trying to join artworks and artists is a big pain because of needing to split and unnest the constituent id. this query is way too slow to run on the entire dataset live. we could create a join table which is a common thing in analytics.

ya! that makes sense. I can take a crack at making that table

ok, added a table, though not sure if I'm missing some DuckDB syntax for doing joins? look at the cell below. Also, does it make sense to make that lookup table a parquet file too? Its fairly small (1.8MB)

alright, i made it a parquet file and fixed a bug with the artists csv/parquet file. i changed to use an explicit join as well.

maybe we should rename to something more specific like `dbMoMA`?

it's been a bit of a convention to use db for instantiating a database client, we do it in the other SQL datasets as well. part of the motivation is that if someone wants to copy existing code/cell and use for their own data in another notebook its less effort to switch