[jira] [Commented] (DRILL-4007) json reader treats empty list inconsistently,

["Jacques Nadeau (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/DRILL-4007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14986707#comment-14986707 ] 

Jacques Nadeau commented on DRILL-4007:
---------------------------------------

This may be a user bug but isn't a bug in what I would say is the traditional sense. This is a very known limitation of the Drill system. Drill doesn't currently distinguish between: an untyped null list: an untyped null value, a typed null value and a nonexistent field. This is the current design of the system.

Addressing this isn't a small undertaking and a number of people are already working on it. The problem here has very little to do with the JSON reader.  There are a number of things that need to be solved before this can be addressed. The umbrella issue that comes to mind immediately is DRILL-3229, now added as a dependency for this JIRA. Additionally, we have to resolve how to deal with untyped nulls, how we resolve them in the context of JDBC and ODBC, dealing with schema changes between untyped nulls and typed values, etc. 

> json reader treats empty list inconsistently,
> ---------------------------------------------
>
>                 Key: DRILL-4007
>                 URL: https://issues.apache.org/jira/browse/DRILL-4007
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Storage - JSON
>            Reporter: Sean Hsuan-Yi Chu
>            Assignee: Sean Hsuan-Yi Chu
>            Priority: Critical
>         Attachments: a.json, b.json
>
>
> Depending on where the empty list shows up, the empty list could be treated as empty-list or just a null. 
> Running the following query on the folder with files in the attachment can reproduce the observation:
> {code}
> +--------+
> |   a    |
> +--------+
> | null   |
> | ["b"]  |
> | []     |
> +--------+
> {code}
> Note that both first and third records come from
> {code}
> {“a”:[]}
> {code}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)