[
https://issues.apache.org/jira/browse/DRILL-1241?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14098071#comment-14098071
]
Aman Sinha commented on DRILL-1241:
-----------------------------------
Every operator should return an output schema even if there are no records (see
previously fixed bug DRILL-884). In this particular bug, the LIMIT operator
was not obeying that policy. Fixed by adding the appropriate handling for such
case in LimitRecordBatch. Added various unit tests with Limit 0 on scans,
joins, aggregations; also tested with non-zero Offsets.
> Query that returns 0 rows gives alarming error
> ----------------------------------------------
>
> Key: DRILL-1241
> URL: https://issues.apache.org/jira/browse/DRILL-1241
> Project: Apache Drill
> Issue Type: Bug
> Reporter: Julian Hyde
> Assignee: DrillCommitter
> Fix For: 0.5.0
>
> Attachments:
> 0001-DRILL-1241-Ensure-that-Limit-produces-at-least-1-bat.patch
>
>
> When I run a query with 'limit 0' I got an error that sounded like an
> internal protocol error.
> {code:sql}
> 0: jdbc:drill:zk=local> SELECT * FROM cp.`employee.json` limit 0;
> Query failed: Failure while running fragment. The incoming iterator returned
> a state of NONE on the first batch. There should always be at least one batch
> output before returning NONE [a7416ee7-1d9c-4591-a927-2664455f3ef1]
> Error: exception while executing query: Failure while trying to get next
> result batch. (state=,code=0)
> {code}
> Should be more like a user error, since Drill cannot handle queries that
> return 0 rows. Or maybe (if as I surmise the problem is that Drill just
> doesn't know the names of the columns) generate a default column
> 'unknown_result_set_schema' and no rows.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
