dom/base/nsTextFragment.h

   1 /* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
   2 /* vim: set ts=8 sts=2 et sw=2 tw=80: */
   3 /* This Source Code Form is subject to the terms of the Mozilla Public
   4  * License, v. 2.0. If a copy of the MPL was not distributed with this
   5  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
   6
   7 /*
   8  * A class which represents a fragment of text (eg inside a text
   9  * node); if only codepoints below 256 are used, the text is stored as
  10  * a char*; otherwise the text is stored as a char16_t*
  11  */
  12
  13 #ifndef nsTextFragment_h___
  14 #define nsTextFragment_h___
  15
  16 #include "mozilla/Attributes.h"
  17 #include "mozilla/MemoryReporting.h"
  18
  19 #include "nsCharTraits.h"
  20 #include "nsString.h"
  21 #include "nsStringBuffer.h"
  22 #include "nsReadableUtils.h"
  23 #include "nsISupportsImpl.h"
  24
  25 // XXX should this normalize the code to keep a \u0000 at the end?
  26
  27 // XXX nsTextFragmentPool?
  28
  29 /**
  30  * A fragment of text. If mIs2b is 1 then the m2b pointer is valid
  31  * otherwise the m1b pointer is valid. If m1b is used then each byte
  32  * of data represents a single ucs2 character with the high byte being
  33  * zero.
  34  *
  35  * This class does not have a virtual destructor therefore it is not
  36  * meant to be subclassed.
  37  */
  38 class nsTextFragment final {
  39  public:
  40   static nsresult Init();
  41   static void Shutdown();
  42
  43   /**
  44    * Default constructor. Initialize the fragment to be empty.
  45    */
  46   nsTextFragment() : m1b(nullptr), mAllBits(0) {
  47     MOZ_COUNT_CTOR(nsTextFragment);
  48     NS_ASSERTION(sizeof(FragmentBits) == 4, "Bad field packing!");
  49   }
  50
  51   ~nsTextFragment();
  52
  53   /**
  54    * Change the contents of this fragment to be a copy of the
  55    * the argument fragment, or to "" if unable to allocate enough memory.
  56    */
  57   nsTextFragment& operator=(const nsTextFragment& aOther);
  58
  59   /**
  60    * Return true if this fragment is represented by char16_t data
  61    */
  62   bool Is2b() const { return mState.mIs2b; }
  63
  64   /**
  65    * Return true if this fragment contains Bidi text
  66    * For performance reasons this flag is only set if explicitely requested (by
  67    * setting the aUpdateBidi argument on SetTo or Append to true).
  68    */
  69   bool IsBidi() const { return mState.mIsBidi; }
  70
  71   /**
  72    * Get a pointer to constant char16_t data.
  73    */
  74   const char16_t* Get2b() const {
  75     MOZ_ASSERT(Is2b(), "not 2b text");
  76     return static_cast<char16_t*>(m2b->Data());
  77   }
  78
  79   /**
  80    * Get a pointer to constant char data.
  81    */
  82   const char* Get1b() const {
  83     NS_ASSERTION(!Is2b(), "not 1b text");
  84     return (const char*)m1b;
  85   }
  86
  87   /**
  88    * Get the length of the fragment. The length is the number of logical
  89    * characters, not the number of bytes to store the characters.
  90    */
  91   uint32_t GetLength() const { return mState.mLength; }
  92
  93 #define NS_MAX_TEXT_FRAGMENT_LENGTH (static_cast<uint32_t>(0x1FFFFFFF))
  94
  95   bool CanGrowBy(size_t n) const {
  96     return n < (1 << 29) && mState.mLength + n < (1 << 29);
  97   }
  98
  99   /**
 100    * Change the contents of this fragment to be a copy of the given
 101    * buffer. If aUpdateBidi is true, contents of the fragment will be scanned,
 102    * and mState.mIsBidi will be turned on if it includes any Bidi characters.
 103    * If aForce2b is true, aBuffer will be stored as char16_t as is.  Then,
 104    * you can access the value faster but may waste memory if all characters
 105    * are less than U+0100.
 106    */
 107   bool SetTo(const char16_t* aBuffer, uint32_t aLength, bool aUpdateBidi,
 108              bool aForce2b);
 109
 110   bool SetTo(const nsString& aString, bool aUpdateBidi, bool aForce2b) {
 111     if (MOZ_UNLIKELY(aString.Length() > NS_MAX_TEXT_FRAGMENT_LENGTH)) {
 112       return false;
 113     }
 114     ReleaseText();
 115     if (aForce2b && !aUpdateBidi) {
 116       nsStringBuffer* buffer = nsStringBuffer::FromString(aString);
 117       if (buffer) {
 118         NS_ADDREF(m2b = buffer);
 119         mState.mInHeap = true;
 120         mState.mIs2b = true;
 121         mState.mLength = aString.Length();
 122         return true;
 123       }
 124     }
 125
 126     return SetTo(aString.get(), aString.Length(), aUpdateBidi, aForce2b);
 127   }
 128
 129   /**
 130    * Append aData to the end of this fragment. If aUpdateBidi is true, contents
 131    * of the fragment will be scanned, and mState.mIsBidi will be turned on if
 132    * it includes any Bidi characters.
 133    * If aForce2b is true, the string will be stored as char16_t as is.  Then,
 134    * you can access the value faster but may waste memory if all characters
 135    * are less than U+0100.
 136    */
 137   bool Append(const char16_t* aBuffer, uint32_t aLength, bool aUpdateBidi,
 138               bool aForce2b);
 139
 140   /**
 141    * Append the contents of this string fragment to aString
 142    */
 143   void AppendTo(nsAString& aString) const {
 144     if (!AppendTo(aString, mozilla::fallible)) {
 145       aString.AllocFailed(aString.Length() + GetLength());
 146     }
 147   }
 148
 149   /**
 150    * Append the contents of this string fragment to aString
 151    * @return false if an out of memory condition is detected, true otherwise
 152    */
 153   [[nodiscard]] bool AppendTo(nsAString& aString,
 154                               const mozilla::fallible_t& aFallible) const {
 155     if (mState.mIs2b) {
 156       if (aString.IsEmpty()) {
 157         m2b->ToString(mState.mLength, aString);
 158         return true;
 159       }
 160       bool ok = aString.Append(Get2b(), mState.mLength, aFallible);
 161       if (!ok) {
 162         return false;
 163       }
 164
 165       return true;
 166     } else {
 167       return AppendASCIItoUTF16(Substring(m1b, mState.mLength), aString,
 168                                 aFallible);
 169     }
 170   }
 171
 172   /**
 173    * Append a substring of the contents of this string fragment to aString.
 174    * @param aOffset where to start the substring in this text fragment
 175    * @param aLength the length of the substring
 176    */
 177   void AppendTo(nsAString& aString, uint32_t aOffset, uint32_t aLength) const {
 178     if (!AppendTo(aString, aOffset, aLength, mozilla::fallible)) {
 179       aString.AllocFailed(aString.Length() + aLength);
 180     }
 181   }
 182
 183   /**
 184    * Append a substring of the contents of this string fragment to aString.
 185    * @param aString the string in which to append
 186    * @param aOffset where to start the substring in this text fragment
 187    * @param aLength the length of the substring
 188    * @return false if an out of memory condition is detected, true otherwise
 189    */
 190   [[nodiscard]] bool AppendTo(nsAString& aString, uint32_t aOffset,
 191                               uint32_t aLength,
 192                               const mozilla::fallible_t& aFallible) const {
 193     if (mState.mIs2b) {
 194       bool ok = aString.Append(Get2b() + aOffset, aLength, aFallible);
 195       if (!ok) {
 196         return false;
 197       }
 198
 199       return true;
 200     } else {
 201       return AppendASCIItoUTF16(Substring(m1b + aOffset, aLength), aString,
 202                                 aFallible);
 203     }
 204   }
 205
 206   /**
 207    * Make a copy of the fragments contents starting at offset for
 208    * count characters. The offset and count will be adjusted to
 209    * lie within the fragments data. The fragments data is converted if
 210    * necessary.
 211    */
 212   void CopyTo(char16_t* aDest, uint32_t aOffset, uint32_t aCount);
 213
 214   /**
 215    * Return the character in the text-fragment at the given
 216    * index. This always returns a char16_t.
 217    */
 218   char16_t CharAt(uint32_t aIndex) const {
 219     MOZ_ASSERT(aIndex < mState.mLength, "bad index");
 220     return mState.mIs2b ? Get2b()[aIndex]
 221                         : static_cast<unsigned char>(m1b[aIndex]);
 222   }
 223
 224   /**
 225    * IsHighSurrogateFollowedByLowSurrogateAt() returns true if character at
 226    * aIndex is high surrogate and it's followed by low surrogate.
 227    */
 228   inline bool IsHighSurrogateFollowedByLowSurrogateAt(uint32_t aIndex) const {
 229     MOZ_ASSERT(aIndex < mState.mLength);
 230     if (!mState.mIs2b || aIndex + 1 >= mState.mLength) {
 231       return false;
 232     }
 233     return NS_IS_SURROGATE_PAIR(Get2b()[aIndex], Get2b()[aIndex + 1]);
 234   }
 235
 236   /**
 237    * IsLowSurrogateFollowingHighSurrogateAt() returns true if character at
 238    * aIndex is low surrogate and it follows high surrogate.
 239    */
 240   inline bool IsLowSurrogateFollowingHighSurrogateAt(uint32_t aIndex) const {
 241     MOZ_ASSERT(aIndex < mState.mLength);
 242     if (!mState.mIs2b || !aIndex) {
 243       return false;
 244     }
 245     return NS_IS_SURROGATE_PAIR(Get2b()[aIndex - 1], Get2b()[aIndex]);
 246   }
 247
 248   /**
 249    * ScalarValueAt() returns a Unicode scalar value at aIndex.  If the character
 250    * at aIndex is a high surrogate followed by low surrogate, returns character
 251    * code for the pair.  If the index is low surrogate, or a high surrogate but
 252    * not in a pair, returns 0.
 253    */
 254   inline char32_t ScalarValueAt(uint32_t aIndex) const {
 255     MOZ_ASSERT(aIndex < mState.mLength);
 256     if (!mState.mIs2b) {
 257       return static_cast<unsigned char>(m1b[aIndex]);
 258     }
 259     char16_t ch = Get2b()[aIndex];
 260     if (!IS_SURROGATE(ch)) {
 261       return ch;
 262     }
 263     if (aIndex + 1 < mState.mLength && NS_IS_HIGH_SURROGATE(ch)) {
 264       char16_t nextCh = Get2b()[aIndex + 1];
 265       if (NS_IS_LOW_SURROGATE(nextCh)) {
 266         return SURROGATE_TO_UCS4(ch, nextCh);
 267       }
 268     }
 269     return 0;
 270   }
 271
 272   void SetBidi(bool aBidi) { mState.mIsBidi = aBidi; }
 273
 274   struct FragmentBits {
 275     // uint32_t to ensure that the values are unsigned, because we
 276     // want 0/1, not 0/-1!
 277     // Making these bool causes Windows to not actually pack them,
 278     // which causes crashes because we assume this structure is no more than
 279     // 32 bits!
 280     uint32_t mInHeap : 1;
 281     uint32_t mIs2b : 1;
 282     uint32_t mIsBidi : 1;
 283     // Note that when you change the bits of mLength, you also need to change
 284     // NS_MAX_TEXT_FRAGMENT_LENGTH.
 285     uint32_t mLength : 29;
 286   };
 287
 288   size_t SizeOfExcludingThis(mozilla::MallocSizeOf aMallocSizeOf) const;
 289
 290   /**
 291    * Check whether the text in this fragment is the same as the text in the
 292    * other fragment.
 293    */
 294   [[nodiscard]] bool TextEquals(const nsTextFragment& aOther) const;
 295
 296  private:
 297   void ReleaseText();
 298
 299   /**
 300    * Scan the contents of the fragment and turn on mState.mIsBidi if it
 301    * includes any Bidi characters.
 302    */
 303   void UpdateBidiFlag(const char16_t* aBuffer, uint32_t aLength);
 304
 305   union {
 306     nsStringBuffer* m2b;
 307     const char* m1b;  // This is const since it can point to shared data
 308   };
 309
 310   union {
 311     uint32_t mAllBits;
 312     FragmentBits mState;
 313   };
 314 };
 315
 316 #endif /* nsTextFragment_h___ */