| 1 | package felix.optimizer; |
| 2 | |
| 3 | import java.util.ArrayList; |
| 4 | import java.util.Arrays; |
| 5 | import java.util.HashMap; |
| 6 | import java.util.HashSet; |
| 7 | import java.util.Iterator; |
| 8 | import java.util.concurrent.ExecutorService; |
| 9 | import java.util.concurrent.Executors; |
| 10 | |
| 11 | |
| 12 | import tuffy.db.RDB; |
| 13 | import tuffy.mln.Literal; |
| 14 | import tuffy.mln.Term; |
| 15 | import tuffy.mln.Type; |
| 16 | import tuffy.ra.ConjunctiveQuery; |
| 17 | import tuffy.ra.Expression; |
| 18 | import tuffy.util.Config; |
| 19 | import tuffy.util.Timer; |
| 20 | import tuffy.util.UIMan; |
| 21 | |
| 22 | import felix.dstruct.ConcurrentOperatorsBucket; |
| 23 | import felix.dstruct.DataMovementOperator; |
| 24 | import felix.dstruct.FelixPredicate; |
| 25 | import felix.dstruct.StatOperator; |
| 26 | import felix.optimizer.CostModel; |
| 27 | import felix.optimizer.CostModel.resultTuple; |
| 28 | import felix.society.TaskList; |
| 29 | import felix.society.TaskSet; |
| 30 | import felix.task.OptimizeDMOTask; |
| 31 | import felix.util.FelixConfig; |
| 32 | import felix.util.FelixUIMan; |
| 33 | |
| 34 | /** |
| 35 | * An object of DMOOptimer takes inputs as a DMO, analyzes its |
| 36 | * logic plan, and fill in its physical plan. Current version of the |
| 37 | * DMOOptimizer will only optimize the materialization trade-off. |
| 38 | * Note that, this class does not touch the DB directly, instead |
| 39 | * it uses an instance of {@link CostModel}. |
| 40 | * |
| 41 | * @author Ce Zhang |
| 42 | * |
| 43 | */ |
| 44 | public class DMOOptimizer { |
| 45 | |
| 46 | /** |
| 47 | * Cost model used to optimize the DMO. |
| 48 | */ |
| 49 | public CostModel cm; |
| 50 | |
| 51 | /** |
| 52 | * Database connection. |
| 53 | */ |
| 54 | RDB db; |
| 55 | |
| 56 | /** |
| 57 | * The constructor. |
| 58 | * @param _cm |
| 59 | */ |
| 60 | public DMOOptimizer(CostModel _cm){ |
| 61 | cm = _cm; |
| 62 | db = RDB.getRDBbyConfig(Config.db_schema); |
| 63 | } |
| 64 | |
| 65 | /** |
| 66 | * Close the database connection used in this DMOOptimizer. |
| 67 | */ |
| 68 | public void close(){ |
| 69 | db.close(); |
| 70 | } |
| 71 | |
| 72 | boolean picasso = false; |
| 73 | |
| 74 | |
| 75 | /** |
| 76 | * Optimize all DMOs appearing in the given {@link ConcurrentOperatorsBucket}. |
| 77 | * @param cob |
| 78 | */ |
| 79 | public void optimizeDMO(ConcurrentOperatorsBucket cob){ |
| 80 | |
| 81 | // optimize their DMOs |
| 82 | ExecutorService pool = Executors.newFixedThreadPool(Config.getNumThreads()); |
| 83 | |
| 84 | TaskList tasks = new TaskList(); |
| 85 | |
| 86 | TaskSet taskset1 = new TaskSet(); |
| 87 | |
| 88 | for(StatOperator op : cob.getOperators()){ |
| 89 | taskset1.addSubTask(new OptimizeDMOTask(op, this)); |
| 90 | } |
| 91 | |
| 92 | tasks.addSubTask(taskset1); |
| 93 | try { |
| 94 | tasks.execute(pool); |
| 95 | } catch (Exception e) { |
| 96 | e.printStackTrace(); |
| 97 | } |
| 98 | |
| 99 | pool.shutdown(); |
| 100 | } |
| 101 | |
| 102 | /** |
| 103 | * Returns signature of query set. |
| 104 | * @param sets |
| 105 | * @return |
| 106 | */ |
| 107 | public String giveMeTheSignatureOfTwoQuery(HashMap<Literal, ConjunctiveQuery> sets){ |
| 108 | |
| 109 | String signature = ""; |
| 110 | |
| 111 | String[] qs = new String[sets.values().size()]; |
| 112 | int ct = 0; |
| 113 | for(ConjunctiveQuery query : sets.values()){ |
| 114 | |
| 115 | String[] l = new String[query.body.size()]; |
| 116 | |
| 117 | for(int i=0;i<l.length;i++){ |
| 118 | l[i] = query.body.get(i).toString(); |
| 119 | } |
| 120 | |
| 121 | Arrays.sort(l); |
| 122 | String ss = ""; |
| 123 | for(String s : l){ |
| 124 | ss = ss + s + ","; |
| 125 | } |
| 126 | |
| 127 | qs[ct++] = ss; |
| 128 | |
| 129 | } |
| 130 | |
| 131 | Arrays.sort(qs); |
| 132 | for(String s : qs){ |
| 133 | signature = signature + s + " | "; |
| 134 | } |
| 135 | |
| 136 | return signature; |
| 137 | } |
| 138 | |
| 139 | /** |
| 140 | * Optimize the materialization trade-off of the given DMO. |
| 141 | * @param dmo |
| 142 | */ |
| 143 | public void optimizeMateralization(DataMovementOperator dmo){ |
| 144 | try{ |
| 145 | |
| 146 | ConjunctiveQuery rule = dmo.logicQueryPlan.objectConjunctiveQuery; |
| 147 | |
| 148 | double bb = dmo.predictedBB; |
| 149 | double bf = dmo.PredictedBF; |
| 150 | double ff = dmo.PredictedFF; |
| 151 | |
| 152 | // cost for full view. |
| 153 | double view = cm.getFullViewCost(rule, bb, bf, ff, dmo.whichToBound); |
| 154 | |
| 155 | // cost for full materalization. |
| 156 | double full = cm.getFullMatCost(rule, bb, bf, ff, dmo.whichToBound); |
| 157 | |
| 158 | HashMap<ConjunctiveQuery, HashMap<Literal, ConjunctiveQuery>> comb = |
| 159 | new HashMap<ConjunctiveQuery, HashMap<Literal, ConjunctiveQuery>>(); |
| 160 | ArrayList<String> additionalSelList = new ArrayList<String>(); |
| 161 | HashMap<String, String> typeMapping = new HashMap<String, String>(); |
| 162 | |
| 163 | for(String s : dmo.additionalSelList){ |
| 164 | if(s.equals("weight")){ |
| 165 | if(rule.sourceClause.hasEmbeddedWeight()){ |
| 166 | additionalSelList.add(rule.sourceClause.getVarWeight()); |
| 167 | typeMapping.put(rule.sourceClause.getVarWeight(), "double_"); |
| 168 | } |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | // all binary decomposition of this conjunctive query. |
| 173 | comb = this.generateAllPossiblePlans(rule.head, rule.body, additionalSelList, typeMapping); |
| 174 | |
| 175 | ConjunctiveQuery bestQuery = null; |
| 176 | HashMap<Literal, ConjunctiveQuery> fullplan = null; |
| 177 | ConjunctiveQuery fullQuery = null; |
| 178 | |
| 179 | resultTuple smallestCost = cm.getNewEmptyResultTuple(); |
| 180 | smallestCost.totalCost = Double.MAX_VALUE; |
| 181 | |
| 182 | if(comb != null){ |
| 183 | for(ConjunctiveQuery cq : comb.keySet()){ |
| 184 | |
| 185 | ConjunctiveQuery q1 = null; |
| 186 | ConjunctiveQuery q2 = null; |
| 187 | |
| 188 | for(Literal sub : cq.body){ |
| 189 | if(q1 == null) |
| 190 | q1 = comb.get(cq).get(sub); |
| 191 | else |
| 192 | q2 = comb.get(cq).get(sub); |
| 193 | } |
| 194 | |
| 195 | if(q1 != null){ |
| 196 | q1.addConstraintAll(rule.getConstraint(q1.allVariable)); |
| 197 | } |
| 198 | |
| 199 | if(q2 != null){ |
| 200 | q2.addConstraintAll(rule.getConstraint(q2.allVariable)); |
| 201 | } |
| 202 | |
| 203 | if(q2 == null){ |
| 204 | fullQuery = cq; |
| 205 | fullplan = comb.get(cq); |
| 206 | continue; |
| 207 | } |
| 208 | |
| 209 | //TODO: |
| 210 | HashSet<Expression> cons = new HashSet<Expression>(); |
| 211 | cons.addAll(q1.getConstraint()); |
| 212 | cons.addAll(q2.getConstraint()); |
| 213 | if(rule.getConstraint().size() > cons.size()){ |
| 214 | continue; |
| 215 | } |
| 216 | |
| 217 | resultTuple cost = cm.getJoinCostBetweenTwoMaterializedTable( |
| 218 | cq.head, q1, q2, bb, bf, ff, dmo.whichToBound); |
| 219 | |
| 220 | if(cost == null){ |
| 221 | continue; |
| 222 | } |
| 223 | |
| 224 | if(FelixConfig.pickRandom== false && (smallestCost.totalCost == Double.MAX_VALUE || |
| 225 | (cost.totalCost <= smallestCost.totalCost/1)) ){ |
| 226 | |
| 227 | bestQuery = cq; |
| 228 | smallestCost = cost; |
| 229 | } |
| 230 | |
| 231 | if(FelixConfig.pickRandom==true && Math.random()<0.8 && cost.totalCost < 300000){ |
| 232 | |
| 233 | bestQuery = cq; |
| 234 | smallestCost = cost; |
| 235 | } |
| 236 | |
| 237 | } |
| 238 | }else{ |
| 239 | fullplan = new HashMap<Literal, ConjunctiveQuery>(); |
| 240 | fullQuery = new ConjunctiveQuery(); |
| 241 | fullQuery.setHead(rule.head); |
| 242 | } |
| 243 | |
| 244 | if(picasso == true){ |
| 245 | |
| 246 | System.out.print("postgreUnit=" + cm.postgreUnit + "\tmemoryTradeOff=" + cm.memoryTradeOff + "\t"); |
| 247 | |
| 248 | if(smallestCost.totalCost <= view && smallestCost.totalCost <= full){ |
| 249 | |
| 250 | System.out.println(this.giveMeTheSignatureOfTwoQuery(comb.get(bestQuery))); |
| 251 | |
| 252 | }else if (view <= smallestCost.totalCost && view <= full){ |
| 253 | System.out.println("VIEW"); |
| 254 | }else{ |
| 255 | System.out.println("FULL"); |
| 256 | } |
| 257 | |
| 258 | |
| 259 | }else{ |
| 260 | |
| 261 | //TODO: |
| 262 | //if(Config.gp == true){ |
| 263 | // FelixConfig.allMat = true; |
| 264 | //} |
| 265 | |
| 266 | |
| 267 | // select the best one from hybrid/full-mat/full-view |
| 268 | FelixUIMan.printobj(2,0,rule); |
| 269 | if(comb != null){ |
| 270 | FelixUIMan.printobj(2,0,smallestCost); |
| 271 | if(bestQuery != null){ |
| 272 | FelixUIMan.printobj(2,0,comb.get(bestQuery)); |
| 273 | } |
| 274 | } |
| 275 | FelixUIMan.println(2,0,"full = " + full); |
| 276 | FelixUIMan.println(2,0,"view = " + view); |
| 277 | |
| 278 | if(FelixConfig.allView){ |
| 279 | |
| 280 | FelixUIMan.println(1,0,">>> Regard the following query as view: \n" + "\t" + rule); |
| 281 | |
| 282 | // SchedulerTest.planCosts.add(view); |
| 283 | rule.isView = true; |
| 284 | dmo.physicalQueryPlan.objectConjunctiveQuery = rule; |
| 285 | }else{ |
| 286 | |
| 287 | if(FelixConfig.allMat){ |
| 288 | |
| 289 | FelixUIMan.println(1,0,">>> Regard the following query as materialized table: \n" + "\t" + rule); |
| 290 | |
| 291 | // SchedulerTest.planCosts.add(full); |
| 292 | |
| 293 | rule.isView = false; |
| 294 | |
| 295 | fullQuery.isView = true; |
| 296 | fullQuery.type = rule.type; |
| 297 | fullQuery.setWeight(rule.getWeight()); |
| 298 | fullQuery.sourceClause = rule.sourceClause; |
| 299 | |
| 300 | fullQuery.inverseEmbededWeight = rule.inverseEmbededWeight; |
| 301 | |
| 302 | dmo.physicalQueryPlan.objectConjunctiveQuery = fullQuery; |
| 303 | if(fullplan.size() > 0){ |
| 304 | dmo.physicalQueryPlan.datalogQueries.add(fullplan.get(fullQuery.body.get(0))); |
| 305 | fullplan.get(fullQuery.body.get(0)).head.getPred().prepareDB(db); |
| 306 | } |
| 307 | |
| 308 | }else{ |
| 309 | |
| 310 | if(smallestCost.totalCost <= view && smallestCost.totalCost <= full){ |
| 311 | |
| 312 | // SchedulerTest.planCosts.add(smallestCost.totalCost); |
| 313 | |
| 314 | ConjunctiveQuery q1 = null; |
| 315 | ConjunctiveQuery q2 = null; |
| 316 | |
| 317 | if(bestQuery == null){ |
| 318 | throw new Exception("Errors!"); |
| 319 | } |
| 320 | |
| 321 | for(Literal sub : comb.get(bestQuery).keySet()){ |
| 322 | if(q1 == null) q1 = comb.get(bestQuery).get(sub); |
| 323 | else q2 = comb.get(bestQuery).get(sub); |
| 324 | } |
| 325 | |
| 326 | ConjunctiveQuery rsCQ = new ConjunctiveQuery(); |
| 327 | rsCQ.isView = true; |
| 328 | rsCQ.type = rule.type; |
| 329 | rsCQ.setWeight(rule.getWeight()); |
| 330 | rsCQ.sourceClause = rule.sourceClause; |
| 331 | rsCQ.inverseEmbededWeight = rule.inverseEmbededWeight; |
| 332 | |
| 333 | rsCQ.setHead(bestQuery.head); |
| 334 | |
| 335 | if(q1.body.size() == 1){ |
| 336 | rsCQ.addBodyLit(q1.body.get(0)); |
| 337 | }else{ |
| 338 | rsCQ.addBodyLit(q1.head); |
| 339 | dmo.physicalQueryPlan.datalogQueries.add(q1); |
| 340 | //q1.head.getPred().createTable(); |
| 341 | q1.head.getPred().prepareDB(db); |
| 342 | } |
| 343 | |
| 344 | if(q2.body.size() == 1){ |
| 345 | rsCQ.addBodyLit(q2.body.get(0)); |
| 346 | }else{ |
| 347 | rsCQ.addBodyLit(q2.head); |
| 348 | dmo.physicalQueryPlan.datalogQueries.add(q2); |
| 349 | q2.head.getPred().prepareDB(db); |
| 350 | } |
| 351 | |
| 352 | dmo.physicalQueryPlan.objectConjunctiveQuery = rsCQ; |
| 353 | |
| 354 | q1.sourceClause = rule.sourceClause; |
| 355 | q2.sourceClause = rule.sourceClause; |
| 356 | |
| 357 | FelixUIMan.println(1,0,">>> Regard the following query as partially materialized view: \n" |
| 358 | + "\t" + rule + "\n" + "\t" + q1 + "\n" + "\t" + q2); |
| 359 | |
| 360 | }else if (view <= smallestCost.totalCost && view <= full){ |
| 361 | |
| 362 | FelixUIMan.println(1,0,">>> Regard the following query as view: \n" + "\t" + rule); |
| 363 | |
| 364 | // SchedulerTest.planCosts.add(view); |
| 365 | |
| 366 | rule.isView = true; |
| 367 | dmo.physicalQueryPlan.objectConjunctiveQuery = rule; |
| 368 | |
| 369 | }else{ |
| 370 | |
| 371 | FelixUIMan.println(1,0,">>> Regard the following query as materialized table: \n" + "\t" + rule); |
| 372 | |
| 373 | // SchedulerTest.planCosts.add(full); |
| 374 | |
| 375 | rule.isView = false; |
| 376 | |
| 377 | fullQuery.isView = true; |
| 378 | fullQuery.type = rule.type; |
| 379 | fullQuery.setWeight(rule.getWeight()); |
| 380 | fullQuery.sourceClause = rule.sourceClause; |
| 381 | |
| 382 | fullQuery.inverseEmbededWeight = rule.inverseEmbededWeight; |
| 383 | |
| 384 | dmo.physicalQueryPlan.objectConjunctiveQuery = fullQuery; |
| 385 | if(fullplan.size() > 0){ |
| 386 | dmo.physicalQueryPlan.datalogQueries.add(fullplan.get(fullQuery.body.get(0))); |
| 387 | fullplan.get(fullQuery.body.get(0)).head.getPred().prepareDB(db); |
| 388 | } |
| 389 | } |
| 390 | } |
| 391 | } |
| 392 | } |
| 393 | |
| 394 | |
| 395 | |
| 396 | }catch(Exception e){ |
| 397 | e.printStackTrace(); |
| 398 | } |
| 399 | } |
| 400 | |
| 401 | |
| 402 | /** |
| 403 | * Given a literal as goal, a set of literals as subgoals, generate all possible binary decompositions |
| 404 | * of this set of subgoals. The decomposition looks like: |
| 405 | * |
| 406 | * <br/> |
| 407 | * |
| 408 | * Q(...) :- Q1(...), Q2(...) <br/> |
| 409 | * Q1(...) :- g1, g2, ... <br/> |
| 410 | * Q2(...) :- g1', g2', ... <br/> |
| 411 | * |
| 412 | * <br/> |
| 413 | * |
| 414 | * Q1 and Q2 are materialized, and Q will be regarded as a view. If Q2 is null, this |
| 415 | * is equivalent to the fully-materialized case. |
| 416 | * |
| 417 | * @param head goal |
| 418 | * @param subgoals set of subgoals |
| 419 | * @param additionalSelList terms needed to be maintained in the variable list of Q1 and Q2. |
| 420 | * By default all variables needed to compute Q from Q1 and Q2 will be automatically maintained, |
| 421 | * however, if you want to maintain others, add them in this list. |
| 422 | * @param typeMapping If you want some variables in Q1 and Q2 have special types (e.g., double instead |
| 423 | * of constant ID), put them in this map. |
| 424 | * @return Mappings from Q to Q1 and Q2. |
| 425 | */ |
| 426 | public HashMap<ConjunctiveQuery, HashMap<Literal, ConjunctiveQuery>> generateAllPossiblePlans(Literal head, |
| 427 | ArrayList<Literal> subgoals, ArrayList<String> additionalSelList, HashMap<String, String> typeMapping){ |
| 428 | |
| 429 | HashMap<ConjunctiveQuery, HashMap<Literal, ConjunctiveQuery>> forReturn = |
| 430 | new HashMap<ConjunctiveQuery, HashMap<Literal, ConjunctiveQuery>>(); |
| 431 | |
| 432 | //ONLY CONSIDERING BI-PARTITIONING NOW |
| 433 | Integer[] biPar = new Integer[subgoals.size()]; |
| 434 | for(int i=0;i<biPar.length;i++) biPar[i] = 0; |
| 435 | |
| 436 | if(subgoals.size() == 0){ |
| 437 | return null; |
| 438 | } |
| 439 | |
| 440 | Integer ct = (int) Math.pow(2, subgoals.size()); |
| 441 | while(--ct >= 0){ |
| 442 | |
| 443 | |
| 444 | HashMap<Literal, ConjunctiveQuery> list = new HashMap<Literal, ConjunctiveQuery>(); |
| 445 | |
| 446 | ConjunctiveQuery q1 = new ConjunctiveQuery(); |
| 447 | ConjunctiveQuery q2 = new ConjunctiveQuery(); |
| 448 | ConjunctiveQuery q = new ConjunctiveQuery(); |
| 449 | q.setHead(head);//q.head = head; |
| 450 | |
| 451 | HashSet<Term> varSet1 = new HashSet<Term>(); |
| 452 | HashSet<Term> varSet2 = new HashSet<Term>(); |
| 453 | HashSet<String> termName1 = new HashSet<String>(); |
| 454 | HashSet<String> termName2 = new HashSet<String>(); |
| 455 | FelixPredicate p1 = new FelixPredicate(FelixPredicate.getNextTmpPredicateName(), false); |
| 456 | FelixPredicate p2 = new FelixPredicate(FelixPredicate.getNextTmpPredicateName(), false); |
| 457 | |
| 458 | HashMap<Term, Type> term2TypeMapping = new HashMap<Term, Type>(); |
| 459 | |
| 460 | for(int i=0;i<biPar.length;i++){ |
| 461 | if(biPar[i] == 0){ |
| 462 | //q1.body.add(subgoals.get(i)); |
| 463 | q1.addBodyLit(subgoals.get(i)); |
| 464 | varSet1.addAll(subgoals.get(i).getTerms()); |
| 465 | for(int j=0;j<subgoals.get(i).getTerms().size();j++){ |
| 466 | term2TypeMapping.put(subgoals.get(i).getTerms().get(j), subgoals.get(i).getPred().getTypeAt(j)); |
| 467 | } |
| 468 | |
| 469 | }else{ |
| 470 | //q2.body.add(subgoals.get(i)); |
| 471 | q2.addBodyLit(subgoals.get(i)); |
| 472 | varSet2.addAll(subgoals.get(i).getTerms()); |
| 473 | for(int j=0;j<subgoals.get(i).getTerms().size();j++){ |
| 474 | term2TypeMapping.put(subgoals.get(i).getTerms().get(j), subgoals.get(i).getPred().getTypeAt(j)); |
| 475 | } |
| 476 | } |
| 477 | } |
| 478 | |
| 479 | // generate next 0-1 combination. |
| 480 | biPar[0] ++; |
| 481 | for(int i=0;i<biPar.length-1;i++){ |
| 482 | if(biPar[i] == 2){ |
| 483 | biPar[i] = 0; |
| 484 | biPar[i+1] ++; |
| 485 | } |
| 486 | } |
| 487 | |
| 488 | if(q1.body.size() == 0){ |
| 489 | continue; |
| 490 | } |
| 491 | |
| 492 | |
| 493 | HashSet<String> allIntermediateTerms = new HashSet<String>(); |
| 494 | HashSet<String> allIntermediateTerms2 = new HashSet<String>(); |
| 495 | |
| 496 | for(Term t : varSet1) allIntermediateTerms.add(t.var()); |
| 497 | for(Term t : varSet2) allIntermediateTerms2.add(t.var()); |
| 498 | allIntermediateTerms.retainAll(allIntermediateTerms2); |
| 499 | for(Term t : head.getTerms()) allIntermediateTerms.add(t.var()); |
| 500 | allIntermediateTerms.addAll(additionalSelList); |
| 501 | |
| 502 | Literal q1Head = new Literal(p1, true); |
| 503 | for(Term t : varSet1){ |
| 504 | if(t.isVariable() == true && allIntermediateTerms.contains(t.var())){ |
| 505 | if(!termName1.contains(t.var())){ |
| 506 | q1Head.appendTerm(t); |
| 507 | termName1.add(t.var()); |
| 508 | if(typeMapping.containsKey(t.var())){ |
| 509 | p1.appendArgument(new Type(typeMapping.get(t.var()))); |
| 510 | }else{ |
| 511 | p1.appendArgument(term2TypeMapping.get(t)); |
| 512 | } |
| 513 | } |
| 514 | } |
| 515 | } |
| 516 | q1.setHead(q1Head); |
| 517 | |
| 518 | |
| 519 | Literal q2Head = new Literal(p2, true); |
| 520 | for(Term t : varSet2){ |
| 521 | if(t.isVariable() == true && allIntermediateTerms.contains(t.var())){ |
| 522 | if(!termName2.contains(t.var())){ |
| 523 | q2Head.appendTerm(t); |
| 524 | termName2.add(t.var()); |
| 525 | if(typeMapping.containsKey(t.var())){ |
| 526 | p2.appendArgument(new Type(typeMapping.get(t.var()))); |
| 527 | }else{ |
| 528 | p2.appendArgument(term2TypeMapping.get(t)); |
| 529 | } |
| 530 | } |
| 531 | |
| 532 | } |
| 533 | } |
| 534 | q2.setHead(q2Head); |
| 535 | |
| 536 | if(q2.head.getTerms().size() == 0 && q1.head.getTerms().size() == 0){ |
| 537 | continue; |
| 538 | } |
| 539 | |
| 540 | // full |
| 541 | if(q2.head.getTerms().size() == 0){ |
| 542 | list.put(q1.head, q1); |
| 543 | q.addBodyLit(q1.head); |
| 544 | forReturn.put(q, list); |
| 545 | continue; |
| 546 | } |
| 547 | |
| 548 | list.put(q1.head, q1); |
| 549 | list.put(q2.head, q2); |
| 550 | |
| 551 | q.addBodyLit(q1.head); |
| 552 | q.addBodyLit(q2.head); |
| 553 | |
| 554 | forReturn.put(q, list); |
| 555 | |
| 556 | |
| 557 | } |
| 558 | |
| 559 | |
| 560 | return forReturn; |
| 561 | } |
| 562 | } |