Roll android_tools deps to the latest.
[chromium-blink-merge.git] / base / strings / string_number_conversions.cc
blob59dc93be8ade0ac54cd73f3e053c8ce2b3b77c1e
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 return static_cast<UINT2>(value < 0 ? -value : value);
49 // This set of templates is very similar to the above templates, but
50 // for testing whether an integer is negative.
51 template <typename INT2, bool NEG2>
52 struct TestNegT {};
53 template <typename INT2>
54 struct TestNegT<INT2, false> {
55 static bool TestNeg(INT2 value) {
56 // value is unsigned, and can never be negative.
57 return false;
60 template <typename INT2>
61 struct TestNegT<INT2, true> {
62 static bool TestNeg(INT2 value) {
63 return value < 0;
67 static STR IntToString(INT value) {
68 // log10(2) ~= 0.3 bytes needed per bit or per byte log10(2**8) ~= 2.4.
69 // So round up to allocate 3 output characters per byte, plus 1 for '-'.
70 const int kOutputBufSize = 3 * sizeof(INT) + 1;
72 // Allocate the whole string right away, we will right back to front, and
73 // then return the substr of what we ended up using.
74 STR outbuf(kOutputBufSize, 0);
76 bool is_neg = TestNegT<INT, NEG>::TestNeg(value);
77 // Even though is_neg will never be true when INT is parameterized as
78 // unsigned, even the presence of the unary operation causes a warning.
79 UINT res = ToUnsignedT<INT, UINT, NEG>::ToUnsigned(value);
81 typename STR::iterator it(outbuf.end());
82 do {
83 --it;
84 DCHECK(it != outbuf.begin());
85 *it = static_cast<typename STR::value_type>((res % 10) + '0');
86 res /= 10;
87 } while (res != 0);
88 if (is_neg) {
89 --it;
90 DCHECK(it != outbuf.begin());
91 *it = static_cast<typename STR::value_type>('-');
93 return STR(it, outbuf.end());
97 // Utility to convert a character to a digit in a given base
98 template<typename CHAR, int BASE, bool BASE_LTE_10> class BaseCharToDigit {
101 // Faster specialization for bases <= 10
102 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, true> {
103 public:
104 static bool Convert(CHAR c, uint8* digit) {
105 if (c >= '0' && c < '0' + BASE) {
106 *digit = c - '0';
107 return true;
109 return false;
113 // Specialization for bases where 10 < base <= 36
114 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, false> {
115 public:
116 static bool Convert(CHAR c, uint8* digit) {
117 if (c >= '0' && c <= '9') {
118 *digit = c - '0';
119 } else if (c >= 'a' && c < 'a' + BASE - 10) {
120 *digit = c - 'a' + 10;
121 } else if (c >= 'A' && c < 'A' + BASE - 10) {
122 *digit = c - 'A' + 10;
123 } else {
124 return false;
126 return true;
130 template<int BASE, typename CHAR> bool CharToDigit(CHAR c, uint8* digit) {
131 return BaseCharToDigit<CHAR, BASE, BASE <= 10>::Convert(c, digit);
134 // There is an IsWhitespace for wchars defined in string_util.h, but it is
135 // locale independent, whereas the functions we are replacing were
136 // locale-dependent. TBD what is desired, but for the moment let's not introduce
137 // a change in behaviour.
138 template<typename CHAR> class WhitespaceHelper {
141 template<> class WhitespaceHelper<char> {
142 public:
143 static bool Invoke(char c) {
144 return 0 != isspace(static_cast<unsigned char>(c));
148 template<> class WhitespaceHelper<char16> {
149 public:
150 static bool Invoke(char16 c) {
151 return 0 != iswspace(c);
155 template<typename CHAR> bool LocalIsWhitespace(CHAR c) {
156 return WhitespaceHelper<CHAR>::Invoke(c);
159 // IteratorRangeToNumberTraits should provide:
160 // - a typedef for iterator_type, the iterator type used as input.
161 // - a typedef for value_type, the target numeric type.
162 // - static functions min, max (returning the minimum and maximum permitted
163 // values)
164 // - constant kBase, the base in which to interpret the input
165 template<typename IteratorRangeToNumberTraits>
166 class IteratorRangeToNumber {
167 public:
168 typedef IteratorRangeToNumberTraits traits;
169 typedef typename traits::iterator_type const_iterator;
170 typedef typename traits::value_type value_type;
172 // Generalized iterator-range-to-number conversion.
174 static bool Invoke(const_iterator begin,
175 const_iterator end,
176 value_type* output) {
177 bool valid = true;
179 while (begin != end && LocalIsWhitespace(*begin)) {
180 valid = false;
181 ++begin;
184 if (begin != end && *begin == '-') {
185 if (!std::numeric_limits<value_type>::is_signed) {
186 valid = false;
187 } else if (!Negative::Invoke(begin + 1, end, output)) {
188 valid = false;
190 } else {
191 if (begin != end && *begin == '+') {
192 ++begin;
194 if (!Positive::Invoke(begin, end, output)) {
195 valid = false;
199 return valid;
202 private:
203 // Sign provides:
204 // - a static function, CheckBounds, that determines whether the next digit
205 // causes an overflow/underflow
206 // - a static function, Increment, that appends the next digit appropriately
207 // according to the sign of the number being parsed.
208 template<typename Sign>
209 class Base {
210 public:
211 static bool Invoke(const_iterator begin, const_iterator end,
212 typename traits::value_type* output) {
213 *output = 0;
215 if (begin == end) {
216 return false;
219 // Note: no performance difference was found when using template
220 // specialization to remove this check in bases other than 16
221 if (traits::kBase == 16 && end - begin > 2 && *begin == '0' &&
222 (*(begin + 1) == 'x' || *(begin + 1) == 'X')) {
223 begin += 2;
226 for (const_iterator current = begin; current != end; ++current) {
227 uint8 new_digit = 0;
229 if (!CharToDigit<traits::kBase>(*current, &new_digit)) {
230 return false;
233 if (current != begin) {
234 if (!Sign::CheckBounds(output, new_digit)) {
235 return false;
237 *output *= traits::kBase;
240 Sign::Increment(new_digit, output);
242 return true;
246 class Positive : public Base<Positive> {
247 public:
248 static bool CheckBounds(value_type* output, uint8 new_digit) {
249 if (*output > static_cast<value_type>(traits::max() / traits::kBase) ||
250 (*output == static_cast<value_type>(traits::max() / traits::kBase) &&
251 new_digit > traits::max() % traits::kBase)) {
252 *output = traits::max();
253 return false;
255 return true;
257 static void Increment(uint8 increment, value_type* output) {
258 *output += increment;
262 class Negative : public Base<Negative> {
263 public:
264 static bool CheckBounds(value_type* output, uint8 new_digit) {
265 if (*output < traits::min() / traits::kBase ||
266 (*output == traits::min() / traits::kBase &&
267 new_digit > 0 - traits::min() % traits::kBase)) {
268 *output = traits::min();
269 return false;
271 return true;
273 static void Increment(uint8 increment, value_type* output) {
274 *output -= increment;
279 template<typename ITERATOR, typename VALUE, int BASE>
280 class BaseIteratorRangeToNumberTraits {
281 public:
282 typedef ITERATOR iterator_type;
283 typedef VALUE value_type;
284 static value_type min() {
285 return std::numeric_limits<value_type>::min();
287 static value_type max() {
288 return std::numeric_limits<value_type>::max();
290 static const int kBase = BASE;
293 template<typename ITERATOR>
294 class BaseHexIteratorRangeToIntTraits
295 : public BaseIteratorRangeToNumberTraits<ITERATOR, int, 16> {
298 template<typename ITERATOR>
299 class BaseHexIteratorRangeToUIntTraits
300 : public BaseIteratorRangeToNumberTraits<ITERATOR, uint32, 16> {
303 template<typename ITERATOR>
304 class BaseHexIteratorRangeToInt64Traits
305 : public BaseIteratorRangeToNumberTraits<ITERATOR, int64, 16> {
308 template<typename ITERATOR>
309 class BaseHexIteratorRangeToUInt64Traits
310 : public BaseIteratorRangeToNumberTraits<ITERATOR, uint64, 16> {
313 typedef BaseHexIteratorRangeToIntTraits<StringPiece::const_iterator>
314 HexIteratorRangeToIntTraits;
316 typedef BaseHexIteratorRangeToUIntTraits<StringPiece::const_iterator>
317 HexIteratorRangeToUIntTraits;
319 typedef BaseHexIteratorRangeToInt64Traits<StringPiece::const_iterator>
320 HexIteratorRangeToInt64Traits;
322 typedef BaseHexIteratorRangeToUInt64Traits<StringPiece::const_iterator>
323 HexIteratorRangeToUInt64Traits;
325 template<typename STR>
326 bool HexStringToBytesT(const STR& input, std::vector<uint8>* output) {
327 DCHECK_EQ(output->size(), 0u);
328 size_t count = input.size();
329 if (count == 0 || (count % 2) != 0)
330 return false;
331 for (uintptr_t i = 0; i < count / 2; ++i) {
332 uint8 msb = 0; // most significant 4 bits
333 uint8 lsb = 0; // least significant 4 bits
334 if (!CharToDigit<16>(input[i * 2], &msb) ||
335 !CharToDigit<16>(input[i * 2 + 1], &lsb))
336 return false;
337 output->push_back((msb << 4) | lsb);
339 return true;
342 template <typename VALUE, int BASE>
343 class StringPieceToNumberTraits
344 : public BaseIteratorRangeToNumberTraits<StringPiece::const_iterator,
345 VALUE,
346 BASE> {
349 template <typename VALUE>
350 bool StringToIntImpl(const StringPiece& input, VALUE* output) {
351 return IteratorRangeToNumber<StringPieceToNumberTraits<VALUE, 10> >::Invoke(
352 input.begin(), input.end(), output);
355 template <typename VALUE, int BASE>
356 class StringPiece16ToNumberTraits
357 : public BaseIteratorRangeToNumberTraits<StringPiece16::const_iterator,
358 VALUE,
359 BASE> {
362 template <typename VALUE>
363 bool String16ToIntImpl(const StringPiece16& input, VALUE* output) {
364 return IteratorRangeToNumber<StringPiece16ToNumberTraits<VALUE, 10> >::Invoke(
365 input.begin(), input.end(), output);
368 } // namespace
370 std::string IntToString(int value) {
371 return IntToStringT<std::string, int, unsigned int, true>::
372 IntToString(value);
375 string16 IntToString16(int value) {
376 return IntToStringT<string16, int, unsigned int, true>::
377 IntToString(value);
380 std::string UintToString(unsigned int value) {
381 return IntToStringT<std::string, unsigned int, unsigned int, false>::
382 IntToString(value);
385 string16 UintToString16(unsigned int value) {
386 return IntToStringT<string16, unsigned int, unsigned int, false>::
387 IntToString(value);
390 std::string Int64ToString(int64 value) {
391 return IntToStringT<std::string, int64, uint64, true>::
392 IntToString(value);
395 string16 Int64ToString16(int64 value) {
396 return IntToStringT<string16, int64, uint64, true>::IntToString(value);
399 std::string Uint64ToString(uint64 value) {
400 return IntToStringT<std::string, uint64, uint64, false>::
401 IntToString(value);
404 string16 Uint64ToString16(uint64 value) {
405 return IntToStringT<string16, uint64, uint64, false>::
406 IntToString(value);
409 std::string DoubleToString(double value) {
410 // According to g_fmt.cc, it is sufficient to declare a buffer of size 32.
411 char buffer[32];
412 dmg_fp::g_fmt(buffer, value);
413 return std::string(buffer);
416 bool StringToInt(const StringPiece& input, int* output) {
417 return StringToIntImpl(input, output);
420 bool StringToInt(const StringPiece16& input, int* output) {
421 return String16ToIntImpl(input, output);
424 bool StringToUint(const StringPiece& input, unsigned* output) {
425 return StringToIntImpl(input, output);
428 bool StringToUint(const StringPiece16& input, unsigned* output) {
429 return String16ToIntImpl(input, output);
432 bool StringToInt64(const StringPiece& input, int64* output) {
433 return StringToIntImpl(input, output);
436 bool StringToInt64(const StringPiece16& input, int64* output) {
437 return String16ToIntImpl(input, output);
440 bool StringToUint64(const StringPiece& input, uint64* output) {
441 return StringToIntImpl(input, output);
444 bool StringToUint64(const StringPiece16& input, uint64* output) {
445 return String16ToIntImpl(input, output);
448 bool StringToSizeT(const StringPiece& input, size_t* output) {
449 return StringToIntImpl(input, output);
452 bool StringToSizeT(const StringPiece16& input, size_t* output) {
453 return String16ToIntImpl(input, output);
456 bool StringToDouble(const std::string& input, double* output) {
457 // Thread-safe? It is on at least Mac, Linux, and Windows.
458 ScopedClearErrno clear_errno;
460 char* endptr = NULL;
461 *output = dmg_fp::strtod(input.c_str(), &endptr);
463 // Cases to return false:
464 // - If errno is ERANGE, there was an overflow or underflow.
465 // - If the input string is empty, there was nothing to parse.
466 // - If endptr does not point to the end of the string, there are either
467 // characters remaining in the string after a parsed number, or the string
468 // does not begin with a parseable number. endptr is compared to the
469 // expected end given the string's stated length to correctly catch cases
470 // where the string contains embedded NUL characters.
471 // - If the first character is a space, there was leading whitespace
472 return errno == 0 &&
473 !input.empty() &&
474 input.c_str() + input.length() == endptr &&
475 !isspace(input[0]);
478 // Note: if you need to add String16ToDouble, first ask yourself if it's
479 // really necessary. If it is, probably the best implementation here is to
480 // convert to 8-bit and then use the 8-bit version.
482 // Note: if you need to add an iterator range version of StringToDouble, first
483 // ask yourself if it's really necessary. If it is, probably the best
484 // implementation here is to instantiate a string and use the string version.
486 std::string HexEncode(const void* bytes, size_t size) {
487 static const char kHexChars[] = "0123456789ABCDEF";
489 // Each input byte creates two output hex characters.
490 std::string ret(size * 2, '\0');
492 for (size_t i = 0; i < size; ++i) {
493 char b = reinterpret_cast<const char*>(bytes)[i];
494 ret[(i * 2)] = kHexChars[(b >> 4) & 0xf];
495 ret[(i * 2) + 1] = kHexChars[b & 0xf];
497 return ret;
500 bool HexStringToInt(const StringPiece& input, int* output) {
501 return IteratorRangeToNumber<HexIteratorRangeToIntTraits>::Invoke(
502 input.begin(), input.end(), output);
505 bool HexStringToUInt(const StringPiece& input, uint32* output) {
506 return IteratorRangeToNumber<HexIteratorRangeToUIntTraits>::Invoke(
507 input.begin(), input.end(), output);
510 bool HexStringToInt64(const StringPiece& input, int64* output) {
511 return IteratorRangeToNumber<HexIteratorRangeToInt64Traits>::Invoke(
512 input.begin(), input.end(), output);
515 bool HexStringToUInt64(const StringPiece& input, uint64* output) {
516 return IteratorRangeToNumber<HexIteratorRangeToUInt64Traits>::Invoke(
517 input.begin(), input.end(), output);
520 bool HexStringToBytes(const std::string& input, std::vector<uint8>* output) {
521 return HexStringToBytesT(input, output);
524 } // namespace base