maedhroz commented on code in PR #3649:
URL: https://github.com/apache/cassandra/pull/3649#discussion_r1833407074
##########
src/java/org/apache/cassandra/index/sai/plan/StorageAttachedIndexSearcher.java:
##########
@@ -400,36 +430,43 @@ private UnfilteredRowIterator applyIndexFilter(PrimaryKey
key, UnfilteredRowIter
{
queryContext.rowsFiltered++;
- if (tree.isSatisfiedBy(partition.partitionKey(), (Row)
unfiltered, staticRow))
+ if (tree.isSatisfiedBy(partitionKey, (Row) unfiltered,
staticRow))
{
- matchingRows.add(unfiltered);
+ matches.add(unfiltered);
hasMatch = true;
+
+ if (topK)
+ {
+ PrimaryKey shadowed =
keyFactory.hasClusteringColumns()
+ ?
keyFactory.create(partitionKey, ((Row) unfiltered).clustering())
+ :
keyFactory.create(partitionKey);
+ keysToShadow.remove(shadowed);
+ }
}
}
}
- if (!hasMatch)
+ if (topK)
{
- queryContext.rowsFiltered++;
+ // If any rows match the filter, there should be no need to
shadow the static primary key:
+ if (hasMatch && keyFactory.hasClusteringColumns())
+ keysToShadow.remove(keyFactory.create(partitionKey,
Clustering.STATIC_CLUSTERING));
- if (tree.isSatisfiedBy(key.partitionKey(), staticRow,
staticRow))
- hasMatch = true;
+ // Record primary keys shadowed by expired TTLs, row
tombstones, or range tombstones:
+ if (!keysToShadow.isEmpty())
+
queryContext.vectorContext().recordShadowedPrimaryKeys(keysToShadow);
Review Comment:
@jasonstack It looks like `VectorUpdateDeleteTest` does touch every codepath
in this method. Indeed, it did break down a couple times during my refactor.
What I've done here is loosely inspired by your suggestion, albeit with some
protections to avoid allocations in the non-topK case.
I've also removed what really does look like an unnecessary second round of
`isSatisfiedBy()` calls using just the static row. Before CASSANDRA-19034, this
actually happened before the `while` loop that filters the rows, but it was
never strictly necessary. I suppose I'll see if CI disagrees, but I think this
make sense.
WDYT?
--
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]
