private SequenceFeatures sequenceFeatureStore;
+ /*
+ * A cursor holding the approximate current view position to the sequence,
+ * as determined by findIndex or findPosition or findPositions.
+ * Using a cursor as a hint allows these methods to be more performant for
+ * large sequences.
+ */
+ private SequenceCursor cursor;
+
+ /*
+ * A number that should be incremented whenever the sequence is edited.
+ * If the value matches the cursor token, then we can trust the cursor,
+ * if not then it should be recomputed.
+ */
+ private int changeCount;
+
/**
* Creates a new Sequence object.
*
{
this.sequence = seq.toCharArray();
checkValidRange();
+ sequenceChanged();
}
@Override
return this.description;
}
- /*
- * (non-Javadoc)
- *
- * @see jalview.datamodel.SequenceI#findIndex(int)
+ /**
+ * {@inheritDoc}
*/
@Override
public int findIndex(int pos)
{
- // returns the alignment position for a residue
+ /*
+ * use a valid, hopefully nearby, cursor if available
+ */
+ if (isValidCursor(cursor))
+ {
+ return findIndex(pos, cursor);
+ }
+
int j = start;
int i = 0;
// Rely on end being at least as long as the length of the sequence.
{
j++;
}
-
i++;
}
- if ((j == end) && (j < pos))
+ if (j == end && j < pos)
{
return end + 1;
}
- else
+
+ updateCursor(pos, i);
+ return i;
+ }
+
+ /**
+ * Updates the cursor to the latest found residue and column position
+ *
+ * @param residuePos
+ * (start..)
+ * @param column
+ * (1..)
+ */
+ protected void updateCursor(int residuePos, int column)
+ {
+ cursor = new SequenceCursor(this, residuePos, column, this.changeCount);
+ }
+
+ /**
+ * Answers the aligned column position (1..) for the given residue position
+ * (start..) given a 'hint' of a residue/column location in the neighbourhood.
+ * The hint may be left of, at, or to the right of the required position.
+ *
+ * @param pos
+ * @param curs
+ * @return
+ */
+ protected int findIndex(int pos, SequenceCursor curs)
+ {
+ if (!isValidCursor(curs))
+ {
+ /*
+ * wrong or invalidated cursor, compute de novo
+ */
+ return findIndex(pos);
+ }
+
+ if (curs.residuePosition == pos)
+ {
+ return curs.columnPosition;
+ }
+
+ /*
+ * move left or right to find pos from hint.position
+ */
+ int col = curs.columnPosition - 1; // convert from base 1 to 0-based array
+ // index
+ int newPos = curs.residuePosition;
+ int delta = newPos > pos ? -1 : 1;
+
+ while (newPos != pos)
{
- return i;
+ col += delta; // shift one column left or right
+ if (col < 0 || col == sequence.length)
+ {
+ break;
+ }
+ if (!Comparison.isGap(sequence[col]))
+ {
+ newPos += delta;
+ }
}
+
+ col++; // convert back to base 1
+ updateCursor(pos, col);
+
+ return col;
}
+ /**
+ * {@inheritDoc}
+ */
@Override
- public int findPosition(int i)
+ public int findPosition(final int column)
{
+ /*
+ * use a valid, hopefully nearby, cursor if available
+ */
+ if (isValidCursor(cursor))
+ {
+ return findPosition(column + 1, cursor);
+ }
+
+ // TODO recode this more naturally i.e. count residues only
+ // as they are found, not 'in anticipation'
+
+ int lastPosFound = 0;
+ int lastPosFoundColumn = 0;
+ int seqlen = sequence.length;
+ if (seqlen > 0 && !Comparison.isGap(sequence[0]))
+ {
+ lastPosFound = start;
+ lastPosFoundColumn = 0;
+ }
+
int j = 0;
int pos = start;
- int seqlen = sequence.length;
- while ((j < i) && (j < seqlen))
+
+ while (j < column && j < seqlen)
{
if (!Comparison.isGap(sequence[j]))
{
+ lastPosFound = pos;
+ lastPosFoundColumn = j;
pos++;
}
-
j++;
}
+ if (j < seqlen && !Comparison.isGap(sequence[j]))
+ {
+ lastPosFound = pos;
+ lastPosFoundColumn = j;
+ }
+
+ /*
+ * update the cursor to the last residue position found (if any)
+ * (converting column position to base 1)
+ */
+ if (lastPosFound != 0)
+ {
+ updateCursor(lastPosFound, lastPosFoundColumn + 1);
+ }
return pos;
}
/**
+ * Answers true if the given cursor is not null, is for this sequence object,
+ * and has a token value that matches this object's changeCount, else false.
+ * This allows us to ignore a cursor as 'stale' if the sequence has been
+ * modified since the cursor was created.
+ *
+ * @param curs
+ * @return
+ */
+ protected boolean isValidCursor(SequenceCursor curs)
+ {
+ if (curs == null || curs.sequence != this || curs.token != changeCount)
+ {
+ return false;
+ }
+ /*
+ * sanity check against range
+ */
+ if (curs.columnPosition < 0 || curs.columnPosition >= sequence.length)
+ {
+ return false;
+ }
+ if (curs.residuePosition < start || curs.residuePosition > end)
+ {
+ return false;
+ }
+ return true;
+ }
+
+ /**
+ * Answers the sequence position (start..) for the given aligned column
+ * position (1..), given a hint of a cursor in the neighbourhood. The cursor
+ * may lie left of, at, or to the right of the column position.
+ *
+ * @param col
+ * @param curs
+ * @return
+ */
+ protected int findPosition(final int col, SequenceCursor curs)
+ {
+ if (!isValidCursor(curs))
+ {
+ /*
+ * wrong or invalidated cursor, compute de novo
+ */
+ return findPosition(col - 1);// ugh back to base 0
+ }
+
+ if (curs.columnPosition == col)
+ {
+ cursor = curs; // in case this method becomes public
+ return curs.residuePosition; // easy case :-)
+ }
+
+ /*
+ * move left or right to find pos from cursor position
+ */
+ int column = curs.columnPosition - 1; // to base 0
+ int newPos = curs.residuePosition;
+ int delta = curs.columnPosition > col ? -1 : 1;
+ boolean gapped = false;
+ int lastFoundPosition = curs.residuePosition;
+ int lastFoundPositionColumn = curs.columnPosition;
+
+ while (column != col - 1)
+ {
+ column += delta; // shift one column left or right
+ if (column < 0 || column == sequence.length)
+ {
+ break;
+ }
+ gapped = Comparison.isGap(sequence[column]);
+ if (!gapped)
+ {
+ newPos += delta;
+ lastFoundPosition = newPos;
+ lastFoundPositionColumn = column + 1;
+ }
+ }
+
+ if (cursor == null || lastFoundPosition != cursor.residuePosition)
+ {
+ updateCursor(lastFoundPosition, lastFoundPositionColumn);
+ }
+
+ /*
+ * hack to give position to the right if on a gap
+ * or beyond the length of the sequence (see JAL-2562)
+ */
+ if (delta > 0 && (gapped || column >= sequence.length))
+ {
+ newPos++;
+ }
+
+ return newPos;
+ }
+
+ /**
* {@inheritDoc}
*/
@Override
public Range findPositions(int fromCol, int toCol)
{
+ if (cursor != null && cursor.sequence == this
+ && cursor.token == changeCount)
+ {
+ return findPositions(fromCol, toCol, cursor);
+ }
+
/*
* count residues before fromCol
*/
*/
int firstPos = 0;
int lastPos = 0;
- int firstPosCol = 0;
boolean foundFirst = false;
while (j <= toCol && j < seqlen)
if (!foundFirst)
{
firstPos = count;
- firstPosCol = j;
foundFirst = true;
}
lastPos = count;
}
/**
+ * Returns the range of sequence positions included in the given alignment
+ * position range. If no positions are included (the range is entirely gaps),
+ * then returns null. The cursor parameter may provide a starting position in
+ * the neighbourhood of the search (which may be left of, right of, or
+ * overlapping the search region).
+ *
+ * @param fromCol
+ * start column of region (0..)
+ * @param toCol
+ * end column of region (0..)
+ * @param curs
+ * @return
+ */
+ protected Range findPositions(int fromCol, int toCol, SequenceCursor curs)
+ {
+ if (!isValidCursor(curs))
+ {
+ /*
+ * wrong or invalidated cursor, compute de novo
+ */
+ return findPositions(fromCol, toCol);
+ }
+
+ /*
+ * keep this simple...first step from cursor to fromCol...
+ */
+ final int seqlen = sequence.length;
+ int resNo = curs.residuePosition;
+ int col = curs.columnPosition - 1; // from base 1 to base 0
+ if (col != fromCol)
+ {
+ int delta = col > fromCol ? -1 : 1;
+ while (col != fromCol && col >= 0 && col < seqlen)
+ {
+ if (!Comparison.isGap(sequence[col]))
+ {
+ resNo += delta;
+ }
+ col += delta;
+ }
+ }
+
+ if (col < fromCol || col == seqlen)
+ {
+ /*
+ * sequence lies to the left of the target region
+ */
+ return null;
+ }
+
+ /*
+ * resNo is now the residue at fromCol (if not gapped), else the one
+ * before it (if delta == 1), else the one after (if delta == -1);
+ * we want the residue before fromCol
+ */
+ if (!Comparison.isGap(sequence[fromCol]))
+ {
+ resNo--;
+ }
+ else if (curs.columnPosition > fromCol)
+ {
+ resNo -= 2;
+ }
+
+ /*
+ * now first and last residues between fromCol and toCol
+ */
+ int firstPos = 0;
+ int lastPos = 0;
+ boolean foundFirst = false;
+
+ while (col <= toCol && col < seqlen)
+ {
+ if (!Comparison.isGap(sequence[col]))
+ {
+ resNo++;
+ if (!foundFirst)
+ {
+ firstPos = resNo;
+ foundFirst = true;
+ }
+ lastPos = resNo;
+ }
+ col++;
+ }
+
+ if (firstPos == 0)
+ {
+ /*
+ * no residues in this range
+ */
+ return null;
+ }
+
+ return new Range(firstPos, lastPos);
+ }
+
+ /**
* Returns an int array where indices correspond to each residue in the
* sequence and the element value gives its position in the alignment
*
start = newstart;
end = newend;
sequence = tmp;
+ sequenceChanged();
}
@Override
}
sequence = tmp;
+ sequenceChanged();
}
@Override
}
return sequenceFeatureStore.findFeatures(from, to, types);
}
+
+ /**
+ * Invalidates any stale cursors (forcing recalculation) by incrementing the
+ * token that has to match the one presented by the cursor
+ */
+ @Override
+ public void sequenceChanged()
+ {
+ changeCount++;
+ }
}