[
https://issues.apache.org/jira/browse/DRILL-451?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13947243#comment-13947243
]
Steven Phillips commented on DRILL-451:
---------------------------------------
I don't think this data is valid. Although drill can handle dynamic, changing
schema, I think within a single file, a particular field can only have one
type. That seems like a reasonable requirement.
If anyone else has thoughts, feel free to chime in.
> JSON read fails if rows have different number of fields
> -------------------------------------------------------
>
> Key: DRILL-451
> URL: https://issues.apache.org/jira/browse/DRILL-451
> Project: Apache Drill
> Issue Type: Bug
> Reporter: Ramana Inukonda Nagaraj
>
> Query:
> 0: jdbc:drill:> select * from dfs.`/user/root/employeeObject.json`;
> Notice the input file has different fields.
> Json file:
> {
> "first": "Jimmy1",
> "age": 29,
> "sex": "M",
> "salary": 6300.10
> }
> {
> "first": "Jimmy",
> "last": "James",
> "age": 29,
> "sex": "M",
> "salary": 6300.10
> }
> {
> "first": "Jimmy",
> "sex": "M",
> "salary": 6300.10
> "wealth":100000
> }
> This works though:
> {
> "first": "Jimmy1",
> "age": 29,
> "sex": "M",
> "salary": 6300.10
> }
> {
> "first": "Jimmy",
> "last": "James",
> "age": 29,
> "sex": "M",
> "mom": "testString",
> "salary": 6300.10
> }
> 0: jdbc:drill:> select * from dfs.`/user/root/employeeObject.json`;
> +------------+------------+------------+------------+------------+------------+
> | first | age | sex | salary | last | mom
> |
> +------------+------------+------------+------------+------------+------------+
> | Jimmy1 | 29 | M | 6300.1 | null | null
> |
> | Jimmy | 29 | M | 6300.1 | James | testString
> |
> +------------+------------+------------+------------+------------+------------+
> After playing around a bit, looks like this only happens if the extra field
> is a float. Works perfectly if extra fields are int or string.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
