blambov commented on code in PR #4300:
URL: https://github.com/apache/cassandra/pull/4300#discussion_r2276249630
##########
src/java/org/apache/cassandra/db/compaction/CompactionTask.java:
##########
@@ -102,6 +110,43 @@ public boolean
reduceScopeForLimitedSpace(Set<SSTableReader> nonExpiredSSTables,
return false;
}
+ private void
maybeNotifyIndexersAboutRowsInFullyExpiredSSTables(Set<SSTableReader>
fullyExpiredSSTables)
+ {
+ if (fullyExpiredSSTables.isEmpty())
+ return;
+
+ List<Index> indexes = cfs.indexManager.listIndexes()
+ .stream()
+
.filter(Index::notifyIndexerAboutRowsInFullyExpiredSSTables)
+ .collect(Collectors.toList());
+
+ if (indexes.isEmpty())
+ return;
+
+ for (SSTableReader expiredSSTable : fullyExpiredSSTables)
+ {
+ expiredSSTable.getScanner()
+ .forEachRemaining(partition ->
+ {
+
partition.forEachRemaining(unfiltered -> {
Review Comment:
We definitely don't want to materialize the whole partition, but the number
rows in the partition should not matter here. The number of "versions", i.e.
sources for each merged row, in this case is always one (only the obsoleted
sstable). Thus we should be able to move the transaction initialization as it
is to the per-partition section.
As the row is deleted, something like
```
transaction.onRowMerge(null, (Row) unfiltered);
```
should be the correct usage.
If this doesn't work, let's not spend any more time on it and go back to
your original implementation.
--
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]
