tesseract  4.1.1
trie.cpp
Go to the documentation of this file.
1 /* -*-C-*-
2  ********************************************************************************
3  *
4  * File: trie.cpp (Formerly trie.c)
5  * Description: Functions to build a trie data structure.
6  * Author: Mark Seaman, OCR Technology
7  *
8  * (c) Copyright 1987, Hewlett-Packard Company.
9  ** Licensed under the Apache License, Version 2.0 (the "License");
10  ** you may not use this file except in compliance with the License.
11  ** You may obtain a copy of the License at
12  ** http://www.apache.org/licenses/LICENSE-2.0
13  ** Unless required by applicable law or agreed to in writing, software
14  ** distributed under the License is distributed on an "AS IS" BASIS,
15  ** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
16  ** See the License for the specific language governing permissions and
17  ** limitations under the License.
18  *
19  *********************************************************************************/
20 /*----------------------------------------------------------------------
21  I n c l u d e s
22 ----------------------------------------------------------------------*/
23 
24 #include "trie.h"
25 
26 #include "callcpp.h"
27 #include "dawg.h"
28 #include "dict.h"
29 #include "genericvector.h"
30 #include "helpers.h"
31 #include "kdpair.h"
32 
33 namespace tesseract {
34 
35 const char kDoNotReverse[] = "RRP_DO_NO_REVERSE";
36 const char kReverseIfHasRTL[] = "RRP_REVERSE_IF_HAS_RTL";
37 const char kForceReverse[] = "RRP_FORCE_REVERSE";
38 
39 const char * const RTLReversePolicyNames[] = {
43 };
44 
45 const char Trie::kAlphaPatternUnicode[] = "\u2000";
46 const char Trie::kDigitPatternUnicode[] = "\u2001";
47 const char Trie::kAlphanumPatternUnicode[] = "\u2002";
48 const char Trie::kPuncPatternUnicode[] = "\u2003";
49 const char Trie::kLowerPatternUnicode[] = "\u2004";
50 const char Trie::kUpperPatternUnicode[] = "\u2005";
51 
52 const char *Trie::get_reverse_policy_name(RTLReversePolicy reverse_policy) {
53  return RTLReversePolicyNames[reverse_policy];
54 }
55 
56 // Reset the Trie to empty.
57 void Trie::clear() {
59  nodes_.clear();
61  num_edges_ = 0;
62  new_dawg_node(); // Need to allocate node 0.
63 }
64 
65 bool Trie::edge_char_of(NODE_REF node_ref, NODE_REF next_node,
66  int direction, bool word_end, UNICHAR_ID unichar_id,
67  EDGE_RECORD **edge_ptr, EDGE_INDEX *edge_index) const {
68  if (debug_level_ == 3) {
69  tprintf("edge_char_of() given node_ref " REFFORMAT " next_node " REFFORMAT
70  " direction %d word_end %d unichar_id %d, exploring node:\n",
71  node_ref, next_node, direction, word_end, unichar_id);
72  if (node_ref != NO_EDGE) {
73  print_node(node_ref, nodes_[node_ref]->forward_edges.size());
74  }
75  }
76  if (node_ref == NO_EDGE) return false;
77  assert(node_ref < nodes_.size());
78  EDGE_VECTOR &vec = (direction == FORWARD_EDGE) ?
79  nodes_[node_ref]->forward_edges : nodes_[node_ref]->backward_edges;
80  int vec_size = vec.size();
81  if (node_ref == 0 && direction == FORWARD_EDGE) { // binary search
82  EDGE_INDEX start = 0;
83  EDGE_INDEX end = vec_size - 1;
84  EDGE_INDEX k;
85  int compare;
86  while (start <= end) {
87  k = (start + end) >> 1; // (start + end) / 2
88  compare = given_greater_than_edge_rec(next_node, word_end,
89  unichar_id, vec[k]);
90  if (compare == 0) { // given == vec[k]
91  *edge_ptr = &(vec[k]);
92  *edge_index = k;
93  return true;
94  } else if (compare == 1) { // given > vec[k]
95  start = k + 1;
96  } else { // given < vec[k]
97  end = k - 1;
98  }
99  }
100  } else { // linear search
101  for (int i = 0; i < vec_size; ++i) {
102  EDGE_RECORD &edge_rec = vec[i];
103  if (edge_rec_match(next_node, word_end, unichar_id,
104  next_node_from_edge_rec(edge_rec),
105  end_of_word_from_edge_rec(edge_rec),
106  unichar_id_from_edge_rec(edge_rec))) {
107  *edge_ptr = &(edge_rec);
108  *edge_index = i;
109  return true;
110  }
111  }
112  }
113  return false; // not found
114 }
115 
116 bool Trie::add_edge_linkage(NODE_REF node1, NODE_REF node2, bool marker_flag,
117  int direction, bool word_end,
118  UNICHAR_ID unichar_id) {
119  EDGE_VECTOR *vec = (direction == FORWARD_EDGE) ?
120  &(nodes_[node1]->forward_edges) : &(nodes_[node1]->backward_edges);
121  int search_index;
122  if (node1 == 0 && direction == FORWARD_EDGE) {
123  search_index = 0; // find the index to make the add sorted
124  while (search_index < vec->size() &&
125  given_greater_than_edge_rec(node2, word_end, unichar_id,
126  (*vec)[search_index]) == 1) {
127  search_index++;
128  }
129  } else {
130  search_index = vec->size(); // add is unsorted, so index does not matter
131  }
132  EDGE_RECORD edge_rec;
133  link_edge(&edge_rec, node2, marker_flag, direction, word_end, unichar_id);
134  if (node1 == 0 && direction == BACKWARD_EDGE &&
136  EDGE_INDEX edge_index = root_back_freelist_.pop_back();
137  (*vec)[edge_index] = edge_rec;
138  } else if (search_index < vec->size()) {
139  vec->insert(edge_rec, search_index);
140  } else {
141  vec->push_back(edge_rec);
142  }
143  if (debug_level_ > 1) {
144  tprintf("new edge in nodes_[" REFFORMAT "]: ", node1);
145  print_edge_rec(edge_rec);
146  tprintf("\n");
147  }
148  num_edges_++;
149  return true;
150 }
151 
153  NODE_REF the_next_node,
154  bool marker_flag,
155  UNICHAR_ID unichar_id) {
156  EDGE_RECORD *back_edge_ptr;
157  EDGE_INDEX back_edge_index;
158  ASSERT_HOST(edge_char_of(the_next_node, NO_EDGE, BACKWARD_EDGE, false,
159  unichar_id, &back_edge_ptr, &back_edge_index));
160  if (marker_flag) {
161  *back_edge_ptr |= (MARKER_FLAG << flag_start_bit_);
162  *edge_ptr |= (MARKER_FLAG << flag_start_bit_);
163  }
164  // Mark both directions as end of word.
165  *back_edge_ptr |= (WERD_END_FLAG << flag_start_bit_);
166  *edge_ptr |= (WERD_END_FLAG << flag_start_bit_);
167 }
168 
170  const GenericVector<bool> *repetitions) {
171  if (word.length() <= 0) return false; // can't add empty words
172  if (repetitions != nullptr) ASSERT_HOST(repetitions->size() == word.length());
173  // Make sure the word does not contain invalid unchar ids.
174  for (int i = 0; i < word.length(); ++i) {
175  if (word.unichar_id(i) < 0 ||
176  word.unichar_id(i) >= unicharset_size_) return false;
177  }
178 
179  EDGE_RECORD *edge_ptr;
180  NODE_REF last_node = 0;
181  NODE_REF the_next_node;
182  bool marker_flag = false;
183  EDGE_INDEX edge_index;
184  int i;
185  int32_t still_finding_chars = true;
186  int32_t word_end = false;
187  bool add_failed = false;
188  bool found;
189 
190  if (debug_level_ > 1) word.print("\nAdding word: ");
191 
192  UNICHAR_ID unichar_id;
193  for (i = 0; i < word.length() - 1; ++i) {
194  unichar_id = word.unichar_id(i);
195  marker_flag = (repetitions != nullptr) ? (*repetitions)[i] : false;
196  if (debug_level_ > 1) tprintf("Adding letter %d\n", unichar_id);
197  if (still_finding_chars) {
198  found = edge_char_of(last_node, NO_EDGE, FORWARD_EDGE, word_end,
199  unichar_id, &edge_ptr, &edge_index);
200  if (found && debug_level_ > 1) {
201  tprintf("exploring edge " REFFORMAT " in node " REFFORMAT "\n",
202  edge_index, last_node);
203  }
204  if (!found) {
205  still_finding_chars = false;
206  } else if (next_node_from_edge_rec(*edge_ptr) == 0) {
207  // We hit the end of an existing word, but the new word is longer.
208  // In this case we have to disconnect the existing word from the
209  // backwards root node, mark the current position as end-of-word
210  // and add new nodes for the increased length. Disconnecting the
211  // existing word from the backwards root node requires a linear
212  // search, so it is much faster to add the longest words first,
213  // to avoid having to come here.
214  word_end = true;
215  still_finding_chars = false;
216  remove_edge(last_node, 0, word_end, unichar_id);
217  } else {
218  // We have to add a new branch here for the new word.
219  if (marker_flag) set_marker_flag_in_edge_rec(edge_ptr);
220  last_node = next_node_from_edge_rec(*edge_ptr);
221  }
222  }
223  if (!still_finding_chars) {
224  the_next_node = new_dawg_node();
225  if (debug_level_ > 1)
226  tprintf("adding node " REFFORMAT "\n", the_next_node);
227  if (the_next_node == 0) {
228  add_failed = true;
229  break;
230  }
231  if (!add_new_edge(last_node, the_next_node,
232  marker_flag, word_end, unichar_id)) {
233  add_failed = true;
234  break;
235  }
236  word_end = false;
237  last_node = the_next_node;
238  }
239  }
240  the_next_node = 0;
241  unichar_id = word.unichar_id(i);
242  marker_flag = (repetitions != nullptr) ? (*repetitions)[i] : false;
243  if (debug_level_ > 1) tprintf("Adding letter %d\n", unichar_id);
244  if (still_finding_chars &&
245  edge_char_of(last_node, NO_EDGE, FORWARD_EDGE, false,
246  unichar_id, &edge_ptr, &edge_index)) {
247  // An extension of this word already exists in the trie, so we
248  // only have to add the ending flags in both directions.
249  add_word_ending(edge_ptr, next_node_from_edge_rec(*edge_ptr),
250  marker_flag, unichar_id);
251  } else {
252  // Add a link to node 0. All leaves connect to node 0 so the back links can
253  // be used in reduction to a dawg. This root backward node has one edge
254  // entry for every word, (except prefixes of longer words) so it is huge.
255  if (!add_failed &&
256  !add_new_edge(last_node, the_next_node, marker_flag, true, unichar_id))
257  add_failed = true;
258  }
259  if (add_failed) {
260  tprintf("Re-initializing document dictionary...\n");
261  clear();
262  return false;
263  } else {
264  return true;
265  }
266 }
267 
269  auto *node = new TRIE_NODE_RECORD();
270  nodes_.push_back(node);
271  return nodes_.length() - 1;
272 }
273 
274 // Sort function to sort words by decreasing order of length.
275 static int sort_strings_by_dec_length(const void* v1, const void* v2) {
276  const auto *s1 = static_cast<const STRING *>(v1);
277  const auto *s2 = static_cast<const STRING *>(v2);
278  return s2->length() - s1->length();
279 }
280 
281 bool Trie::read_and_add_word_list(const char *filename,
282  const UNICHARSET &unicharset,
283  Trie::RTLReversePolicy reverse_policy) {
284  GenericVector<STRING> word_list;
285  if (!read_word_list(filename, &word_list)) return false;
286  word_list.sort(sort_strings_by_dec_length);
287  return add_word_list(word_list, unicharset, reverse_policy);
288 }
289 
290 bool Trie::read_word_list(const char *filename,
291  GenericVector<STRING>* words) {
292  FILE *word_file;
293  char line_str[CHARS_PER_LINE];
294  int word_count = 0;
295 
296  word_file = fopen(filename, "rb");
297  if (word_file == nullptr) return false;
298 
299  while (fgets(line_str, sizeof(line_str), word_file) != nullptr) {
300  chomp_string(line_str); // remove newline
301  STRING word_str(line_str);
302  ++word_count;
303  if (debug_level_ && word_count % 10000 == 0)
304  tprintf("Read %d words so far\n", word_count);
305  words->push_back(word_str);
306  }
307  if (debug_level_)
308  tprintf("Read %d words total.\n", word_count);
309  fclose(word_file);
310  return true;
311 }
312 
314  const UNICHARSET &unicharset,
315  Trie::RTLReversePolicy reverse_policy) {
316  for (int i = 0; i < words.size(); ++i) {
317  WERD_CHOICE word(words[i].string(), unicharset);
318  if (word.length() == 0 || word.contains_unichar_id(INVALID_UNICHAR_ID))
319  continue;
320  if ((reverse_policy == RRP_REVERSE_IF_HAS_RTL &&
321  word.has_rtl_unichar_id()) ||
322  reverse_policy == RRP_FORCE_REVERSE) {
324  }
325  if (!word_in_dawg(word)) {
326  add_word_to_dawg(word);
327  if (!word_in_dawg(word)) {
328  tprintf("Error: word '%s' not in DAWG after adding it\n",
329  words[i].string());
330  return false;
331  }
332  }
333  }
334  return true;
335 }
336 
344  unicharset->unichar_insert(kPuncPatternUnicode);
350  initialized_patterns_ = true;
351  unicharset_size_ = unicharset->size();
352 }
353 
355  const UNICHARSET &unicharset,
356  GenericVector<UNICHAR_ID> *vec) const {
357  bool is_alpha = unicharset.get_isalpha(unichar_id);
358  if (is_alpha) {
361  if (unicharset.get_islower(unichar_id)) {
363  } else if (unicharset.get_isupper(unichar_id)) {
365  }
366  }
367  if (unicharset.get_isdigit(unichar_id)) {
369  if (!is_alpha) vec->push_back(alphanum_pattern_);
370  }
371  if (unicharset.get_ispunctuation(unichar_id)) {
372  vec->push_back(punc_pattern_);
373  }
374 }
375 
377  if (ch == 'c') {
378  return alpha_pattern_;
379  } else if (ch == 'd') {
380  return digit_pattern_;
381  } else if (ch == 'n') {
382  return alphanum_pattern_;
383  } else if (ch == 'p') {
384  return punc_pattern_;
385  } else if (ch == 'a') {
386  return lower_pattern_;
387  } else if (ch == 'A') {
388  return upper_pattern_;
389  } else {
390  return INVALID_UNICHAR_ID;
391  }
392 }
393 
394 bool Trie::read_pattern_list(const char *filename,
395  const UNICHARSET &unicharset) {
396  if (!initialized_patterns_) {
397  tprintf("please call initialize_patterns() before read_pattern_list()\n");
398  return false;
399  }
400 
401  FILE *pattern_file = fopen(filename, "rb");
402  if (pattern_file == nullptr) {
403  tprintf("Error opening pattern file %s\n", filename);
404  return false;
405  }
406 
407  int pattern_count = 0;
408  char string[CHARS_PER_LINE];
409  while (fgets(string, CHARS_PER_LINE, pattern_file) != nullptr) {
410  chomp_string(string); // remove newline
411  // Parse the pattern and construct a unichar id vector.
412  // Record the number of repetitions of each unichar in the parallel vector.
413  WERD_CHOICE word(&unicharset);
414  GenericVector<bool> repetitions_vec;
415  const char *str_ptr = string;
416  int step = unicharset.step(str_ptr);
417  bool failed = false;
418  while (step > 0) {
419  UNICHAR_ID curr_unichar_id = INVALID_UNICHAR_ID;
420  if (step == 1 && *str_ptr == '\\') {
421  ++str_ptr;
422  if (*str_ptr == '\\') { // regular '\' unichar that was escaped
423  curr_unichar_id = unicharset.unichar_to_id(str_ptr, step);
424  } else {
425  if (word.length() < kSaneNumConcreteChars) {
426  tprintf("Please provide at least %d concrete characters at the"
427  " beginning of the pattern\n", kSaneNumConcreteChars);
428  failed = true;
429  break;
430  }
431  // Parse character class from expression.
432  curr_unichar_id = character_class_to_pattern(*str_ptr);
433  }
434  } else {
435  curr_unichar_id = unicharset.unichar_to_id(str_ptr, step);
436  }
437  if (curr_unichar_id == INVALID_UNICHAR_ID) {
438  failed = true;
439  break; // failed to parse this pattern
440  }
441  word.append_unichar_id(curr_unichar_id, 1, 0.0, 0.0);
442  repetitions_vec.push_back(false);
443  str_ptr += step;
444  step = unicharset.step(str_ptr);
445  // Check if there is a repetition pattern specified after this unichar.
446  if (step == 1 && *str_ptr == '\\' && *(str_ptr+1) == '*') {
447  repetitions_vec[repetitions_vec.size()-1] = true;
448  str_ptr += 2;
449  step = unicharset.step(str_ptr);
450  }
451  }
452  if (failed) {
453  tprintf("Invalid user pattern %s\n", string);
454  continue;
455  }
456  // Insert the pattern into the trie.
457  if (debug_level_ > 2) {
458  tprintf("Inserting expanded user pattern %s\n",
459  word.debug_string().string());
460  }
461  if (!this->word_in_dawg(word)) {
462  this->add_word_to_dawg(word, &repetitions_vec);
463  if (!this->word_in_dawg(word)) {
464  tprintf("Error: failed to insert pattern '%s'\n", string);
465  }
466  }
467  ++pattern_count;
468  }
469  if (debug_level_) {
470  tprintf("Read %d valid patterns from %s\n", pattern_count, filename);
471  }
472  fclose(pattern_file);
473  return true;
474 }
475 
476 void Trie::remove_edge_linkage(NODE_REF node1, NODE_REF node2, int direction,
477  bool word_end, UNICHAR_ID unichar_id) {
478  EDGE_RECORD *edge_ptr = nullptr;
479  EDGE_INDEX edge_index = 0;
480  ASSERT_HOST(edge_char_of(node1, node2, direction, word_end,
481  unichar_id, &edge_ptr, &edge_index));
482  if (debug_level_ > 1) {
483  tprintf("removed edge in nodes_[" REFFORMAT "]: ", node1);
484  print_edge_rec(*edge_ptr);
485  tprintf("\n");
486  }
487  if (direction == FORWARD_EDGE) {
488  nodes_[node1]->forward_edges.remove(edge_index);
489  } else if (node1 == 0) {
490  KillEdge(&nodes_[node1]->backward_edges[edge_index]);
491  root_back_freelist_.push_back(edge_index);
492  } else {
493  nodes_[node1]->backward_edges.remove(edge_index);
494  }
495  --num_edges_;
496 }
497 
498 // Some optimizations employed in add_word_to_dawg and trie_to_dawg:
499 // 1 Avoid insertion sorting or bubble sorting the tail root node
500 // (back links on node 0, a list of all the leaves.). The node is
501 // huge, and sorting it with n^2 time is terrible.
502 // 2 Avoid using GenericVector::remove on the tail root node.
503 // (a) During add of words to the trie, zero-out the unichars and
504 // keep a freelist of spaces to re-use.
505 // (b) During reduction, just zero-out the unichars of deleted back
506 // links, skipping zero entries while searching.
507 // 3 Avoid linear search of the tail root node. This has to be done when
508 // a suffix is added to an existing word. Adding words by decreasing
509 // length avoids this problem entirely. Words can still be added in
510 // any order, but it is faster to add the longest first.
512  root_back_freelist_.clear(); // Will be invalided by trie_to_dawg.
513  if (debug_level_ > 2) {
514  print_all("Before reduction:", MAX_NODE_EDGES_DISPLAY);
515  }
516  auto reduced_nodes = new bool[nodes_.size()];
517  for (int i = 0; i < nodes_.size(); i++) reduced_nodes[i] = false;
518  this->reduce_node_input(0, reduced_nodes);
519  delete[] reduced_nodes;
520 
521  if (debug_level_ > 2) {
522  print_all("After reduction:", MAX_NODE_EDGES_DISPLAY);
523  }
524  // Build a translation map from node indices in nodes_ vector to
525  // their target indices in EDGE_ARRAY.
526  auto *node_ref_map = new NODE_REF[nodes_.size() + 1];
527  int i, j;
528  node_ref_map[0] = 0;
529  for (i = 0; i < nodes_.size(); ++i) {
530  node_ref_map[i+1] = node_ref_map[i] + nodes_[i]->forward_edges.size();
531  }
532  int num_forward_edges = node_ref_map[i];
533 
534  // Convert nodes_ vector into EDGE_ARRAY translating the next node references
535  // in edges using node_ref_map. Empty nodes and backward edges are dropped.
536  auto edge_array = new EDGE_RECORD[num_forward_edges];
537  EDGE_ARRAY edge_array_ptr = edge_array;
538  for (i = 0; i < nodes_.size(); ++i) {
539  TRIE_NODE_RECORD *node_ptr = nodes_[i];
540  int end = node_ptr->forward_edges.size();
541  for (j = 0; j < end; ++j) {
542  EDGE_RECORD &edge_rec = node_ptr->forward_edges[j];
543  NODE_REF node_ref = next_node_from_edge_rec(edge_rec);
544  ASSERT_HOST(node_ref < nodes_.size());
545  UNICHAR_ID unichar_id = unichar_id_from_edge_rec(edge_rec);
546  link_edge(edge_array_ptr, node_ref_map[node_ref], false, FORWARD_EDGE,
547  end_of_word_from_edge_rec(edge_rec), unichar_id);
548  if (j == end - 1) set_marker_flag_in_edge_rec(edge_array_ptr);
549  ++edge_array_ptr;
550  }
551  }
552  delete[] node_ref_map;
553 
554  return new SquishedDawg(edge_array, num_forward_edges, type_, lang_,
556 }
557 
559  const EDGE_RECORD &edge1,
560  const EDGE_RECORD &edge2) {
561  if (debug_level_ > 1) {
562  tprintf("\nCollapsing node %" PRIi64 ":\n", node);
564  tprintf("Candidate edges: ");
565  print_edge_rec(edge1);
566  tprintf(", ");
567  print_edge_rec(edge2);
568  tprintf("\n\n");
569  }
570  NODE_REF next_node1 = next_node_from_edge_rec(edge1);
571  NODE_REF next_node2 = next_node_from_edge_rec(edge2);
572  TRIE_NODE_RECORD *next_node2_ptr = nodes_[next_node2];
573  // Translate all edges going to/from next_node2 to go to/from next_node1.
574  EDGE_RECORD *edge_ptr = nullptr;
575  EDGE_INDEX edge_index;
576  int i;
577  // The backward link in node to next_node2 will be zeroed out by the caller.
578  // Copy all the backward links in next_node2 to node next_node1
579  for (i = 0; i < next_node2_ptr->backward_edges.size(); ++i) {
580  const EDGE_RECORD &bkw_edge = next_node2_ptr->backward_edges[i];
581  NODE_REF curr_next_node = next_node_from_edge_rec(bkw_edge);
582  UNICHAR_ID curr_unichar_id = unichar_id_from_edge_rec(bkw_edge);
583  int curr_word_end = end_of_word_from_edge_rec(bkw_edge);
584  bool marker_flag = marker_flag_from_edge_rec(bkw_edge);
585  add_edge_linkage(next_node1, curr_next_node, marker_flag, BACKWARD_EDGE,
586  curr_word_end, curr_unichar_id);
587  // Relocate the corresponding forward edge in curr_next_node
588  ASSERT_HOST(edge_char_of(curr_next_node, next_node2, FORWARD_EDGE,
589  curr_word_end, curr_unichar_id,
590  &edge_ptr, &edge_index));
591  set_next_node_in_edge_rec(edge_ptr, next_node1);
592  }
593  int next_node2_num_edges = (next_node2_ptr->forward_edges.size() +
594  next_node2_ptr->backward_edges.size());
595  if (debug_level_ > 1) {
596  tprintf("removed %d edges from node " REFFORMAT "\n",
597  next_node2_num_edges, next_node2);
598  }
599  next_node2_ptr->forward_edges.clear();
600  next_node2_ptr->backward_edges.clear();
601  num_edges_ -= next_node2_num_edges;
602  return true;
603 }
604 
606  UNICHAR_ID unichar_id,
607  NODE_REF node,
608  EDGE_VECTOR* backward_edges,
609  NODE_MARKER reduced_nodes) {
610  if (debug_level_ > 1)
611  tprintf("reduce_lettered_edges(edge=" REFFORMAT ")\n", edge_index);
612  // Compare each of the edge pairs with the given unichar_id.
613  bool did_something = false;
614  for (int i = edge_index; i < backward_edges->size() - 1; ++i) {
615  // Find the first edge that can be eliminated.
616  UNICHAR_ID curr_unichar_id = INVALID_UNICHAR_ID;
617  while (i < backward_edges->size()) {
618  if (!DeadEdge((*backward_edges)[i])) {
619  curr_unichar_id = unichar_id_from_edge_rec((*backward_edges)[i]);
620  if (curr_unichar_id != unichar_id) return did_something;
621  if (can_be_eliminated((*backward_edges)[i])) break;
622  }
623  ++i;
624  }
625  if (i == backward_edges->size()) break;
626  const EDGE_RECORD &edge_rec = (*backward_edges)[i];
627  // Compare it to the rest of the edges with the given unichar_id.
628  for (int j = i + 1; j < backward_edges->size(); ++j) {
629  const EDGE_RECORD &next_edge_rec = (*backward_edges)[j];
630  if (DeadEdge(next_edge_rec)) continue;
631  UNICHAR_ID next_id = unichar_id_from_edge_rec(next_edge_rec);
632  if (next_id != unichar_id) break;
633  if (end_of_word_from_edge_rec(next_edge_rec) ==
634  end_of_word_from_edge_rec(edge_rec) &&
635  can_be_eliminated(next_edge_rec) &&
636  eliminate_redundant_edges(node, edge_rec, next_edge_rec)) {
637  reduced_nodes[next_node_from_edge_rec(edge_rec)] = false;
638  did_something = true;
639  KillEdge(&(*backward_edges)[j]);
640  }
641  }
642  }
643  return did_something;
644 }
645 
647  int num_edges = edges->size();
648  if (num_edges <= 1) return;
650  sort_vec.reserve(num_edges);
651  for (int i = 0; i < num_edges; ++i) {
653  unichar_id_from_edge_rec((*edges)[i]), (*edges)[i]));
654  }
655  sort_vec.sort();
656  for (int i = 0; i < num_edges; ++i)
657  (*edges)[i] = sort_vec[i].data;
658 }
659 
661  NODE_MARKER reduced_nodes) {
662  EDGE_VECTOR &backward_edges = nodes_[node]->backward_edges;
663  sort_edges(&backward_edges);
664  if (debug_level_ > 1) {
665  tprintf("reduce_node_input(node=" REFFORMAT ")\n", node);
667  }
668 
669  EDGE_INDEX edge_index = 0;
670  while (edge_index < backward_edges.size()) {
671  if (DeadEdge(backward_edges[edge_index])) continue;
672  UNICHAR_ID unichar_id =
673  unichar_id_from_edge_rec(backward_edges[edge_index]);
674  while (reduce_lettered_edges(edge_index, unichar_id, node,
675  &backward_edges, reduced_nodes));
676  while (++edge_index < backward_edges.size()) {
677  UNICHAR_ID id = unichar_id_from_edge_rec(backward_edges[edge_index]);
678  if (!DeadEdge(backward_edges[edge_index]) && id != unichar_id) break;
679  }
680  }
681  reduced_nodes[node] = true; // mark as reduced
682 
683  if (debug_level_ > 1) {
684  tprintf("Node " REFFORMAT " after reduction:\n", node);
686  }
687 
688  for (int i = 0; i < backward_edges.size(); ++i) {
689  if (DeadEdge(backward_edges[i])) continue;
690  NODE_REF next_node = next_node_from_edge_rec(backward_edges[i]);
691  if (next_node != 0 && !reduced_nodes[next_node]) {
692  reduce_node_input(next_node, reduced_nodes);
693  }
694  }
695 }
696 
697 void Trie::print_node(NODE_REF node, int max_num_edges) const {
698  if (node == NO_EDGE) return; // nothing to print
699  TRIE_NODE_RECORD *node_ptr = nodes_[node];
700  int num_fwd = node_ptr->forward_edges.size();
701  int num_bkw = node_ptr->backward_edges.size();
702  EDGE_VECTOR *vec;
703  for (int dir = 0; dir < 2; ++dir) {
704  if (dir == 0) {
705  vec = &(node_ptr->forward_edges);
706  tprintf(REFFORMAT " (%d %d): ", node, num_fwd, num_bkw);
707  } else {
708  vec = &(node_ptr->backward_edges);
709  tprintf("\t");
710  }
711  int i;
712  for (i = 0; (dir == 0 ? i < num_fwd : i < num_bkw) &&
713  i < max_num_edges; ++i) {
714  if (DeadEdge((*vec)[i])) continue;
715  print_edge_rec((*vec)[i]);
716  tprintf(" ");
717  }
718  if (dir == 0 ? i < num_fwd : i < num_bkw) tprintf("...");
719  tprintf("\n");
720  }
721 }
722 
723 } // namespace tesseract
tesseract::Dawg::given_greater_than_edge_rec
int given_greater_than_edge_rec(NODE_REF next_node, bool word_end, UNICHAR_ID unichar_id, const EDGE_RECORD &edge_rec) const
Definition: dawg.h:247
chomp_string
void chomp_string(char *str)
Definition: helpers.h:77
tesseract::Trie::kAlphanumPatternUnicode
static const char kAlphanumPatternUnicode[]
Definition: trie.h:71
UNICHARSET::size
int size() const
Definition: unicharset.h:341
STRING::string
const char * string() const
Definition: strngs.cpp:194
tesseract::Dawg::unicharset_size_
int unicharset_size_
Definition: dawg.h:308
WERD_CHOICE::has_rtl_unichar_id
bool has_rtl_unichar_id() const
Definition: ratngs.cpp:435
TRIE_NODE_RECORD::forward_edges
EDGE_VECTOR forward_edges
Definition: trie.h:43
callcpp.h
WERD_CHOICE::debug_string
const STRING debug_string() const
Definition: ratngs.h:495
kdpair.h
tesseract::Trie::next_node
NODE_REF next_node(EDGE_REF edge_ref) const override
Definition: trie.h:127
tesseract::Trie::kSaneNumConcreteChars
static const int kSaneNumConcreteChars
Definition: trie.h:65
tesseract::Trie::link_edge
void link_edge(EDGE_RECORD *edge, NODE_REF nxt, bool repeats, int direction, bool word_end, UNICHAR_ID unichar_id)
Definition: trie.h:300
UNICHARSET::step
int step(const char *str) const
Definition: unicharset.cpp:233
tesseract::KDPairInc
Definition: kdpair.h:51
tesseract::Dawg::perm_
PermuterType perm_
Permuter code that should be used if the word is found in this Dawg.
Definition: dawg.h:300
GenericVector::sort
void sort()
Definition: genericvector.h:1146
tesseract::Trie::reduce_node_input
void reduce_node_input(NODE_REF node, NODE_MARKER reduced_nodes)
Definition: trie.cpp:660
tesseract::Trie::initialized_patterns_
bool initialized_patterns_
Definition: trie.h:427
tesseract::Trie::DeadEdge
bool DeadEdge(const EDGE_RECORD &edge_rec) const
Definition: trie.h:152
tesseract::Trie::print_edge_rec
void print_edge_rec(const EDGE_RECORD &edge_rec) const
Definition: trie.h:311
BACKWARD_EDGE
#define BACKWARD_EDGE
Definition: dawg.h:82
MAX_NODE_EDGES_DISPLAY
#define MAX_NODE_EDGES_DISPLAY
Definition: dawg.h:83
CHARS_PER_LINE
#define CHARS_PER_LINE
Definition: dict.h:38
EDGE_RECORD
uint64_t EDGE_RECORD
Definition: dawg.h:49
tesseract::Dawg::end_of_word_from_edge_rec
bool end_of_word_from_edge_rec(const EDGE_RECORD &edge_rec) const
Returns true if this edge marks the end of a word.
Definition: dawg.h:222
tesseract::Dawg::unichar_id_from_edge_rec
UNICHAR_ID unichar_id_from_edge_rec(const EDGE_RECORD &edge_rec) const
Returns UNICHAR_ID recorded in this edge.
Definition: dawg.h:226
tesseract::Trie::alpha_pattern_
UNICHAR_ID alpha_pattern_
Definition: trie.h:421
UNICHARSET::get_isalpha
bool get_isalpha(UNICHAR_ID unichar_id) const
Definition: unicharset.h:491
tesseract
Definition: altorenderer.cpp:25
tesseract::Trie::eliminate_redundant_edges
bool eliminate_redundant_edges(NODE_REF node, const EDGE_RECORD &edge1, const EDGE_RECORD &edge2)
Definition: trie.cpp:558
REFFORMAT
#define REFFORMAT
Definition: dawg.h:89
tesseract::Trie::punc_pattern_
UNICHAR_ID punc_pattern_
Definition: trie.h:424
tesseract::Dawg::word_in_dawg
bool word_in_dawg(const WERD_CHOICE &word) const
Returns true if the given word is in the Dawg.
Definition: dawg.cpp:65
GenericVector::remove
void remove(int index)
Definition: genericvector.h:803
tesseract::Dawg::edge_rec_match
bool edge_rec_match(NODE_REF next_node, bool word_end, UNICHAR_ID unichar_id, NODE_REF other_next_node, bool other_word_end, UNICHAR_ID other_unichar_id) const
Definition: dawg.h:268
tesseract::Trie::add_new_edge
bool add_new_edge(NODE_REF node1, NODE_REF node2, bool repeats, bool word_end, UNICHAR_ID unichar_id)
Definition: trie.h:350
STRING::length
int32_t length() const
Definition: strngs.cpp:189
tesseract::Dawg::next_node_from_edge_rec
NODE_REF next_node_from_edge_rec(const EDGE_RECORD &edge_rec) const
Returns the next node visited by following this edge.
Definition: dawg.h:209
tesseract::Trie::lower_pattern_
UNICHAR_ID lower_pattern_
Definition: trie.h:425
UNICHARSET::get_isdigit
bool get_isdigit(UNICHAR_ID unichar_id) const
Definition: unicharset.h:512
GenericVector< EDGE_RECORD >
tesseract::Trie::edge_char_of
EDGE_REF edge_char_of(NODE_REF node_ref, UNICHAR_ID unichar_id, bool word_end) const override
Definition: trie.h:98
UNICHARSET::get_islower
bool get_islower(UNICHAR_ID unichar_id) const
Definition: unicharset.h:498
tesseract::Trie::read_word_list
bool read_word_list(const char *filename, GenericVector< STRING > *words)
Definition: trie.cpp:290
tesseract::Trie::KillEdge
void KillEdge(EDGE_RECORD *edge_rec) const
Definition: trie.h:148
tesseract::Dawg::lang_
STRING lang_
Definition: dawg.h:297
tesseract::Trie::read_pattern_list
bool read_pattern_list(const char *filename, const UNICHARSET &unicharset)
Definition: trie.cpp:394
tesseract::kForceReverse
const char kForceReverse[]
Definition: trie.cpp:37
genericvector.h
UNICHAR_ID
int UNICHAR_ID
Definition: unichar.h:34
WERD_CHOICE::unichar_id
UNICHAR_ID unichar_id(int index) const
Definition: ratngs.h:305
tprintf
DLLSYM void tprintf(const char *format,...)
Definition: tprintf.cpp:35
tesseract::Trie::upper_pattern_
UNICHAR_ID upper_pattern_
Definition: trie.h:426
GenericVector::clear
void clear()
Definition: genericvector.h:895
UNICHARSET::unichar_to_id
UNICHAR_ID unichar_to_id(const char *const unichar_repr) const
Definition: unicharset.cpp:210
tesseract::Trie::RRP_FORCE_REVERSE
@ RRP_FORCE_REVERSE
Definition: trie.h:61
ASSERT_HOST
#define ASSERT_HOST(x)
Definition: errcode.h:88
GenericVector::insert
void insert(const T &t, int index)
Definition: genericvector.h:788
GenericVector::pop_back
T pop_back()
Definition: genericvector.h:772
tesseract::Dawg::set_marker_flag_in_edge_rec
void set_marker_flag_in_edge_rec(EDGE_RECORD *edge_rec)
Sets this edge record to be the last one in a sequence of edges.
Definition: dawg.h:237
TRIE_NODE_RECORD
Definition: trie.h:42
tesseract::Dawg::set_next_node_in_edge_rec
void set_next_node_in_edge_rec(EDGE_RECORD *edge_rec, EDGE_REF value)
Sets the next node link for this edge in the Dawg.
Definition: dawg.h:231
tesseract::Trie::print_all
void print_all(const char *msg, int max_num_edges)
Definition: trie.h:328
GenericVector::reserve
void reserve(int size)
Definition: genericvector.h:717
NODE_MARKER
bool * NODE_MARKER
Definition: trie.h:39
tesseract::Dawg::type_
DawgType type_
Definition: dawg.h:298
WERD_CHOICE::append_unichar_id
void append_unichar_id(UNICHAR_ID unichar_id, int blob_count, float rating, float certainty)
Definition: ratngs.cpp:472
WERD_END_FLAG
#define WERD_END_FLAG
Definition: dawg.h:86
helpers.h
tesseract::Trie::print_node
void print_node(NODE_REF node, int max_num_edges) const override
Definition: trie.cpp:697
tesseract::Trie::new_dawg_node
NODE_REF new_dawg_node()
Definition: trie.cpp:268
tesseract::kDoNotReverse
const char kDoNotReverse[]
Definition: trie.cpp:35
tesseract::Trie::add_word_ending
void add_word_ending(EDGE_RECORD *edge, NODE_REF the_next_node, bool repeats, UNICHAR_ID unichar_id)
Definition: trie.cpp:152
tesseract::Trie::sort_edges
void sort_edges(EDGE_VECTOR *edges)
Definition: trie.cpp:646
UNICHARSET::unichar_insert
void unichar_insert(const char *const unichar_repr, OldUncleanUnichars old_style)
Definition: unicharset.cpp:626
NODE_REF
int64_t NODE_REF
Definition: dawg.h:52
MARKER_FLAG
#define MARKER_FLAG
Definition: dawg.h:84
GenericVector::length
int length() const
Definition: genericvector.h:86
tesseract::Trie::root_back_freelist_
GenericVector< EDGE_INDEX > root_back_freelist_
Definition: trie.h:415
tesseract::Dawg::flag_start_bit_
int flag_start_bit_
Definition: dawg.h:309
tesseract::Trie::read_and_add_word_list
bool read_and_add_word_list(const char *filename, const UNICHARSET &unicharset, Trie::RTLReversePolicy reverse)
Definition: trie.cpp:281
trie.h
FORWARD_EDGE
#define FORWARD_EDGE
Definition: dawg.h:81
tesseract::Trie::can_be_eliminated
bool can_be_eliminated(const EDGE_RECORD &edge_rec)
Definition: trie.h:320
tesseract::Trie::trie_to_dawg
SquishedDawg * trie_to_dawg()
Definition: trie.cpp:511
tesseract::Trie::add_word_list
bool add_word_list(const GenericVector< STRING > &words, const UNICHARSET &unicharset, Trie::RTLReversePolicy reverse_policy)
Definition: trie.cpp:313
tesseract::Trie::unichar_id_to_patterns
void unichar_id_to_patterns(UNICHAR_ID unichar_id, const UNICHARSET &unicharset, GenericVector< UNICHAR_ID > *vec) const override
Definition: trie.cpp:354
tesseract::Trie::alphanum_pattern_
UNICHAR_ID alphanum_pattern_
Definition: trie.h:423
tesseract::Trie::num_edges_
uint64_t num_edges_
Definition: trie.h:416
tesseract::Trie::nodes_
TRIE_NODES nodes_
Definition: trie.h:413
GenericVector::size
int size() const
Definition: genericvector.h:72
tesseract::Trie::kAlphaPatternUnicode
static const char kAlphaPatternUnicode[]
Definition: trie.h:69
TRIE_NODE_RECORD::backward_edges
EDGE_VECTOR backward_edges
Definition: trie.h:44
EDGE_INDEX
int64_t EDGE_INDEX
Definition: trie.h:38
EDGE_ARRAY
EDGE_RECORD * EDGE_ARRAY
Definition: dawg.h:50
tesseract::Trie::get_reverse_policy_name
static const char * get_reverse_policy_name(RTLReversePolicy reverse_policy)
Definition: trie.cpp:52
WERD_CHOICE::reverse_and_mirror_unichar_ids
void reverse_and_mirror_unichar_ids()
Definition: ratngs.cpp:369
tesseract::Trie::kPuncPatternUnicode
static const char kPuncPatternUnicode[]
Definition: trie.h:72
tesseract::Trie::kDigitPatternUnicode
static const char kDigitPatternUnicode[]
Definition: trie.h:70
tesseract::RTLReversePolicyNames
const char *const RTLReversePolicyNames[]
Definition: trie.cpp:39
GenericVector::delete_data_pointers
void delete_data_pointers()
Definition: genericvector.h:912
tesseract::Trie::remove_edge_linkage
void remove_edge_linkage(NODE_REF node1, NODE_REF node2, int direction, bool word_end, UNICHAR_ID unichar_id)
Definition: trie.cpp:476
dict.h
tesseract::SquishedDawg
Definition: dawg.h:407
tesseract::Trie::add_word_to_dawg
bool add_word_to_dawg(const WERD_CHOICE &word, const GenericVector< bool > *repetitions)
Definition: trie.cpp:169
dawg.h
tesseract::Trie::RTLReversePolicy
RTLReversePolicy
Definition: trie.h:58
tesseract::Trie::add_edge_linkage
bool add_edge_linkage(NODE_REF node1, NODE_REF node2, bool repeats, int direction, bool word_end, UNICHAR_ID unichar_id)
Definition: trie.cpp:116
WERD_CHOICE::length
int length() const
Definition: ratngs.h:293
UNICHARSET::get_ispunctuation
bool get_ispunctuation(UNICHAR_ID unichar_id) const
Definition: unicharset.h:519
tesseract::Trie::initialize_patterns
void initialize_patterns(UNICHARSET *unicharset)
Definition: trie.cpp:337
WERD_CHOICE::contains_unichar_id
bool contains_unichar_id(UNICHAR_ID unichar_id) const
Definition: ratngs.cpp:330
tesseract::Trie::RRP_REVERSE_IF_HAS_RTL
@ RRP_REVERSE_IF_HAS_RTL
Definition: trie.h:60
tesseract::Trie::reduce_lettered_edges
bool reduce_lettered_edges(EDGE_INDEX edge_index, UNICHAR_ID unichar_id, NODE_REF node, EDGE_VECTOR *backward_edges, NODE_MARKER reduced_nodes)
Definition: trie.cpp:605
tesseract::Dawg::debug_level_
int debug_level_
Definition: dawg.h:312
STRING
Definition: strngs.h:45
tesseract::Trie::digit_pattern_
UNICHAR_ID digit_pattern_
Definition: trie.h:422
tesseract::Trie::character_class_to_pattern
UNICHAR_ID character_class_to_pattern(char ch)
Definition: trie.cpp:376
UNICHARSET::get_isupper
bool get_isupper(UNICHAR_ID unichar_id) const
Definition: unicharset.h:505
WERD_CHOICE::print
void print() const
Definition: ratngs.h:570
tesseract::Trie::remove_edge
void remove_edge(NODE_REF node1, NODE_REF node2, bool word_end, UNICHAR_ID unichar_id)
Definition: trie.h:375
tesseract::Trie::kLowerPatternUnicode
static const char kLowerPatternUnicode[]
Definition: trie.h:73
UNICHARSET
Definition: unicharset.h:145
GenericVector::empty
bool empty() const
Definition: genericvector.h:91
WERD_CHOICE
Definition: ratngs.h:263
tesseract::Trie::kUpperPatternUnicode
static const char kUpperPatternUnicode[]
Definition: trie.h:74
tesseract::Dawg::marker_flag_from_edge_rec
bool marker_flag_from_edge_rec(const EDGE_RECORD &edge_rec) const
Returns the marker flag of this edge.
Definition: dawg.h:213
tesseract::Trie::clear
void clear()
Definition: trie.cpp:57
tesseract::kReverseIfHasRTL
const char kReverseIfHasRTL[]
Definition: trie.cpp:36
GenericVector::push_back
int push_back(T object)
Definition: genericvector.h:837