blob: 41319b35fd66dbaad3d709416a0c22b3f45b5beb [file] [log] [blame]
kalman@chromium.orgd2d0b6b2012-01-26 00:27:291// Copyright (c) 2012 The Chromium Authors. All rights reserved.
brettw@chromium.org528c56d2010-07-30 19:28:442// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.
4
5#include "base/string_number_conversions.h"
6
michaelbai@google.com3132e35c2011-07-07 20:46:507#include <ctype.h>
brettw@chromium.org528c56d2010-07-30 19:28:448#include <errno.h>
9#include <stdlib.h>
tzik@chromium.orge7972d12011-06-18 11:53:1410#include <wctype.h>
brettw@chromium.org528c56d2010-07-30 19:28:4411
erikwright@chromium.orgcfcf2da2010-10-21 14:05:3812#include <limits>
13
brettw@chromium.org528c56d2010-07-30 19:28:4414#include "base/logging.h"
15#include "base/third_party/dmg_fp/dmg_fp.h"
16#include "base/utf_string_conversions.h"
17
18namespace base {
19
20namespace {
21
22template <typename STR, typename INT, typename UINT, bool NEG>
23struct IntToStringT {
24 // This is to avoid a compiler warning about unary minus on unsigned type.
25 // For example, say you had the following code:
26 // template <typename INT>
27 // INT abs(INT value) { return value < 0 ? -value : value; }
28 // Even though if INT is unsigned, it's impossible for value < 0, so the
29 // unary minus will never be taken, the compiler will still generate a
30 // warning. We do a little specialization dance...
31 template <typename INT2, typename UINT2, bool NEG2>
32 struct ToUnsignedT {};
33
34 template <typename INT2, typename UINT2>
35 struct ToUnsignedT<INT2, UINT2, false> {
36 static UINT2 ToUnsigned(INT2 value) {
37 return static_cast<UINT2>(value);
38 }
39 };
40
41 template <typename INT2, typename UINT2>
42 struct ToUnsignedT<INT2, UINT2, true> {
43 static UINT2 ToUnsigned(INT2 value) {
44 return static_cast<UINT2>(value < 0 ? -value : value);
45 }
46 };
47
48 // This set of templates is very similar to the above templates, but
49 // for testing whether an integer is negative.
50 template <typename INT2, bool NEG2>
51 struct TestNegT {};
52 template <typename INT2>
53 struct TestNegT<INT2, false> {
54 static bool TestNeg(INT2 value) {
55 // value is unsigned, and can never be negative.
56 return false;
57 }
58 };
59 template <typename INT2>
60 struct TestNegT<INT2, true> {
61 static bool TestNeg(INT2 value) {
62 return value < 0;
63 }
64 };
65
66 static STR IntToString(INT value) {
67 // log10(2) ~= 0.3 bytes needed per bit or per byte log10(2**8) ~= 2.4.
68 // So round up to allocate 3 output characters per byte, plus 1 for '-'.
69 const int kOutputBufSize = 3 * sizeof(INT) + 1;
70
71 // Allocate the whole string right away, we will right back to front, and
72 // then return the substr of what we ended up using.
73 STR outbuf(kOutputBufSize, 0);
74
75 bool is_neg = TestNegT<INT, NEG>::TestNeg(value);
76 // Even though is_neg will never be true when INT is parameterized as
77 // unsigned, even the presence of the unary operation causes a warning.
78 UINT res = ToUnsignedT<INT, UINT, NEG>::ToUnsigned(value);
79
80 for (typename STR::iterator it = outbuf.end();;) {
81 --it;
82 DCHECK(it != outbuf.begin());
83 *it = static_cast<typename STR::value_type>((res % 10) + '0');
84 res /= 10;
85
86 // We're done..
87 if (res == 0) {
88 if (is_neg) {
89 --it;
90 DCHECK(it != outbuf.begin());
91 *it = static_cast<typename STR::value_type>('-');
92 }
93 return STR(it, outbuf.end());
94 }
95 }
96 NOTREACHED();
97 return STR();
98 }
99};
100
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38101// Utility to convert a character to a digit in a given base
102template<typename CHAR, int BASE, bool BASE_LTE_10> class BaseCharToDigit {
brettw@chromium.org528c56d2010-07-30 19:28:44103};
104
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38105// Faster specialization for bases <= 10
106template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, true> {
brettw@chromium.org528c56d2010-07-30 19:28:44107 public:
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38108 static bool Convert(CHAR c, uint8* digit) {
109 if (c >= '0' && c < '0' + BASE) {
110 *digit = c - '0';
111 return true;
brettw@chromium.org528c56d2010-07-30 19:28:44112 }
brettw@chromium.org528c56d2010-07-30 19:28:44113 return false;
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38114 }
115};
116
117// Specialization for bases where 10 < base <= 36
118template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, false> {
119 public:
120 static bool Convert(CHAR c, uint8* digit) {
121 if (c >= '0' && c <= '9') {
122 *digit = c - '0';
123 } else if (c >= 'a' && c < 'a' + BASE - 10) {
124 *digit = c - 'a' + 10;
125 } else if (c >= 'A' && c < 'A' + BASE - 10) {
126 *digit = c - 'A' + 10;
127 } else {
128 return false;
129 }
130 return true;
131 }
132};
133
134template<int BASE, typename CHAR> bool CharToDigit(CHAR c, uint8* digit) {
135 return BaseCharToDigit<CHAR, BASE, BASE <= 10>::Convert(c, digit);
brettw@chromium.org528c56d2010-07-30 19:28:44136}
137
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38138// There is an IsWhitespace for wchars defined in string_util.h, but it is
139// locale independent, whereas the functions we are replacing were
140// locale-dependent. TBD what is desired, but for the moment let's not introduce
141// a change in behaviour.
142template<typename CHAR> class WhitespaceHelper {
143};
144
145template<> class WhitespaceHelper<char> {
146 public:
147 static bool Invoke(char c) {
erikwright@chromium.orgeae2a822010-11-02 19:30:02148 return 0 != isspace(static_cast<unsigned char>(c));
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38149 }
150};
151
152template<> class WhitespaceHelper<char16> {
153 public:
154 static bool Invoke(char16 c) {
155 return 0 != iswspace(c);
156 }
157};
158
159template<typename CHAR> bool LocalIsWhitespace(CHAR c) {
160 return WhitespaceHelper<CHAR>::Invoke(c);
161}
162
163// IteratorRangeToNumberTraits should provide:
164// - a typedef for iterator_type, the iterator type used as input.
165// - a typedef for value_type, the target numeric type.
166// - static functions min, max (returning the minimum and maximum permitted
167// values)
168// - constant kBase, the base in which to interpret the input
169template<typename IteratorRangeToNumberTraits>
170class IteratorRangeToNumber {
171 public:
172 typedef IteratorRangeToNumberTraits traits;
173 typedef typename traits::iterator_type const_iterator;
174 typedef typename traits::value_type value_type;
175
176 // Generalized iterator-range-to-number conversion.
177 //
178 static bool Invoke(const_iterator begin,
179 const_iterator end,
180 value_type* output) {
181 bool valid = true;
182
183 while (begin != end && LocalIsWhitespace(*begin)) {
184 valid = false;
185 ++begin;
186 }
187
188 if (begin != end && *begin == '-') {
189 if (!Negative::Invoke(begin + 1, end, output)) {
190 valid = false;
191 }
192 } else {
193 if (begin != end && *begin == '+') {
194 ++begin;
195 }
196 if (!Positive::Invoke(begin, end, output)) {
197 valid = false;
198 }
199 }
200
201 return valid;
202 }
203
204 private:
205 // Sign provides:
206 // - a static function, CheckBounds, that determines whether the next digit
207 // causes an overflow/underflow
208 // - a static function, Increment, that appends the next digit appropriately
209 // according to the sign of the number being parsed.
210 template<typename Sign>
211 class Base {
212 public:
213 static bool Invoke(const_iterator begin, const_iterator end,
214 typename traits::value_type* output) {
215 *output = 0;
216
217 if (begin == end) {
218 return false;
219 }
220
221 // Note: no performance difference was found when using template
222 // specialization to remove this check in bases other than 16
ali.akbar@gmail.com331580b2011-08-12 20:13:20223 if (traits::kBase == 16 && end - begin > 2 && *begin == '0' &&
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38224 (*(begin + 1) == 'x' || *(begin + 1) == 'X')) {
225 begin += 2;
226 }
227
228 for (const_iterator current = begin; current != end; ++current) {
229 uint8 new_digit = 0;
230
231 if (!CharToDigit<traits::kBase>(*current, &new_digit)) {
232 return false;
233 }
234
235 if (current != begin) {
236 if (!Sign::CheckBounds(output, new_digit)) {
237 return false;
238 }
239 *output *= traits::kBase;
240 }
241
242 Sign::Increment(new_digit, output);
243 }
244 return true;
245 }
246 };
247
248 class Positive : public Base<Positive> {
249 public:
250 static bool CheckBounds(value_type* output, uint8 new_digit) {
251 if (*output > static_cast<value_type>(traits::max() / traits::kBase) ||
252 (*output == static_cast<value_type>(traits::max() / traits::kBase) &&
253 new_digit > traits::max() % traits::kBase)) {
254 *output = traits::max();
255 return false;
256 }
257 return true;
258 }
259 static void Increment(uint8 increment, value_type* output) {
260 *output += increment;
261 }
262 };
263
264 class Negative : public Base<Negative> {
265 public:
266 static bool CheckBounds(value_type* output, uint8 new_digit) {
267 if (*output < traits::min() / traits::kBase ||
268 (*output == traits::min() / traits::kBase &&
269 new_digit > 0 - traits::min() % traits::kBase)) {
270 *output = traits::min();
271 return false;
272 }
273 return true;
274 }
275 static void Increment(uint8 increment, value_type* output) {
276 *output -= increment;
277 }
278 };
279};
280
281template<typename ITERATOR, typename VALUE, int BASE>
282class BaseIteratorRangeToNumberTraits {
283 public:
284 typedef ITERATOR iterator_type;
285 typedef VALUE value_type;
286 static value_type min() {
287 return std::numeric_limits<value_type>::min();
288 }
289 static value_type max() {
290 return std::numeric_limits<value_type>::max();
291 }
292 static const int kBase = BASE;
293};
294
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38295template<typename ITERATOR>
296class BaseHexIteratorRangeToIntTraits
297 : public BaseIteratorRangeToNumberTraits<ITERATOR, int, 16> {
298 public:
299 // Allow parsing of 0xFFFFFFFF, which is technically an overflow
300 static unsigned int max() {
301 return std::numeric_limits<unsigned int>::max();
302 }
303};
304
tedvessenes@gmail.comeb72b272011-12-19 16:10:55305typedef BaseHexIteratorRangeToIntTraits<StringPiece::const_iterator>
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38306 HexIteratorRangeToIntTraits;
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38307
brettw@chromium.org528c56d2010-07-30 19:28:44308template<typename STR>
309bool HexStringToBytesT(const STR& input, std::vector<uint8>* output) {
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38310 DCHECK_EQ(output->size(), 0u);
brettw@chromium.org528c56d2010-07-30 19:28:44311 size_t count = input.size();
312 if (count == 0 || (count % 2) != 0)
313 return false;
314 for (uintptr_t i = 0; i < count / 2; ++i) {
315 uint8 msb = 0; // most significant 4 bits
316 uint8 lsb = 0; // least significant 4 bits
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38317 if (!CharToDigit<16>(input[i * 2], &msb) ||
318 !CharToDigit<16>(input[i * 2 + 1], &lsb))
brettw@chromium.org528c56d2010-07-30 19:28:44319 return false;
320 output->push_back((msb << 4) | lsb);
321 }
322 return true;
323}
324
tedvessenes@gmail.comeb72b272011-12-19 16:10:55325template <typename VALUE, int BASE>
326class StringPieceToNumberTraits
327 : public BaseIteratorRangeToNumberTraits<StringPiece::const_iterator,
328 VALUE,
329 BASE> {};
330
331template <typename VALUE>
332bool StringToIntImpl(const StringPiece& input, VALUE* output) {
333 return IteratorRangeToNumber<StringPieceToNumberTraits<VALUE, 10> >::Invoke(
334 input.begin(), input.end(), output);
335}
336
337template <typename VALUE, int BASE>
338class StringPiece16ToNumberTraits
339 : public BaseIteratorRangeToNumberTraits<StringPiece16::const_iterator,
340 VALUE,
341 BASE> {};
342
343template <typename VALUE>
344bool String16ToIntImpl(const StringPiece16& input, VALUE* output) {
345 return IteratorRangeToNumber<StringPiece16ToNumberTraits<VALUE, 10> >::Invoke(
346 input.begin(), input.end(), output);
347}
348
brettw@chromium.org528c56d2010-07-30 19:28:44349} // namespace
350
351std::string IntToString(int value) {
352 return IntToStringT<std::string, int, unsigned int, true>::
353 IntToString(value);
354}
355
356string16 IntToString16(int value) {
357 return IntToStringT<string16, int, unsigned int, true>::
358 IntToString(value);
359}
360
361std::string UintToString(unsigned int value) {
362 return IntToStringT<std::string, unsigned int, unsigned int, false>::
363 IntToString(value);
364}
365
366string16 UintToString16(unsigned int value) {
367 return IntToStringT<string16, unsigned int, unsigned int, false>::
368 IntToString(value);
369}
370
371std::string Int64ToString(int64 value) {
372 return IntToStringT<std::string, int64, uint64, true>::
373 IntToString(value);
374}
375
376string16 Int64ToString16(int64 value) {
377 return IntToStringT<string16, int64, uint64, true>::IntToString(value);
378}
379
380std::string Uint64ToString(uint64 value) {
381 return IntToStringT<std::string, uint64, uint64, false>::
382 IntToString(value);
383}
384
385string16 Uint64ToString16(uint64 value) {
386 return IntToStringT<string16, uint64, uint64, false>::
387 IntToString(value);
388}
389
390std::string DoubleToString(double value) {
391 // According to g_fmt.cc, it is sufficient to declare a buffer of size 32.
392 char buffer[32];
393 dmg_fp::g_fmt(buffer, value);
394 return std::string(buffer);
395}
396
tedvessenes@gmail.comeb72b272011-12-19 16:10:55397bool StringToInt(const StringPiece& input, int* output) {
398 return StringToIntImpl(input, output);
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38399}
400
tedvessenes@gmail.comeb72b272011-12-19 16:10:55401bool StringToInt(const StringPiece16& input, int* output) {
402 return String16ToIntImpl(input, output);
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38403}
404
yoz@chromium.orgd1bafad2012-01-28 01:02:17405bool StringToUint(const StringPiece& input, unsigned* output) {
406 return StringToIntImpl(input, output);
407}
408
409bool StringToUint(const StringPiece16& input, unsigned* output) {
410 return String16ToIntImpl(input, output);
411}
412
tedvessenes@gmail.comeb72b272011-12-19 16:10:55413bool StringToInt64(const StringPiece& input, int64* output) {
414 return StringToIntImpl(input, output);
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38415}
416
tedvessenes@gmail.comeb72b272011-12-19 16:10:55417bool StringToInt64(const StringPiece16& input, int64* output) {
418 return String16ToIntImpl(input, output);
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38419}
420
yoz@chromium.orgd1bafad2012-01-28 01:02:17421bool StringToUint64(const StringPiece& input, uint64* output) {
kalman@chromium.orgd2d0b6b2012-01-26 00:27:29422 return StringToIntImpl(input, output);
423}
424
yoz@chromium.orgd1bafad2012-01-28 01:02:17425bool StringToUint64(const StringPiece16& input, uint64* output) {
kalman@chromium.orgd2d0b6b2012-01-26 00:27:29426 return String16ToIntImpl(input, output);
427}
428
brettw@chromium.org528c56d2010-07-30 19:28:44429bool StringToDouble(const std::string& input, double* output) {
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38430 errno = 0; // Thread-safe? It is on at least Mac, Linux, and Windows.
431 char* endptr = NULL;
432 *output = dmg_fp::strtod(input.c_str(), &endptr);
433
434 // Cases to return false:
435 // - If errno is ERANGE, there was an overflow or underflow.
436 // - If the input string is empty, there was nothing to parse.
437 // - If endptr does not point to the end of the string, there are either
438 // characters remaining in the string after a parsed number, or the string
439 // does not begin with a parseable number. endptr is compared to the
440 // expected end given the string's stated length to correctly catch cases
441 // where the string contains embedded NUL characters.
442 // - If the first character is a space, there was leading whitespace
443 return errno == 0 &&
444 !input.empty() &&
445 input.c_str() + input.length() == endptr &&
446 !isspace(input[0]);
brettw@chromium.org528c56d2010-07-30 19:28:44447}
448
449// Note: if you need to add String16ToDouble, first ask yourself if it's
450// really necessary. If it is, probably the best implementation here is to
451// convert to 8-bit and then use the 8-bit version.
452
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38453// Note: if you need to add an iterator range version of StringToDouble, first
454// ask yourself if it's really necessary. If it is, probably the best
455// implementation here is to instantiate a string and use the string version.
456
brettw@chromium.org528c56d2010-07-30 19:28:44457std::string HexEncode(const void* bytes, size_t size) {
458 static const char kHexChars[] = "0123456789ABCDEF";
459
460 // Each input byte creates two output hex characters.
461 std::string ret(size * 2, '\0');
462
463 for (size_t i = 0; i < size; ++i) {
464 char b = reinterpret_cast<const char*>(bytes)[i];
465 ret[(i * 2)] = kHexChars[(b >> 4) & 0xf];
466 ret[(i * 2) + 1] = kHexChars[b & 0xf];
467 }
468 return ret;
469}
470
tedvessenes@gmail.comeb72b272011-12-19 16:10:55471bool HexStringToInt(const StringPiece& input, int* output) {
erikwright@chromium.orgcfcf2da2010-10-21 14:05:38472 return IteratorRangeToNumber<HexIteratorRangeToIntTraits>::Invoke(
473 input.begin(), input.end(), output);
474}
475
brettw@chromium.org528c56d2010-07-30 19:28:44476bool HexStringToBytes(const std::string& input, std::vector<uint8>* output) {
477 return HexStringToBytesT(input, output);
478}
479
480} // namespace base