[DRILL-6390] UnsupportedOperationException: Unable to get holder type for minor type [LATE] and mode [OPTIONAL]

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.11.0
• Component/s: Execution - Data Types, Storage - JSON
• Assignee: Unassigned
• Reporter: Trung Ta
• Created: 08/May/18 09:10
• Updated: 13/Dec/18 23:45

Description

I tried this query on some of our JSON files that are stored in a S3 Bucket, in which I tried to convert JSON objects (application, application_fee) into varchar:

SELECT
     CONCAT(
         '"', CONVERT_TO(_stg.data.application, 'JSON'), '"'
         , ',', '"', CONVERT_TO(_stg.data.application_fee, 'JSON'), '"'
     ) as JSONs
 FROM (
     SELECT
         flatten(js.data) AS data
     FROM
         s3_bucket.`<PATH_TO_FILE_IN_S3>` AS js
     ) _stg

On some of the datasets the query failed and I got this error message

org.apache.drill.common.exceptions.UserRemoteException: SYSTEM ERROR: UnsupportedOperationException: Unable to get holder type for minor type [LATE] and mode [OPTIONAL] Fragment 0:0 [Error Id: 5e74e125-de22-46fa-8d31-233d0fc40140

The dataset, on which the query failed looks like following:

{
    "data": 
        [
            {
                "application": null,
                "application_fee": null,
                "amount": 1
            },
            {
                "application": 
                    {
                         "id": "some_id",
                         "name": "Some App",
                         "object": "application"
                    },
                "application_fee": null,
                "amount": 5
            }
        ],
    "has_more": false,
    "object": "list",
    "url": "/v1/charges"
}

I ran some further tests on the query and found out that the query only fails at object "application_fee", which is null in both arrays of object "data". Which might explain why I never ran into this error in the other datasets (no object arrays there are null).