How does this compare to Postgres + parquet FDW? Is zero copy feasible in Postgres with FDWs?
I am not very familiar with the Postgres Parquet FDW, but here is an educated guess!
Postgres is a row store engine rather than a column store, so I believe there will need to be quite a lot of translation for Postgres to be able to process parquet data (DuckDB and parquet are both columnar). My hypothesis is that DuckDB would be significantly faster! However, feel free to benchmark things!
I should also add that there is a duckdb fdw, so you could have DuckDB read from your parquet files and do faster transformations before you pull your data into Postgres!