You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Chris Mathews (JIRA)" <ji...@apache.org> on 2015/10/13 18:12:05 UTC
[jira] [Commented] (DRILL-3922) select * on json document with dot
in key name results in failure
[ https://issues.apache.org/jira/browse/DRILL-3922?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14955200#comment-14955200 ]
Chris Mathews commented on DRILL-3922:
--------------------------------------
I think this is related since the same error message is returned; so I do not think this is just a JSON issue.
>From the following examples it can be seen that when accessing parquet files the results are inconsistent, but a similar error message is returned:
{noformat:bgColor=#B2D1A7}
0: jdbc:drill:> select count(z_timestamp) from zen.pm.ERICSSON_WRAN;
+-----------+
| EXPR$0 |
+-----------+
| 10972390 |
+-----------+
1 row selected (1.444 seconds)
{noformat}
{noformat:bgColor=#B2D1A7}
0: jdbc:drill:> select count(`H3_VS.RB.PS.Int.DL.384`) from zen.pm.ERICSSON_WRAN;
Error: SYSTEM ERROR: UnsupportedOperationException: Field references must be singular names.
{noformat}
{noformat:bgColor=#DEA897}
0: jdbc:drill:> select `H3_VS.RB.PS.Int.DL.384` from zen.pm.ERICSSON_WRAN limit 1;
Error: SYSTEM ERROR: UnsupportedOperationException: Field references must be singular names.
Fragment 1:0
{noformat}
{noformat:bgColor=#B2D1A7}
0: jdbc:drill:> select `H3_VS.RB.PS.Int.DL.384` as `xyz` from zen.pm.ERICSSON_WRAN limit 1;
+-------+
| xyz |
+-------+
| null |
+-------+
1 row selected (3.749 seconds)
{noformat}
And using another parquet file with column names containing dot:
{noformat:bgColor=#B2D1A7}
0: jdbc:drill:> select `z.date` from zen.root.cjm;
+----------------+
| z.date |
+----------------+
| 1429592610462 |
+----------------+
1 row selected (0.133 seconds)
{noformat}
{noformat:bgColor=#B2D1A7}
0: jdbc:drill:> select `z.date` as `H3_VS.RB.PS.Int.DL.384` from zen.root.cjm;
+-------------------------+
| H3_VS.RB.PS.Int.DL.384 |
+-------------------------+
| 1429592610462 |
+-------------------------+
1 row selected (0.11 seconds)
{noformat}
{noformat:bgColor=#DEA897}
0: jdbc:drill:> select count(`z.date`) from zen.root.cjm;
Error: SYSTEM ERROR: UnsupportedOperationException: Field references must be singular names.
{noformat}
{noformat:bgColor=#B2D1A7}
0: jdbc:drill:> select count(*) from zen.root.cjm;
+---------+
| EXPR$0 |
+---------+
| 1 |
+---------+
1 row selected (0.141 seconds)
{noformat}
> select * on json document with dot in key name results in failure
> -----------------------------------------------------------------
>
> Key: DRILL-3922
> URL: https://issues.apache.org/jira/browse/DRILL-3922
> Project: Apache Drill
> Issue Type: Bug
> Components: Storage - JSON
> Affects Versions: 1.1.0, 1.2.0
> Reporter: John Omernik
> Fix For: Future
>
>
> If there is a dot in the key name of a json field, and the user issues a select * from `path/to/file.json` it will result a system error: "Unsupported Operation Exception: Field references must be singular names"
> To reproduce:
> {"hello":"goodbye", "yousayyes":"Isayno"}
> Works fine
> {"hello.yoko":"goodbye", "yousayyes":"Isayno"}
> Fails.
> Note, this is only if you do select * if you do select `hello.yoko` from file then it will work fine.
> The select * is needed to help with data exploration.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)