Tajo already has a support for self-describing data formats like JSON, Parquet, or ORC. While they are capable of providing schema information by themselves, users must define schema to query on them with the current implementation. To solve this inconvenience, we have to improve our query planner to support self-describing data formats well.
First, we need to allow omitting schema definition for the create table statement. When a query is submitted for a self-describing table, the columns which don't exist in that table will be filled with Nulls.