Git commits messages and history

A reflection on how to use Git commits to document changes and decisions

Posted on May 14, 2018 6 minute read

Git is a very powerful tool, that can be used to more than just archiving code and give you the possibility to revert and track changes. If you use it wisely, even if you don’t have the git-foo powers your project can benefit from it. Sometimes, while browsing around some code on github or gitlab I struggle to find some reasoning for changes, why those files we’re changed or why something was implemented that way. The climax is when you found a commit message with tons of changes, spread across several files and the commit message: misc fixes or another vague title, that says absolute nothing about the change.

Transposing this to a company project, I believe it’s even more important, because people come and go to the projects and we sometimes struggle to find the reasoning for changing or why the develop choose that approach instead of another one. If there’s no place with that information, we’re left in the dark and may start to do some changes and will wither be rejected at review time (in the best case) or released in production and lead to unexpected behaviour (in case the change was made due some restrictions).

So what can we do to avoid this and save precious developer hours or, for someone like me, who’s very found of cheese and tend to forget the reasoning’s after sometime? Use, what you already need to write: the commit message!

Unleash the power of the commit messages

In order to make the best use the of Git utilities to work with commits history we must follow some rules. If you’re not used to them, it may looks too overwhelming at the beginning, but after a fer commits you’ll start to see the light and benefit from them. Also a good place to start enforcing these is in the code review process, including the format and content of commit messages as part of the review process.

What to include in the commit message?

Every commit message should answer three simple questions:

  • Why?
  • How?
  • What?

Why this change is needed?

This change may fix a bug, add a feature, improve performance, reliability, stability, readability, etc. This information will help the reviewer or person looking through the project history to identify the reason that lead to the code change.

How this change will fulfil his purpose?

High level description of what have been done and what approach was taken to fulfil the requirements. If there’s alternative approaches that we’re tried and failed, I like to enumerate them along with their downsides to better justify the approach taken.

What will this change affect the system?

What are the side-effects of this change.

The format of the commit messages

In this section I will give an example of what is considered a good commit message. Note that I’m not the creator of this format, it’s just a mere observation of other projects best practices with some reasoning around.

Short summary with less than 50 chars

Body of the commit message, with the explanation of why we need this change, what changed and what are
the side effects. Use at most 72 chars per line and in the imperative, like "Fix bug" and not "Fixed bug".

You can split your message by paragraphs, but make sure you add an empty line between them.

 - Use bullet points whenever you need to enumerate things

Issue references or commands

you may wonder that the last line means, but if you use GitHub or GitLab you can use Issue #1 to mention the issue that this commit is related or Closes #1 to state that this commit will fix and close the issue, Make sure you search for the available commands from your provider.

Reasoning about some conventions here

50 chars summary

This will be converted as the Pull/Merge request title, also allow to quickly identify the changes

Empty line after summary

That’s the obvious separation between the title and the body

72 char lines

  • git log does not do any special special wrapping of the commit messages. With an 80 char terminal, we’ll have to have room for identation.
  • GitHub/GitLab don’t break lines
  • An exception to this rule can be applied when quoting something or other text with a special formatting.

Use of the imperative

Because it matches with commit messages generated by commands like git merge and git revert and consistency is a odd thing.

Useful tips for a sane commit history

Never user git commit -m

This way, you will ever have an proper editor window

Never mix two changes in the same commit

If your commit addresses one issue, solve that issue in only one commit, if you need other changes, add as much commit messages as you need. For example if an issue needs an updated on you dependencies, add one commit to update it and then another to address the issue. Remember that in Git history, like the real world, order is important.

Make sure every commit is green

Every commit should pass you CI,

Use squash and fixup

By the sum of the previous two, sometimes we need to have intermediary commits with smaller changes or experiments. It’s OK to have them, but remember to squash or fixup them, before the Pull/Merge Request.

Bonus content

Use the force… but with care

When you need to do a push with --force to override some changes, make sure you use --force-with-lease to update the branch with possible changes from others, otherwise you will override your teammate’s work with your command.

How to squash commits

Imagine that you have three temporary commits and want to merge all the changes in one commit with a single message. You start with:

git rebase -i HEAD~3

meaning you want to interactively (-i) change the last three commits. At this point you’ll get:

pick abfc1db01f Fix test case 
pick 70e3d1ea9d Fix typo 
pick b54e5706e7 Improve documentation

# Rebase 4b010b0815..3e46cf7533 onto 4b010b0815 (3 commands)
#
# Commands:
# p, pick = use commit
# r, reword = use commit, but edit the commit message
# e, edit = use commit, but stop for amending
# s, squash = use commit, but meld into previous commit
# f, fixup = like "squash", but discard this commit's log message
# x, exec = run command (the rest of the line) using shell
# d, drop = remove commit
#
# These lines can be re-ordered; they are executed from top to bottom.
#
# If you remove a line here THAT COMMIT WILL BE LOST.
#
# However, if you remove everything, the rebase will be aborted.
#
# Note that empty commits are commented out

So you’ll change the comman from pick to squash (s) or fixup (f) if you want to discard the commit messages. After saving you’ll have an editor to input your commit message and voilá.

If you have pushed the commits to your remote (server), you’ll need to do a --force-with-lease to override the old commits.

How to configure Vim as your git editor and enforce this rules

Add the editor to the core section of your .gitconfig file:

[core]
  editor = vim

Personally I would recommend to use vim-fugitive plugin, but adding this line to your .vimrc also work as a minimalist solution: autocmd Filetype gitcommit setlocal spell textwidth=72

To know more about vim-fugitive check it’s page

If you want to ask some question or share your thoughts please contact me from the links on the left.

To share this post, use these links: