icuSources/i18n/double-conversion-utils.h

   1 // © 2018 and later: Unicode, Inc. and others.
   2 // License & terms of use: http://www.unicode.org/copyright.html
   3 //
   4 // From the double-conversion library. Original license:
   5 //
   6 // Copyright 2010 the V8 project authors. All rights reserved.
   7 // Redistribution and use in source and binary forms, with or without
   8 // modification, are permitted provided that the following conditions are
   9 // met:
  10 //
  11 //     * Redistributions of source code must retain the above copyright
  12 //       notice, this list of conditions and the following disclaimer.
  13 //     * Redistributions in binary form must reproduce the above
  14 //       copyright notice, this list of conditions and the following
  15 //       disclaimer in the documentation and/or other materials provided
  16 //       with the distribution.
  17 //     * Neither the name of Google Inc. nor the names of its
  18 //       contributors may be used to endorse or promote products derived
  19 //       from this software without specific prior written permission.
  20 //
  21 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  24 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  25 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  26 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  27 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  28 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  29 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  30 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  31 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32
  33 // ICU PATCH: ifdef around UCONFIG_NO_FORMATTING
  34 #include "unicode/utypes.h"
  35 #if !UCONFIG_NO_FORMATTING
  36
  37 #ifndef DOUBLE_CONVERSION_UTILS_H_
  38 #define DOUBLE_CONVERSION_UTILS_H_
  39
  40 #include <stdlib.h>
  41 #include <string.h>
  42
  43 // ICU PATCH: Use U_ASSERT instead of <assert.h>
  44 #include "uassert.h"
  45 #define ASSERT U_ASSERT
  46
  47 #ifndef UNIMPLEMENTED
  48 #define UNIMPLEMENTED() (abort())
  49 #endif
  50 #ifndef DOUBLE_CONVERSION_NO_RETURN
  51 #ifdef _MSC_VER
  52 #define DOUBLE_CONVERSION_NO_RETURN __declspec(noreturn)
  53 #else
  54 #define DOUBLE_CONVERSION_NO_RETURN __attribute__((noreturn))
  55 #endif
  56 #endif
  57 #ifndef UNREACHABLE
  58 #ifdef _MSC_VER
  59 void DOUBLE_CONVERSION_NO_RETURN abort_noreturn();
  60 inline void abort_noreturn() { abort(); }
  61 #define UNREACHABLE()   (abort_noreturn())
  62 #else
  63 #define UNREACHABLE()   (abort())
  64 #endif
  65 #endif
  66
  67
  68 // Double operations detection based on target architecture.
  69 // Linux uses a 80bit wide floating point stack on x86. This induces double
  70 // rounding, which in turn leads to wrong results.
  71 // An easy way to test if the floating-point operations are correct is to
  72 // evaluate: 89255.0/1e22. If the floating-point stack is 64 bits wide then
  73 // the result is equal to 89255e-22.
  74 // The best way to test this, is to create a division-function and to compare
  75 // the output of the division with the expected result. (Inlining must be
  76 // disabled.)
  77 // On Linux,x86 89255e-22 != Div_double(89255.0/1e22)
  78 // ICU PATCH: Enable ARM32 & ARM64 builds for Windows with 'defined(_M_ARM) || defined(_M_ARM64)'.
  79 #if defined(_M_X64) || defined(__x86_64__) || \
  80     defined(__ARMEL__) || defined(__avr32__) || defined(_M_ARM) || defined(_M_ARM64) || \
  81     defined(__hppa__) || defined(__ia64__) || \
  82     defined(__mips__) || \
  83     defined(__powerpc__) || defined(__ppc__) || defined(__ppc64__) || \
  84     defined(_POWER) || defined(_ARCH_PPC) || defined(_ARCH_PPC64) || \
  85     defined(__sparc__) || defined(__sparc) || defined(__s390__) || \
  86     defined(__SH4__) || defined(__alpha__) || \
  87     defined(_MIPS_ARCH_MIPS32R2) || \
  88     defined(__AARCH64EL__) || defined(__aarch64__) || \
  89     defined(__riscv)
  90 #define DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS 1
  91 #elif defined(__mc68000__)
  92 #undef DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS
  93 #elif defined(_M_IX86) || defined(__i386__) || defined(__i386)
  94 #if defined(_WIN32)
  95 // Windows uses a 64bit wide floating point stack.
  96 #define DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS 1
  97 #else
  98 #undef DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS
  99 #endif  // _WIN32
 100 #else
 101 #error Target architecture was not detected as supported by Double-Conversion.
 102 #endif
 103
 104 #if defined(__GNUC__)
 105 #define DOUBLE_CONVERSION_UNUSED __attribute__((unused))
 106 #else
 107 #define DOUBLE_CONVERSION_UNUSED
 108 #endif
 109
 110 #if defined(_WIN32) && !defined(__MINGW32__)
 111
 112 typedef signed char int8_t;
 113 typedef unsigned char uint8_t;
 114 typedef short int16_t;  // NOLINT
 115 typedef unsigned short uint16_t;  // NOLINT
 116 typedef int int32_t;
 117 typedef unsigned int uint32_t;
 118 typedef __int64 int64_t;
 119 typedef unsigned __int64 uint64_t;
 120 // intptr_t and friends are defined in crtdefs.h through stdio.h.
 121
 122 #else
 123
 124 #include <stdint.h>
 125
 126 #endif
 127
 128 typedef uint16_t uc16;
 129
 130 // The following macro works on both 32 and 64-bit platforms.
 131 // Usage: instead of writing 0x1234567890123456
 132 //      write UINT64_2PART_C(0x12345678,90123456);
 133 #define UINT64_2PART_C(a, b) (((static_cast<uint64_t>(a) << 32) + 0x##b##u))
 134
 135
 136 // The expression ARRAY_SIZE(a) is a compile-time constant of type
 137 // size_t which represents the number of elements of the given
 138 // array. You should only use ARRAY_SIZE on statically allocated
 139 // arrays.
 140 #ifndef ARRAY_SIZE
 141 #define ARRAY_SIZE(a)                                   \
 142   ((sizeof(a) / sizeof(*(a))) /                         \
 143   static_cast<size_t>(!(sizeof(a) % sizeof(*(a)))))
 144 #endif
 145
 146 // A macro to disallow the evil copy constructor and operator= functions
 147 // This should be used in the private: declarations for a class
 148 #ifndef DISALLOW_COPY_AND_ASSIGN
 149 #define DISALLOW_COPY_AND_ASSIGN(TypeName)      \
 150   TypeName(const TypeName&);                    \
 151   void operator=(const TypeName&)
 152 #endif
 153
 154 // A macro to disallow all the implicit constructors, namely the
 155 // default constructor, copy constructor and operator= functions.
 156 //
 157 // This should be used in the private: declarations for a class
 158 // that wants to prevent anyone from instantiating it. This is
 159 // especially useful for classes containing only static methods.
 160 #ifndef DISALLOW_IMPLICIT_CONSTRUCTORS
 161 #define DISALLOW_IMPLICIT_CONSTRUCTORS(TypeName) \
 162   TypeName();                                    \
 163   DISALLOW_COPY_AND_ASSIGN(TypeName)
 164 #endif
 165
 166 // ICU PATCH: Wrap in ICU namespace
 167 U_NAMESPACE_BEGIN
 168
 169 namespace double_conversion {
 170
 171 static const int kCharSize = sizeof(char);
 172
 173 // Returns the maximum of the two parameters.
 174 template <typename T>
 175 static T Max(T a, T b) {
 176   return a < b ? b : a;
 177 }
 178
 179
 180 // Returns the minimum of the two parameters.
 181 template <typename T>
 182 static T Min(T a, T b) {
 183   return a < b ? a : b;
 184 }
 185
 186
 187 inline int StrLength(const char* string) {
 188   size_t length = strlen(string);
 189   ASSERT(length == static_cast<size_t>(static_cast<int>(length)));
 190   return static_cast<int>(length);
 191 }
 192
 193 // This is a simplified version of V8's Vector class.
 194 template <typename T>
 195 class Vector {
 196  public:
 197   Vector() : start_(NULL), length_(0) {}
 198   Vector(T* data, int len) : start_(data), length_(len) {
 199     ASSERT(len == 0 || (len > 0 && data != NULL));
 200   }
 201
 202   // Returns a vector using the same backing storage as this one,
 203   // spanning from and including 'from', to but not including 'to'.
 204   Vector<T> SubVector(int from, int to) {
 205     ASSERT(to <= length_);
 206     ASSERT(from < to);
 207     ASSERT(0 <= from);
 208     return Vector<T>(start() + from, to - from);
 209   }
 210
 211   // Returns the length of the vector.
 212   int length() const { return length_; }
 213
 214   // Returns whether or not the vector is empty.
 215   bool is_empty() const { return length_ == 0; }
 216
 217   // Returns the pointer to the start of the data in the vector.
 218   T* start() const { return start_; }
 219
 220   // Access individual vector elements - checks bounds in debug mode.
 221   T& operator[](int index) const {
 222     ASSERT(0 <= index && index < length_);
 223     return start_[index];
 224   }
 225
 226   T& first() { return start_[0]; }
 227
 228   T& last() { return start_[length_ - 1]; }
 229
 230  private:
 231   T* start_;
 232   int length_;
 233 };
 234
 235
 236 // Helper class for building result strings in a character buffer. The
 237 // purpose of the class is to use safe operations that checks the
 238 // buffer bounds on all operations in debug mode.
 239 class StringBuilder {
 240  public:
 241   StringBuilder(char* buffer, int buffer_size)
 242       : buffer_(buffer, buffer_size), position_(0) { }
 243
 244   ~StringBuilder() { if (!is_finalized()) Finalize(); }
 245
 246   int size() const { return buffer_.length(); }
 247
 248   // Get the current position in the builder.
 249   int position() const {
 250     ASSERT(!is_finalized());
 251     return position_;
 252   }
 253
 254   // Reset the position.
 255   void Reset() { position_ = 0; }
 256
 257   // Add a single character to the builder. It is not allowed to add
 258   // 0-characters; use the Finalize() method to terminate the string
 259   // instead.
 260   void AddCharacter(char c) {
 261     ASSERT(c != '\0');
 262     ASSERT(!is_finalized() && position_ < buffer_.length());
 263     buffer_[position_++] = c;
 264   }
 265
 266   // Add an entire string to the builder. Uses strlen() internally to
 267   // compute the length of the input string.
 268   void AddString(const char* s) {
 269     AddSubstring(s, StrLength(s));
 270   }
 271
 272   // Add the first 'n' characters of the given string 's' to the
 273   // builder. The input string must have enough characters.
 274   void AddSubstring(const char* s, int n) {
 275     ASSERT(!is_finalized() && position_ + n < buffer_.length());
 276     ASSERT(static_cast<size_t>(n) <= strlen(s));
 277     memmove(&buffer_[position_], s, n * kCharSize);
 278     position_ += n;
 279   }
 280
 281
 282   // Add character padding to the builder. If count is non-positive,
 283   // nothing is added to the builder.
 284   void AddPadding(char c, int count) {
 285     for (int i = 0; i < count; i++) {
 286       AddCharacter(c);
 287     }
 288   }
 289
 290   // Finalize the string by 0-terminating it and returning the buffer.
 291   char* Finalize() {
 292     ASSERT(!is_finalized() && position_ < buffer_.length());
 293     buffer_[position_] = '\0';
 294     // Make sure nobody managed to add a 0-character to the
 295     // buffer while building the string.
 296     ASSERT(strlen(buffer_.start()) == static_cast<size_t>(position_));
 297     position_ = -1;
 298     ASSERT(is_finalized());
 299     return buffer_.start();
 300   }
 301
 302  private:
 303   Vector<char> buffer_;
 304   int position_;
 305
 306   bool is_finalized() const { return position_ < 0; }
 307
 308   DISALLOW_IMPLICIT_CONSTRUCTORS(StringBuilder);
 309 };
 310
 311 // The type-based aliasing rule allows the compiler to assume that pointers of
 312 // different types (for some definition of different) never alias each other.
 313 // Thus the following code does not work:
 314 //
 315 // float f = foo();
 316 // int fbits = *(int*)(&f);
 317 //
 318 // The compiler 'knows' that the int pointer can't refer to f since the types
 319 // don't match, so the compiler may cache f in a register, leaving random data
 320 // in fbits.  Using C++ style casts makes no difference, however a pointer to
 321 // char data is assumed to alias any other pointer.  This is the 'memcpy
 322 // exception'.
 323 //
 324 // Bit_cast uses the memcpy exception to move the bits from a variable of one
 325 // type of a variable of another type.  Of course the end result is likely to
 326 // be implementation dependent.  Most compilers (gcc-4.2 and MSVC 2005)
 327 // will completely optimize BitCast away.
 328 //
 329 // There is an additional use for BitCast.
 330 // Recent gccs will warn when they see casts that may result in breakage due to
 331 // the type-based aliasing rule.  If you have checked that there is no breakage
 332 // you can use BitCast to cast one pointer type to another.  This confuses gcc
 333 // enough that it can no longer see that you have cast one pointer type to
 334 // another thus avoiding the warning.
 335 template <class Dest, class Source>
 336 inline Dest BitCast(const Source& source) {
 337   // Compile time assertion: sizeof(Dest) == sizeof(Source)
 338   // A compile error here means your Dest and Source have different sizes.
 339   DOUBLE_CONVERSION_UNUSED
 340       typedef char VerifySizesAreEqual[sizeof(Dest) == sizeof(Source) ? 1 : -1];
 341
 342   Dest dest;
 343   memmove(&dest, &source, sizeof(dest));
 344   return dest;
 345 }
 346
 347 template <class Dest, class Source>
 348 inline Dest BitCast(Source* source) {
 349   return BitCast<Dest>(reinterpret_cast<uintptr_t>(source));
 350 }
 351
 352 }  // namespace double_conversion
 353
 354 // ICU PATCH: Close ICU namespace
 355 U_NAMESPACE_END
 356
 357 #endif  // DOUBLE_CONVERSION_UTILS_H_
 358 #endif // ICU PATCH: close #if !UCONFIG_NO_FORMATTING