Using GIT every day is quite normal for a developer. We work with GIT, we trust it to keep our files safe and we trust it to share them too. A long time ago, when I was on my first internship at Easter-eggs I wrote a small web service to automate a customer's GIT workflow.
I learned a lot about GIT at that time. Not only the command line interface, but also some internals and structures. Since then, I love it, and I use it every time I deal with text files… well, I'm always dealing with text files, so I'm ALWAYS dealing with GIT.
But, even if I'm loving it, I have to say : GIT is also painful:
- You have a long list of command to type to do what you want to do
- Lots of repetition through a single day developing in a team
- Through repetition, us, poor human beings, make mistakes
- More command to revert those errors
- More command to complete the task you wanted to de before you made those errors
- More error due to more commands…
It's difficult to make GIT act the way you want, you have to learn a lot for that. I saw a lot of programmer who actually adopted a simpler approach
So let's try to learn some ways to do better, some ways to master your repository and not going crazy about it.
Stash and Discard
First painpoints in your workflow will happen locally, on your machine. You'll have to organize your commits before pushing them to the team. Let's avoid some dangerous shortcut. Never commit everything in your directory, carefully select your changes and put a clean and clear commit message to help people understand your patch.
Once you got clear commits, and are ready to push, or rebase on remote branches, you'll often find yourself in a situation were you have some more changes that you can't send. It may be configurations, debug logging instruction, or even some code commented out during your work. This files and changes will prevent GIT from performing further operation including merges and rebases.
So, let's use the most important script to me: git-stash. It'll let you save your working tree in a separate commit, not in your current branch. You can then do whatever you want on a clean working tree without worying of all the little changes you might want to keep. Nobody will ever notice your stash.
You just have to apply your stash once you finish manipulating your repository and you're good to go.
But wait, this isn't magic and you'll have to deal with possible conflict between your stashed changes and your local files. I often find myself in a situation were I got multiple stash on my repos, and I can't remember what's in it.
There is no solution right now to deal with this with a simple installation of GIT. You can find some addons to help you, but in my opinions, best ways to avoid possible issues is to apply simple rules to keep a sane repository:
- Configuration files in .gitignore file. You don't want to commit configuration variables (sometimes contains credentials)
- If you add local files and can't commit them even in .gitignore, put them in a ignored directory referenced in .gitignore
- Force yourself to always work on a clear and updated repository. Rebase often on team's work.
- Communicate with your team ! Share your solutions, there's pretty good chances your coworkers have the same issues !
Then, logically, when you want to commit and push your work, you quickly take habits with stash. You multiply commands on your terminal and end up with some aliases or meta-command to help you.
Repository, remotes and branches
Another complexity that often occurs is implied by remotes and branches. You don't work alone, and you need to use your coworkers' work. It's easy to lose grip.
Switching repositories, remotes and branches is hard, don't count on your memory, it sucks at remembering things correctly… humans…
The simplest solution is the same as everyone does with their code: conventions also known as Branching models. Except here, all companies have different ways to do that, and you often find in a single team several ways of doing things.
First solution : A single repository
Simpler solution is to remove complexity through a single repository. You avoid this particular issue and some team may like it.
The main advantage of this approach is your commits, they are bigger, but contains everything you need and you can add or remove thing by editing single commits.
I personally discourage this as it won't scale with whatever your team do next.
If your company grows, chances are that your repos will grow too. You add teams on some other project and everything inside your repository. Rebasing always and always on a repos that admit so much changes will be harassing.
Second solution : Hooks
You can also forbid any human to do thing and automate your team workflow with git hooks. This will automate some task, like submitting a pull request or cleaning up the project. You can automate your tests with the commange git commit to ensure your commit has run your tests.
You can basically add anything you want to your workflow automatically with hooks.
Your team will perform good operations, but they won't understand them. Worse, they won't be able to detect when it goes wrong. Write your hooks in bash and there's pretty good chances that some of your users will fail executing them.
So, git-hooks might be a good thing, but check the following before telling everyone to use it :
- Everyone have to be able to execute your scripts
- Everyone must understand what it does
Third solution, the hardest : Learning
By far, learning GIT is the better solution to all your problems. If your team understand how GIT works, there won't be any trouble using it. You'll eventually find people doing complex things in unexpeted way.
This might sounds obvious, but it isn't. I've seen team where everyone know how to use GIT but no one ever used it the same way before. Team members were all so sure to understand what they were doing that nobody listen to other team member.
This is the best advice I could ever make here: Take the time teach every person in your team. It will take some weeks, or month to have ervyone on the same page, but it worth it !
So, yes, GIT is awesome and everyone is using it. But it's also a complex tools, it will hurts your feelings. You can make mistake, you can loose time.
There won't be any magic tool to do things you want right away. All we can do is a bunch of script and conventions that suits you and your team.
The way we're using GIT tells us more about how we organise ourselve than the work we do. And the greatest tool you can imagine to be more productive with GIT as a team is learning.
Don't hesitate to communicate on tools, tips or any conventions you saw on other teams: lots of developper are seeking help with GIT and every plugins could save their day.