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);
359 * A helper method that adds to the result list any features from the
360 * collection provided whose feature group matches the specified group
366 private void addFeaturesForGroup(String group,
367 Collection<SequenceFeature> sfs, List<SequenceFeature> result)
373 for (SequenceFeature sf : sfs)
375 String featureGroup = sf.getFeatureGroup();
376 if (group == null && featureGroup == null
377 || group != null && group.equals(featureGroup))
385 * Adds the feature to the list of non-positional features (with lazy
386 * instantiation of the list if it is null), and returns true. The feature
387 * group is added to the set of distinct feature groups for non-positional
388 * features. This method allows duplicate features, so test before calling to
393 protected boolean addNonPositionalFeature(SequenceFeature feature)
395 if (nonPositionalFeatures == null)
397 nonPositionalFeatures = new ArrayList<>();
400 nonPositionalFeatures.add(feature);
402 nonPositionalFeatureGroups.add(feature.getFeatureGroup());
408 * Answers true if this store contains the given feature (testing by
409 * SequenceFeature.equals), else false
414 public boolean contains(SequenceFeature feature)
416 if (feature.isNonPositional())
418 return containsNonPositionalFeature(feature);
421 if (feature.isContactFeature())
423 return containsContactFeature(feature);
426 return containsPositionalFeature(feature);
429 private boolean containsPositionalFeature(SequenceFeature feature)
431 return features == null || feature.begin > maxStart ? false
432 : features.contains(feature);
436 * Answers true if this store already contains a contact feature equal to the
437 * given feature (by {@code SequenceFeature.equals()} test), else false
442 private boolean containsContactFeature(SequenceFeature feature)
444 return contactFeatureStarts != null && feature.begin <= maxContactStart
445 && listContains(contactFeatureStarts, feature);
449 * Answers true if this store already contains a non-positional feature equal
450 * to the given feature (by {@code SequenceFeature.equals()} test), else false
455 private boolean containsNonPositionalFeature(SequenceFeature feature)
457 return nonPositionalFeatures == null ? false
458 : nonPositionalFeatures.contains(feature);
462 * Deletes the given feature from the store, returning true if it was found
463 * (and deleted), else false. This method makes no assumption that the feature
464 * is in the 'expected' place in the store, in case it has been modified since
469 public synchronized boolean delete(SequenceFeature sf)
471 boolean removed = false;
474 * try contact positions (and if found, delete
475 * from both lists of contact positions)
477 if (!removed && contactFeatureStarts != null)
479 removed = contactFeatureStarts.remove(sf);
482 contactFeatureEnds.remove(sf);
487 * if not found, try non-positional features
489 if (!removed && nonPositionalFeatures != null)
491 removed = nonPositionalFeatures.remove(sf);
495 * if not found, try nested features
497 if (!removed && features != null)
499 removed = features.remove(sf);
510 public List<SequenceFeature> findFeatures(long start, long end)
512 return findFeatures(start, end, null);
516 * Returns a (possibly empty) list of features whose extent overlaps the given
517 * range. The returned list is not ordered. Contact features are included if
518 * either of the contact points lies within the range. If the {@code result}
519 * parameter is not null, new entries are added to this list and the (possibly
520 * extended) list returned.
523 * start position of overlap range (inclusive)
525 * end position of overlap range (inclusive)
529 public List<SequenceFeature> findFeatures(long start, long end,
530 List<SequenceFeature> result)
534 result = new ArrayList<>();
537 findContactFeatures(start, end, result);
538 features.findOverlaps(start, end, result);
543 public List<SequenceFeature> getContactFeatures()
545 return getContactFeatures(new ArrayList<>());
549 * Answers a list of all contact features. If there are none, returns an
550 * immutable empty list.
554 public List<SequenceFeature> getContactFeatures(
555 List<SequenceFeature> result)
557 if (contactFeatureStarts != null)
559 result.addAll(contactFeatureStarts);
565 * Answers the number of positional (or non-positional) features stored.
566 * Contact features count as 1.
571 public int getFeatureCount(boolean positional)
575 return nonPositionalFeatures == null ? 0
576 : nonPositionalFeatures.size();
581 if (contactFeatureStarts != null)
583 // note a contact feature (start/end) counts as one
584 size += contactFeatureStarts.size();
587 if (features != null)
589 size += features.size();
595 * Answers the set of distinct feature groups stored, possibly including null,
596 * as an unmodifiable view of the set. The parameter determines whether the
597 * groups for positional or for non-positional features are returned.
599 * @param positionalFeatures
602 public Set<String> getFeatureGroups(boolean positionalFeatures)
604 if (positionalFeatures)
606 return Collections.unmodifiableSet(positionalFeatureGroups);
610 return nonPositionalFeatureGroups == null
611 ? Collections.<String> emptySet()
612 : Collections.unmodifiableSet(nonPositionalFeatureGroups);
617 * Answers a list of all either positional or non-positional features whose
618 * feature group matches the given group (which may be null)
624 public List<SequenceFeature> getFeaturesForGroup(boolean positional,
627 List<SequenceFeature> result = new ArrayList<>();
630 * if we know features don't include the target group, no need
631 * to inspect them for matches
633 if (positional && !positionalFeatureGroups.contains(group)
634 || !positional && !nonPositionalFeatureGroups.contains(group))
641 addFeaturesForGroup(group, contactFeatureStarts, result);
642 addFeaturesForGroup(group, features, result);
646 addFeaturesForGroup(group, nonPositionalFeatures, result);
652 * Answers the maximum score held for positional or non-positional features.
653 * This may be Float.NaN if there are no features, are none has a non-NaN
659 public float getMaximumScore(boolean positional)
661 return positional ? positionalMaxScore : nonPositionalMaxScore;
665 * Answers the minimum score held for positional or non-positional features.
666 * This may be Float.NaN if there are no features, are none has a non-NaN
672 public float getMinimumScore(boolean positional)
674 return positional ? positionalMinScore : nonPositionalMinScore;
678 * Answers a (possibly empty) list of all non-positional features
682 public List<SequenceFeature> getNonPositionalFeatures()
684 List<SequenceFeature> result = new ArrayList<>();
685 getNonPositionalFeatures(result);
690 * Adds any stored non-positional features to the result list
694 public void getNonPositionalFeatures(List<SequenceFeature> result)
696 if (nonPositionalFeatures != null)
698 result.addAll(nonPositionalFeatures);
703 * Returns a (possibly empty) list of all positional features stored
707 public List<SequenceFeature> getPositionalFeatures()
709 List<SequenceFeature> result = new ArrayList<>();
710 getPositionalFeatures(result);
716 * Adds all positional features stored to the result list, in no guaranteed
717 * order, and with no check for duplicates
719 public void getPositionalFeatures(List<SequenceFeature> result)
722 * add any contact features - from the list by start position
724 if (contactFeatureStarts != null)
726 result.addAll(contactFeatureStarts);
730 * add any nested features
732 if (features != null)
734 result.addAll(features);
739 * Answers the total length of positional features (or zero if there are
740 * none). Contact features contribute a value of 1 to the total.
744 public int getTotalFeatureLength()
750 * Answers true if this store has no features, else false
754 public boolean isEmpty()
756 boolean hasFeatures = (contactFeatureStarts != null
757 && !contactFeatureStarts.isEmpty())
758 || (nonPositionalFeatures != null
759 && !nonPositionalFeatures.isEmpty())
760 || (features != null && features.size() > 0);
766 * Rescan all features to recompute any cached values after an entry has been
767 * deleted. This is expected to be an infrequent event, so performance here is
770 protected synchronized void rescanAfterDelete()
772 positionalFeatureGroups.clear();
773 nonPositionalFeatureGroups.clear();
775 positionalMinScore = Float.NaN;
776 positionalMaxScore = Float.NaN;
777 nonPositionalMinScore = Float.NaN;
778 nonPositionalMaxScore = Float.NaN;
780 * scan non-positional features for groups and scores
782 if (nonPositionalFeatures != null)
784 for (int i = 0, n = nonPositionalFeatures.size(); i < n; i++)
786 SequenceFeature sf = nonPositionalFeatures.get(i);
787 nonPositionalFeatureGroups.add(sf.getFeatureGroup());
788 float score = sf.getScore();
789 nonPositionalMinScore = min(nonPositionalMinScore, score);
790 nonPositionalMaxScore = max(nonPositionalMaxScore, score);
794 rescanPositional(contactFeatureStarts);
795 rescanPositional(features);
799 * Scans the given features and updates cached feature groups, minimum and
800 * maximum feature score, and total feature extent (length) for positional
805 private void rescanPositional(Collection<SequenceFeature> sfs)
811 for (SequenceFeature sf : sfs)
813 positionalFeatureGroups.add(sf.getFeatureGroup());
814 float score = sf.getScore();
815 positionalMinScore = min(positionalMinScore, score);
816 positionalMaxScore = max(positionalMaxScore, score);
817 totalExtent += getFeatureLength(sf);
822 * Adds the shift amount to the start and end of all positional features whose
823 * start position is at or after fromPosition. Returns true if at least one
824 * feature was shifted, else false.
826 * @param fromPosition
830 public synchronized boolean shiftFeatures(int fromPosition, int shiftBy)
833 * Because begin and end are final fields (to ensure the data store's
834 * integrity), we have to delete each feature and re-add it as amended.
835 * (Although a simple shift of all values would preserve data integrity!)
837 boolean modified = false;
838 List<SequenceFeature> list = getPositionalFeatures();
839 for (int i = 0, n = list.size(); i < n; i++)
841 SequenceFeature sf = list.get(i);
842 if (sf.getBegin() >= fromPosition)
845 int newBegin = sf.getBegin() + shiftBy;
846 int newEnd = sf.getEnd() + shiftBy;
849 * sanity check: don't shift left of the first residue
853 newBegin = Math.max(1, newBegin);
854 SequenceFeature sf2 = new SequenceFeature(sf, newBegin, newEnd,
855 sf.getFeatureGroup(), sf.getScore());
865 * Answers the position (0, 1...) in the list of the first entry whose end
866 * position is not less than {@ pos}. If no such entry is found, answers the
867 * length of the list.
873 protected int findFirstEnd(List<SequenceFeature> list, long pos)
875 return BinarySearcher.findFirst(list, false, Compare.GE, (int) pos);
879 * Adds contact features to the result list where either the second or the
880 * first contact position lies within the target range
886 protected void findContactFeatures(long from, long to,
887 List<SequenceFeature> result)
889 if (contactFeatureStarts != null)
891 findContactStartOverlaps(from, to, result);
892 findContactEndOverlaps(from, to, result);
897 * Adds to the result list any contact features whose end (second contact
898 * point), but not start (first contact point), lies in the query from-to
905 private void findContactEndOverlaps(long from, long to,
906 List<SequenceFeature> result)
909 * find the first contact feature (if any)
910 * whose end point is not before the target range
912 int index = findFirstEnd(contactFeatureEnds, from);
914 int n = contactFeatureEnds.size();
917 SequenceFeature sf = contactFeatureEnds.get(index);
918 if (!sf.isContactFeature())
920 System.err.println("Error! non-contact feature type " + sf.getType()
921 + " in contact features list");
926 int begin = sf.getBegin();
927 if (begin >= from && begin <= to)
930 * this feature's first contact position lies in the search range
931 * so we don't include it in results a second time
937 if (sf.getEnd() > to)
940 * this feature (and all following) has end point after the target range
946 * feature has end >= from and end <= to
947 * i.e. contact end point lies within overlap search range
955 * Adds contact features whose start position lies in the from-to range to the
962 private void findContactStartOverlaps(long from, long to,
963 List<SequenceFeature> result)
965 int index = BinarySearcher.findFirst(contactFeatureStarts, true,
966 Compare.GE, (int) from);
968 while (index < contactFeatureStarts.size())
970 SequenceFeature sf = contactFeatureStarts.get(index);
971 if (!sf.isContactFeature())
973 System.err.println("Error! non-contact feature " + sf.toString()
974 + " in contact features list");
978 if (sf.getBegin() > to)
981 * this feature's start (and all following) follows the target range
987 * feature has begin >= from and begin <= to
988 * i.e. contact start point lies within overlap search range