We're thrilled to get feedback and code contributions on pcodec!

If you think you've found a bug or other issue, please file a Github issue.

If you have a feature request or general question, it's best to join our Discord for a quick response. If you're opposed to creating a Discord account, Github issues also work.

Code Contribution

• If you're thinking of implementing something, it's best to chat with us about it first. That way we can vet your idea and make sure your efforts won't be in vain.
• Before making a PR, make sure to
  • Test your code; cargo test and cargo clippy.
  • Format it; cargo fmt.
  • Run the benchmarks. This verifies compression and decompression works for each synthetic dataset, which occasionally catches strange cases the tests miss. Also, if your change might affect performance, compare relevant runtimes to the benchmarks on the main branch.

Looking for ideas on what to contribute? Grep through the repo for concrete TODOs, or look at our project ideas for harder, underspecified problems.

Testing

Rust

cargo test

Python

From a venv with maturin installed, and from pco_python/, run maturin develop to recompile and install the python package into your venv. Then you can run pytest --doctest-glob='*.md'.

To build the Sphinx documentation locally, make sure sphinx and toml are installed, and run:

rm -rf docs/_build && \
  maturin develop && \
  sphinx-apidoc -o docs/ . && \
  pushd docs && \
  make html && \
  open _build/html/index.html && \
  popd

Java

With Maven installed, and from pco_java/, run mvn test -P rust if you want to recompile the Rust code, or just mvn test if not.

Deploying Packages

Rust / Crates.io

pco and pco_cli are manually deployed with cargo publish from a local clone of the repo.

Python / PyPi

pco_python is packaged by a Github workflow whenever the release name contains "Python". This runs a lot of maturin builds, each of which produces a dynamic library for a targets (in the sense of OS / hardware tuples). If all builds succeed, each such package is automatically and separately published to PyPi.

Java / JVM / Maven Central Repository

pco_java is similarly packaged by a Github workflow whenever the release name contains "Java". Unlike Python, libraries for all platforms get bundled into a single JAR before being published to Maven Central Repository staging. After the upload, someone (probably @mwlon) needs to log into the Maven Central Repository website and promote it from staging to prod if it looks good.