Checkout the R bindings for DuckDB[0]. You should find that it does the same thing (i.e. run SQL against a dataframe/file on disk) much faster for many SQL operations.

[0] https://duckdb.org/docs/api/r