The semantic versioning specification (https://semver.org) gives us a tool sanely evolve software. However, as a library grows and perhaps gets split into multiple smaller ones, it can be difficult to keep track of the semver impact of a series of changes.
semverlog
aims to solve this problem by allowing you to describe each change and its semver impact in a separate file under the .changes
directory.
- Zero configuration:
semverlog
implements a very opinionated workflow. This allows it to operate without a configuration file, avoiding clutter in your repository. - Scalable: With
semverlog
, each change (typically introduced in a single PR), is described in a separate file. This avoids merge conflicts in otherwise often touched files likeCHANGELOG.md
or manifest files. - Correct: One of the biggest challenges in maintaining a large library is correctly tracking semver.
With
semverlog
, you can record the semver impact of your change together with the code. This makes releases stress-free. - Flexible: Despite being free of configuration files,
semverlog
aims to flexible where possible. For example, we don't integrate with package managers, insteadsemverlog
requires you to pass in the current version of the package as an argument. - Contributor-friendly: You can use
semverlog
as a maintainer without forcing your contributors into it. The change file format is simple and the name doesn't matter, meaning contributors can create entries without installing the tool.
- Within your repository, create a directory named
.changes
. - Add a markdown file (name irrelevant) for each change (i.e. pull request).
Upon release time, call semverlog compute-bump-level <CURRENT_VERSION>
to compute the next version bump.
To generate a changelog entry, call semverlog compile-changelog <NEW_VERSION>
.
This will compile all entries within .changes
together, sorted by priority and creation time.
A change file is a markdown file with a yaml front-matter:
---
kind: added|fixed|changed|removed|deprecated|security
breaking: true|false
priority: 0-10
---
<CHANGE_TEXT>
breaking
and priority
are optional.
By default, the changed
and removed
kind are considered a breaking change, unless overridden with breaking: false
.
<CHANGE_TEXT>
will be output verbatim upon semverlog compile-changelog
.
It is recommended to stick to paragraphs and not add any extra headings to avoid formatting problems.
If you want to add headings, take into account that all changes are within sections of level-3 (###
) headings, i.e. ### Added
etc so you want to start your headings at level 4 within the changelog entry.
- Squash-merge your PRs: This will ensure the change file is committed together with the actual change.
- Delete change files within
.changes
after each release.semverlog
will always consider the entire.changes
directory.
- changie: Much more features but also more configuration
- cargo-changelog: No integration with semver (yet)