Roll src/third_party/WebKit d9c6159:8139f33 (svn 201974:201975)
[chromium-blink-merge.git] / base / strings / string_number_conversions.cc
blobb6b65d2a3bf8ce53dc1fd463006ce1bc88877377
1 // Copyright (c) 2012 The Chromium 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.
5 #include "base/strings/string_number_conversions.h"
7 #include <ctype.h>
8 #include <errno.h>
9 #include <stdlib.h>
10 #include <wctype.h>
12 #include <limits>
14 #include "base/logging.h"
15 #include "base/scoped_clear_errno.h"
16 #include "base/strings/utf_string_conversions.h"
17 #include "base/third_party/dmg_fp/dmg_fp.h"
19 namespace base {
21 namespace {
23 template <typename STR, typename INT, typename UINT, bool NEG>
24 struct IntToStringT {
25 // This is to avoid a compiler warning about unary minus on unsigned type.
26 // For example, say you had the following code:
27 // template <typename INT>
28 // INT abs(INT value) { return value < 0 ? -value : value; }
29 // Even though if INT is unsigned, it's impossible for value < 0, so the
30 // unary minus will never be taken, the compiler will still generate a
31 // warning. We do a little specialization dance...
32 template <typename INT2, typename UINT2, bool NEG2>
33 struct ToUnsignedT {};
35 template <typename INT2, typename UINT2>
36 struct ToUnsignedT<INT2, UINT2, false> {
37 static UINT2 ToUnsigned(INT2 value) {
38 return static_cast<UINT2>(value);
42 template <typename INT2, typename UINT2>
43 struct ToUnsignedT<INT2, UINT2, true> {
44 static UINT2 ToUnsigned(INT2 value) {
45 if (value >= 0) {
46 return value;
47 } else {
48 // Avoid integer overflow when negating INT_MIN.
49 return static_cast<UINT2>(-(value + 1)) + 1;
54 // This set of templates is very similar to the above templates, but
55 // for testing whether an integer is negative.
56 template <typename INT2, bool NEG2>
57 struct TestNegT {};
58 template <typename INT2>
59 struct TestNegT<INT2, false> {
60 static bool TestNeg(INT2 value) {
61 // value is unsigned, and can never be negative.
62 return false;
65 template <typename INT2>
66 struct TestNegT<INT2, true> {
67 static bool TestNeg(INT2 value) {
68 return value < 0;
72 static STR IntToString(INT value) {
73 // log10(2) ~= 0.3 bytes needed per bit or per byte log10(2**8) ~= 2.4.
74 // So round up to allocate 3 output characters per byte, plus 1 for '-'.
75 const int kOutputBufSize = 3 * sizeof(INT) + 1;
77 // Allocate the whole string right away, we will right back to front, and
78 // then return the substr of what we ended up using.
79 STR outbuf(kOutputBufSize, 0);
81 bool is_neg = TestNegT<INT, NEG>::TestNeg(value);
82 // Even though is_neg will never be true when INT is parameterized as
83 // unsigned, even the presence of the unary operation causes a warning.
84 UINT res = ToUnsignedT<INT, UINT, NEG>::ToUnsigned(value);
86 typename STR::iterator it(outbuf.end());
87 do {
88 --it;
89 DCHECK(it != outbuf.begin());
90 *it = static_cast<typename STR::value_type>((res % 10) + '0');
91 res /= 10;
92 } while (res != 0);
93 if (is_neg) {
94 --it;
95 DCHECK(it != outbuf.begin());
96 *it = static_cast<typename STR::value_type>('-');
98 return STR(it, outbuf.end());
102 // Utility to convert a character to a digit in a given base
103 template<typename CHAR, int BASE, bool BASE_LTE_10> class BaseCharToDigit {
106 // Faster specialization for bases <= 10
107 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, true> {
108 public:
109 static bool Convert(CHAR c, uint8* digit) {
110 if (c >= '0' && c < '0' + BASE) {
111 *digit = static_cast<uint8>(c - '0');
112 return true;
114 return false;
118 // Specialization for bases where 10 < base <= 36
119 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, false> {
120 public:
121 static bool Convert(CHAR c, uint8* digit) {
122 if (c >= '0' && c <= '9') {
123 *digit = c - '0';
124 } else if (c >= 'a' && c < 'a' + BASE - 10) {
125 *digit = c - 'a' + 10;
126 } else if (c >= 'A' && c < 'A' + BASE - 10) {
127 *digit = c - 'A' + 10;
128 } else {
129 return false;
131 return true;
135 template<int BASE, typename CHAR> bool CharToDigit(CHAR c, uint8* digit) {
136 return BaseCharToDigit<CHAR, BASE, BASE <= 10>::Convert(c, digit);
139 // There is an IsUnicodeWhitespace for wchars defined in string_util.h, but it
140 // is locale independent, whereas the functions we are replacing were
141 // locale-dependent. TBD what is desired, but for the moment let's not
142 // introduce a change in behaviour.
143 template<typename CHAR> class WhitespaceHelper {
146 template<> class WhitespaceHelper<char> {
147 public:
148 static bool Invoke(char c) {
149 return 0 != isspace(static_cast<unsigned char>(c));
153 template<> class WhitespaceHelper<char16> {
154 public:
155 static bool Invoke(char16 c) {
156 return 0 != iswspace(c);
160 template<typename CHAR> bool LocalIsWhitespace(CHAR c) {
161 return WhitespaceHelper<CHAR>::Invoke(c);
164 // IteratorRangeToNumberTraits should provide:
165 // - a typedef for iterator_type, the iterator type used as input.
166 // - a typedef for value_type, the target numeric type.
167 // - static functions min, max (returning the minimum and maximum permitted
168 // values)
169 // - constant kBase, the base in which to interpret the input
170 template<typename IteratorRangeToNumberTraits>
171 class IteratorRangeToNumber {
172 public:
173 typedef IteratorRangeToNumberTraits traits;
174 typedef typename traits::iterator_type const_iterator;
175 typedef typename traits::value_type value_type;
177 // Generalized iterator-range-to-number conversion.
179 static bool Invoke(const_iterator begin,
180 const_iterator end,
181 value_type* output) {
182 bool valid = true;
184 while (begin != end && LocalIsWhitespace(*begin)) {
185 valid = false;
186 ++begin;
189 if (begin != end && *begin == '-') {
190 if (!std::numeric_limits<value_type>::is_signed) {
191 valid = false;
192 } else if (!Negative::Invoke(begin + 1, end, output)) {
193 valid = false;
195 } else {
196 if (begin != end && *begin == '+') {
197 ++begin;
199 if (!Positive::Invoke(begin, end, output)) {
200 valid = false;
204 return valid;
207 private:
208 // Sign provides:
209 // - a static function, CheckBounds, that determines whether the next digit
210 // causes an overflow/underflow
211 // - a static function, Increment, that appends the next digit appropriately
212 // according to the sign of the number being parsed.
213 template<typename Sign>
214 class Base {
215 public:
216 static bool Invoke(const_iterator begin, const_iterator end,
217 typename traits::value_type* output) {
218 *output = 0;
220 if (begin == end) {
221 return false;
224 // Note: no performance difference was found when using template
225 // specialization to remove this check in bases other than 16
226 if (traits::kBase == 16 && end - begin > 2 && *begin == '0' &&
227 (*(begin + 1) == 'x' || *(begin + 1) == 'X')) {
228 begin += 2;
231 for (const_iterator current = begin; current != end; ++current) {
232 uint8 new_digit = 0;
234 if (!CharToDigit<traits::kBase>(*current, &new_digit)) {
235 return false;
238 if (current != begin) {
239 if (!Sign::CheckBounds(output, new_digit)) {
240 return false;
242 *output *= traits::kBase;
245 Sign::Increment(new_digit, output);
247 return true;
251 class Positive : public Base<Positive> {
252 public:
253 static bool CheckBounds(value_type* output, uint8 new_digit) {
254 if (*output > static_cast<value_type>(traits::max() / traits::kBase) ||
255 (*output == static_cast<value_type>(traits::max() / traits::kBase) &&
256 new_digit > traits::max() % traits::kBase)) {
257 *output = traits::max();
258 return false;
260 return true;
262 static void Increment(uint8 increment, value_type* output) {
263 *output += increment;
267 class Negative : public Base<Negative> {
268 public:
269 static bool CheckBounds(value_type* output, uint8 new_digit) {
270 if (*output < traits::min() / traits::kBase ||
271 (*output == traits::min() / traits::kBase &&
272 new_digit > 0 - traits::min() % traits::kBase)) {
273 *output = traits::min();
274 return false;
276 return true;
278 static void Increment(uint8 increment, value_type* output) {
279 *output -= increment;
284 template<typename ITERATOR, typename VALUE, int BASE>
285 class BaseIteratorRangeToNumberTraits {
286 public:
287 typedef ITERATOR iterator_type;
288 typedef VALUE value_type;
289 static value_type min() {
290 return std::numeric_limits<value_type>::min();
292 static value_type max() {
293 return std::numeric_limits<value_type>::max();
295 static const int kBase = BASE;
298 template<typename ITERATOR>
299 class BaseHexIteratorRangeToIntTraits
300 : public BaseIteratorRangeToNumberTraits<ITERATOR, int, 16> {
303 template<typename ITERATOR>
304 class BaseHexIteratorRangeToUIntTraits
305 : public BaseIteratorRangeToNumberTraits<ITERATOR, uint32, 16> {
308 template<typename ITERATOR>
309 class BaseHexIteratorRangeToInt64Traits
310 : public BaseIteratorRangeToNumberTraits<ITERATOR, int64, 16> {
313 template<typename ITERATOR>
314 class BaseHexIteratorRangeToUInt64Traits
315 : public BaseIteratorRangeToNumberTraits<ITERATOR, uint64, 16> {
318 typedef BaseHexIteratorRangeToIntTraits<StringPiece::const_iterator>
319 HexIteratorRangeToIntTraits;
321 typedef BaseHexIteratorRangeToUIntTraits<StringPiece::const_iterator>
322 HexIteratorRangeToUIntTraits;
324 typedef BaseHexIteratorRangeToInt64Traits<StringPiece::const_iterator>
325 HexIteratorRangeToInt64Traits;
327 typedef BaseHexIteratorRangeToUInt64Traits<StringPiece::const_iterator>
328 HexIteratorRangeToUInt64Traits;
330 template<typename STR>
331 bool HexStringToBytesT(const STR& input, std::vector<uint8>* output) {
332 DCHECK_EQ(output->size(), 0u);
333 size_t count = input.size();
334 if (count == 0 || (count % 2) != 0)
335 return false;
336 for (uintptr_t i = 0; i < count / 2; ++i) {
337 uint8 msb = 0; // most significant 4 bits
338 uint8 lsb = 0; // least significant 4 bits
339 if (!CharToDigit<16>(input[i * 2], &msb) ||
340 !CharToDigit<16>(input[i * 2 + 1], &lsb))
341 return false;
342 output->push_back((msb << 4) | lsb);
344 return true;
347 template <typename VALUE, int BASE>
348 class StringPieceToNumberTraits
349 : public BaseIteratorRangeToNumberTraits<StringPiece::const_iterator,
350 VALUE,
351 BASE> {
354 template <typename VALUE>
355 bool StringToIntImpl(const StringPiece& input, VALUE* output) {
356 return IteratorRangeToNumber<StringPieceToNumberTraits<VALUE, 10> >::Invoke(
357 input.begin(), input.end(), output);
360 template <typename VALUE, int BASE>
361 class StringPiece16ToNumberTraits
362 : public BaseIteratorRangeToNumberTraits<StringPiece16::const_iterator,
363 VALUE,
364 BASE> {
367 template <typename VALUE>
368 bool String16ToIntImpl(const StringPiece16& input, VALUE* output) {
369 return IteratorRangeToNumber<StringPiece16ToNumberTraits<VALUE, 10> >::Invoke(
370 input.begin(), input.end(), output);
373 } // namespace
375 std::string IntToString(int value) {
376 return IntToStringT<std::string, int, unsigned int, true>::
377 IntToString(value);
380 string16 IntToString16(int value) {
381 return IntToStringT<string16, int, unsigned int, true>::
382 IntToString(value);
385 std::string UintToString(unsigned int value) {
386 return IntToStringT<std::string, unsigned int, unsigned int, false>::
387 IntToString(value);
390 string16 UintToString16(unsigned int value) {
391 return IntToStringT<string16, unsigned int, unsigned int, false>::
392 IntToString(value);
395 std::string Int64ToString(int64 value) {
396 return IntToStringT<std::string, int64, uint64, true>::IntToString(value);
399 string16 Int64ToString16(int64 value) {
400 return IntToStringT<string16, int64, uint64, true>::IntToString(value);
403 std::string Uint64ToString(uint64 value) {
404 return IntToStringT<std::string, uint64, uint64, false>::IntToString(value);
407 string16 Uint64ToString16(uint64 value) {
408 return IntToStringT<string16, uint64, uint64, false>::IntToString(value);
411 std::string SizeTToString(size_t value) {
412 return IntToStringT<std::string, size_t, size_t, false>::IntToString(value);
415 string16 SizeTToString16(size_t value) {
416 return IntToStringT<string16, size_t, size_t, false>::IntToString(value);
419 std::string DoubleToString(double value) {
420 // According to g_fmt.cc, it is sufficient to declare a buffer of size 32.
421 char buffer[32];
422 dmg_fp::g_fmt(buffer, value);
423 return std::string(buffer);
426 bool StringToInt(const StringPiece& input, int* output) {
427 return StringToIntImpl(input, output);
430 bool StringToInt(const StringPiece16& input, int* output) {
431 return String16ToIntImpl(input, output);
434 bool StringToUint(const StringPiece& input, unsigned* output) {
435 return StringToIntImpl(input, output);
438 bool StringToUint(const StringPiece16& input, unsigned* output) {
439 return String16ToIntImpl(input, output);
442 bool StringToInt64(const StringPiece& input, int64* output) {
443 return StringToIntImpl(input, output);
446 bool StringToInt64(const StringPiece16& input, int64* output) {
447 return String16ToIntImpl(input, output);
450 bool StringToUint64(const StringPiece& input, uint64* output) {
451 return StringToIntImpl(input, output);
454 bool StringToUint64(const StringPiece16& input, uint64* output) {
455 return String16ToIntImpl(input, output);
458 bool StringToSizeT(const StringPiece& input, size_t* output) {
459 return StringToIntImpl(input, output);
462 bool StringToSizeT(const StringPiece16& input, size_t* output) {
463 return String16ToIntImpl(input, output);
466 bool StringToDouble(const std::string& input, double* output) {
467 // Thread-safe? It is on at least Mac, Linux, and Windows.
468 ScopedClearErrno clear_errno;
470 char* endptr = NULL;
471 *output = dmg_fp::strtod(input.c_str(), &endptr);
473 // Cases to return false:
474 // - If errno is ERANGE, there was an overflow or underflow.
475 // - If the input string is empty, there was nothing to parse.
476 // - If endptr does not point to the end of the string, there are either
477 // characters remaining in the string after a parsed number, or the string
478 // does not begin with a parseable number. endptr is compared to the
479 // expected end given the string's stated length to correctly catch cases
480 // where the string contains embedded NUL characters.
481 // - If the first character is a space, there was leading whitespace
482 return errno == 0 &&
483 !input.empty() &&
484 input.c_str() + input.length() == endptr &&
485 !isspace(input[0]);
488 // Note: if you need to add String16ToDouble, first ask yourself if it's
489 // really necessary. If it is, probably the best implementation here is to
490 // convert to 8-bit and then use the 8-bit version.
492 // Note: if you need to add an iterator range version of StringToDouble, first
493 // ask yourself if it's really necessary. If it is, probably the best
494 // implementation here is to instantiate a string and use the string version.
496 std::string HexEncode(const void* bytes, size_t size) {
497 static const char kHexChars[] = "0123456789ABCDEF";
499 // Each input byte creates two output hex characters.
500 std::string ret(size * 2, '\0');
502 for (size_t i = 0; i < size; ++i) {
503 char b = reinterpret_cast<const char*>(bytes)[i];
504 ret[(i * 2)] = kHexChars[(b >> 4) & 0xf];
505 ret[(i * 2) + 1] = kHexChars[b & 0xf];
507 return ret;
510 bool HexStringToInt(const StringPiece& input, int* output) {
511 return IteratorRangeToNumber<HexIteratorRangeToIntTraits>::Invoke(
512 input.begin(), input.end(), output);
515 bool HexStringToUInt(const StringPiece& input, uint32* output) {
516 return IteratorRangeToNumber<HexIteratorRangeToUIntTraits>::Invoke(
517 input.begin(), input.end(), output);
520 bool HexStringToInt64(const StringPiece& input, int64* output) {
521 return IteratorRangeToNumber<HexIteratorRangeToInt64Traits>::Invoke(
522 input.begin(), input.end(), output);
525 bool HexStringToUInt64(const StringPiece& input, uint64* output) {
526 return IteratorRangeToNumber<HexIteratorRangeToUInt64Traits>::Invoke(
527 input.begin(), input.end(), output);
530 bool HexStringToBytes(const std::string& input, std::vector<uint8>* output) {
531 return HexStringToBytesT(input, output);
534 } // namespace base