[ https://issues.apache.org/jira/browse/DRILL-3123?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14547397#comment-14547397 ]
Steven Phillips commented on DRILL-3123: ---------------------------------------- I think this is kind of a strange query, and the partition column feature was not really designed to be used this way. I am inclined to return no "directory" columns at all when doing globbing. > Dir0 has issues when we have a '/' at the beginning of the path > --------------------------------------------------------------- > > Key: DRILL-3123 > URL: https://issues.apache.org/jira/browse/DRILL-3123 > Project: Apache Drill > Issue Type: Bug > Components: Execution - Data Types > Affects Versions: 1.0.0 > Reporter: Rahul Challapalli > Assignee: Daniel Barclay (Drill) > > Follow the below steps : > {code} > hadoop fs -mkdir /drill/testdata/repro1/20150120 > hadoop fs -mkdir /drill/testdata/repro1/20150121 > 1. Add the below workspace : > "repro1": { > "location": "/drill/testdata/repro1", > "writable": true, > "defaultInputFormat": "parquet" > } > 2. Now copy a sample json file into both the above directories > {code} > The below query returns incorrect results : > {code} > select * from dfs.repro1.`/*/sample.json` limit 1; > +-----------+---------+-----------+-----+------+ > | dir0 | dir1 | dir2 | id | val | > +-----------+---------+-----------+-----+------+ > | testdata | repro1 | 20150121 | 1 | 1 | > +-----------+---------+-----------+-----+------+ > {code} > The same query worked from an older build (commit # > d10769f478900ff1868d206086874bdd67a45e7d) > {code} > select * from dfs.repro1.`/*/sample.json` limit 1; > +------------+------------+------------+ > | dir0 | id | val | > +------------+------------+------------+ > | 20150121 | 1 | 1 | > +------------+------------+------------+ > {code} -- This message was sent by Atlassian JIRA (v6.3.4#6332)