This document describes some best practices for collaborating on repositories. Following these practices makes it easier for contributors (new and old) to understand what is expected of them. It should be linked to in the README.md.
There are many good practices that this document does not cover. These include other members of the wider community reviewing pull requests (PRs) they are interested in, and maintainers encouraging and supporting people who open issues to make PRs to solve them. This document facilitates these other good practices by clarifying what can seem a mysterious process to those who are unfamiliar with it.
This document is also only intended for community practices, it is not suitable for solo projects with one maintainer.
Interactions with people in the community must always follow the community standards,
including in pull requests, issues, and discussions.
- PRs should match the existing code style present in the file.
- PRs affecting the public API, including adding new features, must update the public documentation.
- Comments and (possibly internal) docstrings should make the code accessible.
- PRs that change code must have appropriate tests.
- Changes to the code must be made via PR, not pushing to master.
- PRs must have 1 approval before they are merged.
- PR authors should not approve their own PRs.
- PRs should pass CI tests before being merged.
- PRs by people without merge rights must have approval from someone who has merge rights (who will usually then merge the PR).
- PRs by people with merge rights must have approval from someone else, who may or may not have merge rights (and then may merge their own PR).
- PRs by people with merge rights should not be merged by people other than the author (just approved).
- A release should be made as soon as possible after a bugfix PR is merged.
- Care and consideration should be given as to when to make a breaking release.
- If the repository is in a state where there are unreleased changes for an extended period of time in preparation for a release, then the version in the Project.toml should be set to the version number of the intended release, with the -DEV suffix.
- The person who merged the PR should register the new release of the package.
- Collaborator merge rights are typically assigned at an Organisational level for all repositories in a GitHub organisation, or at a Team level for a subset of repositories.
- Before becoming a collaborator it is usual to:
- contribute several PRs,
- review constructively and kindly several PRs,
- contribute meaningfully to several discussions on issues.
- You may ask to be added as a collaborator. It is not rude to ask.
This page offers some further guidance on conventions that can be helpful when collaborating on projects. This is an expansion on the Collaborative Practices, with more details and extra guidance. Anything detailed here should be considered less important than the main Collaborative Practices.
-
You should usually open an issue about a bug or possible improvement before opening a PR with a solution.
-
PRs should do a single thing, so that they are easier to review.
- For example, fix one bug, or update compatibility, rather than fixing a bunch of bugs and updating compatibility and adding a new feature.
-
PRs should add tests which cover the new or fixed functionality.
-
PRs that move code should not also change code, so that they are easier to review.
-
If only moving code, review for correctness is not required.
-
If only changing code, then the diff makes it clear what lines have changed.
-
-
PRs with large improvements to style should not also change functionality.
- This is to avoid making large diffs that are not the focus of the PR.
- While it is often helpful to fix a few typos in comments on the way past, it is different to using a regex or formatter on the whole project to fix spacing around operators.
-
PRs introducing breaking changes should make this clear when opening the PR.
-
You should not push commits which commented-out tests.
- If pushing a commit for which a test is broken, use the
@test_broken
macro. - Commenting out tests while developing locally is okay, but committing a commented-out test increases the risk of it silently not being run when it should be.
- If pushing a commit for which a test is broken, use the
-
You should not squash down commits while review is still on-going.
- Squashing commits prevents the reviewer being able to see what commits are added since the last review.
-
You should help review your PRs, even though you cannot approve your own PRs.
- For instance, start the review process by commenting on why certain bits of the code changed, or highlighting places where you would particularly like reviewer feedback.
-
Review comments should be phrased as questions, as it shows you are open to new ideas.
- For instance, “Why did you change this to X? Doesn’t that prevent Y?” rather than “You should not have changed this, it will prevent Y”.
-
Small review suggestions, such as typo fixes, should make use of the
suggested change
feature.- This makes it easier and more likely for all the smaller changes to be made.
-
Reviewers should continue acting as a reviewer until the PR is merged.
- Follow the extension of SemVer 2.0 encoded in Julia package manager Pkg.jl.
- For a version number X.Y.Z, with Major version X, Minor version Y, Patch version Z:
- Post-1.0.0: for breaking changes increment X, for non-breaking new features increment Y, for bug-fixes increment Z.
- Pre-1.0.0: for breaking changes increment Y, for non-breaking (feature or bug-fix) increment Z.
- Introducing deprecations is not breaking; removing deprecations is breaking.
- There is a cost to making breaking releases - downstream packages have to add support for the new version - so there has to be a bigger benefit to making breaking changes.
Following the Collaborative Practices, when there are unreleased changes in the repository for an extended period of time the version number in the Project.toml should be suffixed with -DEV
.
This makes it clear that there are unreleased changes.
Which is useful for many things, including quickly understanding why a bug is still occurring, and working out if a bugfix may need to be backported.
Some more details on the use of -DEV
.
-
After/during/before the PR making the first change of the release, the version number in the Project.toml should be changed to the intended release number should suffixed
-DEV
.- For instance, if the current version is 0.6.3, then the PR making the breaking change could bump it to 0.7.0-DEV.
- Things are more complex if a breaking change is made after the version has been suffixed with
-DEV
for a non-breaking change.- This should be rare since non-breaking changes should be released as soon as possible.
- If it does occur the following rule applies: if all version numbers to the right of the digit you would increment are zero, then you do not need to change the version; otherwise you do.
-
Follow-up PRs can then be made which do not need to increment the version.
-
Once all the follow-up changes have been made, we can make a PR to drop the
-DEV
suffix and make a new release once this PR is merged. -
Note that locally when using
pkg”dev Foo...”
to install particular unreleased versions to an environment, Pkg ignores suffixes to the version number. Pkg treats0.7.0-DEV
identically to0.7.0
. This means you can update the[compat]
section of a group of packages and test them together.
- Generally changing dependency compatibility should be a non-breaking feature.
- i.e. pre-1.0 change patch version number, post-1.0 change the minor version number.
- For instance, adding or removing compatibility with a particular version of a current dependency, which may or may not require internal code changes.
- This also applies when adding or removing packages as dependencies.
- The new feature in question is the ability to use with a different set of packages.
- Changing a dependency to resolve a bug is a bug-fix.
- i.e. pre/post-1.0 change patch version number.
- For instance, if a bug in a downstream dependency is causing a problem in your package restricting compat to not allow that version would be a bug-fix.
- Changing compatibility with dependencies may be a breaking release, if it breaks the user-facing interface.
That is to say if the dependency’s API leaks into your API.
There are three ways that this can happen:
- Reexporting a function that has changed.
- Returning an object of a type that’s behaviour has changed.
- Subtyping an object that has changed.
- Changing Julia version compatibility must be a non-breaking feature.
- It cannot alone be breaking, since Julia versions that are now unsupported will just never see this newer package release.
- Tagging the change as a Minor release makes it possible to release backported bug fixes for users stuck on the old Julia version.
- if the package is pre-1.0, minor releases count as breaking. Therefore, tag the release as a patch release unless one intends to
support earlier versions of Julia with backports (as needed).
For instance, if the current release is
5.4.0
then we can still go back and release5.3.1
.
- if the package is pre-1.0, minor releases count as breaking. Therefore, tag the release as a patch release unless one intends to
support earlier versions of Julia with backports (as needed).
For instance, if the current release is
- Dropping support for earlier versions of Julia has a cost - it prevents users on those versions, such as the Long-Term Support version, from using newer releases of your package - so there should usually be a compelling reason to drop support.
Do not panic, these things sometimes slip through.
It is important to fix it as soon as possible, as otherwise people start using the breaking change, and reverting it later causes more problems (c.f. Murphy's law).
To fix it:
- Make a PR which reverts the PR that made the breaking change.
- Bump the Patch version number in the Project.toml. It was a bug that a breaking API change was made, so a Patch release is correct to fix it.
- Merge the PR and release the new version.
Once the change is reverted you can take stock and decide what to do. There are generally 2 options:
- Make a new PR to reimplement the feature in a non-breaking way.
- Make a new PR which reverts the reversion, bump the version number to signify it as breaking, and release the new breaking version.
Consider a package which is currently on v1.14.2. I made a PR to add a new feature and tagged release v1.15.0. The next evening, we get bug reports that the new feature actually broke lots of real uses.
Maybe I changed what I thought was an internal function, but one that was actually part of the public API; maybe I accidentally changed the return type, and that was something people depended on. Whatever it was, I broke it, and this was not caught in code review.
To fix it, I revert the change, and then tag release v1.15.1. Hopefully, I also can add a test to prevent that part of the API being broken by mistake.
Now I look at my change again. If I can add the same functionality in a non-breaking way - for example, make a new internal function for my use - then I would do so and tag v1.15.2 or v1.16.0 depending on what had to change. If I cannot make an equivalent non-breaking change, then I would have to make the breaking change and tag v2.0.0.
Say you were updating PackageA to support a new version of a dependency, PackageB. For example, you want PackageA v1.1.0 to support PackageB v0.5 and to discontinue supporting v0.4. But say you forgot to remove the compatibility for v0.4, which now no longer works, but other downstream packages that only use v0.4 are now pulling in PackageA v1.1.0 and getting errors.
Simply releasing a patch for PackageA (v1.1.1) that removes support for v0.4 won't work in this instance because downstream packages will continue to pull in v1.1.0. It might seem sufficient to just pin the downstream packages to use v1.0.0 but there may be a lot of them to fix and you can't be certain you're aware of them all. It also does nothing to prevent new compatibility issues arising in future.
To fix this, you should still release a patch of PackageA (v1.1.1) that removes support for v0.4 of PackageB but you should then mark v1.1.0 of PackageA as broken in the registry.
To do this, simply make a PR to the the registry adding yanked = true
to the Version.toml
file under the version causing issues (in this case v1.1.0).
This marks the release as broken and prevents it from being used by any package from then on.
Many of these guidelines can and should be enforced automatically.
- GitHub: Defining the mergeability of pull requests
- Bitbucket: Suggest or require checks before a merge
- GitLab: Status checks that are required to allow a merge requested [WIP]
Everything on this list can, in theory, break users' code. See XKCD#1172. However, we consider changes to these things to be non-breaking from the perspective of package versioning.
-
Bugs: We may make backwards incompatible behavior changes if the current implementation is clearly broken, that is, if it contradicts the documentation or if a well-understood behavior is not properly implemented due to a bug.
-
Internal changes: Non-public API may be changed or removed. The public API is all exported symbols, plus any unexported symbols that are explicitly documented as part of the public API, for instance documented as part of standard usage in the README or hosted documentation.
-
Exception behavior:
-
Exceptions may be replaced with non-error behavior. For instance, we may change a function to compute a result instead of raising an exception, even if that error is documented.
-
Error message text may change.
-
Exception types may change unless the exception type for a specific error condition is specified in the documentation.
-
-
Floating point numerical details: The specific floating point values may change at any time. Users should rely only on approximate accuracy, numerical stability, or statistical properties, not on the specific bits computed.
-
New exports: Adding a new export is never considered breaking. However, one should consider carefully before exporting a commonly used name that might clash with an existing name (especially, if clashing with
Base
). -
New supertypes:
- A new supertype may be added to an existing hierarchy.
That is, changing
A <: B
toA <: B <: C
orA <: C <: B
. This includes adding a supertype to something without one, i.e. with supertypeAny
. - A
Union
constant may be replaced by an abstract type that covers all elements of the union.
- A new supertype may be added to an existing hierarchy.
That is, changing
-
Changes to the string representation: The output of
print
/string
orshow
/repr
on a type may change at any time. Users should not depend on the exact text, but rather on the meaning of the text. Changing the string representation often breaks downstream packages tests, because it is hard to write test-cases that depend only on meaning (though unit tests with mocking can be shielded from this kind of breaking).
(This guidance on non-breaking changes is inspired by https://www.tensorflow.org/guide/versions.)
As mentioned at the top, community repositories following ColPrac, should link to it in their README.md
.
One way to do that is with a GitHub badge.
[![ColPrac: Contributor's Guide on Collaborative Practices for Community Packages](https://img.shields.io/badge/ColPrac-Contributor's%20Guide-blueviolet)](https://github.com/SciML/ColPrac)
In many-cases ColPrac serves in the places of a CONTRIBUTING.md
, having all the common guidance that you would otherwise put there.
If your package has its own CONTRIBUTING.md
then you should also link to ColPrac there, and indicate how the contents of ColPrac relates to the CONTRIBUTING.md
.
For example by stating:
We follow the ColPrac guide for collaborative practices. New contributors should make sure to read that guide. Below are some additional practices we follow.