/**
* @param buckets list of buckets, sorted from newest to oldest, from which to return the newest
* bucket within thresholds.
* @param minThreshold minimum number of sstables in a bucket to qualify.
* @param maxThreshold maximum number of sstables to compact at once (the returned bucket will be
* trimmed down to this).
* @return a bucket (list) of sstables to compact.
*/
@VisibleForTesting
static List<SSTableReader> newestBucket(
List<List<SSTableReader>> buckets,
int minThreshold,
int maxThreshold,
long now,
long baseTime) {
// If the "incoming window" has at least minThreshold SSTables, choose that one.
// For any other bucket, at least 2 SSTables is enough.
// In any case, limit to maxThreshold SSTables.
Target incomingWindow = getInitialTarget(now, baseTime);
for (List<SSTableReader> bucket : buckets) {
if (bucket.size() >= minThreshold
|| (bucket.size() >= 2 && !incomingWindow.onTarget(bucket.get(0).getMinTimestamp())))
return trimToThreshold(bucket, maxThreshold);
}
return Collections.emptyList();
}
/**
* Group files with similar min timestamp into buckets. Files with recent min timestamps are
* grouped together into buckets designated to short timespans while files with older timestamps
* are grouped into buckets representing longer timespans.
*
* @param files pairs consisting of a file and its min timestamp
* @param timeUnit
* @param base
* @param now
* @return a list of buckets of files. The list is ordered such that the files with newest
* timestamps come first. Each bucket is also a list of files ordered from newest to oldest.
*/
@VisibleForTesting
static <T> List<List<T>> getBuckets(
Collection<Pair<T, Long>> files, long timeUnit, int base, long now) {
// Sort files by age. Newest first.
final List<Pair<T, Long>> sortedFiles = Lists.newArrayList(files);
Collections.sort(
sortedFiles,
Collections.reverseOrder(
new Comparator<Pair<T, Long>>() {
public int compare(Pair<T, Long> p1, Pair<T, Long> p2) {
return p1.right.compareTo(p2.right);
}
}));
List<List<T>> buckets = Lists.newArrayList();
Target target = getInitialTarget(now, timeUnit);
PeekingIterator<Pair<T, Long>> it = Iterators.peekingIterator(sortedFiles.iterator());
outerLoop:
while (it.hasNext()) {
while (!target.onTarget(it.peek().right)) {
// If the file is too new for the target, skip it.
if (target.compareToTimestamp(it.peek().right) < 0) {
it.next();
if (!it.hasNext()) break outerLoop;
} else // If the file is too old for the target, switch targets.
target = target.nextTarget(base);
}
List<T> bucket = Lists.newArrayList();
while (target.onTarget(it.peek().right)) {
bucket.add(it.next().left);
if (!it.hasNext()) break;
}
buckets.add(bucket);
}
return buckets;
}