This guide is intended to help you get started with developing portgraph.
If you find any errors or omissions in this document, please open an issue!
You can setup the development environment you will need:
You can use the git hook in .github/pre-commit
to automatically run the test and check formatting before committing.
To install it, run:
ln -s .github/pre-commit .git/hooks/pre-commit
# Or, to check before pushing instead
ln -s .github/pre-commit .git/hooks/pre-push
To compile and test the rust code, run:
cargo build
cargo test
Finally, if you have rust nightly installed, you can run miri
to detect
undefined behaviour in the code.
cargo +nightly miri test
We use two kinds of benchmarks in this project:
- A wall-clock time benchmark using
criterion
. This measures the time taken to run a function by running it multiple times. - A single-shot instruction count / memory hits benchmark using
iai-callgrind
. This measures the number of instructions executed and the number of cache hits and misses.
Both tools run the same set of test cases.
When profiling and debugging performance issues, you may also want to use samply to visualize the see flame graphs of specific examples.
This is the simplest kind of benchmark. To run the, use:
cargo bench --bench criterion_benches
These benchmarks are useful when running in noisy environments, in addition to being faster than criterion. We run these on CI to track historical performance in bencher.dev.
To run these, you must have valgrind
installed.
Support for Apple Silicon (M1/M2/...) macs is
experimental, so you
will need to manually clone and compile the branch. See
LouisBrunner/valgrind-macos
for instructions.
In addition to valgrind
, you will need to install iai-callgrind
runner. The
pre-build binaries are available on
cargo binstall
.
cargo binstall iai-callgrind-runner
The benchmarks can then be run with:
cargo bench --bench iai_benches
The rustfmt tool is used to enforce a consistent rust coding style. The CI will fail if the code is not formatted correctly.
To format your code, run:
# Format rust code
cargo fmt
We also check for clippy warnings, which are a set of linting rules for rust. To run clippy, run:
cargo clippy --all-targets
We run coverage checks on the CI. Once you submit a PR, you can review the line-by-line coverage report on codecov.
To run the coverage checks locally, install cargo-llvm-cov
, generate the report with:
cargo llvm-cov --lcov > lcov.info
and open it with your favourite coverage viewer. In VSCode, you can use
coverage-gutters
.
We welcome contributions to portgraph! Please open an issue or pull request if you have any questions or suggestions.
PRs should be made against the main
branch, and should pass all CI checks before being merged. This includes using the conventional commits format for the PR title.
The general format of a contribution title should be:
<type>(<scope>)!: <description>
Where the scope is optional, and the !
is only included if this is a semver breaking change that requires a major version bump.
We accept the following contribution types:
- feat: New features.
- fix: Bug fixes.
- docs: Improvements to the documentation.
- style: Formatting, missing semi colons, etc; no code change.
- refactor: Refactoring code without changing behaviour.
- perf: Code refactoring focused on improving performance.
- test: Adding missing tests, refactoring tests; no production code change.
- ci: CI related changes. These changes are not published in the changelog.
- chore: Updating build tasks, package manager configs, etc. These changes are not published in the changelog.
- revert: Reverting previous commits.