/src/tesseract/src/classify/shapetable.cpp
Line | Count | Source |
1 | | // Copyright 2010 Google Inc. All Rights Reserved. |
2 | | // Author: rays@google.com (Ray Smith) |
3 | | /////////////////////////////////////////////////////////////////////// |
4 | | // File: shapetable.cpp |
5 | | // Description: Class to map a classifier shape index to unicharset |
6 | | // indices and font indices. |
7 | | // Author: Ray Smith |
8 | | // |
9 | | // (C) Copyright 2010, Google Inc. |
10 | | // Licensed under the Apache License, Version 2.0 (the "License"); |
11 | | // you may not use this file except in compliance with the License. |
12 | | // You may obtain a copy of the License at |
13 | | // http://www.apache.org/licenses/LICENSE-2.0 |
14 | | // Unless required by applicable law or agreed to in writing, software |
15 | | // distributed under the License is distributed on an "AS IS" BASIS, |
16 | | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
17 | | // See the License for the specific language governing permissions and |
18 | | // limitations under the License. |
19 | | // |
20 | | /////////////////////////////////////////////////////////////////////// |
21 | | |
22 | | #include "shapetable.h" |
23 | | |
24 | | #include "bitvector.h" |
25 | | #include "fontinfo.h" |
26 | | #include "intfeaturespace.h" |
27 | | #include "unicharset.h" |
28 | | #include "unicity_table.h" |
29 | | |
30 | | #include <algorithm> |
31 | | |
32 | | namespace tesseract { |
33 | | |
34 | | // Helper function to get the index of the first result with the required |
35 | | // unichar_id. If the results are sorted by rating, this will also be the |
36 | | // best result with the required unichar_id. |
37 | | // Returns -1 if the unichar_id is not found |
38 | | int ShapeRating::FirstResultWithUnichar(const std::vector<ShapeRating> &results, |
39 | 0 | const ShapeTable &shape_table, UNICHAR_ID unichar_id) { |
40 | 0 | size_t r = 0; |
41 | 0 | for (const auto &result : results) { |
42 | 0 | const auto shape_id = result.shape_id; |
43 | 0 | const Shape &shape = shape_table.GetShape(shape_id); |
44 | 0 | if (shape.ContainsUnichar(unichar_id)) { |
45 | 0 | return r; |
46 | 0 | } |
47 | 0 | ++r; |
48 | 0 | } |
49 | 0 | return -1; |
50 | 0 | } |
51 | | |
52 | | // Helper function to get the index of the first result with the required |
53 | | // unichar_id. If the results are sorted by rating, this will also be the |
54 | | // best result with the required unichar_id. |
55 | | // Returns -1 if the unichar_id is not found |
56 | | int UnicharRating::FirstResultWithUnichar(const std::vector<UnicharRating> &results, |
57 | 0 | UNICHAR_ID unichar_id) { |
58 | 0 | size_t r = 0; |
59 | 0 | for (const auto &result : results) { |
60 | 0 | if (result.unichar_id == unichar_id) { |
61 | 0 | return r; |
62 | 0 | } |
63 | 0 | ++r; |
64 | 0 | } |
65 | 0 | return -1; |
66 | 0 | } |
67 | | |
68 | | // Writes to the given file. Returns false in case of error. |
69 | 0 | bool UnicharAndFonts::Serialize(FILE *fp) const { |
70 | 0 | return tesseract::Serialize(fp, &unichar_id) && tesseract::Serialize(fp, font_ids); |
71 | 0 | } |
72 | | |
73 | | // Reads from the given file. Returns false in case of error. |
74 | 14.0k | bool UnicharAndFonts::DeSerialize(TFile *fp) { |
75 | 14.0k | return fp->DeSerialize(&unichar_id) && fp->DeSerialize(font_ids); |
76 | 14.0k | } |
77 | | |
78 | | // Sort function to sort a pair of UnicharAndFonts by unichar_id. |
79 | 0 | int UnicharAndFonts::SortByUnicharId(const void *v1, const void *v2) { |
80 | 0 | const auto *p1 = static_cast<const UnicharAndFonts *>(v1); |
81 | 0 | const auto *p2 = static_cast<const UnicharAndFonts *>(v2); |
82 | 0 | return p1->unichar_id - p2->unichar_id; |
83 | 0 | } |
84 | | |
85 | 0 | bool UnicharAndFonts::StdSortByUnicharId(const UnicharAndFonts &v1, const UnicharAndFonts &v2) { |
86 | 0 | return v1.unichar_id < v2.unichar_id; |
87 | 0 | } |
88 | | |
89 | | // Writes to the given file. Returns false in case of error. |
90 | 0 | bool Shape::Serialize(FILE *fp) const { |
91 | 0 | uint8_t sorted = unichars_sorted_; |
92 | 0 | return tesseract::Serialize(fp, &sorted) && tesseract::Serialize(fp, unichars_); |
93 | 0 | } |
94 | | // Reads from the given file. Returns false in case of error. |
95 | | |
96 | 14.0k | bool Shape::DeSerialize(TFile *fp) { |
97 | 14.0k | uint8_t sorted; |
98 | 14.0k | if (!fp->DeSerialize(&sorted)) { |
99 | 0 | return false; |
100 | 0 | } |
101 | 14.0k | unichars_sorted_ = sorted != 0; |
102 | 14.0k | return fp->DeSerialize(unichars_); |
103 | 14.0k | } |
104 | | |
105 | | // Adds a font_id for the given unichar_id. If the unichar_id is not |
106 | | // in the shape, it is added. |
107 | 0 | void Shape::AddToShape(int unichar_id, int font_id) { |
108 | 0 | for (auto &unichar : unichars_) { |
109 | 0 | if (unichar.unichar_id == unichar_id) { |
110 | | // Found the unichar in the shape table. |
111 | 0 | std::vector<int> &font_list = unichar.font_ids; |
112 | 0 | for (int f : font_list) { |
113 | 0 | if (f == font_id) { |
114 | 0 | return; // Font is already there. |
115 | 0 | } |
116 | 0 | } |
117 | 0 | font_list.push_back(font_id); |
118 | 0 | return; |
119 | 0 | } |
120 | 0 | } |
121 | | // Unichar_id is not in shape, so add it to shape. |
122 | 0 | unichars_.emplace_back(unichar_id, font_id); |
123 | 0 | unichars_sorted_ = unichars_.size() <= 1; |
124 | 0 | } |
125 | | |
126 | | // Adds everything in other to this. |
127 | 0 | void Shape::AddShape(const Shape &other) { |
128 | 0 | for (const auto &unichar : other.unichars_) { |
129 | 0 | for (auto font_id : unichar.font_ids) { |
130 | 0 | AddToShape(unichar.unichar_id, font_id); |
131 | 0 | } |
132 | 0 | } |
133 | 0 | unichars_sorted_ = unichars_.size() <= 1; |
134 | 0 | } |
135 | | |
136 | | // Returns true if the shape contains the given unichar_id, font_id pair. |
137 | 0 | bool Shape::ContainsUnicharAndFont(int unichar_id, int font_id) const { |
138 | 0 | for (const auto &unichar : unichars_) { |
139 | 0 | if (unichar.unichar_id == unichar_id) { |
140 | | // Found the unichar, so look for the font. |
141 | 0 | auto &font_list = unichar.font_ids; |
142 | 0 | for (int f : font_list) { |
143 | 0 | if (f == font_id) { |
144 | 0 | return true; |
145 | 0 | } |
146 | 0 | } |
147 | 0 | return false; |
148 | 0 | } |
149 | 0 | } |
150 | 0 | return false; |
151 | 0 | } |
152 | | |
153 | | // Returns true if the shape contains the given unichar_id, ignoring font. |
154 | 0 | bool Shape::ContainsUnichar(int unichar_id) const { |
155 | 0 | for (const auto &unichar : unichars_) { |
156 | 0 | if (unichar.unichar_id == unichar_id) { |
157 | 0 | return true; |
158 | 0 | } |
159 | 0 | } |
160 | 0 | return false; |
161 | 0 | } |
162 | | |
163 | | // Returns true if the shape contains the given font, ignoring unichar_id. |
164 | 0 | bool Shape::ContainsFont(int font_id) const { |
165 | 0 | for (const auto &unichar : unichars_) { |
166 | 0 | auto &font_list = unichar.font_ids; |
167 | 0 | for (int f : font_list) { |
168 | 0 | if (f == font_id) { |
169 | 0 | return true; |
170 | 0 | } |
171 | 0 | } |
172 | 0 | } |
173 | 0 | return false; |
174 | 0 | } |
175 | | // Returns true if the shape contains the given font properties, ignoring |
176 | | // unichar_id. |
177 | 0 | bool Shape::ContainsFontProperties(const FontInfoTable &font_table, uint32_t properties) const { |
178 | 0 | for (const auto &unichar : unichars_) { |
179 | 0 | auto &font_list = unichar.font_ids; |
180 | 0 | for (int f : font_list) { |
181 | 0 | if (font_table.at(f).properties == properties) { |
182 | 0 | return true; |
183 | 0 | } |
184 | 0 | } |
185 | 0 | } |
186 | 0 | return false; |
187 | 0 | } |
188 | | // Returns true if the shape contains multiple different font properties, |
189 | | // ignoring unichar_id. |
190 | 0 | bool Shape::ContainsMultipleFontProperties(const FontInfoTable &font_table) const { |
191 | 0 | uint32_t properties = font_table.at(unichars_[0].font_ids[0]).properties; |
192 | 0 | for (const auto &unichar : unichars_) { |
193 | 0 | auto &font_list = unichar.font_ids; |
194 | 0 | for (int f : font_list) { |
195 | 0 | if (font_table.at(f).properties != properties) { |
196 | 0 | return true; |
197 | 0 | } |
198 | 0 | } |
199 | 0 | } |
200 | 0 | return false; |
201 | 0 | } |
202 | | |
203 | | // Returns true if this shape is equal to other (ignoring order of unichars |
204 | | // and fonts). |
205 | 0 | bool Shape::operator==(const Shape &other) const { |
206 | 0 | return IsSubsetOf(other) && other.IsSubsetOf(*this); |
207 | 0 | } |
208 | | |
209 | | // Returns true if this is a subset (including equal) of other. |
210 | 0 | bool Shape::IsSubsetOf(const Shape &other) const { |
211 | 0 | for (const auto &unichar : unichars_) { |
212 | 0 | int unichar_id = unichar.unichar_id; |
213 | 0 | const std::vector<int> &font_list = unichar.font_ids; |
214 | 0 | for (int f : font_list) { |
215 | 0 | if (!other.ContainsUnicharAndFont(unichar_id, f)) { |
216 | 0 | return false; |
217 | 0 | } |
218 | 0 | } |
219 | 0 | } |
220 | 0 | return true; |
221 | 0 | } |
222 | | |
223 | | // Returns true if the lists of unichar ids are the same in this and other, |
224 | | // ignoring fonts. |
225 | | // NOT const, as it will sort the unichars on demand. |
226 | 0 | bool Shape::IsEqualUnichars(Shape *other) { |
227 | 0 | if (unichars_.size() != other->unichars_.size()) { |
228 | 0 | return false; |
229 | 0 | } |
230 | 0 | if (!unichars_sorted_) { |
231 | 0 | SortUnichars(); |
232 | 0 | } |
233 | 0 | if (!other->unichars_sorted_) { |
234 | 0 | other->SortUnichars(); |
235 | 0 | } |
236 | 0 | for (unsigned c = 0; c < unichars_.size(); ++c) { |
237 | 0 | if (unichars_[c].unichar_id != other->unichars_[c].unichar_id) { |
238 | 0 | return false; |
239 | 0 | } |
240 | 0 | } |
241 | 0 | return true; |
242 | 0 | } |
243 | | |
244 | | // Sorts the unichars_ vector by unichar. |
245 | 0 | void Shape::SortUnichars() { |
246 | 0 | std::sort(unichars_.begin(), unichars_.end(), UnicharAndFonts::StdSortByUnicharId); |
247 | 0 | unichars_sorted_ = true; |
248 | 0 | } |
249 | | |
250 | 0 | ShapeTable::ShapeTable() : unicharset_(nullptr), num_fonts_(0) {} |
251 | 4 | ShapeTable::ShapeTable(const UNICHARSET &unicharset) : unicharset_(&unicharset), num_fonts_(0) {} |
252 | | |
253 | | // Writes to the given file. Returns false in case of error. |
254 | 0 | bool ShapeTable::Serialize(FILE *fp) const { |
255 | 0 | return tesseract::Serialize(fp, shape_table_); |
256 | 0 | } |
257 | | // Reads from the given file. Returns false in case of error. |
258 | | |
259 | 4 | bool ShapeTable::DeSerialize(TFile *fp) { |
260 | 4 | if (!fp->DeSerialize(shape_table_)) { |
261 | 0 | return false; |
262 | 0 | } |
263 | 4 | num_fonts_ = 0; |
264 | 4 | return true; |
265 | 4 | } |
266 | | |
267 | | // Returns the number of fonts used in this ShapeTable, computing it if |
268 | | // necessary. |
269 | 0 | int ShapeTable::NumFonts() const { |
270 | 0 | if (num_fonts_ <= 0) { |
271 | 0 | for (auto shape_id : shape_table_) { |
272 | 0 | const Shape &shape = *shape_id; |
273 | 0 | for (int c = 0; c < shape.size(); ++c) { |
274 | 0 | for (auto font_id : shape[c].font_ids) { |
275 | 0 | if (font_id >= num_fonts_) { |
276 | 0 | num_fonts_ = font_id + 1; |
277 | 0 | } |
278 | 0 | } |
279 | 0 | } |
280 | 0 | } |
281 | 0 | } |
282 | 0 | return num_fonts_; |
283 | 0 | } |
284 | | |
285 | | // Re-indexes the class_ids in the shapetable according to the given map. |
286 | | // Useful in conjunction with set_unicharset. |
287 | 0 | void ShapeTable::ReMapClassIds(const std::vector<int> &unicharset_map) { |
288 | 0 | for (auto shape : shape_table_) { |
289 | 0 | for (int c = 0; c < shape->size(); ++c) { |
290 | 0 | shape->SetUnicharId(c, unicharset_map[(*shape)[c].unichar_id]); |
291 | 0 | } |
292 | 0 | } |
293 | 0 | } |
294 | | |
295 | | // Returns a string listing the classes/fonts in a shape. |
296 | 0 | std::string ShapeTable::DebugStr(unsigned shape_id) const { |
297 | 0 | if (shape_id >= shape_table_.size()) { |
298 | 0 | return "INVALID_UNICHAR_ID"; |
299 | 0 | } |
300 | 0 | const Shape &shape = GetShape(shape_id); |
301 | 0 | std::string result; |
302 | 0 | result += "Shape" + std::to_string(shape_id); |
303 | 0 | if (shape.size() > 100) { |
304 | 0 | result += " Num unichars=" + std::to_string(shape.size()); |
305 | 0 | return result; |
306 | 0 | } |
307 | 0 | for (int c = 0; c < shape.size(); ++c) { |
308 | 0 | result += " c_id=" + std::to_string(shape[c].unichar_id); |
309 | 0 | result += "="; |
310 | 0 | result += unicharset_->id_to_unichar(shape[c].unichar_id); |
311 | 0 | if (shape.size() < 10) { |
312 | 0 | result += ", " + std::to_string(shape[c].font_ids.size()); |
313 | 0 | result += " fonts ="; |
314 | 0 | int num_fonts = shape[c].font_ids.size(); |
315 | 0 | if (num_fonts > 10) { |
316 | 0 | result += " " + std::to_string(shape[c].font_ids[0]); |
317 | 0 | result += " ... " + std::to_string(shape[c].font_ids[num_fonts - 1]); |
318 | 0 | } else { |
319 | 0 | for (int f = 0; f < num_fonts; ++f) { |
320 | 0 | result += " " + std::to_string(shape[c].font_ids[f]); |
321 | 0 | } |
322 | 0 | } |
323 | 0 | } |
324 | 0 | } |
325 | 0 | return result; |
326 | 0 | } |
327 | | |
328 | | // Returns a debug string summarizing the table. |
329 | 0 | std::string ShapeTable::SummaryStr() const { |
330 | 0 | int max_unichars = 0; |
331 | 0 | int num_multi_shapes = 0; |
332 | 0 | int num_master_shapes = 0; |
333 | 0 | for (unsigned s = 0; s < shape_table_.size(); ++s) { |
334 | 0 | if (MasterDestinationIndex(s) != s) { |
335 | 0 | continue; |
336 | 0 | } |
337 | 0 | ++num_master_shapes; |
338 | 0 | int shape_size = GetShape(s).size(); |
339 | 0 | if (shape_size > 1) { |
340 | 0 | ++num_multi_shapes; |
341 | 0 | } |
342 | 0 | if (shape_size > max_unichars) { |
343 | 0 | max_unichars = shape_size; |
344 | 0 | } |
345 | 0 | } |
346 | 0 | std::string result; |
347 | 0 | result += "Number of shapes = " + std::to_string(num_master_shapes); |
348 | 0 | result += " max unichars = " + std::to_string(max_unichars); |
349 | 0 | result += " number with multiple unichars = " + std::to_string(num_multi_shapes); |
350 | 0 | return result; |
351 | 0 | } |
352 | | |
353 | | // Adds a new shape starting with the given unichar_id and font_id. |
354 | | // Returns the assigned index. |
355 | 0 | unsigned ShapeTable::AddShape(int unichar_id, int font_id) { |
356 | 0 | auto index = shape_table_.size(); |
357 | 0 | auto *shape = new Shape; |
358 | 0 | shape->AddToShape(unichar_id, font_id); |
359 | 0 | shape_table_.push_back(shape); |
360 | 0 | num_fonts_ = std::max(num_fonts_, font_id + 1); |
361 | 0 | return index; |
362 | 0 | } |
363 | | |
364 | | // Adds a copy of the given shape unless it is already present. |
365 | | // Returns the assigned index or index of existing shape if already present. |
366 | 0 | unsigned ShapeTable::AddShape(const Shape &other) { |
367 | 0 | unsigned index; |
368 | 0 | for (index = 0; index < shape_table_.size() && !(other == *shape_table_[index]); ++index) { |
369 | 0 | continue; |
370 | 0 | } |
371 | 0 | if (index == shape_table_.size()) { |
372 | 0 | auto *shape = new Shape(other); |
373 | 0 | shape_table_.push_back(shape); |
374 | 0 | } |
375 | 0 | num_fonts_ = 0; |
376 | 0 | return index; |
377 | 0 | } |
378 | | |
379 | | // Removes the shape given by the shape index. |
380 | 0 | void ShapeTable::DeleteShape(unsigned shape_id) { |
381 | 0 | delete shape_table_[shape_id]; |
382 | 0 | shape_table_.erase(shape_table_.begin() + shape_id); |
383 | 0 | } |
384 | | |
385 | | // Adds a font_id to the given existing shape index for the given |
386 | | // unichar_id. If the unichar_id is not in the shape, it is added. |
387 | 0 | void ShapeTable::AddToShape(unsigned shape_id, int unichar_id, int font_id) { |
388 | 0 | Shape &shape = *shape_table_[shape_id]; |
389 | 0 | shape.AddToShape(unichar_id, font_id); |
390 | 0 | num_fonts_ = std::max(num_fonts_, font_id + 1); |
391 | 0 | } |
392 | | |
393 | | // Adds the given shape to the existing shape with the given index. |
394 | 0 | void ShapeTable::AddShapeToShape(unsigned shape_id, const Shape &other) { |
395 | 0 | Shape &shape = *shape_table_[shape_id]; |
396 | 0 | shape.AddShape(other); |
397 | 0 | num_fonts_ = 0; |
398 | 0 | } |
399 | | |
400 | | // Returns the id of the shape that contains the given unichar and font. |
401 | | // If not found, returns -1. |
402 | | // If font_id < 0, the font_id is ignored and the first shape that matches |
403 | | // the unichar_id is returned. |
404 | 0 | int ShapeTable::FindShape(int unichar_id, int font_id) const { |
405 | 0 | for (unsigned s = 0; s < shape_table_.size(); ++s) { |
406 | 0 | const Shape &shape = GetShape(s); |
407 | 0 | for (int c = 0; c < shape.size(); ++c) { |
408 | 0 | if (shape[c].unichar_id == unichar_id) { |
409 | 0 | if (font_id < 0) { |
410 | 0 | return s; // We don't care about the font. |
411 | 0 | } |
412 | 0 | for (auto f : shape[c].font_ids) { |
413 | 0 | if (f == font_id) { |
414 | 0 | return s; |
415 | 0 | } |
416 | 0 | } |
417 | 0 | } |
418 | 0 | } |
419 | 0 | } |
420 | 0 | return -1; |
421 | 0 | } |
422 | | |
423 | | // Returns the first unichar_id and font_id in the given shape. |
424 | 0 | void ShapeTable::GetFirstUnicharAndFont(unsigned shape_id, int *unichar_id, int *font_id) const { |
425 | 0 | const UnicharAndFonts &unichar_and_fonts = (*shape_table_[shape_id])[0]; |
426 | 0 | *unichar_id = unichar_and_fonts.unichar_id; |
427 | 0 | *font_id = unichar_and_fonts.font_ids[0]; |
428 | 0 | } |
429 | | |
430 | | // Expands all the classes/fonts in the shape individually to build |
431 | | // a ShapeTable. |
432 | 0 | int ShapeTable::BuildFromShape(const Shape &shape, const ShapeTable &master_shapes) { |
433 | 0 | BitVector shape_map(master_shapes.NumShapes()); |
434 | 0 | for (int u_ind = 0; u_ind < shape.size(); ++u_ind) { |
435 | 0 | for (auto font_id : shape[u_ind].font_ids) { |
436 | 0 | int c = shape[u_ind].unichar_id; |
437 | 0 | int master_id = master_shapes.FindShape(c, font_id); |
438 | 0 | if (master_id >= 0) { |
439 | 0 | shape_map.SetBit(master_id); |
440 | 0 | } else if (FindShape(c, font_id) < 0) { |
441 | 0 | AddShape(c, font_id); |
442 | 0 | } |
443 | 0 | } |
444 | 0 | } |
445 | 0 | int num_masters = 0; |
446 | 0 | for (unsigned s = 0; s < master_shapes.NumShapes(); ++s) { |
447 | 0 | if (shape_map[s]) { |
448 | 0 | AddShape(master_shapes.GetShape(s)); |
449 | 0 | ++num_masters; |
450 | 0 | } |
451 | 0 | } |
452 | 0 | return num_masters; |
453 | 0 | } |
454 | | |
455 | | // Returns true if the shapes are already merged. |
456 | 0 | bool ShapeTable::AlreadyMerged(unsigned shape_id1, unsigned shape_id2) const { |
457 | 0 | return MasterDestinationIndex(shape_id1) == MasterDestinationIndex(shape_id2); |
458 | 0 | } |
459 | | |
460 | | // Returns true if any shape contains multiple unichars. |
461 | 0 | bool ShapeTable::AnyMultipleUnichars() const { |
462 | 0 | auto num_shapes = NumShapes(); |
463 | 0 | for (unsigned s1 = 0; s1 < num_shapes; ++s1) { |
464 | 0 | if (MasterDestinationIndex(s1) != s1) { |
465 | 0 | continue; |
466 | 0 | } |
467 | 0 | if (GetShape(s1).size() > 1) { |
468 | 0 | return true; |
469 | 0 | } |
470 | 0 | } |
471 | 0 | return false; |
472 | 0 | } |
473 | | |
474 | | // Returns the maximum number of unichars over all shapes. |
475 | 1.85M | int ShapeTable::MaxNumUnichars() const { |
476 | 1.85M | int max_num_unichars = 0; |
477 | 1.85M | int num_shapes = NumShapes(); |
478 | 6.51G | for (int s = 0; s < num_shapes; ++s) { |
479 | 6.51G | if (GetShape(s).size() > max_num_unichars) { |
480 | 1.85M | max_num_unichars = GetShape(s).size(); |
481 | 1.85M | } |
482 | 6.51G | } |
483 | 1.85M | return max_num_unichars; |
484 | 1.85M | } |
485 | | |
486 | | // Merges shapes with a common unichar over the [start, end) interval. |
487 | | // Assumes single unichar per shape. |
488 | 0 | void ShapeTable::ForceFontMerges(unsigned start, unsigned end) { |
489 | 0 | for (unsigned s1 = start; s1 < end; ++s1) { |
490 | 0 | if (MasterDestinationIndex(s1) == s1 && GetShape(s1).size() == 1) { |
491 | 0 | int unichar_id = GetShape(s1)[0].unichar_id; |
492 | 0 | for (auto s2 = s1 + 1; s2 < end; ++s2) { |
493 | 0 | if (MasterDestinationIndex(s2) == s2 && GetShape(s2).size() == 1 && |
494 | 0 | unichar_id == GetShape(s2)[0].unichar_id) { |
495 | 0 | MergeShapes(s1, s2); |
496 | 0 | } |
497 | 0 | } |
498 | 0 | } |
499 | 0 | } |
500 | 0 | ShapeTable compacted(*unicharset_); |
501 | 0 | compacted.AppendMasterShapes(*this, nullptr); |
502 | 0 | *this = compacted; |
503 | 0 | } |
504 | | |
505 | | // Returns the number of unichars in the master shape. |
506 | 0 | unsigned ShapeTable::MasterUnicharCount(unsigned shape_id) const { |
507 | 0 | int master_id = MasterDestinationIndex(shape_id); |
508 | 0 | return GetShape(master_id).size(); |
509 | 0 | } |
510 | | |
511 | | // Returns the sum of the font counts in the master shape. |
512 | 0 | int ShapeTable::MasterFontCount(unsigned shape_id) const { |
513 | 0 | int master_id = MasterDestinationIndex(shape_id); |
514 | 0 | const Shape &shape = GetShape(master_id); |
515 | 0 | int font_count = 0; |
516 | 0 | for (int c = 0; c < shape.size(); ++c) { |
517 | 0 | font_count += shape[c].font_ids.size(); |
518 | 0 | } |
519 | 0 | return font_count; |
520 | 0 | } |
521 | | |
522 | | // Returns the number of unichars that would result from merging the shapes. |
523 | 0 | int ShapeTable::MergedUnicharCount(unsigned shape_id1, unsigned shape_id2) const { |
524 | | // Do it the easy way for now. |
525 | 0 | int master_id1 = MasterDestinationIndex(shape_id1); |
526 | 0 | int master_id2 = MasterDestinationIndex(shape_id2); |
527 | 0 | Shape combined_shape(*shape_table_[master_id1]); |
528 | 0 | combined_shape.AddShape(*shape_table_[master_id2]); |
529 | 0 | return combined_shape.size(); |
530 | 0 | } |
531 | | |
532 | | // Merges two shape_ids, leaving shape_id2 marked as merged. |
533 | 0 | void ShapeTable::MergeShapes(unsigned shape_id1, unsigned shape_id2) { |
534 | 0 | auto master_id1 = MasterDestinationIndex(shape_id1); |
535 | 0 | auto master_id2 = MasterDestinationIndex(shape_id2); |
536 | | // Point master_id2 (and all merged shapes) to master_id1. |
537 | 0 | shape_table_[master_id2]->set_destination_index(master_id1); |
538 | | // Add all the shapes of master_id2 to master_id1. |
539 | 0 | shape_table_[master_id1]->AddShape(*shape_table_[master_id2]); |
540 | 0 | } |
541 | | |
542 | | // Swaps two shape_ids. |
543 | 0 | void ShapeTable::SwapShapes(unsigned shape_id1, unsigned shape_id2) { |
544 | 0 | Shape *tmp = shape_table_[shape_id1]; |
545 | 0 | shape_table_[shape_id1] = shape_table_[shape_id2]; |
546 | 0 | shape_table_[shape_id2] = tmp; |
547 | 0 | } |
548 | | |
549 | | // Returns the destination of this shape, (if merged), taking into account |
550 | | // the fact that the destination may itself have been merged. |
551 | 0 | unsigned ShapeTable::MasterDestinationIndex(unsigned shape_id) const { |
552 | 0 | auto dest_id = shape_table_[shape_id]->destination_index(); |
553 | 0 | if (static_cast<unsigned>(dest_id) == shape_id || dest_id < 0) { |
554 | 0 | return shape_id; // Is master already. |
555 | 0 | } |
556 | 0 | auto master_id = shape_table_[dest_id]->destination_index(); |
557 | 0 | if (master_id == dest_id || master_id < 0) { |
558 | 0 | return dest_id; // Dest is the master and shape_id points to it. |
559 | 0 | } |
560 | 0 | master_id = MasterDestinationIndex(master_id); |
561 | 0 | return master_id; |
562 | 0 | } |
563 | | |
564 | | // Returns false if the unichars in neither shape is a subset of the other. |
565 | 0 | bool ShapeTable::SubsetUnichar(unsigned shape_id1, unsigned shape_id2) const { |
566 | 0 | const Shape &shape1 = GetShape(shape_id1); |
567 | 0 | const Shape &shape2 = GetShape(shape_id2); |
568 | 0 | int c1, c2; |
569 | 0 | for (c1 = 0; c1 < shape1.size(); ++c1) { |
570 | 0 | int unichar_id1 = shape1[c1].unichar_id; |
571 | 0 | if (!shape2.ContainsUnichar(unichar_id1)) { |
572 | 0 | break; |
573 | 0 | } |
574 | 0 | } |
575 | 0 | for (c2 = 0; c2 < shape2.size(); ++c2) { |
576 | 0 | int unichar_id2 = shape2[c2].unichar_id; |
577 | 0 | if (!shape1.ContainsUnichar(unichar_id2)) { |
578 | 0 | break; |
579 | 0 | } |
580 | 0 | } |
581 | 0 | return c1 == shape1.size() || c2 == shape2.size(); |
582 | 0 | } |
583 | | |
584 | | // Returns false if the unichars in neither shape is a subset of the other. |
585 | 0 | bool ShapeTable::MergeSubsetUnichar(int merge_id1, int merge_id2, unsigned shape_id) const { |
586 | 0 | const Shape &merge1 = GetShape(merge_id1); |
587 | 0 | const Shape &merge2 = GetShape(merge_id2); |
588 | 0 | const Shape &shape = GetShape(shape_id); |
589 | 0 | int cm1, cm2, cs; |
590 | 0 | for (cs = 0; cs < shape.size(); ++cs) { |
591 | 0 | int unichar_id = shape[cs].unichar_id; |
592 | 0 | if (!merge1.ContainsUnichar(unichar_id) && !merge2.ContainsUnichar(unichar_id)) { |
593 | 0 | break; // Shape is not a subset of the merge. |
594 | 0 | } |
595 | 0 | } |
596 | 0 | for (cm1 = 0; cm1 < merge1.size(); ++cm1) { |
597 | 0 | int unichar_id1 = merge1[cm1].unichar_id; |
598 | 0 | if (!shape.ContainsUnichar(unichar_id1)) { |
599 | 0 | break; // Merge is not a subset of shape |
600 | 0 | } |
601 | 0 | } |
602 | 0 | for (cm2 = 0; cm2 < merge2.size(); ++cm2) { |
603 | 0 | int unichar_id2 = merge2[cm2].unichar_id; |
604 | 0 | if (!shape.ContainsUnichar(unichar_id2)) { |
605 | 0 | break; // Merge is not a subset of shape |
606 | 0 | } |
607 | 0 | } |
608 | 0 | return cs == shape.size() || (cm1 == merge1.size() && cm2 == merge2.size()); |
609 | 0 | } |
610 | | |
611 | | // Returns true if the unichar sets are equal between the shapes. |
612 | 0 | bool ShapeTable::EqualUnichars(unsigned shape_id1, unsigned shape_id2) const { |
613 | 0 | const Shape &shape1 = GetShape(shape_id1); |
614 | 0 | const Shape &shape2 = GetShape(shape_id2); |
615 | 0 | for (int c1 = 0; c1 < shape1.size(); ++c1) { |
616 | 0 | int unichar_id1 = shape1[c1].unichar_id; |
617 | 0 | if (!shape2.ContainsUnichar(unichar_id1)) { |
618 | 0 | return false; |
619 | 0 | } |
620 | 0 | } |
621 | 0 | for (int c2 = 0; c2 < shape2.size(); ++c2) { |
622 | 0 | int unichar_id2 = shape2[c2].unichar_id; |
623 | 0 | if (!shape1.ContainsUnichar(unichar_id2)) { |
624 | 0 | return false; |
625 | 0 | } |
626 | 0 | } |
627 | 0 | return true; |
628 | 0 | } |
629 | | |
630 | | // Returns true if the unichar sets are equal between the shapes. |
631 | 0 | bool ShapeTable::MergeEqualUnichars(int merge_id1, int merge_id2, unsigned shape_id) const { |
632 | 0 | const Shape &merge1 = GetShape(merge_id1); |
633 | 0 | const Shape &merge2 = GetShape(merge_id2); |
634 | 0 | const Shape &shape = GetShape(shape_id); |
635 | 0 | for (int cs = 0; cs < shape.size(); ++cs) { |
636 | 0 | auto unichar_id = shape[cs].unichar_id; |
637 | 0 | if (!merge1.ContainsUnichar(unichar_id) && !merge2.ContainsUnichar(unichar_id)) { |
638 | 0 | return false; // Shape has a unichar that appears in neither merge. |
639 | 0 | } |
640 | 0 | } |
641 | 0 | for (int cm1 = 0; cm1 < merge1.size(); ++cm1) { |
642 | 0 | auto unichar_id1 = merge1[cm1].unichar_id; |
643 | 0 | if (!shape.ContainsUnichar(unichar_id1)) { |
644 | 0 | return false; // Merge has a unichar that is not in shape. |
645 | 0 | } |
646 | 0 | } |
647 | 0 | for (int cm2 = 0; cm2 < merge2.size(); ++cm2) { |
648 | 0 | auto unichar_id2 = merge2[cm2].unichar_id; |
649 | 0 | if (!shape.ContainsUnichar(unichar_id2)) { |
650 | 0 | return false; // Merge has a unichar that is not in shape. |
651 | 0 | } |
652 | 0 | } |
653 | 0 | return true; |
654 | 0 | } |
655 | | |
656 | | // Returns true if there is a common unichar between the shapes. |
657 | 0 | bool ShapeTable::CommonUnichars(unsigned shape_id1, unsigned shape_id2) const { |
658 | 0 | const Shape &shape1 = GetShape(shape_id1); |
659 | 0 | const Shape &shape2 = GetShape(shape_id2); |
660 | 0 | for (int c1 = 0; c1 < shape1.size(); ++c1) { |
661 | 0 | auto unichar_id1 = shape1[c1].unichar_id; |
662 | 0 | if (shape2.ContainsUnichar(unichar_id1)) { |
663 | 0 | return true; |
664 | 0 | } |
665 | 0 | } |
666 | 0 | return false; |
667 | 0 | } |
668 | | |
669 | | // Returns true if there is a common font id between the shapes. |
670 | 0 | bool ShapeTable::CommonFont(unsigned shape_id1, unsigned shape_id2) const { |
671 | 0 | const Shape &shape1 = GetShape(shape_id1); |
672 | 0 | const Shape &shape2 = GetShape(shape_id2); |
673 | 0 | for (int c1 = 0; c1 < shape1.size(); ++c1) { |
674 | 0 | const std::vector<int> &font_list1 = shape1[c1].font_ids; |
675 | 0 | for (int f : font_list1) { |
676 | 0 | if (shape2.ContainsFont(f)) { |
677 | 0 | return true; |
678 | 0 | } |
679 | 0 | } |
680 | 0 | } |
681 | 0 | return false; |
682 | 0 | } |
683 | | |
684 | | // Appends the master shapes from other to this. |
685 | | // If not nullptr, shape_map is set to map other shape_ids to this's shape_ids. |
686 | 0 | void ShapeTable::AppendMasterShapes(const ShapeTable &other, std::vector<int> *shape_map) { |
687 | 0 | if (shape_map != nullptr) { |
688 | 0 | shape_map->clear(); |
689 | 0 | shape_map->resize(other.NumShapes(), -1); |
690 | 0 | } |
691 | 0 | for (unsigned s = 0; s < other.shape_table_.size(); ++s) { |
692 | 0 | if (other.shape_table_[s]->destination_index() < 0) { |
693 | 0 | int index = AddShape(*other.shape_table_[s]); |
694 | 0 | if (shape_map != nullptr) { |
695 | 0 | (*shape_map)[s] = index; |
696 | 0 | } |
697 | 0 | } |
698 | 0 | } |
699 | 0 | } |
700 | | |
701 | | // Returns the number of master shapes remaining after merging. |
702 | 0 | int ShapeTable::NumMasterShapes() const { |
703 | 0 | int num_shapes = 0; |
704 | 0 | for (auto s : shape_table_) { |
705 | 0 | if (s->destination_index() < 0) { |
706 | 0 | ++num_shapes; |
707 | 0 | } |
708 | 0 | } |
709 | 0 | return num_shapes; |
710 | 0 | } |
711 | | |
712 | | // Adds the unichars of the given shape_id to the vector of results. Any |
713 | | // unichar_id that is already present just has the fonts added to the |
714 | | // font set for that result without adding a new entry in the vector. |
715 | | // NOTE: it is assumed that the results are given to this function in order |
716 | | // of decreasing rating. |
717 | | // The unichar_map vector indicates the index of the results entry containing |
718 | | // each unichar, or -1 if the unichar is not yet included in results. |
719 | | void ShapeTable::AddShapeToResults(const ShapeRating &shape_rating, std::vector<int> *unichar_map, |
720 | 0 | std::vector<UnicharRating> *results) const { |
721 | 0 | if (shape_rating.joined) { |
722 | 0 | AddUnicharToResults(UNICHAR_JOINED, shape_rating.rating, unichar_map, results); |
723 | 0 | } |
724 | 0 | if (shape_rating.broken) { |
725 | 0 | AddUnicharToResults(UNICHAR_BROKEN, shape_rating.rating, unichar_map, results); |
726 | 0 | } |
727 | 0 | const Shape &shape = GetShape(shape_rating.shape_id); |
728 | 0 | for (int u = 0; u < shape.size(); ++u) { |
729 | 0 | int result_index = |
730 | 0 | AddUnicharToResults(shape[u].unichar_id, shape_rating.rating, unichar_map, results); |
731 | 0 | for (auto font_id : shape[u].font_ids) { |
732 | 0 | (*results)[result_index].fonts.emplace_back(font_id, |
733 | 0 | IntCastRounded(shape_rating.rating * INT16_MAX)); |
734 | 0 | } |
735 | 0 | } |
736 | 0 | } |
737 | | |
738 | | // Adds the given unichar_id to the results if needed, updating unichar_map |
739 | | // and returning the index of unichar in results. |
740 | | int ShapeTable::AddUnicharToResults(int unichar_id, float rating, std::vector<int> *unichar_map, |
741 | 0 | std::vector<UnicharRating> *results) const { |
742 | 0 | int result_index = unichar_map->at(unichar_id); |
743 | 0 | if (result_index < 0) { |
744 | 0 | UnicharRating result(unichar_id, rating); |
745 | 0 | result_index = results->size(); |
746 | 0 | results->push_back(result); |
747 | 0 | (*unichar_map)[unichar_id] = result_index; |
748 | 0 | } |
749 | 0 | return result_index; |
750 | 0 | } |
751 | | |
752 | | } // namespace tesseract |