Making and Accepting Pull Requests#

Making a Pull Request#

In the pull request description (editable on GitHub), you should include the following information (some may be omitted if it is e.g. a small bug fix and not a new feature or design change):

  • What does the change do? (description of the new features, changes, or refactorings)

  • Where should the reviewer start the review? (what is the central module that changed, etc.)

  • What is the use case, if it is a new feature?

  • give an example of use or screenshot/plot if applicable

  • Are any new dependencies required? (dependencies should be kept to a minimum, so all new dependences need to be accepted by management)

  • is there a relevant issue open that this addresses? (use the #ISSUENUMBER syntax to link it)

Note that you can include syntax-highlighted code examples by using 3 back-tics:

```python

code here

```

Warning

Note that the continous integration system will run only if you open a pull request from a branch in the main repository. Pull requests opened from a branch from a forked repository cannot run the continous integration system since they cannot access the repository secrets used in some of the workflow (e.g. to download proprietary data).

Continuous integration system#

The Travis continuos integration (CI) system runs tests for pushes on a pull request or the master branch. At the moment the CI system runs the following workflows:

  • lint, which runs the pre-commit hook over the code (runs for all pull requests and push commits to the master branch)

  • pyflakes, which runs pyflakes over the code (runs for all pull requests and push commits to the master branch)

  • docs, which tests if the documentation can be built correctly (runs for all pull requests and push commits to the master branch)

  • tests, which runs unit tests over a set of test files (runs for all pull requests and push commit to the master branch)

  • release-drafter, which drafts release notes (runs only for pushes on the master branch). Note that the release drafter will produce different sections based on the label applied to the pull requests merged into the master branch. See categories in the release drafter configuration file

  • deploy, which deploys automatically a new release to PyPI whenever a tag called v* (e.g. v0.3.4) is pushed to the master branch

Note

If you apply the documentation-only label on a pull request, only the docs workflow will run! This allows to have a faster CI run when changes are made only to the documentation part.

Keep in mind#

  • make sure you remember to update the documentation as well as the code! (see the docs/ directory), and make sure it builds with no errors i.e. run make doc and check the local documentation in your browser by opening the file docs/_build/html/index.html

  • pull requests that cause tests to fail on Travis will not be accepted until those tests pass. For this, check that the tests are passing on your local repository by running make test before pushing new changes to the PR

  • the code style will be checked. If you have pre-commit installed, this will be checked automatically when you try to add a commit, possibly stopping the commit if the checks are not passing

Accepting a Pull Request#

magic-cta-pipe maintainers must do a code review before accepting any pull request. During the review the reviewer can ask for changes to be made, and the requester can simply push them to the branch associated with the request and they will automatically appear (no new pull request needed). The following guidelines should be used to facilitate the review procedure:

  • Perform a scientific or conceptual Review if the request introduces new features, algorithms, or design changes

  • Look at the use case for the proposed change.

    • if the use case is missing, ask for one

    • does it make sense? Is it connected to a goal, requirement, or specification?

  • Perform a Code Review

    • Check that all automatic checks succeeded, if not notify the author and give guidance on how to fix the identified issues.

    • Check that all functions and classes have API documentation in the correct format.

    • Check that there are at least basic unit tests for the added functionality / fixed bug.

    • Check that the API (function and class definitions) is clear and easy to understand.

    • Check for common coding mistakes.

    • Check for obvious performance issues.

    • Check that the code uses the existing features of ctapipe.

    • Check that the code doesn’t introduce new features that are already present in another form.