adds comments JAL-3383
[jalview.git] / src / jalview / datamodel / features / FeatureStore.java
index 686ac26..e60aec5 100644 (file)
@@ -23,7 +23,9 @@ package jalview.datamodel.features;
 import jalview.datamodel.SequenceFeature;
 
 import java.util.ArrayList;
+import java.util.Arrays;
 import java.util.Collections;
+import java.util.Comparator;
 import java.util.HashSet;
 import java.util.List;
 import java.util.Set;
@@ -90,9 +92,7 @@ public class FeatureStore
 
   float nonPositionalMaxScore;
 
-  private SequenceFeature[] temp = new SequenceFeature[3];
-
-  private boolean isTainted;
+  private ArrayList<SequenceFeature> featuresList;
 
   /**
    * Constructor
@@ -100,6 +100,7 @@ public class FeatureStore
   public FeatureStore()
   {
     features = new IntervalStore<>();
+    featuresList = new ArrayList<>();
     positionalFeatureGroups = new HashSet<>();
     nonPositionalFeatureGroups = new HashSet<>();
     positionalMinScore = Float.NaN;
@@ -118,6 +119,7 @@ public class FeatureStore
    * 
    * @param feature
    */
+
   public boolean addFeature(SequenceFeature feature)
   {
     if (contains(feature))
@@ -186,18 +188,17 @@ public class FeatureStore
   {
     if (feature.isNonPositional())
     {
-      return nonPositionalFeatures == null ? false : nonPositionalFeatures
-              .contains(feature);
+      return nonPositionalFeatures == null ? false
+              : nonPositionalFeatures.contains(feature);
     }
 
     if (feature.isContactFeature())
     {
-      return contactFeatureStarts == null ? false : listContains(
-              contactFeatureStarts, feature);
+      return contactFeatureStarts == null ? false
+              : listContains(contactFeatureStarts, feature);
     }
 
-    return features == null ? false : features
-            .contains(feature);
+    return features == null ? false : features.contains(feature);
   }
 
   /**
@@ -254,7 +255,7 @@ public class FeatureStore
       features = new IntervalStore<>();
     }
     features.add(feature);
-    isTainted = true;
+    featuresList.add(feature);
   }
 
   /**
@@ -285,7 +286,6 @@ public class FeatureStore
             f -> f.getBegin() >= feature.getBegin());
     contactFeatureStarts.add(insertPosition, feature);
 
-
     /*
      * insert into list sorted by end (second contact position):
      * binary search the sorted list to find the insertion point
@@ -349,6 +349,7 @@ public class FeatureStore
    *          end position of overlap range (inclusive)
    * @return
    */
+
   public List<SequenceFeature> findOverlappingFeatures(long start, long end)
   {
     List<SequenceFeature> result = new ArrayList<>();
@@ -408,8 +409,8 @@ public class FeatureStore
       SequenceFeature sf = contactFeatureEnds.get(index);
       if (!sf.isContactFeature())
       {
-        System.err.println("Error! non-contact feature type "
-                + sf.getType() + " in contact features list");
+        System.err.println("Error! non-contact feature type " + sf.getType()
+                + " in contact features list");
         index++;
         continue;
       }
@@ -488,6 +489,7 @@ public class FeatureStore
    * 
    * @return
    */
+
   public List<SequenceFeature> getPositionalFeatures()
   {
     List<SequenceFeature> result = new ArrayList<>();
@@ -517,6 +519,7 @@ public class FeatureStore
    * 
    * @return
    */
+
   public List<SequenceFeature> getContactFeatures()
   {
     if (contactFeatureStarts == null)
@@ -532,6 +535,7 @@ public class FeatureStore
    * 
    * @return
    */
+
   public List<SequenceFeature> getNonPositionalFeatures()
   {
     if (nonPositionalFeatures == null)
@@ -549,6 +553,7 @@ public class FeatureStore
    * 
    * @param sf
    */
+
   public synchronized boolean delete(SequenceFeature sf)
   {
     boolean removed = false;
@@ -583,6 +588,7 @@ public class FeatureStore
     if (!removed && features != null)
     {
       removed = features.remove(sf);
+      featuresList.remove(sf);
     }
 
     if (removed)
@@ -607,7 +613,6 @@ public class FeatureStore
     positionalMaxScore = Float.NaN;
     nonPositionalMinScore = Float.NaN;
     nonPositionalMaxScore = Float.NaN;
-    isTainted = true;
     /*
      * scan non-positional features for groups and scores
      */
@@ -677,13 +682,13 @@ public class FeatureStore
    * 
    * @return
    */
+
   public boolean isEmpty()
   {
     boolean hasFeatures = (contactFeatureStarts != null
-            && !contactFeatureStarts
-                    .isEmpty())
-            || (nonPositionalFeatures != null && !nonPositionalFeatures
-                    .isEmpty())
+            && !contactFeatureStarts.isEmpty())
+            || (nonPositionalFeatures != null
+                    && !nonPositionalFeatures.isEmpty())
             || (features != null && features.size() > 0);
 
     return !hasFeatures;
@@ -697,6 +702,7 @@ public class FeatureStore
    * @param positionalFeatures
    * @return
    */
+
   public Set<String> getFeatureGroups(boolean positionalFeatures)
   {
     if (positionalFeatures)
@@ -705,9 +711,9 @@ public class FeatureStore
     }
     else
     {
-      return nonPositionalFeatureGroups == null ? Collections
-              .<String> emptySet() : Collections
-              .unmodifiableSet(nonPositionalFeatureGroups);
+      return nonPositionalFeatureGroups == null
+              ? Collections.<String> emptySet()
+              : Collections.unmodifiableSet(nonPositionalFeatureGroups);
     }
   }
 
@@ -718,12 +724,13 @@ public class FeatureStore
    * @param positional
    * @return
    */
+
   public int getFeatureCount(boolean positional)
   {
     if (!positional)
     {
-      return nonPositionalFeatures == null ? 0 : nonPositionalFeatures
-              .size();
+      return nonPositionalFeatures == null ? 0
+              : nonPositionalFeatures.size();
     }
 
     int size = 0;
@@ -748,6 +755,7 @@ public class FeatureStore
    * 
    * @return
    */
+
   public int getTotalFeatureLength()
   {
     return totalExtent;
@@ -761,6 +769,7 @@ public class FeatureStore
    * @param positional
    * @return
    */
+
   public float getMinimumScore(boolean positional)
   {
     return positional ? positionalMinScore : nonPositionalMinScore;
@@ -774,6 +783,7 @@ public class FeatureStore
    * @param positional
    * @return
    */
+
   public float getMaximumScore(boolean positional)
   {
     return positional ? positionalMaxScore : nonPositionalMaxScore;
@@ -787,6 +797,7 @@ public class FeatureStore
    * @param group
    * @return
    */
+
   public List<SequenceFeature> getFeaturesForGroup(boolean positional,
           String group)
   {
@@ -807,8 +818,8 @@ public class FeatureStore
     for (SequenceFeature sf : sfs)
     {
       String featureGroup = sf.getFeatureGroup();
-      if (group == null && featureGroup == null || group != null
-              && group.equals(featureGroup))
+      if (group == null && featureGroup == null
+              || group != null && group.equals(featureGroup))
       {
         result.add(sf);
       }
@@ -825,6 +836,7 @@ public class FeatureStore
    * @param shiftBy
    * @return
    */
+
   public synchronized boolean shiftFeatures(int fromPosition, int shiftBy)
   {
     /*
@@ -857,39 +869,157 @@ public class FeatureStore
     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.
-   * Uses isTainted field to know when to reconstruct its temporary array.
    * 
    * @param pos
    * @return list of SequenceFeatures
    * @author Bob Hanson 2019.07.30
    */
-  public void findOverlappingFeatures(int pos, List<SequenceFeature> result)
+
+  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 (features != null)
+    if (featuresList != null)
     {
-      int n = features.size();
-      if (isTainted)
+      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)
       {
-        isTainted = false;
-        if (temp.length < n)
-        {
-          temp = new SequenceFeature[n << 1];
-        }
-        features.toArray(temp);
+        ithis.containedBy = getContainedBy(intervals[i - 1], ithis);
+      }
+      if (ithis.end > maxEnd)
+      {
+        maxEnd = ithis.end;
       }
-      findOverlaps(temp, n, pos, result);
     }
   }
 
   /**
+   * 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 &gt;= 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
@@ -936,27 +1066,97 @@ public class FeatureStore
   }
 
   /**
-   * Brute force point-interval overlap test
+   * Find all overlaps; special case when there is only one feature. The
+   * required array of start-sorted SequenceFeature is created lazily.
    * 
    * @param features
-   * @param n
    * @param pos
    * @param result
    */
-  private static void findOverlaps(SequenceFeature[] features, int n,
-          int pos,
+  private void findOverlaps(List<SequenceFeature> features, int pos,
           List<SequenceFeature> result)
   {
-    // BH I know, brute force. We need a single-position overlap
-    // method for IntervalStore, I think.
-    for (int i = n; --i >= 0;)
+    int n = featuresList.size();
+    if (n == 1)
     {
-      SequenceFeature f = features[i];
-      if (f.begin <= pos && f.end >= pos)
+      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(f);
+        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]);
   }
 
+
 }