Thank you for taking the time to contribute. We appreciate it!
The GA4GH schemas define an API for sharing genomic and phenotypic data. Supporting the schemas is a reference server, which acts as a useful, complete implementation of the API, and a compliance suite, which provides data and functions to test an API implementation. Development of these three aspects of the project are coordinated, so that changes to the schemas are reflected in the creation of a working prototype within the reference server and compliance tests that establish conformance. Coordination in this way ensures the alliance is building software and standards that work for genomics.
There are two ways to contribute to the effort - via issues, which are used for discussion, and pull requests, which are concrete proposals of change.
The project's Issues Page is a forum to discuss both major and minor issues related to developing the standards, formats, and APIs. It also serves as the means for collaborating with the group and discussing contributions that will ultimately lead to changes to the formats and APIs. See the Issue section below for specifics on how issues are resolved by the community. Examples of the type of issues that can be submitted are:
- Identify use cases that will shape the standards and APIs
- How to add or delete objects and/or object attributes
- How a particular attribute should be defined
- Report bugs you encounter when using the reference implementations
The way to contribute development effort and code to the project is via GitHub pull requests. GitHub provides a nice overview on how to create a pull request. Contributions typically require pull requests to each of the schemas, server and compliance repositories, although pull requests to the server may merely improve the code without affecting the API, and therefore changing the schemas or compliance tests. A set of branches across the repositories each with the same name is a branch set, e.g. the master branch in each repository forms the master branch set.
Some general rules to follow:
- Fork the main project into your personal GitHub space to work on.
- Create a branch for each update that you're working on. These branches are often called "feature" or "topic" branches. Any changes that you push to your feature branch will automatically be shown in the pull request.
- If necessary, replicate the last two steps' fork-and-branch process for each of the compliance and server repositories to create a branch set - each constituent repository branch will be identically named.
- Coordinate pull requests across the branch set by making them as simultaneously as possible, and cross referencing them.
- Keep your pull requests as small as possible. Large pull requests are hard to review. Try to break up your changes into self-contained and incremental pull requests.
- The first line of commit messages should be a short (<80 character) summary, followed by an empty line and then any details that you want to share about the commit.
- Please try to follow the existing syntax style
When you submit or change your pull request, the Travis build system will automatically run tests to ensure valid schema syntax. If your pull request fails to pass tests, review the test log, make changes and then push them to your feature branch to be tested again.
Once a pull request or issue have been submitted, anyone can comment or vote on to express their opinion following the Apache voting system. Quick summary:
- +1 something you agree with
- -1 if you have a strong objection to an issue, which will be taken very seriously. A -1 vote should provide an alternative solution.
- +0 or -0 for neutral comments or weak opinions.
- It's okay to have input without voting
- Silence gives assent
A pull request with at least two +1 votes, no -1 votes, that has been open for at least 3 days, and whose cross-referenced pull requests to server and compliance have similarly been upvoted is ready to be merged. The merge should be done by someone from a different organization than the submitter. (We sometimes waive the 3 days for cosmetic-only changes -- use good judgment.) A pull request to either the schemas, servers or compliance repository that involves changes to the others should not be merged without coordinating, mergable pull requests to the other repositories. Conversely, when merging a pull request the other pull requests in the branch set must be merged at the same time (In practise, when merging a branch set including the schemas repository, merge the pull request to schemas first to avoid the continuous integration build issues). If an issue gets any -1 votes, the comments on the issue need to reach consensus before the issue can be resolved one way or the other. There isn't any strict time limit on a contentious issue.
The project will strive for full consensus on everything until it runs into a problem with this model.
The current code conventions for the source files are as follows:
- Follow the protocol buffers style guide
- Use two-space indentation, and no tabs.
- Hard-wrap code to 80 characters per line.
- Comments may use reStructuredText mark up.
The goal of GA4GH is to define an interoperable API specification. To achieve this, the intent, rationale, and semantics of all data objects and operations need to be clearly and precisely defined. Decisions that are not captured in documentation are lost.
All schemas defined in GA4GH must include normative documentation. This should consist of overview and design documentation as well as documentation in the schemas. This documentation should explain the goals, overall design, and rationale for decisions that were made. It must be addressed to both client and server developer audiences. It may cite published papers or stable web documentation.
Overview documentation should be in ReStructuredText markdown format (*.rst) in the doc/source/ directory. This format was chosen for inclusion in a Sphinx-based documentation system that is under development. Graphics are encouraged and should have a source to drawings in SVG format that will be converted to PNG by the documentation build.
Developers are encouraged to get feedback on design documentation before developing and implementing schemas.
If you wish to collaborate on a new feature with other GA4GH members you can ask that a topic branch set be created. This will generally involve the creation of identically named topic branches for each of the schema, compliance and server repositories. Since Github does not allow pull requests against branches that do not yet exist, you will have to create an issue asking for the topic branch set to be created.
Once a topic branch set exists, pull requests can be made against it in the usual way. It may also be brought up to date with new changes merged into master by anyone with commit access, if the changes produce merely a fast-forward merge for each constituent branch. However, if changes from the master branch create a new merge commit in or or more of the repositories, that commit needs to be reviewed in a pull request.
Changes made in a topic branch set can be merged into master by creating and then resolving in the normal way a pull request against the master branch set, irrespective of ownership by a task-team (see below) or not.
Topic branch sets that have been merged into master and that are no longer being developed upon should be deleted (they will still appear in the git history), this can be achieved by consensus of those working on the topic branch set, e.g. a specific task-team.
Frequently topic branch sets are developed by members working within a task team of the Data Working Group. Topic branch sets developed by a task team should be prefixed with the task team's name followed by a dash, e.g. reads-foo, refVar-bar. To enable speedy development, task-teams may set their own rules for contribution to the topic branch sets they own. This allows task-teams to develop features in the view of the larger group, while potentially being unencumbered by the more lengthy process of standard issue resolution.
From time to time the group will make a release. This is achieved by creating a branch set including all the repositories named "release-foo", where foo is the release name. Only bug fixes are allowed to release branch sets. To refer to a specific version of a release branch set either the commit id can be used, or alternatively (better), a tag can be created (which should be replicated across repositories).
As projects mature, the need to have a standing Data Working Group task team with regular teleconferences and meetings will decline. Mature task teams will enter a *maintenance mode which will entail the following:
- A task team chair will be appointed to regularly review the
ga4gh#schemasGithub Issues for issue that would effect the outcomes of the retired task team. - The task team chair will tag those issues with the retired group's label.
- Minor pull requests (e.g. documentation enhancements) would follow the same issue resolution process as outlined above.
- Major pull requeusts (e.g. API additions or changes) would be additionally labeled 'major change', and the retired task team chair will reach out to that list of DWG members and request comment on the issue.