2 * Jalview - A Sequence Alignment Editor and Viewer ($$Version-Rel$$)
3 * Copyright (C) $$Year-Rel$$ The Jalview Authors
5 * This file is part of Jalview.
7 * Jalview is free software: you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation, either version 3
10 * of the License, or (at your option) any later version.
12 * Jalview is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty
14 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR
15 * PURPOSE. See the GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with Jalview. If not, see <http://www.gnu.org/licenses/>.
19 * The Jalview Authors are detailed in the 'AUTHORS' file.
21 package jalview.datamodel.features;
23 import jalview.datamodel.SequenceFeature;
25 import java.util.ArrayList;
26 import java.util.Collection;
27 import java.util.Collections;
28 import java.util.HashSet;
29 import java.util.List;
32 import intervalstore.api.IntervalStoreI;
33 import intervalstore.impl.BinarySearcher;
34 import intervalstore.impl.BinarySearcher.Compare;
36 public class FeatureStore
38 public enum IntervalStoreType
41 * original NCList-based IntervalStore
43 INTERVAL_STORE_NCLIST,
46 * linked-list IntervalStore
48 INTERVAL_STORE_LINKED_LIST,
51 * NCList as array buffer IntervalStore
53 INTERVAL_STORE_NCARRAY
57 * track largest start for quick insertion of ordered features
59 protected int maxStart = -1;
61 protected int maxContactStart = -1;
64 * Non-positional features have no (zero) start/end position.
65 * Kept as a separate list in case this criterion changes in future.
67 List<SequenceFeature> nonPositionalFeatures;
70 * contact features ordered by first contact position
72 List<SequenceFeature> contactFeatureStarts;
75 * contact features ordered by second contact position
77 List<SequenceFeature> contactFeatureEnds;
80 * IntervalStore holds remaining features and provides efficient
81 * query for features overlapping any given interval
83 IntervalStoreI<SequenceFeature> features;
86 * Feature groups represented in stored positional features
87 * (possibly including null)
89 Set<String> positionalFeatureGroups;
92 * Feature groups represented in stored non-positional features
93 * (possibly including null)
95 Set<String> nonPositionalFeatureGroups;
98 * the total length of all positional features; contact features count 1 to
99 * the total and 1 to size(), consistent with an average 'feature length' of 1
103 float positionalMinScore;
105 float positionalMaxScore;
107 float nonPositionalMinScore;
109 float nonPositionalMaxScore;
112 * Answers the 'length' of the feature, counting 0 for non-positional features
113 * and 1 for contact features
118 protected static int getFeatureLength(SequenceFeature feature)
120 if (feature.isNonPositional())
124 if (feature.isContactFeature())
128 return 1 + feature.getEnd() - feature.getBegin();
132 * Answers true if the list contains the feature, else false. This method is
133 * optimised for the condition that the list is sorted on feature start
134 * position ascending, and will give unreliable results if this does not hold.
140 public static boolean listContains(List<SequenceFeature> list,
141 SequenceFeature feature)
143 if (list == null || feature == null)
149 * locate the first entry in the list which does not precede the feature
151 int begin = feature.begin;
152 int pos = BinarySearcher.findFirst(list, true, Compare.GE, begin);
153 int len = list.size();
156 SequenceFeature sf = list.get(pos);
157 if (sf.begin > begin)
159 return false; // no match found
161 if (sf.equals(feature))
171 * A helper method to return the maximum of two floats, where a non-NaN value
172 * is treated as 'greater than' a NaN value (unlike Math.max which does the
178 protected static float max(float f1, float f2)
182 return Float.isNaN(f2) ? f1 : f2;
186 return Float.isNaN(f2) ? f1 : Math.max(f1, f2);
191 * A helper method to return the minimum of two floats, where a non-NaN value
192 * is treated as 'less than' a NaN value (unlike Math.min which does the
198 protected static float min(float f1, float f2)
202 return Float.isNaN(f2) ? f1 : f2;
206 return Float.isNaN(f2) ? f1 : Math.min(f1, f2);
211 * Constructor that defaults to using NCList IntervalStore
213 public FeatureStore()
215 this(IntervalStoreType.INTERVAL_STORE_NCLIST);
219 * Constructor that allows an alternative IntervalStore implementation to be
222 public FeatureStore(IntervalStoreType intervalStoreType)
224 features = getIntervalStore(intervalStoreType);
225 positionalFeatureGroups = new HashSet<>();
226 nonPositionalFeatureGroups = new HashSet<>();
227 positionalMinScore = Float.NaN;
228 positionalMaxScore = Float.NaN;
229 nonPositionalMinScore = Float.NaN;
230 nonPositionalMaxScore = Float.NaN;
232 // only construct nonPositionalFeatures or contactFeatures if needed
235 private IntervalStoreI<SequenceFeature> getIntervalStore(
236 IntervalStoreType type)
241 case INTERVAL_STORE_NCLIST:
242 return new intervalstore.impl.IntervalStore<>();
243 case INTERVAL_STORE_NCARRAY:
244 return new intervalstore.nonc.IntervalStore<>();
245 case INTERVAL_STORE_LINKED_LIST:
246 return new intervalstore.nonc.IntervalStore0<>();
251 * Add a contact feature to the lists that hold them ordered by start (first
252 * contact) and by end (second contact) position, ensuring the lists remain
253 * ordered, and returns true. This method allows duplicate features to be
254 * added, so test before calling to avoid this.
259 protected synchronized boolean addContactFeature(SequenceFeature feature)
261 if (contactFeatureStarts == null)
263 contactFeatureStarts = new ArrayList<>();
264 contactFeatureEnds = new ArrayList<>();
268 * insert into list sorted by start (first contact position):
269 * binary search the sorted list to find the insertion point
271 int insertAt = BinarySearcher.findFirst(contactFeatureStarts, true,
272 Compare.GE, feature.begin);
273 contactFeatureStarts.add(insertAt, feature);
275 * insert into list sorted by end (second contact position):
276 * binary search the sorted list to find the insertion point
278 contactFeatureEnds.add(findFirstEnd(contactFeatureEnds, feature.end),
285 * Adds one sequence feature to the store, and returns true, unless the
286 * feature is already contained in the store, in which case this method
287 * returns false. Containment is determined by SequenceFeature.equals()
292 public boolean addFeature(SequenceFeature feature)
294 if (feature.isContactFeature())
296 if (containsContactFeature(feature))
300 positionalFeatureGroups.add(feature.getFeatureGroup());
301 if (feature.begin > maxContactStart)
303 maxContactStart = feature.begin;
305 addContactFeature(feature);
307 else if (feature.isNonPositional())
309 if (containsNonPositionalFeature(feature))
314 addNonPositionalFeature(feature);
318 if (!features.add(feature, false))
322 positionalFeatureGroups.add(feature.getFeatureGroup());
323 if (feature.begin > maxStart)
325 maxStart = feature.begin;
330 * record the total extent of positional features, to make
331 * getTotalFeatureLength possible; we count the length of a
332 * contact feature as 1
334 totalExtent += getFeatureLength(feature);
337 * record the minimum and maximum score for positional
338 * and non-positional features
340 float score = feature.getScore();
341 if (!Float.isNaN(score))
343 if (feature.isNonPositional())
345 nonPositionalMinScore = min(nonPositionalMinScore, score);
346 nonPositionalMaxScore = max(nonPositionalMaxScore, score);
350 positionalMinScore = min(positionalMinScore, score);
351 positionalMaxScore = max(positionalMaxScore, score);
358 private void addFeaturesForGroup(String group,
359 Collection<SequenceFeature> sfs, List<SequenceFeature> result)
365 for (SequenceFeature sf : sfs)
367 String featureGroup = sf.getFeatureGroup();
368 if (group == null && featureGroup == null
369 || group != null && group.equals(featureGroup))
377 * Adds the feature to the list of non-positional features (with lazy
378 * instantiation of the list if it is null), and returns true. The feature
379 * group is added to the set of distinct feature groups for non-positional
380 * features. This method allows duplicate features, so test before calling to
385 protected boolean addNonPositionalFeature(SequenceFeature feature)
387 if (nonPositionalFeatures == null)
389 nonPositionalFeatures = new ArrayList<>();
392 nonPositionalFeatures.add(feature);
394 nonPositionalFeatureGroups.add(feature.getFeatureGroup());
400 * Answers true if this store contains the given feature (testing by
401 * SequenceFeature.equals), else false
406 public boolean contains(SequenceFeature feature)
408 if (feature.isNonPositional())
410 return containsNonPositionalFeature(feature);
413 if (feature.isContactFeature())
415 return containsContactFeature(feature);
418 return containsPositionalFeature(feature);
421 private boolean containsPositionalFeature(SequenceFeature feature)
423 return features == null || feature.begin > maxStart ? false
424 : features.contains(feature);
428 * Answers true if this store already contains a contact feature equal to the
429 * given feature (by {@code SequenceFeature.equals()} test), else false
434 private boolean containsContactFeature(SequenceFeature feature)
436 return contactFeatureStarts != null && feature.begin <= maxContactStart
437 && listContains(contactFeatureStarts, feature);
441 * Answers true if this store already contains a non-positional feature equal
442 * to the given feature (by {@code SequenceFeature.equals()} test), else false
447 private boolean containsNonPositionalFeature(SequenceFeature feature)
449 return nonPositionalFeatures == null ? false
450 : nonPositionalFeatures.contains(feature);
454 * Deletes the given feature from the store, returning true if it was found
455 * (and deleted), else false. This method makes no assumption that the feature
456 * is in the 'expected' place in the store, in case it has been modified since
461 public synchronized boolean delete(SequenceFeature sf)
463 boolean removed = false;
466 * try contact positions (and if found, delete
467 * from both lists of contact positions)
469 if (!removed && contactFeatureStarts != null)
471 removed = contactFeatureStarts.remove(sf);
474 contactFeatureEnds.remove(sf);
479 * if not found, try non-positional features
481 if (!removed && nonPositionalFeatures != null)
483 removed = nonPositionalFeatures.remove(sf);
487 * if not found, try nested features
489 if (!removed && features != null)
491 removed = features.remove(sf);
502 public List<SequenceFeature> findOverlappingFeatures(long start, long end)
504 return findOverlappingFeatures(start, end, null);
507 public List<SequenceFeature> getContactFeatures()
509 return getContactFeatures(new ArrayList<>());
513 * Answers a list of all contact features. If there are none, returns an
514 * immutable empty list.
518 public List<SequenceFeature> getContactFeatures(
519 List<SequenceFeature> result)
521 if (contactFeatureStarts != null)
523 result.addAll(contactFeatureStarts);
529 * Answers the number of positional (or non-positional) features stored.
530 * Contact features count as 1.
535 public int getFeatureCount(boolean positional)
539 return nonPositionalFeatures == null ? 0
540 : nonPositionalFeatures.size();
543 return (contactFeatureStarts == null ? 0 : contactFeatureStarts.size())
548 * Answers the set of distinct feature groups stored, possibly including null,
549 * as an unmodifiable view of the set. The parameter determines whether the
550 * groups for positional or for non-positional features are returned.
552 * @param positionalFeatures
555 public Set<String> getFeatureGroups(boolean positionalFeatures)
557 if (positionalFeatures)
559 return Collections.unmodifiableSet(positionalFeatureGroups);
563 return nonPositionalFeatureGroups == null
564 ? Collections.<String> emptySet()
565 : Collections.unmodifiableSet(nonPositionalFeatureGroups);
569 public Collection<SequenceFeature> getFeatures()
575 * Answers a list of all either positional or non-positional features whose
576 * feature group matches the given group (which may be null)
582 public List<SequenceFeature> getFeaturesForGroup(boolean positional,
585 List<SequenceFeature> result = new ArrayList<>();
588 * if we know features don't include the target group, no need
589 * to inspect them for matches
591 if (positional && !positionalFeatureGroups.contains(group)
592 || !positional && !nonPositionalFeatureGroups.contains(group))
599 addFeaturesForGroup(group, contactFeatureStarts, result);
600 addFeaturesForGroup(group, features, result);
604 addFeaturesForGroup(group, nonPositionalFeatures, result);
610 * Answers the maximum score held for positional or non-positional features.
611 * This may be Float.NaN if there are no features, are none has a non-NaN
617 public float getMaximumScore(boolean positional)
619 return positional ? positionalMaxScore : nonPositionalMaxScore;
623 * Answers the minimum score held for positional or non-positional features.
624 * This may be Float.NaN if there are no features, are none has a non-NaN
630 public float getMinimumScore(boolean positional)
632 return positional ? positionalMinScore : nonPositionalMinScore;
635 public List<SequenceFeature> getNonPositionalFeatures()
637 return getNonPositionalFeatures(new ArrayList<>());
641 * Answers a list of all non-positional features. If there are none, returns
642 * an immutable empty list.
646 public List<SequenceFeature> getNonPositionalFeatures(
647 List<SequenceFeature> result)
649 if (nonPositionalFeatures != null)
651 result.addAll(nonPositionalFeatures);
656 public List<SequenceFeature> getPositionalFeatures()
658 return getPositionalFeatures(new ArrayList<>());
662 * Answers a list of all positional features stored, in no guaranteed order
666 public List<SequenceFeature> getPositionalFeatures(
667 List<SequenceFeature> result)
671 * add any contact features - from the list by start position
673 if (contactFeatureStarts != null)
675 result.addAll(contactFeatureStarts);
679 * add any nested features
681 if (features != null)
683 result.addAll(features);
690 * Answers the total length of positional features (or zero if there are
691 * none). Contact features contribute a value of 1 to the total.
695 public int getTotalFeatureLength()
701 * Answers true if this store has no features, else false
705 public boolean isEmpty()
707 boolean hasFeatures = (contactFeatureStarts != null
708 && !contactFeatureStarts.isEmpty())
709 || (nonPositionalFeatures != null
710 && !nonPositionalFeatures.isEmpty())
711 || features.size() > 0;
717 * Rescan all features to recompute any cached values after an entry has been
718 * deleted. This is expected to be an infrequent event, so performance here is
721 protected synchronized void rescanAfterDelete()
723 positionalFeatureGroups.clear();
724 nonPositionalFeatureGroups.clear();
726 positionalMinScore = Float.NaN;
727 positionalMaxScore = Float.NaN;
728 nonPositionalMinScore = Float.NaN;
729 nonPositionalMaxScore = Float.NaN;
731 * scan non-positional features for groups and scores
733 if (nonPositionalFeatures != null)
735 List<SequenceFeature> list = nonPositionalFeatures;
736 for (int i = 0, n = list.size(); i < n; i++)
738 SequenceFeature sf = list.get(i);
739 nonPositionalFeatureGroups.add(sf.getFeatureGroup());
740 float score = sf.getScore();
741 nonPositionalMinScore = min(nonPositionalMinScore, score);
742 nonPositionalMaxScore = max(nonPositionalMaxScore, score);
747 * scan positional features for groups, scores and extents
750 rescanPositional(contactFeatureStarts);
751 rescanPositional(features);
754 private void rescanPositional(Collection<SequenceFeature> sfs)
760 for (SequenceFeature sf : sfs)
762 positionalFeatureGroups.add(sf.getFeatureGroup());
763 float score = sf.getScore();
764 positionalMinScore = min(positionalMinScore, score);
765 positionalMaxScore = max(positionalMaxScore, score);
766 totalExtent += getFeatureLength(sf);
771 * Adds the shift amount to the start and end of all positional features whose
772 * start position is at or after fromPosition. Returns true if at least one
773 * feature was shifted, else false.
775 * @param fromPosition
779 public synchronized boolean shiftFeatures(int fromPosition, int shiftBy)
782 * Because begin and end are final fields (to ensure the data store's
783 * integrity), we have to delete each feature and re-add it as amended.
784 * (Although a simple shift of all values would preserve data integrity!)
786 boolean modified = false;
787 List<SequenceFeature> list = getPositionalFeatures();
788 for (int i = 0, n = list.size(); i < n; i++)
790 SequenceFeature sf = list.get(i);
791 if (sf.getBegin() >= fromPosition)
794 int newBegin = sf.getBegin() + shiftBy;
795 int newEnd = sf.getEnd() + shiftBy;
798 * sanity check: don't shift left of the first residue
802 newBegin = Math.max(1, newBegin);
803 SequenceFeature sf2 = new SequenceFeature(sf, newBegin, newEnd,
804 sf.getFeatureGroup(), sf.getScore());
814 * Answers the position (0, 1...) in the list of the first entry whose end
815 * position is not less than {@ pos}. If no such entry is found, answers the
816 * length of the list.
822 protected int findFirstEnd(List<SequenceFeature> list, long pos)
824 return BinarySearcher.findFirst(list, false, Compare.GE, (int) pos);
828 * Adds contact features to the result list where either the second or the
829 * first contact position lies within the target range
835 protected void findContactFeatures(long from, long to,
836 List<SequenceFeature> result)
838 if (contactFeatureStarts != null)
840 findContactStartOverlaps(from, to, result);
841 findContactEndOverlaps(from, to, result);
846 * Adds to the result list any contact features whose end (second contact
847 * point), but not start (first contact point), lies in the query from-to
854 private void findContactEndOverlaps(long from, long to,
855 List<SequenceFeature> result)
858 * find the first contact feature (if any)
859 * whose end point is not before the target range
861 int index = findFirstEnd(contactFeatureEnds, from);
863 int n = contactFeatureEnds.size();
866 SequenceFeature sf = contactFeatureEnds.get(index);
867 if (!sf.isContactFeature())
869 System.err.println("Error! non-contact feature type " + sf.getType()
870 + " in contact features list");
875 int begin = sf.getBegin();
876 if (begin >= from && begin <= to)
879 * this feature's first contact position lies in the search range
880 * so we don't include it in results a second time
886 if (sf.getEnd() > to)
889 * this feature (and all following) has end point after the target range
895 * feature has end >= from and end <= to
896 * i.e. contact end point lies within overlap search range
904 * Adds contact features whose start position lies in the from-to range to the
911 private void findContactStartOverlaps(long from, long to,
912 List<SequenceFeature> result)
914 int index = BinarySearcher.findFirst(contactFeatureStarts, true,
915 Compare.GE, (int) from);
917 while (index < contactFeatureStarts.size())
919 SequenceFeature sf = contactFeatureStarts.get(index);
920 if (!sf.isContactFeature())
922 System.err.println("Error! non-contact feature " + sf.toString()
923 + " in contact features list");
927 if (sf.getBegin() > to)
930 * this feature's start (and all following) follows the target range
936 * feature has begin >= from and begin <= to
937 * i.e. contact start point lies within overlap search range
945 * Returns a (possibly empty) list of features whose extent overlaps the given
946 * range. The returned list is not ordered. Contact features are included if
947 * either of the contact points lies within the range. If the {@code result}
948 * parameter is not null, new entries are added to this list and the (possibly
949 * extended) list returned.
952 * start position of overlap range (inclusive)
954 * end position of overlap range (inclusive)
958 public List<SequenceFeature> findOverlappingFeatures(long start, long end,
959 List<SequenceFeature> result)
963 result = new ArrayList<>();
966 findContactFeatures(start, end, result);
967 features.findOverlaps(start, end, result);