1 | // Copyright 2012 Georg-August-Universität Göttingen, Germany |
---|
2 | // |
---|
3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
---|
4 | // you may not use this file except in compliance with the License. |
---|
5 | // You may obtain a copy of the License at |
---|
6 | // |
---|
7 | // http://www.apache.org/licenses/LICENSE-2.0 |
---|
8 | // |
---|
9 | // Unless required by applicable law or agreed to in writing, software |
---|
10 | // distributed under the License is distributed on an "AS IS" BASIS, |
---|
11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
---|
12 | // See the License for the specific language governing permissions and |
---|
13 | // limitations under the License. |
---|
14 | |
---|
15 | package de.ugoe.cs.autoquest.tasktrees.temporalrelation; |
---|
16 | |
---|
17 | import java.util.HashMap; |
---|
18 | import java.util.Iterator; |
---|
19 | import java.util.LinkedList; |
---|
20 | import java.util.List; |
---|
21 | |
---|
22 | import de.ugoe.cs.autoquest.tasktrees.taskequality.TaskEquality; |
---|
23 | import de.ugoe.cs.autoquest.tasktrees.taskequality.TaskEqualityRuleManager; |
---|
24 | import de.ugoe.cs.autoquest.tasktrees.treeifc.IEventTask; |
---|
25 | import de.ugoe.cs.autoquest.tasktrees.treeifc.IIteration; |
---|
26 | import de.ugoe.cs.autoquest.tasktrees.treeifc.ISelection; |
---|
27 | import de.ugoe.cs.autoquest.tasktrees.treeifc.ISequence; |
---|
28 | import de.ugoe.cs.autoquest.tasktrees.treeifc.ITask; |
---|
29 | import de.ugoe.cs.autoquest.tasktrees.treeifc.ITaskBuilder; |
---|
30 | import de.ugoe.cs.autoquest.tasktrees.treeifc.ITaskFactory; |
---|
31 | import de.ugoe.cs.autoquest.tasktrees.treeifc.ITaskInstance; |
---|
32 | import de.ugoe.cs.autoquest.tasktrees.treeifc.ITaskInstanceList; |
---|
33 | import de.ugoe.cs.autoquest.tasktrees.treeifc.IUserSession; |
---|
34 | import de.ugoe.cs.autoquest.usageprofiles.SymbolComparator; |
---|
35 | import de.ugoe.cs.autoquest.usageprofiles.Trie; |
---|
36 | import de.ugoe.cs.autoquest.usageprofiles.TrieProcessor; |
---|
37 | import de.ugoe.cs.util.StopWatch; |
---|
38 | |
---|
39 | /** |
---|
40 | * <p> |
---|
41 | * TODO comment |
---|
42 | * </p> |
---|
43 | * |
---|
44 | * @author Patrick Harms |
---|
45 | */ |
---|
46 | class SequenceForTaskDetectionRule implements ISessionScopeRule { |
---|
47 | |
---|
48 | /** |
---|
49 | * <p> |
---|
50 | * the task factory to be used for creating substructures for the temporal |
---|
51 | * relationships identified during rule |
---|
52 | * </p> |
---|
53 | */ |
---|
54 | private ITaskFactory taskFactory; |
---|
55 | /** |
---|
56 | * <p> |
---|
57 | * the task builder to be used for creating substructures for the temporal relationships |
---|
58 | * identified during rule application |
---|
59 | * </p> |
---|
60 | */ |
---|
61 | private ITaskBuilder taskBuilder; |
---|
62 | |
---|
63 | /** |
---|
64 | * <p> |
---|
65 | * the task comparator to be used for comparing tasks |
---|
66 | * </p> |
---|
67 | */ |
---|
68 | private TaskComparator taskComparator; |
---|
69 | |
---|
70 | /** |
---|
71 | * <p> |
---|
72 | * instantiates the rule and initializes it with a task equality rule manager and the minimal |
---|
73 | * task equality identified sublist must have to consider them as iterated. |
---|
74 | * </p> |
---|
75 | */ |
---|
76 | SequenceForTaskDetectionRule(TaskEqualityRuleManager taskEqualityRuleManager, |
---|
77 | TaskEquality minimalTaskEquality, |
---|
78 | ITaskFactory taskFactory, |
---|
79 | ITaskBuilder taskBuilder) |
---|
80 | { |
---|
81 | this.taskFactory = taskFactory; |
---|
82 | this.taskBuilder = taskBuilder; |
---|
83 | |
---|
84 | this.taskComparator = new TaskComparator(taskEqualityRuleManager, minimalTaskEquality); |
---|
85 | } |
---|
86 | |
---|
87 | /* (non-Javadoc) |
---|
88 | * @see java.lang.Object#toString() |
---|
89 | */ |
---|
90 | @Override |
---|
91 | public String toString() { |
---|
92 | return "SequenceForTaskDetectionRule"; |
---|
93 | } |
---|
94 | |
---|
95 | /* (non-Javadoc) |
---|
96 | * @see de.ugoe.cs.autoquest.tasktrees.temporalrelation.ISessionScopeRule#apply(java.util.List) |
---|
97 | */ |
---|
98 | @Override |
---|
99 | public RuleApplicationResult apply(List<IUserSession> sessions) { |
---|
100 | RuleApplicationData appData = new RuleApplicationData(sessions); |
---|
101 | |
---|
102 | // this is the real rule application. Loop while something is replaced. |
---|
103 | do { |
---|
104 | System.out.println(); |
---|
105 | |
---|
106 | appData.getStopWatch().start("whole loop"); |
---|
107 | detectAndReplaceIterations(appData); |
---|
108 | |
---|
109 | appData.getStopWatch().start("task replacement"); |
---|
110 | detectAndReplaceTasks(appData); |
---|
111 | appData.getStopWatch().stop("task replacement"); |
---|
112 | appData.getStopWatch().stop("whole loop"); |
---|
113 | |
---|
114 | //((TaskTreeNodeComparator) taskComparator).getStopWatch().dumpStatistics(System.out); |
---|
115 | //((TaskTreeNodeComparator) taskComparator).getStopWatch().reset(); |
---|
116 | |
---|
117 | appData.getStopWatch().dumpStatistics(System.out); |
---|
118 | appData.getStopWatch().reset(); |
---|
119 | |
---|
120 | } |
---|
121 | while (appData.detectedAndReplacedTasks()); |
---|
122 | |
---|
123 | System.out.println |
---|
124 | ("created " + appData.getResult().getNewlyCreatedTasks().size() + |
---|
125 | " new tasks and " + appData.getResult().getNewlyCreatedTaskInstances().size() + |
---|
126 | " appropriate instances\n"); |
---|
127 | |
---|
128 | if ((appData.getResult().getNewlyCreatedTasks().size() > 0) || |
---|
129 | (appData.getResult().getNewlyCreatedTaskInstances().size() > 0)) |
---|
130 | { |
---|
131 | appData.getResult().setRuleApplicationStatus(RuleApplicationStatus.FINISHED); |
---|
132 | } |
---|
133 | |
---|
134 | return appData.getResult(); |
---|
135 | } |
---|
136 | |
---|
137 | /** |
---|
138 | * @param appData |
---|
139 | */ |
---|
140 | private void detectAndReplaceIterations(RuleApplicationData appData) { |
---|
141 | System.out.print("detecting iterations"); |
---|
142 | appData.getStopWatch().start("detecting iterations"); |
---|
143 | |
---|
144 | List<IUserSession> sessions = appData.getSessions(); |
---|
145 | int iteratedTasks = 0; |
---|
146 | |
---|
147 | ITask iteratedTask = null; |
---|
148 | |
---|
149 | do { |
---|
150 | iteratedTask = searchIteratedTask(sessions); |
---|
151 | |
---|
152 | if (iteratedTask != null) { |
---|
153 | replaceIterationsOf(iteratedTask, sessions, appData); |
---|
154 | iteratedTasks++; |
---|
155 | } |
---|
156 | } |
---|
157 | while (iteratedTask != null); |
---|
158 | |
---|
159 | appData.getStopWatch().stop("detecting iterations"); |
---|
160 | System.out.println(" --> found " + iteratedTasks + " iterated tasks"); |
---|
161 | } |
---|
162 | |
---|
163 | /** |
---|
164 | * |
---|
165 | */ |
---|
166 | private ITask searchIteratedTask(List<IUserSession> sessions) { |
---|
167 | for (IUserSession session : sessions) { |
---|
168 | for (int i = 0; i < (session.size() - 1); i++) { |
---|
169 | if (taskComparator.equals(session.get(i), session.get(i + 1))) { |
---|
170 | return session.get(i).getTask(); |
---|
171 | } |
---|
172 | } |
---|
173 | } |
---|
174 | |
---|
175 | return null; |
---|
176 | } |
---|
177 | |
---|
178 | /** |
---|
179 | * |
---|
180 | */ |
---|
181 | /* private ITask searchIteratedTask(List<IUserSession> sessions) { |
---|
182 | int minNoOfRepetitions = 2; |
---|
183 | int minNoOfIterationOccurrences = 1; |
---|
184 | |
---|
185 | Map<ITask, Integer> iterationsCounter = new HashMap<ITask, Integer>(); |
---|
186 | |
---|
187 | for (IUserSession session : sessions) { |
---|
188 | for (int i = 0; i < (session.size() - minNoOfRepetitions + 1); i++) { |
---|
189 | ITask task = session.get(i).getTask(); |
---|
190 | |
---|
191 | // check if the task is iterated |
---|
192 | boolean isIterated = true; |
---|
193 | for (int j = i + 1; j < i + minNoOfRepetitions; j++) { |
---|
194 | if (!taskComparator.equals(task, session.get(j).getTask())) { |
---|
195 | isIterated = false; |
---|
196 | break; |
---|
197 | } |
---|
198 | } |
---|
199 | |
---|
200 | if (isIterated) { |
---|
201 | Integer currentCounter = null; |
---|
202 | |
---|
203 | for (Map.Entry<ITask, Integer> entry : iterationsCounter.entrySet()) { |
---|
204 | if (taskComparator.equals(task, entry.getKey())) { |
---|
205 | currentCounter = entry.getValue(); |
---|
206 | break; |
---|
207 | } |
---|
208 | } |
---|
209 | |
---|
210 | if (currentCounter == null) { |
---|
211 | currentCounter = 1; |
---|
212 | iterationsCounter.put(task, currentCounter); |
---|
213 | } |
---|
214 | else { |
---|
215 | currentCounter++; |
---|
216 | iterationsCounter.put(task, currentCounter); |
---|
217 | } |
---|
218 | |
---|
219 | if (currentCounter >= minNoOfIterationOccurrences) { |
---|
220 | return task; |
---|
221 | } |
---|
222 | } |
---|
223 | } |
---|
224 | } |
---|
225 | |
---|
226 | return null; |
---|
227 | }*/ |
---|
228 | |
---|
229 | /** |
---|
230 | * |
---|
231 | */ |
---|
232 | private void replaceIterationsOf(ITask iteratedTask, |
---|
233 | List<IUserSession> sessions, |
---|
234 | RuleApplicationData appData) |
---|
235 | { |
---|
236 | IIteration iteration = taskFactory.createNewIteration(); |
---|
237 | ITaskInstance iterationInstance = null; |
---|
238 | |
---|
239 | List<ITaskInstance> iterationInstances = new LinkedList<ITaskInstance>(); |
---|
240 | |
---|
241 | for (IUserSession session : sessions) { |
---|
242 | int index = 0; |
---|
243 | while (index < session.size()) { |
---|
244 | if (taskComparator.equals(iteratedTask, session.get(index).getTask())) { |
---|
245 | if (iterationInstance == null) { |
---|
246 | iterationInstance = taskFactory.createNewTaskInstance(iteration); |
---|
247 | iterationInstances.add(iterationInstance); |
---|
248 | taskBuilder.addTaskInstance(session, index, iterationInstance); |
---|
249 | index++; |
---|
250 | } |
---|
251 | |
---|
252 | taskBuilder.addChild(iterationInstance, session.get(index)); |
---|
253 | taskBuilder.removeTaskInstance(session, index); |
---|
254 | } |
---|
255 | else { |
---|
256 | if (iterationInstance != null) { |
---|
257 | iterationInstance = null; |
---|
258 | } |
---|
259 | index++; |
---|
260 | } |
---|
261 | } |
---|
262 | } |
---|
263 | |
---|
264 | harmonizeIterationInstancesModel(iteration, iterationInstances); |
---|
265 | } |
---|
266 | |
---|
267 | /** |
---|
268 | * |
---|
269 | */ |
---|
270 | private void harmonizeIterationInstancesModel(IIteration iteration, |
---|
271 | List<ITaskInstance> iterationInstances) |
---|
272 | { |
---|
273 | List<ITask> iteratedTaskVariants = new LinkedList<ITask>(); |
---|
274 | |
---|
275 | // merge the lexically different variants of iterated task to a unique list |
---|
276 | for (ITaskInstance iterationInstance : iterationInstances) { |
---|
277 | for (ITaskInstance executionVariant : iterationInstance) { |
---|
278 | ITask candidate = executionVariant.getTask(); |
---|
279 | |
---|
280 | boolean found = false; |
---|
281 | for (ITask taskVariant : iteratedTaskVariants) { |
---|
282 | if (taskComparator.areLexicallyEqual(taskVariant, candidate)) { |
---|
283 | taskBuilder.setTask(executionVariant, taskVariant); |
---|
284 | found = true; |
---|
285 | break; |
---|
286 | } |
---|
287 | } |
---|
288 | |
---|
289 | if (!found) { |
---|
290 | iteratedTaskVariants.add(candidate); |
---|
291 | } |
---|
292 | } |
---|
293 | } |
---|
294 | |
---|
295 | // if there are more than one lexically different variant of iterated tasks, adapt the |
---|
296 | // iteration model to be a selection of different variants. In this case also adapt |
---|
297 | // the generated iteration instances to correctly contain selection instances. If there |
---|
298 | // is only one variant of an iterated task, simply set this as the marked task of the |
---|
299 | // iteration. In this case, the instances can be preserved as is |
---|
300 | if (iteratedTaskVariants.size() > 1) { |
---|
301 | ISelection selection = taskFactory.createNewSelection(); |
---|
302 | |
---|
303 | for (ITask variant : iteratedTaskVariants) { |
---|
304 | taskBuilder.addChild(selection, variant); |
---|
305 | } |
---|
306 | |
---|
307 | taskBuilder.setMarkedTask(iteration, selection); |
---|
308 | |
---|
309 | for (ITaskInstance instance : iterationInstances) { |
---|
310 | for (int i = 0; i < instance.size(); i++) { |
---|
311 | ITaskInstance selectionInstance = taskFactory.createNewTaskInstance(selection); |
---|
312 | taskBuilder.addChild(selectionInstance, instance.get(i)); |
---|
313 | taskBuilder.setTaskInstance(instance, i, selectionInstance); |
---|
314 | } |
---|
315 | } |
---|
316 | } |
---|
317 | else { |
---|
318 | taskBuilder.setMarkedTask(iteration, iteratedTaskVariants.get(0)); |
---|
319 | } |
---|
320 | } |
---|
321 | |
---|
322 | /** |
---|
323 | * @param appData |
---|
324 | */ |
---|
325 | private void detectAndReplaceTasks(RuleApplicationData appData) { |
---|
326 | System.out.println("detecting and replacing tasks"); |
---|
327 | appData.getStopWatch().start("detecting tasks"); |
---|
328 | |
---|
329 | getSequencesOccuringMostOften(appData); |
---|
330 | |
---|
331 | appData.getStopWatch().stop("detecting tasks"); |
---|
332 | appData.getStopWatch().start("replacing tasks"); |
---|
333 | |
---|
334 | replaceSequencesOccurringMostOften(appData); |
---|
335 | |
---|
336 | appData.getStopWatch().stop("replacing tasks"); |
---|
337 | System.out.println("detected and replaced " + appData.getLastFoundTasks().size() + " tasks"); |
---|
338 | } |
---|
339 | |
---|
340 | /** |
---|
341 | * @return |
---|
342 | */ |
---|
343 | private void getSequencesOccuringMostOften(RuleApplicationData appData) { |
---|
344 | System.out.println("determining most prominent tasks"); |
---|
345 | |
---|
346 | Tasks tasks; |
---|
347 | boolean createNewTrie = (appData.getLastTrie() == null) || |
---|
348 | appData.detectedAndReplacedTasks(); // tree has changed |
---|
349 | |
---|
350 | do { |
---|
351 | if (createNewTrie) { |
---|
352 | createNewTrie(appData); |
---|
353 | } |
---|
354 | |
---|
355 | MaxCountAndLongestTasksFinder finder = new MaxCountAndLongestTasksFinder(); |
---|
356 | appData.getLastTrie().process(finder); |
---|
357 | |
---|
358 | tasks = finder.getFoundTasks(); |
---|
359 | |
---|
360 | createNewTrie = false; |
---|
361 | |
---|
362 | for (List<ITaskInstance> task : tasks) { |
---|
363 | if (task.size() >= appData.getTrainedSequenceLength()) { |
---|
364 | // Trie must be recreated with a longer sequence length to be sure that |
---|
365 | // the found sequences occurring most often are found in their whole length |
---|
366 | appData.setTrainedSequenceLength(appData.getTrainedSequenceLength() + 1); |
---|
367 | createNewTrie = true; |
---|
368 | break; |
---|
369 | } |
---|
370 | } |
---|
371 | } |
---|
372 | while (createNewTrie); |
---|
373 | |
---|
374 | appData.setLastFoundTasks(tasks); |
---|
375 | |
---|
376 | System.out.println("found " + appData.getLastFoundTasks().size() + " tasks occurring " + |
---|
377 | appData.getLastFoundTasks().getOccurrenceCount() + " times"); |
---|
378 | } |
---|
379 | |
---|
380 | /** |
---|
381 | * @param parent |
---|
382 | * @return |
---|
383 | */ |
---|
384 | private void createNewTrie(RuleApplicationData appData) { |
---|
385 | System.out.println("training trie with a maximum sequence length of " + |
---|
386 | appData.getTrainedSequenceLength()); |
---|
387 | |
---|
388 | appData.getStopWatch().start("training trie"); |
---|
389 | appData.setLastTrie(new Trie<ITaskInstance>(taskComparator)); |
---|
390 | |
---|
391 | List<IUserSession> sessions = appData.getSessions(); |
---|
392 | |
---|
393 | for (IUserSession session : sessions) { |
---|
394 | trainTrie(session, appData); |
---|
395 | } |
---|
396 | |
---|
397 | appData.getStopWatch().stop("training trie"); |
---|
398 | } |
---|
399 | |
---|
400 | /** |
---|
401 | * @param trie |
---|
402 | * @param parent |
---|
403 | */ |
---|
404 | private void trainTrie(IUserSession session, RuleApplicationData appData) { |
---|
405 | List<ITaskInstance> children = session.getExecutedTasks(); |
---|
406 | |
---|
407 | if ((children != null) && (children.size() > 0)) { |
---|
408 | appData.getLastTrie().train(children, appData.getTrainedSequenceLength()); |
---|
409 | } |
---|
410 | } |
---|
411 | |
---|
412 | /** |
---|
413 | * @param appData |
---|
414 | */ |
---|
415 | private void replaceSequencesOccurringMostOften(RuleApplicationData appData) { |
---|
416 | appData.detectedAndReplacedTasks(false); |
---|
417 | |
---|
418 | if ((appData.getLastFoundTasks().size() > 0) && |
---|
419 | (appData.getLastFoundTasks().getOccurrenceCount() > 1)) |
---|
420 | { |
---|
421 | System.out.println("replacing tasks occurrences"); |
---|
422 | |
---|
423 | for (List<ITaskInstance> task : appData.getLastFoundTasks()) { |
---|
424 | ISequence sequence = taskFactory.createNewSequence(); |
---|
425 | |
---|
426 | System.out.println("replacing " + sequence.getId() + ": " + task); |
---|
427 | |
---|
428 | List<ITaskInstance> sequenceInstances = |
---|
429 | replaceTaskOccurrences(task, appData.getSessions(), sequence); |
---|
430 | |
---|
431 | harmonizeSequenceInstancesModel(sequence, sequenceInstances,task.size()); |
---|
432 | appData.detectedAndReplacedTasks(sequenceInstances.size() > 0); |
---|
433 | |
---|
434 | if (sequenceInstances.size() < appData.getLastFoundTasks().getOccurrenceCount()) { |
---|
435 | System.out.println(sequence.getId() + ": replaced task only " + |
---|
436 | sequenceInstances.size() + " times instead of expected " + |
---|
437 | appData.getLastFoundTasks().getOccurrenceCount()); |
---|
438 | } |
---|
439 | } |
---|
440 | } |
---|
441 | |
---|
442 | } |
---|
443 | |
---|
444 | /** |
---|
445 | * |
---|
446 | */ |
---|
447 | private void harmonizeSequenceInstancesModel(ISequence sequence, |
---|
448 | List<ITaskInstance> sequenceInstances, |
---|
449 | int sequenceLength) |
---|
450 | { |
---|
451 | |
---|
452 | // ensure for each subtask that lexically different variants are preserved |
---|
453 | for (int subTaskIndex = 0; subTaskIndex < sequenceLength; subTaskIndex++) { |
---|
454 | List<ITask> subTaskVariants = new LinkedList<ITask>(); |
---|
455 | |
---|
456 | for (ITaskInstance sequenceInstance : sequenceInstances) { |
---|
457 | ITask candidate = sequenceInstance.get(subTaskIndex).getTask(); |
---|
458 | |
---|
459 | boolean found = false; |
---|
460 | |
---|
461 | for (int i = 0; i < subTaskVariants.size(); i++) { |
---|
462 | if (taskComparator.areLexicallyEqual(subTaskVariants.get(i), candidate)) { |
---|
463 | taskBuilder.setTask |
---|
464 | (sequenceInstance.get(subTaskIndex), subTaskVariants.get(i)); |
---|
465 | |
---|
466 | found = true; |
---|
467 | break; |
---|
468 | } |
---|
469 | } |
---|
470 | |
---|
471 | if (!found) { |
---|
472 | subTaskVariants.add(candidate); |
---|
473 | } |
---|
474 | } |
---|
475 | |
---|
476 | // if there are more than one lexically different variant of the sub task at |
---|
477 | // the considered position, adapt the sequence model at that position to have |
---|
478 | // a selection of the different variants. In this case also adapt the |
---|
479 | // generated sequence instances to correctly contain selection instances. If |
---|
480 | // there is only one variant of sub tasks at the given position, simply set |
---|
481 | // this variant as the sub task of the selection. In this case, the instances |
---|
482 | // can be preserved as is |
---|
483 | if (subTaskVariants.size() > 1) { |
---|
484 | ISelection selection = taskFactory.createNewSelection(); |
---|
485 | |
---|
486 | for (ITask variant : subTaskVariants) { |
---|
487 | taskBuilder.addChild(selection, variant); |
---|
488 | } |
---|
489 | |
---|
490 | taskBuilder.addChild(sequence, selection); |
---|
491 | |
---|
492 | for (ITaskInstance instance : sequenceInstances) { |
---|
493 | ITaskInstance selectionInstance = |
---|
494 | taskFactory.createNewTaskInstance(selection); |
---|
495 | taskBuilder.addChild(selectionInstance, instance.get(subTaskIndex)); |
---|
496 | taskBuilder.setTaskInstance(instance, subTaskIndex, selectionInstance); |
---|
497 | } |
---|
498 | } |
---|
499 | else if (subTaskVariants.size() == 1) { |
---|
500 | taskBuilder.addChild(sequence, subTaskVariants.get(0)); |
---|
501 | } |
---|
502 | } |
---|
503 | } |
---|
504 | |
---|
505 | /** |
---|
506 | * @param tree |
---|
507 | */ |
---|
508 | private List<ITaskInstance> replaceTaskOccurrences(List<ITaskInstance> task, |
---|
509 | List<IUserSession> sessions, |
---|
510 | ISequence temporalTaskModel) |
---|
511 | { |
---|
512 | List<ITaskInstance> sequenceInstances = new LinkedList<ITaskInstance>(); |
---|
513 | |
---|
514 | for (IUserSession session : sessions) { |
---|
515 | int index = -1; |
---|
516 | |
---|
517 | do { |
---|
518 | index = getSubListIndex(session, task, ++index); |
---|
519 | |
---|
520 | if (index > -1) { |
---|
521 | sequenceInstances.add |
---|
522 | (RuleUtils.createNewSubSequenceInRange |
---|
523 | (session, index, index + task.size() - 1, temporalTaskModel, |
---|
524 | taskFactory, taskBuilder)); |
---|
525 | } |
---|
526 | } |
---|
527 | while (index > -1); |
---|
528 | } |
---|
529 | |
---|
530 | return sequenceInstances; |
---|
531 | } |
---|
532 | |
---|
533 | /** |
---|
534 | * @param trie |
---|
535 | * @param object |
---|
536 | * @return |
---|
537 | */ |
---|
538 | private int getSubListIndex(ITaskInstanceList list, |
---|
539 | List<ITaskInstance> subList, |
---|
540 | int startIndex) |
---|
541 | { |
---|
542 | boolean matchFound; |
---|
543 | int result = -1; |
---|
544 | |
---|
545 | for (int i = startIndex; i <= list.size() - subList.size(); i++) { |
---|
546 | matchFound = true; |
---|
547 | |
---|
548 | for (int j = 0; j < subList.size(); j++) { |
---|
549 | if (!taskComparator.equals(list.get(i + j), subList.get(j))) { |
---|
550 | matchFound = false; |
---|
551 | break; |
---|
552 | } |
---|
553 | } |
---|
554 | |
---|
555 | if (matchFound) { |
---|
556 | result = i; |
---|
557 | break; |
---|
558 | } |
---|
559 | } |
---|
560 | |
---|
561 | return result; |
---|
562 | } |
---|
563 | |
---|
564 | /** |
---|
565 | * @param trie |
---|
566 | * @param object |
---|
567 | * @return |
---|
568 | */ |
---|
569 | private int getSubListIndex(List<ITaskInstance> list, |
---|
570 | List<ITaskInstance> subList, |
---|
571 | int startIndex) |
---|
572 | { |
---|
573 | boolean matchFound; |
---|
574 | int result = -1; |
---|
575 | |
---|
576 | for (int i = startIndex; i <= list.size() - subList.size(); i++) { |
---|
577 | matchFound = true; |
---|
578 | |
---|
579 | for (int j = 0; j < subList.size(); j++) { |
---|
580 | if (!taskComparator.equals(list.get(i + j), subList.get(j))) { |
---|
581 | matchFound = false; |
---|
582 | break; |
---|
583 | } |
---|
584 | } |
---|
585 | |
---|
586 | if (matchFound) { |
---|
587 | result = i; |
---|
588 | break; |
---|
589 | } |
---|
590 | } |
---|
591 | |
---|
592 | return result; |
---|
593 | } |
---|
594 | |
---|
595 | /** |
---|
596 | * @author Patrick Harms |
---|
597 | */ |
---|
598 | private class MaxCountAndLongestTasksFinder implements TrieProcessor<ITaskInstance> { |
---|
599 | |
---|
600 | /** |
---|
601 | * |
---|
602 | */ |
---|
603 | private int currentCount; |
---|
604 | |
---|
605 | /** |
---|
606 | * |
---|
607 | */ |
---|
608 | private List<List<ITaskInstance>> foundTasks = new LinkedList<List<ITaskInstance>>(); |
---|
609 | |
---|
610 | /** |
---|
611 | * |
---|
612 | */ |
---|
613 | public MaxCountAndLongestTasksFinder() { |
---|
614 | super(); |
---|
615 | this.currentCount = 0; |
---|
616 | } |
---|
617 | |
---|
618 | /* (non-Javadoc) |
---|
619 | * @see de.ugoe.cs.autoquest.usageprofiles.TrieProcessor#process(java.util.List, int) |
---|
620 | */ |
---|
621 | @Override |
---|
622 | public TrieProcessor.Result process(List<ITaskInstance> foundTask, int count) { |
---|
623 | if (foundTask.size() < 2) { |
---|
624 | // ignore single tasks |
---|
625 | return TrieProcessor.Result.CONTINUE; |
---|
626 | } |
---|
627 | |
---|
628 | if (count < 2) { |
---|
629 | // ignore singular occurrences |
---|
630 | return TrieProcessor.Result.SKIP_NODE; |
---|
631 | } |
---|
632 | |
---|
633 | if (this.currentCount > count) { |
---|
634 | // ignore this children of this task, as they may have only smaller counts than |
---|
635 | // the already found tasks |
---|
636 | return TrieProcessor.Result.SKIP_NODE; |
---|
637 | } |
---|
638 | |
---|
639 | if (this.currentCount < count) { |
---|
640 | // the provided task occurs more often that all tasks found so far. |
---|
641 | // clear all found tasks and use the new count as the one searched for |
---|
642 | foundTasks.clear(); |
---|
643 | this.currentCount = count; |
---|
644 | } |
---|
645 | |
---|
646 | if (this.currentCount == count) { |
---|
647 | // the task is of interest. Sort it into the other found tasks so that |
---|
648 | // the longest come first |
---|
649 | boolean added = false; |
---|
650 | for (int i = 0; i < foundTasks.size(); i++) { |
---|
651 | if (foundTasks.get(i).size() < foundTask.size()) { |
---|
652 | // defensive copy |
---|
653 | foundTasks.add(i, new LinkedList<ITaskInstance>(foundTask)); // defensive copy |
---|
654 | added = true; |
---|
655 | break; |
---|
656 | } |
---|
657 | } |
---|
658 | |
---|
659 | if (!added) { |
---|
660 | foundTasks.add(new LinkedList<ITaskInstance>(foundTask)); // defensive copy |
---|
661 | } |
---|
662 | } |
---|
663 | |
---|
664 | return TrieProcessor.Result.CONTINUE; |
---|
665 | } |
---|
666 | |
---|
667 | /** |
---|
668 | * @return |
---|
669 | */ |
---|
670 | public Tasks getFoundTasks() { |
---|
671 | removePermutationsOfShorterTasks(); |
---|
672 | return new Tasks(currentCount, foundTasks); |
---|
673 | } |
---|
674 | |
---|
675 | /** |
---|
676 | * |
---|
677 | */ |
---|
678 | private void removePermutationsOfShorterTasks() { |
---|
679 | // now iterate the sorted list and for each task remove all other tasks that are shorter |
---|
680 | // (i.e. come later in the sorted list) and that represent a subpart of the task |
---|
681 | for (int i = 0; i < foundTasks.size(); i++) { |
---|
682 | for (int j = i + 1; j < foundTasks.size();) { |
---|
683 | if (foundTasks.get(j).size() < foundTasks.get(i).size()) { |
---|
684 | // found a task that is a potential subtask. Check for this and remove the |
---|
685 | // subtask if needed |
---|
686 | List<ITaskInstance> longTask = foundTasks.get(i); |
---|
687 | List<ITaskInstance> shortTask = foundTasks.get(j); |
---|
688 | |
---|
689 | if (getSubListIndex(longTask, shortTask, 0) > -1) { |
---|
690 | foundTasks.remove(j); |
---|
691 | } |
---|
692 | else { |
---|
693 | j++; |
---|
694 | } |
---|
695 | } |
---|
696 | else { |
---|
697 | j++; |
---|
698 | } |
---|
699 | } |
---|
700 | } |
---|
701 | } |
---|
702 | |
---|
703 | } |
---|
704 | |
---|
705 | /** |
---|
706 | * |
---|
707 | */ |
---|
708 | private static class RuleApplicationData { |
---|
709 | |
---|
710 | /** |
---|
711 | * |
---|
712 | */ |
---|
713 | private List<IUserSession> sessions; |
---|
714 | |
---|
715 | /** |
---|
716 | * |
---|
717 | */ |
---|
718 | private Trie<ITaskInstance> lastTrie; |
---|
719 | |
---|
720 | /** |
---|
721 | * default and minimum trained sequence length is 3 |
---|
722 | */ |
---|
723 | private int trainedSequenceLength = 3; |
---|
724 | |
---|
725 | /** |
---|
726 | * |
---|
727 | */ |
---|
728 | private Tasks lastFoundTasks = new Tasks(Integer.MAX_VALUE, null); |
---|
729 | |
---|
730 | /** |
---|
731 | * |
---|
732 | */ |
---|
733 | private boolean detectedAndReplacedTasks; |
---|
734 | |
---|
735 | /** |
---|
736 | * |
---|
737 | */ |
---|
738 | private RuleApplicationResult result = new RuleApplicationResult(); |
---|
739 | |
---|
740 | /** |
---|
741 | * |
---|
742 | */ |
---|
743 | private StopWatch stopWatch = new StopWatch(); |
---|
744 | |
---|
745 | /** |
---|
746 | * |
---|
747 | */ |
---|
748 | private RuleApplicationData(List<IUserSession> sessions) { |
---|
749 | this.sessions = sessions; |
---|
750 | } |
---|
751 | |
---|
752 | /** |
---|
753 | * @return the tree |
---|
754 | */ |
---|
755 | private List<IUserSession> getSessions() { |
---|
756 | return sessions; |
---|
757 | } |
---|
758 | |
---|
759 | /** |
---|
760 | * @param lastTrie the lastTrie to set |
---|
761 | */ |
---|
762 | private void setLastTrie(Trie<ITaskInstance> lastTrie) { |
---|
763 | this.lastTrie = lastTrie; |
---|
764 | } |
---|
765 | |
---|
766 | /** |
---|
767 | * @return the lastTrie |
---|
768 | */ |
---|
769 | private Trie<ITaskInstance> getLastTrie() { |
---|
770 | return lastTrie; |
---|
771 | } |
---|
772 | |
---|
773 | /** |
---|
774 | * @param trainedSequenceLength the trainedSequenceLength to set |
---|
775 | */ |
---|
776 | private void setTrainedSequenceLength(int trainedSequenceLength) { |
---|
777 | this.trainedSequenceLength = trainedSequenceLength; |
---|
778 | } |
---|
779 | |
---|
780 | /** |
---|
781 | * @return the trainedSequenceLength |
---|
782 | */ |
---|
783 | private int getTrainedSequenceLength() { |
---|
784 | return trainedSequenceLength; |
---|
785 | } |
---|
786 | |
---|
787 | /** |
---|
788 | * @param lastFoundSequences the lastFoundSequences to set |
---|
789 | */ |
---|
790 | private void setLastFoundTasks(Tasks lastFoundSequences) { |
---|
791 | this.lastFoundTasks = lastFoundSequences; |
---|
792 | } |
---|
793 | |
---|
794 | /** |
---|
795 | * @return the lastFoundSequences |
---|
796 | */ |
---|
797 | private Tasks getLastFoundTasks() { |
---|
798 | return lastFoundTasks; |
---|
799 | } |
---|
800 | |
---|
801 | /** |
---|
802 | * |
---|
803 | */ |
---|
804 | private void detectedAndReplacedTasks(boolean detectedAndReplacedTasks) { |
---|
805 | this.detectedAndReplacedTasks = detectedAndReplacedTasks; |
---|
806 | } |
---|
807 | |
---|
808 | /** |
---|
809 | * |
---|
810 | */ |
---|
811 | private boolean detectedAndReplacedTasks() { |
---|
812 | return detectedAndReplacedTasks; |
---|
813 | } |
---|
814 | |
---|
815 | /** |
---|
816 | * @return the result |
---|
817 | */ |
---|
818 | private RuleApplicationResult getResult() { |
---|
819 | return result; |
---|
820 | } |
---|
821 | |
---|
822 | /** |
---|
823 | * @return the stopWatch |
---|
824 | */ |
---|
825 | private StopWatch getStopWatch() { |
---|
826 | return stopWatch; |
---|
827 | } |
---|
828 | |
---|
829 | } |
---|
830 | |
---|
831 | |
---|
832 | /** |
---|
833 | * @author Patrick Harms |
---|
834 | */ |
---|
835 | private static class Tasks implements Iterable<List<ITaskInstance>> { |
---|
836 | |
---|
837 | /** |
---|
838 | * |
---|
839 | */ |
---|
840 | private int occurrenceCount; |
---|
841 | |
---|
842 | /** |
---|
843 | * |
---|
844 | */ |
---|
845 | private List<List<ITaskInstance>> sequences; |
---|
846 | |
---|
847 | /** |
---|
848 | * @param occurrenceCount |
---|
849 | * @param sequences |
---|
850 | */ |
---|
851 | private Tasks(int occurrenceCount, List<List<ITaskInstance>> sequences) { |
---|
852 | super(); |
---|
853 | this.occurrenceCount = occurrenceCount; |
---|
854 | this.sequences = sequences; |
---|
855 | } |
---|
856 | |
---|
857 | /** |
---|
858 | * @return |
---|
859 | */ |
---|
860 | private int getOccurrenceCount() { |
---|
861 | return occurrenceCount; |
---|
862 | } |
---|
863 | |
---|
864 | /** |
---|
865 | * @return |
---|
866 | */ |
---|
867 | private int size() { |
---|
868 | return this.sequences.size(); |
---|
869 | } |
---|
870 | |
---|
871 | /** |
---|
872 | * |
---|
873 | */ |
---|
874 | |
---|
875 | /* (non-Javadoc) |
---|
876 | * @see java.lang.Iterable#iterator() |
---|
877 | */ |
---|
878 | @Override |
---|
879 | public Iterator<List<ITaskInstance>> iterator() { |
---|
880 | return this.sequences.iterator(); |
---|
881 | } |
---|
882 | |
---|
883 | } |
---|
884 | |
---|
885 | /** |
---|
886 | * |
---|
887 | */ |
---|
888 | private static class TaskComparator implements SymbolComparator<ITaskInstance> { |
---|
889 | |
---|
890 | /** */ |
---|
891 | private static final long serialVersionUID = 1L; |
---|
892 | |
---|
893 | /** */ |
---|
894 | private Comparer comparer; |
---|
895 | |
---|
896 | /** */ |
---|
897 | private Comparer lexicalComparer; |
---|
898 | |
---|
899 | /** */ |
---|
900 | private HashMap<Long, Boolean> equalityBuffer = new HashMap<Long, Boolean>(); |
---|
901 | |
---|
902 | /** */ |
---|
903 | private HashMap<Long, Boolean> lexicalEqualityBuffer; |
---|
904 | |
---|
905 | /** |
---|
906 | * |
---|
907 | */ |
---|
908 | public TaskComparator(TaskEqualityRuleManager taskEqualityRuleManager, |
---|
909 | TaskEquality minimalNodeEquality) |
---|
910 | { |
---|
911 | super(); |
---|
912 | |
---|
913 | if (minimalNodeEquality == TaskEquality.LEXICALLY_EQUAL) { |
---|
914 | comparer = new LexicalComparer(taskEqualityRuleManager); |
---|
915 | } |
---|
916 | else if (minimalNodeEquality == TaskEquality.SYNTACTICALLY_EQUAL) { |
---|
917 | comparer = new SyntacticalComparer(taskEqualityRuleManager); |
---|
918 | } |
---|
919 | else if (minimalNodeEquality == TaskEquality.SEMANTICALLY_EQUAL) { |
---|
920 | comparer = new SemanticalComparer(taskEqualityRuleManager); |
---|
921 | } |
---|
922 | else { |
---|
923 | comparer = new DefaultComparer(taskEqualityRuleManager, minimalNodeEquality); |
---|
924 | } |
---|
925 | |
---|
926 | if (minimalNodeEquality == TaskEquality.LEXICALLY_EQUAL) { |
---|
927 | lexicalComparer = comparer; |
---|
928 | lexicalEqualityBuffer = equalityBuffer; |
---|
929 | } |
---|
930 | else { |
---|
931 | lexicalComparer = new LexicalComparer(taskEqualityRuleManager); |
---|
932 | lexicalEqualityBuffer = new HashMap<Long, Boolean>(); |
---|
933 | } |
---|
934 | } |
---|
935 | |
---|
936 | /* (non-Javadoc) |
---|
937 | * @see de.ugoe.cs.autoquest.tasktrees.temporalrelation.SymbolComparator#equals(java.lang.Object, java.lang.Object) |
---|
938 | */ |
---|
939 | @Override |
---|
940 | public boolean equals(ITaskInstance taskInstance1, ITaskInstance taskInstance2) { |
---|
941 | return equals(taskInstance1.getTask(), taskInstance2.getTask()); |
---|
942 | } |
---|
943 | |
---|
944 | /** |
---|
945 | * |
---|
946 | */ |
---|
947 | public boolean equals(ITask task1, ITask task2) { |
---|
948 | Boolean result; |
---|
949 | |
---|
950 | if (task1 != task2) { |
---|
951 | if ((task1 instanceof IEventTask) && (task2 instanceof IEventTask)) { |
---|
952 | long key = ((long) System.identityHashCode(task1)) << 32; |
---|
953 | key += System.identityHashCode(task2); |
---|
954 | |
---|
955 | result = equalityBuffer.get(key); |
---|
956 | |
---|
957 | if (result == null) { |
---|
958 | result = comparer.compare(task1, task2); |
---|
959 | equalityBuffer.put(key, result); |
---|
960 | } |
---|
961 | } |
---|
962 | else { |
---|
963 | result = false; |
---|
964 | } |
---|
965 | } |
---|
966 | else { |
---|
967 | result = true; |
---|
968 | } |
---|
969 | |
---|
970 | return result; |
---|
971 | } |
---|
972 | |
---|
973 | /** |
---|
974 | * |
---|
975 | */ |
---|
976 | public boolean areLexicallyEqual(ITask task1, ITask task2) { |
---|
977 | Boolean result; |
---|
978 | |
---|
979 | if (task1 != task2) { |
---|
980 | long key = ((long) System.identityHashCode(task1)) << 32; |
---|
981 | key += System.identityHashCode(task2); |
---|
982 | |
---|
983 | result = lexicalEqualityBuffer.get(key); |
---|
984 | |
---|
985 | if (result == null) { |
---|
986 | result = lexicalComparer.compare(task1, task2); |
---|
987 | lexicalEqualityBuffer.put(key, result); |
---|
988 | } |
---|
989 | } |
---|
990 | else { |
---|
991 | result = true; |
---|
992 | } |
---|
993 | |
---|
994 | return result; |
---|
995 | } |
---|
996 | } |
---|
997 | |
---|
998 | /** |
---|
999 | * |
---|
1000 | */ |
---|
1001 | private static interface Comparer { |
---|
1002 | |
---|
1003 | /** |
---|
1004 | * |
---|
1005 | */ |
---|
1006 | boolean compare(ITask task1, ITask task2); |
---|
1007 | } |
---|
1008 | |
---|
1009 | /** |
---|
1010 | * |
---|
1011 | */ |
---|
1012 | private static class LexicalComparer implements Comparer { |
---|
1013 | |
---|
1014 | /** |
---|
1015 | * <p> |
---|
1016 | * the task equality manager needed for comparing tasks with each other |
---|
1017 | * </p> |
---|
1018 | */ |
---|
1019 | private TaskEqualityRuleManager taskEqualityRuleManager; |
---|
1020 | |
---|
1021 | /** |
---|
1022 | * |
---|
1023 | */ |
---|
1024 | public LexicalComparer(TaskEqualityRuleManager taskEqualityRuleManager) { |
---|
1025 | this.taskEqualityRuleManager = taskEqualityRuleManager; |
---|
1026 | } |
---|
1027 | |
---|
1028 | /** |
---|
1029 | * |
---|
1030 | */ |
---|
1031 | public boolean compare(ITask task1, ITask task2) { |
---|
1032 | return taskEqualityRuleManager.areLexicallyEqual(task1, task2); |
---|
1033 | } |
---|
1034 | } |
---|
1035 | |
---|
1036 | /** |
---|
1037 | * |
---|
1038 | */ |
---|
1039 | private static class SyntacticalComparer implements Comparer { |
---|
1040 | |
---|
1041 | /** |
---|
1042 | * <p> |
---|
1043 | * the task equality manager needed for comparing tasks with each other |
---|
1044 | * </p> |
---|
1045 | */ |
---|
1046 | private TaskEqualityRuleManager taskEqualityRuleManager; |
---|
1047 | |
---|
1048 | /** |
---|
1049 | * |
---|
1050 | */ |
---|
1051 | public SyntacticalComparer(TaskEqualityRuleManager taskEqualityRuleManager) { |
---|
1052 | this.taskEqualityRuleManager = taskEqualityRuleManager; |
---|
1053 | } |
---|
1054 | |
---|
1055 | /** |
---|
1056 | * |
---|
1057 | */ |
---|
1058 | public boolean compare(ITask task1, ITask task2) { |
---|
1059 | return taskEqualityRuleManager.areSyntacticallyEqual(task1, task2); |
---|
1060 | } |
---|
1061 | } |
---|
1062 | |
---|
1063 | /** |
---|
1064 | * |
---|
1065 | */ |
---|
1066 | private static class SemanticalComparer implements Comparer { |
---|
1067 | |
---|
1068 | /** |
---|
1069 | * <p> |
---|
1070 | * the task equality manager needed for comparing tasks with each other |
---|
1071 | * </p> |
---|
1072 | */ |
---|
1073 | private TaskEqualityRuleManager taskEqualityRuleManager; |
---|
1074 | |
---|
1075 | /** |
---|
1076 | * |
---|
1077 | */ |
---|
1078 | public SemanticalComparer(TaskEqualityRuleManager taskEqualityRuleManager) { |
---|
1079 | this.taskEqualityRuleManager = taskEqualityRuleManager; |
---|
1080 | } |
---|
1081 | |
---|
1082 | /** |
---|
1083 | * |
---|
1084 | */ |
---|
1085 | public boolean compare(ITask task1, ITask task2) { |
---|
1086 | return taskEqualityRuleManager.areSemanticallyEqual(task1, task2); |
---|
1087 | } |
---|
1088 | } |
---|
1089 | |
---|
1090 | /** |
---|
1091 | * |
---|
1092 | */ |
---|
1093 | private static class DefaultComparer implements Comparer { |
---|
1094 | |
---|
1095 | /** |
---|
1096 | * <p> |
---|
1097 | * the task equality manager needed for comparing tasks with each other |
---|
1098 | * </p> |
---|
1099 | */ |
---|
1100 | private TaskEqualityRuleManager taskEqualityRuleManager; |
---|
1101 | |
---|
1102 | /** |
---|
1103 | * <p> |
---|
1104 | * the minimal task equality two identified sublists need to have to consider them as equal |
---|
1105 | * </p> |
---|
1106 | */ |
---|
1107 | private TaskEquality minimalNodeEquality; |
---|
1108 | |
---|
1109 | /** |
---|
1110 | * |
---|
1111 | */ |
---|
1112 | public DefaultComparer(TaskEqualityRuleManager taskEqualityRuleManager, |
---|
1113 | TaskEquality minimalNodeEquality) |
---|
1114 | { |
---|
1115 | this.taskEqualityRuleManager = taskEqualityRuleManager; |
---|
1116 | this.minimalNodeEquality = minimalNodeEquality; |
---|
1117 | } |
---|
1118 | |
---|
1119 | /** |
---|
1120 | * |
---|
1121 | */ |
---|
1122 | public boolean compare(ITask task1, ITask task2) { |
---|
1123 | return taskEqualityRuleManager.areAtLeastEqual(task1, task2, minimalNodeEquality); |
---|
1124 | } |
---|
1125 | } |
---|
1126 | } |
---|