Introduction#

We welcome and encourage contributions of all kinds, from all levels, such as:

  1. Tickets with issue reports or feature requests

  2. Discussions

  3. Documentation improvements

  4. Code, both PR and (especially) PR Review.

In addition to submitting new PRs, we have a healthy tradition of community members reviewing each other’s PRs. Doing so is a great way to help the community as well as get more familiar with Rust and the relevant codebases.

Development Environment#

Start with the Development Environment Quick Start.

For more detail, see the full development environment guide and the testing guide.

Finding and Creating Issues to Work On#

You can find a curated good-first-issue list to help you get started. You can read about how we plan larger projects in the Roadmap and Improvement Proposals section.

Open Contribution and Assigning tickets#

DataFusion is an open contribution project, and thus there is no particular project imposed deadline for completing issues or restrictions on who can work on an issue, nor limits to how many people can work on an issue at the same time.

Contributors drive the project forward based on their own priorities and interests and thus you are free to work on any issue that interests you.

If someone is already working on an issue that you want or need but hasn’t been able to finish it yet, feel free to help them out.

If there is an existing open PR for an issue you plan to work on, please review that PR before opening a new one. Duplicate, unacknowledged PRs consume valuable reviewer time and we may close them. If there is an existing PR, please identify it in the PR description and explain why you are opening a new one and not helping with the previous one. In general it is both polite and will help avoid unnecessary duplication of work if you also leave a note on an issue when you start working on it.

If you want to work on an issue which is not already assigned to someone and has no comment indicating someone is already working on it, you can assign the issue to yourself by submitting a single word comment take. However, if you are unable to make progress please unassign the issue by commenting a single word untake.

Developer’s guide#

Pull Request Overview#

We welcome pull requests (PRs) from anyone in the community.

DataFusion is a rapidly evolving project and we try to review and merge PRs quickly.

Review bandwidth is currently our most limited resource, and we highly encourage reviews by the broader community. If you are waiting for your PR to be reviewed, consider helping review other PRs that are waiting. Such review both helps the reviewer to learn the codebase and become more expert, as well as helps identify issues in the PR (such as lack of test coverage), that can be addressed and make future reviews faster and more efficient.

The lifecycle of a PR is:

  1. Create a PR targeting the main branch.

  2. For new contributors a committer must first trigger the CI tasks. Please mention the members from committers list in the PR to help trigger the CI

  3. Your PR will be reviewed. Please respond to all feedback on the PR: you don’t have to change the code, but you should acknowledge the feedback. PRs waiting for the feedback for more than a few days will be marked as draft.

  4. Once the PR is approved, one of the committers will merge your PR, typically within 24 hours. We leave approved “major” changes (see below) open for 24 hours prior to merging, and sometimes leave “minor” PRs open for the same time to permit additional feedback.

Note that the above time frames are estimates. Due to limited committer bandwidth, it may take longer to merge your PR. Please wait patiently. If it has been several days you can friendly ping the committer who approved your PR to help remind them to merge it.

Creating Pull Requests#

With coding agents, the number of open PRs now far exceeds our review capacity. To help reviewers focus on fewer PRs, users without write access are limited to 3 open, non-draft PRs at a time. If you reach the limit, wait for existing PRs to be merged or close them until you have fewer than 3 before opening another.

When possible, we recommend splitting your contributions into multiple smaller focused PRs rather than large PRs (500+ lines) because:

  1. The PR is more likely to be reviewed quickly – our reviewers struggle to find the contiguous time needed to review large PRs.

  2. The PR discussions tend to be more focused and less likely to get lost among several different threads.

  3. It is often easier to accept and act on feedback when it comes early in a small change, before a particular approach has been polished too much.

If you are concerned that a larger design will be lost in a string of small PRs, creating a large draft PR that shows how they all work together can help.

Note all commits in a PR are squashed when merged to the main branch so there is one commit per PR after merge.

For larger PRs, it is often helpful to leave a review on your own PR with comments calling out important changes or specific important choices. These annotations can help reviewers quickly find areas they should focus on, thus speeding up review.

Release Management and Backports#

Contributor-facing guidance for release branches, patch releases, and backports is documented in the Release Management guide.

Before Submitting a PR#

Before submitting a PR, run the standard non-functional checks. PRs must pass before merge.

./dev/rust_lint.sh
# use `--write` to automatically fix some formatting and lint errors
# ./dev/rust_lint.sh --write --allow-dirty

Please ensure your PR follows the testing guide. In particular:

AI-Assisted Contributions#

We welcome AI-assisted PRs, but not unreviewed “AI dumps”. See the AI Policy page for what we expect from authors and reviewers who use AI tools.

Conventional Commits & Labeling PRs#

We generate change logs for each release using an automated process that will categorize PRs based on the title and/or the GitHub labels attached to the PR.

We follow the Conventional Commits specification to categorize PRs based on the title. This most often simply means looking for titles starting with prefixes such as fix:, feat:, docs:, or chore:. We do not enforce this convention but encourage its use if you want your PR to feature in the correct section of the changelog.

The change log generator will also look at GitHub labels such as bug, enhancement, or api change, and labels do take priority over the conventional commit approach, allowing maintainers to re-categorize PRs after they have been merged.

Reviewing Pull Requests#

See the Reviewing Pull Requests guide for what we look for when reviewing PRs and how to prepare your own for review.

Performance Improvements#

Performance improvements are always welcome: performance is a key DataFusion feature.

In general, the performance improvement from a change should be “enough” to justify any added code complexity. How much is “enough” is a judgement made by the committers, but generally means that the improvement should be noticeable in a real-world scenario and is greater than the noise of the benchmarking system.

To help committers evaluate the potential improvement, performance PRs should in general be accompanied by benchmark results that demonstrate the improvement.

The best way to demonstrate a performance improvement is with the existing benchmarks:

If there is no suitable existing benchmark, you can create a new one. It helps to isolate the effects of your change by creating a separate PR with the benchmark, and then a PR with the code change that improves the benchmark.

“Major” and “Minor” PRs#

Since we are a worldwide community, we have contributors in many timezones who review and comment. To ensure anyone who wishes has an opportunity to review a PR, our committers try to ensure that at least 24 hours passes between when a “major” PR is approved and when it is merged.

A “major” PR means there is a substantial change in design or a change in the API. Committers apply their best judgment to determine what constitutes a substantial change. A “minor” PR might be merged without a 24 hour delay, again subject to the judgment of the committer. Examples of potential “minor” PRs are:

  1. Documentation improvements/additions

  2. Small bug fixes

  3. Non-controversial build-related changes (clippy, version upgrades etc.)

  4. Smaller non-controversial feature additions

The good thing about open code and open development is that any issues in one change can almost always be fixed with a follow on PR.

Stale PRs#

Pull requests will be marked with a stale label after 60 days of inactivity and then closed 7 days after that. Commenting on the PR will remove the stale label.

CI Runners#

Runs-On#

We use Runs-On for some actions in the main repository, which run in the ASF AWS account to speed up CI. In forks, these actions run on the default GitHub runners since forks do not have access to ASF infrastructure.

To configure them, we use the following format:

runs-on: ${{ github.repository_owner == 'apache' && format('runs-on={0},family=m8a,cpu=16,image=ubuntu24-full-x64,extras=s3-cache,disk=large,tag=datafusion', github.run_id) || 'ubuntu-latest' }}

This is a conditional expression that uses Runs-On custom runners for the main repository and falls back to the standard GitHub runners for forks. Runs-On configuration follows the Runs-On pattern.

For those actions we also use the Runs-On action, which adds support for external caching and reports job metrics:

- uses: runs-on/action@cd2b598b0515d39d78c38a02d529db87d2196d1e

For the standard GitHub runners, this action will do nothing.

Spot Instances#

By default, Runs-On actions run as spot instances, which means they might occasionally be interrupted. In the CI you would see:

Error: The operation was canceled.

According to Runs-On, spot instance termination is extremely rare for instances running for less than 1h. Those actions will be restarted automatically.

GitHub Runners#

We also use standard GitHub runners for some actions in the main repository; these are also runnable in forks.