[
https://issues.apache.org/jira/browse/DRILL-1259?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14087358#comment-14087358
]
Steven Phillips commented on DRILL-1259:
----------------------------------------
This is something we can discuss, but I'm of the opinion that fields in a json
document should be treated like a map, without a well defined order. If a query
selects specific columns, they should be returned in the correct order, but for
* queries, I don't think there should be any expectation for what order the
columns will come back in, the exception would be storage plugins/formats that
have a rigidly defined schema.
> Drill changing the order of the columns when the first record has null value
> for the column
> -------------------------------------------------------------------------------------------
>
> Key: DRILL-1259
> URL: https://issues.apache.org/jira/browse/DRILL-1259
> Project: Apache Drill
> Issue Type: Bug
> Components: Storage - JSON
> Reporter: Rahul Challapalli
>
> git.commit.id.abbrev=98b208e
> JSON File Used :
> {code}
> {
> "a" : null,
> "b" : 1
> }
> {
> "a" : 1,
> "b" : 1
> }
> {code}
> Query : The order has changed in the below result
> {code}
> select * from `temp.json`;
> +------------+------------+
> | b | a |
> +------------+------------+
> | 1 | null |
> | 1 | 1 |
> +------------+------------+
> {code}
> However with the below dataset where the null field is in the second row
> drill preserves the order
> {code}
> {
> "a" : 1,
> "b" : 1
> }
> {
> "a" : null,
> "b" : 1
> }
> {code}
> Query:
> {code}
> select * from `temp.json`;
> +------------+------------+
> | a | b |
> +------------+------------+
> | 1 | 1 |
> | null | 1 |
> +------------+------------+
> {code}
--
This message was sent by Atlassian JIRA
(v6.2#6252)
