Parsing nested csv data for apache druid

I am trying to parse nested tsv structure. It is a little convoluted data. For example,
column_x, column_y, column_z
1, 2, {‘key’: ‘value’, ‘key1’:{‘key2’: ‘value2’}}

need to parse column_z as json field and feed data to druid.

Read about flatten, which only works on json data. How to write a ingestion spec for such a case,i.e, to filter column and parse json.

You can not currently do that without some preprocessing (to filter out the column and ingest as JSON) ahead of time