amogh-jahagirdar commented on code in PR #4578:
URL: https://github.com/apache/iceberg/pull/4578#discussion_r885240699
##########
core/src/main/java/org/apache/iceberg/RemoveSnapshots.java:
##########
@@ -161,21 +175,120 @@ public List<Snapshot> apply() {
private TableMetadata internalApply() {
this.base = ops.refresh();
+ if (base.snapshots().isEmpty()) {
+ return base;
+ }
Set<Long> idsToRetain = Sets.newHashSet();
- List<Long> ancestorIds = SnapshotUtil.ancestorIds(base.currentSnapshot(),
base::snapshot);
- if (minNumSnapshots >= ancestorIds.size()) {
- idsToRetain.addAll(ancestorIds);
- } else {
- idsToRetain.addAll(ancestorIds.subList(0, minNumSnapshots));
+
+ // Identify dangling snapshots which should be retained
+ Set<Long> danglingSnapshotsToRetain =
computeDanglingSnapshotsToRetain(base.refs().values());
+ idsToRetain.addAll(danglingSnapshotsToRetain);
+
+ // Identify refs that should be removed
+ Map<String, SnapshotRef> retainedRefs = computeRetainedRefs(base.refs());
+ Map<Long, List<String>> retainedIdToRefs = Maps.newHashMap();
+ for (Map.Entry<String, SnapshotRef> retainedRefEntry :
retainedRefs.entrySet()) {
+ long snapshotId = retainedRefEntry.getValue().snapshotId();
+ retainedIdToRefs.putIfAbsent(snapshotId, Lists.newArrayList());
+ retainedIdToRefs.get(snapshotId).add(retainedRefEntry.getKey());
+ idsToRetain.add(snapshotId);
+ }
+
+ for (long idToRemove : idsToRemove) {
+ List<String> refsForId = retainedIdToRefs.get(idToRemove);
+ Preconditions.checkArgument(refsForId == null,
+ "Cannot expire %s. Still referenced by refs: %s", idToRemove,
refsForId);
+ }
+
+ Set<Long> branchSnapshotsToRetain =
computeAllBranchSnapshotsToRetain(retainedRefs.values());
+ idsToRetain.addAll(branchSnapshotsToRetain);
+ TableMetadata.Builder updatedMetaBuilder = TableMetadata.buildFrom(base);
+
+ base.snapshots().stream()
+ .map(Snapshot::snapshotId)
+ .filter(snapshot -> !idsToRetain.contains(snapshot))
+ .forEach(idsToRemove::add);
+ updatedMetaBuilder.removeSnapshots(idsToRemove);
+
+ base.refs().keySet().stream()
+ .filter(ref -> !retainedRefs.containsKey(ref))
+ .forEach(updatedMetaBuilder::removeRef);
+
+ return updatedMetaBuilder.build();
+ }
+
+ private Map<String, SnapshotRef> computeRetainedRefs(Map<String,
SnapshotRef> refs) {
+ Map<String, SnapshotRef> retainedRefs = Maps.newHashMap();
+ for (Map.Entry<String, SnapshotRef> refEntry : refs.entrySet()) {
+ String name = refEntry.getKey();
+ SnapshotRef ref = refEntry.getValue();
+ if (name.equals(SnapshotRef.MAIN_BRANCH)) {
+ retainedRefs.put(name, ref);
+ continue;
+ }
+
+ Snapshot snapshot = base.snapshot(ref.snapshotId());
+ long maxRefAgeMs = ref.maxRefAgeMs() != null ? ref.maxRefAgeMs() :
defaultMaxRefAgeMs;
+ if (snapshot != null) {
+ long refAgeMs = now - snapshot.timestampMillis();
+ if (refAgeMs <= maxRefAgeMs) {
+ retainedRefs.put(name, ref);
+ }
+ } else {
+ LOG.warn("Removing invalid ref {}: snapshot {} does not exist", name,
ref.snapshotId());
+ }
+ }
+
+ return retainedRefs;
+ }
+
+ private Set<Long> computeAllBranchSnapshotsToRetain(Collection<SnapshotRef>
refs) {
+ Set<Long> branchSnapshotsToRetain = Sets.newHashSet();
+ for (SnapshotRef ref : refs) {
+ if (ref.isBranch()) {
+ long expireSnapshotsOlderThan = ref.maxSnapshotAgeMs() != null ? now -
ref.maxSnapshotAgeMs() :
+ defaultExpireOlderThan;
+ int minSnapshotsToKeep = ref.minSnapshotsToKeep() != null ?
ref.minSnapshotsToKeep() :
+ defaultMinNumSnapshots;
+ branchSnapshotsToRetain.addAll(
+ computeBranchSnapshotsToRetain(ref.snapshotId(),
expireSnapshotsOlderThan, minSnapshotsToKeep));
+ }
+ }
+
+ return branchSnapshotsToRetain;
+ }
+
+ private Set<Long> computeDanglingSnapshotsToRetain(Collection<SnapshotRef>
refs) {
+ Set<Long> snapshotsPartOfBranches = Sets.newHashSet();
+ for (SnapshotRef ref : refs) {
+ if (ref.isBranch()) {
+ Iterable<Snapshot> branchSnapshots =
SnapshotUtil.ancestorsOf(ref.snapshotId(), base::snapshot);
+ branchSnapshots.forEach(snapshot ->
snapshotsPartOfBranches.add(snapshot.snapshotId()));
Review Comment:
We can be more efficient (time wise, at the tradeoff of memory) by
maintaining the mapping between a branch and all the ancestors for the branch
so that it can be reused across both the dangling snapshots and the branch
snapshots to retain computation. Otherwise we have to get the ancestors
multiple times.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
