compact method
Compact the commit log. If dryRun is true, yields each
commit-id that WOULD be removed without performing the
deletion. If false, performs the compaction and yields each
commit-id as it is removed.
The Hive impl prunes duplicate entries — same atKey, older
commitId. Under the one-entry-per-atKey invariant (enforced
inline on commit and by the startup dedup migration) that set
is empty in practice, so this is a legacy-data safety net.
SQL backends will do a DELETE WHERE (and possibly VACUUM)
instead.
Implementation
@override
Stream<int> compact(bool dryRun) async* {
// One row per atKey by construction — there are no duplicate entries
// to prune, so compaction is a no-op on this backend.
}