1 //===-- StringRef.cpp - Lightweight String References ---------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #include "llvm/ADT/StringRef.h"
10 #include "llvm/ADT/APFloat.h"
11 #include "llvm/ADT/APInt.h"
12 #include "llvm/ADT/Hashing.h"
13 #include "llvm/ADT/StringExtras.h"
14 #include "llvm/ADT/edit_distance.h"
15 #include "llvm/Support/Error.h"
16 #include <bitset>
17 
18 using namespace llvm;
19 
20 // MSVC emits references to this into the translation units which reference it.
21 #ifndef _MSC_VER
22 constexpr size_t StringRef::npos;
23 #endif
24 
25 // strncasecmp() is not available on non-POSIX systems, so define an
26 // alternative function here.
27 static int ascii_strncasecmp(const char *LHS, const char *RHS, size_t Length) {
28   for (size_t I = 0; I < Length; ++I) {
29     unsigned char LHC = toLower(LHS[I]);
30     unsigned char RHC = toLower(RHS[I]);
31     if (LHC != RHC)
32       return LHC < RHC ? -1 : 1;
33   }
34   return 0;
35 }
36 
37 int StringRef::compare_insensitive(StringRef RHS) const {
38   if (int Res = ascii_strncasecmp(Data, RHS.Data, std::min(Length, RHS.Length)))
39     return Res;
40   if (Length == RHS.Length)
41     return 0;
42   return Length < RHS.Length ? -1 : 1;
43 }
44 
45 bool StringRef::starts_with_insensitive(StringRef Prefix) const {
46   return Length >= Prefix.Length &&
47       ascii_strncasecmp(Data, Prefix.Data, Prefix.Length) == 0;
48 }
49 
50 bool StringRef::ends_with_insensitive(StringRef Suffix) const {
51   return Length >= Suffix.Length &&
52       ascii_strncasecmp(end() - Suffix.Length, Suffix.Data, Suffix.Length) == 0;
53 }
54 
55 size_t StringRef::find_insensitive(char C, size_t From) const {
56   char L = toLower(C);
57   return find_if([L](char D) { return toLower(D) == L; }, From);
58 }
59 
60 /// compare_numeric - Compare strings, handle embedded numbers.
61 int StringRef::compare_numeric(StringRef RHS) const {
62   for (size_t I = 0, E = std::min(Length, RHS.Length); I != E; ++I) {
63     // Check for sequences of digits.
64     if (isDigit(Data[I]) && isDigit(RHS.Data[I])) {
65       // The longer sequence of numbers is considered larger.
66       // This doesn't really handle prefixed zeros well.
67       size_t J;
68       for (J = I + 1; J != E + 1; ++J) {
69         bool ld = J < Length && isDigit(Data[J]);
70         bool rd = J < RHS.Length && isDigit(RHS.Data[J]);
71         if (ld != rd)
72           return rd ? -1 : 1;
73         if (!rd)
74           break;
75       }
76       // The two number sequences have the same length (J-I), just memcmp them.
77       if (int Res = compareMemory(Data + I, RHS.Data + I, J - I))
78         return Res < 0 ? -1 : 1;
79       // Identical number sequences, continue search after the numbers.
80       I = J - 1;
81       continue;
82     }
83     if (Data[I] != RHS.Data[I])
84       return (unsigned char)Data[I] < (unsigned char)RHS.Data[I] ? -1 : 1;
85   }
86   if (Length == RHS.Length)
87     return 0;
88   return Length < RHS.Length ? -1 : 1;
89 }
90 
91 // Compute the edit distance between the two given strings.
92 unsigned StringRef::edit_distance(llvm::StringRef Other,
93                                   bool AllowReplacements,
94                                   unsigned MaxEditDistance) const {
95   return llvm::ComputeEditDistance(ArrayRef(data(), size()),
96                                    ArrayRef(Other.data(), Other.size()),
97                                    AllowReplacements, MaxEditDistance);
98 }
99 
100 unsigned llvm::StringRef::edit_distance_insensitive(
101     StringRef Other, bool AllowReplacements, unsigned MaxEditDistance) const {
102   return llvm::ComputeMappedEditDistance(
103       ArrayRef(data(), size()), ArrayRef(Other.data(), Other.size()),
104       llvm::toLower, AllowReplacements, MaxEditDistance);
105 }
106 
107 //===----------------------------------------------------------------------===//
108 // String Operations
109 //===----------------------------------------------------------------------===//
110 
111 std::string StringRef::lower() const {
112   return std::string(map_iterator(begin(), toLower),
113                      map_iterator(end(), toLower));
114 }
115 
116 std::string StringRef::upper() const {
117   return std::string(map_iterator(begin(), toUpper),
118                      map_iterator(end(), toUpper));
119 }
120 
121 //===----------------------------------------------------------------------===//
122 // String Searching
123 //===----------------------------------------------------------------------===//
124 
125 
126 /// find - Search for the first string \arg Str in the string.
127 ///
128 /// \return - The index of the first occurrence of \arg Str, or npos if not
129 /// found.
130 size_t StringRef::find(StringRef Str, size_t From) const {
131   if (From > Length)
132     return npos;
133 
134   const char *Start = Data + From;
135   size_t Size = Length - From;
136 
137   const char *Needle = Str.data();
138   size_t N = Str.size();
139   if (N == 0)
140     return From;
141   if (Size < N)
142     return npos;
143   if (N == 1) {
144     const char *Ptr = (const char *)::memchr(Start, Needle[0], Size);
145     return Ptr == nullptr ? npos : Ptr - Data;
146   }
147 
148   const char *Stop = Start + (Size - N + 1);
149 
150   if (N == 2) {
151     // Provide a fast path for newline finding (CRLF case) in InclusionRewriter.
152     // Not the most optimized strategy, but getting memcmp inlined should be
153     // good enough.
154     do {
155       if (std::memcmp(Start, Needle, 2) == 0)
156         return Start - Data;
157       ++Start;
158     } while (Start < Stop);
159     return npos;
160   }
161 
162   // For short haystacks or unsupported needles fall back to the naive algorithm
163   if (Size < 16 || N > 255) {
164     do {
165       if (std::memcmp(Start, Needle, N) == 0)
166         return Start - Data;
167       ++Start;
168     } while (Start < Stop);
169     return npos;
170   }
171 
172   // Build the bad char heuristic table, with uint8_t to reduce cache thrashing.
173   uint8_t BadCharSkip[256];
174   std::memset(BadCharSkip, N, 256);
175   for (unsigned i = 0; i != N-1; ++i)
176     BadCharSkip[(uint8_t)Str[i]] = N-1-i;
177 
178   do {
179     uint8_t Last = Start[N - 1];
180     if (LLVM_UNLIKELY(Last == (uint8_t)Needle[N - 1]))
181       if (std::memcmp(Start, Needle, N - 1) == 0)
182         return Start - Data;
183 
184     // Otherwise skip the appropriate number of bytes.
185     Start += BadCharSkip[Last];
186   } while (Start < Stop);
187 
188   return npos;
189 }
190 
191 size_t StringRef::find_insensitive(StringRef Str, size_t From) const {
192   StringRef This = substr(From);
193   while (This.size() >= Str.size()) {
194     if (This.starts_with_insensitive(Str))
195       return From;
196     This = This.drop_front();
197     ++From;
198   }
199   return npos;
200 }
201 
202 size_t StringRef::rfind_insensitive(char C, size_t From) const {
203   From = std::min(From, Length);
204   size_t i = From;
205   while (i != 0) {
206     --i;
207     if (toLower(Data[i]) == toLower(C))
208       return i;
209   }
210   return npos;
211 }
212 
213 /// rfind - Search for the last string \arg Str in the string.
214 ///
215 /// \return - The index of the last occurrence of \arg Str, or npos if not
216 /// found.
217 size_t StringRef::rfind(StringRef Str) const {
218   return std::string_view(*this).rfind(Str);
219 }
220 
221 size_t StringRef::rfind_insensitive(StringRef Str) const {
222   size_t N = Str.size();
223   if (N > Length)
224     return npos;
225   for (size_t i = Length - N + 1, e = 0; i != e;) {
226     --i;
227     if (substr(i, N).equals_insensitive(Str))
228       return i;
229   }
230   return npos;
231 }
232 
233 /// find_first_of - Find the first character in the string that is in \arg
234 /// Chars, or npos if not found.
235 ///
236 /// Note: O(size() + Chars.size())
237 StringRef::size_type StringRef::find_first_of(StringRef Chars,
238                                               size_t From) const {
239   std::bitset<1 << CHAR_BIT> CharBits;
240   for (char C : Chars)
241     CharBits.set((unsigned char)C);
242 
243   for (size_type i = std::min(From, Length), e = Length; i != e; ++i)
244     if (CharBits.test((unsigned char)Data[i]))
245       return i;
246   return npos;
247 }
248 
249 /// find_first_not_of - Find the first character in the string that is not
250 /// \arg C or npos if not found.
251 StringRef::size_type StringRef::find_first_not_of(char C, size_t From) const {
252   return std::string_view(*this).find_first_not_of(C, From);
253 }
254 
255 /// find_first_not_of - Find the first character in the string that is not
256 /// in the string \arg Chars, or npos if not found.
257 ///
258 /// Note: O(size() + Chars.size())
259 StringRef::size_type StringRef::find_first_not_of(StringRef Chars,
260                                                   size_t From) const {
261   std::bitset<1 << CHAR_BIT> CharBits;
262   for (char C : Chars)
263     CharBits.set((unsigned char)C);
264 
265   for (size_type i = std::min(From, Length), e = Length; i != e; ++i)
266     if (!CharBits.test((unsigned char)Data[i]))
267       return i;
268   return npos;
269 }
270 
271 /// find_last_of - Find the last character in the string that is in \arg C,
272 /// or npos if not found.
273 ///
274 /// Note: O(size() + Chars.size())
275 StringRef::size_type StringRef::find_last_of(StringRef Chars,
276                                              size_t From) const {
277   std::bitset<1 << CHAR_BIT> CharBits;
278   for (char C : Chars)
279     CharBits.set((unsigned char)C);
280 
281   for (size_type i = std::min(From, Length) - 1, e = -1; i != e; --i)
282     if (CharBits.test((unsigned char)Data[i]))
283       return i;
284   return npos;
285 }
286 
287 /// find_last_not_of - Find the last character in the string that is not
288 /// \arg C, or npos if not found.
289 StringRef::size_type StringRef::find_last_not_of(char C, size_t From) const {
290   for (size_type i = std::min(From, Length) - 1, e = -1; i != e; --i)
291     if (Data[i] != C)
292       return i;
293   return npos;
294 }
295 
296 /// find_last_not_of - Find the last character in the string that is not in
297 /// \arg Chars, or npos if not found.
298 ///
299 /// Note: O(size() + Chars.size())
300 StringRef::size_type StringRef::find_last_not_of(StringRef Chars,
301                                                  size_t From) const {
302   std::bitset<1 << CHAR_BIT> CharBits;
303   for (char C : Chars)
304     CharBits.set((unsigned char)C);
305 
306   for (size_type i = std::min(From, Length) - 1, e = -1; i != e; --i)
307     if (!CharBits.test((unsigned char)Data[i]))
308       return i;
309   return npos;
310 }
311 
312 void StringRef::split(SmallVectorImpl<StringRef> &A,
313                       StringRef Separator, int MaxSplit,
314                       bool KeepEmpty) const {
315   StringRef S = *this;
316 
317   // Count down from MaxSplit. When MaxSplit is -1, this will just split
318   // "forever". This doesn't support splitting more than 2^31 times
319   // intentionally; if we ever want that we can make MaxSplit a 64-bit integer
320   // but that seems unlikely to be useful.
321   while (MaxSplit-- != 0) {
322     size_t Idx = S.find(Separator);
323     if (Idx == npos)
324       break;
325 
326     // Push this split.
327     if (KeepEmpty || Idx > 0)
328       A.push_back(S.slice(0, Idx));
329 
330     // Jump forward.
331     S = S.slice(Idx + Separator.size(), npos);
332   }
333 
334   // Push the tail.
335   if (KeepEmpty || !S.empty())
336     A.push_back(S);
337 }
338 
339 void StringRef::split(SmallVectorImpl<StringRef> &A, char Separator,
340                       int MaxSplit, bool KeepEmpty) const {
341   StringRef S = *this;
342 
343   // Count down from MaxSplit. When MaxSplit is -1, this will just split
344   // "forever". This doesn't support splitting more than 2^31 times
345   // intentionally; if we ever want that we can make MaxSplit a 64-bit integer
346   // but that seems unlikely to be useful.
347   while (MaxSplit-- != 0) {
348     size_t Idx = S.find(Separator);
349     if (Idx == npos)
350       break;
351 
352     // Push this split.
353     if (KeepEmpty || Idx > 0)
354       A.push_back(S.slice(0, Idx));
355 
356     // Jump forward.
357     S = S.slice(Idx + 1, npos);
358   }
359 
360   // Push the tail.
361   if (KeepEmpty || !S.empty())
362     A.push_back(S);
363 }
364 
365 //===----------------------------------------------------------------------===//
366 // Helpful Algorithms
367 //===----------------------------------------------------------------------===//
368 
369 /// count - Return the number of non-overlapped occurrences of \arg Str in
370 /// the string.
371 size_t StringRef::count(StringRef Str) const {
372   size_t Count = 0;
373   size_t Pos = 0;
374   size_t N = Str.size();
375   // TODO: For an empty `Str` we return 0 for legacy reasons. Consider changing
376   //       this to `Length + 1` which is more in-line with the function
377   //       description.
378   if (!N)
379     return 0;
380   while ((Pos = find(Str, Pos)) != npos) {
381     ++Count;
382     Pos += N;
383   }
384   return Count;
385 }
386 
387 static unsigned GetAutoSenseRadix(StringRef &Str) {
388   if (Str.empty())
389     return 10;
390 
391   if (Str.startswith("0x") || Str.startswith("0X")) {
392     Str = Str.substr(2);
393     return 16;
394   }
395 
396   if (Str.startswith("0b") || Str.startswith("0B")) {
397     Str = Str.substr(2);
398     return 2;
399   }
400 
401   if (Str.startswith("0o")) {
402     Str = Str.substr(2);
403     return 8;
404   }
405 
406   if (Str[0] == '0' && Str.size() > 1 && isDigit(Str[1])) {
407     Str = Str.substr(1);
408     return 8;
409   }
410 
411   return 10;
412 }
413 
414 bool llvm::consumeUnsignedInteger(StringRef &Str, unsigned Radix,
415                                   unsigned long long &Result) {
416   // Autosense radix if not specified.
417   if (Radix == 0)
418     Radix = GetAutoSenseRadix(Str);
419 
420   // Empty strings (after the radix autosense) are invalid.
421   if (Str.empty()) return true;
422 
423   // Parse all the bytes of the string given this radix.  Watch for overflow.
424   StringRef Str2 = Str;
425   Result = 0;
426   while (!Str2.empty()) {
427     unsigned CharVal;
428     if (Str2[0] >= '0' && Str2[0] <= '9')
429       CharVal = Str2[0] - '0';
430     else if (Str2[0] >= 'a' && Str2[0] <= 'z')
431       CharVal = Str2[0] - 'a' + 10;
432     else if (Str2[0] >= 'A' && Str2[0] <= 'Z')
433       CharVal = Str2[0] - 'A' + 10;
434     else
435       break;
436 
437     // If the parsed value is larger than the integer radix, we cannot
438     // consume any more characters.
439     if (CharVal >= Radix)
440       break;
441 
442     // Add in this character.
443     unsigned long long PrevResult = Result;
444     Result = Result * Radix + CharVal;
445 
446     // Check for overflow by shifting back and seeing if bits were lost.
447     if (Result / Radix < PrevResult)
448       return true;
449 
450     Str2 = Str2.substr(1);
451   }
452 
453   // We consider the operation a failure if no characters were consumed
454   // successfully.
455   if (Str.size() == Str2.size())
456     return true;
457 
458   Str = Str2;
459   return false;
460 }
461 
462 bool llvm::consumeSignedInteger(StringRef &Str, unsigned Radix,
463                                 long long &Result) {
464   unsigned long long ULLVal;
465 
466   // Handle positive strings first.
467   if (Str.empty() || Str.front() != '-') {
468     if (consumeUnsignedInteger(Str, Radix, ULLVal) ||
469         // Check for value so large it overflows a signed value.
470         (long long)ULLVal < 0)
471       return true;
472     Result = ULLVal;
473     return false;
474   }
475 
476   // Get the positive part of the value.
477   StringRef Str2 = Str.drop_front(1);
478   if (consumeUnsignedInteger(Str2, Radix, ULLVal) ||
479       // Reject values so large they'd overflow as negative signed, but allow
480       // "-0".  This negates the unsigned so that the negative isn't undefined
481       // on signed overflow.
482       (long long)-ULLVal > 0)
483     return true;
484 
485   Str = Str2;
486   Result = -ULLVal;
487   return false;
488 }
489 
490 /// GetAsUnsignedInteger - Workhorse method that converts a integer character
491 /// sequence of radix up to 36 to an unsigned long long value.
492 bool llvm::getAsUnsignedInteger(StringRef Str, unsigned Radix,
493                                 unsigned long long &Result) {
494   if (consumeUnsignedInteger(Str, Radix, Result))
495     return true;
496 
497   // For getAsUnsignedInteger, we require the whole string to be consumed or
498   // else we consider it a failure.
499   return !Str.empty();
500 }
501 
502 bool llvm::getAsSignedInteger(StringRef Str, unsigned Radix,
503                               long long &Result) {
504   if (consumeSignedInteger(Str, Radix, Result))
505     return true;
506 
507   // For getAsSignedInteger, we require the whole string to be consumed or else
508   // we consider it a failure.
509   return !Str.empty();
510 }
511 
512 bool StringRef::consumeInteger(unsigned Radix, APInt &Result) {
513   StringRef Str = *this;
514 
515   // Autosense radix if not specified.
516   if (Radix == 0)
517     Radix = GetAutoSenseRadix(Str);
518 
519   assert(Radix > 1 && Radix <= 36);
520 
521   // Empty strings (after the radix autosense) are invalid.
522   if (Str.empty()) return true;
523 
524   // Skip leading zeroes.  This can be a significant improvement if
525   // it means we don't need > 64 bits.
526   while (!Str.empty() && Str.front() == '0')
527     Str = Str.substr(1);
528 
529   // If it was nothing but zeroes....
530   if (Str.empty()) {
531     Result = APInt(64, 0);
532     *this = Str;
533     return false;
534   }
535 
536   // (Over-)estimate the required number of bits.
537   unsigned Log2Radix = 0;
538   while ((1U << Log2Radix) < Radix) Log2Radix++;
539   bool IsPowerOf2Radix = ((1U << Log2Radix) == Radix);
540 
541   unsigned BitWidth = Log2Radix * Str.size();
542   if (BitWidth < Result.getBitWidth())
543     BitWidth = Result.getBitWidth(); // don't shrink the result
544   else if (BitWidth > Result.getBitWidth())
545     Result = Result.zext(BitWidth);
546 
547   APInt RadixAP, CharAP; // unused unless !IsPowerOf2Radix
548   if (!IsPowerOf2Radix) {
549     // These must have the same bit-width as Result.
550     RadixAP = APInt(BitWidth, Radix);
551     CharAP = APInt(BitWidth, 0);
552   }
553 
554   // Parse all the bytes of the string given this radix.
555   Result = 0;
556   while (!Str.empty()) {
557     unsigned CharVal;
558     if (Str[0] >= '0' && Str[0] <= '9')
559       CharVal = Str[0]-'0';
560     else if (Str[0] >= 'a' && Str[0] <= 'z')
561       CharVal = Str[0]-'a'+10;
562     else if (Str[0] >= 'A' && Str[0] <= 'Z')
563       CharVal = Str[0]-'A'+10;
564     else
565       break;
566 
567     // If the parsed value is larger than the integer radix, the string is
568     // invalid.
569     if (CharVal >= Radix)
570       break;
571 
572     // Add in this character.
573     if (IsPowerOf2Radix) {
574       Result <<= Log2Radix;
575       Result |= CharVal;
576     } else {
577       Result *= RadixAP;
578       CharAP = CharVal;
579       Result += CharAP;
580     }
581 
582     Str = Str.substr(1);
583   }
584 
585   // We consider the operation a failure if no characters were consumed
586   // successfully.
587   if (size() == Str.size())
588     return true;
589 
590   *this = Str;
591   return false;
592 }
593 
594 bool StringRef::getAsInteger(unsigned Radix, APInt &Result) const {
595   StringRef Str = *this;
596   if (Str.consumeInteger(Radix, Result))
597     return true;
598 
599   // For getAsInteger, we require the whole string to be consumed or else we
600   // consider it a failure.
601   return !Str.empty();
602 }
603 
604 bool StringRef::getAsDouble(double &Result, bool AllowInexact) const {
605   APFloat F(0.0);
606   auto StatusOrErr = F.convertFromString(*this, APFloat::rmNearestTiesToEven);
607   if (errorToBool(StatusOrErr.takeError()))
608     return true;
609 
610   APFloat::opStatus Status = *StatusOrErr;
611   if (Status != APFloat::opOK) {
612     if (!AllowInexact || !(Status & APFloat::opInexact))
613       return true;
614   }
615 
616   Result = F.convertToDouble();
617   return false;
618 }
619 
620 // Implementation of StringRef hashing.
621 hash_code llvm::hash_value(StringRef S) {
622   return hash_combine_range(S.begin(), S.end());
623 }
624 
625 unsigned DenseMapInfo<StringRef, void>::getHashValue(StringRef Val) {
626   assert(Val.data() != getEmptyKey().data() &&
627          "Cannot hash the empty key!");
628   assert(Val.data() != getTombstoneKey().data() &&
629          "Cannot hash the tombstone key!");
630   return (unsigned)(hash_value(Val));
631 }
632