mfbt/Latin1.h

   1 /* This Source Code Form is subject to the terms of the Mozilla Public
   2  * License, v. 2.0. If a copy of the MPL was not distributed with this
   3  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
   4
   5 /* Latin-1 operations (i.e. a byte is the corresponding code point).
   6  * (Note: this is *not* the same as the encoding of windows-1252 or
   7  * latin1 content on the web. In Web terms, this encoding
   8  * corresponds to "isomorphic decode" / "isomorphic encoding" from
   9  * the Infra Standard.)
  10  */
  11
  12 #ifndef mozilla_Latin1_h
  13 #define mozilla_Latin1_h
  14
  15 #include <type_traits>
  16
  17 #include "mozilla/JsRust.h"
  18 #include "mozilla/Span.h"
  19
  20 #if MOZ_HAS_JSRUST()
  21 #  include "encoding_rs_mem.h"
  22 #endif
  23
  24 namespace mozilla {
  25
  26 namespace detail {
  27
  28 // It's important for optimizations that Latin1ness checks
  29 // and inflation/deflation function use the same short
  30 // string limit. The limit is 16, because that's the shortest
  31 // that inflates/deflates using SIMD.
  32 constexpr size_t kShortStringLimitForInlinePaths = 16;
  33
  34 template <typename Char>
  35 class MakeUnsignedChar {
  36  public:
  37   using Type = std::make_unsigned_t<Char>;
  38 };
  39
  40 template <>
  41 class MakeUnsignedChar<char16_t> {
  42  public:
  43   using Type = char16_t;
  44 };
  45
  46 template <>
  47 class MakeUnsignedChar<char32_t> {
  48  public:
  49   using Type = char32_t;
  50 };
  51
  52 }  // namespace detail
  53
  54 /**
  55  * Returns true iff |aChar| is Latin-1 but not ASCII, i.e. in the range
  56  * [0x80, 0xFF].
  57  */
  58 template <typename Char>
  59 constexpr bool IsNonAsciiLatin1(Char aChar) {
  60   using UnsignedChar = typename detail::MakeUnsignedChar<Char>::Type;
  61   auto uc = static_cast<UnsignedChar>(aChar);
  62   return uc >= 0x80 && uc <= 0xFF;
  63 }
  64
  65 #if MOZ_HAS_JSRUST()
  66
  67 /**
  68  * Returns |true| iff |aString| contains only Latin1 characters, that is,
  69  * characters in the range [U+0000, U+00FF].
  70  *
  71  * @param aString a potentially-invalid UTF-16 string to scan
  72  */
  73 inline bool IsUtf16Latin1(mozilla::Span<const char16_t> aString) {
  74   size_t length = aString.Length();
  75   const char16_t* ptr = aString.Elements();
  76   // For short strings, calling into Rust is a pessimization, and the SIMD
  77   // code won't have a chance to kick in anyway.
  78   // 16 is a bit larger than logically necessary for this function alone,
  79   // but it's important that the limit here matches the limit used in
  80   // LossyConvertUtf16toLatin1!
  81   if (length < mozilla::detail::kShortStringLimitForInlinePaths) {
  82     char16_t accu = 0;
  83     for (size_t i = 0; i < length; i++) {
  84       accu |= ptr[i];
  85     }
  86     return accu < 0x100;
  87   }
  88   return encoding_mem_is_utf16_latin1(ptr, length);
  89 }
  90
  91 /**
  92  * Returns |true| iff |aString| is valid UTF-8 containing only Latin-1
  93  * characters.
  94  *
  95  * If you know that the argument is always absolutely guaranteed to be valid
  96  * UTF-8, use the faster UnsafeIsValidUtf8Latin1() instead.
  97  *
  98  * @param aString potentially-invalid UTF-8 string to scan
  99  */
 100 inline bool IsUtf8Latin1(mozilla::Span<const char> aString) {
 101   return encoding_mem_is_utf8_latin1(aString.Elements(), aString.Length());
 102 }
 103
 104 /**
 105  * Returns |true| iff |aString|, which MUST be valid UTF-8, contains only
 106  * Latin1 characters, that is, characters in the range [U+0000, U+00FF].
 107  * (If |aString| might not be valid UTF-8, use |IsUtf8Latin1| instead.)
 108  *
 109  * @param aString known-valid UTF-8 string to scan
 110  */
 111 inline bool UnsafeIsValidUtf8Latin1(mozilla::Span<const char> aString) {
 112   return encoding_mem_is_str_latin1(aString.Elements(), aString.Length());
 113 }
 114
 115 /**
 116  * Returns the index of first byte that starts an invalid byte
 117  * sequence or a non-Latin1 byte sequence in a potentially-invalid UTF-8
 118  * string, or the length of the string if there are neither.
 119  *
 120  * If you know that the argument is always absolutely guaranteed to be valid
 121  * UTF-8, use the faster UnsafeValidUtf8Lati1UpTo() instead.
 122  *
 123  * @param aString potentially-invalid UTF-8 string to scan
 124  */
 125 inline size_t Utf8Latin1UpTo(mozilla::Span<const char> aString) {
 126   return encoding_mem_utf8_latin1_up_to(aString.Elements(), aString.Length());
 127 }
 128
 129 /**
 130  * Returns the index of first byte that starts a non-Latin1 byte
 131  * sequence in a known-valid UTF-8 string, or the length of the
 132  * string if there are none. (If the string might not be valid
 133  * UTF-8, use Utf8Latin1UpTo() instead.)
 134  *
 135  * @param aString known-valid UTF-8 string to scan
 136  */
 137 inline size_t UnsafeValidUtf8Lati1UpTo(mozilla::Span<const char> aString) {
 138   return encoding_mem_str_latin1_up_to(aString.Elements(), aString.Length());
 139 }
 140
 141 /**
 142  * If all the code points in the input are below U+0100, converts to Latin1,
 143  * i.e. unsigned byte value is Unicode scalar value. If there are code points
 144  * above U+00FF, produces unspecified garbage in a memory-safe way. The
 145  * nature of the garbage must not be relied upon.
 146  *
 147  * The length of aDest must not be less than the length of aSource.
 148  */
 149 inline void LossyConvertUtf16toLatin1(mozilla::Span<const char16_t> aSource,
 150                                       mozilla::Span<char> aDest) {
 151   const char16_t* srcPtr = aSource.Elements();
 152   size_t srcLen = aSource.Length();
 153   char* dstPtr = aDest.Elements();
 154   size_t dstLen = aDest.Length();
 155   // Avoid function call overhead when SIMD isn't used anyway
 156   // If you change the length limit here, be sure to change
 157   // IsUtf16Latin1 and IsAscii to match so that optimizations don't
 158   // fail!
 159   if (srcLen < mozilla::detail::kShortStringLimitForInlinePaths) {
 160     MOZ_ASSERT(dstLen >= srcLen);
 161     uint8_t* unsignedPtr = reinterpret_cast<uint8_t*>(dstPtr);
 162     const char16_t* end = srcPtr + srcLen;
 163     while (srcPtr < end) {
 164       *unsignedPtr = static_cast<uint8_t>(*srcPtr);
 165       ++srcPtr;
 166       ++unsignedPtr;
 167     }
 168     return;
 169   }
 170   encoding_mem_convert_utf16_to_latin1_lossy(srcPtr, srcLen, dstPtr, dstLen);
 171 }
 172
 173 /**
 174  * If all the code points in the input are below U+0100, converts to Latin1,
 175  * i.e. unsigned byte value is Unicode scalar value. If there are code points
 176  * above U+00FF, produces unspecified garbage in a memory-safe way. The
 177  * nature of the garbage must not be relied upon.
 178  *
 179  * Returns the number of code units written.
 180  *
 181  * The length of aDest must not be less than the length of aSource.
 182  */
 183 inline size_t LossyConvertUtf8toLatin1(mozilla::Span<const char> aSource,
 184                                        mozilla::Span<char> aDest) {
 185   return encoding_mem_convert_utf8_to_latin1_lossy(
 186       aSource.Elements(), aSource.Length(), aDest.Elements(), aDest.Length());
 187 }
 188
 189 /**
 190  * Converts each byte of |aSource|, interpreted as a Unicode scalar value
 191  * having that unsigned value, to its UTF-8 representation in |aDest|.
 192  *
 193  * Returns the number of code units written.
 194  *
 195  * The length of aDest must be at least twice the length of aSource.
 196  */
 197 inline size_t ConvertLatin1toUtf8(mozilla::Span<const char> aSource,
 198                                   mozilla::Span<char> aDest) {
 199   return encoding_mem_convert_latin1_to_utf8(
 200       aSource.Elements(), aSource.Length(), aDest.Elements(), aDest.Length());
 201 }
 202
 203 /**
 204  * Converts bytes whose unsigned value is interpreted as Unicode code point
 205  * (i.e. U+0000 to U+00FF, inclusive) to UTF-8 with potentially insufficient
 206  * output space.
 207  *
 208  * Returns the number of bytes read and the number of bytes written.
 209  *
 210  * If the output isn't large enough, not all input is consumed.
 211  *
 212  * The conversion is guaranteed to be complete if the length of aDest is
 213  * at least the length of aSource times two.
 214  *
 215  * The output is always valid UTF-8 ending on scalar value boundary
 216  * even in the case of partial conversion.
 217  *
 218  * The semantics of this function match the semantics of
 219  * TextEncoder.encodeInto.
 220  * https://encoding.spec.whatwg.org/#dom-textencoder-encodeinto
 221  */
 222 inline std::tuple<size_t, size_t> ConvertLatin1toUtf8Partial(
 223     mozilla::Span<const char> aSource, mozilla::Span<char> aDest) {
 224   size_t srcLen = aSource.Length();
 225   size_t dstLen = aDest.Length();
 226   encoding_mem_convert_latin1_to_utf8_partial(aSource.Elements(), &srcLen,
 227                                               aDest.Elements(), &dstLen);
 228   return std::make_tuple(srcLen, dstLen);
 229 }
 230
 231 /**
 232  * Converts Latin-1 code points (i.e. each byte is the identical code
 233  * point) from |aSource| to UTF-16 code points in |aDest|.
 234  *
 235  * The length of aDest must not be less than the length of aSource.
 236  */
 237 inline void ConvertLatin1toUtf16(mozilla::Span<const char> aSource,
 238                                  mozilla::Span<char16_t> aDest) {
 239   const char* srcPtr = aSource.Elements();
 240   size_t srcLen = aSource.Length();
 241   char16_t* dstPtr = aDest.Elements();
 242   size_t dstLen = aDest.Length();
 243   // Avoid function call overhead when SIMD isn't used anyway
 244   if (srcLen < mozilla::detail::kShortStringLimitForInlinePaths) {
 245     MOZ_ASSERT(dstLen >= srcLen);
 246     const uint8_t* unsignedPtr = reinterpret_cast<const uint8_t*>(srcPtr);
 247     const uint8_t* end = unsignedPtr + srcLen;
 248     while (unsignedPtr < end) {
 249       *dstPtr = *unsignedPtr;
 250       ++unsignedPtr;
 251       ++dstPtr;
 252     }
 253     return;
 254   }
 255   encoding_mem_convert_latin1_to_utf16(srcPtr, srcLen, dstPtr, dstLen);
 256 }
 257
 258 #endif
 259
 260 };  // namespace mozilla
 261
 262 #endif  // mozilla_Latin1_h