; 68 } 69 // The two number sequences have the same length (J-I), just memcmp them. 70 if (int Res = compareMemory(Data + I, RHS.Data + I, J - I)) 71 return Res < 0 ? -1 : 1; 72 // Identical number sequences, continue search after the numbers. 73 I = J - 1; 74 continue; 75 } 76 if (Data[I] != RHS.Data[I]) 77 return (unsigned char)Data[I] < (unsigned char)RHS.Data[I] ? -1 : 1; 78 } 79 if (Length == RHS.Length) 80 return 0; 81 return Length < RHS.Length ? -1 : 1; 82} 83 84// Compute the edit distance between the two given strings. 85unsigned StringRef::edit_distance(llvm::StringRef Other, 86 bool AllowReplacements, 87 unsigned MaxEditDistance) { 88 return llvm::ComputeEditDistance( 89 llvm::ArrayRef<char>(data(), size()), 90 llvm::ArrayRef<char>(Other.data(), Other.size()), 91 AllowReplacements, MaxEditDistance); 92} 93 94//===----------------------------------------------------------------------===// 95// String Operations 96//===----------------------------------------------------------------------===// 97 98std::string StringRef::lower() const { 99 std::string Result(size(), char()); 100 for (size_type i = 0, e = size(); i != e; ++i) { 101 Result[i] = ascii_tolower(Data[i]); 102 } 103 return Result; 104} 105 106std::string StringRef::upper() const { 107 std::string Result(size(), char()); 108 for (size_type i = 0, e = size(); i != e; ++i) { 109 Result[i] = ascii_toupper(Data[i]); 110 } 111 return Result; 112} 113 114//===----------------------------------------------------------------------===// 115// String Searching 116//===----------------------------------------------------------------------===// 117 118 119/// find - Search for the first string \arg Str in the string. 120/// 121/// \return - The index of the first occurrence of \arg Str, or npos if not 122/// found. 123size_t StringRef::find(StringRef Str, size_t From) const { 124 size_t N = Str.size(); 125 if (N > Length) 126 return npos; 127 128 // For short haystacks or unsupported needles fall back to the naive algorithm 129 if (Length < 16 || N > 255 || N == 0) { 130 for (size_t e = Length - N + 1, i = min(From, e); i != e; ++i) 131 if (substr(i, N).equals(Str)) 132 return i; 133 return npos; 134 } 135 136 if (From >= Length) 137 return npos; 138 139 // Build the bad char heuristic table, with uint8_t to reduce cache thrashing. 140 uint8_t BadCharSkip[256]; 141 std::memset(BadCharSkip, N, 256); 142 for (unsigned i = 0; i != N-1; ++i) 143 BadCharSkip[(uint8_t)Str[i]] = N-1-i; 144 145 unsigned Len = Length-From, Pos = From; 146 while (Len >= N) { 147 if (substr(Pos, N).equals(Str)) // See if this is the correct substring. 148 return Pos; 149 150 // Otherwise skip the appropriate number of bytes. 151 uint8_t Skip = BadCharSkip[(uint8_t)(*this)[Pos+N-1]]; 152 Len -= Skip; 153 Pos += Skip; 154 } 155 156 return npos; 157} 158 159/// rfind - Search for the last string \arg Str in the string. 160/// 161/// \return - The index of the last occurrence of \arg Str, or npos if not 162/// found. 163size_t StringRef::rfind(StringRef Str) const { 164 size_t N = Str.size(); 165 if (N > Length) 166 return npos; 167 for (size_t i = Length - N + 1, e = 0; i != e;) { 168 --i; 169 if (substr(i, N).equals(Str)) 170 return i; 171 } 172 return npos; 173} 174 175/// find_first_of - Find the first character in the string that is in \arg 176/// Chars, or npos if not found. 177/// 178/// Note: O(size() + Chars.size()) 179StringRef::size_type StringRef::find_first_of(StringRef Chars, 180 size_t From) const { 181 std::bitset<1 << CHAR_BIT> CharBits; 182 for (size_type i = 0; i != Chars.size(); ++i) 183 CharBits.set((unsigned char)Chars[i]); 184 185 for (size_type i = min(From, Length), e = Length; i != e; ++i) 186 if (CharBits.test((unsigned char)Data[i])) 187 return i; 188 return npos; 189} 190 191/// find_first_not_of - Find the first character in the string that is not 192/// \arg C or npos if not found. 193StringRef::size_type StringRef::find_first_not_of(char C, size_t From) const { 194 for (size_type i = min(From, Length), e = Length; i != e; ++i) 195 if (Data[i] != C) 196 return i; 197 return npos; 198} 199 200/// find_first_not_of - Find the first character in the string that is not 201/// in the string \arg Chars, or npos if not found. 202/// 203/// Note: O(size() + Chars.size()) 204StringRef::size_type StringRef::find_first_not_of(StringRef Chars, 205 size_t From) const { 206 std::bitset<1 << CHAR_BIT> CharBits; 207 for (size_type i = 0; i != Chars.size(); ++i) 208 CharBits.set((unsigned char)Chars[i]); 209 210 for (size_type i = min(From, Length), e = Length; i != e; ++i) 211 if (!CharBits.test((unsigned char)Data[i])) 212 return i; 213 return npos; 214} 215 216/// find_last_of - Find the last character in the string that is in \arg C, 217/// or npos if not found. 218/// 219/// Note: O(size() + Chars.size()) 220StringRef::size_type StringRef::find_last_of(StringRef Chars, 221 size_t From) const { 222 std::bitset<1 << CHAR_BIT> CharBits; 223 for (size_type i = 0; i != Chars.size(); ++i) 224 CharBits.set((unsigned char)Chars[i]); 225 226 for (size_type i = min(From, Length) - 1, e = -1; i != e; --i) 227 if (CharBits.test((unsigned char)Data[i])) 228 return i; 229 return npos; 230} 231 232//===----------------------------------------------------------------------===// 233// Helpful Algorithms 234//===----------------------------------------------------------------------===// 235 236/// count - Return the number of non-overlapped occurrences of \arg Str in 237/// the string. 238size_t StringRef::count(StringRef Str) const { 239 size_t Count = 0; 240 size_t N = Str.size(); 241 if (N > Length) 242 return 0; 243 for (size_t i = 0, e = Length - N + 1; i != e; ++i) 244 if (substr(i, N).equals(Str)) 245 ++Count; 246 return Count; 247} 248 249static unsigned GetAutoSenseRadix(StringRef &Str) { 250 if (Str.startswith("0x")) { 251 Str = Str.substr(2); 252 return 16; 253 } else if (Str.startswith("0b")) { 254 Str = Str.substr(2); 255 return 2; 256 } else if (Str.startswith("0")) { 257 return 8; 258 } else { 259 return 10; 260 } 261} 262 263 264/// GetAsUnsignedInteger - Workhorse method that converts a integer character 265/// sequence of radix up to 36 to an unsigned long long value. 266static bool GetAsUnsignedInteger(StringRef Str, unsigned Radix, 267 unsigned long long &Result) { 268 // Autosense radix if not specified. 269 if (Radix == 0) 270 Radix = GetAutoSenseRadix(Str); 271 272 // Empty strings (after the radix autosense) are invalid. 273 if (Str.empty()) return true; 274 275 // Parse all the bytes of the string given this radix. Watch for overflow. 276 Result = 0; 277 while (!Str.empty()) { 278 unsigned CharVal; 279 if (Str[0] >= '0' && Str[0] <= '9') 280 CharVal = Str[0]-'0'; 281 else if (Str[0] >= 'a' && Str[0] <= 'z') 282 CharVal = Str[0]-'a'+10; 283 else if (Str[0] >= 'A' && Str[0] <= 'Z') 284 CharVal = Str[0]-'A'+10; 285 else 286 return true; 287 288 // If the parsed value is larger than the integer radix, the string is 289 // invalid. 290 if (CharVal >= Radix) 291 return true; 292 293 // Add in this character. 294 unsigned long long PrevResult = Result; 295 Result = Result*Radix+CharVal; 296 297 // Check for overflow. 298 if (Result < PrevResult) 299 return true; 300 301 Str = Str.substr(1); 302 } 303 304 return false; 305} 306 307bool StringRef::getAsInteger(unsigned Radix, unsigned long long &Result) const { 308 return GetAsUnsignedInteger(*this, Radix, Result); 309} 310 311 312bool StringRef::getAsInteger(unsigned Radix, long long &Result) const { 313 unsigned long long ULLVal; 314 315 // Handle positive strings first. 316 if (empty() || front() != '-') { 317 if (GetAsUnsignedInteger(*this, Radix, ULLVal) || 318 // Check for value so large it overflows a signed value. 319 (long long)ULLVal < 0) 320 return true; 321 Result = ULLVal; 322 return false; 323 } 324 325 // Get the positive part of the value. 326 if (GetAsUnsignedInteger(substr(1), Radix, ULLVal) || 327 // Reject values so large they'd overflow as negative signed, but allow 328 // "-0". This negates the unsigned so that the negative isn't undefined 329 // on signed overflow. 330 (long long)-ULLVal > 0) 331 return true; 332 333 Result = -ULLVal; 334 return false; 335} 336 337bool StringRef::getAsInteger(unsigned Radix, int &Result) const { 338 long long Val; 339 if (getAsInteger(Radix, Val) || 340 (int)Val != Val) 341 return true; 342 Result = Val; 343 return false; 344} 345 346bool StringRef::getAsInteger(unsigned Radix, unsigned &Result) const { 347 unsigned long long Val; 348 if (getAsInteger(Radix, Val) || 349 (unsigned)Val != Val) 350 return true; 351 Result = Val; 352 return false; 353} 354 355bool StringRef::getAsInteger(unsigned Radix, APInt &Result) const { 356 StringRef Str = *this; 357 358 // Autosense radix if not specified. 359 if (Radix == 0) 360 Radix = GetAutoSenseRadix(Str); 361 362 assert(Radix > 1 && Radix <= 36); 363 364 // Empty strings (after the radix autosense) are invalid. 365 if (Str.empty()) return true; 366 367 // Skip leading zeroes. This can be a significant improvement if 368 // it means we don't need > 64 bits. 369 while (!Str.empty() && Str.front() == '0') 370 Str = Str.substr(1); 371 372 // If it was nothing but zeroes.... 373 if (Str.empty()) { 374 Result = APInt(64, 0); 375 return false; 376 } 377 378 // (Over-)estimate the required number of bits. 379 unsigned Log2Radix = 0; 380 while ((1U << Log2Radix) < Radix) Log2Radix++; 381 bool IsPowerOf2Radix = ((1U << Log2Radix) == Radix); 382 383 unsigned BitWidth = Log2Radix * Str.size(); 384 if (BitWidth < Result.getBitWidth()) 385 BitWidth = Result.getBitWidth(); // don't shrink the result 386 else 387 Result = Result.zext(BitWidth); 388 389 APInt RadixAP, CharAP; // unused unless !IsPowerOf2Radix 390 if (!IsPowerOf2Radix) { 391 // These must have the same bit-width as Result. 392 RadixAP = APInt(BitWidth, Radix); 393 CharAP = APInt(BitWidth, 0); 394 } 395 396 // Parse all the bytes of the string given this radix. 397 Result = 0; 398 while (!Str.empty()) { 399 unsigned CharVal; 400 if (Str[0] >= '0' && Str[0] <= '9') 401 CharVal = Str[0]-'0'; 402 else if (Str[0] >= 'a' && Str[0] <= 'z') 403 CharVal = Str[0]-'a'+10; 404 else if (Str[0] >= 'A' && Str[0] <= 'Z') 405 CharVal = Str[0]-'A'+10; 406 else 407 return true; 408 409 // If the parsed value is larger than the integer radix, the string is 410 // invalid. 411 if (CharVal >= Radix) 412 return true; 413 414 // Add in this character. 415 if (IsPowerOf2Radix) { 416 Result <<= Log2Radix; 417 Result |= CharVal; 418 } else { 419 Result *= RadixAP; 420 CharAP = CharVal; 421 Result += CharAP; 422 } 423 424 Str = Str.substr(1); 425 } 426 427 return false; 428} 429