I'm using Apache Iceberg which performs queries on Parquet files. The data saves a lot of types of events and their respective numeric values. I'm hesitating whether it's more efficient to have one column for event type or have a column for each event.
VS
Dedicated column for each event type:

I'm wondering though if I want to run a group by query for col1, col2 by some event_type in theory if there was a dedicated column for that event_type then the db would have to scan less data as opposed to having scan more data when all event types are stored in a single column.
What is the best practice when designing schemas for such data in columnar data storage?
