| File: | GrammarApplicator_runRules.cpp |
| Warning: | line 702, column 7 Value stored to 'did_test' is never read |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
| 1 | /* |
| 2 | * Copyright (C) 2007-2024, GrammarSoft ApS |
| 3 | * Developed by Tino Didriksen <mail@tinodidriksen.com> |
| 4 | * Design by Eckhard Bick <eckhard.bick@mail.dk>, Tino Didriksen <mail@tinodidriksen.com> |
| 5 | * |
| 6 | * This program is free software: you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License as published by |
| 8 | * the Free Software Foundation, either version 3 of the License, or |
| 9 | * (at your option) any later version. |
| 10 | * |
| 11 | * This program is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU General Public License |
| 17 | * along with this progam. If not, see <https://www.gnu.org/licenses/>. |
| 18 | */ |
| 19 | |
| 20 | #include "GrammarApplicator.hpp" |
| 21 | #include "Strings.hpp" |
| 22 | #include "Tag.hpp" |
| 23 | #include "Grammar.hpp" |
| 24 | #include "Window.hpp" |
| 25 | #include "SingleWindow.hpp" |
| 26 | #include "Reading.hpp" |
| 27 | #include "ContextualTest.hpp" |
| 28 | #include "version.hpp" |
| 29 | #include "process.hpp" |
| 30 | |
| 31 | namespace CG3 { |
| 32 | |
| 33 | enum { |
| 34 | RV_NOTHING = 1, |
| 35 | RV_SOMETHING = 2, |
| 36 | RV_DELIMITED = 4, |
| 37 | RV_TRACERULE = 8, |
| 38 | }; |
| 39 | |
| 40 | bool GrammarApplicator::doesWordformsMatch(const Tag* cword, const Tag* rword) { |
| 41 | if (rword && rword != cword) { |
| 42 | if (rword->type & T_REGEXP) { |
| 43 | if (!doesTagMatchRegexp(cword->hash, *rword)) { |
| 44 | return false; |
| 45 | } |
| 46 | } |
| 47 | else if (rword->type & T_CASE_INSENSITIVE) { |
| 48 | if (!doesTagMatchIcase(cword->hash, *rword)) { |
| 49 | return false; |
| 50 | } |
| 51 | } |
| 52 | else { |
| 53 | return false; |
| 54 | } |
| 55 | } |
| 56 | return true; |
| 57 | } |
| 58 | |
| 59 | bool GrammarApplicator::updateRuleToCohorts(Cohort& c, const uint32_t& rsit) { |
| 60 | // Check whether this rule is in the allowed rule list from cmdline flag --rule(s) |
| 61 | if (!valid_rules.empty() && !valid_rules.contains(rsit)) { |
| 62 | return false; |
| 63 | } |
| 64 | SingleWindow* current = c.parent; |
| 65 | const Rule* r = grammar->rule_by_number[rsit]; |
| 66 | if (!doesWordformsMatch(c.wordform, r->wordform)) { |
| 67 | return false; |
| 68 | } |
| 69 | if (current->rule_to_cohorts.size() < rsit+1) { |
| 70 | indexSingleWindow(*current); |
| 71 | } |
| 72 | CohortSet& cohortset = current->rule_to_cohorts[rsit]; |
| 73 | std::vector<size_t> csi; |
| 74 | for (size_t i = 0; i < cohortsets.size(); ++i) { |
| 75 | if (cohortsets[i] != &cohortset) { |
| 76 | continue; |
| 77 | } |
| 78 | csi.push_back(i); |
| 79 | } |
| 80 | if (!csi.empty()) { |
| 81 | auto cap = cohortset.capacity(); |
| 82 | std::vector<CohortSet::const_iterator*> ends; |
| 83 | std::vector<std::pair<CohortSet::const_iterator*,Cohort*>> chs; |
| 84 | for (size_t i = 0; i < csi.size(); ++i) { |
| 85 | if (*rocits[csi[i]] == cohortset.end()) { |
| 86 | ends.push_back(rocits[csi[i]]); |
| 87 | } |
| 88 | else { |
| 89 | chs.push_back(std::pair(rocits[csi[i]], **rocits[csi[i]])); |
| 90 | } |
| 91 | } |
| 92 | cohortset.insert(&c); |
| 93 | for (auto it : ends) { |
| 94 | *it = cohortset.end(); |
| 95 | } |
| 96 | if (cap != cohortset.capacity()) { |
| 97 | for (auto& it : chs) { |
| 98 | *it.first = cohortset.find(it.second); |
| 99 | } |
| 100 | } |
| 101 | } |
| 102 | else { |
| 103 | cohortset.insert(&c); |
| 104 | } |
| 105 | return current->valid_rules.insert(rsit); |
| 106 | } |
| 107 | |
| 108 | bool GrammarApplicator::updateValidRules(const uint32IntervalVector& rules, uint32IntervalVector& intersects, const uint32_t& hash, Reading& reading) { |
| 109 | size_t os = intersects.size(); |
| 110 | auto it = grammar->rules_by_tag.find(hash); |
| 111 | if (it != grammar->rules_by_tag.end()) { |
| 112 | Cohort& c = *(reading.parent); |
| 113 | for (auto rsit : (it->second)) { |
| 114 | if (updateRuleToCohorts(c, rsit) && rules.contains(rsit)) { |
| 115 | intersects.insert(rsit); |
| 116 | } |
| 117 | } |
| 118 | } |
| 119 | return (os != intersects.size()); |
| 120 | } |
| 121 | |
| 122 | void GrammarApplicator::indexSingleWindow(SingleWindow& current) { |
| 123 | current.valid_rules.clear(); |
| 124 | current.rule_to_cohorts.resize(grammar->rule_by_number.size()); |
| 125 | for (auto& cs : current.rule_to_cohorts) { |
| 126 | cs.clear(); |
| 127 | } |
| 128 | |
| 129 | for (auto c : current.cohorts) { |
| 130 | for (uint32_t psit = 0; psit < c->possible_sets.size(); ++psit) { |
| 131 | if (c->possible_sets.test(psit) == false) { |
| 132 | continue; |
| 133 | } |
| 134 | auto rules_it = grammar->rules_by_set.find(psit); |
| 135 | if (rules_it == grammar->rules_by_set.end()) { |
| 136 | continue; |
| 137 | } |
| 138 | for (auto rsit : rules_it->second) { |
| 139 | updateRuleToCohorts(*c, rsit); |
| 140 | } |
| 141 | } |
| 142 | } |
| 143 | } |
| 144 | |
| 145 | TagList GrammarApplicator::getTagList(const Set& theSet, bool unif_mode) const { |
| 146 | TagList theTags; |
| 147 | getTagList(theSet, theTags, unif_mode); |
| 148 | return theTags; |
| 149 | } |
| 150 | |
| 151 | void GrammarApplicator::getTagList(const Set& theSet, TagList& theTags, bool unif_mode) const { |
| 152 | if (theSet.type & ST_SET_UNIFY) { |
| 153 | const auto& usets = (*context_stack.back().unif_sets)[theSet.number]; |
| 154 | const Set& pSet = *(grammar->sets_list[theSet.sets[0]]); |
| 155 | for (auto iter : pSet.sets) { |
| 156 | if (usets.count(iter)) { |
| 157 | getTagList(*(grammar->sets_list[iter]), theTags); |
| 158 | } |
| 159 | } |
| 160 | } |
| 161 | else if (theSet.type & ST_TAG_UNIFY) { |
| 162 | for (auto iter : theSet.sets) { |
| 163 | getTagList(*(grammar->sets_list[iter]), theTags, true); |
| 164 | } |
| 165 | } |
| 166 | else if (!theSet.sets.empty()) { |
| 167 | for (auto iter : theSet.sets) { |
| 168 | getTagList(*(grammar->sets_list[iter]), theTags, unif_mode); |
| 169 | } |
| 170 | } |
| 171 | else if (unif_mode) { |
| 172 | auto unif_tags = context_stack.back().unif_tags; |
| 173 | auto iter = unif_tags->find(theSet.number); |
| 174 | if (iter != unif_tags->end()) { |
| 175 | trie_getTagList(theSet.trie, theTags, iter->second); |
| 176 | trie_getTagList(theSet.trie_special, theTags, iter->second); |
| 177 | } |
| 178 | } |
| 179 | else { |
| 180 | trie_getTagList(theSet.trie, theTags); |
| 181 | trie_getTagList(theSet.trie_special, theTags); |
| 182 | } |
| 183 | // Eliminate consecutive duplicates. Not all duplicates, since AddCohort and Append may have multiple readings with repeated tags |
| 184 | for (auto ot = theTags.begin(); theTags.size() > 1 && ot != theTags.end(); ++ot) { |
| 185 | auto it = ot; |
| 186 | ++it; |
| 187 | for (; it != theTags.end() && std::distance(ot, it) == 1;) { |
| 188 | if (*ot == *it) { |
| 189 | it = theTags.erase(it); |
| 190 | } |
| 191 | else { |
| 192 | ++it; |
| 193 | } |
| 194 | } |
| 195 | } |
| 196 | } |
| 197 | |
| 198 | Reading* GrammarApplicator::get_sub_reading(Reading* tr, int sub_reading) { |
| 199 | if (sub_reading == 0) { |
| 200 | return tr; |
| 201 | } |
| 202 | |
| 203 | if (sub_reading == GSR_ANY) { |
| 204 | // If there aren't any sub-readings, the primary reading is the same as the amalgamation of all readings |
| 205 | if (tr->next == nullptr) { |
| 206 | return tr; |
| 207 | } |
| 208 | |
| 209 | subs_any.emplace_back(Reading()); |
| 210 | Reading* reading = &subs_any.back(); |
| 211 | *reading = *tr; |
| 212 | reading->next = nullptr; |
| 213 | while (tr->next) { |
| 214 | tr = tr->next; |
| 215 | reading->tags_list.push_back(0); |
| 216 | reading->tags_list.insert(reading->tags_list.end(), tr->tags_list.begin(), tr->tags_list.end()); |
| 217 | for (auto tag : tr->tags) { |
| 218 | reading->tags.insert(tag); |
| 219 | reading->tags_bloom.insert(tag); |
| 220 | } |
| 221 | for (auto tag : tr->tags_plain) { |
| 222 | reading->tags_plain.insert(tag); |
| 223 | reading->tags_plain_bloom.insert(tag); |
| 224 | } |
| 225 | for (auto tag : tr->tags_textual) { |
| 226 | reading->tags_textual.insert(tag); |
| 227 | reading->tags_textual_bloom.insert(tag); |
| 228 | } |
| 229 | reading->tags_numerical.insert(tr->tags_numerical.begin(), tr->tags_numerical.end()); |
| 230 | if (tr->mapped) { |
| 231 | reading->mapped = true; |
| 232 | } |
| 233 | if (tr->mapping) { |
| 234 | reading->mapping = tr->mapping; |
| 235 | } |
| 236 | if (tr->matched_target) { |
| 237 | reading->matched_target = true; |
| 238 | } |
| 239 | if (tr->matched_tests) { |
| 240 | reading->matched_tests = true; |
| 241 | } |
| 242 | } |
| 243 | reading->rehash(); |
| 244 | return reading; |
| 245 | } |
| 246 | |
| 247 | if (sub_reading > 0) { |
| 248 | for (int i = 0; i < sub_reading && tr; ++i) { |
| 249 | tr = tr->next; |
| 250 | } |
| 251 | } |
| 252 | else if (sub_reading < 0) { |
| 253 | int ntr = 0; |
| 254 | Reading* ttr = tr; |
| 255 | while (ttr) { |
| 256 | ttr = ttr->next; |
| 257 | --ntr; |
| 258 | } |
| 259 | if (!tr->next) { |
| 260 | tr = nullptr; |
| 261 | } |
| 262 | for (auto i = ntr; i < sub_reading && tr; ++i) { |
| 263 | tr = tr->next; |
| 264 | } |
| 265 | } |
| 266 | return tr; |
| 267 | } |
| 268 | |
| 269 | #define TRACE \ |
| 270 | do { \ |
| 271 | get_apply_to().subreading->hit_by.push_back(rule->number); \ |
| 272 | if (rule->sub_reading == 32767) { \ |
| 273 | get_apply_to().reading->hit_by.push_back(rule->number); \ |
| 274 | } \ |
| 275 | } while (0) |
| 276 | |
| 277 | #define FILL_TAG_LIST(taglist)do { Reading& reading = *get_apply_to().subreading; for ( auto it = (taglist)->begin(); it != (taglist)->end();) { if (reading.tags.find((*it)->hash) == reading.tags.end()) { auto tt = *it; it = (taglist)->erase(it); if (tt->type & T_SPECIAL) { if (context_stack.back().regexgrps == nullptr ) { context_stack.back().regexgrps = ®exgrps_store[used_regex ]; } auto stag = doesTagMatchReading(reading, *tt, false, true ); if (stag) { (taglist)->insert(it, grammar->single_tags .find(stag)->second); } } continue; } ++it; } } while (0) \ |
| 278 | do { \ |
| 279 | Reading& reading = *get_apply_to().subreading; \ |
| 280 | for (auto it = (taglist)->begin(); it != (taglist)->end();) { \ |
| 281 | if (reading.tags.find((*it)->hash) == reading.tags.end()) { \ |
| 282 | auto tt = *it; \ |
| 283 | it = (taglist)->erase(it); \ |
| 284 | if (tt->type & T_SPECIAL) { \ |
| 285 | if (context_stack.back().regexgrps == nullptr) { \ |
| 286 | context_stack.back().regexgrps = ®exgrps_store[used_regex]; \ |
| 287 | } \ |
| 288 | auto stag = doesTagMatchReading(reading, *tt, false, true); \ |
| 289 | if (stag) { \ |
| 290 | (taglist)->insert(it, grammar->single_tags.find(stag)->second); \ |
| 291 | } \ |
| 292 | } \ |
| 293 | continue; \ |
| 294 | } \ |
| 295 | ++it; \ |
| 296 | } \ |
| 297 | } while (0) |
| 298 | |
| 299 | #define FILL_TAG_LIST_RAW(taglist)do { Reading& reading = *get_apply_to().subreading; for ( auto& tt : *(taglist)) { if (tt->type & T_SPECIAL) { if (context_stack.back().regexgrps == nullptr) { context_stack .back().regexgrps = ®exgrps_store[used_regex]; } auto stag = doesTagMatchReading(reading, *tt, false, true); if (stag) { tt = grammar->single_tags.find(stag)->second; } } } } while (0) \ |
| 300 | do { \ |
| 301 | Reading& reading = *get_apply_to().subreading; \ |
| 302 | for (auto& tt : *(taglist)) { \ |
| 303 | if (tt->type & T_SPECIAL) { \ |
| 304 | if (context_stack.back().regexgrps == nullptr) { \ |
| 305 | context_stack.back().regexgrps = ®exgrps_store[used_regex]; \ |
| 306 | } \ |
| 307 | auto stag = doesTagMatchReading(reading, *tt, false, true); \ |
| 308 | if (stag) { \ |
| 309 | tt = grammar->single_tags.find(stag)->second; \ |
| 310 | } \ |
| 311 | } \ |
| 312 | } \ |
| 313 | } while (0) |
| 314 | |
| 315 | #define APPEND_TAGLIST_TO_READING(taglist, reading)do { for (auto tter : (taglist)) { while (tter->type & T_VARSTRING) { tter = generateVarstringTag(tter); } auto hash = tter->hash; if (tter->type & T_MAPPING || tter-> tag[0] == grammar->mapping_prefix) { mappings->push_back (tter); } else { hash = addTagToReading((reading), tter); } if (updateValidRules(rules, intersects, hash, reading)) { iter_rules = intersects.find(rule->number); iter_rules_end = intersects .end(); } } } while (0) \ |
| 316 | do { \ |
| 317 | for (auto tter : (taglist)) { \ |
| 318 | while (tter->type & T_VARSTRING) { \ |
| 319 | tter = generateVarstringTag(tter); \ |
| 320 | } \ |
| 321 | auto hash = tter->hash; \ |
| 322 | if (tter->type & T_MAPPING || tter->tag[0] == grammar->mapping_prefix) { \ |
| 323 | mappings->push_back(tter); \ |
| 324 | } \ |
| 325 | else { \ |
| 326 | hash = addTagToReading((reading), tter); \ |
| 327 | } \ |
| 328 | if (updateValidRules(rules, intersects, hash, reading)) { \ |
| 329 | iter_rules = intersects.find(rule->number); \ |
| 330 | iter_rules_end = intersects.end(); \ |
| 331 | } \ |
| 332 | } \ |
| 333 | } while (0) |
| 334 | |
| 335 | #define VARSTRINGIFY(tag)do { while ((tag)->type & T_VARSTRING) { (tag) = generateVarstringTag ((tag)); } } while (0) \ |
| 336 | do { \ |
| 337 | while ((tag)->type & T_VARSTRING) { \ |
| 338 | (tag) = generateVarstringTag((tag)); \ |
| 339 | } \ |
| 340 | } \ |
| 341 | while (0) |
| 342 | |
| 343 | |
| 344 | bool GrammarApplicator::runSingleRule(SingleWindow& current, const Rule& rule, RuleCallback reading_cb, RuleCallback cohort_cb) { |
| 345 | finish_cohort_loop = true; |
| 346 | bool anything_changed = false; |
| 347 | KEYWORDS type = rule.type; |
| 348 | const Set& set = *(grammar->sets_list[rule.target]); |
| 349 | CohortSet* cohortset = ¤t.rule_to_cohorts[rule.number]; |
| 350 | |
| 351 | auto override_cohortset = [&]() { |
| 352 | if (in_nested) { |
| 353 | if (!current.nested_rule_to_cohorts) { |
| 354 | current.nested_rule_to_cohorts.reset(new CohortSet()); |
| 355 | } |
| 356 | cohortset = current.nested_rule_to_cohorts.get(); |
| 357 | cohortset->clear(); |
| 358 | cohortset->insert(get_apply_to().cohort); |
| 359 | for (auto& t : set.trie_special) { |
| 360 | if (t.first->type & T_CONTEXT && t.first->context_ref_pos <= context_stack.back().context.size()) { |
| 361 | cohortset->insert(context_stack.back().context[t.first->context_ref_pos - 1]); |
| 362 | } |
| 363 | } |
| 364 | } |
| 365 | }; |
| 366 | override_cohortset(); |
| 367 | cohortsets.push_back(cohortset); |
| 368 | rocits.push_back(nullptr); |
| 369 | |
| 370 | scope_guard popper([&]() { |
| 371 | cohortsets.pop_back(); |
| 372 | rocits.pop_back(); |
| 373 | }); |
| 374 | |
| 375 | if (debug_level > 1) { |
| 376 | std::cerr << "DEBUG: " << cohortset->size() << "/" << current.cohorts.size() << " = " << double(cohortset->size()) / double(current.cohorts.size()) << std::endl; |
| 377 | } |
| 378 | for (auto rocit = cohortset->cbegin(); (!cohortset->empty()) && (rocit != cohortset->cend());) { |
| 379 | rocits.back() = &rocit; |
| 380 | Cohort* cohort = *rocit; |
| 381 | ++rocit; |
| 382 | |
| 383 | finish_reading_loop = true; |
| 384 | |
| 385 | if (debug_level > 1) { |
| 386 | std::cerr << "DEBUG: Trying cohort " << cohort->global_number << ":" << cohort->local_number << std::endl; |
| 387 | } |
| 388 | |
| 389 | // If the current cohort is the initial >>> one, skip it. |
| 390 | if (cohort->local_number == 0) { |
| 391 | continue; |
| 392 | } |
| 393 | // If the cohort is removed, skip it... |
| 394 | // Removed cohorts are still in the precalculated rule_to_cohorts map, |
| 395 | // and it would take time to go through the whole map searching for the cohort. |
| 396 | // Haven't tested whether it is worth it... |
| 397 | if (cohort->type & CT_REMOVED) { |
| 398 | continue; |
| 399 | } |
| 400 | |
| 401 | uint32_t c = cohort->local_number; |
| 402 | // If the cohort is temporarily unavailable due to parentheses, skip it. |
| 403 | if ((cohort->type & CT_ENCLOSED) || cohort->parent != ¤t) { |
| 404 | continue; |
| 405 | } |
| 406 | // If there are no readings, skip it. |
| 407 | // This is unlikely to happen as all cohorts will get a magic reading during input, |
| 408 | // and not many use the unsafe Remove rules. |
| 409 | if (cohort->readings.empty()) { |
| 410 | continue; |
| 411 | } |
| 412 | // If there's no reason to even attempt to restore, just skip it. |
| 413 | if (rule.type == K_RESTORE) { |
| 414 | if ((rule.flags & RF_DELAYED) && cohort->delayed.empty()) { |
| 415 | continue; |
| 416 | } |
| 417 | else if ((rule.flags & RF_IGNORED) && cohort->ignored.empty()) { |
| 418 | continue; |
| 419 | } |
| 420 | else if (!(rule.flags & (RF_DELAYED|RF_IGNORED)) && cohort->deleted.empty()) { |
| 421 | continue; |
| 422 | } |
| 423 | } |
| 424 | // If there is not even a remote chance the target set might match this cohort, skip it. |
| 425 | if (rule.sub_reading == 0 && (rule.target >= cohort->possible_sets.size() || !cohort->possible_sets.test(rule.target))) { |
| 426 | continue; |
| 427 | } |
| 428 | |
| 429 | // If there is only 1 reading left and it is a Select or safe Remove rule, skip it. |
| 430 | if (cohort->readings.size() == 1) { |
| 431 | if (type == K_SELECT) { |
| 432 | continue; |
| 433 | } |
| 434 | if (type == K_REMOVE || type == K_IFF) { |
| 435 | if (cohort->readings.front()->noprint) { |
| 436 | continue; |
| 437 | } |
| 438 | if ((!unsafe || (rule.flags & RF_SAFE)) && !(rule.flags & RF_UNSAFE)) { |
| 439 | continue; |
| 440 | } |
| 441 | } |
| 442 | } |
| 443 | else if (type == K_UNMAP && rule.flags & RF_SAFE) { |
| 444 | continue; |
| 445 | } |
| 446 | // If it's a Delimit rule and we're at the final cohort, skip it. |
| 447 | if (type == K_DELIMIT && c == current.cohorts.size() - 1) { |
| 448 | continue; |
| 449 | } |
| 450 | |
| 451 | // If the rule is only supposed to run inside a parentheses, check if cohort is. |
| 452 | if (rule.flags & RF_ENCL_INNER) { |
| 453 | if (!par_left_pos) { |
| 454 | continue; |
| 455 | } |
| 456 | if (cohort->local_number < par_left_pos || cohort->local_number > par_right_pos) { |
| 457 | continue; |
| 458 | } |
| 459 | } |
| 460 | // ...and if the rule should only run outside parentheses, check if cohort is. |
| 461 | else if (rule.flags & RF_ENCL_OUTER) { |
| 462 | if (par_left_pos && cohort->local_number >= par_left_pos && cohort->local_number <= par_right_pos) { |
| 463 | continue; |
| 464 | } |
| 465 | } |
| 466 | |
| 467 | // If this is SETPARENT SAFE and there's already a parent, skip it. |
| 468 | if (type == K_SETPARENT && (rule.flags & RF_SAFE) && cohort->dep_parent != DEP_NO_PARENT) { |
| 469 | continue; |
| 470 | } |
| 471 | if ((rule.flags & RF_NOPARENT) && cohort->dep_parent != DEP_NO_PARENT) { |
| 472 | continue; |
| 473 | } |
| 474 | |
| 475 | // Check if on previous runs the rule did not match this cohort, and skip if that is the case. |
| 476 | // This cache is cleared if any rule causes any state change in the window. |
| 477 | uint32_t ih = hash_value(rule.number, cohort->global_number); |
| 478 | if (index_ruleCohort_no.contains(ih)) { |
| 479 | continue; |
| 480 | } |
| 481 | index_ruleCohort_no.insert(ih); |
| 482 | |
| 483 | size_t num_active = 0; |
| 484 | size_t num_iff = 0; |
| 485 | |
| 486 | std::vector<Rule_Context> reading_contexts; |
| 487 | reading_contexts.reserve(cohort->readings.size()); |
| 488 | |
| 489 | // Assume that Iff rules are really Remove rules, until proven otherwise. |
| 490 | if (rule.type == K_IFF) { |
| 491 | type = K_REMOVE; |
| 492 | } |
| 493 | |
| 494 | bool did_test = false; |
| 495 | bool test_good = false; |
| 496 | bool matched_target = false; |
| 497 | |
| 498 | clear(readings_plain); |
| 499 | clear(subs_any); |
| 500 | |
| 501 | // Varstring capture groups exist on a per-cohort basis, since we may need them for mapping later. |
| 502 | clear(regexgrps_z); |
| 503 | clear(regexgrps_c); |
| 504 | clear(unif_tags_rs); |
| 505 | clear(unif_sets_rs); |
| 506 | |
| 507 | used_regex = 0; |
| 508 | regexgrps_store.resize(std::max(regexgrps_store.size(), cohort->readings.size())); |
| 509 | regexgrps_z.reserve(std::max(regexgrps_z.size(), cohort->readings.size())); |
| 510 | regexgrps_c.reserve(std::max(regexgrps_c.size(), cohort->readings.size())); |
| 511 | |
| 512 | size_t used_unif = 0; |
| 513 | unif_tags_store.resize(std::max(unif_tags_store.size(), cohort->readings.size() + 1)); |
| 514 | unif_sets_store.resize(std::max(unif_sets_store.size(), cohort->readings.size() + 1)); |
| 515 | |
| 516 | { |
| 517 | Rule_Context context; |
| 518 | context.target.cohort = cohort; |
| 519 | context_stack.push_back(std::move(context)); |
| 520 | } |
| 521 | |
| 522 | auto reset_cohorts = [&]() { |
| 523 | cohortset = ¤t.rule_to_cohorts[rule.number]; |
| 524 | override_cohortset(); |
| 525 | cohortsets.back() = cohortset; |
| 526 | if (get_apply_to().cohort->type & CT_REMOVED) { |
| 527 | rocit = cohortset->lower_bound(current.cohorts[get_apply_to().cohort->local_number]); |
| 528 | } |
| 529 | else { |
| 530 | rocit = cohortset->find(current.cohorts[get_apply_to().cohort->local_number]); |
| 531 | if (rocit != cohortset->end()) { |
| 532 | ++rocit; |
| 533 | } |
| 534 | } |
| 535 | }; |
| 536 | |
| 537 | // Remember the current state so we can compare later to see if anything has changed |
| 538 | const size_t state_num_readings = cohort->readings.size(); |
| 539 | const size_t state_num_removed = cohort->deleted.size(); |
| 540 | const size_t state_num_delayed = cohort->delayed.size(); |
| 541 | const size_t state_num_ignored = cohort->ignored.size(); |
| 542 | |
| 543 | // This loop figures out which readings, if any, that are valid targets for the current rule |
| 544 | // Criteria for valid is that the reading must match both target and all contextual tests |
| 545 | for (size_t i = 0; i < cohort->readings.size(); ++i) { |
| 546 | // ToDo: Switch sub-readings so that they build up a passed in vector<Reading*> |
| 547 | Reading* reading = get_sub_reading(cohort->readings[i], rule.sub_reading); |
| 548 | if (!reading) { |
| 549 | cohort->readings[i]->matched_target = false; |
| 550 | cohort->readings[i]->matched_tests = false; |
| 551 | continue; |
| 552 | } |
| 553 | context_stack.back().target.reading = cohort->readings[i]; |
| 554 | context_stack.back().target.subreading = reading; |
| 555 | |
| 556 | // The state is stored in the readings themselves, so clear the old states |
| 557 | reading->matched_target = false; |
| 558 | reading->matched_tests = false; |
| 559 | |
| 560 | if (reading->mapped && (rule.type == K_MAP || rule.type == K_ADD || rule.type == K_REPLACE)) { |
| 561 | continue; |
| 562 | } |
| 563 | if (reading->mapped && (rule.flags & RF_NOMAPPED)) { |
| 564 | continue; |
| 565 | } |
| 566 | if (reading->noprint && !allow_magic_readings) { |
| 567 | continue; |
| 568 | } |
| 569 | if (reading->immutable && rule.type != K_UNPROTECT) { |
| 570 | if (type == K_SELECT) { |
| 571 | reading->matched_target = true; |
| 572 | reading->matched_tests = true; |
| 573 | reading_contexts.push_back(context_stack.back()); |
| 574 | } |
| 575 | ++num_active; |
| 576 | ++num_iff; |
| 577 | continue; |
| 578 | } |
| 579 | |
| 580 | // Check if any previous reading of this cohort had the same plain signature, and if so just copy their results |
| 581 | // This cache is cleared on a per-cohort basis |
| 582 | did_test = false; |
| 583 | if (!(set.type & (ST_SPECIAL | ST_MAPPING | ST_CHILD_UNIFY)) && !readings_plain.empty()) { |
| 584 | auto rpit = readings_plain.find(reading->hash_plain); |
| 585 | if (rpit != readings_plain.end()) { |
| 586 | reading->matched_target = rpit->second->matched_target; |
| 587 | reading->matched_tests = rpit->second->matched_tests; |
| 588 | if (reading->matched_tests) { |
| 589 | ++num_active; |
| 590 | } |
| 591 | if (regexgrps_c.count(rpit->second->number)) { |
| 592 | regexgrps_c[reading->number]; |
| 593 | regexgrps_c[reading->number] = regexgrps_c[rpit->second->number]; |
| 594 | regexgrps_z[reading->number]; |
| 595 | regexgrps_z[reading->number] = regexgrps_z[rpit->second->number]; |
| 596 | |
| 597 | context_stack.back().regexgrp_ct = regexgrps_z[reading->number]; |
| 598 | context_stack.back().regexgrps = regexgrps_c[reading->number]; |
| 599 | } |
| 600 | context_stack.back().unif_tags = unif_tags_rs[reading->hash_plain]; |
| 601 | context_stack.back().unif_sets = unif_sets_rs[reading->hash_plain]; |
| 602 | did_test = true; |
| 603 | test_good = rpit->second->matched_tests; |
| 604 | reading_contexts.push_back(context_stack.back()); |
| 605 | continue; |
| 606 | } |
| 607 | } |
| 608 | |
| 609 | // Regex capture is done on a per-reading basis, so clear all captured state. |
| 610 | context_stack.back().regexgrp_ct = 0; |
| 611 | context_stack.back().regexgrps = ®exgrps_store[used_regex]; |
| 612 | |
| 613 | // Unification is done on a per-reading basis, so clear all unification state. |
| 614 | context_stack.back().unif_tags = &unif_tags_store[used_unif]; |
| 615 | context_stack.back().unif_sets = &unif_sets_store[used_unif]; |
| 616 | unif_tags_rs[reading->hash_plain] = context_stack.back().unif_tags; |
| 617 | unif_sets_rs[reading->hash_plain] = context_stack.back().unif_sets; |
| 618 | unif_tags_rs[reading->hash] = context_stack.back().unif_tags; |
| 619 | unif_sets_rs[reading->hash] = context_stack.back().unif_sets; |
| 620 | ++used_unif; |
| 621 | |
| 622 | context_stack.back().unif_tags->clear(); |
| 623 | context_stack.back().unif_sets->clear(); |
| 624 | |
| 625 | unif_last_wordform = 0; |
| 626 | unif_last_baseform = 0; |
| 627 | unif_last_textual = 0; |
| 628 | |
| 629 | same_basic = reading->hash_plain; |
| 630 | rule_target = context_target = nullptr; |
| 631 | if (context_stack.size() > 1) { |
| 632 | Cohort* m = context_stack[context_stack.size()-2].mark; |
| 633 | if (m) set_mark(m); |
| 634 | else set_mark(cohort); |
| 635 | } |
| 636 | else { |
| 637 | set_mark(cohort); |
| 638 | } |
| 639 | uint8_t orz = context_stack.back().regexgrp_ct; |
| 640 | for (auto r = cohort->readings[i]; r; r = r->next) { |
| 641 | r->active = true; |
| 642 | } |
| 643 | if (rule.line == 2746) { |
| 644 | cohort = cohort; |
| 645 | } |
| 646 | rule_target = cohort; |
| 647 | // Actually check if the reading is a valid target. First check if rule target matches... |
| 648 | if (rule.target && doesSetMatchReading(*reading, rule.target, (set.type & (ST_CHILD_UNIFY | ST_SPECIAL)) != 0)) { |
| 649 | if (rule.line == 2746) { |
| 650 | cohort = cohort; |
| 651 | } |
| 652 | bool regex_prop = true; |
| 653 | if (orz != context_stack.back().regexgrp_ct) { |
| 654 | did_test = false; |
| 655 | regex_prop = false; |
| 656 | } |
| 657 | rule_target = context_target = cohort; |
| 658 | reading->matched_target = true; |
| 659 | matched_target = true; |
| 660 | bool good = true; |
| 661 | // If we didn't already run the contextual tests, run them now. |
| 662 | if (!did_test) { |
| 663 | context_stack.back().context.clear(); |
| 664 | foreach (it, rule.tests)if (!(rule.tests).empty()) for (auto it = (rule.tests).begin( ), it_end = (rule.tests).end(); it != it_end; ++it) { |
| 665 | ContextualTest* test = *it; |
| 666 | if (rule.flags & RF_RESETX || !(rule.flags & RF_REMEMBERX)) { |
| 667 | set_mark(cohort); |
| 668 | } |
| 669 | seen_barrier = false; |
| 670 | // Keeps track of where we have been, to prevent infinite recursion in trees with loops |
| 671 | dep_deep_seen.clear(); |
| 672 | // Reset the counters for which types of CohortIterator we have in play |
| 673 | std::fill(ci_depths.begin(), ci_depths.end(), UI32(0)); |
| 674 | tmpl_cntx.clear(); |
| 675 | // Run the contextual test... |
| 676 | Cohort* next_test = nullptr; |
| 677 | Cohort* result = nullptr; |
| 678 | Cohort** deep = nullptr; |
| 679 | if (rule.type == K_WITH) { |
| 680 | deep = &result; |
| 681 | merge_with = nullptr; |
| 682 | } |
| 683 | if (!(test->pos & POS_PASS_ORIGIN) && (no_pass_origin || (test->pos & POS_NO_PASS_ORIGIN))) { |
| 684 | next_test = runContextualTest(¤t, c, test, deep, cohort); |
| 685 | } |
| 686 | else { |
| 687 | next_test = runContextualTest(¤t, c, test, deep); |
| 688 | } |
| 689 | context_stack.back().context.push_back(merge_with ? merge_with : result); |
| 690 | test_good = (next_test != nullptr); |
| 691 | |
| 692 | profileRuleContext(test_good, &rule, test); |
| 693 | |
| 694 | if (!test_good) { |
| 695 | good = test_good; |
| 696 | if (it != rule.tests.begin() && !(rule.flags & RF_KEEPORDER)) { |
| 697 | rule.tests.erase(it); |
| 698 | rule.tests.push_front(test); |
| 699 | } |
| 700 | break; |
| 701 | } |
| 702 | did_test = ((set.type & (ST_CHILD_UNIFY | ST_SPECIAL)) == 0 && context_stack.back().unif_tags->empty() && context_stack.back().unif_sets->empty()); |
Value stored to 'did_test' is never read | |
| 703 | } |
| 704 | } |
| 705 | else { |
| 706 | good = test_good; |
| 707 | } |
| 708 | if (good) { |
| 709 | // We've found a match, so Iff should be treated as Select instead of Remove |
| 710 | if (rule.type == K_IFF && type != K_SELECT) { |
| 711 | type = K_SELECT; |
| 712 | if (grammar->has_protect) { |
| 713 | for (size_t j = 0; j < i; ++j) { |
| 714 | Reading* reading = get_sub_reading(cohort->readings[j], rule.sub_reading); |
| 715 | if (reading && reading->immutable) { |
| 716 | reading->matched_target = true; |
| 717 | reading->matched_tests = true; |
| 718 | ++num_active; |
| 719 | ++num_iff; |
| 720 | } |
| 721 | } |
| 722 | } |
| 723 | } |
| 724 | reading->matched_tests = true; |
| 725 | ++num_active; |
| 726 | if (profiler) { |
| 727 | Profiler::Key k{ET_RULE, rule.number + 1 }; |
| 728 | auto& r = profiler->entries[k]; |
| 729 | ++r.num_match; |
| 730 | if (!r.example_window) { |
| 731 | addProfilingExample(r); |
| 732 | } |
| 733 | } |
| 734 | if (!debug_rules.empty() && debug_rules.contains(rule.line)) { |
| 735 | printDebugRule(rule); |
| 736 | } |
| 737 | |
| 738 | if (regex_prop && i && !regexgrps_c.empty()) { |
| 739 | for (auto z = i; z > 0; --z) { |
| 740 | auto it = regexgrps_c.find(cohort->readings[z - 1]->number); |
| 741 | if (it != regexgrps_c.end()) { |
| 742 | regexgrps_c.insert(std::make_pair(reading->number, it->second)); |
| 743 | regexgrps_z.insert(std::make_pair(reading->number, regexgrps_z.find(cohort->readings[z - 1]->number)->second)); |
| 744 | break; |
| 745 | } |
| 746 | } |
| 747 | } |
| 748 | } |
| 749 | else { |
| 750 | context_stack.back().regexgrp_ct = orz; |
| 751 | if (!debug_rules.empty() && debug_rules.contains(rule.line)) { |
| 752 | printDebugRule(rule, true, false); |
| 753 | } |
| 754 | } |
| 755 | ++num_iff; |
| 756 | } |
| 757 | else { |
| 758 | context_stack.back().regexgrp_ct = orz; |
| 759 | if (profiler) { |
| 760 | Profiler::Key k{ ET_RULE, rule.number + 1 }; |
| 761 | ++profiler->entries[k].num_fail; |
| 762 | } |
| 763 | if (!debug_rules.empty() && debug_rules.contains(rule.line)) { |
| 764 | printDebugRule(rule, false, false); |
| 765 | } |
| 766 | } |
| 767 | readings_plain.insert(std::make_pair(reading->hash_plain, reading)); |
| 768 | for (auto r = cohort->readings[i]; r; r = r->next) { |
| 769 | r->active = false; |
| 770 | } |
| 771 | |
| 772 | if (reading != cohort->readings[i]) { |
| 773 | cohort->readings[i]->matched_target = reading->matched_target; |
| 774 | cohort->readings[i]->matched_tests = reading->matched_tests; |
| 775 | } |
| 776 | if (context_stack.back().regexgrp_ct) { |
| 777 | regexgrps_c[reading->number] = context_stack.back().regexgrps; |
| 778 | regexgrps_z[reading->number] = context_stack.back().regexgrp_ct; |
| 779 | ++used_regex; |
| 780 | } |
| 781 | reading_contexts.push_back(context_stack.back()); |
| 782 | } |
| 783 | |
| 784 | if (state_num_readings != cohort->readings.size() || state_num_removed != cohort->deleted.size() || state_num_delayed != cohort->delayed.size() || state_num_ignored != cohort->ignored.size()) { |
| 785 | anything_changed = true; |
| 786 | cohort->type &= ~CT_NUM_CURRENT; |
| 787 | } |
| 788 | |
| 789 | // If none of the readings were valid targets, remove this cohort from the rule's possible cohorts. |
| 790 | if (num_active == 0 && (num_iff == 0 || rule.type != K_IFF)) { |
| 791 | if (!matched_target) { |
| 792 | --rocit; // We have already incremented rocit earlier, so take one step back... |
| 793 | rocit = cohortset->erase(rocit); // ...and one step forward again |
| 794 | } |
| 795 | context_stack.pop_back(); |
| 796 | continue; |
| 797 | } |
| 798 | |
| 799 | // All readings were valid targets, which means there is nothing to do for Select or safe Remove rules. |
| 800 | if (num_active == cohort->readings.size()) { |
| 801 | if (type == K_SELECT) { |
| 802 | context_stack.pop_back(); |
| 803 | continue; |
| 804 | } |
| 805 | if (type == K_REMOVE && (!unsafe || (rule.flags & RF_SAFE)) && !(rule.flags & RF_UNSAFE)) { |
| 806 | context_stack.pop_back(); |
| 807 | continue; |
| 808 | } |
| 809 | } |
| 810 | |
| 811 | for (auto& ctx : reading_contexts) { |
| 812 | if (!ctx.target.subreading->matched_target) { |
| 813 | continue; |
| 814 | } |
| 815 | if (!ctx.target.subreading->matched_tests && rule.type != K_IFF) { |
| 816 | continue; |
| 817 | } |
| 818 | context_stack.back() = ctx; |
| 819 | reset_cohorts_for_loop = false; |
| 820 | reading_cb(); |
| 821 | if (!finish_cohort_loop) { |
| 822 | context_stack.pop_back(); |
| 823 | return anything_changed; |
| 824 | } |
| 825 | if (reset_cohorts_for_loop) { |
| 826 | reset_cohorts(); |
| 827 | break; |
| 828 | } |
| 829 | if (!finish_reading_loop) { |
| 830 | break; |
| 831 | } |
| 832 | } |
| 833 | |
| 834 | reset_cohorts_for_loop = false; |
| 835 | cohort_cb(); |
| 836 | if (!finish_cohort_loop) { |
| 837 | context_stack.pop_back(); |
| 838 | return anything_changed; |
| 839 | } |
| 840 | if (reset_cohorts_for_loop) { |
| 841 | reset_cohorts(); |
| 842 | } |
| 843 | context_stack.pop_back(); |
| 844 | } |
| 845 | return anything_changed; |
| 846 | } |
| 847 | |
| 848 | /** |
| 849 | * Applies the passed rules to the passed SingleWindow. |
| 850 | * |
| 851 | * This function is called at least N*M times where N is number of sections in the grammar and M is the number of windows in the input. |
| 852 | * Possibly many more times, since if a section changes the state of the window the section is run again. |
| 853 | * Only when no further changes are caused at a level does it progress to next level. |
| 854 | * |
| 855 | * The loops in this function are increasingly explosive, despite efforts to contain them. |
| 856 | * In the https://visl.sdu.dk/cg3_performance.html test data, this function is called 1015 times. |
| 857 | * The first loop (rules) is executed 3101728 times. |
| 858 | * The second loop (cohorts) is executed 11087278 times. |
| 859 | * The third loop (finding readings) is executed 11738927 times; of these, 1164585 (10%) match the rule target. |
| 860 | * The fourth loop (contextual test) is executed 1184009 times; of those, 1156322 (97%) fail their contexts. |
| 861 | * The fifth loop (acting on readings) is executed 41540 times. |
| 862 | * |
| 863 | * @param[in,out] current The window to apply rules on |
| 864 | * @param[in] rules The rules to apply |
| 865 | */ |
| 866 | uint32_t GrammarApplicator::runRulesOnSingleWindow(SingleWindow& current, const uint32IntervalVector& rules) { |
| 867 | uint32_t retval = RV_NOTHING; |
| 868 | bool section_did_something = false; |
| 869 | bool delimited = false; |
| 870 | |
| 871 | // ToDo: Now that numbering is used, can't this be made a normal max? Hm, maybe not since --sections can still force another order...but if we're smart, then we re-enumerate rules based on --sections |
| 872 | uint32IntervalVector intersects = current.valid_rules.intersect(rules); |
| 873 | ReadingList removed; |
| 874 | ReadingList selected; |
| 875 | |
| 876 | if (debug_level > 1) { |
| 877 | std::cerr << "DEBUG: Trying window " << current.number << std::endl; |
| 878 | } |
| 879 | |
| 880 | current.parent->cohort_map[0] = current.cohorts.front(); |
| 881 | |
| 882 | foreach (iter_rules, intersects)if (!(intersects).empty()) for (auto iter_rules = (intersects ).begin(), iter_rules_end = (intersects).end(); iter_rules != iter_rules_end; ++iter_rules) { |
| 883 | // Conditionally re-sort the rule-to-cohort mapping when the current rule is finished, regardless of how it finishes |
| 884 | struct Sorter { |
| 885 | SingleWindow& current; |
| 886 | bool do_sort = false; |
| 887 | |
| 888 | Sorter(SingleWindow& current) |
| 889 | : current(current) |
| 890 | {} |
| 891 | |
| 892 | ~Sorter() { |
| 893 | if (do_sort) { |
| 894 | for (auto& cs : current.rule_to_cohorts) { |
| 895 | cs.sort(); |
| 896 | } |
| 897 | } |
| 898 | } |
| 899 | } sorter(current); |
| 900 | |
| 901 | repeat_rule: |
| 902 | bool rule_did_something = false; |
| 903 | uint32_t j = (*iter_rules); |
| 904 | |
| 905 | // Check whether this rule is in the allowed rule list from cmdline flag --rule(s) |
| 906 | if (!valid_rules.empty() && !valid_rules.contains(j)) { |
| 907 | continue; |
| 908 | } |
| 909 | |
| 910 | current_rule = grammar->rule_by_number[j]; |
| 911 | Rule* rule = grammar->rule_by_number[j]; |
| 912 | if (rule->type == K_IGNORE) { |
| 913 | continue; |
| 914 | } |
| 915 | if (debug_level > 1) { |
| 916 | std::cerr << "DEBUG: Trying rule " << rule->line << std::endl; |
| 917 | } |
| 918 | |
| 919 | if (!apply_mappings && (rule->type == K_MAP || rule->type == K_ADD || rule->type == K_REPLACE)) { |
| 920 | continue; |
| 921 | } |
| 922 | if (!apply_corrections && (rule->type == K_SUBSTITUTE || rule->type == K_APPEND)) { |
| 923 | continue; |
| 924 | } |
| 925 | // If there are parentheses and the rule is marked as only run on the final pass, skip if this is not it. |
| 926 | if (current.has_enclosures) { |
| 927 | if ((rule->flags & RF_ENCL_FINAL) && !did_final_enclosure) { |
| 928 | continue; |
| 929 | } |
| 930 | if (did_final_enclosure && !(rule->flags & RF_ENCL_FINAL)) { |
| 931 | continue; |
| 932 | } |
| 933 | } |
| 934 | |
| 935 | bool readings_changed = false; |
| 936 | bool should_repeat = false; |
| 937 | bool should_bail = false; |
| 938 | |
| 939 | auto reindex = [&](SingleWindow* which = nullptr) { |
| 940 | if (!which) { |
| 941 | which = ¤t; |
| 942 | } |
| 943 | foreach (iter, which->cohorts)if (!(which->cohorts).empty()) for (auto iter = (which-> cohorts).begin(), iter_end = (which->cohorts).end(); iter != iter_end; ++iter) { |
| 944 | (*iter)->local_number = UI32(std::distance(which->cohorts.begin(), iter)); |
| 945 | } |
| 946 | gWindow->rebuildCohortLinks(); |
| 947 | }; |
| 948 | |
| 949 | auto collect_subtree = [&](CohortSet& cs, Cohort* head, uint32_t cset) { |
| 950 | if (cset) { |
| 951 | for (auto iter : current.cohorts) { |
| 952 | // Always consider the initial cohort a match |
| 953 | if (iter->global_number == head->global_number) { |
| 954 | cs.insert(iter); |
| 955 | } |
| 956 | else if (iter->dep_parent == head->global_number && doesSetMatchCohortNormal(*iter, cset)) { |
| 957 | cs.insert(iter); |
| 958 | } |
| 959 | } |
| 960 | CohortSet more; |
| 961 | for (auto iter : current.cohorts) { |
| 962 | for (auto cht : cs) { |
| 963 | // Do not grab the whole tree from the root, in case WithChild is not (*) |
| 964 | if (cht->global_number == head->global_number) { |
| 965 | continue; |
| 966 | } |
| 967 | if (isChildOf(iter, cht)) { |
| 968 | more.insert(iter); |
| 969 | } |
| 970 | } |
| 971 | } |
| 972 | cs.insert(more.begin(), more.end()); |
| 973 | } |
| 974 | else { |
| 975 | cs.insert(head); |
| 976 | } |
| 977 | }; |
| 978 | |
| 979 | auto add_cohort = [&](Cohort* cohort, size_t& spacesInAddedWf) { |
| 980 | Cohort* cCohort = alloc_cohort(¤t); |
| 981 | cCohort->global_number = gWindow->cohort_counter++; |
| 982 | |
| 983 | Tag* wf = nullptr; |
| 984 | std::vector<TagList> readings; |
| 985 | auto theTags = ss_taglist.get(); |
| 986 | getTagList(*rule->maplist, theTags); |
| 987 | |
| 988 | for (auto& tter : *theTags) { |
| 989 | if (tter->type & T_VSTR) { |
| 990 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 991 | } |
| 992 | } |
| 993 | |
| 994 | for (auto tter : *theTags) { |
| 995 | if(tter->type & T_WORDFORM) { |
| 996 | spacesInAddedWf = std::count_if(tter->tag.begin(), tter->tag.end(), [](UChar c){ return c == ' '; }); |
| 997 | } |
| 998 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 999 | if (tter->type & T_WORDFORM) { |
| 1000 | cCohort->wordform = tter; |
| 1001 | wf = tter; |
| 1002 | continue; |
| 1003 | } |
| 1004 | if (!wf) { |
| 1005 | u_fprintfu_fprintf_72(ux_stderr, "Error: There must be a wordform before any other tags in ADDCOHORT/MERGECOHORTS on line %u before input line %u.\n", rule->line, numLines); |
| 1006 | CG3Quit(1); |
| 1007 | } |
| 1008 | if (tter->type & T_BASEFORM) { |
| 1009 | readings.resize(readings.size() + 1); |
| 1010 | readings.back().push_back(wf); |
| 1011 | } |
| 1012 | if (readings.empty()) { |
| 1013 | u_fprintfu_fprintf_72(ux_stderr, "Error: There must be a baseform after the wordform in ADDCOHORT/MERGECOHORTS on line %u before input line %u.\n", rule->line, numLines); |
| 1014 | CG3Quit(1); |
| 1015 | } |
| 1016 | readings.back().push_back(tter); |
| 1017 | } |
| 1018 | |
| 1019 | for (auto& tags : readings) { |
| 1020 | for (size_t i = 0; i < tags.size(); ++i) { |
| 1021 | if (tags[i]->hash == grammar->tag_any) { |
| 1022 | auto& nt = cohort->readings.front()->tags_list; |
| 1023 | if (nt.size() <= 2) { |
| 1024 | continue; |
| 1025 | } |
| 1026 | tags.reserve(tags.size() + nt.size() - 2); |
| 1027 | tags[i] = grammar->single_tags[nt[2]]; |
| 1028 | for (size_t j = 3, k = 1; j < nt.size(); ++j) { |
| 1029 | if (grammar->single_tags[nt[j]]->type & T_DEPENDENCY) { |
| 1030 | continue; |
| 1031 | } |
| 1032 | tags.insert(tags.begin() + i + k, grammar->single_tags[nt[j]]); |
| 1033 | ++k; |
| 1034 | } |
| 1035 | } |
| 1036 | } |
| 1037 | } |
| 1038 | |
| 1039 | for (auto& rit : readings) { |
| 1040 | Reading* cReading = alloc_reading(cCohort); |
| 1041 | ++numReadings; |
| 1042 | insert_if_exists(cReading->parent->possible_sets, grammar->sets_any); |
| 1043 | cReading->hit_by.push_back(rule->number); |
| 1044 | cReading->noprint = false; |
| 1045 | TagList mappings; |
| 1046 | for (auto tter : rit) { |
| 1047 | uint32_t hash = tter->hash; |
| 1048 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 1049 | if (tter->type & T_MAPPING || tter->tag[0] == grammar->mapping_prefix) { |
| 1050 | mappings.push_back(tter); |
| 1051 | } |
| 1052 | else { |
| 1053 | hash = addTagToReading(*cReading, hash); |
| 1054 | } |
| 1055 | if (updateValidRules(rules, intersects, hash, *cReading)) { |
| 1056 | iter_rules = intersects.find(rule->number); |
| 1057 | iter_rules_end = intersects.end(); |
| 1058 | } |
| 1059 | } |
| 1060 | if (!mappings.empty()) { |
| 1061 | splitMappings(mappings, *cCohort, *cReading); |
| 1062 | } |
| 1063 | cCohort->appendReading(cReading); |
| 1064 | } |
| 1065 | |
| 1066 | current.parent->cohort_map[cCohort->global_number] = cCohort; |
| 1067 | current.parent->dep_window[cCohort->global_number] = cCohort; |
| 1068 | if (grammar->addcohort_attach && (rule->type == K_ADDCOHORT_BEFORE || rule->type == K_ADDCOHORT_AFTER)) { |
| 1069 | attachParentChild(*cohort, *cCohort); |
| 1070 | } |
| 1071 | |
| 1072 | if (cCohort->readings.empty()) { |
| 1073 | initEmptyCohort(*cCohort); |
| 1074 | if (trace) { |
| 1075 | auto r = cCohort->readings.front(); |
| 1076 | r->hit_by.push_back(rule->number); |
| 1077 | r->noprint = false; |
| 1078 | } |
| 1079 | } |
| 1080 | |
| 1081 | CohortSet cohorts; |
| 1082 | collect_subtree(cohorts, cohort, rule->childset1); |
| 1083 | |
| 1084 | if (rule->type == K_ADDCOHORT_BEFORE) { |
| 1085 | current.cohorts.insert(current.cohorts.begin() + cohorts.front()->local_number, cCohort); |
| 1086 | current.all_cohorts.insert(std::find(current.all_cohorts.begin() + cohorts.front()->local_number, current.all_cohorts.end(), cohorts.front()), cCohort); |
| 1087 | } |
| 1088 | else { |
| 1089 | current.cohorts.insert(current.cohorts.begin() + cohorts.back()->local_number + 1, cCohort); |
| 1090 | current.all_cohorts.insert(std::find(current.all_cohorts.begin() + cohorts.back()->local_number, current.all_cohorts.end(), cohorts.back()) + 1, cCohort); |
| 1091 | } |
| 1092 | |
| 1093 | foreach (iter, current.cohorts)if (!(current.cohorts).empty()) for (auto iter = (current.cohorts ).begin(), iter_end = (current.cohorts).end(); iter != iter_end ; ++iter) { |
| 1094 | (*iter)->local_number = UI32(std::distance(current.cohorts.begin(), iter)); |
| 1095 | } |
| 1096 | gWindow->rebuildCohortLinks(); |
| 1097 | |
| 1098 | return cCohort; |
| 1099 | }; |
| 1100 | |
| 1101 | auto rem_cohort = [&](Cohort* cohort) { |
| 1102 | auto& current = *cohort->parent; |
| 1103 | for (auto iter : cohort->readings) { |
| 1104 | iter->hit_by.push_back(rule->number); |
| 1105 | iter->deleted = true; |
| 1106 | if (trace) { |
| 1107 | iter->noprint = false; |
| 1108 | } |
| 1109 | } |
| 1110 | // Remove the cohort from all rules |
| 1111 | for (auto& cs : current.rule_to_cohorts) { |
| 1112 | cs.erase(cohort); |
| 1113 | } |
| 1114 | // Forward all children of this cohort to the parent of this cohort |
| 1115 | // ToDo: Named relations must be erased |
| 1116 | while (!cohort->dep_children.empty()) { |
| 1117 | uint32_t ch = cohort->dep_children.back(); |
| 1118 | if (cohort->dep_parent == DEP_NO_PARENT) { |
| 1119 | attachParentChild(*gWindow->cohort_map[0], *gWindow->cohort_map[ch], true, true); |
| 1120 | } |
| 1121 | else { |
| 1122 | attachParentChild(*gWindow->cohort_map[cohort->dep_parent], *gWindow->cohort_map[ch], true, true); |
| 1123 | } |
| 1124 | cohort->dep_children.erase(ch); |
| 1125 | } |
| 1126 | cohort->type |= CT_REMOVED; |
| 1127 | cohort->detach(); |
| 1128 | for (auto& cm : gWindow->cohort_map) { |
| 1129 | cm.second->dep_children.erase(cohort->dep_self); |
| 1130 | } |
| 1131 | gWindow->cohort_map.erase(cohort->global_number); |
| 1132 | current.cohorts.erase(current.cohorts.begin() + cohort->local_number); |
| 1133 | foreach (iter, current.cohorts)if (!(current.cohorts).empty()) for (auto iter = (current.cohorts ).begin(), iter_end = (current.cohorts).end(); iter != iter_end ; ++iter) { |
| 1134 | (*iter)->local_number = UI32(std::distance(current.cohorts.begin(), iter)); |
| 1135 | } |
| 1136 | |
| 1137 | if (current.cohorts.size() == 1 && ¤t != gWindow->current) { |
| 1138 | // This window is now empty, so remove it entirely from consideration so rules can look past it |
| 1139 | cohort = current.cohorts[0]; |
| 1140 | |
| 1141 | // Remove the cohort from all rules |
| 1142 | for (auto& cs : current.rule_to_cohorts) { |
| 1143 | cs.erase(cohort); |
| 1144 | } |
| 1145 | cohort->detach(); |
| 1146 | for (auto& cm : gWindow->cohort_map) { |
| 1147 | cm.second->dep_children.erase(cohort->dep_self); |
| 1148 | } |
| 1149 | gWindow->cohort_map.erase(cohort->global_number); |
| 1150 | free_cohort(cohort); |
| 1151 | |
| 1152 | if (current.previous) { |
| 1153 | current.previous->text += current.text + current.text_post; |
| 1154 | current.previous->all_cohorts.insert(current.previous->all_cohorts.end(), current.all_cohorts.begin() + 1, current.all_cohorts.end()); |
| 1155 | } |
| 1156 | else if (current.next) { |
| 1157 | current.next->text = current.text_post + current.next->text; |
| 1158 | current.next->all_cohorts.insert(current.previous->all_cohorts.begin() + 1, current.all_cohorts.begin() + 1, current.all_cohorts.end()); |
| 1159 | } |
| 1160 | current.all_cohorts.clear(); |
| 1161 | |
| 1162 | for (size_t i = 0; i < gWindow->previous.size(); ++i) { |
| 1163 | if (gWindow->previous[i] == ¤t) { |
| 1164 | free_swindow(gWindow->previous[i]); |
| 1165 | gWindow->previous.erase(gWindow->previous.begin() + i); |
| 1166 | break; |
| 1167 | } |
| 1168 | } |
| 1169 | for (size_t i = 0; i < gWindow->next.size(); ++i) { |
| 1170 | if (gWindow->next[i] == ¤t) { |
| 1171 | free_swindow(gWindow->next[i]); |
| 1172 | gWindow->next.erase(gWindow->next.begin() + i); |
| 1173 | break; |
| 1174 | } |
| 1175 | } |
| 1176 | |
| 1177 | gWindow->rebuildSingleWindowLinks(); |
| 1178 | } |
| 1179 | |
| 1180 | gWindow->rebuildCohortLinks(); |
| 1181 | }; |
| 1182 | |
| 1183 | auto ignore_cohort = [&](Cohort* cohort) { |
| 1184 | auto& current = *cohort->parent; |
| 1185 | for (auto iter : cohort->readings) { |
| 1186 | iter->hit_by.push_back(rule->number); |
| 1187 | } |
| 1188 | for (auto& cs : current.rule_to_cohorts) { |
| 1189 | cs.erase(cohort); |
| 1190 | } |
| 1191 | cohort->type |= CT_IGNORED; |
| 1192 | cohort->detach(); |
| 1193 | gWindow->cohort_map.erase(cohort->global_number); |
| 1194 | current.cohorts.erase(current.cohorts.begin() + cohort->local_number); |
| 1195 | }; |
| 1196 | |
| 1197 | auto make_relation_rtag = [&](Tag* tag, uint32_t id) { |
| 1198 | UChar tmp[256] = { 0 }; |
| 1199 | u_sprintfu_sprintf_72(tmp, "R:%S:%u", tag->tag.data(), id); |
| 1200 | auto nt = addTag(tmp); |
| 1201 | return nt; |
| 1202 | }; |
| 1203 | |
| 1204 | auto add_relation_rtag = [&](Cohort* cohort, Tag* tag, uint32_t id) { |
| 1205 | auto nt = make_relation_rtag(tag, id); |
| 1206 | for (auto& r : cohort->readings) { |
| 1207 | addTagToReading(*r, nt); |
| 1208 | } |
| 1209 | }; |
| 1210 | |
| 1211 | auto set_relation_rtag = [&](Cohort* cohort, Tag* tag, uint32_t id) { |
| 1212 | auto nt = make_relation_rtag(tag, id); |
| 1213 | for (auto& r : cohort->readings) { |
| 1214 | for (auto it = r->tags_list.begin(); it != r->tags_list.end();) { |
| 1215 | const auto& utag = grammar->single_tags[*it]->tag; |
| 1216 | if (utag[0] == 'R' && utag[1] == ':' && utag.size() > 2 + tag->tag.size() && utag[2 + tag->tag.size()] == ':' && utag.compare(2, tag->tag.size(), tag->tag) == 0) { |
| 1217 | r->tags.erase(*it); |
| 1218 | r->tags_textual.erase(*it); |
| 1219 | r->tags_numerical.erase(*it); |
| 1220 | r->tags_plain.erase(*it); |
| 1221 | it = r->tags_list.erase(it); |
| 1222 | } |
| 1223 | else { |
| 1224 | ++it; |
| 1225 | } |
| 1226 | } |
| 1227 | addTagToReading(*r, nt); |
| 1228 | } |
| 1229 | }; |
| 1230 | |
| 1231 | auto rem_relation_rtag = [&](Cohort* cohort, Tag* tag, uint32_t id) { |
| 1232 | auto nt = make_relation_rtag(tag, id); |
| 1233 | for (auto& r : cohort->readings) { |
| 1234 | delTagFromReading(*r, nt); |
| 1235 | } |
| 1236 | }; |
| 1237 | |
| 1238 | auto insert_taglist_to_reading = [&](auto& iter, auto& taglist, auto& reading, auto& mappings) { |
| 1239 | for (auto tag : taglist) { |
| 1240 | if (tag->type & T_VARSTRING) { |
| 1241 | tag = generateVarstringTag(tag); |
| 1242 | } |
| 1243 | if (tag->hash == grammar->tag_any) { |
| 1244 | break; |
| 1245 | } |
| 1246 | if (tag->type & T_MAPPING || tag->tag[0] == grammar->mapping_prefix) { |
| 1247 | mappings->push_back(tag); |
| 1248 | } |
| 1249 | else { |
| 1250 | iter = reading.tags_list.insert(iter, tag->hash); |
| 1251 | ++iter; |
| 1252 | } |
| 1253 | if (updateValidRules(rules, intersects, tag->hash, reading)) { |
| 1254 | iter_rules = intersects.find(rule->number); |
| 1255 | iter_rules_end = intersects.end(); |
| 1256 | } |
| 1257 | } |
| 1258 | reflowReading(reading); |
| 1259 | }; |
| 1260 | |
| 1261 | auto cohort_cb = [&]() { |
| 1262 | if (rule->type == K_SELECT || (rule->type == K_IFF && !selected.empty())) { |
| 1263 | Cohort* target = get_apply_to().cohort; |
| 1264 | if (selected.size() < target->readings.size() && !selected.empty()) { |
| 1265 | ReadingList drop; |
| 1266 | size_t si = 0; |
| 1267 | for (size_t ri = 0; ri < target->readings.size(); ri++) { |
| 1268 | // Manually trace, since reading_cb doesn't get called on non-matching readings |
| 1269 | Reading* rd = target->readings[ri]; |
| 1270 | if (rule->sub_reading != 32767) { |
| 1271 | rd = get_sub_reading(rd, rule->sub_reading); |
| 1272 | } |
| 1273 | if (rd) { |
| 1274 | rd->hit_by.push_back(rule->number); |
| 1275 | } |
| 1276 | if (si < selected.size() && target->readings[ri] == selected[si]) { |
| 1277 | si++; |
| 1278 | } |
| 1279 | else { |
| 1280 | target->readings[ri]->deleted = true; |
| 1281 | drop.push_back(target->readings[ri]); |
| 1282 | } |
| 1283 | } |
| 1284 | target->readings.swap(selected); |
| 1285 | if (rule->flags & RF_DELAYED) { |
| 1286 | target->delayed.insert(target->delayed.end(), drop.begin(), drop.end()); |
| 1287 | } |
| 1288 | else if (rule->flags & RF_IGNORED) { |
| 1289 | target->ignored.insert(target->ignored.end(), drop.begin(), drop.end()); |
| 1290 | } |
| 1291 | else { |
| 1292 | target->deleted.insert(target->deleted.end(), drop.begin(), drop.end()); |
| 1293 | } |
| 1294 | readings_changed = true; |
| 1295 | } |
| 1296 | selected.clear(); |
| 1297 | } |
| 1298 | else if (rule->type == K_REMOVE || rule->type == K_IFF) { |
| 1299 | if (!removed.empty() && (removed.size() < get_apply_to().cohort->readings.size() || (unsafe && !(rule->flags & RF_SAFE)) || (rule->flags & RF_UNSAFE))) { |
| 1300 | if (rule->flags & RF_DELAYED) { |
| 1301 | get_apply_to().cohort->delayed.insert(get_apply_to().cohort->delayed.end(), removed.begin(), removed.end()); |
| 1302 | } |
| 1303 | else if (rule->flags & RF_IGNORED) { |
| 1304 | get_apply_to().cohort->ignored.insert(get_apply_to().cohort->ignored.end(), removed.begin(), removed.end()); |
| 1305 | } |
| 1306 | else { |
| 1307 | get_apply_to().cohort->deleted.insert(get_apply_to().cohort->deleted.end(), removed.begin(), removed.end()); |
| 1308 | } |
| 1309 | size_t oz = get_apply_to().cohort->readings.size(); |
| 1310 | while (!removed.empty()) { |
| 1311 | removed.back()->deleted = true; |
| 1312 | for (size_t i = 0; i < oz; ++i) { |
| 1313 | if (get_apply_to().cohort->readings[i] == removed.back()) { |
| 1314 | --oz; |
| 1315 | std::swap(get_apply_to().cohort->readings[i], get_apply_to().cohort->readings[oz]); |
| 1316 | } |
| 1317 | } |
| 1318 | removed.pop_back(); |
| 1319 | } |
| 1320 | get_apply_to().cohort->readings.resize(oz); |
| 1321 | if (debug_level > 0) { |
| 1322 | std::cerr << "DEBUG: Rule " << rule->line << " hit cohort " << get_apply_to().cohort->local_number << std::endl; |
| 1323 | } |
| 1324 | readings_changed = true; |
| 1325 | } |
| 1326 | if (get_apply_to().cohort->readings.empty()) { |
| 1327 | initEmptyCohort(*get_apply_to().cohort); |
| 1328 | } |
| 1329 | selected.clear(); |
| 1330 | } |
| 1331 | else if (rule->type == K_JUMP) { |
| 1332 | auto to = getTagList(*rule->maplist).front(); |
| 1333 | VARSTRINGIFY(to)do { while ((to)->type & T_VARSTRING) { (to) = generateVarstringTag ((to)); } } while (0); |
| 1334 | auto it = grammar->anchors.find(to->hash); |
| 1335 | if (it == grammar->anchors.end()) { |
| 1336 | u_fprintfu_fprintf_72(ux_stderr, "Warning: JUMP on line %u could not find anchor '%S'.\n", rule->line, to->tag.data()); |
| 1337 | } |
| 1338 | else { |
| 1339 | iter_rules = intersects.lower_bound(it->second); |
| 1340 | finish_cohort_loop = false; |
| 1341 | should_repeat = true; |
| 1342 | } |
| 1343 | } |
| 1344 | else if (rule->type == K_REMVARIABLE) { |
| 1345 | auto names = getTagList(*rule->maplist); |
| 1346 | for (auto tag : names) { |
| 1347 | VARSTRINGIFY(tag)do { while ((tag)->type & T_VARSTRING) { (tag) = generateVarstringTag ((tag)); } } while (0); |
| 1348 | auto it = variables.begin(); |
| 1349 | if (tag->type & T_REGEXP) { |
| 1350 | it = std::find_if(it, variables.end(), [&](auto& kv) { return doesTagMatchRegexp(kv.first, *tag); }); |
| 1351 | } |
| 1352 | else if (tag->type & T_CASE_INSENSITIVE) { |
| 1353 | it = std::find_if(it, variables.end(), [&](auto& kv) { return doesTagMatchIcase(kv.first, *tag); }); |
| 1354 | } |
| 1355 | else { |
| 1356 | it = variables.find(tag->hash); |
| 1357 | } |
| 1358 | if (it != variables.end()) { |
| 1359 | if (rule->flags & RF_OUTPUT) { |
| 1360 | current.variables_output.insert(it->first); |
| 1361 | } |
| 1362 | variables.erase(it); |
| 1363 | //u_fprintf(ux_stderr, "Info: RemVariable fired for %S.\n", tag->tag.data()); |
| 1364 | } |
| 1365 | } |
| 1366 | } |
| 1367 | else if (rule->type == K_SETVARIABLE) { |
| 1368 | auto names = getTagList(*rule->maplist); |
| 1369 | auto values = getTagList(*rule->sublist); |
| 1370 | VARSTRINGIFY(names.front())do { while ((names.front())->type & T_VARSTRING) { (names .front()) = generateVarstringTag((names.front())); } } while ( 0); |
| 1371 | VARSTRINGIFY(values.front())do { while ((values.front())->type & T_VARSTRING) { (values .front()) = generateVarstringTag((values.front())); } } while (0); |
| 1372 | variables[names.front()->hash] = values.front()->hash; |
| 1373 | if (rule->flags & RF_OUTPUT) { |
| 1374 | current.variables_output.insert(names.front()->hash); |
| 1375 | } |
| 1376 | //u_fprintf(ux_stderr, "Info: SetVariable fired for %S.\n", names.front()->tag.data()); |
| 1377 | } |
| 1378 | else if (rule->type == K_DELIMIT) { |
| 1379 | auto cohort = get_apply_to().cohort; |
| 1380 | if (cohort->parent->cohorts.size() > cohort->local_number + 1) { |
| 1381 | delimitAt(current, cohort); |
| 1382 | delimited = true; |
| 1383 | readings_changed = true; |
| 1384 | } |
| 1385 | } |
| 1386 | else if (rule->type == K_EXTERNAL_ONCE || rule->type == K_EXTERNAL_ALWAYS) { |
| 1387 | if (rule->type == K_EXTERNAL_ONCE && !current.hit_external.insert(rule->line).second) { |
| 1388 | return; |
| 1389 | } |
| 1390 | |
| 1391 | auto ei = externals.find(rule->varname); |
| 1392 | if (ei == externals.end()) { |
| 1393 | Tag* ext = grammar->single_tags.find(rule->varname)->second; |
| 1394 | UErrorCode err = U_ZERO_ERROR; |
| 1395 | u_strToUTF8u_strToUTF8_72(&cbuffers[0][0], SI32(CG3_BUFFER_SIZE - 1), nullptr, ext->tag.data(), SI32(ext->tag.size()), &err); |
| 1396 | |
| 1397 | Process& es = externals[rule->varname]; |
| 1398 | try { |
| 1399 | es.start(&cbuffers[0][0]); |
| 1400 | writeRaw(es, CG3_EXTERNAL_PROTOCOL); |
| 1401 | } |
| 1402 | catch (std::exception& e) { |
| 1403 | u_fprintfu_fprintf_72(ux_stderr, "Error: External on line %u resulted in error: %s\n", rule->line, e.what()); |
| 1404 | CG3Quit(1); |
| 1405 | } |
| 1406 | ei = externals.find(rule->varname); |
| 1407 | } |
| 1408 | |
| 1409 | pipeOutSingleWindow(current, ei->second); |
| 1410 | pipeInSingleWindow(current, ei->second); |
| 1411 | |
| 1412 | indexSingleWindow(current); |
| 1413 | readings_changed = true; |
| 1414 | index_ruleCohort_no.clear(); |
| 1415 | intersects = current.valid_rules.intersect(rules); |
| 1416 | iter_rules = intersects.find(rule->number); |
| 1417 | iter_rules_end = intersects.end(); |
| 1418 | reset_cohorts_for_loop = true; |
| 1419 | } |
| 1420 | else if (rule->type == K_REMCOHORT) { |
| 1421 | // REMCOHORT-IGNORED |
| 1422 | if (rule->flags & RF_IGNORED) { |
| 1423 | CohortSet cohorts; |
| 1424 | collect_subtree(cohorts, get_apply_to().cohort, rule->childset1); |
| 1425 | for (auto c : reversed(cohorts)) { |
| 1426 | ignore_cohort(c); |
| 1427 | } |
| 1428 | reindex(); |
| 1429 | reflowDependencyWindow(); |
| 1430 | } |
| 1431 | else { |
| 1432 | rem_cohort(get_apply_to().cohort); |
| 1433 | } |
| 1434 | |
| 1435 | // If we just removed the last cohort, add <<< to the new last cohort |
| 1436 | if (get_apply_to().cohort->readings.front()->tags.count(endtag)) { |
| 1437 | for (auto r : current.cohorts.back()->readings) { |
| 1438 | addTagToReading(*r, endtag); |
| 1439 | if (updateValidRules(rules, intersects, endtag, *r)) { |
| 1440 | iter_rules = intersects.find(rule->number); |
| 1441 | iter_rules_end = intersects.end(); |
| 1442 | } |
| 1443 | } |
| 1444 | index_ruleCohort_no.clear(); |
| 1445 | } |
| 1446 | readings_changed = true; |
| 1447 | reset_cohorts_for_loop = true; |
| 1448 | } |
| 1449 | }; |
| 1450 | |
| 1451 | RuleCallback reading_cb = [&]() { |
| 1452 | if (rule->type == K_SELECT || (rule->type == K_IFF && get_apply_to().subreading->matched_tests)) { |
| 1453 | selected.push_back(get_apply_to().reading); |
| 1454 | index_ruleCohort_no.clear(); |
| 1455 | } |
| 1456 | else if (rule->type == K_REMOVE || rule->type == K_IFF) { |
| 1457 | if (rule->type == K_REMOVE && (rule->flags & RF_UNMAPLAST) && removed.size() == get_apply_to().cohort->readings.size() - 1) { |
| 1458 | if (unmapReading(*get_apply_to().subreading, rule->number)) { |
| 1459 | readings_changed = true; |
| 1460 | } |
| 1461 | } |
| 1462 | else { |
| 1463 | TRACE; |
| 1464 | removed.push_back(get_apply_to().reading); |
| 1465 | } |
| 1466 | index_ruleCohort_no.clear(); |
| 1467 | } |
| 1468 | else if (rule->type == K_PROTECT) { |
| 1469 | TRACE; |
| 1470 | get_apply_to().subreading->immutable = true; |
| 1471 | } |
| 1472 | else if (rule->type == K_UNPROTECT) { |
| 1473 | TRACE; |
| 1474 | get_apply_to().subreading->immutable = false; |
| 1475 | } |
| 1476 | else if (rule->type == K_UNMAP) { |
| 1477 | if (unmapReading(*get_apply_to().subreading, rule->number)) { |
| 1478 | index_ruleCohort_no.clear(); |
| 1479 | readings_changed = true; |
| 1480 | } |
| 1481 | } |
| 1482 | else if (rule->type == K_ADDCOHORT_AFTER || rule->type == K_ADDCOHORT_BEFORE) { |
| 1483 | index_ruleCohort_no.clear(); |
| 1484 | TRACE; |
| 1485 | |
| 1486 | size_t spacesInAddedWf = 0; // not used here |
| 1487 | auto cCohort = add_cohort(get_apply_to().cohort, spacesInAddedWf); |
| 1488 | |
| 1489 | // If the new cohort is now the last cohort, add <<< to it and remove <<< from previous last cohort |
| 1490 | if (current.cohorts.back() == cCohort) { |
| 1491 | for (auto r : current.cohorts[current.cohorts.size() - 2]->readings) { |
| 1492 | delTagFromReading(*r, endtag); |
| 1493 | } |
| 1494 | for (auto r : current.cohorts.back()->readings) { |
| 1495 | addTagToReading(*r, endtag); |
| 1496 | if (updateValidRules(rules, intersects, endtag, *r)) { |
| 1497 | iter_rules = intersects.find(rule->number); |
| 1498 | iter_rules_end = intersects.end(); |
| 1499 | } |
| 1500 | } |
| 1501 | } |
| 1502 | indexSingleWindow(current); |
| 1503 | readings_changed = true; |
| 1504 | |
| 1505 | reset_cohorts_for_loop = true; |
| 1506 | } |
| 1507 | else if (rule->type == K_SPLITCOHORT) { |
| 1508 | index_ruleCohort_no.clear(); |
| 1509 | |
| 1510 | std::vector<std::pair<Cohort*, std::vector<TagList>>> cohorts; |
| 1511 | |
| 1512 | auto theTags = ss_taglist.get(); |
| 1513 | getTagList(*rule->maplist, theTags); |
| 1514 | |
| 1515 | for (auto& tter : *theTags) { |
| 1516 | if (tter->type & T_VSTR) { |
| 1517 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 1518 | } |
| 1519 | } |
| 1520 | |
| 1521 | Tag* wf = nullptr; |
| 1522 | for (auto tter : *theTags) { |
| 1523 | if (tter->type & T_WORDFORM) { |
| 1524 | cohorts.resize(cohorts.size() + 1); |
| 1525 | cohorts.back().first = alloc_cohort(¤t); |
| 1526 | cohorts.back().first->global_number = gWindow->cohort_counter++; |
| 1527 | wf = tter; |
| 1528 | VARSTRINGIFY(wf)do { while ((wf)->type & T_VARSTRING) { (wf) = generateVarstringTag ((wf)); } } while (0); |
| 1529 | cohorts.back().first->wordform = wf; |
| 1530 | continue; |
| 1531 | } |
| 1532 | if (!wf) { |
| 1533 | u_fprintfu_fprintf_72(ux_stderr, "Error: There must be a wordform before any other tags in SPLITCOHORT on line %u before input line %u.\n", rule->line, numLines); |
| 1534 | CG3Quit(1); |
| 1535 | } |
| 1536 | } |
| 1537 | |
| 1538 | uint32_t rel_trg = DEP_NO_PARENT; |
| 1539 | std::vector<std::pair<uint32_t, uint32_t>> cohort_dep(cohorts.size()); |
| 1540 | cohort_dep.front().second = DEP_NO_PARENT; |
| 1541 | cohort_dep.back().first = DEP_NO_PARENT; |
| 1542 | cohort_dep.back().second = UI32(cohort_dep.size() - 1); |
| 1543 | for (size_t i = 1; i < cohort_dep.size() - 1; ++i) { |
| 1544 | cohort_dep[i].second = UI32(i); |
| 1545 | } |
| 1546 | |
| 1547 | size_t i = 0; |
| 1548 | std::vector<TagList>* readings = &cohorts.front().second; |
| 1549 | Tag* bf = nullptr; |
| 1550 | for (auto tter : *theTags) { |
| 1551 | if (tter->type & T_WORDFORM) { |
| 1552 | ++i; |
| 1553 | bf = nullptr; |
| 1554 | continue; |
| 1555 | } |
| 1556 | if (tter->type & T_BASEFORM) { |
| 1557 | readings = &cohorts[i - 1].second; |
| 1558 | readings->resize(readings->size() + 1); |
| 1559 | readings->back().push_back(cohorts[i - 1].first->wordform); |
| 1560 | bf = tter; |
| 1561 | } |
| 1562 | if (!bf) { |
| 1563 | u_fprintfu_fprintf_72(ux_stderr, "Error: There must be a baseform after the wordform in SPLITCOHORT on line %u before input line %u.\n", rule->line, numLines); |
| 1564 | CG3Quit(1); |
| 1565 | } |
| 1566 | |
| 1567 | UChar dep_self[12] = {}; |
| 1568 | UChar dep_parent[12] = {}; |
| 1569 | if (u_sscanfu_sscanf_72(tter->tag.data(), "%[0-9cd]->%[0-9pm]", &dep_self, &dep_parent) == 2) { |
| 1570 | if (dep_self[0] == 'c' || dep_self[0] == 'd') { |
| 1571 | cohort_dep[i - 1].first = DEP_NO_PARENT; |
| 1572 | if (rel_trg == DEP_NO_PARENT) { |
| 1573 | rel_trg = UI32(i - 1); |
| 1574 | } |
| 1575 | } |
| 1576 | else if (u_sscanfu_sscanf_72(dep_self, "%i", &cohort_dep[i - 1].first) != 1) { |
| 1577 | u_fprintfu_fprintf_72(ux_stderr, "Error: SPLITCOHORT dependency mapping dep_self was not valid on line %u before input line %u.\n", rule->line, numLines); |
| 1578 | CG3Quit(1); |
| 1579 | } |
| 1580 | if (dep_parent[0] == 'p' || dep_parent[0] == 'm') { |
| 1581 | cohort_dep[i - 1].second = DEP_NO_PARENT; |
| 1582 | } |
| 1583 | else if (u_sscanfu_sscanf_72(dep_parent, "%i", &cohort_dep[i - 1].second) != 1) { |
| 1584 | u_fprintfu_fprintf_72(ux_stderr, "Error: SPLITCOHORT dependency mapping dep_parent was not valid on line %u before input line %u.\n", rule->line, numLines); |
| 1585 | CG3Quit(1); |
| 1586 | } |
| 1587 | continue; |
| 1588 | } |
| 1589 | if (tter->tag.size() == 3 && tter->tag[0] == 'R' && tter->tag[1] == ':' && tter->tag[2] == '*') { |
| 1590 | rel_trg = UI32(i - 1); |
| 1591 | continue; |
| 1592 | } |
| 1593 | readings->back().push_back(tter); |
| 1594 | } |
| 1595 | |
| 1596 | if (rel_trg == DEP_NO_PARENT) { |
| 1597 | rel_trg = UI32(cohorts.size() - 1); |
| 1598 | } |
| 1599 | |
| 1600 | for (size_t i = 0; i < cohorts.size(); ++i) { |
| 1601 | Cohort* cCohort = cohorts[i].first; |
| 1602 | readings = &cohorts[i].second; |
| 1603 | |
| 1604 | for (auto tags : *readings) { |
| 1605 | Reading* cReading = alloc_reading(cCohort); |
| 1606 | ++numReadings; |
| 1607 | insert_if_exists(cReading->parent->possible_sets, grammar->sets_any); |
| 1608 | cReading->hit_by.push_back(rule->number); |
| 1609 | cReading->noprint = false; |
| 1610 | TagList mappings; |
| 1611 | |
| 1612 | for (size_t i = 0; i < tags.size(); ++i) { |
| 1613 | if (tags[i]->hash == grammar->tag_any) { |
| 1614 | uint32Vector& nt = get_apply_to().cohort->readings.front()->tags_list; |
| 1615 | if (nt.size() <= 2) { |
| 1616 | continue; |
| 1617 | } |
| 1618 | tags.reserve(tags.size() + nt.size() - 2); |
| 1619 | tags[i] = grammar->single_tags[nt[2]]; |
| 1620 | for (size_t j = 3, k = 1; j < nt.size(); ++j) { |
| 1621 | if (grammar->single_tags[nt[j]]->type & T_DEPENDENCY) { |
| 1622 | continue; |
| 1623 | } |
| 1624 | tags.insert(tags.begin() + i + k, grammar->single_tags[nt[j]]); |
| 1625 | ++k; |
| 1626 | } |
| 1627 | } |
| 1628 | } |
| 1629 | |
| 1630 | for (auto tter : tags) { |
| 1631 | uint32_t hash = tter->hash; |
| 1632 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 1633 | if (tter->type & T_MAPPING || tter->tag[0] == grammar->mapping_prefix) { |
| 1634 | mappings.push_back(tter); |
| 1635 | } |
| 1636 | else { |
| 1637 | hash = addTagToReading(*cReading, hash); |
| 1638 | } |
| 1639 | if (updateValidRules(rules, intersects, hash, *cReading)) { |
| 1640 | iter_rules = intersects.find(rule->number); |
| 1641 | iter_rules_end = intersects.end(); |
| 1642 | } |
| 1643 | } |
| 1644 | if (!mappings.empty()) { |
| 1645 | splitMappings(mappings, *cCohort, *cReading); |
| 1646 | } |
| 1647 | cCohort->appendReading(cReading); |
| 1648 | } |
| 1649 | |
| 1650 | if (cCohort->readings.empty()) { |
| 1651 | initEmptyCohort(*cCohort); |
| 1652 | } |
| 1653 | |
| 1654 | current.parent->dep_window[cCohort->global_number] = cCohort; |
| 1655 | current.parent->cohort_map[cCohort->global_number] = cCohort; |
| 1656 | |
| 1657 | current.cohorts.insert(current.cohorts.begin() + get_apply_to().cohort->local_number + i + 1, cCohort); |
| 1658 | current.all_cohorts.insert(std::find(current.all_cohorts.begin() + get_apply_to().cohort->local_number, current.all_cohorts.end(), get_apply_to().cohort) + i + 1, cCohort); |
| 1659 | } |
| 1660 | |
| 1661 | // Move text from the to-be-deleted cohort to the last new cohort |
| 1662 | std::swap(cohorts.back().first->text, get_apply_to().cohort->text); |
| 1663 | |
| 1664 | for (size_t i = 0; i < cohorts.size(); ++i) { |
| 1665 | Cohort* cCohort = cohorts[i].first; |
| 1666 | |
| 1667 | if (cohort_dep[i].first == DEP_NO_PARENT) { |
| 1668 | while (!get_apply_to().cohort->dep_children.empty()) { |
| 1669 | uint32_t ch = get_apply_to().cohort->dep_children.back(); |
| 1670 | attachParentChild(*cCohort, *current.parent->cohort_map[ch], true, true); |
| 1671 | get_apply_to().cohort->dep_children.erase(ch); // Just in case the attachment can't be made for some reason |
| 1672 | } |
| 1673 | } |
| 1674 | |
| 1675 | if (cohort_dep[i].second == DEP_NO_PARENT) { |
| 1676 | if (current.parent->cohort_map.count(get_apply_to().cohort->dep_parent)) { |
| 1677 | attachParentChild(*current.parent->cohort_map[get_apply_to().cohort->dep_parent], *cCohort, true, true); |
| 1678 | } |
| 1679 | } |
| 1680 | else { |
| 1681 | attachParentChild(*current.parent->cohort_map[cohorts.front().first->global_number + cohort_dep[i].second - 1], *cCohort, true, true); |
| 1682 | } |
| 1683 | |
| 1684 | // Re-attach all named relations to the dependency tail or R:* cohort |
| 1685 | if (rel_trg == i && (get_apply_to().cohort->type & CT_RELATED)) { |
| 1686 | cCohort->setRelated(); |
| 1687 | cCohort->relations.swap(get_apply_to().cohort->relations); |
| 1688 | |
| 1689 | std::pair<SingleWindow**, size_t> swss[3] = { |
| 1690 | std::make_pair(&gWindow->previous[0], gWindow->previous.size()), |
| 1691 | std::make_pair(&gWindow->current, static_cast<size_t>(1)), |
| 1692 | std::make_pair(&gWindow->next[0], gWindow->next.size()), |
| 1693 | }; |
| 1694 | for (auto sws : swss) { |
| 1695 | for (size_t sw = 0; sw < sws.second; ++sw) { |
| 1696 | for (auto ch : sws.first[sw]->cohorts) { |
| 1697 | for (auto& rel : ch->relations) { |
| 1698 | if (rel.second.count(get_apply_to().cohort->global_number)) { |
| 1699 | rel.second.erase(get_apply_to().cohort->global_number); |
| 1700 | rel.second.insert(cCohort->global_number); |
| 1701 | } |
| 1702 | } |
| 1703 | } |
| 1704 | } |
| 1705 | } |
| 1706 | } |
| 1707 | } |
| 1708 | |
| 1709 | // Remove the source cohort |
| 1710 | for (auto iter : get_apply_to().cohort->readings) { |
| 1711 | iter->hit_by.push_back(rule->number); |
| 1712 | iter->deleted = true; |
| 1713 | } |
| 1714 | get_apply_to().cohort->type |= CT_REMOVED; |
| 1715 | get_apply_to().cohort->detach(); |
| 1716 | for (auto& cm : current.parent->cohort_map) { |
| 1717 | cm.second->dep_children.erase(get_apply_to().cohort->dep_self); |
| 1718 | } |
| 1719 | current.parent->cohort_map.erase(get_apply_to().cohort->global_number); |
| 1720 | current.cohorts.erase(current.cohorts.begin() + get_apply_to().cohort->local_number); |
| 1721 | |
| 1722 | reindex(); |
| 1723 | indexSingleWindow(current); |
| 1724 | readings_changed = true; |
| 1725 | |
| 1726 | reset_cohorts_for_loop = true; |
| 1727 | } |
| 1728 | else if (rule->type == K_ADD || rule->type == K_MAP) { |
| 1729 | TRACE; |
| 1730 | auto state_hash = get_apply_to().subreading->hash; |
| 1731 | index_ruleCohort_no.clear(); |
| 1732 | auto& reading = *(get_apply_to().subreading); |
| 1733 | reading.noprint = false; |
| 1734 | auto mappings = ss_taglist.get(); |
| 1735 | auto theTags = ss_taglist.get(); |
| 1736 | getTagList(*rule->maplist, theTags); |
| 1737 | |
| 1738 | bool did_insert = false; |
| 1739 | if (rule->childset1) { |
| 1740 | bool found_spot = false; |
| 1741 | auto spot_tags = ss_taglist.get(); |
| 1742 | getTagList(*grammar->sets_list[rule->childset1], spot_tags); |
| 1743 | FILL_TAG_LIST(spot_tags)do { Reading& reading = *get_apply_to().subreading; for ( auto it = (spot_tags)->begin(); it != (spot_tags)->end( );) { if (reading.tags.find((*it)->hash) == reading.tags.end ()) { auto tt = *it; it = (spot_tags)->erase(it); if (tt-> type & T_SPECIAL) { if (context_stack.back().regexgrps == nullptr) { context_stack.back().regexgrps = ®exgrps_store [used_regex]; } auto stag = doesTagMatchReading(reading, *tt, false, true); if (stag) { (spot_tags)->insert(it, grammar ->single_tags.find(stag)->second); } } continue; } ++it ; } } while (0); |
| 1744 | auto it = reading.tags_list.begin(); |
| 1745 | for (; it != reading.tags_list.end(); ++it) { |
| 1746 | bool found = true; |
| 1747 | auto tmp = it; |
| 1748 | for (auto tag : *spot_tags) { |
| 1749 | if (*tmp != tag->hash) { |
| 1750 | found = false; |
| 1751 | break; |
| 1752 | } |
| 1753 | ++tmp; |
| 1754 | } |
| 1755 | if (found) { |
| 1756 | found_spot = true; |
| 1757 | break; |
| 1758 | } |
| 1759 | } |
| 1760 | if (found_spot) { |
| 1761 | if (rule->flags & RF_AFTER) { |
| 1762 | std::advance(it, spot_tags->size()); |
| 1763 | } |
| 1764 | if (it != reading.tags_list.end()) { |
| 1765 | insert_taglist_to_reading(it, *theTags, reading, mappings); |
| 1766 | did_insert = true; |
| 1767 | } |
| 1768 | } |
| 1769 | } |
| 1770 | |
| 1771 | if (!did_insert) { |
| 1772 | APPEND_TAGLIST_TO_READING(*theTags, reading)do { for (auto tter : (*theTags)) { while (tter->type & T_VARSTRING) { tter = generateVarstringTag(tter); } auto hash = tter->hash; if (tter->type & T_MAPPING || tter-> tag[0] == grammar->mapping_prefix) { mappings->push_back (tter); } else { hash = addTagToReading((reading), tter); } if (updateValidRules(rules, intersects, hash, reading)) { iter_rules = intersects.find(rule->number); iter_rules_end = intersects .end(); } } } while (0); |
| 1773 | } |
| 1774 | if (!mappings->empty()) { |
| 1775 | splitMappings(mappings, *get_apply_to().cohort, reading, rule->type == K_MAP); |
| 1776 | } |
| 1777 | if (rule->type == K_MAP) { |
| 1778 | reading.mapped = true; |
| 1779 | } |
| 1780 | if (reading.hash != state_hash) { |
| 1781 | readings_changed = true; |
| 1782 | } |
| 1783 | } |
| 1784 | else if (rule->type == K_RESTORE) { |
| 1785 | bool did_restore = false; |
| 1786 | auto move_rs = [&](ReadingList& rl) { |
| 1787 | for (size_t i = 0; i < rl.size();) { |
| 1788 | if (doesSetMatchReading(*rl[i], rule->maplist->number)) { |
| 1789 | rl[i]->deleted = false; |
| 1790 | rl[i]->hit_by.push_back(rule->number); |
| 1791 | get_apply_to().cohort->readings.push_back(rl[i]); |
| 1792 | rl.erase(rl.begin() + i); |
| 1793 | did_restore = true; |
| 1794 | } |
| 1795 | else { |
| 1796 | ++i; |
| 1797 | } |
| 1798 | } |
| 1799 | }; |
| 1800 | |
| 1801 | if (rule->flags & RF_DELAYED) { |
| 1802 | move_rs(get_apply_to().cohort->delayed); |
| 1803 | } |
| 1804 | else if (rule->flags & RF_IGNORED) { |
| 1805 | move_rs(get_apply_to().cohort->ignored); |
| 1806 | } |
| 1807 | else { |
| 1808 | move_rs(get_apply_to().cohort->deleted); |
| 1809 | } |
| 1810 | |
| 1811 | if (did_restore) { |
| 1812 | TRACE; |
| 1813 | } |
| 1814 | finish_reading_loop = false; |
| 1815 | } |
| 1816 | else if (rule->type == K_REPLACE) { |
| 1817 | auto state_hash = get_apply_to().subreading->hash; |
| 1818 | index_ruleCohort_no.clear(); |
| 1819 | TRACE; |
| 1820 | get_apply_to().subreading->noprint = false; |
| 1821 | get_apply_to().subreading->tags_list.clear(); |
| 1822 | get_apply_to().subreading->tags_list.push_back(get_apply_to().cohort->wordform->hash); |
| 1823 | get_apply_to().subreading->tags_list.push_back(get_apply_to().subreading->baseform); |
| 1824 | reflowReading(*get_apply_to().subreading); |
| 1825 | auto mappings = ss_taglist.get(); |
| 1826 | auto theTags = ss_taglist.get(); |
| 1827 | getTagList(*rule->maplist, theTags); |
| 1828 | |
| 1829 | APPEND_TAGLIST_TO_READING(*theTags, *get_apply_to().subreading)do { for (auto tter : (*theTags)) { while (tter->type & T_VARSTRING) { tter = generateVarstringTag(tter); } auto hash = tter->hash; if (tter->type & T_MAPPING || tter-> tag[0] == grammar->mapping_prefix) { mappings->push_back (tter); } else { hash = addTagToReading((*get_apply_to().subreading ), tter); } if (updateValidRules(rules, intersects, hash, *get_apply_to ().subreading)) { iter_rules = intersects.find(rule->number ); iter_rules_end = intersects.end(); } } } while (0); |
| 1830 | |
| 1831 | if (!mappings->empty()) { |
| 1832 | splitMappings(mappings, *get_apply_to().cohort, *get_apply_to().subreading, true); |
| 1833 | } |
| 1834 | if (get_apply_to().subreading->hash != state_hash) { |
| 1835 | readings_changed = true; |
| 1836 | } |
| 1837 | } |
| 1838 | else if (rule->type == K_SUBSTITUTE) { |
| 1839 | // ToDo: Check whether this substitution will do nothing at all to the end result |
| 1840 | // ToDo: Not actually...instead, test whether any reading in the cohort already is the end result |
| 1841 | |
| 1842 | auto state_hash = get_apply_to().subreading->hash; |
| 1843 | auto theTags = ss_taglist.get(); |
| 1844 | getTagList(*rule->sublist, theTags); |
| 1845 | |
| 1846 | // Modify the list of tags to remove to be the actual list of tags present, including matching regex and icase tags |
| 1847 | FILL_TAG_LIST(theTags)do { Reading& reading = *get_apply_to().subreading; for ( auto it = (theTags)->begin(); it != (theTags)->end();) { if (reading.tags.find((*it)->hash) == reading.tags.end()) { auto tt = *it; it = (theTags)->erase(it); if (tt->type & T_SPECIAL) { if (context_stack.back().regexgrps == nullptr ) { context_stack.back().regexgrps = ®exgrps_store[used_regex ]; } auto stag = doesTagMatchReading(reading, *tt, false, true ); if (stag) { (theTags)->insert(it, grammar->single_tags .find(stag)->second); } } continue; } ++it; } } while (0); |
| 1848 | |
| 1849 | // Perform the tag removal, remembering the position of the final removed tag for use as insertion spot |
| 1850 | size_t tpos = std::numeric_limits<size_t>::max(); |
| 1851 | bool plain = true; |
| 1852 | for (size_t i = 0; i < get_apply_to().subreading->tags_list.size();) { |
| 1853 | auto& remter = get_apply_to().subreading->tags_list[i]; |
| 1854 | |
| 1855 | if (plain && remter == (*theTags->begin())->hash) { |
| 1856 | if (get_apply_to().subreading->baseform == remter) { |
| 1857 | get_apply_to().subreading->baseform = 0; |
| 1858 | } |
| 1859 | remter = substtag; |
| 1860 | tpos = i; |
| 1861 | for (size_t j = 1; j < theTags->size() && i < get_apply_to().subreading->tags_list.size(); ++j, ++i) { |
| 1862 | auto& remter = get_apply_to().subreading->tags_list[i]; |
| 1863 | auto tter = (*theTags)[j]->hash; |
| 1864 | if (remter != tter) { |
| 1865 | plain = false; |
| 1866 | break; |
| 1867 | } |
| 1868 | get_apply_to().subreading->tags_list.erase(get_apply_to().subreading->tags_list.begin() + i); |
| 1869 | get_apply_to().subreading->tags.erase(tter); |
| 1870 | if (get_apply_to().subreading->baseform == tter) { |
| 1871 | get_apply_to().subreading->baseform = 0; |
| 1872 | } |
| 1873 | } |
| 1874 | continue; |
| 1875 | } |
| 1876 | |
| 1877 | for (auto tter : *theTags) { |
| 1878 | if (remter != tter->hash) { |
| 1879 | continue; |
| 1880 | } |
| 1881 | tpos = i; |
| 1882 | remter = substtag; |
| 1883 | get_apply_to().subreading->tags.erase(tter->hash); |
| 1884 | if (get_apply_to().subreading->baseform == tter->hash) { |
| 1885 | get_apply_to().subreading->baseform = 0; |
| 1886 | } |
| 1887 | } |
| 1888 | |
| 1889 | ++i; |
| 1890 | } |
| 1891 | |
| 1892 | // Should Substitute really do nothing if no tags were removed? 2013-10-21, Eckhard says this is expected behavior. |
| 1893 | if (tpos != std::numeric_limits<size_t>::max()) { |
| 1894 | if (!plain) { |
| 1895 | for (size_t i = 0; i < get_apply_to().subreading->tags_list.size() && i < tpos;) { |
| 1896 | if (get_apply_to().subreading->tags_list[i] == substtag) { |
| 1897 | get_apply_to().subreading->tags_list.erase(get_apply_to().subreading->tags_list.begin() + i); |
| 1898 | --tpos; |
| 1899 | } |
| 1900 | else { |
| 1901 | ++i; |
| 1902 | } |
| 1903 | } |
| 1904 | } |
| 1905 | |
| 1906 | Tag* wf = nullptr; |
| 1907 | index_ruleCohort_no.clear(); |
| 1908 | TRACE; |
| 1909 | get_apply_to().subreading->noprint = false; |
| 1910 | if (tpos >= get_apply_to().subreading->tags_list.size()) { |
| 1911 | tpos = get_apply_to().subreading->tags_list.size() - 1; |
| 1912 | } |
| 1913 | ++tpos; |
| 1914 | auto mappings = ss_taglist.get(); |
| 1915 | auto theTags = ss_taglist.get(); |
| 1916 | getTagList(*rule->maplist, theTags); |
| 1917 | |
| 1918 | for (size_t i = 0; i < get_apply_to().subreading->tags_list.size();) { |
| 1919 | if (get_apply_to().subreading->tags_list[i] == substtag) { |
| 1920 | get_apply_to().subreading->tags_list.erase(get_apply_to().subreading->tags_list.begin() + i); |
| 1921 | tpos = i; |
| 1922 | |
| 1923 | for (auto tag : *theTags) { |
| 1924 | if (tag->type & T_VARSTRING) { |
| 1925 | tag = generateVarstringTag(tag); |
| 1926 | } |
| 1927 | if (tag->hash == grammar->tag_any) { |
| 1928 | break; |
| 1929 | } |
| 1930 | if (tag->type & T_MAPPING || tag->tag[0] == grammar->mapping_prefix) { |
| 1931 | mappings->push_back(tag); |
| 1932 | } |
| 1933 | else { |
| 1934 | if (tag->type & T_WORDFORM) { |
| 1935 | wf = tag; |
| 1936 | } |
| 1937 | get_apply_to().subreading->tags_list.insert(get_apply_to().subreading->tags_list.begin() + tpos, tag->hash); |
| 1938 | ++tpos; |
| 1939 | } |
| 1940 | if (updateValidRules(rules, intersects, tag->hash, *get_apply_to().subreading)) { |
| 1941 | iter_rules = intersects.find(rule->number); |
| 1942 | iter_rules_end = intersects.end(); |
| 1943 | } |
| 1944 | } |
| 1945 | } |
| 1946 | else { |
| 1947 | ++i; |
| 1948 | } |
| 1949 | } |
| 1950 | reflowReading(*get_apply_to().subreading); |
| 1951 | |
| 1952 | if (!mappings->empty()) { |
| 1953 | splitMappings(mappings, *get_apply_to().cohort, *get_apply_to().subreading, true); |
| 1954 | } |
| 1955 | if (wf && wf != get_apply_to().subreading->parent->wordform) { |
| 1956 | for (auto r : get_apply_to().subreading->parent->readings) { |
| 1957 | delTagFromReading(*r, get_apply_to().subreading->parent->wordform); |
| 1958 | addTagToReading(*r, wf); |
| 1959 | } |
| 1960 | for (auto r : get_apply_to().subreading->parent->deleted) { |
| 1961 | delTagFromReading(*r, get_apply_to().subreading->parent->wordform); |
| 1962 | addTagToReading(*r, wf); |
| 1963 | } |
| 1964 | for (auto r : get_apply_to().subreading->parent->delayed) { |
| 1965 | delTagFromReading(*r, get_apply_to().subreading->parent->wordform); |
| 1966 | addTagToReading(*r, wf); |
| 1967 | } |
| 1968 | get_apply_to().subreading->parent->wordform = wf; |
| 1969 | for (auto r : grammar->wf_rules) { |
| 1970 | if (doesWordformsMatch(wf, r->wordform)) { |
| 1971 | current.rule_to_cohorts[r->number].insert(get_apply_to().cohort); |
| 1972 | intersects.insert(r->number); |
| 1973 | } |
| 1974 | else { |
| 1975 | current.rule_to_cohorts[r->number].erase(get_apply_to().cohort); |
| 1976 | } |
| 1977 | } |
| 1978 | updateValidRules(rules, intersects, wf->hash, *get_apply_to().subreading); |
| 1979 | iter_rules = intersects.find(rule->number); |
| 1980 | iter_rules_end = intersects.end(); |
| 1981 | } |
| 1982 | } |
| 1983 | if (get_apply_to().subreading->hash != state_hash) { |
| 1984 | readings_changed = true; |
| 1985 | } |
| 1986 | } |
| 1987 | else if (rule->type == K_APPEND) { |
| 1988 | index_ruleCohort_no.clear(); |
| 1989 | TRACE; |
| 1990 | |
| 1991 | Tag* bf = nullptr; |
| 1992 | std::vector<TagList> readings; |
| 1993 | auto theTags = ss_taglist.get(); |
| 1994 | getTagList(*rule->maplist, theTags); |
| 1995 | |
| 1996 | for (auto& tter : *theTags) { |
| 1997 | if (tter->type & T_VSTR) { |
| 1998 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 1999 | } |
| 2000 | } |
| 2001 | |
| 2002 | for (auto tter : *theTags) { |
| 2003 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 2004 | if (tter->type & T_BASEFORM) { |
| 2005 | bf = tter; |
| 2006 | readings.resize(readings.size() + 1); |
| 2007 | } |
| 2008 | if (bf == nullptr) { |
| 2009 | u_fprintfu_fprintf_72(ux_stderr, "Error: There must be a baseform before any other tags in APPEND on line %u.\n", rule->line); |
| 2010 | CG3Quit(1); |
| 2011 | } |
| 2012 | readings.back().push_back(tter); |
| 2013 | } |
| 2014 | |
| 2015 | for (const auto& rit : readings) { |
| 2016 | Reading* cReading = alloc_reading(get_apply_to().cohort); |
| 2017 | ++numReadings; |
| 2018 | insert_if_exists(cReading->parent->possible_sets, grammar->sets_any); |
| 2019 | addTagToReading(*cReading, get_apply_to().cohort->wordform); |
| 2020 | cReading->hit_by.push_back(rule->number); |
| 2021 | cReading->noprint = false; |
| 2022 | TagList mappings; |
| 2023 | for (auto tter : rit) { |
| 2024 | uint32_t hash = tter->hash; |
| 2025 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 2026 | if (tter->type & T_MAPPING || tter->tag[0] == grammar->mapping_prefix) { |
| 2027 | mappings.push_back(tter); |
| 2028 | } |
| 2029 | else { |
| 2030 | hash = addTagToReading(*cReading, tter); |
| 2031 | } |
| 2032 | if (updateValidRules(rules, intersects, hash, *cReading)) { |
| 2033 | iter_rules = intersects.find(rule->number); |
| 2034 | iter_rules_end = intersects.end(); |
| 2035 | } |
| 2036 | } |
| 2037 | if (!mappings.empty()) { |
| 2038 | splitMappings(mappings, *get_apply_to().cohort, *cReading); |
| 2039 | } |
| 2040 | get_apply_to().cohort->appendReading(cReading); |
| 2041 | } |
| 2042 | |
| 2043 | if (get_apply_to().cohort->readings.size() > 1) { |
| 2044 | foreach (rit, get_apply_to().cohort->readings)if (!(get_apply_to().cohort->readings).empty()) for (auto rit = (get_apply_to().cohort->readings).begin(), rit_end = (get_apply_to ().cohort->readings).end(); rit != rit_end; ++rit) { |
| 2045 | if ((*rit)->noprint) { |
| 2046 | free_reading(*rit); |
| 2047 | rit = get_apply_to().cohort->readings.erase(rit); |
| 2048 | rit_end = get_apply_to().cohort->readings.end(); |
| 2049 | } |
| 2050 | } |
| 2051 | } |
| 2052 | |
| 2053 | readings_changed = true; |
| 2054 | finish_reading_loop = false; |
| 2055 | } |
| 2056 | else if (rule->type == K_COPY) { |
| 2057 | // ToDo: Maybe just goto Substitute directly? |
| 2058 | Reading* cReading = get_apply_to().cohort->allocateAppendReading(*get_apply_to().reading); |
| 2059 | ++numReadings; |
| 2060 | index_ruleCohort_no.clear(); |
| 2061 | TRACE; |
| 2062 | cReading->hit_by.push_back(rule->number); |
| 2063 | cReading->noprint = false; |
| 2064 | |
| 2065 | if (rule->sublist) { |
| 2066 | auto excepts = ss_taglist.get(); |
| 2067 | getTagList(*rule->sublist, excepts); |
| 2068 | FILL_TAG_LIST_RAW(excepts)do { Reading& reading = *get_apply_to().subreading; for ( auto& tt : *(excepts)) { if (tt->type & T_SPECIAL) { if (context_stack.back().regexgrps == nullptr) { context_stack .back().regexgrps = ®exgrps_store[used_regex]; } auto stag = doesTagMatchReading(reading, *tt, false, true); if (stag) { tt = grammar->single_tags.find(stag)->second; } } } } while (0); |
| 2069 | for (auto r = cReading; r; r = r->next) { |
| 2070 | for (auto tter : *excepts) { |
| 2071 | delTagFromReading(*r, tter); |
| 2072 | } |
| 2073 | } |
| 2074 | } |
| 2075 | |
| 2076 | auto mappings = ss_taglist.get(); |
| 2077 | auto theTags = ss_taglist.get(); |
| 2078 | getTagList(*rule->maplist, theTags); |
| 2079 | |
| 2080 | bool did_insert = false; |
| 2081 | if (rule->childset1) { |
| 2082 | auto spot_tags = ss_taglist.get(); |
| 2083 | getTagList(*grammar->sets_list[rule->childset1], spot_tags); |
| 2084 | FILL_TAG_LIST(spot_tags)do { Reading& reading = *get_apply_to().subreading; for ( auto it = (spot_tags)->begin(); it != (spot_tags)->end( );) { if (reading.tags.find((*it)->hash) == reading.tags.end ()) { auto tt = *it; it = (spot_tags)->erase(it); if (tt-> type & T_SPECIAL) { if (context_stack.back().regexgrps == nullptr) { context_stack.back().regexgrps = ®exgrps_store [used_regex]; } auto stag = doesTagMatchReading(reading, *tt, false, true); if (stag) { (spot_tags)->insert(it, grammar ->single_tags.find(stag)->second); } } continue; } ++it ; } } while (0); |
| 2085 | auto it = cReading->tags_list.begin(); |
| 2086 | for (; it != cReading->tags_list.end(); ++it) { |
| 2087 | bool found = true; |
| 2088 | auto tmp = it; |
| 2089 | for (auto tag : *spot_tags) { |
| 2090 | if (*tmp != tag->hash) { |
| 2091 | found = false; |
| 2092 | break; |
| 2093 | } |
| 2094 | ++tmp; |
| 2095 | } |
| 2096 | if (found) { |
| 2097 | break; |
| 2098 | } |
| 2099 | } |
| 2100 | if (rule->flags & RF_AFTER) { |
| 2101 | std::advance(it, spot_tags->size()); |
| 2102 | } |
| 2103 | if (it != cReading->tags_list.end()) { |
| 2104 | insert_taglist_to_reading(it, *theTags, *cReading, mappings); |
| 2105 | did_insert = true; |
| 2106 | } |
| 2107 | } |
| 2108 | |
| 2109 | if (!did_insert) { |
| 2110 | APPEND_TAGLIST_TO_READING(*theTags, *cReading)do { for (auto tter : (*theTags)) { while (tter->type & T_VARSTRING) { tter = generateVarstringTag(tter); } auto hash = tter->hash; if (tter->type & T_MAPPING || tter-> tag[0] == grammar->mapping_prefix) { mappings->push_back (tter); } else { hash = addTagToReading((*cReading), tter); } if (updateValidRules(rules, intersects, hash, *cReading)) { iter_rules = intersects.find(rule->number); iter_rules_end = intersects .end(); } } } while (0); |
| 2111 | } |
| 2112 | if (!mappings->empty()) { |
| 2113 | splitMappings(mappings, *get_apply_to().cohort, *cReading, true); |
| 2114 | } |
| 2115 | readings_changed = true; |
| 2116 | reflowReading(*cReading); |
| 2117 | } |
| 2118 | else if (rule->type == K_MERGECOHORTS) { |
| 2119 | index_ruleCohort_no.clear(); |
| 2120 | |
| 2121 | CohortSet withs; |
| 2122 | Cohort* target = get_apply_to().cohort; |
| 2123 | withs.insert(target); |
| 2124 | Cohort* merge_at = target; |
| 2125 | for (auto it : rule->dep_tests) { |
| 2126 | auto& at = context_stack.back().attach_to; |
| 2127 | at.cohort = nullptr; |
| 2128 | at.reading = nullptr; |
| 2129 | at.subreading = nullptr; |
| 2130 | merge_with = nullptr; |
| 2131 | set_mark(target); |
| 2132 | dep_deep_seen.clear(); |
| 2133 | tmpl_cntx.clear(); |
| 2134 | Cohort* attach = nullptr; |
| 2135 | bool test_good = (runContextualTest(target->parent, target->local_number, it, &attach) && attach); |
| 2136 | |
| 2137 | profileRuleContext(test_good, rule, it); |
| 2138 | |
| 2139 | if (!test_good) { |
| 2140 | finish_reading_loop = false; |
| 2141 | return; |
| 2142 | } |
| 2143 | if (get_attach_to().cohort) { |
| 2144 | merge_at = get_attach_to().cohort; |
| 2145 | if (merge_with) { |
| 2146 | withs.insert(merge_with); |
| 2147 | } |
| 2148 | } |
| 2149 | else if (merge_with) { |
| 2150 | withs.insert(merge_with); |
| 2151 | } |
| 2152 | else { |
| 2153 | withs.insert(attach); |
| 2154 | } |
| 2155 | } |
| 2156 | |
| 2157 | size_t spacesInAddedWf = 0; |
| 2158 | context_stack.back().target.cohort = add_cohort(merge_at, spacesInAddedWf); |
| 2159 | |
| 2160 | for (auto c : withs) { |
| 2161 | size_t foundSpace = c->text.find_first_of(' '); |
| 2162 | while(spacesInAddedWf && foundSpace != std::string::npos) { |
| 2163 | c->text.erase(foundSpace, 1); |
| 2164 | foundSpace = c->text.find_first_of(' '); |
| 2165 | spacesInAddedWf--; |
| 2166 | } |
| 2167 | rem_cohort(c); |
| 2168 | } |
| 2169 | |
| 2170 | // If the last cohort was removed or inserted after, add <<< to the new end |
| 2171 | if (current.cohorts.back()->readings.front()->tags.count(endtag) == 0) { |
| 2172 | for (auto r : current.cohorts[current.cohorts.size() - 2]->readings) { |
| 2173 | delTagFromReading(*r, endtag); |
| 2174 | } |
| 2175 | for (auto r : current.cohorts.back()->readings) { |
| 2176 | addTagToReading(*r, endtag); |
| 2177 | if (updateValidRules(rules, intersects, endtag, *r)) { |
| 2178 | iter_rules = intersects.find(rule->number); |
| 2179 | iter_rules_end = intersects.end(); |
| 2180 | } |
| 2181 | } |
| 2182 | } |
| 2183 | indexSingleWindow(current); |
| 2184 | readings_changed = true; |
| 2185 | |
| 2186 | reset_cohorts_for_loop = true; |
| 2187 | } |
| 2188 | else if (rule->type == K_COPYCOHORT) { |
| 2189 | Cohort* attach = nullptr; |
| 2190 | Cohort* cohort = context_stack.back().target.cohort; |
| 2191 | uint32_t c = cohort->local_number; |
| 2192 | dep_deep_seen.clear(); |
| 2193 | tmpl_cntx.clear(); |
| 2194 | context_stack.back().attach_to.cohort = nullptr; |
| 2195 | context_stack.back().attach_to.reading = nullptr; |
| 2196 | context_stack.back().attach_to.subreading = nullptr; |
| 2197 | if (runContextualTest(¤t, c, rule->dep_target, &attach) && attach) { |
| 2198 | profileRuleContext(true, rule, rule->dep_target); |
| 2199 | |
| 2200 | if (get_attach_to().cohort) { |
| 2201 | attach = get_attach_to().cohort; |
| 2202 | } |
| 2203 | context_target = attach; |
| 2204 | bool good = true; |
| 2205 | for (auto it : rule->dep_tests) { |
| 2206 | context_stack.back().mark = attach; |
| 2207 | dep_deep_seen.clear(); |
| 2208 | tmpl_cntx.clear(); |
| 2209 | bool test_good = (runContextualTest(attach->parent, attach->local_number, it) != nullptr); |
| 2210 | |
| 2211 | profileRuleContext(test_good, rule, it); |
| 2212 | |
| 2213 | if (!test_good) { |
| 2214 | good = test_good; |
| 2215 | break; |
| 2216 | } |
| 2217 | } |
| 2218 | |
| 2219 | if (!good || cohort == attach || cohort->local_number == 0) { |
| 2220 | return; |
| 2221 | } |
| 2222 | |
| 2223 | auto childset = rule->childset2; |
| 2224 | if (rule->flags & RF_REVERSE) { |
| 2225 | std::swap(cohort, attach); |
| 2226 | childset = rule->childset1; |
| 2227 | } |
| 2228 | |
| 2229 | Cohort* cCohort = alloc_cohort(attach->parent); |
| 2230 | cCohort->global_number = gWindow->cohort_counter++; |
| 2231 | cCohort->wordform = cohort->wordform; |
| 2232 | insert_if_exists(cCohort->possible_sets, grammar->sets_any); |
| 2233 | |
| 2234 | auto theTags = ss_taglist.get(); |
| 2235 | getTagList(*rule->maplist, theTags); |
| 2236 | |
| 2237 | for (auto& tter : *theTags) { |
| 2238 | if (tter->type & T_VSTR) { |
| 2239 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 2240 | } |
| 2241 | } |
| 2242 | |
| 2243 | auto excepts = ss_taglist.get(); |
| 2244 | if (rule->sublist) { |
| 2245 | getTagList(*rule->sublist, excepts); |
| 2246 | FILL_TAG_LIST_RAW(excepts)do { Reading& reading = *get_apply_to().subreading; for ( auto& tt : *(excepts)) { if (tt->type & T_SPECIAL) { if (context_stack.back().regexgrps == nullptr) { context_stack .back().regexgrps = ®exgrps_store[used_regex]; } auto stag = doesTagMatchReading(reading, *tt, false, true); if (stag) { tt = grammar->single_tags.find(stag)->second; } } } } while (0); |
| 2247 | } |
| 2248 | |
| 2249 | std::vector<Reading*> rs; |
| 2250 | for (auto r : cohort->readings) { |
| 2251 | rs.clear(); |
| 2252 | for (; r; r = r->next) { |
| 2253 | auto cReading = alloc_reading(cCohort); |
| 2254 | ++numReadings; |
| 2255 | cReading->hit_by.push_back(rule->number); |
| 2256 | cReading->noprint = false; |
| 2257 | TagList mappings; |
| 2258 | for (auto hash : r->tags_list) { |
| 2259 | auto tter = grammar->single_tags[hash]; |
| 2260 | if (tter->type & T_MAPPING || tter->tag[0] == grammar->mapping_prefix) { |
| 2261 | mappings.push_back(tter); |
| 2262 | } |
| 2263 | else { |
| 2264 | hash = addTagToReading(*cReading, hash); |
| 2265 | } |
| 2266 | if (updateValidRules(rules, intersects, hash, *cReading)) { |
| 2267 | iter_rules = intersects.find(rule->number); |
| 2268 | iter_rules_end = intersects.end(); |
| 2269 | } |
| 2270 | } |
| 2271 | for (auto tter : *theTags) { |
| 2272 | auto hash = tter->hash; |
| 2273 | if (hash == grammar->tag_any) { |
| 2274 | continue; |
| 2275 | } |
| 2276 | if (tter->type & T_MAPPING || tter->tag[0] == grammar->mapping_prefix) { |
| 2277 | mappings.push_back(tter); |
| 2278 | } |
| 2279 | else { |
| 2280 | hash = addTagToReading(*cReading, hash); |
| 2281 | } |
| 2282 | if (updateValidRules(rules, intersects, hash, *cReading)) { |
| 2283 | iter_rules = intersects.find(rule->number); |
| 2284 | iter_rules_end = intersects.end(); |
| 2285 | } |
| 2286 | } |
| 2287 | if (!mappings.empty()) { |
| 2288 | splitMappings(mappings, *cCohort, *cReading); |
| 2289 | } |
| 2290 | rs.push_back(cReading); |
| 2291 | } |
| 2292 | auto rn = rs.front(); |
| 2293 | for (size_t j = 1; j < rs.size(); ++j) { |
| 2294 | rn->next = rs[j]; |
| 2295 | rn = rn->next; |
| 2296 | } |
| 2297 | cCohort->appendReading(rs.front()); |
| 2298 | } |
| 2299 | |
| 2300 | if (cCohort->readings.empty()) { |
| 2301 | initEmptyCohort(*cCohort); |
| 2302 | if (trace) { |
| 2303 | auto r = cCohort->readings.front(); |
| 2304 | r->hit_by.push_back(rule->number); |
| 2305 | r->noprint = false; |
| 2306 | } |
| 2307 | } |
| 2308 | |
| 2309 | for (auto r : cCohort->readings) { |
| 2310 | for (; r; r = r->next) { |
| 2311 | for (auto tter : *excepts) { |
| 2312 | delTagFromReading(*r, tter); |
| 2313 | } |
| 2314 | } |
| 2315 | } |
| 2316 | |
| 2317 | if (cohort->wread) { |
| 2318 | cCohort->wread = alloc_reading(cCohort); |
| 2319 | for (auto hash : cohort->wread->tags_list) { |
| 2320 | hash = addTagToReading(*cCohort->wread, hash); |
| 2321 | if (updateValidRules(rules, intersects, hash, *cCohort->wread)) { |
| 2322 | iter_rules = intersects.find(rule->number); |
| 2323 | iter_rules_end = intersects.end(); |
| 2324 | } |
| 2325 | } |
| 2326 | } |
| 2327 | |
| 2328 | current.parent->cohort_map[cCohort->global_number] = cCohort; |
| 2329 | current.parent->dep_window[cCohort->global_number] = cCohort; |
| 2330 | |
| 2331 | CohortSet edges; |
| 2332 | collect_subtree(edges, attach, childset); |
| 2333 | |
| 2334 | if (rule->flags & RF_BEFORE) { |
| 2335 | attach->parent->cohorts.insert(attach->parent->cohorts.begin() + edges.front()->local_number, cCohort); |
| 2336 | attach->parent->all_cohorts.insert(std::find(attach->parent->all_cohorts.begin() + edges.front()->local_number, attach->parent->all_cohorts.end(), edges.front()), cCohort); |
| 2337 | attachParentChild(*edges.front(), *cCohort); |
| 2338 | } |
| 2339 | else { |
| 2340 | attach->parent->cohorts.insert(attach->parent->cohorts.begin() + edges.back()->local_number + 1, cCohort); |
| 2341 | attach->parent->all_cohorts.insert(std::find(attach->parent->all_cohorts.begin() + edges.back()->local_number, attach->parent->all_cohorts.end(), edges.back()) + 1, cCohort); |
| 2342 | attachParentChild(*edges.back(), *cCohort); |
| 2343 | } |
| 2344 | |
| 2345 | reindex(attach->parent); |
| 2346 | indexSingleWindow(*attach->parent); |
| 2347 | readings_changed = true; |
| 2348 | reset_cohorts_for_loop = true; |
| 2349 | } |
| 2350 | } |
| 2351 | else if (rule->type == K_SETPARENT || rule->type == K_SETCHILD || rule->type == K_ADDRELATION || rule->type == K_SETRELATION || rule->type == K_REMRELATION || rule->type == K_ADDRELATIONS || rule->type == K_SETRELATIONS || rule->type == K_REMRELATIONS) { |
| 2352 | auto dep_target_cb = [&]() -> bool { |
| 2353 | Cohort* target = context_stack.back().target.cohort; |
| 2354 | Cohort* attach = context_stack.back().attach_to.cohort; |
| 2355 | swapper<Cohort*> sw((rule->flags & RF_REVERSE) != 0, target, attach); |
| 2356 | if (rule->type == K_SETPARENT || rule->type == K_SETCHILD) { |
| 2357 | bool attached = false; |
| 2358 | if (rule->type == K_SETPARENT) { |
| 2359 | attached = attachParentChild(*attach, *target, (rule->flags & RF_ALLOWLOOP) != 0, (rule->flags & RF_ALLOWCROSS) != 0); |
| 2360 | } |
| 2361 | else { |
| 2362 | attached = attachParentChild(*target, *attach, (rule->flags & RF_ALLOWLOOP) != 0, (rule->flags & RF_ALLOWCROSS) != 0); |
| 2363 | } |
| 2364 | if (attached) { |
| 2365 | index_ruleCohort_no.clear(); |
| 2366 | // force TRACE to use target |
| 2367 | Cohort* at_was = context_stack.back().attach_to.cohort; |
| 2368 | context_stack.back().attach_to.cohort = nullptr; |
| 2369 | TRACE; |
| 2370 | context_stack.back().attach_to.cohort = at_was; |
| 2371 | context_stack.back().target.subreading->noprint = false; |
| 2372 | has_dep = true; |
| 2373 | readings_changed = true; |
| 2374 | } |
| 2375 | return attached; |
| 2376 | } |
| 2377 | else if (rule->type == K_ADDRELATION || rule->type == K_SETRELATION || rule->type == K_REMRELATION) { |
| 2378 | bool rel_did_anything = false; |
| 2379 | auto theTags = ss_taglist.get(); |
| 2380 | getTagList(*rule->maplist, theTags); |
| 2381 | for (auto tter : *theTags) { |
| 2382 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 2383 | if (rule->type == K_ADDRELATION) { |
| 2384 | attach->setRelated(); |
| 2385 | target->setRelated(); |
| 2386 | rel_did_anything |= target->addRelation(tter->hash, attach->global_number); |
| 2387 | add_relation_rtag(target, tter, attach->global_number); |
| 2388 | } |
| 2389 | else if (rule->type == K_SETRELATION) { |
| 2390 | attach->setRelated(); |
| 2391 | target->setRelated(); |
| 2392 | rel_did_anything |= target->setRelation(tter->hash, attach->global_number); |
| 2393 | set_relation_rtag(target, tter, attach->global_number); |
| 2394 | } |
| 2395 | else { |
| 2396 | rel_did_anything |= target->remRelation(tter->hash, attach->global_number); |
| 2397 | rem_relation_rtag(target, tter, attach->global_number); |
| 2398 | } |
| 2399 | } |
| 2400 | if (rel_did_anything) { |
| 2401 | index_ruleCohort_no.clear(); |
| 2402 | // force TRACE to use target |
| 2403 | Cohort* at_was = context_stack.back().attach_to.cohort; |
| 2404 | context_stack.back().attach_to.cohort = nullptr; |
| 2405 | TRACE; |
| 2406 | context_stack.back().attach_to.cohort = at_was; |
| 2407 | context_stack.back().target.subreading->noprint = false; |
| 2408 | readings_changed = true; |
| 2409 | } |
| 2410 | // don't scan onwards if failed |
| 2411 | return true; |
| 2412 | } |
| 2413 | else if (rule->type == K_ADDRELATIONS || rule->type == K_SETRELATIONS || rule->type == K_REMRELATIONS) { |
| 2414 | bool rel_did_anything = false; |
| 2415 | |
| 2416 | auto sublist = ss_taglist.get(); |
| 2417 | getTagList(*rule->sublist, sublist); |
| 2418 | |
| 2419 | auto maplist = ss_taglist.get(); |
| 2420 | getTagList(*rule->maplist, maplist); |
| 2421 | |
| 2422 | for (auto tter : *maplist) { |
| 2423 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 2424 | if (rule->type == K_ADDRELATIONS) { |
| 2425 | target->setRelated(); |
| 2426 | rel_did_anything |= target->addRelation(tter->hash, attach->global_number); |
| 2427 | add_relation_rtag(target, tter, attach->global_number); |
| 2428 | } |
| 2429 | else if (rule->type == K_SETRELATIONS) { |
| 2430 | target->setRelated(); |
| 2431 | rel_did_anything |= target->setRelation(tter->hash, attach->global_number); |
| 2432 | set_relation_rtag(target, tter, attach->global_number); |
| 2433 | } |
| 2434 | else { |
| 2435 | rel_did_anything |= target->remRelation(tter->hash, attach->global_number); |
| 2436 | rem_relation_rtag(target, tter, attach->global_number); |
| 2437 | } |
| 2438 | } |
| 2439 | for (auto tter : *sublist) { |
| 2440 | VARSTRINGIFY(tter)do { while ((tter)->type & T_VARSTRING) { (tter) = generateVarstringTag ((tter)); } } while (0); |
| 2441 | if (rule->type == K_ADDRELATIONS) { |
| 2442 | attach->setRelated(); |
| 2443 | rel_did_anything |= attach->addRelation(tter->hash, target->global_number); |
| 2444 | add_relation_rtag(attach, tter, target->global_number); |
| 2445 | } |
| 2446 | else if (rule->type == K_SETRELATIONS) { |
| 2447 | attach->setRelated(); |
| 2448 | rel_did_anything |= attach->setRelation(tter->hash, target->global_number); |
| 2449 | set_relation_rtag(attach, tter, target->global_number); |
| 2450 | } |
| 2451 | else { |
| 2452 | rel_did_anything |= attach->remRelation(tter->hash, target->global_number); |
| 2453 | rem_relation_rtag(attach, tter, target->global_number); |
| 2454 | } |
| 2455 | } |
| 2456 | if (rel_did_anything) { |
| 2457 | index_ruleCohort_no.clear(); |
| 2458 | // force TRACE to use target |
| 2459 | Cohort* at_was = context_stack.back().attach_to.cohort; |
| 2460 | context_stack.back().attach_to.cohort = nullptr; |
| 2461 | TRACE; |
| 2462 | context_stack.back().attach_to.cohort = at_was; |
| 2463 | context_stack.back().target.subreading->noprint = false; |
| 2464 | readings_changed = true; |
| 2465 | } |
| 2466 | // don't scan onwards if failed |
| 2467 | return true; |
| 2468 | } |
| 2469 | return true; |
| 2470 | }; |
| 2471 | int32_t orgoffset = rule->dep_target->offset; |
| 2472 | auto seen_targets = ss_u32sv.get(); |
| 2473 | |
| 2474 | ReadingSpec orgtarget = context_stack.back().target; |
| 2475 | while (true) { |
| 2476 | auto utags = ss_utags.get(); |
| 2477 | auto usets = ss_usets.get(); |
| 2478 | *utags = *context_stack.back().unif_tags; |
| 2479 | *usets = *context_stack.back().unif_sets; |
| 2480 | |
| 2481 | Cohort* attach = nullptr; |
| 2482 | Cohort* target = context_stack.back().target.cohort; |
| 2483 | seen_targets->insert(target->global_number); |
| 2484 | dep_deep_seen.clear(); |
| 2485 | tmpl_cntx.clear(); |
| 2486 | context_stack.back().attach_to.cohort = nullptr; |
| 2487 | context_stack.back().attach_to.reading = nullptr; |
| 2488 | context_stack.back().attach_to.subreading = nullptr; |
| 2489 | seen_barrier = false; |
| 2490 | if (runContextualTest(target->parent, target->local_number, rule->dep_target, &attach) && attach) { |
| 2491 | profileRuleContext(true, rule, rule->dep_target); |
| 2492 | |
| 2493 | bool break_after = seen_barrier || (rule->flags & RF_NEAREST); |
| 2494 | if (get_attach_to().cohort) { |
| 2495 | attach = get_attach_to().cohort; |
| 2496 | } |
| 2497 | context_target = attach; |
| 2498 | bool good = true; |
| 2499 | for (auto it : rule->dep_tests) { |
| 2500 | context_stack.back().mark = attach; |
| 2501 | dep_deep_seen.clear(); |
| 2502 | tmpl_cntx.clear(); |
| 2503 | bool test_good = (runContextualTest(attach->parent, attach->local_number, it) != nullptr); |
| 2504 | |
| 2505 | profileRuleContext(test_good, rule, it); |
| 2506 | |
| 2507 | if (!test_good) { |
| 2508 | good = test_good; |
| 2509 | break; |
| 2510 | } |
| 2511 | } |
| 2512 | if (!get_attach_to().cohort) { |
| 2513 | context_stack.back().attach_to.cohort = attach; |
| 2514 | } |
| 2515 | if (good) { |
| 2516 | ReadingSpec temp = context_stack.back().target; |
| 2517 | context_stack.back().target = orgtarget; |
| 2518 | bool attached = dep_target_cb(); |
| 2519 | if (attached) { |
| 2520 | break; |
| 2521 | } |
| 2522 | else { |
| 2523 | context_stack.back().target = temp; |
| 2524 | } |
| 2525 | } |
| 2526 | if (break_after) { |
| 2527 | break; |
| 2528 | } |
| 2529 | if (seen_targets->count(attach->global_number)) { |
| 2530 | // We've found a cohort we have seen before... |
| 2531 | // We assume running the test again would result in the same, so don't bother. |
| 2532 | break; |
| 2533 | } |
| 2534 | // Did not successfully attach due to loop restrictions; look onwards from here |
| 2535 | context_stack.back().target = context_stack.back().attach_to; |
| 2536 | context_stack.back().unif_tags->swap(utags); |
| 2537 | context_stack.back().unif_sets->swap(usets); |
| 2538 | if (rule->dep_target->offset != 0) { |
| 2539 | // Temporarily set offset to +/- 1 |
| 2540 | rule->dep_target->offset = ((rule->dep_target->offset < 0) ? -1 : 1); |
| 2541 | } |
| 2542 | } |
| 2543 | else { |
| 2544 | break; |
| 2545 | } |
| 2546 | } |
| 2547 | rule->dep_target->offset = orgoffset; |
| 2548 | finish_reading_loop = false; |
| 2549 | } |
| 2550 | else if (rule->type == K_MOVE_AFTER || rule->type == K_MOVE_BEFORE || rule->type == K_SWITCH) { |
| 2551 | // this is a per-cohort rule |
| 2552 | finish_reading_loop = false; |
| 2553 | // Calculate hash of current state to later compare whether this move/switch actually did anything |
| 2554 | uint32_t phash = 0; |
| 2555 | uint32_t chash = 0; |
| 2556 | for (const auto& c : current.cohorts) { |
| 2557 | phash = hash_value(c->global_number, phash); |
| 2558 | chash = hash_value(c->readings[0]->hash, chash); |
| 2559 | } |
| 2560 | |
| 2561 | // ToDo: ** tests will not correctly work for MOVE/SWITCH; cannot move cohorts between windows |
| 2562 | Cohort* attach = nullptr; |
| 2563 | Cohort* cohort = context_stack.back().target.cohort; |
| 2564 | uint32_t c = cohort->local_number; |
| 2565 | dep_deep_seen.clear(); |
| 2566 | tmpl_cntx.clear(); |
| 2567 | context_stack.back().attach_to.cohort = nullptr; |
| 2568 | context_stack.back().attach_to.reading = nullptr; |
| 2569 | context_stack.back().attach_to.subreading = nullptr; |
| 2570 | if (runContextualTest(¤t, c, rule->dep_target, &attach) && attach && cohort->parent == attach->parent) { |
| 2571 | profileRuleContext(true, rule, rule->dep_target); |
| 2572 | |
| 2573 | if (get_attach_to().cohort) { |
| 2574 | attach = get_attach_to().cohort; |
| 2575 | } |
| 2576 | context_target = attach; |
| 2577 | bool good = true; |
| 2578 | for (auto it : rule->dep_tests) { |
| 2579 | context_stack.back().mark = attach; |
| 2580 | dep_deep_seen.clear(); |
| 2581 | tmpl_cntx.clear(); |
| 2582 | bool test_good = (runContextualTest(attach->parent, attach->local_number, it) != nullptr); |
| 2583 | |
| 2584 | profileRuleContext(test_good, rule, it); |
| 2585 | |
| 2586 | if (!test_good) { |
| 2587 | good = test_good; |
| 2588 | break; |
| 2589 | } |
| 2590 | } |
| 2591 | |
| 2592 | if (!good || cohort == attach || cohort->local_number == 0) { |
| 2593 | return; |
| 2594 | } |
| 2595 | |
| 2596 | swapper<Cohort*> sw((rule->flags & RF_REVERSE) != 0, attach, cohort); |
| 2597 | CohortSet cohorts; |
| 2598 | |
| 2599 | if (rule->type == K_SWITCH) { |
| 2600 | if (attach->local_number == 0) { |
| 2601 | return; |
| 2602 | } |
| 2603 | current.cohorts[cohort->local_number] = attach; |
| 2604 | current.cohorts[attach->local_number] = cohort; |
| 2605 | cohorts.insert(attach); |
| 2606 | cohorts.insert(cohort); |
| 2607 | auto ac_c = std::find(current.all_cohorts.begin() + cohort->local_number, current.all_cohorts.end(), cohort); |
| 2608 | auto ac_a = std::find(current.all_cohorts.begin() + attach->local_number, current.all_cohorts.end(), attach); |
| 2609 | *ac_c = attach; |
| 2610 | *ac_a = cohort; |
| 2611 | } |
| 2612 | else { |
| 2613 | CohortSet edges; |
| 2614 | collect_subtree(edges, attach, rule->childset2); |
| 2615 | collect_subtree(cohorts, cohort, rule->childset1); |
| 2616 | |
| 2617 | bool need_clean = false; |
| 2618 | for (auto iter : cohorts) { |
| 2619 | if (edges.count(iter)) { |
| 2620 | need_clean = true; |
| 2621 | break; |
| 2622 | } |
| 2623 | } |
| 2624 | |
| 2625 | if (need_clean) { |
| 2626 | if (isChildOf(cohort, attach)) { |
| 2627 | edges.erase(cohorts.rbegin(), cohorts.rend()); |
| 2628 | } |
| 2629 | else /* if (isChildOf(attach, cohort)) */ { |
| 2630 | cohorts.erase(edges.rbegin(), edges.rend()); |
| 2631 | } |
| 2632 | } |
| 2633 | if (cohorts.empty() || edges.empty()) { |
| 2634 | finish_reading_loop = false; |
| 2635 | return; |
| 2636 | } |
| 2637 | |
| 2638 | for (auto c : reversed(cohorts)) { |
| 2639 | current.cohorts.erase(current.cohorts.begin() + c->local_number); |
| 2640 | current.all_cohorts.erase(std::find(current.all_cohorts.begin() + c->local_number, current.all_cohorts.end(), c)); |
| 2641 | } |
| 2642 | |
| 2643 | foreach (iter, current.cohorts)if (!(current.cohorts).empty()) for (auto iter = (current.cohorts ).begin(), iter_end = (current.cohorts).end(); iter != iter_end ; ++iter) { |
| 2644 | (*iter)->local_number = UI32(std::distance(current.cohorts.begin(), iter)); |
| 2645 | } |
| 2646 | |
| 2647 | for (auto iter : edges) { |
| 2648 | if (iter->parent != get_apply_to().cohort->parent) { |
| 2649 | u_fprintfu_fprintf_72(ux_stderr, "Error: Move/Switch on line %u tried to move across window boundaries.\n", rule->line); |
| 2650 | CG3Quit(1); |
| 2651 | } |
| 2652 | for (auto cohort : cohorts) { |
| 2653 | if (iter == cohort) { |
| 2654 | u_fprintfu_fprintf_72(ux_stderr, "Error: Move/Switch on line %u tried to move to a removed position.\n", rule->line); |
| 2655 | CG3Quit(1); |
| 2656 | } |
| 2657 | } |
| 2658 | } |
| 2659 | |
| 2660 | uint32_t spot = 0; |
| 2661 | auto ac_spot = current.all_cohorts.begin(); |
| 2662 | if (rule->type == K_MOVE_BEFORE) { |
| 2663 | spot = edges.front()->local_number; |
| 2664 | if (spot == 0) { |
| 2665 | spot = 1; |
| 2666 | } |
| 2667 | ac_spot = std::find(current.all_cohorts.begin() + edges.front()->local_number, current.all_cohorts.end(), edges.front()); |
| 2668 | if ((*ac_spot)->local_number == 0) { |
| 2669 | ++ac_spot; |
| 2670 | } |
| 2671 | } |
| 2672 | else if (rule->type == K_MOVE_AFTER) { |
| 2673 | spot = edges.back()->local_number + 1; |
| 2674 | ac_spot = std::find(current.all_cohorts.begin() + edges.front()->local_number, current.all_cohorts.end(), edges.back()); |
| 2675 | ++ac_spot; |
| 2676 | } |
| 2677 | |
| 2678 | if (spot > current.cohorts.size()) { |
| 2679 | u_fprintfu_fprintf_72(ux_stderr, "Error: Move/Switch on line %u tried to move out of bounds.\n", rule->line); |
| 2680 | CG3Quit(1); |
| 2681 | } |
| 2682 | |
| 2683 | for (auto c : reversed(cohorts)) { |
| 2684 | current.cohorts.insert(current.cohorts.begin() + spot, c); |
| 2685 | current.all_cohorts.insert(ac_spot, c); |
| 2686 | } |
| 2687 | } |
| 2688 | reindex(); |
| 2689 | |
| 2690 | // Compare whether this move/switch actually did anything |
| 2691 | uint32_t phash_n = 0; |
| 2692 | uint32_t chash_n = 0; |
| 2693 | for (const auto& c : current.cohorts) { |
| 2694 | phash_n = hash_value(c->global_number, phash_n); |
| 2695 | chash_n = hash_value(c->readings[0]->hash, chash_n); |
| 2696 | } |
| 2697 | |
| 2698 | if (phash != phash_n || chash != chash_n) { |
| 2699 | if (++rule_hits[rule->number] > current.cohorts.size() * 100) { |
| 2700 | u_fprintfu_fprintf_72(ux_stderr, "Warning: Move/Switch endless loop detected for rule on line %u around input line %u - bailing out!\n", rule->line, get_apply_to().cohort->line_number); |
| 2701 | should_bail = true; |
| 2702 | finish_cohort_loop = false; |
| 2703 | return; |
| 2704 | } |
| 2705 | |
| 2706 | for (auto c : cohorts) { |
| 2707 | for (auto iter : c->readings) { |
| 2708 | iter->hit_by.push_back(rule->number); |
| 2709 | } |
| 2710 | } |
| 2711 | readings_changed = true; |
| 2712 | sorter.do_sort = true; |
| 2713 | } |
| 2714 | } |
| 2715 | } |
| 2716 | else if (rule->type == K_WITH) { |
| 2717 | TRACE; |
| 2718 | bool any_readings_changed = false; |
| 2719 | readings_changed = false; |
| 2720 | in_nested = true; |
| 2721 | for (auto& sr : rule->sub_rules) { |
| 2722 | Rule* cur_was = current_rule; |
| 2723 | Rule* rule_was = rule; |
| 2724 | current_rule = sr; |
| 2725 | rule = sr; |
| 2726 | bool result = false; |
| 2727 | do { |
| 2728 | readings_changed = false; |
| 2729 | result = runSingleRule(current, *rule, reading_cb, cohort_cb); |
| 2730 | any_readings_changed = any_readings_changed || result || readings_changed; |
| 2731 | } while ((result || readings_changed) && (rule->flags & RF_REPEAT) != 0) ; |
| 2732 | current_rule = cur_was; |
| 2733 | rule = rule_was; |
| 2734 | } |
| 2735 | in_nested = false; |
| 2736 | readings_changed = any_readings_changed; |
| 2737 | finish_reading_loop = false; |
| 2738 | } |
| 2739 | else if (rule->type != K_REMCOHORT) { |
| 2740 | TRACE; |
| 2741 | } |
| 2742 | }; |
| 2743 | |
| 2744 | removed.resize(0); |
| 2745 | selected.resize(0); |
| 2746 | bool rv = runSingleRule(current, *rule, reading_cb, cohort_cb); |
| 2747 | if (rv || readings_changed) { |
| 2748 | if (!(rule->flags & RF_NOITERATE) && section_max_count != 1) { |
| 2749 | section_did_something = true; |
| 2750 | } |
| 2751 | rule_did_something = true; |
| 2752 | } |
| 2753 | if (should_bail) { |
| 2754 | goto bailout; |
| 2755 | } |
| 2756 | if (should_repeat) { |
| 2757 | goto repeat_rule; |
| 2758 | } |
| 2759 | |
| 2760 | if (rule_did_something) { |
| 2761 | if (trace_rules.contains(rule->line)) { |
| 2762 | retval |= RV_TRACERULE; |
| 2763 | } |
| 2764 | } |
| 2765 | if (delimited) { |
| 2766 | break; |
| 2767 | } |
| 2768 | if (rule_did_something && (rule->flags & RF_REPEAT)) { |
| 2769 | index_ruleCohort_no.clear(); |
| 2770 | goto repeat_rule; |
| 2771 | } |
| 2772 | |
| 2773 | if (false) { |
| 2774 | bailout: |
| 2775 | rule_hits[rule->number] = 0; |
| 2776 | index_ruleCohort_no.clear(); |
| 2777 | } |
| 2778 | |
| 2779 | if (retval & RV_TRACERULE) { |
| 2780 | break; |
| 2781 | } |
| 2782 | } |
| 2783 | |
| 2784 | if (section_did_something) { |
| 2785 | retval |= RV_SOMETHING; |
| 2786 | } |
| 2787 | if (delimited) { |
| 2788 | retval |= RV_DELIMITED; |
| 2789 | } |
| 2790 | return retval; |
| 2791 | } |
| 2792 | |
| 2793 | uint32_t GrammarApplicator::runGrammarOnSingleWindow(SingleWindow& current) { |
| 2794 | if (!grammar->before_sections.empty() && !no_before_sections) { |
| 2795 | uint32_t rv = runRulesOnSingleWindow(current, runsections[-1]); |
| 2796 | if (rv & (RV_DELIMITED | RV_TRACERULE)) { |
| 2797 | return rv; |
| 2798 | } |
| 2799 | } |
| 2800 | |
| 2801 | if (!grammar->rules.empty() && !no_sections) { |
| 2802 | std::map<uint32_t, uint32_t> counter; |
| 2803 | // Caveat: This may look as if it is not recursing previous sections, but those rules are preprocessed into the successive sections so they are actually run. |
| 2804 | auto iter = runsections.begin(); |
| 2805 | auto iter_end = runsections.end(); |
| 2806 | for (size_t pass = 0; iter != iter_end; ++pass) { |
| 2807 | if (iter->first < 0 || (section_max_count && counter[iter->first] >= section_max_count)) { |
| 2808 | ++iter; |
| 2809 | continue; |
| 2810 | } |
| 2811 | uint32_t rv = 0; |
| 2812 | if (debug_level > 0) { |
| 2813 | std::cerr << "Running section " << iter->first << " (rules " << *(iter->second.begin()) << " through " << *(--(iter->second.end())) << ") on window " << current.number << std::endl; |
| 2814 | } |
| 2815 | rv = runRulesOnSingleWindow(current, iter->second); |
| 2816 | ++counter[iter->first]; |
| 2817 | if (rv & (RV_DELIMITED | RV_TRACERULE)) { |
| 2818 | return rv; |
| 2819 | } |
| 2820 | if (!(rv & RV_SOMETHING)) { |
| 2821 | ++iter; |
| 2822 | pass = 0; |
| 2823 | } |
| 2824 | if (pass >= 1000) { |
| 2825 | u_fprintfu_fprintf_72(ux_stderr, "Warning: Endless loop detected before input line %u. Window contents was:", numLines); |
| 2826 | UString tag; |
| 2827 | for (size_t i = 1; i < current.cohorts.size(); ++i) { |
| 2828 | Tag* t = current.cohorts[i]->wordform; |
| 2829 | tag.assign(t->tag.begin() + 2, t->tag.begin() + t->tag.size() - 2); |
| 2830 | u_fprintfu_fprintf_72(ux_stderr, " %S", tag.data()); |
| 2831 | } |
| 2832 | u_fprintfu_fprintf_72(ux_stderr, "\n"); |
| 2833 | u_fflushu_fflush_72(ux_stderr); |
| 2834 | break; |
| 2835 | } |
| 2836 | } |
| 2837 | } |
| 2838 | |
| 2839 | if (!grammar->after_sections.empty() && !no_after_sections) { |
| 2840 | uint32_t rv = runRulesOnSingleWindow(current, runsections[-2]); |
| 2841 | if (rv & (RV_DELIMITED | RV_TRACERULE)) { |
| 2842 | return rv; |
| 2843 | } |
| 2844 | } |
| 2845 | |
| 2846 | return 0; |
| 2847 | } |
| 2848 | |
| 2849 | void GrammarApplicator::runGrammarOnWindow() { |
| 2850 | SingleWindow* current = gWindow->current; |
| 2851 | did_final_enclosure = false; |
| 2852 | |
| 2853 | for (const auto& vit : current->variables_set) { |
| 2854 | variables[vit.first] = vit.second; |
| 2855 | } |
| 2856 | for (auto vit : current->variables_rem) { |
| 2857 | variables.erase(vit); |
| 2858 | } |
| 2859 | variables[mprefix_key] = mprefix_value; |
| 2860 | |
| 2861 | if (has_dep) { |
| 2862 | reflowDependencyWindow(); |
| 2863 | if (!input_eof && !gWindow->next.empty() && gWindow->next.back()->cohorts.size() > 1) { |
| 2864 | for (auto cohort : gWindow->next.back()->cohorts) { |
| 2865 | gWindow->dep_window[cohort->global_number] = cohort; |
| 2866 | } |
| 2867 | } |
| 2868 | } |
| 2869 | if (has_relations) { |
| 2870 | reflowRelationWindow(); |
| 2871 | } |
| 2872 | |
| 2873 | if (!grammar->parentheses.empty()) { |
| 2874 | label_scanParentheses: |
| 2875 | reverse_foreach (iter, current->cohorts)if (!(current->cohorts).empty()) for (auto iter = (current ->cohorts).rbegin(), iter_end = (current->cohorts).rend (); iter != iter_end; ++iter) { |
| 2876 | Cohort* c = *iter; |
| 2877 | if (c->is_pleft == 0) { |
| 2878 | continue; |
| 2879 | } |
| 2880 | auto p = grammar->parentheses.find(c->is_pleft); |
| 2881 | if (p != grammar->parentheses.end()) { |
| 2882 | auto right = iter.base(); |
| 2883 | --right; |
| 2884 | --right; |
| 2885 | c = *right; |
| 2886 | ++right; |
| 2887 | bool found = false; |
| 2888 | CohortVector encs; |
| 2889 | for (; right != current->cohorts.end(); ++right) { |
| 2890 | Cohort* s = *right; |
| 2891 | encs.push_back(s); |
| 2892 | if (s->is_pright == p->second) { |
| 2893 | found = true; |
| 2894 | break; |
| 2895 | } |
| 2896 | } |
| 2897 | if (found) { |
| 2898 | auto left = iter.base(); |
| 2899 | --left; |
| 2900 | uint32_t lc = (*left)->local_number; |
| 2901 | ++right; |
| 2902 | for (; right != current->cohorts.end(); ++right) { |
| 2903 | *left = *right; |
| 2904 | (*left)->local_number = lc; |
| 2905 | ++lc; |
| 2906 | ++left; |
| 2907 | } |
| 2908 | current->cohorts.resize(current->cohorts.size() - encs.size()); |
| 2909 | auto ec = std::find(current->all_cohorts.begin() + encs.front()->local_number, current->all_cohorts.end(), encs.front()); |
| 2910 | --ec; |
| 2911 | do { |
| 2912 | ++ec; |
| 2913 | (*ec)->type |= CT_ENCLOSED; |
| 2914 | ++((*ec)->enclosed); |
| 2915 | } while (*ec != encs.back()); |
| 2916 | current->has_enclosures = true; |
| 2917 | goto label_scanParentheses; |
| 2918 | } |
| 2919 | } |
| 2920 | } |
| 2921 | } |
| 2922 | |
| 2923 | par_left_tag = 0; |
| 2924 | par_right_tag = 0; |
| 2925 | par_left_pos = 0; |
| 2926 | par_right_pos = 0; |
| 2927 | uint32_t pass = 0; |
| 2928 | |
| 2929 | label_runGrammarOnWindow_begin: |
| 2930 | while (!gWindow->previous.empty() && gWindow->previous.size() > num_windows) { |
| 2931 | SingleWindow* tmp = gWindow->previous.front(); |
| 2932 | printSingleWindow(tmp, *ux_stdout); |
| 2933 | free_swindow(tmp); |
| 2934 | gWindow->previous.erase(gWindow->previous.begin()); |
| 2935 | } |
| 2936 | |
| 2937 | rule_hits.clear(); |
| 2938 | index_ruleCohort_no.clear(); |
| 2939 | current = gWindow->current; |
| 2940 | indexSingleWindow(*current); |
| 2941 | current->hit_external.clear(); |
| 2942 | gWindow->rebuildCohortLinks(); // ToDo: Hack. This can be done better... |
| 2943 | |
| 2944 | ++pass; |
| 2945 | if (pass > 1000) { |
| 2946 | u_fprintfu_fprintf_72(ux_stderr, "Warning: Endless loop detected before input line %u. Window contents was:", numLines); |
| 2947 | UString tag; |
| 2948 | for (size_t i = 1; i < current->cohorts.size(); ++i) { |
| 2949 | Tag* t = current->cohorts[i]->wordform; |
| 2950 | tag.assign(t->tag.begin() + 2, t->tag.begin() + t->tag.size() - 2); |
| 2951 | u_fprintfu_fprintf_72(ux_stderr, " %S", tag.data()); |
| 2952 | } |
| 2953 | u_fprintfu_fprintf_72(ux_stderr, "\n"); |
| 2954 | u_fflushu_fflush_72(ux_stderr); |
| 2955 | return; |
| 2956 | } |
| 2957 | |
| 2958 | if (trace_encl) { |
| 2959 | uint32_t hitpass = std::numeric_limits<uint32_t>::max() - pass; |
| 2960 | for (auto& c : current->cohorts) { |
| 2961 | for (auto rit : c->readings) { |
| 2962 | rit->hit_by.push_back(hitpass); |
| 2963 | } |
| 2964 | } |
| 2965 | } |
| 2966 | |
| 2967 | uint32_t rv = runGrammarOnSingleWindow(*current); |
| 2968 | if (rv & RV_DELIMITED) { |
| 2969 | goto label_runGrammarOnWindow_begin; |
| 2970 | } |
| 2971 | |
| 2972 | label_unpackEnclosures: |
| 2973 | if (current->has_enclosures) { |
| 2974 | size_t nc = current->all_cohorts.size(); |
| 2975 | for (size_t i = 0; i < nc; ++i) { |
| 2976 | Cohort* c = current->all_cohorts[i]; |
| 2977 | if (c->enclosed == 1) { |
| 2978 | size_t la = i; |
| 2979 | for (; la > 0; --la) { |
| 2980 | if (!(current->all_cohorts[la - 1]->type & (CT_ENCLOSED | CT_REMOVED | CT_IGNORED))) { |
| 2981 | --la; |
| 2982 | break; |
| 2983 | } |
| 2984 | } |
| 2985 | size_t ni = current->all_cohorts[la]->local_number; |
| 2986 | |
| 2987 | size_t ra = i; |
| 2988 | size_t ne = 0; |
| 2989 | for (; ra < nc; ++ra) { |
| 2990 | if (!(current->all_cohorts[ra]->type & (CT_ENCLOSED | CT_REMOVED | CT_IGNORED))) { |
| 2991 | break; |
| 2992 | } |
| 2993 | --(current->all_cohorts[ra]->enclosed); |
| 2994 | if (current->all_cohorts[ra]->enclosed == 0) { |
| 2995 | current->all_cohorts[ra]->type &= ~CT_ENCLOSED; |
| 2996 | ++ne; |
| 2997 | } |
| 2998 | } |
| 2999 | |
| 3000 | current->cohorts.resize(current->cohorts.size() + ne, nullptr); |
| 3001 | for (size_t j = current->cohorts.size() - 1; j > ni + ne; --j) { |
| 3002 | current->cohorts[j] = current->cohorts[j - ne]; |
| 3003 | current->cohorts[j]->local_number = UI32(j); |
| 3004 | current->cohorts[j - ne] = nullptr; |
| 3005 | } |
| 3006 | for (size_t j = 0; i < ra; ++i) { |
| 3007 | if (current->all_cohorts[i]->enclosed == 0) { |
| 3008 | current->cohorts[ni + j + 1] = current->all_cohorts[i]; |
| 3009 | current->cohorts[ni + j + 1]->local_number = UI32(ni + j + 1); |
| 3010 | current->cohorts[ni + j + 1]->parent = current; |
| 3011 | ++j; |
| 3012 | } |
| 3013 | } |
| 3014 | par_left_tag = current->all_cohorts[la + 1]->is_pleft; |
| 3015 | par_right_tag = current->all_cohorts[ra - 1]->is_pright; |
| 3016 | par_left_pos = UI32(ni + 1); |
| 3017 | par_right_pos = UI32(ni + ne); |
| 3018 | if (rv & RV_TRACERULE) { |
| 3019 | goto label_unpackEnclosures; |
| 3020 | } |
| 3021 | goto label_runGrammarOnWindow_begin; |
| 3022 | } |
| 3023 | } |
| 3024 | if (!did_final_enclosure) { |
| 3025 | par_left_tag = 0; |
| 3026 | par_right_tag = 0; |
| 3027 | par_left_pos = 0; |
| 3028 | par_right_pos = 0; |
| 3029 | did_final_enclosure = true; |
| 3030 | if (rv & RV_TRACERULE) { |
| 3031 | goto label_unpackEnclosures; |
| 3032 | } |
| 3033 | goto label_runGrammarOnWindow_begin; |
| 3034 | } |
| 3035 | } |
| 3036 | |
| 3037 | bool should_reflow = false; |
| 3038 | for (size_t i = current->all_cohorts.size(); i > 0; --i) { |
| 3039 | auto cohort = current->all_cohorts[i - 1]; |
| 3040 | if (cohort->type & CT_IGNORED) { |
| 3041 | for (auto ins = i; ins > 0; --ins) { |
| 3042 | if (!(current->all_cohorts[ins - 1]->type & (CT_REMOVED | CT_ENCLOSED | CT_IGNORED))) { |
| 3043 | current->cohorts.insert(current->cohorts.begin() + current->all_cohorts[ins - 1]->local_number + 1, cohort); |
| 3044 | cohort->type &= ~CT_IGNORED; |
| 3045 | current->parent->cohort_map.insert(std::make_pair(cohort->global_number, cohort)); |
| 3046 | should_reflow = true; |
| 3047 | break; |
| 3048 | } |
| 3049 | } |
| 3050 | } |
| 3051 | } |
| 3052 | if (should_reflow) { |
| 3053 | for (size_t i = 0; i < current->cohorts.size(); ++i) { |
| 3054 | current->cohorts[i]->local_number = UI32(i); |
| 3055 | } |
| 3056 | reflowDependencyWindow(); |
| 3057 | } |
| 3058 | } |
| 3059 | } |
| 3060 | |
| 3061 | // This helps the all_vislcg3.cpp profiling builds |
| 3062 | #undef TRACE |