Milvus is an open source project and we welcome every contributor. Contributions to Milvus are expected to adhere to our Code of Conduct.
This document outlines the conventions about development workflow, commit message formatting, best practices, and other resources you may need while contributing code to Milvus. File an issue or reach out to our engineering team at Slack if you encounter any problem in participating in contribution to Milvus.
For more instruction on setting up your development environment, refer to Milvus Development Guide.
/build: Stores the files to compile Milvus in container and run E2E test.
/cmd: Stores entry module of Milvus service and some toolkits.
/configs: Stores configuration files of Milvus.
/deployments: Stores checklists to deploy Milvus.
/docs: Stores design docs, developing guide, and monitoring docs of Milvus.
/internal: Stores source code of Milvus, including all Golang and CPP codes.
/scripts: Stores scripts for compilation, installation, code check, and test of Milvus.
/tests: Stores E2E and smoke tests for Milvus.
CONTRIBUTING.md: Contains code contribution guideline.
DEVELOPMENT.md: Contains guideline to build Milvus from source code.
Makefile: Entry to run scripts stored under /scripts directory for compilation, test, and code check.
LICENSE: License of the Milvus project.
Generally, we follow the "fork-and-pull" GitHub workflow.
- Fork the repo on GitHub;
- Commit changes to your own branch;
- Submit a pull request (PR) so that we can review your changes.
Remember to sync your forked repository before submitting proposed changes upstream.
See Make Your First Contribution for detailed instructions.
To manage the stadard GitHub PR process, the Milvus project integrates a prow robot (a Kubernetes-based CI/CD system), which will automatically add labels and assign reviewers to your PR. You may also use commands in comment to manage the process. See Command Reference for more information.
Below are the common issues you may encounter in a PR and the corresponding solutions:
DCO check failed - See DCO page to troubleshoot the problem.
Test cases failed - Deal with the test cases that fail on your PR, or re-run jobs to solve the problem if the test cases are irrelevant to the changes you introduced.
Eligible reviewers will be assigned to your PR once it is opened for a thorough code review, looking at correctness, bugs, opportunities for improvement, documentation and comments, and style.
If your PR does not get a response, you can find corresponding reviewers in the #pr-reviews channel on Slack.
Keeping a consistent style for code, code comments, commit messages, and PR descriptions will greatly accelerate your PR review process. We highly recommend you refer to and comply to the following style guides when you put together your pull requests:
Coding style: refer to the Effictive Go Style Guide
Commit message and PR description style: refer to good commit messages
Below, we listed some tips for you while contributing code to Milvus:
Divide a large change into a logical series of smaller ones, each of which is an easily understandable patch.
Squash your commits to help us maintain a clearer Git history.
Write clear and meaningful Git commit messages.
Do not include
fixes #<issue-number>keywords in your commit messages. State them in the PR description instead.
Compile a clear and detailed PR description explaining the reasons for the changes, and ensure that there is sufficient information for the reviewer to understand your PR.
If your PR completely fixes a specific issue, include
fixes #<issue-number>in the PR body. This will automatically close the issue you mentioned when the PR is merged.
Assign appropriate reviewers to your PR under the guidance of the robot in your PR page.
It is contributors' responsibility to ensure that the submitted code changes pass the tests. Contact @milvus-io/sig-testing in comment for help if you encounter a problem.
Code contributed to Milvus will need to complete a variety of tests, each of which has a different objective:
Unit tests: Verifies if functions operate as expected. Golang includes an in-built ability for unit testing via testing package. Each PR has to pass all unit tests before it can be merged. Each unit test code is stored together with the corresponding Golang file, for example: the functions in milvus/internal/allocator/global_id.go will be tested in milvus/internal/allocator/global_id_test.go.
Integration tests: A subset of the end-to-end test to ensure that the PR can be merged with efficiency. These tests are written in Python, stored under the milvus/tests/python_test and milvus/tests20/python_client directories. Run integration tests:
pytest --tags=smoke .
- End-to-end tests: The complete functional test cases. These tests are written in Python, stored under the milvus/tests/python_test and milvus/tests20/python_client directories. Run end-to-end tests:
Continuous integration will run unit tests and integration tests on PRs, and the results will be updated in GitHub Pull Request status. Any failure will block the code merging.