Scyros

Sound, reproducible and scalable studies of GitHub repositories.

Scyros samples public repositories on GitHub, filters them, downloads their code at a fixed commit and extracts the functions you want to study. It is a command-line tool written in Rust, built so that other researchers can repeat your study and get the same data.

Get started

Paste the following into a terminal:

macOS and Linux

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/fxpl/scyros/releases/latest/download/scyros-installer.sh | sh

Windows (PowerShell)

powershell -ExecutionPolicy Bypass -c "irm https://github.com/fxpl/scyros/releases/latest/download/scyros-installer.ps1 | iex"

Cargo (needs Rust 1.95 or newer)

cargo install scyros

Nix (with flakes enabled)

nix profile install github:fxpl/scyros

Then run scyros --help. New to Scyros? Follow the tutorial or read the docs.

Research

Papers that use Scyros.

TACAS 2026

Verifying Floating-Point Programs in Stainless

Andrea Gilot, Axel Bergström and Eva Darulova

An extension of the Stainless deductive verifier with floating-point support. The extension is evaluated on a new set of benchmarks extracted with Scyros from the dataset mined in the OOPSLA 2026 empirical study of floating-point code. These benchmarks show that Stainless can verify specifications of real-world code and produce counterexamples when the specifications do not hold.

Paper Benchmarks

OOPSLA 2026

Floating-Point Usage on GitHub: A Large-Scale Study of Statically Typed Languages

Andrea Gilot, Tobias Wrigstad and Eva Darulova

An empirical study of how floating-point code is used on GitHub. This study uses Scyros to mine 450k repositories containing statically typed floating-point code, and extracts 10 million floating-point functions from them. It shows that floating-point code is used in the majority of engineered repositories on GitHub, and that the benchmarks used in the literature to evaluate automated reasoning techniques for floating-point arithmetic do not fully represent the floating-point code used in practice. The study also shows how Scyros and the mined dataset can be used to generate new floating-point benchmarks that reflect real-world code.

Paper Dataset Benchmarks

Citing Scyros

Until a paper about Scyros itself is published, please cite the study it was introduced in:

@article{gilot2026largescalestudyfloatingpointusage,
  author    = {Gilot, Andrea and Wrigstad, Tobias and Darulova, Eva},
  title     = {Floating-Point Usage on {GitHub}: A Large-Scale Study of
               Statically Typed Languages},
  journal   = {Proc. ACM Program. Lang.},
  volume    = {10},
  number    = {OOPSLA1},
  articleno = {95},
  numpages  = {28},
  year      = {2026},
  month     = apr,
  publisher = {Association for Computing Machinery},
  doi       = {10.1145/3798203},
  url       = {https://doi.org/10.1145/3798203}
}