Schemaless queries

[collimarco]

I wonder if Ballista needs a pre-defined table schema in order to query thousands of similar Parquet files stored on S3.

Basically the Parquet files have a similar schema, but since they were added at different dates, they can have some differences (e.g. new columns, columns removed, a "status" column that changed from int to string, etc.)

1. Is it possible to use Ballista to query those "similar" files? Or all the Parquet files need to have the exact same schema?
2. Do you need to create a table/view and define the exact fields contained in those S3 Parquet files (e.g. with CREATE TABLE), or you can simply query them and Ballista infers their schema?