[jira] [Commented] (DRILL-2183) Flatten behavior not consistent with rest of drill when we use it on a non-existent field

["Jacques Nadeau (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/DRILL-2183?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14311760#comment-14311760 ] 

Jacques Nadeau commented on DRILL-2183:
---------------------------------------

Remember, flatten is an operator, not a function.  What you're looking for is that flatten(null) => null rather than error.  Since we're defining FLATTEN semantics, I guess that could make sense.  Only question is whether that is a little weird.  As of right now FLATTEN(<<EMPTY_LIST>>) removes the record.  It seems like FLATTEN(<<NULL>>) would behave similarly.  So we should pick one pattern or the other for both operations.

> Flatten behavior not consistent with rest of drill when we use it on a non-existent field
> -----------------------------------------------------------------------------------------
>
>                 Key: DRILL-2183
>                 URL: https://issues.apache.org/jira/browse/DRILL-2183
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Execution - Relational Operators
>            Reporter: Rahul Challapalli
>            Priority: Minor
>
> git.commit.id.abbrev=c54bd6a
> Projecting a non-existent column
> {code}
> 0: jdbc:drill:schema=dfs_eea> select field_not_present from `data1.json`;
> +-------------------+
> | field_not_present |
> +-------------------+
> | null              |
> | null              |
> +-------------------+
> {code}
> Using a non existent column in a function :
> {code}
> 0: jdbc:drill:schema=dfs_eea> select MAX(field_not_present) from `data1.json`;
> +------------+
> |   EXPR$0   |
> +------------+
> | null       |
> +------------+
> 1 row selected (0.07 seconds)
> {code}
> Using a non-existent column in a filter
> {code}
> 0: jdbc:drill:schema=dfs_eea> select MAX(uid) from `data1.json` where field_not_present > 1;
> +------------+
> |   EXPR$0   |
> +------------+
> +------------+
> {code}
> However when we use flatten things work differently
> {code}
> 0: jdbc:drill:schema=dfs_eea> select uid , flatten(field_not_present) from `data1.json`;
> Query failed: RemoteRpcException: Failure while running fragment., org.apache.drill.exec.vector.NullableIntVector cannot be cast to org.apache.drill.exec.vector.RepeatedVector [ afe463ac-2523-43b2-882f-814ae4a116e1 on qa-node191.qa.lab:31010 ]
> [ afe463ac-2523-43b2-882f-814ae4a116e1 on qa-node191.qa.lab:31010 ]
> Error: exception while executing query: Failure while executing query. (state=,code=0)
> {code}
> Even if we do not make it consistent, we should at-least make the error message more clear



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)