This is everything you need to know about having better Git commit messages.
I listed 10 rules for having a professional commit message.
The first and the most important rule for a Git commit message is this! Each commit message must have a complete, meaningful and independent description. Commit message has to be valuable for its readers. As a reader, when I'm looking at the Git's log of your repository, All the commit messages must be clear for me and I should be aware about the changes of each commit, just by a simple look!
I prepared some examples of bad commit messages and I explained why they are bad, so look at these:
❌ First commit
🔷 Why it's bad?
It's not important for the reader to know about the number of commits in its message. Because he/she can be aware about it, by a looking to the Git's log. Also he/she can't have any idea about what's happening in the codebase in this commit!
❌ Add some codes
🔷 Why it's bad?
When you work on a codebase, you can add, delete or modify the code and there is no any other options! So, it has not any benefits for the reader! You must say what you added!
❌ Refactoring and cleaning codes
🔷 Why it's bad?
How can I be informed that which section of the codebase has been modified? Or how can I guess what type of refactoring has been applied on the codebase? Your commit message have to include the answer of these questions in self!
❌ Bug resolved
🔷 Why it's bad?
Which bug is resolved? must the reader guess the bug?! Of course not, so please explain which bug is resolved explicitly.
❌ I added a new feature
🔷 Why it's bad?
According to the previous examples, here we haven't any clear vision about the modifications of the codebase! Btw, there is no need to use pronoun at the first of this commit message. Git saves the author of each commit by itself and we haven't to do a duplicated action!
If you want to write a meaningful message for a commit, avoid writing lengthy descriptions. Very summarized texts too! A sign of a good commit message is a suitable length most of the time. Maybe you want to know why we should avoid long/short messages in our commit messages? Well, here's an example.
Assume my friend asks me to lend him some money.
❌ Please lend me some money!
It's an example of a short message. When my friend tells it, I ask two related questions: "How much money do you need?" and "When will you get it back?". He could make me clear about them, so there was no need to these questions.
❌ Hey Ahmad, get on the bus as soon as it arrives at the bus station. Then go to your bank, fill out a form, and then get the money from the bank and lend me $20. I'm going to return it in three weeks.
This is an example of a long message. It's an example of a long message. Don't tell me how can I provide some money! Just tell me your request.
✅ Ahmad, please lend me 20 dollars. I will return it in 3 weeks.
Perfect! It's that I want to hear!
🔑 The text of your message, must be the gist of your commit.
🔑 If you think that you need a long text to tell the gist of your commit, probably you have a big commit that should be separated in 2 or more smaller commits.
When you want to write a commit message, suppose you give Git an order. Do not ask or inform Git! Just command directly!
For example, suppose you implemented a function that can transform the characters of a text into the upper case characters.
❌ Implementing a function that transforms letters to upper case.
❌ I added a function to transform letters to upper case.
❌ A function that transforms letters to upper case has been added.
✅ Implement a function to transform letters to upper case.
This rule complements the Neither Long nor Short rule. No need to write everything in details.
Always remember, you must tell What and Why something added/removed/modified in the codebase, not How! Also, no need to write the bit or low-priority modifications.
For example, assume you want to refactor a function named getUser in your codebase. After refactoring it, you may notice that the related file has an unused variable inside of itself, so you decide to remove it. When you want to commit your changes, there is no need to write something like this:
❌ Refactor the 'getUser' function and delete an unused variable in its related file.
Why it's not necessary to refer to the delete variable? Because it wasn't the primary goal of this commit. It was a minor modification that we could fix along with another commit. Of course, you can delete this unused variable in a separate commit, or maybe you want to make a new task for your project to find all unused variables and remove them all in a commit!
I prefer this one:
✅ Refactor the 'getUser' function.
This one is an optional offer for your commit messages but can make them very readable and powerful. According to the added modifications in the codebase in your commit, you can use a suitable and matched prefix for its message. In most cases, prefixes are used at the first of commit messages.
Here is a list of the most usable prefixes that I often see in software teams:
[BUGFIX]: For resolving any bug and issues from the codebase.[TEST]: For writing any type of tests (Unit / End-to-End / UserStory / Integration / Regression / ...)[FEATURE]: For adding a new feature and implementation[REFACTOR]: For improving current codes for any purpose like better readability, performance, etc.[UPDATE]: For updating libraries and dependencies, also updating current features by documentation.[BASE]: For installing new libraries and dependencies and setting up the infrastructures of the codebase, like configuring the project and its packages.[DOCS]: For adding or modifying the documentations.
The above cases are only some offers for you so, there isn't any force to use them. You can have your own!
🔑 The text of your prefixes can be lowercase. Please follow a fixed rule. For example, if you want to write your prefixes in lowercase, write them all in lowercase forever in the related repository.
Recently I've seen developers using emojis in their commit messages. In my opinion, it's not a good practice! Say Why? Because of two reasons:
- They don't have a clear message.
- Emojis have specific Unicode, maybe we have a terminal that doesn't support those Unicode, so we can't see any emojis in our terminal.
One of my colleagues disagreed with me about the case one. She believed that each emoji has a special meaning, and we can use them in our commit messages. I asked her to give an example. She showed the emoji of a worm (🐛) and said that everybody knows that this emoji refers to a bug! I asked her what should be used for features, refactoring, testing, and ...? She introduced a website to me and she said they provide lots of emojis and have a unique meaning for all of them.
It's the address of that website: Gitmoji
I disagree with my colleague because this website is not a reference for git commits! It's not the product of a famous company with high-level standards! Only one tasteful person has presented some emojis and has written explanations for each of them according to his taste. And nothing more! One day, if terminals support showing images, I can create a website for using images in the commit messages! For example, I can provide the picture of a scared man and say that refers to a bug! It's not standard :)
This case is similar to Conceptual Prefixes with a main difference: The Conceptual Prefixes have a straight and clear message for their readers and they directly say their content. But the Referral Prefixes point to a external description, like a task on Jira, a card on Trello, an issue on Gitlab or anything else! Referral Prefixes usually be used in the large teams and teams that they use Scrum methodology.
For using a referral prefix, there are some common styles that I listed:
[#<CARD-NUMBER>],[#<ISSUE-NUMBER>]or[#<TASK-NUMBER>][<CARD-NUMBER>],[<ISSUE-NUMBER>]or[<TASK-NUMBER>](<CARD-NUMBER>),(<ISSUE-NUMBER>)or(<TASK-NUMBER>)
As an example of the above cases:
[#217] Resolve the bug of 'fetchUsers' function that causes some users to be missed
[#217] the first part of this commit can refer to a task number 217. It's no matter which tasks manager you are using. You can always use these referral prefixes to point your tasks/cards/issues.
🔑 You can use both Conceptual Prefixes and Referral Prefixes together. It is related to your decision.
I listed some examples for the usage of conceptual prefixes and referral prefixes side by side:
-
[#REFACTOR](217) Use Array's methods like 'forEach' and 'map' instead of simple 'for' and 'while' loops in the codebase -
[REFACTOR-217] Use Array's methods like 'forEach' and 'map' instead of simple 'for' and 'while' loops in the codebase -
[#217][REFACTOR] Use Array's methods like 'forEach' and 'map' instead of simple 'for' and 'while' loops in the codebase
🔑 The above examples are just some template and they are not some rules. You can use them in any other styles.
Sometimes commit messages contain references to variables, functions, methods, classes, and specialized terms. Whenever you need to write a specialized term or the name of a variable/function/method/class or anything similar, wrap its name in a couple of backticks or quotation marks. It helps the reader detect that the wrapped word is different from the others and there is a reference to the codebase.
❌ Upgrade Jest and React Testing Library to the latest versions
✅ Upgrade 'Jest' and 'React Testing Library' to the latest versions
❌ Use for-in loop instead of for-of loop inside calculatePrice function
✅ Use 'for-of' loop instead of 'for-in' loop inside the 'calculatePrice' function
-
Write your messages just in English! Not any other language.
-
Don't put dot (
.) at the end of your messages.
❌ Install 'Redux' and 'Redux Toolkit' libraries.
✅ Install 'Redux' and 'Redux Toolkit' libraries
- Capitalize the first letter of the commit message.
❌ write integration test for 'Settings' module
✅ Write integration test for 'Settings' module
-
Wrap lines at 72 characters. (Always there exists some exceptions)
-
Don't use urls in the messages.
If you have a tour in the log of lots of the famous repositories, you can see a lot of differences between commit messages. For example, some of them have prefixes, and some other have'nt. Some of them include emojis, some of them no. Some start with a capitalized letter, some of them no, etc. It usually happens because teams and people don't follow typical rules. Always have a complete and clear guideline for yourself and your team. By doing it, you can prevent having inconsistent commit messages.
You can use some external tools and linters for double-checking the commits. Linters and these external tools, ensure that no commit message will conflict with the rules you applied. They force you and your team members to follow the rules.
If this repository helped you, don't forgive getting a star ;)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent