Function UTF8Decode

cpp/src/arrow/util/utf8_internal.h:345–390

}

static inline bool UTF8Decode(const uint8_t** data, uint32_t* codepoint) {
  const uint8_t* str = *data;
  if (*str < 0x80) {  // ascii
    *codepoint = *str++;
  } else if (ARROW_PREDICT_FALSE(*str < 0xC0)) {  // invalid non-ascii char
    return false;
  } else if (*str < 0xE0) {
    uint8_t code_unit_1 = (*str++) & 0x1F;  // take last 5 bits
    if (ARROW_PREDICT_FALSE(!Utf8IsContinuation(*str))) {
      return false;
    }
    uint8_t code_unit_2 = (*str++) & 0x3F;  // take last 6 bits
    *codepoint = (code_unit_1 << 6) + code_unit_2;
  } else if (*str < 0xF0) {
    uint8_t code_unit_1 = (*str++) & 0x0F;  // take last 4 bits
    if (ARROW_PREDICT_FALSE(!Utf8IsContinuation(*str))) {
      return false;
    }
    uint8_t code_unit_2 = (*str++) & 0x3F;  // take last 6 bits
    if (ARROW_PREDICT_FALSE(!Utf8IsContinuation(*str))) {
      return false;
    }
    uint8_t code_unit_3 = (*str++) & 0x3F;  // take last 6 bits
    *codepoint = (code_unit_1 << 12) + (code_unit_2 << 6) + code_unit_3;
  } else if (*str < 0xF8) {
    uint8_t code_unit_1 = (*str++) & 0x07;  // take last 3 bits
    if (ARROW_PREDICT_FALSE(!Utf8IsContinuation(*str))) {
      return false;
    }
    uint8_t code_unit_2 = (*str++) & 0x3F;  // take last 6 bits
    if (ARROW_PREDICT_FALSE(!Utf8IsContinuation(*str))) {
      return false;
    }
    uint8_t code_unit_3 = (*str++) & 0x3F;  // take last 6 bits
    if (ARROW_PREDICT_FALSE(!Utf8IsContinuation(*str))) {
      return false;
    }
    uint8_t code_unit_4 = (*str++) & 0x3F;  // take last 6 bits
    *codepoint =
        (code_unit_1 << 18) + (code_unit_2 << 12) + (code_unit_3 << 6) + code_unit_4;
  } else {  // invalid non-ascii char
    return false;
  }
  *data = str;
  return true;
}

static inline bool UTF8DecodeReverse(const uint8_t** data, uint32_t* codepoint) {
  const uint8_t* str = *data;