Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
1.14.0
-
None
-
None
Description
MergingRecordBatch produces OK_NEW_SCHEMA as part of first output batch post buildSchema phase. This is not necessary since in BuildSchema phase it already has the batch with schema and it always produces container of type SV NONE as well. Because of this behavior query like below doesn't work since there is HashAgg downstream to MergingReceiver which doesn't handle OK_NEW_SCHEMA post buildSchema phase.
SELECT cust.c_custkey, cust.c_name, sum(cust.c_acctbal) FROM (SELECT customer.c_custkey, customer.c_name, customer.c_acctbal FROM dfs.singleJson.`customer.json` customer ORDER BY customer.c_name) cust GROUP BY cust.c_custkey, cust.c_name