Comparison with extendr

Why does savvy exist alongside extendr?

extendr is a great framework with a larger community and a longer history. savvy was created as a simpler, more explicit alternative. It is not meant to be a complete replacement for extendr — rather, it offers a different set of trade-offs that may or may not suit your needs.

Both frameworks are actively maintained and used in production. Choose whichever fits your project and preferences better.

Comparison

Design philosophy

extendr prioritizes ergonomics. It automatically converts between R objects and native Rust types at function boundaries, so your Rust code can look clean and dependency-free (e.g., fn foo(x: &[f64]) -> f64). It also provides a rich set of helper macros (r!(), R!(), list!(), data_frame!(), etc.) to reduce boilerplate.

savvy prioritizes explicitness. Functions receive and return R-specific wrapper types (IntegerSexp, OwnedRealSexp, etc.), and no implicit type conversions happen. This can feel verbose, but it makes the data flow between R and Rust transparent and predictable.

API surface

extendr has a broader API surface, covering more R constructs out of the box (data frames, environments, formulas, language objects, pairlists, etc.) and offering optional integrations with ndarray, serde, and faer.

savvy intentionally provides a smaller API. Complex R constructs like data frames are delegated to R wrapper code. This keeps the Rust side simple but means you write more R glue code.

Stability and maintenance

extendr has a larger community (~500 GitHub stars, 33+ contributors, published in JOSS). It is actively maintained with regular releases.

savvy is maintained by a single developer with limited bandwidth. However, the minimal design means there is less surface area for bugs, and the framework has been stable without much active development. It is used in production by notable projects including polars and SedonaDB.