() {
public int compare(Match m1, Match m2) {
@@ -229,7 +223,5 @@
-
-
- // Harmonize matches: finding double matches and merge them
+ // TODO: Harmonize matches: finding double matches and merge them
/*
Console.traceln(Level.INFO, "harmonizing matches");
@@ -241,5 +233,5 @@
if(i!=j) {
if(matchseqs.get(i).equals(matchseqs.get(j))) {
- matchseqs.get(i).addOccurencesOf(matchseqs.get(j));
+ //matchseqs.get(i).addOccurencesOf(matchseqs.get(j));
matchseqs.remove(j);
@@ -258,10 +250,5 @@
}
Collections.sort(matchseqs, comparator);
- */
-
-
-
-
-
+ */
@@ -269,12 +256,15 @@
for (int i = 0; i < matchseqs.size(); i++) {
// Every pattern consists of 2 sequences, therefore the minimum
- // occurences here is 2.
- // We just need the sequences also occuring in other sequences as
+ // occurrences here is 2.
+ // We just need the sequences also occurring in other sequences as
// well
if (matchseqs.get(i).occurenceCount() > 2) {
- matchseqs.get(i).getFirstSequence().printSequence();
- matchseqs.get(i).getSecondSequence().printSequence();
- System.out.println("Found pattern "
- + matchseqs.get(i).occurenceCount() + " times");
+
+ ITask task = matchAsTask(appData,matchseqs.get(i));
+ System.out.println(task);
+ //matchseqs.get(i).getFirstSequence().printSequence();
+ //matchseqs.get(i).getSecondSequence().printSequence();
+ //System.out.println("Found pattern "
+ // + matchseqs.get(i).occurenceCount() + " times");
System.out.println();
}
@@ -284,20 +274,19 @@
alignments = null;
- /*
- * do {
- *
- * // appData.getStopWatch().start("whole loop"); //
- * detectAndReplaceIterations(appData);
- *
- * // appData.getStopWatch().start("task replacement"); //
- * detectAndReplaceTasks(appData); //
- * appData.getStopWatch().stop("task replacement"); //
- * appData.getStopWatch().stop("whole loop");
- *
- * // appData.getStopWatch().dumpStatistics(System.out); //
- * appData.getStopWatch().reset();
- *
- * } while (appData.detectedAndReplacedTasks());
- */
+ do {
+
+ appData.getStopWatch().start("whole loop"); //
+ detectAndReplaceIterations(appData);
+
+ appData.getStopWatch().start("task replacement"); //
+ //detectAndReplaceTasks(appData); //
+ appData.getStopWatch().stop("task replacement"); //
+ appData.getStopWatch().stop("whole loop");
+
+ appData.getStopWatch().dumpStatistics(System.out); //
+ appData.getStopWatch().reset();
+
+ } while (appData.detectedAndReplacedTasks());
+
Console.println("created "
@@ -316,4 +305,6 @@
}
+
+
/**
*
@@ -344,5 +335,4 @@
List sessions = appData.getSessions();
int sessionNo = 0;
- numberseqs = new ArrayList();
for (IUserSession session : sessions) {
Console.traceln(Level.FINE, "handling " + (++sessionNo) + ". "
@@ -357,4 +347,5 @@
uniqueTasks.addSymbol(taskInstance, taskInstance.getTask());
templist.getSequence()[i] = taskInstance.getTask().getId();
+
} else {
taskBuilder.setTask(taskInstance, task);
@@ -362,7 +353,8 @@
unifiedTasks++;
}
- }
- templist.setId(numberseqs.size());
- numberseqs.add(templist);
+ appData.getNumber2Task().put(templist.getSequence()[i], taskInstance.getTask());
+ }
+ templist.setId(appData.getNumberSequences().size());
+ appData.getNumberSequences().add(templist);
comparator.clearBuffers();
}
@@ -505,4 +497,50 @@
}
+
+ ITask matchAsTask(RuleApplicationData appData,Match m) {
+
+ ISequence sequence = taskFactory.createNewSequence();
+
+ int[] first = m.getFirstSequence().getSequence();
+ int[] second = m.getSecondSequence().getSequence();
+ System.out.println("Number2Task size: " +appData.getNumber2Task().size());
+
+ //Both sequences of a match are equally long
+ for(int i=0;i
@@ -579,5 +617,5 @@
appData.getStopWatch().start("detecting tasks");
- getSequencesOccuringMostOften(appData);
+ //getSequencesOccuringMostOften(appData);
appData.getStopWatch().stop("detecting tasks");
@@ -587,8 +625,12 @@
appData.getStopWatch().stop("replacing tasks");
- Console.traceln(Level.INFO, "detected and replaced "
- + appData.getLastFoundTasks().size() + " tasks occuring "
- + appData.getLastFoundTasks().getOccurrenceCount() + " times");
- }
+
+ //Console.traceln(Level.INFO, "detected and replaced "
+ // + appData.getLastFoundTasks().size() + " tasks occuring "
+ // + appData.getLastFoundTasks().getOccurrenceCount() + " times");
+ }
+
+
+
/**
@@ -597,127 +639,10 @@
* this rule
*/
- private void getSequencesOccuringMostOften(RuleApplicationData appData) {
- Console.traceln(Level.FINE, "determining most prominent tasks");
-
- Tasks tasks;
- boolean createNewTrie = (appData.getLastTrie() == null)
- || appData.detectedAndReplacedTasks(); // tree has changed
-
- do {
- if (createNewTrie) {
- createNewTrie(appData);
- }
-
- MaxCountAndLongestTasksFinder finder = new MaxCountAndLongestTasksFinder();
- appData.getLastTrie().process(finder);
-
- tasks = finder.getFoundTasks();
-
- createNewTrie = false;
-
- for (List task : tasks) {
- if (task.size() >= appData.getTrainedSequenceLength()) {
- // Trie must be recreated with a longer sequence length to
- // be sure that
- // the found sequences occurring most often are found in
- // their whole length
- appData.setTrainedSequenceLength(appData
- .getTrainedSequenceLength() + 1);
- createNewTrie = true;
- break;
- }
- }
- } while (createNewTrie);
-
- // create a normal trie for comparison
- /*
- * System.out.println("creating test trie for comparison");
- *
- * appData.getStopWatch().start("testtrie"); Trie trie =
- * new Trie(identityTaskComparator);
- *
- * for (IUserSession session : appData.getSessions()) {
- * trie.train(session.getExecutedTasks(),
- * appData.getTrainedSequenceLength()); }
- *
- * appData.getStopWatch().stop("testtrie");
- *
- * MaxCountAndLongestTasksFinder finder = new
- * MaxCountAndLongestTasksFinder(); trie.process(finder);
- *
- * boolean allTasksEqual = finder.getFoundTasks().size() ==
- * tasks.size();
- *
- * allTasksEqual &= finder.getFoundTasks().occurrenceCount ==
- * tasks.occurrenceCount;
- *
- * for (List task1 : finder.getFoundTasks()) { boolean
- * foundTask = false; for (List task2 : tasks) { boolean
- * tasksEqual = false; if (task1.size() == task2.size()) { tasksEqual =
- * true; for (int i = 0; i < task1.size(); i++) { if
- * (!identityTaskComparator.equals(task1.get(i).getTask(),
- * task2.get(i).getTask())) { tasksEqual = false; } } }
- *
- * if (tasksEqual) { foundTask = true; break; } }
- *
- * if (!foundTask) { System.out.println("different is " + task1);
- * allTasksEqual = false; break; } }
- *
- * if (!allTasksEqual) { System.out.println(finder.getFoundTasks());
- * System.out.println(tasks);
- *
- * throw new
- * IllegalArgumentException("both tries calculated different subsequences"
- * ); }
- *
- * /*TrieStatisticsDumper dumper = new TrieStatisticsDumper();
- * appData.getLastTrie().process(dumper); dumper.dumpCounters();
- */
-
- appData.setLastFoundTasks(tasks);
-
- Console.traceln(Level.FINE, "found "
- + appData.getLastFoundTasks().size() + " tasks " + "occurring "
- + appData.getLastFoundTasks().getOccurrenceCount() + " times");
- }
-
- /**
- * @param appData
- * the rule application data combining all data used for applying
- * this rule
- */
- private void createNewTrie(RuleApplicationData appData) {
- Console.traceln(
- Level.FINER,
- "training trie with a maximum sequence length of "
- + appData.getTrainedSequenceLength());
-
- appData.getStopWatch().start("training trie");
-
- // we prepared the task instances to refer to unique tasks, if they are
- // treated
- // as equal. Therefore, we just compare the identity of the tasks of the
- // task
- // instances
- appData.setLastTrie(new TaskInstanceTrie(identityTaskHandlingStrategy));
-
- appData.getLastTrie().trainSessions(appData.getSessions(),
- appData.getTrainedSequenceLength());
-
- appData.getStopWatch().stop("training trie");
- }
-
- /**
- * @param appData
- * the rule application data combining all data used for applying
- * this rule
- */
private void replaceSequencesOccurringMostOften(RuleApplicationData appData) {
appData.detectedAndReplacedTasks(false);
- if ((appData.getLastFoundTasks().size() > 0)
- && (appData.getLastFoundTasks().getOccurrenceCount() > 1)) {
+
Console.traceln(Level.FINER, "replacing tasks occurrences");
-
+ /*
for (List task : appData.getLastFoundTasks()) {
ISequence sequence = taskFactory.createNewSequence();
@@ -746,7 +671,7 @@
}
}
- }
-
- }
+ */
+ }
+
/**
@@ -880,214 +805,4 @@
}
- /**
- * @param trie
- * @param object
- * @return
- */
- private int getSubListIndex(List list,
- List subList, int startIndex) {
- boolean matchFound;
- int result = -1;
-
- for (int i = startIndex; i <= list.size() - subList.size(); i++) {
- matchFound = true;
-
- for (int j = 0; j < subList.size(); j++) {
- // we prepared the task instances to refer to unique tasks, if
- // they are treated
- // as equal. Therefore, we just compare the identity of the
- // tasks of the task
- // instances
- if (list.get(i + j) != subList.get(j)) {
- matchFound = false;
- break;
- }
- }
-
- if (matchFound) {
- result = i;
- break;
- }
- }
-
- return result;
- }
-
- /**
- * @author Patrick Harms
- */
- private class MaxCountAndLongestTasksFinder implements
- TrieProcessor {
-
- /**
- *
- */
- private int currentCount;
-
- /**
- *
- */
- private List> foundTasks = new LinkedList>();
-
- /**
- *
- */
- public MaxCountAndLongestTasksFinder() {
- super();
- this.currentCount = 0;
- }
-
- /*
- * (non-Javadoc)
- *
- * @see
- * de.ugoe.cs.autoquest.usageprofiles.TrieProcessor#process(java.util
- * .List, int)
- */
- @Override
- public TrieProcessor.Result process(List foundTask,
- int count) {
- if (foundTask.size() < 2) {
- // ignore single tasks
- return TrieProcessor.Result.CONTINUE;
- }
-
- if (count < 2) {
- // ignore singular occurrences
- return TrieProcessor.Result.SKIP_NODE;
- }
-
- if (this.currentCount > count) {
- // ignore this children of this task, as they may have only
- // smaller counts than
- // the already found tasks
- return TrieProcessor.Result.SKIP_NODE;
- }
-
- if (this.currentCount < count) {
- // the provided task occurs more often that all tasks found so
- // far.
- // clear all found tasks and use the new count as the one
- // searched for
- foundTasks.clear();
- this.currentCount = count;
- }
-
- if (this.currentCount == count) {
- // the task is of interest. Sort it into the other found tasks
- // so that
- // the longest come first
- boolean added = false;
- for (int i = 0; i < foundTasks.size(); i++) {
- if (foundTasks.get(i).size() < foundTask.size()) {
- // defensive copy
- foundTasks.add(i, new LinkedList(
- foundTask)); // defensive copy
- added = true;
- break;
- }
- }
-
- if (!added) {
- foundTasks.add(new LinkedList(foundTask)); // defensive
- // copy
- }
- }
-
- return TrieProcessor.Result.CONTINUE;
- }
-
- /**
- * @return
- */
- public Tasks getFoundTasks() {
- removePermutationsOfShorterTasks();
- return new Tasks(currentCount, foundTasks);
- }
-
- /**
- *
- */
- private void removePermutationsOfShorterTasks() {
- // now iterate the sorted list and for each task remove all other
- // tasks that are shorter
- // (i.e. come later in the sorted list) and that represent a subpart
- // of the task
- for (int i = 0; i < foundTasks.size(); i++) {
- for (int j = i + 1; j < foundTasks.size();) {
- if (foundTasks.get(j).size() < foundTasks.get(i).size()) {
- // found a task that is a potential subtask. Check for
- // this and remove the
- // subtask if needed
- List longTask = foundTasks.get(i);
- List shortTask = foundTasks.get(j);
-
- if (getSubListIndex(longTask, shortTask, 0) > -1) {
- foundTasks.remove(j);
- } else {
- j++;
- }
- } else {
- j++;
- }
- }
- }
- }
-
- }
-
- // /**
- // * @author Patrick Harms
- // */
- // private class TrieStatisticsDumper implements
- // TrieProcessor {
- //
- // /**
- // *
- // */
- // private Map counters = new HashMap();
- //
- // /* (non-Javadoc)
- // * @see
- // de.ugoe.cs.autoquest.usageprofiles.TrieProcessor#process(java.util.List,
- // int)
- // */
- // @Override
- // public TrieProcessor.Result process(List subsequence, int
- // count) {
- // if (subsequence.size() == 1) {
- // Integer value = counters.get(count);
- //
- // if (value == null) {
- // value = 0;
- // }
- //
- // counters.put(count, value + 1);
- //
- // return TrieProcessor.Result.CONTINUE;
- // }
- // else {
- // // ignore singular occurrences
- // return TrieProcessor.Result.SKIP_NODE;
- // }
- // }
- //
- // /**
- // * @return
- // */
- // public void dumpCounters() {
- // int dumpedCounters = 0;
- //
- // int count = 1;
- // while (dumpedCounters < counters.size()) {
- // Integer value = counters.get(count++);
- // if (value != null) {
- // System.out.println(value + " symbols occurred " + count + " times");
- // dumpedCounters++;
- // }
- // }
- // }
- //
- // }
/**
@@ -1096,4 +811,10 @@
private static class RuleApplicationData {
+
+ private HashMap number2task;
+
+
+ private ArrayList numberseqs;
+
/**
*
@@ -1101,18 +822,5 @@
private List sessions;
- /**
- *
- */
- private TaskInstanceTrie lastTrie;
-
- /**
- * default and minimum trained sequence length is 3
- */
- private int trainedSequenceLength = 3;
-
- /**
- *
- */
- private Tasks lastFoundTasks = new Tasks(Integer.MAX_VALUE, null);
+
/**
@@ -1124,10 +832,10 @@
*
*/
- private RuleApplicationResult result = new RuleApplicationResult();
+ private RuleApplicationResult result;
/**
*
*/
- private StopWatch stopWatch = new StopWatch();
+ private StopWatch stopWatch;
/**
@@ -1136,4 +844,8 @@
private RuleApplicationData(List sessions) {
this.sessions = sessions;
+ numberseqs = new ArrayList();
+ number2task = new HashMap();
+ stopWatch= new StopWatch();
+ result = new RuleApplicationResult();
}
@@ -1145,48 +857,10 @@
}
- /**
- * @param lastTrie
- * the lastTrie to set
- */
- private void setLastTrie(TaskInstanceTrie lastTrie) {
- this.lastTrie = lastTrie;
- }
-
- /**
- * @return the lastTrie
- */
- private TaskInstanceTrie getLastTrie() {
- return lastTrie;
- }
-
- /**
- * @param trainedSequenceLength
- * the trainedSequenceLength to set
- */
- private void setTrainedSequenceLength(int trainedSequenceLength) {
- this.trainedSequenceLength = trainedSequenceLength;
- }
-
- /**
- * @return the trainedSequenceLength
- */
- private int getTrainedSequenceLength() {
- return trainedSequenceLength;
- }
-
- /**
- * @param lastFoundSequences
- * the lastFoundSequences to set
- */
- private void setLastFoundTasks(Tasks lastFoundSequences) {
- this.lastFoundTasks = lastFoundSequences;
- }
-
- /**
- * @return the lastFoundSequences
- */
- private Tasks getLastFoundTasks() {
- return lastFoundTasks;
- }
+
+ private ArrayList getNumberSequences() {
+ return numberseqs;
+ }
+
+
/**
@@ -1218,147 +892,12 @@
}
- }
-
- /**
- * @author Patrick Harms
- */
- private static class Tasks implements Iterable> {
-
- /**
- *
- */
- private int occurrenceCount;
-
- /**
- *
- */
- private List> sequences;
-
- /**
- * @param occurrenceCount
- * @param sequences
- */
- private Tasks(int occurrenceCount, List> sequences) {
- super();
- this.occurrenceCount = occurrenceCount;
- this.sequences = sequences;
- }
-
- /**
- * @return
- */
- private int getOccurrenceCount() {
- return occurrenceCount;
- }
-
- /**
- * @return
- */
- private int size() {
- return this.sequences.size();
- }
-
- /**
- *
- */
-
- /*
- * (non-Javadoc)
- *
- * @see java.lang.Iterable#iterator()
- */
- @Override
- public Iterator> iterator() {
- return this.sequences.iterator();
- }
-
- /*
- * (non-Javadoc)
- *
- * @see java.lang.Object#toString()
- */
- @Override
- public String toString() {
- StringBuffer result = new StringBuffer();
- result.append(this.occurrenceCount);
- result.append(" occurrences:\n");
-
- for (List task : sequences) {
- result.append(task);
- result.append("\n");
- }
-
- return result.toString();
- }
-
- }
-
- // methods for internal testing
- // private void checkMatchingOfSessions(List> flattenedSessions,
- // List sessions,
- // String when)
- // {
- // List> currentFlattenedSessions = flattenSessions(sessions);
- // if (flattenedSessions.size() != currentFlattenedSessions.size()) {
- // System.out.println("################## number of sessions changed after "
- // + when);
- // }
- // else {
- // for (int i = 0; i < flattenedSessions.size(); i++) {
- // List expected = flattenedSessions.get(i);
- // List current = currentFlattenedSessions.get(i);
- //
- // if (expected.size() != current.size()) {
- // System.out.println
- // ("################## length of session " + i + " changed after " + when);
- // }
- // else {
- // for (int j = 0; j < expected.size(); j++) {
- // if (!expected.get(j).equals(current.get(j))) {
- // System.out.println("################## event " + j + " of session " +
- // i + " changed after " + when);
- // }
- // }
- // }
- // }
- // }
- // }
- //
- // private List> flattenSessions(List sessions) {
- // List> flattenedSessions = new ArrayList>();
- // for (IUserSession session : sessions) {
- // List flattenedUserSession = new ArrayList();
- // flatten(session, flattenedUserSession);
- // flattenedSessions.add(flattenedUserSession);
- // }
- //
- // return flattenedSessions;
- // }
- //
- // private void flatten(IUserSession iUserSession, List
- // flattenedUserSession) {
- // for (ITaskInstance instance : iUserSession) {
- // flatten(instance, flattenedUserSession);
- // }
- // }
- //
- // private void flatten(ITaskInstance instance, List
- // flattenedUserSession) {
- // if (instance instanceof ITaskInstanceList) {
- // for (ITaskInstance child : (ITaskInstanceList) instance) {
- // flatten(child, flattenedUserSession);
- // }
- // }
- // else if (instance instanceof ISelectionInstance) {
- // flatten(((ISelectionInstance) instance).getChild(),
- // flattenedUserSession);
- // }
- // else if (instance instanceof IOptionalInstance) {
- // flatten(((IOptionalInstance) instance).getChild(), flattenedUserSession);
- // }
- // else if (instance instanceof IEventTaskInstance) {
- // flattenedUserSession.add(((IEventTaskInstance) instance).getEvent());
- // }
- // }
+ private HashMap getNumber2Task() {
+ return number2task;
+ }
+
+ }
+
+
+
}