[
https://issues.apache.org/jira/browse/PHOENIX-4882?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16599066#comment-16599066
]
Thomas D'Silva commented on PHOENIX-4882:
-----------------------------------------
We can probably be smarter about resolving the data table for columns that
aren't present in the local Index. I also see the data table being resolved
from IndexExpressionParseNodeRewriter. We also have this TODO in BaseQueryPlan
{code}
// TODO: is it necessary to re-resolve the table?
TableRef dataTableRef =
FromCompiler.getResolver(
FACTORY.namedTable(null,
TableName.create(parentSchemaName, parentTableName)),
context.getConnection()).resolveTable(parentSchemaName, parentTableName);
PTable dataTable = dataTableRef.getTable();
{code}
> The client re-resolves the table for every projected non-indexed column when
> there's a local index.
> ---------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-4882
> URL: https://issues.apache.org/jira/browse/PHOENIX-4882
> Project: Phoenix
> Issue Type: Improvement
> Reporter: Lars Hofhansl
> Priority: Major
>
> See TupleProjectionCompiler.ColumnRefVisitor.
> Each non-projected column causes a server roundtrip in
> LocalIndexDataColumnRef.
> For wide tables that can be hundreds of time.
> I generally find that the same table is resolved over and over again for
> other reasons as well (I counted 15-25 times during query compilation!!)
> Put a break point in ConnectionQueryServicesImpl.getTable and you'll see.
> [~tdsilva]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
