Konstantin Orlov created IGNITE-28453:
-----------------------------------------
Summary: Avoid index key materialization during index row
verification
Key: IGNITE-28453
URL: https://issues.apache.org/jira/browse/IGNITE-28453
Project: Ignite
Issue Type: Improvement
Components: secondary indexes ai3
Reporter: Konstantin Orlov
Assignee: Konstantin Orlov
At the moment, in order to make sure {{IndexRow}} we've just retrieved from
storage is not leftover from transaction rollback, we extract index key once
again. The problem is that we materialize BinaryTuple just to do byte
comparison, and then throw it away:
{code}
private static boolean indexRowMatches(IndexRow indexRow, BinaryRow
binaryRow, TableSchemaAwareIndexStorage schemaAwareIndexStorage) {
BinaryTuple actualIndexRow =
schemaAwareIndexStorage.indexRowResolver().extractColumns(binaryRow);
return
indexRow.indexColumns().byteBuffer().equals(actualIndexRow.byteBuffer());
}
{code}
It would be better to implement byte comparison avoiding tuple materialization.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
