string.cpp

Go to the documentation of this file.

/*
 * This file is part of OpenTTD.
 * OpenTTD is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, version 2.
 * OpenTTD is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 * See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with OpenTTD. If not, see <http://www.gnu.org/licenses/>.
 */
 
#include "stdafx.h"
#include "debug.h"
#include "core/math_func.hpp"
#include "error_func.h"
#include "string_func.h"
#include "string_base.h"
 
#include "table/control_codes.h"
 
#include <sstream>
#include <iomanip>
 
#ifdef _MSC_VER
# define strncasecmp strnicmp
#endif
 
#ifdef _WIN32
# include "os/windows/win32.h"
#endif
 
#ifdef WITH_UNISCRIBE
# include "os/windows/string_uniscribe.h"
#endif
 
#ifdef WITH_ICU_I18N
/* Required by StrNaturalCompare. */
# include <unicode/ustring.h>
# include "language.h"
# include "gfx_func.h"
#endif /* WITH_ICU_I18N */
 
#if defined(WITH_COCOA)
# include "os/macosx/string_osx.h"
#endif
 
#include "safeguards.h"
 
 
void strecpy(std::span<char> dst, std::string_view src)
{
  /* Ensure source string fits with NUL terminator; dst must be at least 1 character longer than src. */
  if (std::empty(dst) || std::size(src) >= std::size(dst) - 1U) {
#if defined(STRGEN) || defined(SETTINGSGEN)
    FatalError("String too long for destination buffer");
#else /* STRGEN || SETTINGSGEN */
    Debug(misc, 0, "String too long for destination buffer");
    src = src.substr(0, std::size(dst) - 1U);
#endif /* STRGEN || SETTINGSGEN */
  }
 
  auto it = std::copy(std::begin(src), std::end(src), std::begin(dst));
  *it = '\0';
}
 
std::string FormatArrayAsHex(std::span<const uint8_t> data)
{
  std::string str;
  str.reserve(data.size() * 2 + 1);
 
  for (auto b : data) {
    fmt::format_to(std::back_inserter(str), "{:02X}", b);
  }
 
  return str;
}
 
static bool IsSccEncodedCode(char32_t c)
{
  switch (c) {
    case SCC_RECORD_SEPARATOR:
    case SCC_ENCODED:
    case SCC_ENCODED_INTERNAL:
    case SCC_ENCODED_NUMERIC:
    case SCC_ENCODED_STRING:
      return true;
 
    default:
      return false;
  }
}
 
template <class T>
static void StrMakeValid(T &dst, const char *str, const char *last, StringValidationSettings settings)
{
  /* Assume the ABSOLUTE WORST to be in str as it comes from the outside. */
 
  while (str <= last && *str != '\0') {
    size_t len = Utf8EncodedCharLen(*str);
    char32_t c;
    /* If the first byte does not look like the first byte of an encoded
     * character, i.e. encoded length is 0, then this byte is definitely bad
     * and it should be skipped.
     * When the first byte looks like the first byte of an encoded character,
     * then the remaining bytes in the string are checked whether the whole
     * encoded character can be there. If that is not the case, this byte is
     * skipped.
     * Finally we attempt to decode the encoded character, which does certain
     * extra validations to see whether the correct number of bytes were used
     * to encode the character. If that is not the case, the byte is probably
     * invalid and it is skipped. We could emit a question mark, but then the
     * logic below cannot just copy bytes, it would need to re-encode the
     * decoded characters as the length in bytes may have changed.
     *
     * The goals here is to get as much valid Utf8 encoded characters from the
     * source string to the destination string.
     *
     * Note: a multi-byte encoded termination ('\0') will trigger the encoded
     * char length and the decoded length to differ, so it will be ignored as
     * invalid character data. If it were to reach the termination, then we
     * would also reach the "last" byte of the string and a normal '\0'
     * termination will be placed after it.
     */
    if (len == 0 || str + len > last + 1 || len != Utf8Decode(&c, str)) {
      /* Maybe the next byte is still a valid character? */
      str++;
      continue;
    }
 
    if ((IsPrintable(c) && (c < SCC_SPRITE_START || c > SCC_SPRITE_END)) || ((settings & SVS_ALLOW_CONTROL_CODE) != 0 && IsSccEncodedCode(c))) {
      /* Copy the character back. Even if dst is current the same as str
       * (i.e. no characters have been changed) this is quicker than
       * moving the pointers ahead by len */
      do {
        *dst++ = *str++;
      } while (--len != 0);
    } else if ((settings & SVS_ALLOW_NEWLINE) != 0 && c == '\n') {
      *dst++ = *str++;
    } else {
      if ((settings & SVS_ALLOW_NEWLINE) != 0 && c == '\r' && str[1] == '\n') {
        str += len;
        continue;
      }
      str += len;
      if ((settings & SVS_REPLACE_TAB_CR_NL_WITH_SPACE) != 0 && (c == '\r' || c == '\n' || c == '\t')) {
        /* Replace the tab, carriage return or newline with a space. */
        *dst++ = ' ';
      } else if ((settings & SVS_REPLACE_WITH_QUESTION_MARK) != 0) {
        /* Replace the undesirable character with a question mark */
        *dst++ = '?';
      }
    }
  }
 
  /* String termination, if needed, is left to the caller of this function. */
}
 
void StrMakeValidInPlace(char *str, const char *last, StringValidationSettings settings)
{
  char *dst = str;
  StrMakeValid(dst, str, last, settings);
  *dst = '\0';
}
 
void StrMakeValidInPlace(char *str, StringValidationSettings settings)
{
  /* We know it is '\0' terminated. */
  StrMakeValidInPlace(str, str + strlen(str), settings);
}
 
std::string StrMakeValid(std::string_view str, StringValidationSettings settings)
{
  if (str.empty()) return {};
 
  auto buf = str.data();
  auto last = buf + str.size() - 1;
 
  std::ostringstream dst;
  std::ostreambuf_iterator<char> dst_iter(dst);
  StrMakeValid(dst_iter, buf, last, settings);
 
  return dst.str();
}
 
bool StrValid(std::span<const char> str)
{
  /* Assume the ABSOLUTE WORST to be in str as it comes from the outside. */
  auto it = std::begin(str);
  auto last = std::prev(std::end(str));
 
  while (it <= last && *it != '\0') {
    size_t len = Utf8EncodedCharLen(*it);
    /* Encoded length is 0 if the character isn't known.
     * The length check is needed to prevent Utf8Decode to read
     * over the terminating '\0' if that happens to be placed
     * within the encoding of an UTF8 character. */
    if (len == 0 || it + len > last) return false;
 
    char32_t c;
    len = Utf8Decode(&c, &*it);
    if (!IsPrintable(c) || (c >= SCC_SPRITE_START && c <= SCC_SPRITE_END)) {
      return false;
    }
 
    it += len;
  }
 
  return *it == '\0';
}
 
void StrTrimInPlace(std::string &str)
{
  str = StrTrimView(str);
}
 
std::string_view StrTrimView(std::string_view str)
{
  size_t first_pos = str.find_first_not_of(' ');
  if (first_pos == std::string::npos) {
    return std::string_view{};
  }
  size_t last_pos = str.find_last_not_of(' ');
  return str.substr(first_pos, last_pos - first_pos + 1);
}
 
bool StrStartsWithIgnoreCase(std::string_view str, const std::string_view prefix)
{
  if (str.size() < prefix.size()) return false;
  return StrEqualsIgnoreCase(str.substr(0, prefix.size()), prefix);
}
 
struct CaseInsensitiveCharTraits : public std::char_traits<char> {
  static bool eq(char c1, char c2) { return toupper(c1) == toupper(c2); }
  static bool ne(char c1, char c2) { return toupper(c1) != toupper(c2); }
  static bool lt(char c1, char c2) { return toupper(c1) <  toupper(c2); }
 
  static int compare(const char *s1, const char *s2, size_t n)
  {
    while (n-- != 0) {
      if (toupper(*s1) < toupper(*s2)) return -1;
      if (toupper(*s1) > toupper(*s2)) return 1;
      ++s1; ++s2;
    }
    return 0;
  }
 
  static const char *find(const char *s, size_t n, char a)
  {
    for (; n > 0; --n, ++s) {
      if (toupper(*s) == toupper(a)) return s;
    }
    return nullptr;
  }
};
 
typedef std::basic_string_view<char, CaseInsensitiveCharTraits> CaseInsensitiveStringView;
 
bool StrEndsWithIgnoreCase(std::string_view str, const std::string_view suffix)
{
  if (str.size() < suffix.size()) return false;
  return StrEqualsIgnoreCase(str.substr(str.size() - suffix.size()), suffix);
}
 
int StrCompareIgnoreCase(const std::string_view str1, const std::string_view str2)
{
  CaseInsensitiveStringView ci_str1{ str1.data(), str1.size() };
  CaseInsensitiveStringView ci_str2{ str2.data(), str2.size() };
  return ci_str1.compare(ci_str2);
}
 
bool StrEqualsIgnoreCase(const std::string_view str1, const std::string_view str2)
{
  if (str1.size() != str2.size()) return false;
  return StrCompareIgnoreCase(str1, str2) == 0;
}
 
size_t Utf8StringLength(const char *s)
{
  size_t len = 0;
  const char *t = s;
  while (Utf8Consume(&t) != 0) len++;
  return len;
}
 
size_t Utf8StringLength(const std::string &str)
{
  return Utf8StringLength(str.c_str());
}
 
bool strtolower(std::string &str, std::string::size_type offs)
{
  bool changed = false;
  for (auto ch = str.begin() + offs; ch != str.end(); ++ch) {
    auto new_ch = static_cast<char>(tolower(static_cast<unsigned char>(*ch)));
    changed |= new_ch != *ch;
    *ch = new_ch;
  }
  return changed;
}
 
bool IsValidChar(char32_t key, CharSetFilter afilter)
{
  switch (afilter) {
    case CS_ALPHANUMERAL:   return IsPrintable(key);
    case CS_NUMERAL:        return (key >= '0' && key <= '9');
    case CS_NUMERAL_SPACE:  return (key >= '0' && key <= '9') || key == ' ';
    case CS_NUMERAL_SIGNED: return (key >= '0' && key <= '9') || key == '-';
    case CS_ALPHA:          return IsPrintable(key) && !(key >= '0' && key <= '9');
    case CS_HEXADECIMAL:    return (key >= '0' && key <= '9') || (key >= 'a' && key <= 'f') || (key >= 'A' && key <= 'F');
    default: NOT_REACHED();
  }
}
 
 
/* UTF-8 handling routines */
 
 
size_t Utf8Decode(char32_t *c, const char *s)
{
  assert(c != nullptr);
 
  if (!HasBit(s[0], 7)) {
    /* Single byte character: 0xxxxxxx */
    *c = s[0];
    return 1;
  } else if (GB(s[0], 5, 3) == 6) {
    if (IsUtf8Part(s[1])) {
      /* Double byte character: 110xxxxx 10xxxxxx */
      *c = GB(s[0], 0, 5) << 6 | GB(s[1], 0, 6);
      if (*c >= 0x80) return 2;
    }
  } else if (GB(s[0], 4, 4) == 14) {
    if (IsUtf8Part(s[1]) && IsUtf8Part(s[2])) {
      /* Triple byte character: 1110xxxx 10xxxxxx 10xxxxxx */
      *c = GB(s[0], 0, 4) << 12 | GB(s[1], 0, 6) << 6 | GB(s[2], 0, 6);
      if (*c >= 0x800) return 3;
    }
  } else if (GB(s[0], 3, 5) == 30) {
    if (IsUtf8Part(s[1]) && IsUtf8Part(s[2]) && IsUtf8Part(s[3])) {
      /* 4 byte character: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx */
      *c = GB(s[0], 0, 3) << 18 | GB(s[1], 0, 6) << 12 | GB(s[2], 0, 6) << 6 | GB(s[3], 0, 6);
      if (*c >= 0x10000 && *c <= 0x10FFFF) return 4;
    }
  }
 
  *c = '?';
  return 1;
}
 
 
template <class T>
inline size_t Utf8Encode(T buf, char32_t c)
{
  if (c < 0x80) {
    *buf = c;
    return 1;
  } else if (c < 0x800) {
    *buf++ = 0xC0 + GB(c,  6, 5);
    *buf   = 0x80 + GB(c,  0, 6);
    return 2;
  } else if (c < 0x10000) {
    *buf++ = 0xE0 + GB(c, 12, 4);
    *buf++ = 0x80 + GB(c,  6, 6);
    *buf   = 0x80 + GB(c,  0, 6);
    return 3;
  } else if (c < 0x110000) {
    *buf++ = 0xF0 + GB(c, 18, 3);
    *buf++ = 0x80 + GB(c, 12, 6);
    *buf++ = 0x80 + GB(c,  6, 6);
    *buf   = 0x80 + GB(c,  0, 6);
    return 4;
  }
 
  *buf = '?';
  return 1;
}
 
size_t Utf8Encode(char *buf, char32_t c)
{
  return Utf8Encode<char *>(buf, c);
}
 
size_t Utf8Encode(std::ostreambuf_iterator<char> &buf, char32_t c)
{
  return Utf8Encode<std::ostreambuf_iterator<char> &>(buf, c);
}
 
size_t Utf8Encode(std::back_insert_iterator<std::string> &buf, char32_t c)
{
  return Utf8Encode<std::back_insert_iterator<std::string> &>(buf, c);
}
 
size_t Utf8TrimString(char *s, size_t maxlen)
{
  size_t length = 0;
 
  for (const char *ptr = strchr(s, '\0'); *s != '\0';) {
    size_t len = Utf8EncodedCharLen(*s);
    /* Silently ignore invalid UTF8 sequences, our only concern trimming */
    if (len == 0) len = 1;
 
    /* Take care when a hard cutoff was made for the string and
     * the last UTF8 sequence is invalid */
    if (length + len >= maxlen || (s + len > ptr)) break;
    s += len;
    length += len;
  }
 
  *s = '\0';
  return length;
}
 
#ifdef DEFINE_STRCASESTR
char *strcasestr(const char *haystack, const char *needle)
{
  size_t hay_len = strlen(haystack);
  size_t needle_len = strlen(needle);
  while (hay_len >= needle_len) {
    if (strncasecmp(haystack, needle, needle_len) == 0) return const_cast<char *>(haystack);
 
    haystack++;
    hay_len--;
  }
 
  return nullptr;
}
#endif /* DEFINE_STRCASESTR */
 
static bool IsGarbageCharacter(char32_t c)
{
  if (c >= '0' && c <= '9') return false;
  if (c >= 'A' && c <= 'Z') return false;
  if (c >= 'a' && c <= 'z') return false;
  if (c >= SCC_CONTROL_START && c <= SCC_CONTROL_END) return true;
  if (c >= 0xC0 && c <= 0x10FFFF) return false;
 
  return true;
}
 
static std::string_view SkipGarbage(std::string_view str)
{
  auto first = std::begin(str);
  auto last = std::end(str);
  while (first < last) {
    char32_t c;
    size_t len = Utf8Decode(&c, &*first);
    if (!IsGarbageCharacter(c)) break;
    first += len;
  }
  return {first, last};
}
 
int StrNaturalCompare(std::string_view s1, std::string_view s2, bool ignore_garbage_at_front)
{
  if (ignore_garbage_at_front) {
    s1 = SkipGarbage(s1);
    s2 = SkipGarbage(s2);
  }
 
#ifdef WITH_ICU_I18N
  if (_current_collator) {
    UErrorCode status = U_ZERO_ERROR;
    int result = _current_collator->compareUTF8(icu::StringPiece(s1.data(), s1.size()), icu::StringPiece(s2.data(), s2.size()), status);
    if (U_SUCCESS(status)) return result;
  }
#endif /* WITH_ICU_I18N */
 
#if defined(_WIN32) && !defined(STRGEN) && !defined(SETTINGSGEN)
  int res = OTTDStringCompare(s1, s2);
  if (res != 0) return res - 2; // Convert to normal C return values.
#endif
 
#if defined(WITH_COCOA) && !defined(STRGEN) && !defined(SETTINGSGEN)
  int res = MacOSStringCompare(s1, s2);
  if (res != 0) return res - 2; // Convert to normal C return values.
#endif
 
  /* Do a normal comparison if ICU is missing or if we cannot create a collator. */
  return StrCompareIgnoreCase(s1, s2);
}
 
#ifdef WITH_ICU_I18N
 
#include <unicode/stsearch.h>
 
static int ICUStringContains(const std::string_view str, const std::string_view value, bool case_insensitive)
{
  if (_current_collator) {
    std::unique_ptr<icu::RuleBasedCollator> coll(dynamic_cast<icu::RuleBasedCollator *>(_current_collator->clone()));
    if (coll) {
      UErrorCode status = U_ZERO_ERROR;
      coll->setStrength(case_insensitive ? icu::Collator::SECONDARY : icu::Collator::TERTIARY);
      coll->setAttribute(UCOL_NUMERIC_COLLATION, UCOL_OFF, status);
 
      auto u_str = icu::UnicodeString::fromUTF8(icu::StringPiece(str.data(), str.size()));
      auto u_value = icu::UnicodeString::fromUTF8(icu::StringPiece(value.data(), value.size()));
      icu::StringSearch u_searcher(u_value, u_str, coll.get(), nullptr, status);
      if (U_SUCCESS(status)) {
        auto pos = u_searcher.first(status);
        if (U_SUCCESS(status)) return pos != USEARCH_DONE ? 1 : 0;
      }
    }
  }
 
  return -1;
}
#endif /* WITH_ICU_I18N */
 
[[nodiscard]] bool StrNaturalContains(const std::string_view str, const std::string_view value)
{
#ifdef WITH_ICU_I18N
  int res_u = ICUStringContains(str, value, false);
  if (res_u >= 0) return res_u > 0;
#endif /* WITH_ICU_I18N */
 
#if defined(_WIN32) && !defined(STRGEN) && !defined(SETTINGSGEN)
  int res = Win32StringContains(str, value, false);
  if (res >= 0) return res > 0;
#endif
 
#if defined(WITH_COCOA) && !defined(STRGEN) && !defined(SETTINGSGEN)
  int res = MacOSStringContains(str, value, false);
  if (res >= 0) return res > 0;
#endif
 
  return str.find(value) != std::string_view::npos;
}
 
[[nodiscard]] bool StrNaturalContainsIgnoreCase(const std::string_view str, const std::string_view value)
{
#ifdef WITH_ICU_I18N
  int res_u = ICUStringContains(str, value, true);
  if (res_u >= 0) return res_u > 0;
#endif /* WITH_ICU_I18N */
 
#if defined(_WIN32) && !defined(STRGEN) && !defined(SETTINGSGEN)
  int res = Win32StringContains(str, value, true);
  if (res >= 0) return res > 0;
#endif
 
#if defined(WITH_COCOA) && !defined(STRGEN) && !defined(SETTINGSGEN)
  int res = MacOSStringContains(str, value, true);
  if (res >= 0) return res > 0;
#endif
 
  CaseInsensitiveStringView ci_str{ str.data(), str.size() };
  CaseInsensitiveStringView ci_value{ value.data(), value.size() };
  return ci_str.find(ci_value) != CaseInsensitiveStringView::npos;
}
 
static int ConvertHexNibbleToByte(char c)
{
  if (c >= '0' && c <= '9') return c - '0';
  if (c >= 'A' && c <= 'F') return c + 10 - 'A';
  if (c >= 'a' && c <= 'f') return c + 10 - 'a';
  return -1;
}
 
bool ConvertHexToBytes(std::string_view hex, std::span<uint8_t> bytes)
{
  if (bytes.size() != hex.size() / 2) {
    return false;
  }
 
  /* Hex-string lengths are always divisible by 2. */
  if (hex.size() % 2 != 0) {
    return false;
  }
 
  for (size_t i = 0; i < hex.size() / 2; i++) {
    auto hi = ConvertHexNibbleToByte(hex[i * 2]);
    auto lo = ConvertHexNibbleToByte(hex[i * 2 + 1]);
 
    if (hi < 0 || lo < 0) {
      return false;
    }
 
    bytes[i] = (hi << 4) | lo;
  }
 
  return true;
}
 
#ifdef WITH_UNISCRIBE
 
/* static */ std::unique_ptr<StringIterator> StringIterator::Create()
{
  return std::make_unique<UniscribeStringIterator>();
}
 
#elif defined(WITH_ICU_I18N)
 
#include <unicode/utext.h>
#include <unicode/brkiter.h>
 
class IcuStringIterator : public StringIterator
{
  icu::BreakIterator *char_itr; 
  icu::BreakIterator *word_itr; 
 
  std::vector<UChar> utf16_str;      
  std::vector<size_t> utf16_to_utf8; 
 
public:
  IcuStringIterator() : char_itr(nullptr), word_itr(nullptr)
  {
    UErrorCode status = U_ZERO_ERROR;
    this->char_itr = icu::BreakIterator::createCharacterInstance(icu::Locale(_current_language != nullptr ? _current_language->isocode : "en"), status);
    this->word_itr = icu::BreakIterator::createWordInstance(icu::Locale(_current_language != nullptr ? _current_language->isocode : "en"), status);
 
    this->utf16_str.push_back('\0');
    this->utf16_to_utf8.push_back(0);
  }
 
  ~IcuStringIterator() override
  {
    delete this->char_itr;
    delete this->word_itr;
  }
 
  void SetString(const char *s) override
  {
    const char *string_base = s;
 
    /* Unfortunately current ICU versions only provide rudimentary support
     * for word break iterators (especially for CJK languages) in combination
     * with UTF-8 input. As a work around we have to convert the input to
     * UTF-16 and create a mapping back to UTF-8 character indices. */
    this->utf16_str.clear();
    this->utf16_to_utf8.clear();
 
    while (*s != '\0') {
      size_t idx = s - string_base;
 
      char32_t c = Utf8Consume(&s);
      if (c < 0x10000) {
        this->utf16_str.push_back((UChar)c);
      } else {
        /* Make a surrogate pair. */
        this->utf16_str.push_back((UChar)(0xD800 + ((c - 0x10000) >> 10)));
        this->utf16_str.push_back((UChar)(0xDC00 + ((c - 0x10000) & 0x3FF)));
        this->utf16_to_utf8.push_back(idx);
      }
      this->utf16_to_utf8.push_back(idx);
    }
    this->utf16_str.push_back('\0');
    this->utf16_to_utf8.push_back(s - string_base);
 
    UText text = UTEXT_INITIALIZER;
    UErrorCode status = U_ZERO_ERROR;
    utext_openUChars(&text, this->utf16_str.data(), this->utf16_str.size() - 1, &status);
    this->char_itr->setText(&text, status);
    this->word_itr->setText(&text, status);
    this->char_itr->first();
    this->word_itr->first();
  }
 
  size_t SetCurPosition(size_t pos) override
  {
    /* Convert incoming position to an UTF-16 string index. */
    uint utf16_pos = 0;
    for (uint i = 0; i < this->utf16_to_utf8.size(); i++) {
      if (this->utf16_to_utf8[i] == pos) {
        utf16_pos = i;
        break;
      }
    }
 
    /* isBoundary has the documented side-effect of setting the current
     * position to the first valid boundary equal to or greater than
     * the passed value. */
    this->char_itr->isBoundary(utf16_pos);
    return this->utf16_to_utf8[this->char_itr->current()];
  }
 
  size_t Next(IterType what) override
  {
    int32_t pos;
    switch (what) {
      case ITER_CHARACTER:
        pos = this->char_itr->next();
        break;
 
      case ITER_WORD:
        pos = this->word_itr->following(this->char_itr->current());
        /* The ICU word iterator considers both the start and the end of a word a valid
         * break point, but we only want word starts. Move to the next location in
         * case the new position points to whitespace. */
        while (pos != icu::BreakIterator::DONE &&
            IsWhitespace(Utf16DecodeChar((const uint16_t *)&this->utf16_str[pos]))) {
          int32_t new_pos = this->word_itr->next();
          /* Don't set it to DONE if it was valid before. Otherwise we'll return END
           * even though the iterator wasn't at the end of the string before. */
          if (new_pos == icu::BreakIterator::DONE) break;
          pos = new_pos;
        }
 
        this->char_itr->isBoundary(pos);
        break;
 
      default:
        NOT_REACHED();
    }
 
    return pos == icu::BreakIterator::DONE ? END : this->utf16_to_utf8[pos];
  }
 
  size_t Prev(IterType what) override
  {
    int32_t pos;
    switch (what) {
      case ITER_CHARACTER:
        pos = this->char_itr->previous();
        break;
 
      case ITER_WORD:
        pos = this->word_itr->preceding(this->char_itr->current());
        /* The ICU word iterator considers both the start and the end of a word a valid
         * break point, but we only want word starts. Move to the previous location in
         * case the new position points to whitespace. */
        while (pos != icu::BreakIterator::DONE &&
            IsWhitespace(Utf16DecodeChar((const uint16_t *)&this->utf16_str[pos]))) {
          int32_t new_pos = this->word_itr->previous();
          /* Don't set it to DONE if it was valid before. Otherwise we'll return END
           * even though the iterator wasn't at the start of the string before. */
          if (new_pos == icu::BreakIterator::DONE) break;
          pos = new_pos;
        }
 
        this->char_itr->isBoundary(pos);
        break;
 
      default:
        NOT_REACHED();
    }
 
    return pos == icu::BreakIterator::DONE ? END : this->utf16_to_utf8[pos];
  }
};
 
/* static */ std::unique_ptr<StringIterator> StringIterator::Create()
{
  return std::make_unique<IcuStringIterator>();
}
 
#else
 
class DefaultStringIterator : public StringIterator
{
  const char *string; 
  size_t len;         
  size_t cur_pos;     
 
public:
  DefaultStringIterator() : string(nullptr), len(0), cur_pos(0)
  {
  }
 
  void SetString(const char *s) override
  {
    this->string = s;
    this->len = strlen(s);
    this->cur_pos = 0;
  }
 
  size_t SetCurPosition(size_t pos) override
  {
    assert(this->string != nullptr && pos <= this->len);
    /* Sanitize in case we get a position inside an UTF-8 sequence. */
    while (pos > 0 && IsUtf8Part(this->string[pos])) pos--;
    return this->cur_pos = pos;
  }
 
  size_t Next(IterType what) override
  {
    assert(this->string != nullptr);
 
    /* Already at the end? */
    if (this->cur_pos >= this->len) return END;
 
    switch (what) {
      case ITER_CHARACTER: {
        char32_t c;
        this->cur_pos += Utf8Decode(&c, this->string + this->cur_pos);
        return this->cur_pos;
      }
 
      case ITER_WORD: {
        char32_t c;
        /* Consume current word. */
        size_t offs = Utf8Decode(&c, this->string + this->cur_pos);
        while (this->cur_pos < this->len && !IsWhitespace(c)) {
          this->cur_pos += offs;
          offs = Utf8Decode(&c, this->string + this->cur_pos);
        }
        /* Consume whitespace to the next word. */
        while (this->cur_pos < this->len && IsWhitespace(c)) {
          this->cur_pos += offs;
          offs = Utf8Decode(&c, this->string + this->cur_pos);
        }
 
        return this->cur_pos;
      }
 
      default:
        NOT_REACHED();
    }
 
    return END;
  }
 
  size_t Prev(IterType what) override
  {
    assert(this->string != nullptr);
 
    /* Already at the beginning? */
    if (this->cur_pos == 0) return END;
 
    switch (what) {
      case ITER_CHARACTER:
        return this->cur_pos = Utf8PrevChar(this->string + this->cur_pos) - this->string;
 
      case ITER_WORD: {
        const char *s = this->string + this->cur_pos;
        char32_t c;
        /* Consume preceding whitespace. */
        do {
          s = Utf8PrevChar(s);
          Utf8Decode(&c, s);
        } while (s > this->string && IsWhitespace(c));
        /* Consume preceding word. */
        while (s > this->string && !IsWhitespace(c)) {
          s = Utf8PrevChar(s);
          Utf8Decode(&c, s);
        }
        /* Move caret back to the beginning of the word. */
        if (IsWhitespace(c)) Utf8Consume(&s);
 
        return this->cur_pos = s - this->string;
      }
 
      default:
        NOT_REACHED();
    }
 
    return END;
  }
};
 
#if defined(WITH_COCOA) && !defined(STRGEN) && !defined(SETTINGSGEN)
/* static */ std::unique_ptr<StringIterator> StringIterator::Create()
{
  std::unique_ptr<StringIterator> i = OSXStringIterator::Create();
  if (i != nullptr) return i;
 
  return std::make_unique<DefaultStringIterator>();
}
#else
/* static */ std::unique_ptr<StringIterator> StringIterator::Create()
{
  return std::make_unique<DefaultStringIterator>();
}
#endif /* defined(WITH_COCOA) && !defined(STRGEN) && !defined(SETTINGSGEN) */
 
#endif