Ádám Szita created HIVE-26745:
---------------------------------
Summary: HPL unable to handle Decimal or null values in hplsql mode
Key: HIVE-26745
URL: https://issues.apache.org/jira/browse/HIVE-26745
Project: Hive
Issue Type: Bug
Reporter: Ádám Szita
Assignee: Ádám Szita
Decimal or null values coming from Hive service side are not handled properly
in HPL.
When we're using beeline in HPL mode then the results of queries such as SELECT
CAST are returned in RowSet objects directly from HS2 code. This is due to
HplSqlQueryExecutor expecting RowSet type in OperationRowResult. For this
RowSet interface, a ColumnBasedSet class brings the imlementation which holds a
ColumnBuffer list to hold the result data.
The internal representation however cannot hold BigDecimal values, as seen
here:
https://github.com/apache/hive/blob/master/serde/src/java/org/apache/hadoop/hive/serde2/thrift/ColumnBuffer.java#L396
so while the SELECT CAST query will produce a HiveBigDecimal instance, it will
be converted to String at
https://github.com/apache/hive/blob/master/service/src/java/org/apache/hive/service/cli/ColumnBasedSet.java#L110-L112
before we add this piece of data to the result set.
Further down the line we will see class java.lang.String cannot be casted to
class java.math.BigDecimal.
ColumnBuffer seems like a pretty commonly used code and I don't think we would
want to try and extend this with the ability the carry decimal types. Perhaps
we could create a Jira to recreate the BigDecimal value from the String
representation arriving on the HPL side.
Since the two issues are both about value conversion I'm aiming to fix these in
one go.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
