+
+ /**
+ * 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 f1
+ * @param f2
+ */
+ protected static float min(float f1, float f2)
+ {
+ if (Float.isNaN(f1))
+ {
+ 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
+ {
+ return Float.isNaN(f2) ? f1 : Math.max(f1, f2);
+ }
+ }
+
+ /**
+ * Answers true if this store has no features, else false
+ *
+ * @return
+ */
+
+ public boolean isEmpty()
+ {
+ boolean hasFeatures = (contactFeatureStarts != null
+ && !contactFeatureStarts.isEmpty())
+ || (nonPositionalFeatures != null
+ && !nonPositionalFeatures.isEmpty())
+ || (features != null && features.size() > 0);
+
+ return !hasFeatures;
+ }
+
+ /**
+ * Answers the set of distinct feature groups stored, possibly including null,
+ * as an unmodifiable view of the set. The parameter determines whether the
+ * groups for positional or for non-positional features are returned.
+ *
+ * @param positionalFeatures
+ * @return
+ */
+
+ public Set<String> getFeatureGroups(boolean positionalFeatures)
+ {
+ if (positionalFeatures)
+ {
+ return Collections.unmodifiableSet(positionalFeatureGroups);
+ }
+ else
+ {
+ return nonPositionalFeatureGroups == null
+ ? Collections.<String> emptySet()
+ : Collections.unmodifiableSet(nonPositionalFeatureGroups);
+ }
+ }
+
+ /**
+ * 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 = 0;
+
+ if (contactFeatureStarts != null)
+ {
+ // note a contact feature (start/end) counts as one
+ size += contactFeatureStarts.size();
+ }
+
+ if (features != null)
+ {
+ size += features.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<>();
+
+ /*
+ * 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;
+ }
+
+ /**
+ * Adds the shift amount to the start and end of all positional features whose
+ * start position is at or after fromPosition. Returns true if at least one
+ * feature was shifted, else false.
+ *
+ * @param fromPosition
+ * @param shiftBy
+ * @return
+ */
+
+ public synchronized boolean shiftFeatures(int fromPosition, int shiftBy)
+ {
+ /*
+ * Because begin and end are final fields (to ensure the data store's
+ * integrity), we have to delete each feature and re-add it as amended.
+ * (Although a simple shift of all values would preserve data integrity!)
+ */
+ boolean modified = false;
+ for (SequenceFeature sf : getPositionalFeatures())
+ {
+ if (sf.getBegin() >= fromPosition)
+ {
+ modified = true;
+ int newBegin = sf.getBegin() + shiftBy;
+ int newEnd = sf.getEnd() + shiftBy;
+
+ /*
+ * sanity check: don't shift left of the first residue
+ */
+ if (newEnd > 0)
+ {
+ newBegin = Math.max(1, newBegin);
+ SequenceFeature sf2 = new SequenceFeature(sf, newBegin, newEnd,
+ sf.getFeatureGroup(), sf.getScore());
+ addFeature(sf2);
+ }
+ delete(sf);
+ }
+ }
+ return modified;
+ }
+
+ /////////////////////// added by Bob Hanson ///////////////////////
+
+ // The following methods use a linked list of containment in features
+ // rather than IntervalStore. Implemented only for OverviewPanel, because
+ // only that makes calls for start == end in feature overlap requests.
+ //
+ //
+ // There are two parts --- initialization, and overlap searching.
+ //
+ // Initialization involves two steps:
+ //
+ // (1) sorting of features by start position using a standard Array.sort with
+ // Comparator.
+ // (2) linking of features, effectively nesting them.
+ //
+ // Searching also involves two steps:
+ //
+ // (1) binary search for a position within the sorted features array.
+ // (2) traversing the linked lists with an end check to read out the
+ // overlapped features at this position.
+ //
+ // All of this is done with very simple standard methods.
+
+ // single public method:
+
+ /**
+ * Find all features containing this position.
+ *
+ * @param pos
+ * @return list of SequenceFeatures
+ * @author Bob Hanson 2019.07.30
+ */
+
+ public List<SequenceFeature> findOverlappingFeatures(int pos,
+ List<SequenceFeature> result)
+ {
+ if (result == null)
+ {
+ result = new ArrayList<>();
+ }
+
+ if (contactFeatureStarts != null)
+ {
+ findContacts(contactFeatureStarts, pos, result, true);
+ findContacts(contactFeatureEnds, pos, result, false);
+ }
+ if (featuresList != null)
+ {
+ findOverlaps(featuresList, pos, result);
+ }
+ return result;
+ }
+
+ // Initialization
+
+ /*
+ * contact features ordered by first contact position
+ */
+ private SequenceFeature[] orderedFeatureStarts;
+
+ private void rebuildArrays(int n)
+ {
+ if (startComp == null)
+ {
+ startComp = new StartComparator();
+ }
+ orderedFeatureStarts = new SequenceFeature[n];
+ for (int i = n; --i >= 0;)
+ {
+ SequenceFeature sf = featuresList.get(i);
+ sf.index = i; // for debugging only
+ orderedFeatureStarts[i] = sf;
+ }
+ Arrays.sort(orderedFeatureStarts, startComp);
+ linkFeatures(orderedFeatureStarts);
+ }
+
+ /**
+ * just a standard Comparator
+ */
+ private static StartComparator startComp;
+
+ class StartComparator implements Comparator<SequenceFeature>
+ {
+
+ @Override
+ public int compare(SequenceFeature o1, SequenceFeature o2)
+ {
+ int p1 = o1.begin;
+ int p2 = o2.begin;
+ return (p1 < p2 ? -1 : p1 > p2 ? 1 : 0);
+ }
+
+ }
+
+ /**
+ *
+ * @param intervals
+ */
+ private void linkFeatures(SequenceFeature[] intervals)
+ {
+ if (intervals.length < 2)
+ {
+ return;
+ }
+ int maxEnd = intervals[0].end;
+ for (int i = 1, n = intervals.length; i < n; i++)
+ {
+ SequenceFeature ithis = intervals[i];
+ if (ithis.begin <= maxEnd)
+ {
+ ithis.containedBy = getContainedBy(intervals[i - 1], ithis);
+ }
+ if (ithis.end > maxEnd)
+ {
+ maxEnd = ithis.end;
+ }
+ }
+ }
+
+ /**
+ * Since we are traversing the sorted feature array, all elements prior to the
+ * one we are working on have been fully linked. All we are doing is following
+ * those links until we find the first array feature with a containedBy
+ * element that has an end >= our begin point. It is generally a very short
+ * list -- maybe one or two depths. But it might be more than that.
+ *
+ * @param sf
+ * @param sf0
+ * @return
+ */
+ private SequenceFeature getContainedBy(SequenceFeature sf,
+ SequenceFeature sf0)
+ {
+ int begin = sf0.begin;
+ while (sf != null)
+ {
+ if (begin <= sf.end)
+ {
+ System.out.println("\nFS found " + sf0.index + ":" + sf0
+ + "\nFS in " + sf.index + ":" + sf);
+ return sf;
+ }
+ sf = sf.containedBy;
+ }
+ return null;
+ }
+
+ // Searching for overlapping features at a given position:
+
+ /**
+ * Binary search for contact start or end at a given (Overview) position.
+ *
+ * @param l
+ * @param pos
+ * @param result
+ * @param isStart
+ *
+ * @author Bob Hanson 2019.07.30
+ */
+ private static void findContacts(List<SequenceFeature> l, int pos,
+ List<SequenceFeature> result, boolean isStart)
+ {
+ int low = 0;
+ int high = l.size() - 1;
+ while (low <= high)
+ {
+ int mid = (low + high) >>> 1;
+ SequenceFeature f = l.get(mid);
+ switch (Long.signum((isStart ? f.begin : f.end) - pos))
+ {
+ case -1:
+ low = mid + 1;
+ continue;
+ case 1:
+ high = mid - 1;
+ continue;
+ case 0:
+ int m = mid;
+ result.add(f);
+ // could be "5" in 12345556788 ?
+ while (++mid <= high && (f = l.get(mid)) != null
+ && (isStart ? f.begin : f.end) == pos)
+ {
+ result.add(f);
+ }
+ while (--m >= low && (f = l.get(m)) != null
+ && (isStart ? f.begin : f.end) == pos)
+ {
+ result.add(f);
+ }
+ return;
+ }
+ }
+ }
+
+ /**
+ * Find all overlaps; special case when there is only one feature. The
+ * required array of start-sorted SequenceFeature is created lazily.
+ *
+ * @param features
+ * @param pos
+ * @param result
+ */
+ private void findOverlaps(List<SequenceFeature> features, int pos,
+ List<SequenceFeature> result)
+ {
+ int n = featuresList.size();
+ if (n == 1)
+ {
+ checkOne(featuresList.get(0), pos, result);
+ return;
+ }
+ if (orderedFeatureStarts == null)
+ {
+ rebuildArrays(n);
+ }
+
+ // (1) Find the closest feature to this position.
+
+ SequenceFeature sf = findClosestFeature(orderedFeatureStarts, pos);
+
+ // (2) Traverse the containedBy field, checking for overlap.
+
+ while (sf != null)
+ {
+ if (sf.end >= pos)
+ {
+ result.add(sf);
+ }
+ sf = sf.containedBy;
+ }
+ }
+
+ /**
+ * Quick check when we only have one feature.
+ *
+ * @param sf
+ * @param pos
+ * @param result
+ */
+ private void checkOne(SequenceFeature sf, int pos,
+ List<SequenceFeature> result)
+ {
+ if (sf.begin <= pos && sf.end >= pos)
+ {
+ result.add(sf);
+ }
+ return;
+ }
+
+ /**
+ * A binary search identical to the one used for contact start/end, but here
+ * we return the feature itself.
+ *
+ * @param l
+ * @param pos
+ * @return
+ */
+ private SequenceFeature findClosestFeature(SequenceFeature[] l, int pos)
+ {
+ int low = 0;
+ int high = l.length - 1;
+ while (low <= high)
+ {
+ int mid = (low + high) >>> 1;
+ SequenceFeature f = l[mid];
+ switch (Long.signum(f.begin - pos))
+ {
+ case -1:
+ low = mid + 1;
+ continue;
+ case 1:
+ high = mid - 1;
+ continue;
+ case 0:
+
+ while (++mid <= high && l[mid].begin == pos)
+ {
+ ;
+ }
+ mid--;
+ return l[mid];
+ }
+ }
+ // -1 here?
+ return (high < 0 || low >= l.length ? null : l[high]);
+ }
+
+