The Apache Beam community welcomes contributions from anyone with a passion for data processing! Beam has many different opportunities for contributions – write new examples, add new user-facing libraries (new statistical libraries, new IO connectors, etc), work on the core programming model, build specific runners (Apache Apex, Apache Flink, Apache Spark, Google Cloud Dataflow, etc), or participate on the documentation effort.
We use a review-then-commit workflow in Beam for all contributions.
For larger contributions or those that affect multiple components:
For all contributions:
We look forward to working with you!
If interested, you can also join the other mailing lists.
We use the Apache Software Foundation’s JIRA as an issue tracking and project management tool, as well as a way to communicate among a very diverse and distributed set of contributors. To be able to gather feedback, avoid frustration, and avoid duplicated efforts all Beam-related work should be tracked there.
If you do not already have an Apache JIRA account, sign up here.
If a quick search doesn’t turn up an existing JIRA issue for the work you want to contribute, create it. Please discuss your idea with a committer or the component lead in JIRA or, alternatively, on the developer mailing list.
If there’s an existing JIRA issue for your intended contribution, please comment about your intended work. Once the work is understood, a committer will assign the issue to you. (If you don’t have a JIRA role yet, you’ll be added to the “contributor” role.) If an issue is currently assigned, please check with the current assignee before reassigning.
For moderate or large contributions, you should not start coding or writing a design document unless there is a corresponding JIRA issue assigned to you for that work. Simple changes, like fixing typos, do not require an associated issue.
We don’t have an official IRC channel. Most of the online discussions happen in the Apache Beam Slack channel. If you want access, you need to send an email to the user mailing list to request access.
Chat rooms are great for quick questions or discussions on specialized topics. Remember that we strongly encourage communication via the mailing lists, and we prefer to discuss more complex subjects by email. Developers should be careful to move or duplicate all the official or useful discussions to the issue tracking system and/or the dev mailing list.
To avoid potential frustration during the code review cycle, we encourage you to clearly scope and design non-trivial contributions with the Beam community before you start coding.
Generally, the JIRA issue is the best place to gather relevant design docs, comments, or references. It’s great to explicitly include relevant stakeholders early in the conversation. For designs that may be generally interesting, we also encourage conversations on the developer’s mailing list.
We suggest using Google Docs for sharing designs that may benefit from diagrams or comments. Please remember to make the document world-commentable and add a link to it from the relevant JIRA issue. You may want to start from this template.
To contribute code to Apache Beam, you’ll have to do a few administrative steps once, and then follow a few guidelines for each contribution.
When developing a new
PTransform, consult the PTransform Style Guide.
Apache Software Foundation (ASF) desires that all contributors of ideas, code, or documentation to the Apache projects complete, sign, and submit an Individual Contributor License Agreement (ICLA). The purpose of this agreement is to clearly define the terms under which intellectual property has been contributed to the ASF and thereby allow us to defend the project should there be a legal dispute regarding the software at some future time.
We require you to have an ICLA on file with the Apache Secretary for larger contributions only. For smaller ones, however, we rely on clause five of the Apache License, Version 2.0, describing licensing of intentionally submitted contributions and do not require an ICLA in that case.
We use GitHub’s pull request functionality to review proposed code changes.
If you do not already have a personal GitHub account, sign up here.
Go to the Beam GitHub mirror and fork the repository to your own private account. This will be your private workspace for staging changes.
You are now ready to create the development environment on your local machine. Feel free to repeat these steps on all machines that you want to use for development.
We assume you are using SSH-based authentication with GitHub. If necessary, exchange SSH keys with GitHub by following their instructions.
Clone Beam’s read-only GitHub mirror.
$ git clone https://github.com/apache/beam.git $ cd beam
Add your forked repository as an additional Git remote, where you’ll push your changes.
$ git remote add <GitHub_user> email@example.com:<GitHub_user>/beam.git
You are now ready to start developing!
We recommend setting up a virtual envioment for developing Python SDK. Please see instructions available in Quickstart (Python) for setting up a virtual environment.
Depending on your preferred development environment, you may need to prepare it to develop Beam code.
To configure annotation processing in IntelliJ:
IntelliJ supports checkstyle within the IDE using the Checkstyle-IDEA plugin.
You can also scan an entire module by opening the Checkstyle tools window and clicking the “Check Module” button. The scan should report no errors.
Note: Selecting “Check Project” may report some errors from the archetype modules as they are not configured for Checkstyle validation.
IntelliJ supports code styles within the IDE. Use one of the following to ensure your code style matches the project’s checkstyle enforcements.
Use a recent Eclipse version that includes m2e. Currently we recommend Eclipse Neon. Start Eclipse with a fresh workspace in a separate directory from your checkout.
Install m2e-apt: Beam uses apt annotation processing to provide auto generated code. One example is the usage of Google AutoValue. By default m2e does not support this and you will see compile errors.
Help -> Eclipse Marketplace -> Search for “m2 apt” -> Install m2e-apt 1.2 or higher
Activate the apt processing
Window -> Preferences -> Maven -> Annotation processing -> Switch to Experimental: Delegate annotation processing … -> Ok
Import the beam projects
File -> Import… -> Existing Maven Projects -> Browse to the directory you cloned into and select “beam” -> make sure all beam projects are selected -> Finalize
You now should have all the beam projects imported into eclipse and should see no compile errors.
Eclipse supports checkstyle within the IDE using the Checkstyle plugin.
Eclipse supports code styles within the IDE. Use one of the following to ensure your code style matches the project’s checkstyle enforcements.
You’ll work on your contribution in a branch in your own (forked) repository. Create a local branch, initialized with the state of the branch you expect your changes to be merged into. Keep in mind that we use several branches, including
master, feature-specific, and release-specific branches. If you are unsure, initialize with the state of the
$ git fetch --all $ git checkout -b <my-branch> origin/master
At this point, you can start making and committing changes to this branch in a standard way.
Periodically while you work, and certainly before submitting a pull request, you should update your branch with the most recent changes to the target branch.
$ git pull --rebase
Remember to always use
--rebase parameter to avoid extraneous merge commits.
Then you can push your local, committed changes to your (forked) repository on GitHub. Since rebase may change that branch’s history, you may need to force push. You’ll run:
$ git push <GitHub_user> <my-branch> --force
All code should have appropriate unit testing coverage. New code should have new tests in the same contribution. Bug fixes should include a regression test to prevent the issue from reoccurring.
For contributions to the Java code, run unit tests locally via Maven.
$ mvn clean verify
For contributions to the Python code, you can use command given below to run unit tests locally. If you update any of the cythonized files in Python SDK, you must install “cython” package before running following command to properly test your code. We recommend setting up a virtual environment before testing your code.
$ python setup.py test
You can use following command to run a single test method.
$ python setup.py test -s <module>.<test class>.<test method>
To Check for lint errors locally, install “tox” package and run following command.
$ pip install tox $ tox -e lint
Beam supports running Python SDK tests using Maven. For this, navigate to root directory of your Apache Beam clone and execute following command. Currently this cannot be run from a virtual environment.
$ mvn clean verify -pl sdks/python
Once the initial code is complete and the tests pass, it’s time to start the code review process. We review and discuss all code, no matter who authors it. It’s a great way to build community, since you can learn from other developers, and they become familiar with your contribution. It also builds a strong project by encouraging a high quality bar and keeping code consistent throughout the project.
Organize your commits to make a committer’s job easier when reviewing. Committers normally prefer multiple small pull requests, instead of a single large pull request. Within a pull request, a relatively small number of commits that break the problem into logical steps is preferred. For most pull requests, you’ll squash your changes down to 1 commit. You can use the following command to re-order, squash, edit, or change description of individual commits.
$ git rebase -i origin/master
You’ll then push to your branch on GitHub. Note: when updating your commit after pull request feedback and use squash to get back to one commit, you will need to do a force submit to the branch on your repo.
Navigate to the Beam GitHub mirror to create a pull request. The title of the pull request should be strictly in the following format:
[BEAM-<JIRA-issue-#>] <Title of the pull request>
Please include a descriptive pull request message to help make the comitter’s job easier when reviewing. It’s fine to refer to existing design docs or the contents of the associated JIRA as appropriate.
If you know a good committer to review your pull request, please make a comment like the following. If not, don’t worry – a committer will pick it up.
Hi @<GitHub-committer-username>, can you please take a look?
When choosing a committer to review, think about who is the expert on the relevant code, who the stakeholders are for this change, and who else would benefit from becoming familiar with the code. If you’d appreciate comments from additional folks but already have a main committer, you can explicitly cc them using
During the code review process, don’t rebase your branch or otherwise modify published commits, since this can remove existing comment history and be confusing to the committer when reviewing. When you make a revision, always push it in a new commit.
Our GitHub mirror automatically provides pre-commit testing coverage using Jenkins. Please make sure those tests pass; the contribution cannot be merged otherwise.
Once the committer is happy with the change, they’ll respond with an LGTM (“looks good to me!”). At this point, the committer will take over, possibly make some additional touch ups, and merge your changes into the codebase.
In the case the author is also a committer, either can merge the pull request. Just be sure to communicate clearly whose responsibility it is in this particular case.
Thank you for your contribution to Beam!
Once the pull request is merged into the Beam repository, you can safely delete the branch locally and purge it from your forked repository.
From another local branch, run:
$ git fetch --all $ git branch -d <my-branch> $ git push <GitHub_user> --delete <my-branch>
The community will close stale pull requests in order to keep the project healthy. A pull request becomes stale after its author fails to respond to actionable comments for 60 days. Author of a closed pull request is welcome to reopen the same pull request again in the future. The associated JIRAs will be unassigned from the author but will stay open.
Once the code has been peer reviewed by a committer, the next step is for the committer to merge it into the authoritative Apache repository, not the read-only GitHub mirror. (In the case that the author is also a committer, it is acceptable for either the author of the change or committer who reviewed the change to do the merge. Just be explicit about whose job it is!)
Pull requests should not be merged before the review has received an explicit LGTM from another committer. Exceptions to this rule may be made rarely, on a case-by-case basis only, in the committer’s discretion for situations such as build breakages.
Committers should never commit anything without going through a pull request, since that would bypass test coverage and potentially cause the build to fail due to checkstyle, etc. In addition, pull requests ensure that changes are communicated properly and potential flaws or improvements can be spotted. Always go through the pull request, even if you won’t wait for the code review. Even then, comments can be provided in the pull requests after it has been merged to work on follow-ups.
Committing is currently a manual process, but we are investigating tools to automate pieces of this process.
Add the Apache Git remote in your local clone, by running:
$ git remote add apache https://git-wip-us.apache.org/repos/asf/beam.git
We recommend renaming the
origin remote to
github, to avoid confusion when dealing with this many remotes.
$ git remote rename origin github
github remote, add an additional fetch reference, which will cause every pull request to be made available as a remote branch in your workspace.
$ git config --local --add remote.github.fetch \ '+refs/pull/*/head:refs/remotes/github/pr/*'
You can confirm your configuration by running the following command.
$ git remote -v apache https://git-wip-us.apache.org/repos/asf/beam.git (fetch) apache https://git-wip-us.apache.org/repos/asf/beam.git (push) github https://github.com/apache/beam.git (fetch) github https://github.com/apache/beam.git (push) <username> firstname.lastname@example.org:<username>/beam.git (fetch) <username> email@example.com:<username>/beam.git (push)
If you are merging a larger contribution, please make sure that the contributor has an ICLA on file with the Apache Secretary. You can view the list of committers here, as well as ICLA-signers who aren’t yet committers.
For smaller contributions, however, this is not required. In this case, we rely on clause five of the Apache License, Version 2.0, describing licensing of intentionally submitted contributions.
Before merging, please make sure that Jenkins tests pass, as visible in the GitHub pull request. Do not merge the pull request otherwise.
At some point in the review process, you should take the pull request over and complete any outstanding work that is either minor, stylistic, or otherwise outside the expertise of the contributor.
Fetch references from all remote repositories, and checkout the specific pull request branch.
$ git fetch --all $ git checkout -b finish-pr-<pull-request-#> github/pr/<pull-request-#>
At this point, you can commit any final touches to the pull request. For example, you should:
You will often need the following command, assuming you’ll be merging changes into the
$ git rebase -i apache/master
Please make sure to retain authorship of original commits to give proper credit to the contributor. You are welcome to change their commits slightly (e.g., fix a typo) and squash them, but more substantive changes should be a separate commit and review.
Once you are ready to merge, fetch all remotes, checkout the destination branch and merge the changes.
$ git fetch --all $ git checkout apache/master $ git merge --no-ff -m 'This closes #<pull-request-#>' finish-pr-<pull-request-#>
--no-ff option and the specific commit message “This closes #<pull request #>” – it ensures proper marking in the tooling. It would be nice to include additional information in the merge commit message, such as the title and summary of the pull request.
At this point, you want to ensure everything is right. Test it with
mvn verify. Run
git log --graph, etc. When you are happy with how it looks, push it. This is the point of no return – proceed with caution.
$ git push apache HEAD:master
Done. You can delete the local
finish-pr-<pull-request-#> branch if you like.
The project management committee (PMC) can grant more rights to a contributor, such as commit access or decision power, and recognize them as new committers or PMC members.
The PMC periodically discusses this topic and privately votes to grant more rights to a contributor. If the vote passes, the contributor is invited to accept or reject the nomination. Once accepted, the PMC announces the decision publicly and updates the list of team member accordingly.
The key to the selection process is Meritocracy, literally government by merit. Contributors earn merit in many ways: contributing code, testing releases, participating in documentation effort, answering user questions, debating design proposals, triaging issues, evangelizing the project, growing user base, and any other action that benefits the project as a whole.
Therefore, there isn’t a single committer bar, e.g., a specific number of commits. In most cases, new committers will have a combination of different types of contributions that are hard to compare to each other. Similarly, there isn’t a single PMC bar either. The PMC discusses all contributions from an individual, and evaluates the overall impact across all the dimensions above.
Nothing gives us greater joy than recognizing new committers or PMC members – that’s the only way we can grow. If there’s ever any doubt about this topic, please email dev@ or private@ and we’ll gladly discuss.
The directions above assume you are submitting code to the
master branch. In addition, there are a few other locations where code is maintained. Generally these follow the same engage-design-code-review-commit process as above, with some minor adjustments to commands.
Some larger features are developed on a feature branch before being merged into
master. In particular, this is often used for initial development of new components like SDKs or runners.
To contribute code on a feature branch, use the same process as above, but replace
master with the name of the branch.
Since feature branches are often used for new components, you may find that there is no committer familiar with all the details of the new language or runner. In that case, consider asking someone else familiar with the technology to do an initial review before looping in a committer for a final review and merge.
If you are working on a feature branch, you’ll also want to frequently merge in
master in order to prevent life on the branch from deviating too
far from reality. Like all changes, this should be done via pull request. It
is permitted for a committer to self-merge such a pull request if there are no
conflicts or test failures. If there are any conflicts of tests that need
fixing, then those should get a full review from another committer.
In order for a feature branch to be merged into
master, new components and major features should aim to meet the following guidelines.
A new runner should:
A new SDK should:
The Beam website is in the Beam Site GitHub mirror repository in the
asf-site branch (not
Issues are tracked in the website component in JIRA.
The README file in the website repository has more information on how to set up the required dependencies for your development environment.
The general guidelines for cloning a repository can be adjusted to use the
asf-site branch of
$ git clone -b asf-site https://github.com/apache/beam-site.git $ cd beam-site $ git remote add <GitHub_user> firstname.lastname@example.org:<GitHub_user>/beam-site.git $ git fetch --all $ git checkout -b <my-branch> origin/asf-site
While you are working on your pull request, you can test and develop live by running the following command in the root folder of the website:
$ bundle exec jekyll serve --incremental
Jekyll will start a webserver on port 4000. As you make changes to the content, Jekyll will rebuild it automatically.
In addition, you can run the tests to valid your links using:
$ bundle exec rake test
Both of these commands will cause the
content/ directory to be generated. Merging autogenerated content can get tricky, so please leave this directory out of your commits and pull request by doing:
$ git checkout -- content
When you are ready, submit a pull request using the Beam Site GitHub mirror, including the JIRA issue as usual.
During review, committers will patch in your PR, generate the static
content/, and review the changes.
Follow the same committer process as above, but using repository
apache/beam-site and branch
https://gitbox.apache.org/repos/asf/beam-site.git as the address for
git remote add apache.
In addition, the committer is responsible for doing the final
bundle exec jekyll build to generate the static content, so follow the instructions above to install
This command generates the
content/ directory. The committer should add and commit the content related to the PR.
$ git add content/<files related to the pr> $ git commit -m "Regenerate website"
Finally you should merge the changes into the
asf-site branch and push them into the