*/
public class FeatureStore
{
- Comparator<ContiguousI> startOrdering = new RangeComparator(true);
+ /**
+ * a class providing criteria for performing a binary search of a list
+ */
+ abstract static class SearchCriterion
+ {
+ /**
+ * Answers true if the entry passes the search criterion test
+ *
+ * @param entry
+ * @return
+ */
+ abstract boolean compare(SequenceFeature entry);
- Comparator<ContiguousI> endOrdering = new RangeComparator(false);
+ static SearchCriterion byStart(final long target)
+ {
+ return new SearchCriterion() {
+
+ @Override
+ boolean compare(SequenceFeature entry)
+ {
+ return entry.getBegin() >= target;
+ }
+ };
+ }
+
+ static SearchCriterion byEnd(final long target)
+ {
+ return new SearchCriterion()
+ {
+
+ @Override
+ boolean compare(SequenceFeature entry)
+ {
+ return entry.getEnd() >= target;
+ }
+ };
+ }
+
+ static SearchCriterion byFeature(final ContiguousI to,
+ final Comparator<ContiguousI> rc)
+ {
+ return new SearchCriterion()
+ {
+
+ @Override
+ boolean compare(SequenceFeature entry)
+ {
+ return rc.compare(entry, to) >= 0;
+ }
+ };
+ }
+ }
/*
* Non-positional features have no (zero) start/end position.
*/
Set<String> nonPositionalFeatureGroups;
+ /*
+ * the total length of all positional features; contact features count 1 to
+ * the total and 1 to size(), consistent with an average 'feature length' of 1
+ */
+ int totalExtent;
+
+ float positionalMinScore;
+
+ float positionalMaxScore;
+
+ float nonPositionalMinScore;
+
+ float nonPositionalMaxScore;
+
/**
* Constructor
*/
{
nonNestedFeatures = new ArrayList<SequenceFeature>();
positionalFeatureGroups = new HashSet<String>();
+ nonPositionalFeatureGroups = new HashSet<String>();
+ positionalMinScore = Float.NaN;
+ positionalMaxScore = Float.NaN;
+ nonPositionalMinScore = Float.NaN;
+ nonPositionalMaxScore = Float.NaN;
// we only construct nonPositionalFeatures, contactFeatures
// or the NCList if we need to
*/
public boolean addFeature(SequenceFeature feature)
{
+ /*
+ * keep a record of feature groups
+ */
if (!feature.isNonPositional())
{
positionalFeatureGroups.add(feature.getFeatureGroup());
}
else
{
- if (!nonNestedFeatures.contains(feature))
+ if (!contains(nonNestedFeatures, feature))
{
added = addNonNestedFeature(feature);
if (!added)
}
}
+ if (added)
+ {
+ /*
+ * record the total extent of positional features, to make
+ * getTotalFeatureLength possible; we count the length of a
+ * contact feature as 1
+ */
+ totalExtent += getFeatureLength(feature);
+
+ /*
+ * record the minimum and maximum score for positional
+ * and non-positional features
+ */
+ float score = feature.getScore();
+ if (!Float.isNaN(score))
+ {
+ if (feature.isNonPositional())
+ {
+ nonPositionalMinScore = min(nonPositionalMinScore, score);
+ nonPositionalMaxScore = max(nonPositionalMaxScore, score);
+ }
+ else
+ {
+ positionalMinScore = min(positionalMinScore, score);
+ positionalMaxScore = max(positionalMaxScore, score);
+ }
+ }
+ }
+
return added;
}
/**
+ * Answers the 'length' of the feature, counting 0 for non-positional features
+ * and 1 for contact features
+ *
+ * @param feature
+ * @return
+ */
+ protected static int getFeatureLength(SequenceFeature feature)
+ {
+ if (feature.isNonPositional())
+ {
+ return 0;
+ }
+ if (feature.isContactFeature())
+ {
+ return 1;
+ }
+ return 1 + feature.getEnd() - feature.getBegin();
+ }
+
+ /**
* Adds the feature to the list of non-positional features (with lazy
* instantiation of the list if it is null), and returns true. If the
* non-positional features already include the new feature (by equality test),
if (nonPositionalFeatures == null)
{
nonPositionalFeatures = new ArrayList<SequenceFeature>();
- nonPositionalFeatureGroups = new HashSet<String>();
}
if (nonPositionalFeatures.contains(feature))
{
* contained by) an existing feature. If there is no nesting, the feature is
* added to the list and the method returns true. If nesting is found, the
* feature is not added and the method returns false.
- * <p>
- * Contact features are added at the position of their first contact point
*
* @param feature
* @return
{
synchronized (nonNestedFeatures)
{
- int insertPosition = binarySearchForAdd(nonNestedFeatures, feature);
+ /*
+ * find the first stored feature which doesn't precede the new one
+ */
+ int insertPosition = binarySearch(nonNestedFeatures,
+ SearchCriterion.byFeature(feature, RangeComparator.BY_START_POSITION));
/*
* fail if we detect feature enclosure - of the new feature by
}
/*
- * checks passed - add or append the feature
+ * checks passed - add the feature
*/
- if (insertPosition == nonNestedFeatures.size())
- {
- nonNestedFeatures.add(feature);
- }
- else
- {
- nonNestedFeatures.add(insertPosition, feature);
- }
+ nonNestedFeatures.add(insertPosition, feature);
+
return true;
}
}
}
/**
- * Answers the index of the first element in the given list which follows or
- * matches the given feature in the sort order. If no such element, answers
- * the length of the list.
- *
- * @param list
- * @param feature
- *
- * @return
- */
- protected int binarySearchForAdd(List<SequenceFeature> list, SequenceFeature feature)
- {
- // TODO binary search!
- int i = 0;
- while (i < list.size())
- {
- if (startOrdering.compare(nonNestedFeatures.get(i), feature) >= 0)
- {
- break;
- }
- i++;
- }
- return i;
- }
-
- /**
* Add a contact feature to the lists that hold them ordered by start (first
* contact) and by end (second contact) position, ensuring the lists remain
* ordered, and returns true. If the contact feature lists already contain the
contactFeatureEnds = new ArrayList<SequenceFeature>();
}
- // TODO binary search for insertion points!
- if (contactFeatureStarts.contains(feature))
+ if (contains(contactFeatureStarts, feature))
{
return false;
}
- contactFeatureStarts.add(feature);
- Collections.sort(contactFeatureStarts, startOrdering);
+ /*
+ * binary search the sorted list to find the insertion point
+ */
+ int insertPosition = binarySearch(contactFeatureStarts,
+ SearchCriterion.byFeature(feature,
+ RangeComparator.BY_START_POSITION));
+ contactFeatureStarts.add(insertPosition, feature);
+ // and resort to mak siccar...just in case insertion point not quite right
+ Collections.sort(contactFeatureStarts, RangeComparator.BY_START_POSITION);
+
+ insertPosition = binarySearch(contactFeatureStarts,
+ SearchCriterion.byFeature(feature,
+ RangeComparator.BY_END_POSITION));
contactFeatureEnds.add(feature);
- Collections.sort(contactFeatureEnds, endOrdering);
+ Collections.sort(contactFeatureEnds, RangeComparator.BY_END_POSITION);
return true;
}
/**
- * Returns a (possibly empty) list of entries whose range overlaps the given
+ * Answers true if the list contains the feature, else false. This method is
+ * optimised for the condition that the list is sorted on feature start
+ * position ascending, and will give unreliable results if this does not hold.
+ *
+ * @param features
+ * @param feature
+ * @return
+ */
+ protected static boolean contains(List<SequenceFeature> features,
+ SequenceFeature feature)
+ {
+ if (features == null || feature == null)
+ {
+ return false;
+ }
+
+ /*
+ * locate the first entry in the list which does not precede the feature
+ */
+ int pos = binarySearch(features,
+ SearchCriterion.byFeature(feature, RangeComparator.BY_START_POSITION));
+ int len = features.size();
+ while (pos < len)
+ {
+ SequenceFeature sf = features.get(pos);
+ if (sf.getBegin() > feature.getBegin())
+ {
+ return false; // no match found
+ }
+ if (sf.equals(feature))
+ {
+ return true;
+ }
+ pos++;
+ }
+ return false;
+ }
+
+ /**
+ * Returns a (possibly empty) list of features whose extent overlaps the given
* range. The returned list is not ordered. Contact features are included if
* either of the contact points lies within the range.
*
/**
* Adds contact features to the result list where either the second or the
- * first contact position lies within the target range.
+ * first contact position lies within the target range
*
* @param from
* @param to
}
/**
+ * Adds to the result list any contact features whose end (second contact
+ * point), but not start (first contact point), lies in the query from-to
+ * range
+ *
* @param from
* @param to
* @param result
protected void findContactEndFeatures(long from, long to,
List<SequenceFeature> result)
{
- // TODO binary search for startPosition
- for (int startPosition = 0; startPosition < contactFeatureEnds.size(); startPosition++)
+ /*
+ * find the first contact feature (if any) that does not lie
+ * entirely before the target range
+ */
+ int startPosition = binarySearch(contactFeatureEnds,
+ SearchCriterion.byEnd(from));
+ for (; startPosition < contactFeatureEnds.size(); startPosition++)
{
SequenceFeature sf = contactFeatureEnds.get(startPosition);
if (!sf.isContactFeature())
+ sf.getType() + " in contact features list");
continue;
}
+
int begin = sf.getBegin();
if (begin >= from && begin <= to)
{
*/
continue;
}
+
int end = sf.getEnd();
if (end >= from && end <= to)
{
result.add(sf);
}
- }
- }
-
- /**
- * Returns the index of the first contact feature found whose end (second
- * contact position) is not before the given start position. If no such
- * feature is found, returns the length of the contact features list.
- *
- * @param start
- * @return
- */
- protected int contactsBinarySearch(long start)
- {
- // TODO binary search!!
- int i = 0;
- while (i < contactFeatureEnds.size())
- {
- if (contactFeatureEnds.get(i).getEnd() >= start)
+ if (end > to)
{
break;
}
- i++;
}
-
- return i;
}
/**
- * Adds features to the result list that are at a single position which lies
- * within the target range. Non-positional features (start=end=0) and contact
+ * Adds non-nested features to the result list that lie within the target
+ * range. Non-positional features (start=end=0), contact features and nested
* features are excluded.
*
* @param from
protected void findNonNestedFeatures(long from, long to,
List<SequenceFeature> result)
{
- int startIndex = binarySearch(nonNestedFeatures, from);
+ int startIndex = binarySearch(nonNestedFeatures,
+ SearchCriterion.byEnd(from));
+
findNonNestedFeatures(startIndex, from, to, result);
}
* @return
*/
protected int findNonNestedFeatures(final int startIndex, long from,
- long to,
- List<SequenceFeature> result)
+ long to, List<SequenceFeature> result)
{
int i = startIndex;
while (i < nonNestedFeatures.size())
}
/**
- * Performs a binary search of the (sorted) list to find the index of the
- * first entry whose end position is not less than the target position (i.e.
- * skip all features that properly precede the given position)
- *
- * @param features
- * @param target
- * @return
- */
- protected int binarySearch(List<SequenceFeature> features, long target)
- {
- int width = features.size() / 2;
- int lastpos = width;
- while (width > 0)
- {
- int end = features.get(lastpos).getEnd();
- width = width / 2;
- if (end > target)
- {
- lastpos -= width;
- }
- else
- {
- lastpos += width;
- }
- }
- // todo correct binary search
- return lastpos > 1 ? lastpos - 2 : 0;
- // return lastpos;
- }
-
- /**
* Adds contact features whose start position lies in the from-to range to the
* result list
*
protected void findContactStartFeatures(long from, long to,
List<SequenceFeature> result)
{
- // TODO binary search for startPosition
- for (int startPosition = 0; startPosition < contactFeatureStarts.size(); startPosition++)
+ int startPosition = binarySearch(contactFeatureStarts,
+ SearchCriterion.byStart(from));
+
+ for (; startPosition < contactFeatureStarts.size(); startPosition++)
{
SequenceFeature sf = contactFeatureStarts.get(startPosition);
if (!sf.isContactFeature())
if (removed)
{
- rebuildFeatureGroups(sf.getFeatureGroup(), removedNonPositional);
+ rescanAfterDelete();
}
return removed;
}
/**
- * Check whether the given feature group is still represented, in either
- * positional or non-positional features, and if not, remove it from the set
- * of feature groups
+ * Rescan all features to recompute any cached values after an entry has been
+ * deleted. This is expected to be an infrequent event, so performance here is
+ * not critical.
+ */
+ protected synchronized void rescanAfterDelete()
+ {
+ positionalFeatureGroups.clear();
+ nonPositionalFeatureGroups.clear();
+ totalExtent = 0;
+ positionalMinScore = Float.NaN;
+ positionalMaxScore = Float.NaN;
+ nonPositionalMinScore = Float.NaN;
+ nonPositionalMaxScore = Float.NaN;
+
+ /*
+ * scan non-positional features for groups and scores
+ */
+ for (SequenceFeature sf : getNonPositionalFeatures())
+ {
+ nonPositionalFeatureGroups.add(sf.getFeatureGroup());
+ float score = sf.getScore();
+ nonPositionalMinScore = min(nonPositionalMinScore, score);
+ nonPositionalMaxScore = max(nonPositionalMaxScore, score);
+ }
+
+ /*
+ * scan positional features for groups, scores and extents
+ */
+ for (SequenceFeature sf : getPositionalFeatures())
+ {
+ positionalFeatureGroups.add(sf.getFeatureGroup());
+ float score = sf.getScore();
+ positionalMinScore = min(positionalMinScore, score);
+ positionalMaxScore = max(positionalMaxScore, score);
+ totalExtent += getFeatureLength(sf);
+ }
+ }
+
+ /**
+ * A helper method to return the minimum of two floats, where a non-NaN value
+ * is treated as 'less than' a NaN value (unlike Math.min which does the
+ * opposite)
*
- * @param featureGroup
- * @param nonPositional
+ * @param f1
+ * @param f2
*/
- protected void rebuildFeatureGroups(String featureGroup,
- boolean nonPositional)
+ protected static float min(float f1, float f2)
{
- if (nonPositional && nonPositionalFeatures != null)
+ if (Float.isNaN(f1))
{
- boolean found = false;
- for (SequenceFeature sf : nonPositionalFeatures)
- {
- String group = sf.getFeatureGroup();
- if (featureGroup == group
- || (featureGroup != null && featureGroup.equals(group)))
- {
- found = true;
- break;
- }
- }
- if (!found)
- {
- nonPositionalFeatureGroups.remove(featureGroup);
- }
+ return Float.isNaN(f2) ? f1 : f2;
+ }
+ else
+ {
+ return Float.isNaN(f2) ? f1 : Math.min(f1, f2);
+ }
+ }
+
+ /**
+ * A helper method to return the maximum of two floats, where a non-NaN value
+ * is treated as 'greater than' a NaN value (unlike Math.max which does the
+ * opposite)
+ *
+ * @param f1
+ * @param f2
+ */
+ protected static float max(float f1, float f2)
+ {
+ if (Float.isNaN(f1))
+ {
+ return Float.isNaN(f2) ? f1 : f2;
}
- else if (!findFeatureGroup(featureGroup))
+ else
{
- positionalFeatureGroups.remove(featureGroup);
+ return Float.isNaN(f2) ? f1 : Math.max(f1, f2);
}
}
.unmodifiableSet(nonPositionalFeatureGroups);
}
}
+
+ /**
+ * Performs a binary search of the (sorted) list to find the index of the
+ * first entry which returns true for the given comparator function. Returns
+ * the length of the list if there is no such entry.
+ *
+ * @param features
+ * @param sc
+ * @return
+ */
+ protected static int binarySearch(List<SequenceFeature> features,
+ SearchCriterion sc)
+ {
+ int start = 0;
+ int end = features.size() - 1;
+ int matched = features.size();
+
+ while (start <= end)
+ {
+ int mid = (start + end) / 2;
+ SequenceFeature entry = features.get(mid);
+ boolean compare = sc.compare(entry);
+ if (compare)
+ {
+ matched = mid;
+ end = mid - 1;
+ }
+ else
+ {
+ start = mid + 1;
+ }
+ }
+
+ return matched;
+ }
+
+ /**
+ * Answers the number of positional (or non-positional) features stored.
+ * Contact features count as 1.
+ *
+ * @param positional
+ * @return
+ */
+ public int getFeatureCount(boolean positional)
+ {
+ if (!positional)
+ {
+ return nonPositionalFeatures == null ? 0 : nonPositionalFeatures
+ .size();
+ }
+
+ int size = nonNestedFeatures.size();
+
+ if (contactFeatureStarts != null)
+ {
+ // note a contact feature (start/end) counts as one
+ size += contactFeatureStarts.size();
+ }
+
+ if (nestedFeatures != null)
+ {
+ size += nestedFeatures.size();
+ }
+
+ return size;
+ }
+
+ /**
+ * Answers the total length of positional features (or zero if there are
+ * none). Contact features contribute a value of 1 to the total.
+ *
+ * @return
+ */
+ public int getTotalFeatureLength()
+ {
+ return totalExtent;
+ }
+
+ /**
+ * Answers the minimum score held for positional or non-positional features.
+ * This may be Float.NaN if there are no features, are none has a non-NaN
+ * score.
+ *
+ * @param positional
+ * @return
+ */
+ public float getMinimumScore(boolean positional)
+ {
+ return positional ? positionalMinScore : nonPositionalMinScore;
+ }
+
+ /**
+ * Answers the maximum score held for positional or non-positional features.
+ * This may be Float.NaN if there are no features, are none has a non-NaN
+ * score.
+ *
+ * @param positional
+ * @return
+ */
+ public float getMaximumScore(boolean positional)
+ {
+ return positional ? positionalMaxScore : nonPositionalMaxScore;
+ }
+
+ /**
+ * Answers a list of all either positional or non-positional features whose
+ * feature group matches the given group (which may be null)
+ *
+ * @param positional
+ * @param group
+ * @return
+ */
+ public List<SequenceFeature> getFeaturesForGroup(boolean positional,
+ String group)
+ {
+ List<SequenceFeature> result = new ArrayList<SequenceFeature>();
+
+ /*
+ * if we know features don't include the target group, no need
+ * to inspect them for matches
+ */
+ if (positional && !positionalFeatureGroups.contains(group)
+ || !positional && !nonPositionalFeatureGroups.contains(group))
+ {
+ return result;
+ }
+
+ List<SequenceFeature> sfs = positional ? getPositionalFeatures()
+ : getNonPositionalFeatures();
+ for (SequenceFeature sf : sfs)
+ {
+ String featureGroup = sf.getFeatureGroup();
+ if (group == null && featureGroup == null || group != null
+ && group.equals(featureGroup))
+ {
+ result.add(sf);
+ }
+ }
+ return result;
+ }
}