Function ParseBoxFileStr

deps/tesseract/ccstruct/boxread.cpp:166–223

Parses the given box file string into a page_number, utf8_str, and bounding_box. Returns true on a successful parse. The box file is assumed to contain box definitions, one per line, of the following format for blob-level boxes: and for word/line-level boxes: WordStr # See applyybox

//   WordStr <left> <bottom> <right> <top> <page id> #<space-delimited word str>
// See applyybox.cpp for more information.
bool ParseBoxFileStr(const char* boxfile_str, int* page_number,
                     STRING* utf8_str, TBOX* bounding_box) {
  *bounding_box = TBOX();       // Initialize it to empty.
  *utf8_str = "";
  char uch[kBoxReadBufSize];
  const char *buffptr = boxfile_str;
  // Read the unichar without messing up on Tibetan.
  // According to issue 253 the utf-8 surrogates 85 and A0 are treated
  // as whitespace by sscanf, so it is more reliable to just find
  // ascii space and tab.
  int uch_len = 0;
  // Skip unicode file designation, if present.
  const unsigned char *ubuf = reinterpret_cast<const unsigned char*>(buffptr);
  if (ubuf[0] == 0xef && ubuf[1] == 0xbb && ubuf[2] == 0xbf)
      buffptr += 3;
  // Allow a single blank as the UTF-8 string. Check for empty string and
  // then blindly eat the first character.
  if (*buffptr == '\0') return false;
  do {
    uch[uch_len++] = *buffptr++;
  } while (*buffptr != '\0' && *buffptr != ' ' && *buffptr != '\t' &&
           uch_len < kBoxReadBufSize - 1);
  uch[uch_len] = '\0';
  if (*buffptr != '\0') ++buffptr;
  int x_min, y_min, x_max, y_max;
  *page_number = 0;
  int count = sscanf(buffptr, "%d %d %d %d %d",
                 &x_min, &y_min, &x_max, &y_max, page_number);
  if (count != 5 && count != 4) {
    tprintf("Bad box coordinates in boxfile string! %s\n", ubuf);
    return false;
  }
  // Test for long space-delimited string label.
  if (strcmp(uch, kMultiBlobLabelCode) == 0 &&
      (buffptr = strchr(buffptr, '#')) != NULL) {
    strncpy(uch, buffptr + 1, kBoxReadBufSize - 1);
    uch[kBoxReadBufSize - 1] = '\0';  // Prevent buffer overrun.
    chomp_string(uch);
    uch_len = strlen(uch);
  }
  // Validate UTF8 by making unichars with it.
  int used = 0;
  while (used < uch_len) {
    UNICHAR ch(uch + used, uch_len - used);
    int new_used = ch.utf8_len();
    if (new_used == 0) {
      tprintf("Bad UTF-8 str %s starts with 0x%02x at col %d\n",
              uch + used, uch[used], used + 1);
      return false;
    }
    used += new_used;
  }
  *utf8_str = uch;
  if (x_min > x_max) Swap(&x_min, &x_max);
  if (y_min > y_max) Swap(&y_min, &y_max);
  bounding_box->set_to_given_coords(x_min, y_min, x_max, y_max);
  return true;  // Successfully read a box.
}