You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Explicitly label output views. In a complex program, many of the views may only be used for intermediate results and don't need to be integrated (we also don't need to attach neighborhood circuits to such views). One way to express this would be to explicitly label output views. [SQL] Support LOCAL views, which do not produce circuit outputs #1851
Multiway joins. A multiway join operator will allow us to avoid indexing intermediate results of each pairwise join in a chain. It can be used in multiple contexts, but will make the most difference in queries that compute multiple aggregates, including rolling aggregates.
Implement linear aggregates wherever possible. Linear aggregates use less storage and are more efficient. They are tricky in SQL due to the semantics of NULLs, but can be done according to @mihaibudiu
GC: make sure we are able to GC everything that's GC'able in theory
distinctantijoin[SQL] There is no GC for antijoins #3390emit_finalviewsExplicitly label output views. In a complex program, many of the views may only be used for intermediate results and don't need to be integrated (we also don't need to attach neighborhood circuits to such views). One way to express this would be to explicitly label output views. [SQL] Support LOCAL views, which do not produce circuit outputs #1851
Multiway joins. A multiway join operator will allow us to avoid indexing intermediate results of each pairwise join in a chain. It can be used in multiple contexts, but will make the most difference in queries that compute multiple aggregates, including rolling aggregates.
Implement linear aggregates wherever possible. Linear aggregates use less storage and are more efficient. They are tricky in SQL due to the semantics of NULLs, but can be done according to @mihaibudiu
Optimize away unused columns.