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.List;
29 * An adaption of FeatureStore that is efficient and lightweight, accelerating
30 * processing speed in JavaScript.
33 * @author Bob Hanson 2019.08.03
36 public class FeatureStoreJS extends FeatureStore
38 boolean contactsTainted = true;
42 * internal reference to features as an ArrayList
44 private ArrayList<SequenceFeature> featureList;
47 * contact features ordered by first contact position
49 private SequenceFeature[] orderedFeatureStarts;
52 * indicates that we need to rebuild orderedFeatureStarts and reset index
55 private boolean isTainted = true;
60 public FeatureStoreJS()
62 features = featureList = new ArrayList<>();
66 * Add a contact feature to the lists that hold them ordered by start (first
67 * contact) and by end (second contact) position, ensuring the lists remain
68 * ordered, and returns true. This method allows duplicate features to be
69 * added, so test before calling to avoid this.
75 protected synchronized boolean addContactFeature(SequenceFeature feature)
77 if (contactFeatureStarts == null)
79 contactFeatureStarts = new ArrayList<>();
80 contactFeatureEnds = new ArrayList<>();
82 contactFeatureStarts.add(
83 findFirstBegin(contactFeatureStarts, feature.begin), feature);
84 contactFeatureEnds.add(findFirstEnd(contactFeatureEnds, feature.end),
89 // The following methods use a linked list of containment in SequenceFeature
90 // rather than IntervalStore.
92 // There are two parts --- initialization, and overlap searching.
94 // Initialization involves two steps:
96 // (1) sorting of features by start position using a standard Array.sort with
98 // (2) linking of features, effectively nesting them.
100 // Searching involves three steps:
102 // (1) binary search for a starting point within the sorted features array.
103 // (2) traverse the linked lists with an end check to read out the
104 // overlapped features at this position.
105 // (3) For an interval, find the last feature that starts in this interval,
106 // and add all features up through that feature.
108 // All of this is done with very simple standard methods.
113 * Adds one feature to the IntervalStore that can manage nested features
114 * (creating the IntervalStore if necessary)
117 protected synchronized void addPositionalFeature(SequenceFeature feature)
119 featureList.add(findFirstBegin(featureList, feature.begin), feature);
124 protected boolean containsFeature(SequenceFeature feature)
127 return getEquivalentFeatureIndex(featureList, feature) >= 0;
131 protected boolean findAndRemoveNonContactFeature(SequenceFeature sf)
133 int pos = getEquivalentFeatureIndex(featureList, sf);
138 featureList.remove(pos);
139 return (isTainted = true);
143 * Adds contact features to the result list where either the second or the
144 * first contact position lies within the target range
151 protected void findContactFeatures(long from, long to,
152 List<SequenceFeature> result)
154 getContactStartOverlaps(from, to, result);
155 getContactEndOverlaps(from, to, result);
159 * Locate the first feature start that is at or after this position.
163 static int ntotal = 0, ncaught = 0;
166 protected int findFirstBegin(List<SequenceFeature> list, long pos)
168 int matched = list.size();
169 int end = matched - 1;
170 int start = (end < 0 || list.get(end).begin < pos ? matched : 0);
175 // else if (start > end)
181 // System.out.println(
182 // "FSJS " + ncaught + "/" + ntotal + " " + pos + "/"
183 // + (end < 0 ? -1 : list.get(end).begin) + " "
184 // + start + "/" + matched);
188 int mid = (start + end) / 2;
189 if (list.get(mid).begin >= pos)
203 * Locate the feature start that is after or at this position.
208 protected int findFirstEnd(List<SequenceFeature> list, long pos)
211 int end = list.size() - 1;
212 int matched = list.size();
216 int mid = (start + end) / 2;
217 if (list.get(mid).end >= pos)
231 * Returns a (possibly empty) list of features whose extent overlaps the given
232 * range. The returned list is not ordered. Contact features are included if
233 * either of the contact points lies within the range.
236 * start position of overlap range (inclusive)
238 * end position of overlap range (inclusive)
243 public List<SequenceFeature> findOverlappingFeatures(long start, long end,
244 List<SequenceFeature> result)
248 result = new ArrayList<>();
250 if (contactFeatureStarts != null)
254 getContactPoints(contactFeatureStarts, start, result, true);
255 getContactPoints(contactFeatureEnds, start, result, false);
259 findContactFeatures(start, end, result);
262 if (featureList.size() > 0)
264 getOverlaps(start, end, result);
270 * A binary search identical to the one used for contact start/end, but here
271 * we return the feature itself. Unlike Collection.BinarySearch, all we have
272 * to be is close, not exact, and we make sure if there is a string of
273 * identical starts, then we slide to the end so that we can check all of
280 private int getClosestFeature(SequenceFeature[] l, long pos)
283 int high = l.length - 1;
286 int mid = (low + high) >>> 1;
287 SequenceFeature f = l[mid];
288 switch (Long.signum(f.begin - pos))
298 while (++mid <= high && l[mid].begin == pos)
305 return (high < 0 ? -1 : high);
309 * Adds to the result list any contact features whose end (second contact
310 * point), but not start (first contact point), lies in the query from-to
318 private void getContactEndOverlaps(long from, long to,
319 List<SequenceFeature> result)
321 // find the first contact feature (if any)
322 // with end point not before the target range
324 for (int i = findFirstEnd(contactFeatureEnds,
325 from), n = contactFeatureEnds.size(); i < n; i++)
327 SequenceFeature sf = contactFeatureEnds.get(i);
328 if (sf.begin >= from && sf.begin <= to)
330 // this feature's first contact position lies in the search range
331 // so we don't include it in results a second time
337 // this feature (and all following) has end point after the target range
341 // feature has end >= from and end <= to
342 // i.e. contact end point lies within overlap search range
348 * Binary search for contact start or end at a given (Overview) position.
355 * @author Bob Hanson 2019.07.30
357 private void getContactPoints(List<SequenceFeature> l, long pos,
358 List<SequenceFeature> result, boolean isStart)
361 int high = l.size() - 1;
364 int mid = (low + high) >>> 1;
365 SequenceFeature f = l.get(mid);
366 switch (Long.signum((isStart ? f.begin : f.end) - pos))
377 // could be "5" in 12345556788 ?
378 while (++mid <= high && (f = l.get(mid)) != null
379 && (isStart ? f.begin : f.end) == pos)
383 while (--m >= low && (f = l.get(m)) != null
384 && (isStart ? f.begin : f.end) == pos)
394 * Adds contact features whose start position lies in the from-to range to the
402 private void getContactStartOverlaps(long from, long to,
403 List<SequenceFeature> result)
405 for (int i = findFirstBegin(contactFeatureStarts,
406 from), n = contactFeatureStarts.size(); i < n; i++)
408 SequenceFeature sf = contactFeatureStarts.get(i);
418 * Since we are traversing the sorted feature array in a forward direction,
419 * all elements prior to the one we are working on have been fully linked. All
420 * we are doing is following those links until we find the first array feature
421 * with a containedBy element that has an end >= our begin point. It is
422 * generally a very short list -- maybe one or two depths. But it might be
429 private SequenceFeature getContainedBy(SequenceFeature sf,
432 int begin = sf0.begin;
437 // System.out.println("\nFSJS found " + sf0 + "\nFS in " + sf);
446 * Fast find of known features already on the list; slower removal of
447 * equivalent feature, not necessarily identical.
450 * @return 0-based index for this feature in featureList
453 protected int getEquivalentFeatureIndex(List<SequenceFeature> list,
454 SequenceFeature feature)
456 int pos = feature.index1 - 1;
457 if (!isTainted && pos >= 0)
461 return super.getEquivalentFeatureIndex(list, feature);
465 * Find all overlaps; special case when there is only one feature. The
466 * required array of start-sorted SequenceFeature is created lazily.
472 private void getOverlaps(long start, long end,
473 List<SequenceFeature> result)
475 int n = featureList.size();
481 justCheckOne(featureList.get(0), start, end, result);
486 orderedFeatureStarts = featureList
487 .toArray(new SequenceFeature[featureList.size()]);
488 linkFeatures(orderedFeatureStarts);
494 // (1) Find the closest feature to this position.
496 int index = getClosestFeature(orderedFeatureStarts, start);
497 SequenceFeature sf = (index < 0 ? null : orderedFeatureStarts[index]);
499 // (2) Traverse the containedBy field, checking for overlap.
510 // (3) For an interval, find the last feature that starts in this interval,
511 // and add all features up through that feature.
515 // fill in with all features that start within this interval, fully
517 int index2 = getClosestFeature(orderedFeatureStarts, end);
518 while (++index <= index2)
520 result.add(orderedFeatureStarts[index]);
527 * Quick check when we only have one feature.
534 private void justCheckOne(SequenceFeature sf, long start, long end,
535 List<SequenceFeature> result)
537 if (sf.begin <= end && sf.end >= start)
545 * Run through the sorted sequence array once, building the containedBy linked
546 * list references. Does a check first to make sure there is actually
547 * something out there that is overlapping. A null for sf.containedBy means
548 * there are no overlaps for this feature.
552 private void linkFeatures(SequenceFeature[] features)
554 int n = features.length;
560 features[0].index1 = 1;
563 int maxEnd = features[0].end;
564 for (int i = 1; i < n;)
566 SequenceFeature sf = features[i];
567 if (sf.begin <= maxEnd)
569 sf.containedBy = getContainedBy(features[i - 1], sf);