/**
* Compute the difference between original and revised sequences.
*
* @param orig The original sequence.
* @param rev The revised sequence to be compared with the original.
* @return A Revision object describing the differences.
* @throws DifferenciationFailedException if the diff could not be computed.
*/
public Revision diff(Object[] orig, Object[] rev) throws DifferentiationFailedException {
// create map eqs, such that for each item in both orig and rev
// eqs(item) = firstOccurrence(item, orig);
Map eqs = buildEqSet(orig, rev);
// create an array such that
// indx[i] = NOT_FOUND_i if orig[i] is not in rev
// indx[i] = firstOccurrence(orig[i], orig)
int[] indx = buildIndex(eqs, orig, NOT_FOUND_i);
// create an array such that
// jndx[j] = NOT_FOUND_j if orig[j] is not in rev
// jndx[j] = firstOccurrence(rev[j], orig)
int[] jndx = buildIndex(eqs, rev, NOT_FOUND_j);
// what in effect has been done is to build a unique hash
// for each item that is in both orig and rev
// and to label each item in orig and new with that hash value
// or a marker that the item is not common to both.
eqs = null; // let gc know we're done with this
Revision deltas = new Revision(); // !!! new Revision()
int i = 0;
int j = 0;
// skip matching
// skip leading items that are equal
// could be written
// for (i=0; indx[i] != EOS && indx[i] == jndx[i]; i++);
// j = i;
for (; indx[i] != EOS && indx[i] == jndx[j]; i++, j++) {
/* void */
}
while (indx[i] != jndx[j]) { // only equal if both == EOS
// they are different
int ia = i;
int ja = j;
// size of this delta
do {
// look down rev for a match
// stop at a match
// or if the FO(rev[j]) > FO(orig[i])
// or at the end
while (jndx[j] < 0 || jndx[j] < indx[i]) {
j++;
}
// look down orig for a match
// stop at a match
// or if the FO(orig[i]) > FO(rev[j])
// or at the end
while (indx[i] < 0 || indx[i] < jndx[j]) {
i++;
}
// this doesn't do a compare each line with each other line
// so it won't find all matching lines
} while (indx[i] != jndx[j]);
// on exit we have a match
// they are equal, reverse any exedent matches
// it is possible to overshoot, so count back matching items
while (i > ia && j > ja && indx[i - 1] == jndx[j - 1]) {
--i;
--j;
}
deltas.addDelta(Delta.newDelta(new Chunk(orig, ia, i - ia), new Chunk(rev, ja, j - ja)));
// skip matching
for (; indx[i] != EOS && indx[i] == jndx[j]; i++, j++) {
/* void */
}
}
return deltas;
}
