I had the question of whether you can create a Looker view from a database view (vs. the database table, which is the default). From Looker support, it looks like this is natively possible in Snowflake and for other DB platforms, this is possible by using a derived table, simplified by executing a query against a database view in SQL Runner and then using the “Add to Project” feature. This will create the dimensions and measures dynamically.
My question, now knowing that it’s possible, is this a best practice? Is it better to build the core aggregate DB tables in the ETL and then do any “refined” modeling in Looker? Just struggling a bit on how to split the modeling work between the DB and Looker. We are using Aurora MySQL BTW.