[ci skip]
8.6 KiB
Maintaining Shoulda Matchers
As maintainers of the gem, this is our guide. Most of the steps and guidelines in the Contributing document apply here, including how to set up your environment, write code to fit the code style, run tests, craft commits and manage branches. Beyond this, this document provides some details that would be too low-level for contributors.
Communication
We have several ways that we can communicate with each other:
- In planning major releases, it can be helpful to create a new issue outlining the changes as well as steps needed to launch the release. This serves both as an announcement to the community as well as an area to keep a checklist.
- To track progress for the next release, GitHub milestones are useful.
- To track progress on the movement of issues, labels are useful.
- To communicate small-scale changes, pull requests are effective, as mentioned above.
- To communicate large-scale changes or explain topics, email is best.
Managing the community
As anyone who has played a sim game before, it's important to make your patrons happy. We do this by:
- Answering questions from members of the community
- Closing stale issues and feature requests
- Keeping the community informed by ensuring that the changelog is up to date
- Ensuring that the inline documentation, as well as the docsite, is kept up to date
Workflow
We generally follow GitHub Flow. The master
branch is the main line, and all
branches are cut from and get merged back into this branch. Generally, the
workflow is as follows:
- Cut a feature or bugfix branch from this branch.
- Upon completing a branch, create a PR and ask another maintainer to approve it.
- Try to keep the commit history as clean as possible. Before merging, squash "WIP" or related commits together and rebase as needed.
- Once your PR is approved and you've cleaned up your branch, you're free to merge it in.
Architecture
Besides the matchers, there are files in lib
which you may need to reference
or update:
lib/shoulda/matchers/doublespeak*
— a small handrolled mocking library which is used by thepermit
matcherlib/shoulda/matchers/util*
— extra methods which are used in various places to detect library versions, wrap/indent text, and more
Updating the changelog
After every user-facing change makes it into master, we make a note of it in the
changelog, which for historical reasons is kept in NEWS.md
. The changelog is
sorted in reverse order by release version, with the topmost version as the next
release (tagged as "(Unreleased)").
Within each version, there are five available categories you can divide changes into. They are all optional but they should appear in this order:
- Backward-compatible changes
- Deprecations
- Bug fixes
- Features
- Improvements
Within each category section, the changes relevant to that category are listed in chronological order.
For each change, provide a human-readable description of the change as well as a linked reference to the PR where that change emerged (or the commit ID if no such PR is available). This helps users cross-reference changes if they need to.
Documentation
Generating documentation
As mentioned in the Contributing document, we use YARD for documentation. YARD
is configured via .yardopts
to process the Ruby files in lib/
as well as
NEWS.md
and the Markdown files in docs/
and write the documentation in HTML
form to doc
. This command will do exactly that:
bundle exec yard doc
However, if you're actively updating the documentation, it's more helpful to launch a process that will watch the aforementioned source files for changes and generate the HTML for you automatically:
bundle exec rake docs:autogenerate
Whichever approach you take, you can view the generated docs locally by running:
open doc/index.html
About the docsite
The docfiles that YARD generates are published to the docsite, which is located at:
The docsite is hosted on GitHub Pages*. As such, the gh-pages
branch hosts its
code. Usually you don't need to update this branch directly unless you want to
make changes to the homepage itself. As for the documentation, it is hosted
one level deeper:
https://matchers.shoulda.io/docs
The URL above actually links to a bare-bones HTML page which merely serves to
automatically redirect the visitor to the docs for the latest published version
of the gem. This version is hardcoded in the HTML page, but is also updated
automatically by the docs:publish
and docs:publish_latest
tasks.
* thoughtbot owns https://shoulda.io, and
they've got matchers.shoulda.io
set up on the DNS level as an alias for
thoughtbot.github.io/shoulda-matchers
.
Versioning
Naming a new version
As designated in the README, we follow SemVer 2.0. This offers a meaningful baseline for deciding how to name versions. Generally speaking:
- We bump the "major" part of the version if we're introducing backward-incompatible changes (e.g. changing the API or core behavior, removing parts of the API, or dropping support for a version of Ruby).
- We bump the "minor" part if we're adding a new feature (e.g. adding a new matcher or adding a new qualifier to a matcher).
- We bump the "patch" part if we're merely including bugfixes.
In addition to major, minor, and patch levels, you can also append a
suffix to the version for pre-release versions. We usually use this to issue
release candidates prior to an actual release. A version number in this case
might look like 4.0.0.rc1
.
Releasing a new version
Releasing a new version is very simple:
-
First, you'll want to be given ownership permissions for the Ruby gem itself. If you want to give someone else these rights, you can use:
gem owner shoulda-matchers -a <email address>
-
Next, you'll want to update the
VERSION
constant inlib/shoulda/matchers/version.rb
. This constant is referenced in the gemspec and is used in the Rake tasks to publish the gem on RubyGems as well as generate documentation. -
Finally, you'll want to run:
rake release
This will not only push the gem to RubyGems, but also update the docsite.
Re-publishing docs
In general you'll use the release
task to update the docsite, but there may be
situations where you'll need to do it manually.
You can re-publish the docs for the latest version (as governed by
lib/shoulda/matchers/version.rb
) by running:
bundle exec rake docs:publish_latest
This will update the auto-redirect on the index page to the latest version. For instance, if the latest version were 4.0.0, this command would publish the docs at https://matchers.shoulda.io/docs/v4.0.0 and simultaneously redirect https://matchers.shoulda.io/docs to this location.
However, if you want to publish the docs for a version but manually set the auto-redirected version, you can run this instead:
bundle exec rake docs:publish[version, latest_version]
Here, version
and latest_version
are both version strings. For instance, you
might say:
bundle exec rake docs:publish[4.0.0, 3.7.2]
This would publish the docs for 4.0.0 at https://matchers.shoulda.io/docs/v4.0.0, but redirect https://matchers.shoulda.io/docs to https://matchers.shoulda.io/docs/v3.7.2.
Addendum: Labels
Considering that we work on the gem in our spare time, we've found labels to be useful for cataloguing and marking progress. Over time we've added quite a collection of labels. Here's a quick list:
Labels for issues
- Issue: Bug
- Issue: Feature Request
- Issue: Need to Investigate — if we don't know whether a bug is legitimate or not
- Issue: PR Needed — perhaps unnecessary, but it does signal to the community that we'd love a PR
Labels for PRs
- PR: Bugfix
- PR: Feature
- PR: Good to Merge — most of the time not necessary, but can be helpful in a code freeze before a release to mark PRs that we will include in the next release
- PR: In Progress — used to mark PRs that are still being worked on by the PR author
- PR: Needs Documentation
- PR: Needs Review
- PR: Needs Tests
- PR: Needs Updates Before Merge — along the same lines as the other "Needs" tags, but more generic
Generic labels
- Blocked
- Documentation
- Needs Decision
- Needs Revisiting
- Question
- Rails X
- Ruby X.Y
- UX