todo

https://pandas.pydata.org/docs/development/extending.html#registering-custom-accessors

thoughts

i dunno. maybe the default duckdb is pretty good
- https://duckdb.org/2021/05/14/sql-on-pandas.html
- except, i dont' know if i like the implicit namespace stuff they're doing

https://gist.github.com/ajfriend/6942bdd4a9eb31861f7dfab471f411fb

also, maybe the whole relational API is pretty good?
- https://duckdb.org/2021/05/14/sql-on-pandas.html

idea: garbage-collected, duckdb-backed pandas replacement

avoids marshalling back and forth to pandas
lets users use natural python interface, instead

viewers/editors