[
https://issues.apache.org/jira/browse/DRILL-4166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15046010#comment-15046010
]
Victoria Markman commented on DRILL-4166:
-----------------------------------------
Now that I'm thinking about it, technically this is wrong result.
> Inconsistent return type from COALESCE function with views and parquet data
> sources
> ------------------------------------------------------------------------------------
>
> Key: DRILL-4166
> URL: https://issues.apache.org/jira/browse/DRILL-4166
> Project: Apache Drill
> Issue Type: Bug
> Components: Execution - Data Types
> Affects Versions: 1.4.0
> Reporter: Victoria Markman
>
> j1 is a parquet file that is defined as following:
> {code}
> [Mon Dec 07 15:06:58] # ~/parquet-tools/parquet-schema 0_0_0.parquet
>
> message root {
> optional binary c_varchar (UTF8);
> optional int32 c_integer;
> optional int64 c_bigint;
> optional float c_float;
> optional double c_double;
> optional int32 c_date (DATE);
> optional int32 c_time (TIME_MILLIS);
> optional int64 c_timestamp (TIMESTAMP_MILLIS);
> optional boolean c_boolean;
> optional double d9;
> optional double d18;
> optional double d28;
> optional double d38;
> }
> {code}
> COALESCE function with input arguments date and timestamp returns timestamp
> data type as expected:
> {code}
> 0: jdbc:drill:schema=dfs> create table x as select coalesce(c_date,
> c_timestamp) from j1 limit 1;
> +-----------+----------------------------+
> | Fragment | Number of records written |
> +-----------+----------------------------+
> | 0_0 | 1 |
> +-----------+----------------------------+
> 1 row selected (0.685 seconds)
> [Mon Dec 07 15:06:14] # ~/parquet-tools/parquet-schema 0_0_0.parquet
>
> message root {
> optional int64 EXPR$0 (TIMESTAMP_MILLIS);
> }
> {code}
> However, if you define a view on top of this parquet file and run the same
> query, return type is going to be DATE and not TIMESTAMP:
> {code}
> 0: jdbc:drill:schema=dfs> create view v1 as select cast(c_date as date) as
> c_date, cast(c_timestamp as timestamp) as c_timestamp from j1 limit 1;
> +-------+------------------------------------------------------------+
> | ok | summary |
> +-------+------------------------------------------------------------+
> | true | View 'v1' created successfully in 'dfs.subqueries' schema |
> +-------+------------------------------------------------------------+
> 1 row selected (0.368 seconds)
> 0: jdbc:drill:schema=dfs> describe v1;
> +--------------+------------+--------------+
> | COLUMN_NAME | DATA_TYPE | IS_NULLABLE |
> +--------------+------------+--------------+
> | c_date | DATE | YES |
> | c_timestamp | TIMESTAMP | YES |
> +--------------+------------+--------------+
> 2 rows selected (0.518 seconds)
> [Mon Dec 07 15:11:50] # ~/parquet-tools/parquet-schema 0_0_0.parquet
>
> message root {
> optional int32 EXPR$0 (DATE);
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
