Thanks, the user-research-first part was definitely enlightening and important to do. Our first many iterations were very wrong.
As for large scale datasets - we do cache those sample tables and we only grab the first 10 rows. Also, BigQuery actually has a great API for fetching a set of sample data that we utilize heavily. They made it because of exactly the potential concerns you outlined around columnar stores.
As for large scale datasets - we do cache those sample tables and we only grab the first 10 rows. Also, BigQuery actually has a great API for fetching a set of sample data that we utilize heavily. They made it because of exactly the potential concerns you outlined around columnar stores.