1 //===-- StringRef.cpp - Lightweight String References ---------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
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"
18 // MSVC emits references to this into the translation units which reference it.
20 const size_t StringRef::npos;
23 static char ascii_tolower(char x) {
24 if (x >= 'A' && x <= 'Z')
29 static char ascii_toupper(char x) {
30 if (x >= 'a' && x <= 'z')
35 static bool ascii_isdigit(char x) {
36 return x >= '0' && x <= '9';
39 /// compare_lower - Compare strings, ignoring case.
40 int 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]);
45 return LHC < RHC ? -1 : 1;
48 if (Length == RHS.Length)
50 return Length < RHS.Length ? -1 : 1;
53 /// compare_numeric - Compare strings, handle embedded numbers.
54 int 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.
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]);
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.
76 if (Data[I] != RHS.Data[I])
77 return (unsigned char)Data[I] < (unsigned char)RHS.Data[I] ? -1 : 1;
79 if (Length == RHS.Length)
81 return Length < RHS.Length ? -1 : 1;
84 // Compute the edit distance between the two given strings.
85 unsigned 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);
94 //===----------------------------------------------------------------------===//
96 //===----------------------------------------------------------------------===//
98 std::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]);
106 std::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]);
114 //===----------------------------------------------------------------------===//
116 //===----------------------------------------------------------------------===//
119 /// find - Search for the first string \arg Str in the string.
121 /// \return - The index of the first occurrence of \arg Str, or npos if not
123 size_t StringRef::find(StringRef Str, size_t From) const {
124 size_t N = Str.size();
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))
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;
145 unsigned Len = Length-From, Pos = From;
147 if (substr(Pos, N).equals(Str)) // See if this is the correct substring.
150 // Otherwise skip the appropriate number of bytes.
151 uint8_t Skip = BadCharSkip[(uint8_t)(*this)[Pos+N-1]];
159 /// rfind - Search for the last string \arg Str in the string.
161 /// \return - The index of the last occurrence of \arg Str, or npos if not
163 size_t StringRef::rfind(StringRef Str) const {
164 size_t N = Str.size();
167 for (size_t i = Length - N + 1, e = 0; i != e;) {
169 if (substr(i, N).equals(Str))
175 /// find_first_of - Find the first character in the string that is in \arg
176 /// Chars, or npos if not found.
178 /// Note: O(size() + Chars.size())
179 StringRef::size_type StringRef::find_first_of(StringRef Chars,
181 std::bitset<1 << CHAR_BIT> CharBits;
182 for (size_type i = 0; i != Chars.size(); ++i)
183 CharBits.set((unsigned char)Chars[i]);
185 for (size_type i = min(From, Length), e = Length; i != e; ++i)
186 if (CharBits.test((unsigned char)Data[i]))
191 /// find_first_not_of - Find the first character in the string that is not
192 /// \arg C or npos if not found.
193 StringRef::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)
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.
203 /// Note: O(size() + Chars.size())
204 StringRef::size_type StringRef::find_first_not_of(StringRef Chars,
206 std::bitset<1 << CHAR_BIT> CharBits;
207 for (size_type i = 0; i != Chars.size(); ++i)
208 CharBits.set((unsigned char)Chars[i]);
210 for (size_type i = min(From, Length), e = Length; i != e; ++i)
211 if (!CharBits.test((unsigned char)Data[i]))
216 /// find_last_of - Find the last character in the string that is in \arg C,
217 /// or npos if not found.
219 /// Note: O(size() + Chars.size())
220 StringRef::size_type StringRef::find_last_of(StringRef Chars,
222 std::bitset<1 << CHAR_BIT> CharBits;
223 for (size_type i = 0; i != Chars.size(); ++i)
224 CharBits.set((unsigned char)Chars[i]);
226 for (size_type i = min(From, Length) - 1, e = -1; i != e; --i)
227 if (CharBits.test((unsigned char)Data[i]))
232 void StringRef::split(SmallVectorImpl<StringRef> &A,
233 StringRef Separators, int MaxSplit,
234 bool KeepEmpty) const {
235 StringRef rest = *this;
237 // rest.data() is used to distinguish cases like "a," that splits into
238 // "a" + "" and "a" that splits into "a" + 0.
240 rest.data() != NULL && (MaxSplit < 0 || splits < MaxSplit);
242 std::pair<StringRef, StringRef> p = rest.split(Separators);
244 if (p.first.size() != 0 || KeepEmpty)
245 A.push_back(p.first);
248 // If we have a tail left, add it.
249 if (rest.data() != NULL && (rest.size() != 0 || KeepEmpty))
253 //===----------------------------------------------------------------------===//
254 // Helpful Algorithms
255 //===----------------------------------------------------------------------===//
257 /// count - Return the number of non-overlapped occurrences of \arg Str in
259 size_t StringRef::count(StringRef Str) const {
261 size_t N = Str.size();
264 for (size_t i = 0, e = Length - N + 1; i != e; ++i)
265 if (substr(i, N).equals(Str))
270 static unsigned GetAutoSenseRadix(StringRef &Str) {
271 if (Str.startswith("0x")) {
274 } else if (Str.startswith("0b")) {
277 } else if (Str.startswith("0")) {
285 /// GetAsUnsignedInteger - Workhorse method that converts a integer character
286 /// sequence of radix up to 36 to an unsigned long long value.
287 static bool GetAsUnsignedInteger(StringRef Str, unsigned Radix,
288 unsigned long long &Result) {
289 // Autosense radix if not specified.
291 Radix = GetAutoSenseRadix(Str);
293 // Empty strings (after the radix autosense) are invalid.
294 if (Str.empty()) return true;
296 // Parse all the bytes of the string given this radix. Watch for overflow.
298 while (!Str.empty()) {
300 if (Str[0] >= '0' && Str[0] <= '9')
301 CharVal = Str[0]-'0';
302 else if (Str[0] >= 'a' && Str[0] <= 'z')
303 CharVal = Str[0]-'a'+10;
304 else if (Str[0] >= 'A' && Str[0] <= 'Z')
305 CharVal = Str[0]-'A'+10;
309 // If the parsed value is larger than the integer radix, the string is
311 if (CharVal >= Radix)
314 // Add in this character.
315 unsigned long long PrevResult = Result;
316 Result = Result*Radix+CharVal;
318 // Check for overflow.
319 if (Result < PrevResult)
328 bool StringRef::getAsInteger(unsigned Radix, unsigned long long &Result) const {
329 return GetAsUnsignedInteger(*this, Radix, Result);
333 bool StringRef::getAsInteger(unsigned Radix, long long &Result) const {
334 unsigned long long ULLVal;
336 // Handle positive strings first.
337 if (empty() || front() != '-') {
338 if (GetAsUnsignedInteger(*this, Radix, ULLVal) ||
339 // Check for value so large it overflows a signed value.
340 (long long)ULLVal < 0)
346 // Get the positive part of the value.
347 if (GetAsUnsignedInteger(substr(1), Radix, ULLVal) ||
348 // Reject values so large they'd overflow as negative signed, but allow
349 // "-0". This negates the unsigned so that the negative isn't undefined
350 // on signed overflow.
351 (long long)-ULLVal > 0)
358 bool StringRef::getAsInteger(unsigned Radix, int &Result) const {
360 if (getAsInteger(Radix, Val) ||
367 bool StringRef::getAsInteger(unsigned Radix, unsigned &Result) const {
368 unsigned long long Val;
369 if (getAsInteger(Radix, Val) ||
370 (unsigned)Val != Val)
376 bool StringRef::getAsInteger(unsigned Radix, APInt &Result) const {
377 StringRef Str = *this;
379 // Autosense radix if not specified.
381 Radix = GetAutoSenseRadix(Str);
383 assert(Radix > 1 && Radix <= 36);
385 // Empty strings (after the radix autosense) are invalid.
386 if (Str.empty()) return true;
388 // Skip leading zeroes. This can be a significant improvement if
389 // it means we don't need > 64 bits.
390 while (!Str.empty() && Str.front() == '0')
393 // If it was nothing but zeroes....
395 Result = APInt(64, 0);
399 // (Over-)estimate the required number of bits.
400 unsigned Log2Radix = 0;
401 while ((1U << Log2Radix) < Radix) Log2Radix++;
402 bool IsPowerOf2Radix = ((1U << Log2Radix) == Radix);
404 unsigned BitWidth = Log2Radix * Str.size();
405 if (BitWidth < Result.getBitWidth())
406 BitWidth = Result.getBitWidth(); // don't shrink the result
408 Result = Result.zext(BitWidth);
410 APInt RadixAP, CharAP; // unused unless !IsPowerOf2Radix
411 if (!IsPowerOf2Radix) {
412 // These must have the same bit-width as Result.
413 RadixAP = APInt(BitWidth, Radix);
414 CharAP = APInt(BitWidth, 0);
417 // Parse all the bytes of the string given this radix.
419 while (!Str.empty()) {
421 if (Str[0] >= '0' && Str[0] <= '9')
422 CharVal = Str[0]-'0';
423 else if (Str[0] >= 'a' && Str[0] <= 'z')
424 CharVal = Str[0]-'a'+10;
425 else if (Str[0] >= 'A' && Str[0] <= 'Z')
426 CharVal = Str[0]-'A'+10;
430 // If the parsed value is larger than the integer radix, the string is
432 if (CharVal >= Radix)
435 // Add in this character.
436 if (IsPowerOf2Radix) {
437 Result <<= Log2Radix;