import java.util.ArrayList;
import java.util.Arrays;
+import java.util.BitSet;
import java.util.List;
/**
*
* Use at your own risk!
*
- * TODO: efficient implementation of private posMap method
- *
* TODO: test/ensure that sense of from and to ratio start position is conserved
* (codon start position recovery)
*/
*/
protected int[][] makeFromMap()
{
- // TODO not used - remove??
+ // TODO only used for test - remove??
return posMap(fromShifts, fromRatio, toShifts, toRatio);
}
*/
protected int[][] makeToMap()
{
- // TODO not used - remove??
+ // TODO only used for test - remove??
return posMap(toShifts, toRatio, fromShifts, fromRatio);
}
* @return int[] { from, to pos in range }, int[range.to-range.from+1]
* returning mapped position
*/
- private int[][] posMap(List<int[]> shiftTo, int ratio,
- List<int[]> shiftFrom, int toRatio)
+ private int[][] posMap(List<int[]> shiftTo, int sourceRatio,
+ List<int[]> shiftFrom, int targetRatio)
{
- // TODO not used - remove??
+ // TODO only used for test - remove??
int iv = 0, ivSize = shiftTo.size();
if (iv >= ivSize)
{
int mp[][] = new int[to - from + 2][];
for (int i = 0; i < mp.length; i++)
{
- int[] m = shift(i + from, shiftTo, ratio, shiftFrom, toRatio);
+ int[] m = shift(i + from, shiftTo, sourceRatio, shiftFrom, targetRatio);
if (m != null)
{
if (i == 0)
}
/**
- * find series of intervals mapping from start-end in the From map.
- *
- * @param start
- * position mapped 'to'
- * @param end
- * position mapped 'to'
- * @return series of [start, end] ranges in sequence mapped 'from'
- */
- public int[] locateInFrom(int start, int end)
- {
- // inefficient implementation
- int fromStart[] = shiftTo(start);
- // needs to be inclusive of end of symbol position
- int fromEnd[] = shiftTo(end);
- System.out.println("locateInFrom");
- return getIntervals(fromShifts, fromStart, fromEnd, fromRatio);
- }
-
- /**
- * find series of intervals mapping from start-end in the to map.
- *
- * @param start
- * position mapped 'from'
- * @param end
- * position mapped 'from'
- * @return series of [start, end] ranges in sequence mapped 'to'
- */
- public int[] locateInTo(int start, int end)
- {
- int toStart[] = shiftFrom(start);
- int toEnd[] = shiftFrom(end);
- return getIntervals(toShifts, toStart, toEnd, toRatio);
- }
-
- /**
* like shift - except returns the intervals in the given vector of shifts
* which were spanned in traversing fromStart to fromEnd
*
*/
public int getToPosition(int mpos)
{
- // TODO not used - remove??
int[] mp = shiftTo(mpos);
if (mp != null)
{
}
/**
- * get range of positions in To frame for the mpos word in From
- *
- * @param mpos
- * position in From
- * @return null or int[] first position in To for mpos, last position in to
- * for Mpos
- */
- public int[] getToWord(int mpos)
- {
- int[] mp = shiftTo(mpos);
- if (mp != null)
- {
- return new int[] { mp[0], mp[0] + mp[2] * (getFromRatio() - 1) };
- }
- return null;
- }
-
- /**
- * get From position in the associated reference frame for position pos in the
- * associated sequence.
- *
- * @param pos
- * @return
- */
- public int getMappedPosition(int pos)
- {
- // TODO not used - remove??
- int[] mp = shiftFrom(pos);
- if (mp != null)
- {
- return mp[0];
- }
- return pos;
- }
-
- public int[] getMappedWord(int pos)
- {
- // TODO not used - remove??
- int[] mp = shiftFrom(pos);
- if (mp != null)
- {
- return new int[] { mp[0], mp[0] + mp[2] * (getToRatio() - 1) };
- }
- return null;
- }
-
- /**
*
* @return a MapList whose From range is this maplist's To Range, and vice
* versa
}
/**
- * test for containment rather than equivalence to another mapping
- *
- * @param map
- * to be tested for containment
- * @return true if local or mapped range map contains or is contained by this
- * mapping
- */
- public boolean containsEither(boolean local, MapList map)
- {
- // TODO not used - remove?
- if (local)
- {
- return ((getFromLowest() >= map.getFromLowest()
- && getFromHighest() <= map.getFromHighest())
- || (getFromLowest() <= map.getFromLowest()
- && getFromHighest() >= map.getFromHighest()));
- }
- else
- {
- return ((getToLowest() >= map.getToLowest()
- && getToHighest() <= map.getToHighest())
- || (getToLowest() <= map.getToLowest()
- && getToHighest() >= map.getToHighest()));
- }
- }
-
- /**
* String representation - for debugging, not guaranteed not to change
*/
@Override
List<int[]> toRanges = new ArrayList<>();
for (int[] range : getToRanges())
{
+ int fromLength = Math.abs(range[1] - range[0]) + 1;
int[] transferred = map.locateInTo(range[0], range[1]);
if (transferred == null || transferred.length % 2 != 0)
{
* convert [start1, end1, start2, end2, ...]
* to [[start1, end1], [start2, end2], ...]
*/
+ int toLength = 0;
for (int i = 0; i < transferred.length;)
{
toRanges.add(new int[] { transferred[i], transferred[i + 1] });
+ toLength += Math.abs(transferred[i + 1] - transferred[i]) + 1;
i += 2;
}
+
+ /*
+ * check we mapped the full range - if not, abort
+ */
+ if (fromLength * map.getToRatio() != toLength * map.getFromRatio())
+ {
+ return null;
+ }
}
return new MapList(getFromRanges(), toRanges, outFromRatio, outToRatio);
}
/**
- * Returns the [start, end, start, end, ...] ranges in the 'from' range that
- * map to the given start-end in the 'to' range. Returns null if either
- * {@code start} or {@code end} is not a mapped 'to' range position.
+ * Returns the [start1, end1, start2, end2, ...] positions in the 'from' range
+ * that map to positions between {@code start} and {@code end} in the 'to'
+ * range. Note that for a reverse strand mapping this will return ranges with
+ * end < start. Returns null if no mapped positions are found in start-end.
*
* @param start
* @param end
* @return
*/
- public int[] locateInFrom2(int start, int end)
+ public int[] locateInFrom(int start, int end)
{
- List<int[]> ranges = mapBetween(start, end, toShifts, fromShifts,
+ return mapPositions(start, end, toShifts, fromShifts,
toRatio, fromRatio);
-
- // TODO: or just return the List and adjust calling code to match
- return ranges.isEmpty() ? null : MappingUtils.rangeListToArray(ranges);
}
/**
- * Returns the [start, end, start, end, ...] ranges in the 'to' range that map
- * to the given start-end in the 'from' range. Returns null if either
- * {@code start} or {@code end} is not a mapped 'from' range position.
+ * Returns the [start1, end1, start2, end2, ...] positions in the 'to' range
+ * that map to positions between {@code start} and {@code end} in the 'from'
+ * range. Note that for a reverse strand mapping this will return ranges with
+ * end < start. Returns null if no mapped positions are found in start-end.
*
* @param start
* @param end
* @return
*/
- public int[] locateInTo2(int start, int end)
+ public int[] locateInTo(int start, int end)
{
- List<int[]> ranges = mapBetween(start, end, fromShifts, toShifts,
+ return mapPositions(start, end, fromShifts, toShifts,
fromRatio, toRatio);
-
- return ranges.isEmpty() ? null : MappingUtils.rangeListToArray(ranges);
}
/**
- * A helper method for navigating the mapping. Returns a (possibly empty) list
- * of [start-end] positions in {@code ranges2} that map to positions in
- * {@code ranges1} between {@code start} and {@code end}.
+ * Helper method that returns the [start1, end1, start2, end2, ...] positions
+ * in {@code targetRange} that map to positions between {@code start} and
+ * {@code end} in {@code sourceRange}. Note that for a reverse strand mapping
+ * this will return ranges with end < start. Returns null if no mapped
+ * positions are found in start-end.
*
* @param start
* @param end
- * @param ranges1
- * @param ranges2
- * @param wordLength1
- * @param wordLength2
+ * @param sourceRange
+ * @param targetRange
+ * @param sourceWordLength
+ * @param targetWordLength
* @return
*/
- final static List<int[]> mapBetween(int start, int end,
- List<int[]> ranges1, List<int[]> ranges2, int wordLength1,
- int wordLength2)
+ final static int[] mapPositions(int start, int end,
+ List<int[]> sourceRange, List<int[]> targetRange,
+ int sourceWordLength, int targetWordLength)
{
- /*
- * first traverse ranges1 and record count of mapped positions
- * to any that overlap start-end
- */
- List<int[]> overlaps = findOverlapPositions(ranges1, start, end);
- if (overlaps.isEmpty())
+ if (end < start)
{
- return overlaps;
+ int tmp = end;
+ end = start;
+ start = tmp;
}
/*
- * convert positions to equivalent 'word' positions in ranges
+ * traverse sourceRange and mark offsets in targetRange
+ * of any positions that lie in [start, end]
*/
- mapWords(overlaps, wordLength1, wordLength2);
+ BitSet offsets = getMappedOffsetsForPositions(start, end, sourceRange,
+ sourceWordLength, targetWordLength);
/*
- * walk ranges2 and record the values found at
- * the offsets in 'overlaps'
+ * traverse targetRange and collect positions at the marked offsets
*/
- List<int[]> mapped = new ArrayList<>();
- final int s1 = overlaps.size();
- final int s2 = ranges2.size();
- int rangeIndex = 0;
- int rangeOffset = 0;
- int mappedCount = 0;
-
- for (int i = 0 ; i < s1 ; i++)
- {
- /*
- * for each range in overlaps, walk ranges2 and record the values
- * at the offsets, advancing rangeIndex / Offset
- */
- int [] mappedRange = ranges2.get(rangeIndex);
- int [] overlap = overlaps.get(s1);
- while (mappedCount < overlap[1])
- {
-
- }
- }
-
- return mapped;
+ List<int[]> mapped = getPositionsForOffsets(targetRange, offsets);
+
+ // TODO: or just return the List and adjust calling code to match
+ return mapped.isEmpty() ? null : MappingUtils.rangeListToArray(mapped);
}
/**
- * Converts the start-end positions (counted from zero) in the {@code ranges}
- * list from one word length to another. Start-end positions are expanded if
- * necessary to cover a whole word of length {@code wordLength1}. Positions
- * are then divided by {@code wordLength1} and multiplied by
- * {@code wordLength2} to give equivalent mapped words.
- * <p>
- * Put simply, this converts peptide residue positions to the corresponding
- * codon ranges, and codons - including partial codons - to the corresponding
- * peptide positions; for example
+ * Scans the list of {@code ranges} for any values (positions) that lie
+ * between start and end (inclusive), and records the <em>offsets</em> from
+ * the start of the list as a BitSet. The offset positions are converted to
+ * corresponding words in blocks of {@code wordLength2}.
*
* <pre>
- * [1, 10] with word lengths 3:1 converts (as if bases [0-11]) to [1, 4]
+ * For example:
+ * 1:1 (e.g. gene to CDS):
+ * ranges { [10-20], [31-40] }, wordLengthFrom = wordLength 2 = 1
+ * for start = 1, end = 9, returns a BitSet with no bits set
+ * for start = 1, end = 11, returns a BitSet with bits 0-1 set
+ * for start = 15, end = 35, returns a BitSet with bits 5-15 set
+ * 1:3 (peptide to codon):
+ * ranges { [1-200] }, wordLengthFrom = 1, wordLength 2 = 3
+ * for start = 9, end = 9, returns a BitSet with bits 24-26 set
+ * 3:1 (codon to peptide):
+ * ranges { [101-150], [171-180] }, wordLengthFrom = 3, wordLength 2 = 1
+ * for start = 101, end = 102 (partial first codon), returns a BitSet with bit 0 set
+ * for start = 150, end = 171 (partial 17th codon), returns a BitSet with bit 16 set
+ * 3:1 (circular DNA to peptide):
+ * ranges { [101-150], [21-30] }, wordLengthFrom = 3, wordLength 2 = 1
+ * for start = 24, end = 40 (spans codons 18-20), returns a BitSet with bits 17-19 set
* </pre>
*
- * @param ranges
- * @param wordLength1
- * @param wordLength2
+ * @param start
+ * @param end
+ * @param sourceRange
+ * @param sourceWordLength
+ * @param targetWordLength
* @return
*/
- final static void mapWords(List<int[]> ranges, int wordLength1,
- int wordLength2)
+ protected final static BitSet getMappedOffsetsForPositions(int start,
+ int end, List<int[]> sourceRange, int sourceWordLength, int targetWordLength)
{
- if (wordLength1 == 1 && wordLength2 == 1)
- {
- return; // nothing to do here
- }
- int s = ranges.size();
- for (int i = 0; i < s; i++)
+ BitSet overlaps = new BitSet();
+ int offset = 0;
+ final int s1 = sourceRange.size();
+ for (int i = 0; i < s1; i++)
{
- int[] range = ranges.get(i);
+ int[] range = sourceRange.get(i);
+ final int offset1 = offset;
+ int overlapStartOffset = -1;
+ int overlapEndOffset = -1;
- /*
- * expand range start to the start of a word,
- * and convert to wordLength2
- */
- range[0] -= range[0] % wordLength1;
- range[0] = range[0] / wordLength1 * wordLength2;
+ if (range[1] >= range[0])
+ {
+ /*
+ * forward direction range
+ */
+ if (start <= range[1] && end >= range[0])
+ {
+ /*
+ * overlap
+ */
+ int overlapStart = Math.max(start, range[0]);
+ overlapStartOffset = offset1 + overlapStart - range[0];
+ int overlapEnd = Math.min(end, range[1]);
+ overlapEndOffset = offset1 + overlapEnd - range[0];
+ }
+ }
+ else
+ {
+ /*
+ * reverse direction range
+ */
+ if (start <= range[0] && end >= range[1])
+ {
+ /*
+ * overlap
+ */
+ int overlapStart = Math.max(start, range[1]);
+ int overlapEnd = Math.min(end, range[0]);
+ overlapStartOffset = offset1 + range[0] - overlapEnd;
+ overlapEndOffset = offset1 + range[0] - overlapStart;
+ }
+ }
- /*
- * similar calculation for range end, adding
- * (wordLength2 - 1) for end of mapped word
- */
- range[1] -= range[1] % wordLength1;
- range[1] = range[1] / wordLength1 * wordLength2;
- range[1] += wordLength2 - 1;
+ if (overlapStartOffset > -1)
+ {
+ /*
+ * found an overlap
+ */
+ if (sourceWordLength != targetWordLength)
+ {
+ /*
+ * convert any overlap found to whole words in the target range
+ * (e.g. treat any partial codon overlap as if the whole codon)
+ */
+ overlapStartOffset -= overlapStartOffset % sourceWordLength;
+ overlapStartOffset = overlapStartOffset / sourceWordLength
+ * targetWordLength;
+
+ /*
+ * similar calculation for range end, adding
+ * (wordLength2 - 1) for end of mapped word
+ */
+ overlapEndOffset -= overlapEndOffset % sourceWordLength;
+ overlapEndOffset = overlapEndOffset / sourceWordLength
+ * targetWordLength;
+ overlapEndOffset += targetWordLength - 1;
+ }
+ overlaps.set(overlapStartOffset, overlapEndOffset + 1);
+ }
+ offset += 1 + Math.abs(range[1] - range[0]);
}
+ return overlaps;
}
/**
- * Helper method that returns a (possibly empty) list of offsets in
- * {@code ranges} to subranges that overlap {@code start-end}. The list
- * returned holds counts of the number of positions traversed (inclusive) to
- * reach the overlapping positions, not the overlapping values. Returns null
- * if there are no overlaps.
+ * Returns a (possibly empty) list of the [start-end] values (positions) at
+ * offsets in the {@code targetRange} list that are marked by 'on' bits in the
+ * {@code offsets} bitset.
*
- * @param ranges
- * @param start
- * @param end
+ * @param targetRange
+ * @param offsets
* @return
*/
- final static List<int[]> findOverlapPositions(List<int[]> ranges,
- int start, int end)
+ protected final static List<int[]> getPositionsForOffsets(
+ List<int[]> targetRange, BitSet offsets)
{
- List<int[]> positions = new ArrayList<>();
- int pos = 0;
- int s = ranges.size();
- for (int i = 0; i < s; i++)
+ List<int[]> mapped = new ArrayList<>();
+ if (offsets.isEmpty())
{
- int[] range = ranges.get(i);
- addOverlap(positions, pos, range, start, end);
- pos += 1 + Math.abs(range[1] - range[0]);
+ return mapped;
}
- return positions;
+
+ /*
+ * count of positions preceding ranges[i]
+ */
+ int traversed = 0;
+
+ /*
+ * for each [from-to] range in ranges:
+ * - find subranges (if any) at marked offsets
+ * - add the start-end values at the marked positions
+ */
+ final int toAdd = offsets.cardinality();
+ int added = 0;
+ final int s2 = targetRange.size();
+ for (int i = 0; added < toAdd && i < s2; i++)
+ {
+ int[] range = targetRange.get(i);
+ added += addOffsetPositions(mapped, traversed, range, offsets);
+ traversed += Math.abs(range[1] - range[0]) + 1;
+ }
+ return mapped;
}
/**
- * A helper method that checks whether {@code range} overlaps
- * {@code start-end}, and if so adds the positional offset of the overlap to
- * {@code positions}.
+ * Helper method that adds any start-end subranges of {@code range} that are
+ * at offsets in {@code range} marked by set bits in overlaps.
+ * {@code mapOffset} is added to {@code range} offset positions. Returns the
+ * count of positions added.
*
- * @param positions
- * a list of map offsets to add to
- * @param pos
- * the number of mapped positions already visited
+ * @param mapped
+ * @param mapOffset
* @param range
- * a from-to range (may be forward or reverse)
- * @param start
- * position to test for overlap in range
- * @param end
- * position to test for overlap in range
+ * @param overlaps
* @return
*/
- final static void addOverlap(List<int[]> positions, int pos, int[] range,
- int start, int end)
+ final static int addOffsetPositions(List<int[]> mapped,
+ final int mapOffset, final int[] range, final BitSet overlaps)
{
- if (range[1] >= range[0])
+ final int rangeLength = 1 + Math.abs(range[1] - range[0]);
+ final int step = range[1] < range[0] ? -1 : 1;
+ int offsetStart = 0; // offset into range
+ int added = 0;
+
+ while (offsetStart < rangeLength)
{
/*
- * forward direction range
+ * find the start of the next marked overlap offset;
+ * if there is none, or it is beyond range, then finished
*/
- if (start <= range[1] && end >= range[0])
+ int overlapStart = overlaps.nextSetBit(mapOffset + offsetStart);
+ if (overlapStart == -1 || overlapStart - mapOffset >= rangeLength)
{
/*
- * overlap
+ * no more overlaps, or no more within range[]
*/
- int overlapStart = Math.max(start, range[0]);
- int overlapEnd = Math.min(end, range[1]);
- positions
- .add(new int[]
- { 1 + overlapStart - range[0], 1 + overlapEnd - range[0] });
+ return added;
}
- }
- else
- {
+ overlapStart -= mapOffset;
+
/*
- * reverse direction range
+ * end of the overlap range is just before the next clear bit;
+ * restrict it to end of range if necessary;
+ * note we may add a reverse strand range here (end < start)
*/
- if (start <= range[0] && end >= range[1])
- {
- /*
- * overlap
- */
- int overlapStart = Math.max(start, range[1]);
- int overlapEnd = Math.min(end, range[0]);
- positions
- .add(new int[]
- { 1 + range[0] - overlapStart, 1 + range[0] - overlapEnd });
- }
+ int overlapEnd = overlaps.nextClearBit(mapOffset + overlapStart + 1);
+ overlapEnd = (overlapEnd == -1) ? rangeLength - 1
+ : Math.min(rangeLength - 1, overlapEnd - mapOffset - 1);
+ int startPosition = range[0] + step * overlapStart;
+ int endPosition = range[0] + step * overlapEnd;
+ mapped.add(new int[] { startPosition, endPosition });
+ offsetStart = overlapEnd + 1;
+ added += Math.abs(endPosition - startPosition) + 1;
}
+
+ return added;
}
}