In the last few years, the number of programmers concerned about writing structured commit messages have dramatically grown. As exposed by Tim Pope in article readable commit messages are easy to follow when looking through the project history. Moreover the AngularJS contributing guides introduced conventions that can be used by automation tools to automatically generate useful documentation, or by developers during debugging process.
This document borrows some concepts, conventions and even text mainly from these two sources, extending them in order to provide a sensible guideline for writing commit messages.
- allow generating CHANGELOG.md by script
- allow ignoring commits by git bisect (not important commits like formatting)
- provide better information when browsing the history
Consider the following commit messages:
- fix comment stripping
- fixing broken links
- Bit of refactoring
- Check whether links do exist and throw exception
- Fix sitemap include (to work on case sensitive linux)
- Fix small typo in docs widget (tutorial instructions)
- Fix test for scenario.Application - should remove old iframe
- docs - various doc fixes
- docs - stripping extra new lines
- Replaced double line break with single when text is fetched from Google
- Added support for properties in documentation
On the one hand, looking at the first 5 messages is not possible to identify which part of the code had changed. (The remaining messages try to specify where the change is, but they don't share any convention…)
On the other hand, commits introducing formatting changes (adding/removing spaces/empty lines, indentation), missing semi colons, comments, etc are not interesting to debug code or to documenting changes, since no logic change inside them.
Although it is possible to find more information by checking which files had been changed and performing some diffs, it is not practical when looking in git history.
Structural conventions can speed up this process, by allowing filtering. For example, when bisecting, you can ignore files by doing:
git bisect skip $(git rev-list --grep irrelevant <good place> HEAD)<type>(<scope>): <subject> <meta>
<BLANK LINE>
<body>
<BLANK LINE>
<footer>
The first line of a commit message should not be longer than 72 characters, and all the other lines should have a maximum of 100 characters! This allows the message to be easier to read on Github as well as in various git tools.
Subject line may be prefixed for continuous integration purposes. For example, JIRA requires ticket in the beggining of commit message message: "[LHJ-16] fix(compile): add unit tests for windows"
The usage of markup that can be read as plain text (e.g. markdown, reST, etc), specially in body and footer is optional, but useful when auto generating changelogs.
The <type> of a commit message should be a single word or abbreviation drawn
from an ontology, according to the nature of the project.
This document specifies a programming ontology, with the following elements:
- feat: A new feature
- fix: A bug fix
- docs: Documentation only changes
- style: Changes that do not affect the meaning of the code (white-space, formatting, missing semi-colons, etc)
- refactor: A code change that neither fixes a bug nor adds a feature
- perf: A code change that improves performance
- test: Adding missing tests
- build: Changes to the build/compilation/packaging process or auxiliary tools such as documentation generation
- ci: Changes in the continuous integration/delivery setup
In the case a project decides to use a different ontology, this needs to be specified in its contributing guidelines
(usually a CONTRIBUTING.md file in the root of the project),
either by including a modified copy of the text in this document replacing the items in the list above,
or by linking this document and providing a replacement type list. Example text:
<!-- CONTRIBUTING.md -->
# Commit Message Guidelines
This project follows [specific commit message guidelines](... link to this document), with the following allowed types:
* ...: ...
* ...: ...An example of ontology for a book writing project is: typo, restructure, write, move, etc…
Usually it is convenient to mention exactly which part of the code base changed.
The <scope> token is responsible for providing that information.
For example, imagine that a webserver introduces OAuth login via Facebook, a
valid commit message would be:
feat(auth): introduce sign in via Facebook
Please notice that here auth might refer to a conceptual architectural component
or even the basename of an file in the repository. While the granularity of the
scope can vary, it is important for it to be a part of the "common language"
spoken in the project.
Please notice that in some cases the scope is naturally too broad, and therefore not worthy to mention. Examples of this special case:
docs: replace old website URL
refactor: move folder structure to `src` directory layout
Subject line should contains succinct description of the change.
- use imperative, present tense: “change” not “changed” nor “changes”
- don't capitalize first letter
- no dot (.) at the end
Additionally, the end of subject-line may contain hashtags to facilitate changelog generation and bissecting.
#wip- indicate for contributors the feature being implemented is not complete yet. Should not be included in changelogs (just the last commit for a feature goes to the changelog).#irrelevant- the commit does not add useful information. Used when fixing typos, etc... Should not be included in changelogs.
- just as in
<subject>use imperative, present tense: “change” not “changed” nor “changes” - includes motivation for the change and contrasts with previous behavior
- http://365git.tumblr.com/post/3308646748/writing-git-commit-messages
- http://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html
All breaking changes have to be mentioned in footer with the description of the change, justification and migration notes
BREAKING CHANGE: isolate scope bindings definition has changed and
the inject option for the directive controller injection was removed.
To migrate the code follow the example below:
Before:
scope: {
myAttr: 'attribute',
myBind: 'bind',
myExpression: 'expression',
myEval: 'evaluate',
myAccessor: 'accessor'
}
After:
scope: {
myAttr: '@',
myBind: '@',
myExpression: '&',
// myEval - usually not useful, but in cases where the expression is assignable, you can use '='
myAccessor: '=' // in directive's template change myAccessor() to myAccessor
}
The removed `inject` wasn't generaly useful for directives so there should be no code using it.
Closed bugs should be listed on a separate line in the footer prefixed with "Closes" keyword like this:
Closes #234
or in case of multiple issues:
Closes #123, #245, #992
If the commit reverts a previous commit, it should begin with revert:, followed by the header of the reverted commit. In the body it should say: This reverts commit ., where the hash is the SHA of the commit being reverted.
feat($browser): add onUrlChange event (popstate/hashchange/polling)
New $browser event:
- forward popstate event if available
- forward hashchange event if popstate not available
- do polling when neither popstate nor hashchange available
Breaks $browser.onHashChange, which was removed (use onUrlChange instead)
fix($compile): add unit tests for IE9
Older IEs serialize html uppercased, but IE9 does not...
Would be better to expect case insensitive, unfortunately jasmine does
not allow to user regexps for throw expectations.
Closes #392
Breaks foo.bar api, foo.baz should be used instead
feat(directive): add directives disabled/checked/multiple/readonly/selected
New directives for proper binding these attributes in older browsers (IE).
Added coresponding description, live examples and e2e tests.
Closes #351
style($location): add couple of missing semi colons
docs(guide): update fixed docs from Google Docs
Couple of typos fixed:
- indentation
- batchLogbatchLog -> batchLog
- start periodic checking
- missing brace
feat($compile): simplify isolate scope bindings
Change the isolate scope binding options to:
- @attr - attribute binding (including interpolation)
- =model - by-directional model binding
- &expr - expression execution binding
This change simplifies the terminology as well as
number of choices available to the developer. It
also supports local name aliasing from the parent.
BREAKING CHANGE: isolate scope bindings definition has changed and
the inject option for the directive controller injection was removed.
To migrate the code follow the example below:
Before:
scope: {
myAttr: 'attribute',
myBind: 'bind',
myExpression: 'expression',
myEval: 'evaluate',
myAccessor: 'accessor'
}
After:
scope: {
myAttr: '@',
myBind: '@',
myExpression: '&',
// myEval - usually not useful, but in cases where the expression is assignable, you can use '='
myAccessor: '=' // in directive's template change myAccessor() to myAccessor
}
The removed `inject` wasn't generaly useful for directives so there should be no code using it.
Changelogs may contain three sections: new features, bug fixes, breaking changes. This list could be generated by script when doing a release, along with links to related commits. Of course you can edit this change log before actual release, but it could generate the skeleton.
List of all subjects (first lines in commit message) since last release:
git log <last tag> HEAD --pretty=format:%sNew features in this release
git log <last release> HEAD --grep featThanks to @stephenparish for the original text
Hi.
I currently use git not only for programming, but for writing too. Thinking about that, I suggest a format that permits any use of git as a versioning system, not only versioning of code.
Suggested changes:
<type>drawable from any ontology, thus separating format from content.Types for writing could, for example, be ones such as typo, restructure, write, move, etc.
Some other general comments:
<scope>s<meta>tags may be structured; i.e. drawn from an ontologyWhat are your thoughts?