Skip to content

README.md

Latest commit

 

History

History
52 lines (40 loc) · 2.17 KB

README.md

File metadata and controls

52 lines (40 loc) · 2.17 KB
Raw

Optional Rust acceleration

Skylos works without this extension and uses Python fallbacks when it cannot import the native APIs. The extension is a separate source project that provides the skylos_fast module. Skylos does not declare a fast pip extra; pip install "skylos[fast]" does not install this extension.

Build from source

Use the Python environment in which you run Skylos. You need Python 3.11 or later, a Rust toolchain with Cargo, and a native compiler/linker. On macOS, install the Xcode Command Line Tools with xcode-select --install if they are missing.

From a checkout of this repository:

python -m pip install ./rust
python -c "from skylos.core.fast import FAST_AVAILABLE; print(FAST_AVAILABLE)"
skylos doctor

The build uses Maturin and compiles for that Python interpreter and platform. The availability check should print True, and doctor should report skylos_fast available. Rebuild the extension when changing Python versions or CPU architecture. To return to the Python fallbacks, run python -m pip uninstall skylos-fast in the same environment.

What it accelerates

  • Source-file discovery.
  • Code-clone similarity comparisons.
  • Python class-coupling analysis.
  • Import-cycle detection.

This does not accelerate every check. Native AST visitor acceleration and batch grep are not currently wired into the analyzer.

Can findings change?

Yes. The native extension is experimental and can change scan speed and findings. Clone similarity uses a separate implementation that does not apply the default difflib.SequenceMatcher heuristic for frequently repeated characters. For example, comparing "x" + "a" * 200 with "y" + "a" * 200 scores 0.0 in the Python fallback and about 0.995 in the native implementation. That can change which pairs exceed a clone threshold. Native coupling analysis also uses a separate parser.

The parity workflow builds the extension and checks selected examples against Python implementations. Those tests do not guarantee identical findings for every repository. Keep the same backend when comparing scan results across machines or CI runs.