michaeljmarshall commented on code in PR #4353:
URL: https://github.com/apache/cassandra/pull/4353#discussion_r2683911485
##########
src/java/org/apache/cassandra/index/sai/StorageAttachedIndexGroup.java:
##########
@@ -173,6 +173,15 @@ public Index.Indexer indexerFor(Predicate<Index>
indexSelector,
return indexers.isEmpty() ? null : new Index.Indexer()
{
+ @Override
+ public void begin()
+ {
+ // TODO is this a good solution? I only need this because we
do not always
+ // indexed everything in insertRow and updateRow.
+ for (Index.Indexer indexer : indexers)
+ indexer.begin();
Review Comment:
In an alternate version, I implemented these other methods:
```java
@Override
public void partitionDelete(DeletionTime deletionTime)
{
// Initialize the memtable index to ensure proper SAI views of
the data
indexContext.initializeMemtableIndex(mt);
}
@Override
public void rangeTombstone(RangeTombstone tombstone)
{
// Initialize the memtable index to ensure proper SAI views of
the data
indexContext.initializeMemtableIndex(mt);
}
@Override
public void removeRow(Row row)
{
// Initialize the memtable index to ensure proper SAI views of
the data
indexContext.initializeMemtableIndex(mt);
}
```
the issue here is making sure that the `View` object has a reference to all
relevant memtables. I don't remember why I didn't use this logic when I first
wrote this patch in August, but I think that might be better since it covers
the actual missing cases without adding overhead to the insert/update paths.
--
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]
