+/*
+ * Jalview - A Sequence Alignment Editor and Viewer ($$Version-Rel$$)
+ * Copyright (C) $$Year-Rel$$ The Jalview Authors
+ *
+ * This file is part of Jalview.
+ *
+ * Jalview is free software: you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation, either version 3
+ * of the License, or (at your option) any later version.
+ *
+ * Jalview is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR
+ * PURPOSE. See the GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with Jalview. If not, see <http://www.gnu.org/licenses/>.
+ * The Jalview Authors are detailed in the 'AUTHORS' file.
+ */
package jalview.datamodel.features;
-import jalview.datamodel.ContiguousI;
import jalview.datamodel.SequenceFeature;
import java.util.ArrayList;
import java.util.Collections;
-import java.util.Comparator;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
+import intervalstore.api.IntervalStoreI;
+import intervalstore.impl.BinarySearcher;
+import intervalstore.impl.IntervalStore;
+
/**
* A data store for a set of sequence features that supports efficient lookup of
* features overlapping a given range. Intended for (but not limited to) storage
*/
public class FeatureStore
{
- /**
- * 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);
-
- /**
- * serves a search condition for finding the first feature whose start
- * position follows a given target location
- *
- * @param target
- * @return
- */
- static SearchCriterion byStart(final long target)
- {
- return new SearchCriterion() {
-
- @Override
- boolean compare(SequenceFeature entry)
- {
- return entry.getBegin() >= target;
- }
- };
- }
-
- /**
- * serves a search condition for finding the first feature whose end
- * position is at or follows a given target location
- *
- * @param target
- * @return
- */
- static SearchCriterion byEnd(final long target)
- {
- return new SearchCriterion()
- {
-
- @Override
- boolean compare(SequenceFeature entry)
- {
- return entry.getEnd() >= target;
- }
- };
- }
-
- /**
- * serves a search condition for finding the first feature which follows the
- * given range as determined by a supplied comparator
- *
- * @param target
- * @return
- */
- 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.
* Kept as a separate list in case this criterion changes in future.
List<SequenceFeature> nonPositionalFeatures;
/*
- * An ordered list of features, with the promise that no feature in the list
- * properly contains any other. This constraint allows bounded linear search
- * of the list for features overlapping a region.
- * Contact features are not included in this list.
- */
- List<SequenceFeature> nonNestedFeatures;
-
- /*
* contact features ordered by first contact position
*/
List<SequenceFeature> contactFeatureStarts;
List<SequenceFeature> contactFeatureEnds;
/*
- * Nested Containment List is used to hold any features that are nested
- * within (properly contained by) any other feature. This is a recursive tree
- * which supports depth-first scan for features overlapping a range.
- * It is used here as a 'catch-all' fallback for features that cannot be put
- * into a simple ordered list without invalidating the search methods.
+ * IntervalStore holds remaining features and provides efficient
+ * query for features overlapping any given interval
*/
- NCList<SequenceFeature> nestedFeatures;
+ IntervalStoreI<SequenceFeature> features;
/*
* Feature groups represented in stored positional features
float nonPositionalMaxScore;
+ private SequenceFeature[] temp = new SequenceFeature[3];
+
+ private boolean isTainted;
+
/**
* Constructor
*/
public FeatureStore()
{
- nonNestedFeatures = new ArrayList<SequenceFeature>();
- positionalFeatureGroups = new HashSet<String>();
- nonPositionalFeatureGroups = new HashSet<String>();
+ features = new IntervalStore<>();
+ positionalFeatureGroups = new HashSet<>();
+ nonPositionalFeatureGroups = new HashSet<>();
positionalMinScore = Float.NaN;
positionalMaxScore = Float.NaN;
nonPositionalMinScore = Float.NaN;
nonPositionalMaxScore = Float.NaN;
- // we only construct nonPositionalFeatures, contactFeatures
- // or the NCList if we need to
+ // we only construct nonPositionalFeatures, contactFeatures if we need to
}
/**
positionalFeatureGroups.add(feature.getFeatureGroup());
}
- boolean added = false;
-
if (feature.isContactFeature())
{
- added = addContactFeature(feature);
+ addContactFeature(feature);
}
else if (feature.isNonPositional())
{
- added = addNonPositionalFeature(feature);
+ addNonPositionalFeature(feature);
}
else
{
- added = addNonNestedFeature(feature);
- if (!added)
- {
- /*
- * detected a nested feature - put it in the NCList structure
- */
- added = addNestedFeature(feature);
- }
+ addNestedFeature(feature);
}
- 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 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))
+ /*
+ * record the minimum and maximum score for positional
+ * and non-positional features
+ */
+ float score = feature.getScore();
+ if (!Float.isNaN(score))
+ {
+ if (feature.isNonPositional())
{
- if (feature.isNonPositional())
- {
- nonPositionalMinScore = min(nonPositionalMinScore, score);
- nonPositionalMaxScore = max(nonPositionalMaxScore, score);
- }
- else
- {
- positionalMinScore = min(positionalMinScore, score);
- positionalMaxScore = max(positionalMaxScore, score);
- }
+ nonPositionalMinScore = min(nonPositionalMinScore, score);
+ nonPositionalMaxScore = max(nonPositionalMaxScore, score);
+ }
+ else
+ {
+ positionalMinScore = min(positionalMinScore, score);
+ positionalMaxScore = max(positionalMaxScore, score);
}
}
- return added;
+ return true;
}
/**
contactFeatureStarts, feature);
}
- if (listContains(nonNestedFeatures, feature))
- {
- return true;
- }
-
- return nestedFeatures == null ? false : nestedFeatures
+ return features == null ? false : features
.contains(feature);
}
{
if (nonPositionalFeatures == null)
{
- nonPositionalFeatures = new ArrayList<SequenceFeature>();
+ nonPositionalFeatures = new ArrayList<>();
}
nonPositionalFeatures.add(feature);
}
/**
- * Adds one feature to the NCList that can manage nested features (creating
- * the NCList if necessary), and returns true. If the feature is already
- * stored in the NCList (by equality test), then it is not added, and this
- * method returns false.
+ * Adds one feature to the IntervalStore that can manage nested features
+ * (creating the IntervalStore if necessary)
*/
- protected synchronized boolean addNestedFeature(SequenceFeature feature)
+ protected synchronized void addNestedFeature(SequenceFeature feature)
{
- if (nestedFeatures == null)
+ if (features == null)
{
- nestedFeatures = new NCList<>(feature);
- return true;
+ features = new IntervalStore<>();
}
- return nestedFeatures.add(feature, false);
- }
-
- /**
- * Add a feature to the list of non-nested features, maintaining the ordering
- * of the list. A check is made for whether the feature is nested in (properly
- * 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.
- *
- * @param feature
- * @return
- */
- protected boolean addNonNestedFeature(SequenceFeature feature)
- {
- synchronized (nonNestedFeatures)
- {
- /*
- * 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
- * the one preceding it, or of the next feature by the new one
- */
- if (insertPosition > 0)
- {
- if (encloses(nonNestedFeatures.get(insertPosition - 1), feature))
- {
- return false;
- }
- }
- if (insertPosition < nonNestedFeatures.size())
- {
- if (encloses(feature, nonNestedFeatures.get(insertPosition)))
- {
- return false;
- }
- }
-
- /*
- * checks passed - add the feature
- */
- nonNestedFeatures.add(insertPosition, feature);
-
- return true;
- }
- }
-
- /**
- * Answers true if range1 properly encloses range2, else false
- *
- * @param range1
- * @param range2
- * @return
- */
- protected boolean encloses(ContiguousI range1, ContiguousI range2)
- {
- int begin1 = range1.getBegin();
- int begin2 = range2.getBegin();
- int end1 = range1.getEnd();
- int end2 = range2.getEnd();
- if (begin1 == begin2 && end1 > end2)
- {
- return true;
- }
- if (begin1 < begin2 && end1 >= end2)
- {
- return true;
- }
- return false;
+ features.add(feature);
+ isTainted = true;
}
/**
{
if (contactFeatureStarts == null)
{
- contactFeatureStarts = new ArrayList<SequenceFeature>();
+ contactFeatureStarts = new ArrayList<>();
}
if (contactFeatureEnds == null)
{
- contactFeatureEnds = new ArrayList<SequenceFeature>();
+ contactFeatureEnds = new ArrayList<>();
}
/*
+ * insert into list sorted by start (first contact position):
* binary search the sorted list to find the insertion point
*/
- int insertPosition = binarySearch(contactFeatureStarts,
- SearchCriterion.byFeature(feature,
- RangeComparator.BY_START_POSITION));
+ int insertPosition = BinarySearcher.findFirst(contactFeatureStarts,
+ f -> f.getBegin() >= feature.getBegin());
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, RangeComparator.BY_END_POSITION);
+
+ /*
+ * insert into list sorted by end (second contact position):
+ * binary search the sorted list to find the insertion point
+ */
+ insertPosition = BinarySearcher.findFirst(contactFeatureEnds,
+ f -> f.getEnd() >= feature.getEnd());
+ contactFeatureEnds.add(insertPosition, feature);
return true;
}
/*
* 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 pos = binarySearch(features,
+ // SearchCriterion.byFeature(feature, RangeComparator.BY_START_POSITION));
+ int pos = BinarySearcher.findFirst(features,
+ val -> val.getBegin() >= feature.getBegin());
int len = features.size();
while (pos < len)
{
{
List<SequenceFeature> result = new ArrayList<>();
- findNonNestedFeatures(start, end, result);
-
findContactFeatures(start, end, result);
- if (nestedFeatures != null)
+ if (features != null)
{
- result.addAll(nestedFeatures.findOverlaps(start, end));
+ result.addAll(features.findOverlaps(start, end));
}
return result;
{
if (contactFeatureStarts != null)
{
- findContactStartFeatures(from, to, result);
+ findContactStartOverlaps(from, to, result);
}
if (contactFeatureEnds != null)
{
- findContactEndFeatures(from, to, result);
+ findContactEndOverlaps(from, to, result);
}
}
* @param to
* @param result
*/
- protected void findContactEndFeatures(long from, long to,
+ protected void findContactEndOverlaps(long from, long to,
List<SequenceFeature> result)
{
/*
- * find the first contact feature (if any) that does not lie
- * entirely before the target range
+ * find the first contact feature (if any)
+ * whose end point is not before the target range
*/
- int startPosition = binarySearch(contactFeatureEnds,
- SearchCriterion.byEnd(from));
- for (; startPosition < contactFeatureEnds.size(); startPosition++)
+ int index = BinarySearcher.findFirst(contactFeatureEnds,
+ f -> f.getEnd() >= from);
+
+ while (index < contactFeatureEnds.size())
{
- SequenceFeature sf = contactFeatureEnds.get(startPosition);
+ SequenceFeature sf = contactFeatureEnds.get(index);
if (!sf.isContactFeature())
{
System.err.println("Error! non-contact feature type "
+ sf.getType() + " in contact features list");
+ index++;
continue;
}
* this feature's first contact position lies in the search range
* so we don't include it in results a second time
*/
+ index++;
continue;
}
- int end = sf.getEnd();
- if (end >= from && end <= to)
- {
- result.add(sf);
- }
- if (end > to)
+ if (sf.getEnd() > to)
{
+ /*
+ * this feature (and all following) has end point after the target range
+ */
break;
}
- }
- }
- /**
- * 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
- * @param to
- * @param result
- */
- protected void findNonNestedFeatures(long from, long to,
- List<SequenceFeature> result)
- {
- /*
- * find the first feature whose end position is
- * after the target range start
- */
- int startIndex = binarySearch(nonNestedFeatures,
- SearchCriterion.byEnd(from));
-
- final int startIndex1 = startIndex;
- int i = startIndex1;
- while (i < nonNestedFeatures.size())
- {
- SequenceFeature sf = nonNestedFeatures.get(i);
- if (sf.getBegin() > to)
- {
- break;
- }
- if (sf.getBegin() <= to && sf.getEnd() >= from)
- {
- result.add(sf);
- }
- i++;
+ /*
+ * feature has end >= from and end <= to
+ * i.e. contact end point lies within overlap search range
+ */
+ result.add(sf);
+ index++;
}
}
* @param to
* @param result
*/
- protected void findContactStartFeatures(long from, long to,
+ protected void findContactStartOverlaps(long from, long to,
List<SequenceFeature> result)
{
- int startPosition = binarySearch(contactFeatureStarts,
- SearchCriterion.byStart(from));
+ int index = BinarySearcher.findFirst(contactFeatureStarts,
+ f -> f.getBegin() >= from);
- for (; startPosition < contactFeatureStarts.size(); startPosition++)
+ while (index < contactFeatureStarts.size())
{
- SequenceFeature sf = contactFeatureStarts.get(startPosition);
+ SequenceFeature sf = contactFeatureStarts.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 " + sf.toString()
+ + " in contact features list");
+ index++;
continue;
}
- int begin = sf.getBegin();
- if (begin >= from && begin <= to)
+ if (sf.getBegin() > to)
{
- result.add(sf);
+ /*
+ * this feature's start (and all following) follows the target range
+ */
+ break;
}
+
+ /*
+ * feature has begin >= from and begin <= to
+ * i.e. contact start point lies within overlap search range
+ */
+ result.add(sf);
+ index++;
}
}
*/
public List<SequenceFeature> getPositionalFeatures()
{
- /*
- * add non-nested features (may be all features for many cases)
- */
List<SequenceFeature> result = new ArrayList<>();
- result.addAll(nonNestedFeatures);
/*
* add any contact features - from the list by start position
/*
* add any nested features
*/
- if (nestedFeatures != null)
+ if (features != null)
{
- result.addAll(nestedFeatures.getEntries());
+ result.addAll(features);
}
return result;
*/
public synchronized boolean delete(SequenceFeature sf)
{
- /*
- * try the non-nested positional features first
- */
- boolean removed = nonNestedFeatures.remove(sf);
+ boolean removed = false;
/*
- * if not found, try contact positions (and if found, delete
+ * try contact positions (and if found, delete
* from both lists of contact positions)
*/
if (!removed && contactFeatureStarts != null)
/*
* if not found, try nested features
*/
- if (!removed && nestedFeatures != null)
+ if (!removed && features != null)
{
- removed = nestedFeatures.delete(sf);
+ removed = features.remove(sf);
}
if (removed)
positionalMaxScore = Float.NaN;
nonPositionalMinScore = Float.NaN;
nonPositionalMaxScore = Float.NaN;
-
+ isTainted = true;
/*
* scan non-positional features for groups and scores
*/
*/
public boolean isEmpty()
{
- boolean hasFeatures = !nonNestedFeatures.isEmpty()
- || (contactFeatureStarts != null && !contactFeatureStarts
+ boolean hasFeatures = (contactFeatureStarts != null
+ && !contactFeatureStarts
.isEmpty())
|| (nonPositionalFeatures != null && !nonPositionalFeatures
.isEmpty())
- || (nestedFeatures != null && nestedFeatures.size() > 0);
+ || (features != null && features.size() > 0);
return !hasFeatures;
}
}
/**
- * 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.
*
.size();
}
- int size = nonNestedFeatures.size();
+ int size = 0;
if (contactFeatureStarts != null)
{
size += contactFeatureStarts.size();
}
- if (nestedFeatures != null)
+ if (features != null)
{
- size += nestedFeatures.size();
+ size += features.size();
}
return size;
}
/**
- * Adds the shift value to the start and end of all positional features.
- * Returns true if at least one feature was updated, else false.
+ * 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 shift
+ * @param fromPosition
+ * @param shiftBy
* @return
*/
- public synchronized boolean shiftFeatures(int shift)
+ public synchronized boolean shiftFeatures(int fromPosition, int shiftBy)
{
/*
* Because begin and end are final fields (to ensure the data store's
boolean modified = false;
for (SequenceFeature sf : getPositionalFeatures())
{
- modified = true;
- int newBegin = sf.getBegin() + shift;
- int newEnd = sf.getEnd() + shift;
-
- /*
- * sanity check: don't shift left of the first residue
- */
- if (newEnd > 0)
+ if (sf.getBegin() >= fromPosition)
{
- newBegin = Math.max(1, newBegin);
- SequenceFeature sf2 = new SequenceFeature(sf, newBegin, newEnd,
- sf.getFeatureGroup(), sf.getScore());
- addFeature(sf2);
+ 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);
}
- delete(sf);
}
return modified;
}
+
+ /**
+ * 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)
+ {
+
+ if (contactFeatureStarts != null)
+ {
+ findContacts(contactFeatureStarts, pos, result, true);
+ findContacts(contactFeatureEnds, pos, result, false);
+ }
+ if (features != null)
+ {
+ int n = features.size();
+ if (isTainted)
+ {
+ isTainted = false;
+ if (temp.length < n)
+ {
+ temp = new SequenceFeature[n << 1];
+ }
+ features.toArray(temp);
+ }
+ findOverlaps(temp, n, pos, result);
+ }
+ }
+
+ /**
+ * 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;
+ }
+ }
+ }
+
+ /**
+ * Brute force point-interval overlap test
+ *
+ * @param features
+ * @param n
+ * @param pos
+ * @param result
+ */
+ private static void findOverlaps(SequenceFeature[] features, int n,
+ 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;)
+ {
+ SequenceFeature f = features[i];
+ if (f.begin <= pos && f.end >= pos)
+ {
+ result.add(f);
+ }
+ }
+ }
+
}