third_party/WebKit/Source/wtf/text/StringImpl.h

   1 /*
   2  * Copyright (C) 1999 Lars Knoll (knoll@kde.org)
   3  * Copyright (C) 2005, 2006, 2007, 2008, 2009, 2010, 2013 Apple Inc. All rights reserved.
   4  * Copyright (C) 2009 Google Inc. All rights reserved.
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Library General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Library General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Library General Public License
  17  * along with this library; see the file COPYING.LIB.  If not, write to
  18  * the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
  19  * Boston, MA 02110-1301, USA.
  20  *
  21  */
  22
  23 #ifndef StringImpl_h
  24 #define StringImpl_h
  25
  26 #include "wtf/ASCIICType.h"
  27 #include "wtf/Forward.h"
  28 #include "wtf/HashMap.h"
  29 #include "wtf/StringHasher.h"
  30 #include "wtf/Vector.h"
  31 #include "wtf/WTFExport.h"
  32 #include "wtf/text/Unicode.h"
  33 #include <limits.h>
  34
  35 #if USE(CF)
  36 typedef const struct __CFString * CFStringRef;
  37 #endif
  38
  39 #ifdef __OBJC__
  40 @class NSString;
  41 #endif
  42
  43 namespace WTF {
  44
  45 struct AlreadyHashed;
  46 struct CStringTranslator;
  47 template<typename CharacterType> struct HashAndCharactersTranslator;
  48 struct HashAndUTF8CharactersTranslator;
  49 struct LCharBufferTranslator;
  50 struct CharBufferFromLiteralDataTranslator;
  51 struct SubstringTranslator;
  52 struct UCharBufferTranslator;
  53 template<typename> class RetainPtr;
  54
  55 enum TextCaseSensitivity { TextCaseSensitive, TextCaseInsensitive };
  56
  57 enum StripBehavior { StripExtraWhiteSpace, DoNotStripWhiteSpace };
  58
  59 typedef bool (*CharacterMatchFunctionPtr)(UChar);
  60 typedef bool (*IsWhiteSpaceFunctionPtr)(UChar);
  61 typedef HashMap<unsigned, StringImpl*, AlreadyHashed> StaticStringsTable;
  62
  63 // Define STRING_STATS to turn on run time statistics of string sizes and memory usage
  64 #undef STRING_STATS
  65
  66 #ifdef STRING_STATS
  67 struct StringStats {
  68     inline void add8BitString(unsigned length)
  69     {
  70         ++m_totalNumberStrings;
  71         ++m_number8BitStrings;
  72         m_total8BitData += length;
  73     }
  74
  75     inline void add16BitString(unsigned length)
  76     {
  77         ++m_totalNumberStrings;
  78         ++m_number16BitStrings;
  79         m_total16BitData += length;
  80     }
  81
  82     void removeString(StringImpl*);
  83     void printStats();
  84
  85     static const unsigned s_printStringStatsFrequency = 5000;
  86     static unsigned s_stringRemovesTillPrintStats;
  87
  88     unsigned m_totalNumberStrings;
  89     unsigned m_number8BitStrings;
  90     unsigned m_number16BitStrings;
  91     unsigned long long m_total8BitData;
  92     unsigned long long m_total16BitData;
  93 };
  94
  95 void addStringForStats(StringImpl*);
  96 void removeStringForStats(StringImpl*);
  97
  98 #define STRING_STATS_ADD_8BIT_STRING(length) StringImpl::stringStats().add8BitString(length); addStringForStats(this)
  99 #define STRING_STATS_ADD_16BIT_STRING(length) StringImpl::stringStats().add16BitString(length); addStringForStats(this)
 100 #define STRING_STATS_REMOVE_STRING(string) StringImpl::stringStats().removeString(string); removeStringForStats(this)
 101 #else
 102 #define STRING_STATS_ADD_8BIT_STRING(length) ((void)0)
 103 #define STRING_STATS_ADD_16BIT_STRING(length) ((void)0)
 104 #define STRING_STATS_REMOVE_STRING(string) ((void)0)
 105 #endif
 106
 107 // You can find documentation about this class in this doc:
 108 // https://docs.google.com/document/d/1kOCUlJdh2WJMJGDf-WoEQhmnjKLaOYRbiHz5TiGJl14/edit?usp=sharing
 109 class WTF_EXPORT StringImpl {
 110     WTF_MAKE_NONCOPYABLE(StringImpl);
 111     friend struct WTF::CStringTranslator;
 112     template<typename CharacterType> friend struct WTF::HashAndCharactersTranslator;
 113     friend struct WTF::HashAndUTF8CharactersTranslator;
 114     friend struct WTF::CharBufferFromLiteralDataTranslator;
 115     friend struct WTF::LCharBufferTranslator;
 116     friend struct WTF::SubstringTranslator;
 117     friend struct WTF::UCharBufferTranslator;
 118
 119 private:
 120     // StringImpls are allocated out of the WTF buffer partition.
 121     void* operator new(size_t);
 122     void* operator new(size_t, void* ptr) { return ptr; }
 123     void operator delete(void*);
 124
 125     // Used to construct static strings, which have an special refCount that can never hit zero.
 126     // This means that the static string will never be destroyed, which is important because
 127     // static strings will be shared across threads & ref-counted in a non-threadsafe manner.
 128     enum ConstructEmptyStringTag { ConstructEmptyString };
 129     explicit StringImpl(ConstructEmptyStringTag)
 130         : m_refCount(1)
 131         , m_length(0)
 132         , m_hash(0)
 133         , m_isAtomic(false)
 134         , m_is8Bit(true)
 135         , m_isStatic(true)
 136     {
 137         // Ensure that the hash is computed so that AtomicStringHash can call existingHash()
 138         // with impunity. The empty string is special because it is never entered into
 139         // AtomicString's HashKey, but still needs to compare correctly.
 140         STRING_STATS_ADD_8BIT_STRING(m_length);
 141         hash();
 142     }
 143
 144     enum ConstructEmptyString16BitTag { ConstructEmptyString16Bit };
 145     explicit StringImpl(ConstructEmptyString16BitTag)
 146         : m_refCount(1)
 147         , m_length(0)
 148         , m_hash(0)
 149         , m_isAtomic(false)
 150         , m_is8Bit(false)
 151         , m_isStatic(true)
 152     {
 153         STRING_STATS_ADD_16BIT_STRING(m_length);
 154         hash();
 155     }
 156
 157     // FIXME: there has to be a less hacky way to do this.
 158     enum Force8Bit { Force8BitConstructor };
 159     StringImpl(unsigned length, Force8Bit)
 160         : m_refCount(1)
 161         , m_length(length)
 162         , m_hash(0)
 163         , m_isAtomic(false)
 164         , m_is8Bit(true)
 165         , m_isStatic(false)
 166     {
 167         ASSERT(m_length);
 168         STRING_STATS_ADD_8BIT_STRING(m_length);
 169     }
 170
 171     StringImpl(unsigned length)
 172         : m_refCount(1)
 173         , m_length(length)
 174         , m_hash(0)
 175         , m_isAtomic(false)
 176         , m_is8Bit(false)
 177         , m_isStatic(false)
 178     {
 179         ASSERT(m_length);
 180         STRING_STATS_ADD_16BIT_STRING(m_length);
 181     }
 182
 183     enum StaticStringTag { StaticString };
 184     StringImpl(unsigned length, unsigned hash, StaticStringTag)
 185         : m_refCount(1)
 186         , m_length(length)
 187         , m_hash(hash)
 188         , m_isAtomic(false)
 189         , m_is8Bit(true)
 190         , m_isStatic(true)
 191     {
 192     }
 193
 194 public:
 195     ~StringImpl();
 196
 197     static StringImpl* createStatic(const char* string, unsigned length, unsigned hash);
 198     static void freezeStaticStrings();
 199     static const StaticStringsTable& allStaticStrings();
 200     static unsigned highestStaticStringLength() { return m_highestStaticStringLength; }
 201
 202     static PassRefPtr<StringImpl> create(const UChar*, unsigned length);
 203     static PassRefPtr<StringImpl> create(const LChar*, unsigned length);
 204     static PassRefPtr<StringImpl> create8BitIfPossible(const UChar*, unsigned length);
 205     template<size_t inlineCapacity>
 206     static PassRefPtr<StringImpl> create8BitIfPossible(const Vector<UChar, inlineCapacity>& vector)
 207     {
 208         return create8BitIfPossible(vector.data(), vector.size());
 209     }
 210
 211     ALWAYS_INLINE static PassRefPtr<StringImpl> create(const char* s, unsigned length) { return create(reinterpret_cast<const LChar*>(s), length); }
 212     static PassRefPtr<StringImpl> create(const LChar*);
 213     ALWAYS_INLINE static PassRefPtr<StringImpl> create(const char* s) { return create(reinterpret_cast<const LChar*>(s)); }
 214
 215     static PassRefPtr<StringImpl> createUninitialized(unsigned length, LChar*& data);
 216     static PassRefPtr<StringImpl> createUninitialized(unsigned length, UChar*& data);
 217
 218     // Reallocate the StringImpl. The originalString must be only owned by the PassRefPtr.
 219     // Just like the input pointer of realloc(), the originalString can't be used after this function.
 220     static PassRefPtr<StringImpl> reallocate(PassRefPtr<StringImpl> originalString, unsigned length);
 221
 222     // If this StringImpl has only one reference, we can truncate the string by updating
 223     // its m_length property without actually re-allocating its buffer.
 224     void truncateAssumingIsolated(unsigned length)
 225     {
 226         ASSERT(hasOneRef());
 227         ASSERT(length <= m_length);
 228         m_length = length;
 229     }
 230
 231     unsigned length() const { return m_length; }
 232     bool is8Bit() const { return m_is8Bit; }
 233
 234     ALWAYS_INLINE const LChar* characters8() const { ASSERT(is8Bit()); return reinterpret_cast<const LChar*>(this + 1); }
 235     ALWAYS_INLINE const UChar* characters16() const { ASSERT(!is8Bit()); return reinterpret_cast<const UChar*>(this + 1); }
 236
 237     template <typename CharType>
 238     ALWAYS_INLINE const CharType * getCharacters() const;
 239
 240     size_t sizeInBytes() const;
 241
 242     bool isAtomic() const { return m_isAtomic; }
 243     void setIsAtomic(bool isAtomic) { m_isAtomic = isAtomic; }
 244
 245     bool isStatic() const { return m_isStatic; }
 246
 247 private:
 248     // The high bits of 'hash' are always empty, but we prefer to store our flags
 249     // in the low bits because it makes them slightly more efficient to access.
 250     // So, we shift left and right when setting and getting our hash code.
 251     void setHash(unsigned hash) const
 252     {
 253         ASSERT(!hasHash());
 254         // Multiple clients assume that StringHasher is the canonical string hash function.
 255         ASSERT(hash == (is8Bit() ? StringHasher::computeHashAndMaskTop8Bits(characters8(), m_length) : StringHasher::computeHashAndMaskTop8Bits(characters16(), m_length)));
 256         m_hash = hash;
 257         ASSERT(hash); // Verify that 0 is a valid sentinel hash value.
 258     }
 259
 260     unsigned rawHash() const
 261     {
 262         return m_hash;
 263     }
 264
 265     void destroyIfNotStatic();
 266
 267 public:
 268     bool hasHash() const
 269     {
 270         return rawHash() != 0;
 271     }
 272
 273     unsigned existingHash() const
 274     {
 275         ASSERT(hasHash());
 276         return rawHash();
 277     }
 278
 279     unsigned hash() const
 280     {
 281         if (hasHash())
 282             return existingHash();
 283         return hashSlowCase();
 284     }
 285
 286     ALWAYS_INLINE bool hasOneRef() const
 287     {
 288         return m_refCount == 1;
 289     }
 290
 291     ALWAYS_INLINE void ref()
 292     {
 293         ++m_refCount;
 294     }
 295
 296     ALWAYS_INLINE void deref()
 297     {
 298         if (hasOneRef()) {
 299             destroyIfNotStatic();
 300             return;
 301         }
 302
 303         --m_refCount;
 304     }
 305
 306     static StringImpl* empty();
 307     static StringImpl* empty16Bit();
 308
 309     // FIXME: Does this really belong in StringImpl?
 310     template <typename T> static void copyChars(T* destination, const T* source, unsigned numCharacters)
 311     {
 312         memcpy(destination, source, numCharacters * sizeof(T));
 313     }
 314
 315     ALWAYS_INLINE static void copyChars(UChar* destination, const LChar* source, unsigned numCharacters)
 316     {
 317         for (unsigned i = 0; i < numCharacters; ++i)
 318             destination[i] = source[i];
 319     }
 320
 321     // Some string features, like refcounting and the atomicity flag, are not
 322     // thread-safe. We achieve thread safety by isolation, giving each thread
 323     // its own copy of the string.
 324     PassRefPtr<StringImpl> isolatedCopy() const;
 325
 326     PassRefPtr<StringImpl> substring(unsigned pos, unsigned len = UINT_MAX);
 327
 328     UChar operator[](unsigned i) const
 329     {
 330         ASSERT_WITH_SECURITY_IMPLICATION(i < m_length);
 331         if (is8Bit())
 332             return characters8()[i];
 333         return characters16()[i];
 334     }
 335     UChar32 characterStartingAt(unsigned);
 336
 337     bool containsOnlyWhitespace();
 338
 339     int toIntStrict(bool* ok = 0, int base = 10);
 340     unsigned toUIntStrict(bool* ok = 0, int base = 10);
 341     int64_t toInt64Strict(bool* ok = 0, int base = 10);
 342     uint64_t toUInt64Strict(bool* ok = 0, int base = 10);
 343
 344     int toInt(bool* ok = 0); // ignores trailing garbage
 345     unsigned toUInt(bool* ok = 0); // ignores trailing garbage
 346     int64_t toInt64(bool* ok = 0); // ignores trailing garbage
 347     uint64_t toUInt64(bool* ok = 0); // ignores trailing garbage
 348
 349     // FIXME: Like the strict functions above, these give false for "ok" when there is trailing garbage.
 350     // Like the non-strict functions above, these return the value when there is trailing garbage.
 351     // It would be better if these were more consistent with the above functions instead.
 352     double toDouble(bool* ok = 0);
 353     float toFloat(bool* ok = 0);
 354
 355     PassRefPtr<StringImpl> lower();
 356     PassRefPtr<StringImpl> upper();
 357     PassRefPtr<StringImpl> lower(const AtomicString& localeIdentifier);
 358     PassRefPtr<StringImpl> upper(const AtomicString& localeIdentifier);
 359
 360     PassRefPtr<StringImpl> fill(UChar);
 361     // FIXME: Do we need fill(char) or can we just do the right thing if UChar is ASCII?
 362     PassRefPtr<StringImpl> foldCase();
 363
 364     PassRefPtr<StringImpl> stripWhiteSpace();
 365     PassRefPtr<StringImpl> stripWhiteSpace(IsWhiteSpaceFunctionPtr);
 366     PassRefPtr<StringImpl> simplifyWhiteSpace(StripBehavior stripBehavior = StripExtraWhiteSpace);
 367     PassRefPtr<StringImpl> simplifyWhiteSpace(IsWhiteSpaceFunctionPtr, StripBehavior stripBehavior = StripExtraWhiteSpace);
 368
 369     PassRefPtr<StringImpl> removeCharacters(CharacterMatchFunctionPtr);
 370     template <typename CharType>
 371     ALWAYS_INLINE PassRefPtr<StringImpl> removeCharacters(const CharType* characters, CharacterMatchFunctionPtr);
 372
 373     size_t find(LChar character, unsigned start = 0);
 374     size_t find(char character, unsigned start = 0);
 375     size_t find(UChar character, unsigned start = 0);
 376     size_t find(CharacterMatchFunctionPtr, unsigned index = 0);
 377     size_t find(const LChar*, unsigned index = 0);
 378     ALWAYS_INLINE size_t find(const char* s, unsigned index = 0) { return find(reinterpret_cast<const LChar*>(s), index); }
 379     size_t find(StringImpl*);
 380     size_t find(StringImpl*, unsigned index);
 381     size_t findIgnoringCase(const LChar*, unsigned index = 0);
 382     ALWAYS_INLINE size_t findIgnoringCase(const char* s, unsigned index = 0) { return findIgnoringCase(reinterpret_cast<const LChar*>(s), index); }
 383     size_t findIgnoringCase(StringImpl*, unsigned index = 0);
 384
 385     size_t findNextLineStart(unsigned index = UINT_MAX);
 386
 387     size_t reverseFind(UChar, unsigned index = UINT_MAX);
 388     size_t reverseFind(StringImpl*, unsigned index = UINT_MAX);
 389     size_t reverseFindIgnoringCase(StringImpl*, unsigned index = UINT_MAX);
 390
 391     size_t count(LChar) const;
 392
 393     bool startsWith(StringImpl* str, TextCaseSensitivity caseSensitivity = TextCaseSensitive) { return ((caseSensitivity == TextCaseSensitive) ? reverseFind(str, 0) : reverseFindIgnoringCase(str, 0)) == 0; }
 394     bool startsWith(UChar) const;
 395     bool startsWith(const char*, unsigned matchLength, TextCaseSensitivity) const;
 396     template<unsigned matchLength>
 397     bool startsWith(const char (&prefix)[matchLength], TextCaseSensitivity caseSensitivity = TextCaseSensitive) const { return startsWith(prefix, matchLength - 1, caseSensitivity); }
 398
 399     bool endsWith(StringImpl*, TextCaseSensitivity = TextCaseSensitive);
 400     bool endsWith(UChar) const;
 401     bool endsWith(const char*, unsigned matchLength, TextCaseSensitivity) const;
 402     template<unsigned matchLength>
 403     bool endsWith(const char (&prefix)[matchLength], TextCaseSensitivity caseSensitivity = TextCaseSensitive) const { return endsWith(prefix, matchLength - 1, caseSensitivity); }
 404
 405     PassRefPtr<StringImpl> replace(UChar, UChar);
 406     PassRefPtr<StringImpl> replace(UChar, StringImpl*);
 407     ALWAYS_INLINE PassRefPtr<StringImpl> replace(UChar pattern, const char* replacement, unsigned replacementLength) { return replace(pattern, reinterpret_cast<const LChar*>(replacement), replacementLength); }
 408     PassRefPtr<StringImpl> replace(UChar, const LChar*, unsigned replacementLength);
 409     PassRefPtr<StringImpl> replace(UChar, const UChar*, unsigned replacementLength);
 410     PassRefPtr<StringImpl> replace(StringImpl*, StringImpl*);
 411     PassRefPtr<StringImpl> replace(unsigned index, unsigned len, StringImpl*);
 412     PassRefPtr<StringImpl> upconvertedString();
 413
 414 #if USE(CF)
 415     RetainPtr<CFStringRef> createCFString();
 416 #endif
 417 #ifdef __OBJC__
 418     operator NSString*();
 419 #endif
 420
 421 #ifdef STRING_STATS
 422     ALWAYS_INLINE static StringStats& stringStats() { return m_stringStats; }
 423 #endif
 424     static const UChar latin1CaseFoldTable[256];
 425
 426 private:
 427     template<typename CharType> static size_t allocationSize(unsigned length)
 428     {
 429         RELEASE_ASSERT(length <= ((std::numeric_limits<unsigned>::max() - sizeof(StringImpl)) / sizeof(CharType)));
 430         return sizeof(StringImpl) + length * sizeof(CharType);
 431     }
 432
 433     template <class UCharPredicate> PassRefPtr<StringImpl> stripMatchedCharacters(UCharPredicate);
 434     template <typename CharType, class UCharPredicate> PassRefPtr<StringImpl> simplifyMatchedCharactersToSpace(UCharPredicate, StripBehavior);
 435     NEVER_INLINE unsigned hashSlowCase() const;
 436
 437 #ifdef STRING_STATS
 438     static StringStats m_stringStats;
 439 #endif
 440
 441     static unsigned m_highestStaticStringLength;
 442
 443 #if ENABLE(ASSERT)
 444     void assertHashIsCorrect()
 445     {
 446         ASSERT(hasHash());
 447         ASSERT(existingHash() == StringHasher::computeHashAndMaskTop8Bits(characters8(), length()));
 448     }
 449 #endif
 450
 451 private:
 452     unsigned m_refCount;
 453     unsigned m_length;
 454     mutable unsigned m_hash : 24;
 455     unsigned m_isAtomic : 1;
 456     unsigned m_is8Bit : 1;
 457     unsigned m_isStatic : 1;
 458 };
 459
 460 template <>
 461 ALWAYS_INLINE const LChar* StringImpl::getCharacters<LChar>() const { return characters8(); }
 462
 463 template <>
 464 ALWAYS_INLINE const UChar* StringImpl::getCharacters<UChar>() const { return characters16(); }
 465
 466 WTF_EXPORT bool equal(const StringImpl*, const StringImpl*);
 467 WTF_EXPORT bool equal(const StringImpl*, const LChar*);
 468 inline bool equal(const StringImpl* a, const char* b) { return equal(a, reinterpret_cast<const LChar*>(b)); }
 469 WTF_EXPORT bool equal(const StringImpl*, const LChar*, unsigned);
 470 WTF_EXPORT bool equal(const StringImpl*, const UChar*, unsigned);
 471 inline bool equal(const StringImpl* a, const char* b, unsigned length) { return equal(a, reinterpret_cast<const LChar*>(b), length); }
 472 inline bool equal(const LChar* a, StringImpl* b) { return equal(b, a); }
 473 inline bool equal(const char* a, StringImpl* b) { return equal(b, reinterpret_cast<const LChar*>(a)); }
 474 WTF_EXPORT bool equalNonNull(const StringImpl* a, const StringImpl* b);
 475
 476 template<typename CharType>
 477 ALWAYS_INLINE bool equal(const CharType* a, const CharType* b, unsigned length) { return !memcmp(a, b, length * sizeof(CharType)); }
 478
 479 ALWAYS_INLINE bool equal(const LChar* a, const UChar* b, unsigned length)
 480 {
 481     for (unsigned i = 0; i < length; ++i) {
 482         if (a[i] != b[i])
 483             return false;
 484     }
 485     return true;
 486 }
 487
 488 ALWAYS_INLINE bool equal(const UChar* a, const LChar* b, unsigned length) { return equal(b, a, length); }
 489
 490 WTF_EXPORT bool equalIgnoringCase(const StringImpl*, const StringImpl*);
 491 WTF_EXPORT bool equalIgnoringCase(const StringImpl*, const LChar*);
 492 inline bool equalIgnoringCase(const LChar* a, const StringImpl* b) { return equalIgnoringCase(b, a); }
 493 WTF_EXPORT bool equalIgnoringCase(const LChar*, const LChar*, unsigned);
 494 WTF_EXPORT bool equalIgnoringCase(const UChar*, const LChar*, unsigned);
 495 inline bool equalIgnoringCase(const UChar* a, const char* b, unsigned length) { return equalIgnoringCase(a, reinterpret_cast<const LChar*>(b), length); }
 496 inline bool equalIgnoringCase(const LChar* a, const UChar* b, unsigned length) { return equalIgnoringCase(b, a, length); }
 497 inline bool equalIgnoringCase(const char* a, const UChar* b, unsigned length) { return equalIgnoringCase(b, reinterpret_cast<const LChar*>(a), length); }
 498 inline bool equalIgnoringCase(const char* a, const LChar* b, unsigned length) { return equalIgnoringCase(b, reinterpret_cast<const LChar*>(a), length); }
 499 inline bool equalIgnoringCase(const UChar* a, const UChar* b, int length)
 500 {
 501     ASSERT(length >= 0);
 502     return !Unicode::umemcasecmp(a, b, length);
 503 }
 504 WTF_EXPORT bool equalIgnoringCaseNonNull(const StringImpl*, const StringImpl*);
 505
 506 WTF_EXPORT bool equalIgnoringNullity(StringImpl*, StringImpl*);
 507
 508 template<typename CharacterTypeA, typename CharacterTypeB>
 509 inline bool equalIgnoringASCIICase(const CharacterTypeA* a, const CharacterTypeB* b, unsigned length)
 510 {
 511     for (unsigned i = 0; i < length; ++i) {
 512         if (toASCIILower(a[i]) != toASCIILower(b[i]))
 513             return false;
 514     }
 515     return true;
 516 }
 517
 518 template<typename CharacterTypeA, typename CharacterTypeB>
 519 bool startsWithIgnoringASCIICase(const CharacterTypeA& reference, const CharacterTypeB& prefix)
 520 {
 521     unsigned prefixLength = prefix.length();
 522     if (prefixLength > reference.length())
 523         return false;
 524
 525     if (reference.is8Bit()) {
 526         if (prefix.is8Bit())
 527             return equalIgnoringASCIICase(reference.characters8(), prefix.characters8(), prefixLength);
 528         return equalIgnoringASCIICase(reference.characters8(), prefix.characters16(), prefixLength);
 529     }
 530     if (prefix.is8Bit())
 531         return equalIgnoringASCIICase(reference.characters16(), prefix.characters8(), prefixLength);
 532     return equalIgnoringASCIICase(reference.characters16(), prefix.characters16(), prefixLength);
 533 }
 534
 535 template<typename CharacterType>
 536 inline size_t find(const CharacterType* characters, unsigned length, CharacterType matchCharacter, unsigned index = 0)
 537 {
 538     while (index < length) {
 539         if (characters[index] == matchCharacter)
 540             return index;
 541         ++index;
 542     }
 543     return kNotFound;
 544 }
 545
 546 ALWAYS_INLINE size_t find(const UChar* characters, unsigned length, LChar matchCharacter, unsigned index = 0)
 547 {
 548     return find(characters, length, static_cast<UChar>(matchCharacter), index);
 549 }
 550
 551 inline size_t find(const LChar* characters, unsigned length, UChar matchCharacter, unsigned index = 0)
 552 {
 553     if (matchCharacter & ~0xFF)
 554         return kNotFound;
 555     return find(characters, length, static_cast<LChar>(matchCharacter), index);
 556 }
 557
 558 inline size_t find(const LChar* characters, unsigned length, CharacterMatchFunctionPtr matchFunction, unsigned index = 0)
 559 {
 560     while (index < length) {
 561         if (matchFunction(characters[index]))
 562             return index;
 563         ++index;
 564     }
 565     return kNotFound;
 566 }
 567
 568 inline size_t find(const UChar* characters, unsigned length, CharacterMatchFunctionPtr matchFunction, unsigned index = 0)
 569 {
 570     while (index < length) {
 571         if (matchFunction(characters[index]))
 572             return index;
 573         ++index;
 574     }
 575     return kNotFound;
 576 }
 577
 578 template<typename CharacterType>
 579 inline size_t findNextLineStart(const CharacterType* characters, unsigned length, unsigned index = 0)
 580 {
 581     while (index < length) {
 582         CharacterType c = characters[index++];
 583         if ((c != '\n') && (c != '\r'))
 584             continue;
 585
 586         // There can only be a start of a new line if there are more characters
 587         // beyond the current character.
 588         if (index < length) {
 589             // The 3 common types of line terminators are 1. \r\n (Windows),
 590             // 2. \r (old MacOS) and 3. \n (Unix'es).
 591
 592             if (c == '\n')
 593                 return index; // Case 3: just \n.
 594
 595             CharacterType c2 = characters[index];
 596             if (c2 != '\n')
 597                 return index; // Case 2: just \r.
 598
 599             // Case 1: \r\n.
 600             // But, there's only a start of a new line if there are more
 601             // characters beyond the \r\n.
 602             if (++index < length)
 603                 return index;
 604         }
 605     }
 606     return kNotFound;
 607 }
 608
 609 template<typename CharacterType>
 610 inline size_t reverseFindLineTerminator(const CharacterType* characters, unsigned length, unsigned index = UINT_MAX)
 611 {
 612     if (!length)
 613         return kNotFound;
 614     if (index >= length)
 615         index = length - 1;
 616     CharacterType c = characters[index];
 617     while ((c != '\n') && (c != '\r')) {
 618         if (!index--)
 619             return kNotFound;
 620         c = characters[index];
 621     }
 622     return index;
 623 }
 624
 625 template<typename CharacterType>
 626 inline size_t reverseFind(const CharacterType* characters, unsigned length, CharacterType matchCharacter, unsigned index = UINT_MAX)
 627 {
 628     if (!length)
 629         return kNotFound;
 630     if (index >= length)
 631         index = length - 1;
 632     while (characters[index] != matchCharacter) {
 633         if (!index--)
 634             return kNotFound;
 635     }
 636     return index;
 637 }
 638
 639 ALWAYS_INLINE size_t reverseFind(const UChar* characters, unsigned length, LChar matchCharacter, unsigned index = UINT_MAX)
 640 {
 641     return reverseFind(characters, length, static_cast<UChar>(matchCharacter), index);
 642 }
 643
 644 inline size_t reverseFind(const LChar* characters, unsigned length, UChar matchCharacter, unsigned index = UINT_MAX)
 645 {
 646     if (matchCharacter & ~0xFF)
 647         return kNotFound;
 648     return reverseFind(characters, length, static_cast<LChar>(matchCharacter), index);
 649 }
 650
 651 inline size_t StringImpl::find(LChar character, unsigned start)
 652 {
 653     if (is8Bit())
 654         return WTF::find(characters8(), m_length, character, start);
 655     return WTF::find(characters16(), m_length, character, start);
 656 }
 657
 658 ALWAYS_INLINE size_t StringImpl::find(char character, unsigned start)
 659 {
 660     return find(static_cast<LChar>(character), start);
 661 }
 662
 663 inline size_t StringImpl::find(UChar character, unsigned start)
 664 {
 665     if (is8Bit())
 666         return WTF::find(characters8(), m_length, character, start);
 667     return WTF::find(characters16(), m_length, character, start);
 668 }
 669
 670 inline unsigned lengthOfNullTerminatedString(const UChar* string)
 671 {
 672     size_t length = 0;
 673     while (string[length] != UChar(0))
 674         ++length;
 675     RELEASE_ASSERT(length <= std::numeric_limits<unsigned>::max());
 676     return static_cast<unsigned>(length);
 677 }
 678
 679 template<size_t inlineCapacity>
 680 bool equalIgnoringNullity(const Vector<UChar, inlineCapacity>& a, StringImpl* b)
 681 {
 682     if (!b)
 683         return !a.size();
 684     if (a.size() != b->length())
 685         return false;
 686     if (b->is8Bit())
 687         return equal(a.data(), b->characters8(), b->length());
 688     return equal(a.data(), b->characters16(), b->length());
 689 }
 690
 691 template<typename CharacterType1, typename CharacterType2>
 692 static inline int codePointCompare(unsigned l1, unsigned l2, const CharacterType1* c1, const CharacterType2* c2)
 693 {
 694     const unsigned lmin = l1 < l2 ? l1 : l2;
 695     unsigned pos = 0;
 696     while (pos < lmin && *c1 == *c2) {
 697         ++c1;
 698         ++c2;
 699         ++pos;
 700     }
 701
 702     if (pos < lmin)
 703         return (c1[0] > c2[0]) ? 1 : -1;
 704
 705     if (l1 == l2)
 706         return 0;
 707
 708     return (l1 > l2) ? 1 : -1;
 709 }
 710
 711 static inline int codePointCompare8(const StringImpl* string1, const StringImpl* string2)
 712 {
 713     return codePointCompare(string1->length(), string2->length(), string1->characters8(), string2->characters8());
 714 }
 715
 716 static inline int codePointCompare16(const StringImpl* string1, const StringImpl* string2)
 717 {
 718     return codePointCompare(string1->length(), string2->length(), string1->characters16(), string2->characters16());
 719 }
 720
 721 static inline int codePointCompare8To16(const StringImpl* string1, const StringImpl* string2)
 722 {
 723     return codePointCompare(string1->length(), string2->length(), string1->characters8(), string2->characters16());
 724 }
 725
 726 static inline int codePointCompare(const StringImpl* string1, const StringImpl* string2)
 727 {
 728     if (!string1)
 729         return (string2 && string2->length()) ? -1 : 0;
 730
 731     if (!string2)
 732         return string1->length() ? 1 : 0;
 733
 734     bool string1Is8Bit = string1->is8Bit();
 735     bool string2Is8Bit = string2->is8Bit();
 736     if (string1Is8Bit) {
 737         if (string2Is8Bit)
 738             return codePointCompare8(string1, string2);
 739         return codePointCompare8To16(string1, string2);
 740     }
 741     if (string2Is8Bit)
 742         return -codePointCompare8To16(string2, string1);
 743     return codePointCompare16(string1, string2);
 744 }
 745
 746 static inline bool isSpaceOrNewline(UChar c)
 747 {
 748     // Use isASCIISpace() for basic Latin-1.
 749     // This will include newlines, which aren't included in Unicode DirWS.
 750     return c <= 0x7F ? WTF::isASCIISpace(c) : WTF::Unicode::direction(c) == WTF::Unicode::WhiteSpaceNeutral;
 751 }
 752
 753 inline PassRefPtr<StringImpl> StringImpl::isolatedCopy() const
 754 {
 755     if (is8Bit())
 756         return create(characters8(), m_length);
 757     return create(characters16(), m_length);
 758 }
 759
 760 // TODO(rob.buis) possibly find a better place for this method.
 761 // Turns a UChar32 to uppercase based on localeIdentifier.
 762 WTF_EXPORT UChar32 toUpper(UChar32, const AtomicString& localeIdentifier);
 763
 764 struct StringHash;
 765
 766 // StringHash is the default hash for StringImpl* and RefPtr<StringImpl>
 767 template<typename T> struct DefaultHash;
 768 template<> struct DefaultHash<StringImpl*> {
 769     typedef StringHash Hash;
 770 };
 771 template<> struct DefaultHash<RefPtr<StringImpl>> {
 772     typedef StringHash Hash;
 773 };
 774
 775 }
 776
 777 using WTF::StringImpl;
 778 using WTF::equal;
 779 using WTF::equalNonNull;
 780 using WTF::TextCaseSensitivity;
 781 using WTF::TextCaseSensitive;
 782 using WTF::TextCaseInsensitive;
 783
 784 #endif