Part of the problem here is that the static dataframe is designed to be
used a read only abstraction in Spark, and updating that requires the user
to drop the dataframe holding the reference data and recreate it. And in
order for the join to use the recreated dataframe, the query has to be
I am extremely leery about pushing product on this forum and have refrained
from it in the past. But since you are talking about loading parquet data
into Spark, run some aggregate queries and then write the results to a fast
data store, and specifically asking for product options, it makes