// the find
wesm/feather
Feather: fast, interoperable binary data frame storage for Python, R, and more powered by Apache Arrow
This is the original Feather repo from Wes McKinney — a fast binary columnar format for sharing data frames between Python and R. Development has since moved into Apache Arrow itself, making this repo a historical artifact. If you're reading this README, you should be using `pyarrow.feather` or `arrow::read_feather` in R instead.
The core idea was genuinely good: Arrow's columnar layout makes reads fast because you can memory-map the file and skip deserialization for many numeric types. Cross-language interop between Python and R without CSV round-trips was a real pain point it solved well. The format spec in `doc/FORMAT.md` is straightforward enough that third-party implementations appeared quickly.
The repo is effectively abandoned — it's a redirect notice with legacy bindings that wrap pyarrow anyway. The Python package `feather-format` on PyPI still exists but is just a thin shim; installing it adds a dependency chain for zero benefit over using pyarrow directly. No Parquet support, no schema evolution, and the original format is superseded by Feather V2 (IPC format) which only lives in the Arrow repo. The listed Julia binding is a separate community project that may or may not track Arrow compatibility.