[
https://issues.apache.org/jira/browse/DRILL-3646?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14725193#comment-14725193
]
Hari Sekhon commented on DRILL-3646:
------------------------------------
Show files does work, but then you have to know which files are data sources
and which directories are tables. One could easily set up a specific workspace
where everything inside is a data source to work around this, but it still
seems a little counterintuitive if coming from RDBMS or even Hive to be able to
select from a "phantom" table that doesn't appear in "show tables".
I believe the more intuitive behaviour would be for Drill to show all files and
directories with relevant data formats such as CSV, JSON, *.log, parquet files
etc. I believe this would only slow things down if there are a lot of
subdirectories to do listings on to determine if they're tables?
> Show tables in DFS workspace
> ----------------------------
>
> Key: DRILL-3646
> URL: https://issues.apache.org/jira/browse/DRILL-3646
> Project: Apache Drill
> Issue Type: Bug
> Components: Metadata, Storage - Information Schema
> Affects Versions: 1.1.0
> Reporter: Hari Sekhon
> Assignee: Steven Phillips
>
> Drill does not show tables in a DFS workspace, even when I just created a
> parquet table there using CTAS through Drill itself.
> The output results are blank, a zero row table with blank column header.
> {code}
> 0: jdbc:drill:zk=local> show tables in dfs.hari;
> +--+
> | |
> +--+
> +--+
> No rows selected (0.137 seconds)
> {code}
> although I can still query the table I just created as long as I know it's
> there and query it blindly:
> {code}
> 0: jdbc:drill:zk=local> select count(*) from dfs.hari.auditlogs_parquet_drill;
> +---------+
> | EXPR$0 |
> +---------+
> | 2579 |
> +---------+
> 1 row selected (0.129 seconds)
> {code}
> I can't describe the table so I really do have to query it blindly too (I
> previously raised a different jira for parquet describe support DRILL-3525
> and other formats DRILL-3524 to DRILL-3529):
> {code}
> 0: jdbc:drill:zk=local> select count(*) from dfs.hari.auditlogs_parquet_drill;
> +---------+
> | EXPR$0 |
> +---------+
> | 2579 |
> +---------+
> 1 row selected (0.129 seconds)
> {code}
> This jira is specifically to address the inability to list tables (or perhaps
> files/dirs) in a DFS workspace though.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
