[
https://issues.apache.org/jira/browse/DRILL-2738?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14490403#comment-14490403
]
Venkata krishnan Sowrirajan commented on DRILL-2738:
----------------------------------------------------
May I know why SQL doesn't guarantee this query? Should this query be
re-written like below (casting column[0] as timestamp with a subquery
offsetting the first row)
"select cast(columns[0] as timestamp) `t`, * from (select * from
`guts-csv/CSV/guts_run_lab-app002.csv` offset 1);"
> Offset with casting a column to timestamp not working
> -----------------------------------------------------
>
> Key: DRILL-2738
> URL: https://issues.apache.org/jira/browse/DRILL-2738
> Project: Apache Drill
> Issue Type: Bug
> Affects Versions: 0.7.0
> Reporter: Venkata krishnan Sowrirajan
>
> In the below query, it should skip the first row which is a header and want
> to cast one of the column to timestamp. But it is trying to parse the first
> row to cast it to timestamp. Without casting it to timestamp, simple offset
> query works fine.
> "select cast(columns[0] as timestamp) from
> `guts-csv/CSV/guts_run_lab-app002.csv` offset 1;"
> So I did explain plan on the above query
> explain plan without implementation for select cast(columns[0] as timestamp)
> from `guts-csv/CSV/guts_run_lab-app002.csv` offset 1;
> DrillScreenRel
> DrillLimitRel(offset=[1])
> DrillProjectRel(EXPR$0=[CAST(ITEM($0, 0)):TIMESTAMP(0)])
> DrillScanRel(table=[[fs, drill, guts-csv/CSV/guts_run_lab-app002.csv]],
> groupscan=[EasyGroupScan
> [selectionRoot=/mapr/yarn-test/drill/guts-csv/CSV/guts_run_lab-app002.csv,
> numFiles=1, columns=[`columns`[0]],
> files=[file:/mapr/yarn-test/drill/guts-csv/CSV/guts_run_lab-app002.csv]]])
> In the plan, it looks like it tries to do casting to timestamp and then the
> offset operation which is why its failing.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
