Working with Git
Default branch name
Your repository’s default branch should be called main
.
To configure git to use main
for new repositories, first make sure you are using git 2.28 or later, then run:
git config --global init.defaultBranch main
To migrate existing repositories from master
to main
, you can use
GitHub’s branch renaming
tool.
Before you rename your default branch, you should consider the impact this will have on:
- internal users
- external users
- your continuous integration (CI) system (for example, Concourse pipelines or GitHub actions)
You can configure some CI systems to check out the default branch rather than hardcoding a branch name:
- GitHub Actions supports a
$default-branch
macro - concourse-git-resource: if you do not set
branch
, it will check out the default branch
Commits
Atomic commits
Each commit should be atomic - this means that each commit:
- is self-contained
- only includes changes that achieve a specific step
- should be in a logical order
As Anna Shipman describes it, they should tell a story.
Example
From Anna’s blogpost, we see an example of how a single commit with 51 files changed was broken down into a series of 8 commits implementing those changes instead.
Commit messages
Writing good commit messages is important. Not just for yourself, but for other developers on your project. This includes:
- new (or recently absent) developers who want to get up to speed on progress
- interested external parties who want to follow progress of the project
- people in the public (remember, we code in the open) who want to see our work, or learn from our practices
- any future developers (including yourself) who want to see why a change was made
Capturing context around a change allows people to understand why a particular implementation decision was made, much like an architecture decision record. We’re being kinder to our future selves.
Recommended blog posts on commit messages
- How to Write a Git Commit Message
- 5 useful tips for a better commit message
- Every line of code is always documented
Content
A good commit message briefly summarises the “what” for scanning purposes, but also includes the “why”. If the “what” in the message is not enough, the diff is there as a fallback. This is not true for the “why” of a change - this can be much harder or impossible to reconstruct, but is often of great significance.
Example
Set cache headers
prefer:
Set cache headers
IE 6 was doing foo, so we need to do X.
See http://example.com/why-is-this-broken for more details.
Links to issue trackers
A link to a ticket in an issue tracker should not be seen as an alternative to writing a commit message.
While a link can add some extra context for people reviewing a pull-request, the commit message should stand on its own. There’s no guarantee that the link will continue to work in the future when someone is looking through the commit history to understand why a change was made.
If you are adding a link to a publicly viewable repository, ensure that the linked ticket is publicly viewable (and likely to remain so).
Structure
Commit messages should start with a one-line summary no longer than 50 characters. Various Git tools (including GitHub) use this as the commit summary, so you should format it like an email subject, with a leading capital and no full stop. The Git convention is to write these in the present tense. For example:
Leverage best-of-breed synergies going forward
You should leave a blank line before the rest of the commit message, which you should wrap at around 72 characters: this makes it easier to view commit messages in a terminal.
Example
Taken from Tim Pope’s guidelines.
Capitalized, short (50 chars or less) summary
More detailed explanatory text, if necessary. Wrap it to about 72 characters or so. In some contexts, the first line is treated as the subject of an email and the rest of the text as the body. The blank line separating the summary from the body is critical (unless you omit the body entirely); tools like rebase can get confused if you run the two together.
Write your commit message in the present tense: “Fix bug” and not “Fixed bug.” This convention matches up with commit messages generated by commands like git merge and git revert.
Further paragraphs come after blank lines.
- Bullet points are okay, too
- Typically a hyphen or asterisk is used for the bullet, preceded by a single space, with blank lines in between, but conventions vary here
- Use a hanging indent
Branching/merging conventions
You may often choose to work on a particular feature on a “feature branch”
rather than directly on main
. Indeed, given how cheap branches are in Git,
using branches is positively encouraged.
You are encouraged to make liberal use of Git’s history rewriting
features while working
locally, in order to arrange your commits into appropriate logical chunks that
will make sense to your fellow developers. In particular, you may find
git rebase --interactive
very useful. You are also encouraged to avoid merge
commits and use git rebase main
instead. However, you should not rewrite commits
that have been pushed unless you:
- are very sure that no-one else will be affected by you rewriting the branch history
- have an extremely good reason. For example: someone has committed sensitive information (personally identifiable data, passwords and suchlike) and it needs purging from history
When in doubt you should err towards smaller commits, which can be rebased together later. It’s harder to break large commits out into smaller chunks.
The smaller commits should still be logical chunks, but this will give context
for a more specific change and make git tools like annotate
and log
more
useful.
When merging from a feature branch to main (or any other mainline development
branch), in particular one that has previously been shared with colleagues, you
should use git merge
‘s --no-ff
option to preserve evidence of your feature
branch in the repository history. This advice may be freely ignored for smaller
local feature branches for which a fast-forward merge will look like any other
routine development work on main
.
Do not use git push -f
, use --force-with-lease
instead
Force pushing in git is a subject that attracts all kinds of religious
battles. This advice is not about the merits of force pushing. This is about how
to use git push -f
if and when you do use it.
Let’s say you’re working on a branch, 'foobar’, and you decide to force push to the remote. So you do this:
$ git push -f
If anyone else has committed anything to your branch since you last pulled, you will blow their changes away.
So for a bit of safety, if you ever need to force push please instead do:
$ git push --force-with-lease
--force-with-lease
refuses to update a
branch unless
it is the state that we expect, which is that nobody has updated the remote branch.