You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@drill.apache.org by "Krystal (JIRA)" <ji...@apache.org> on 2017/03/27 23:25:41 UTC
[jira] [Created] (DRILL-5389) select 2 int96 using
convert_from(col, 'TIMESTAMP_IMPALA') function fails
Krystal created DRILL-5389:
------------------------------
Summary: select 2 int96 using convert_from(col, 'TIMESTAMP_IMPALA') function fails
Key: DRILL-5389
URL: https://issues.apache.org/jira/browse/DRILL-5389
Project: Apache Drill
Issue Type: Bug
Components: Storage - Parquet
Affects Versions: 1.9.0, 1.10.0
Reporter: Krystal
I have a table containing 2 int96 time stamp columns. If I select one column at a time, it works.
select convert_from(create_timestamp1, 'TIMESTAMP_IMPALA') from dfs.`/user/hive/warehouse/hive1_parquet` where voter_id=3;
+------------------------+
| EXPR$0 |
+------------------------+
| 2017-04-14 02:27:55.0 |
+------------------------+
select convert_from(create_timestamp2, 'TIMESTAMP_IMPALA') from dfs.`/user/hive/warehouse/hive1_parquet` where voter_id=3;
+------------------------+
| EXPR$0 |
+------------------------+
| 2017-05-30 19:30:11.0 |
+------------------------+
However, if I include both columns on the same select, it fails:
select convert_from(create_timestamp1, 'TIMESTAMP_IMPALA'), convert_from(create_timestamp2, 'TIMESTAMP_IMPALA') from dfs.`/user/hive/warehouse/hive1_parquet` where voter_id=3;
Error: SYSTEM ERROR: ArrayIndexOutOfBoundsException: 0
This is reproducible in drill-1.9 also.
In drill-1.10, setting store.parquet.reader.int96_as_timestamp`=true, the same query works fine.
select create_timestamp1,create_timestamp2 from dfs.`/user/hive/warehouse/hive1_parquet` where voter_id=3;
+------------------------+------------------------+
| create_timestamp1 | create_timestamp2 |
+------------------------+------------------------+
| 2017-04-14 02:27:55.0 | 2017-05-30 19:30:11.0 |
+------------------------+------------------------+
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)