Post

Git Tutorial

Posted Updated
By Robert Caddy 16 min read

Intro to Git and GitHub

General Structure

  1. Setup and Install
  2. Creating new repos
  3. Repository History & Introspection
  4. Git Config & Aliases
  5. Committing
  6. Workflow & Branching
  7. Assorted Useful Git Commands
  8. Other

1. Setup

Installing Git

macOS

  1. Install Homebrew with /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
    • If you use macports or something else then use that to install git
  2. Install git with brew install git
  3. Check that git is installed with git --version, this guide uses some features that were implemented after git 2.3.10 so make sure you have at least that version

Linux

Very similar to the macOS install but you’ll need to use whichever package manager is present for your distro.

Windows

I personally haven’t done this. I recommend you look into the Windows Subsystem for Linux (WSL) for your git and HPC needs

Register for GitHub

Follow the instructions to sign up for GitHub. Note that as a student/researcher you can get free GitHub Pro by filling out the application here.

I encourage you to setup SSH for connecting with GitHub. If you want to digitally sign your commits to show that they’re from you then you can follow the instructions I’ve written here, this is entirely optional.

2. Creating a New Repo

GitHub Documentation

  1. Run git init -b main
    • -b main is optional, it sets the primary branch name to main instead of master. You can make all your git repos default to this by settting
      1
2

      [init]
    defaultBranch =  main

      in your gitconfig file

  2. Run git add . to prepare all the files in the current directory for committing. You can replace the dot with a space seperated list of filenames if you would prefer
  3. Commit the changes with git commit -m "message". You can also omit the -m "message" and write your commit message in the dialog box that opens, then save and exit, that will commit your changes.
  4. Go to GitHub and create a new repo with the same name as your directory then
    1
2
3
4
5
6
7

    # Set the new remote
$ git remote add origin  <REMOTE_URL>
# Verify the new remote URL
$ git remote -v
# Push to new remote. This is how you push any branch to the remote and set
# it as an upstream branch
git push -u origin main

Notes

  • You can add multiple origins, usefull if you’re managing a complex structure of repos and their forks like many PIs are.

3. Committing & Branches

Committing

Git is built around “committing” code and changes. Any time you change something, add a new file, etc. you can then commit that file/change to add it to repository history. Choose which things you want to commit you must first run git add <files/pattern> to prepare whichever files you want for committing then commit those files with git commit -m "message" or git commit.

Git Message Structure

There are some best practices for git commit messages. GitHub gives some line length limitations for optimal display which are noted in the template message below. Here’s a good guide on general structure and ideas for writing good commit messages. There are also some good more automated tools such as Conventional Commits.

Personally I make sure that each commit title answers the questions “If applied, this commit will…”, as long as you’re consistent with your team, clear, and concise it doesn’t really matter what you choose. I am looking into using Conventional Commits though.

You can setup a commit message template by adding the following lines to your .gitconfig file.

1
2

[commit]
    template = /path/to/template/git-message-template.txt

My commit message template

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

# You MUST limit the title to 50 characters and lines to 72 characters
# The 50 and 72 character limits are indicated with a '|' in each prompt
#
# If applied, this commit will...                |


# Why is this change needed?                                           |
#Prior to this change,

# How does it address the issue? (response options, )                  |
# to close multiple each issue must be prefaced with the keyword
#Partially addresses issue #X
#Resolves #X
#Fixes #X
#Closes #X

Git amend

git commit --amend lets you edit the last commit message. If you’ve already pushed that commit you’ll need to force push the change. Editing commit messages for older commits is doable but much more complex.

Branches

Git branches are the main organizational tool for git and their uses are many and varied. Branches are to complex to be simply summarized in a few lines in this document so I’ll refer you to the Atlassion documentation on branches and provide a Git Branches cheatsheet that I use regularly.

Branches should generally be short lived: make a branch, add a small feature, merge the branch, then get rid of it. Long lived branches can be difficult to rebase. Once a feature branch has been merged it shouldn’t be used again. Using an already merged branch can be complex

4. Repository History & Introspection

Git Status

git status you will use all the time to check the status of a repo. Shows current status of rebases, merges, staged files, unstaged files, tracked files, etc. Has flags for targeting specific branches, what info is displayed, etc.

Git Log and Aliases

  • git log shows lots of info but it isn’t organized super well.
  • git log --oneline --decorate --graph --all gives a prettier picture.
  • git show SHA gives all the details on a specific commit

Git Blame

git blame is used for checking who made what modification to your code. Just use GitLens for VS Code or GitHub.

Git Diff

git diff shows all the differences between the unstaged changes and the previous commit + staged changes.

Visual Studio Code (VSC) Extentions

  • GitLens, a million amazing features. Just install it asap. Theres a paid version with a few more features but it’s free for public repos so you should sign up.
  • Git Graph, A very clean and usable graph of the history of the repo. An improved version of git log --oneline --graph
  • Diffs and merge conflicts are much simplier

Ack

A simple tool for recursively searching text files. General usage is just ack "pattern" and it will recuresivley search for the pattern while ignoring common VCS directories, binaries, etc. My alias for it is:

1

alias ack='ack --ignore-dir={_site,.git,.container,._site} --ignore-case --literal'

Why use Ack?

  1. fast
  2. portable
  3. easier than grep
  4. Ignores the files and directories you probably don’t want to search
  5. easy to specify file type with --language. e.g. --cpp or --python
  6. pretty colors and good output formatting
  7. Perl Regex
  8. Similar flags to grep

5. Git Config & Aliases

Link to my gitconfig skeleton

Highlight these Aliases: (run this by showing my gitconfig on the left and commands on the right in a split terminal window)

  • stick - a clearer and more concise version of git log
  • longStick - same as stick without the length limit
  • check - run Repo-check-updates.sh
  • hub - if on macOS then open the GitHub page in the default browser. If not on macOS then print the URL
  • alias root='cd $(git rev-parse --show-toplevel)'
  • If you’re doing anything even moderately complicated I recommend writing a bash script for it and calling that script rather than writing it directly in the gitconfig. Doing the latter can give weird results. Also note, all bash commands in git aliases are run in a separate shell so commands with stuff like cd in it won’t produce the desired effect.

The .gitignore File

Your .gitignore files tells git which files to ignore. Good for compiled code, logs, OS files, etc.

6. Workflow

Once you’ve finished the feature, bug fix, etc. you might find that you have a lot of commits at an intermediate state. Often these intermediate states are poorly designed, broken, or bad in some other way. Ideally we would like to reduce all our work to one, or a couple, high quality commits before merging them. Here’s two options on how to do that; you should always backup your branch before trying either of these methods.

Option 1: Safely Squash and Rebase your Git Commits

This section is based on the post Always Squash and Rebase your Git Commits by Anna Sherman. It’s written in more detail there and I highly recommend reading it

This method works best for a small (<50) commits where you care about the intermediate state and commit messages. It can be more complex but lets you walk through your intermediate commits one-by-one to fix any conflicts.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

# Pull current root branch
git pull main
# Create and checkout a new branch, Note that you can also checkout specific commits
git checkout -b featureBranch
# Set that branch upstream
git push origin featureBranch --set-upstream

# do stuff, make commits, etc

# Squash
git rebase -i HEAD~[number of commits]
# OR
git rebase -i [SHA]
# SHA is the unique identifier for a given commit. It's a 40 character hash
# (SHA=Simple Hashing Algorithm) which is often referenced byt just the first 7
# characters. You'll usually see it just to the left of the commit message

# Check that your main branch is up to date
git checkout main
git pull

# Rebase on to main branch
git checkout featureBranch
git rebase main

# Update remote branch, NOTE THE DANGERS OF force push
git push origin featureBranch --force
# force pull
# git fetch --all
# git reset --hard origin/<branchName>

# Merge featureBranch into main
git checkout main
git merge featureBranch
git push

Squashing to a single commit isn’t required, but maybe squash down to a handful and make sure that each commit builds and passes tests. Remember you can have multiple layers of this: featureBranch off of main and subFeatureBranch off of featureBranch

Option 2: The Nuclear Option

This is the nuclear approach for when you do not care about intermediate state. This is often nice for when you have hundreds of commits and you don’t want to deal with any potential merge conflicts between them and instead just use the latest state. You will destroy all new commits since you branched and just keep the changed files. If you want to know what the commit messages said you will need to get them manually before starting this process. Please backup your branch before doing this.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

# Pull current root branch
git pull main
# Create and checkout a new branch, Note that you can also checkout specific commits
git checkout -b featureBranch
# Set that branch upstream
git push origin featureBranch --set-upstream

# do stuff, make commits, etc

# Reset
# `git branch --show-current` returns the name of the current branch,
#  "featureBranch" in this case.
# `git merge-base <commit 1> <commit 2>` returns the SHA to the most recent
# common ancestor.
#
# Overall this command takes us back to the same state as when `featureBranch`
# branched off of `main` but with all the changes since then ready to stage and
# commit
git reset $(git merge-base main $(git branch --show-current))
git add <whatever you want to add>
git commit

# Check that your main branch is up to date
git checkout main
git pull

# Rebase on to main branch
git checkout featureBranch
git rebase main

# Update remote branch, NOTE THE DANGERS OF force push
git push origin featureBranch --force
# force pull
# git fetch --all
# git reset --hard origin/<branchName>

# Merge featureBranch into main
git checkout main
git merge featureBranch
git push

7. Assorted Useful Git Commands

Git Reset

detailed guide

–soft vs. –mixed vs –hard guide

git reset is an incredibly powerful and useful tool and you should know how to use it. However, it’s less a gun that you can shoot yourself in the foot with and more of a hand-grenade vending machine so make sure to backup your branch before using it.

  • git reset, removes all staged files from staging already
  • git reset --hard, deletes all uncommitted changes to tracked files
  • git reset --soft HEAD~X, undo the last X commits but keep the changes staged. Fantastic for if you mess up a commit.
  • git reset SHA, moves both the current HEAD pointer and the branch pointer to a specific SHA. This is often a more dangerous version of git revert or git cherry-pick and unless you know what you’re doing should be avoided. DON’T USE IT ON SHARED BRANCHES
    • If used with --soft it will delete all the commits after SHA but NOT change the files that were committed and it will stage them for a commit.
    • If used with --mixed (default behavior) it will delete all the commits after SHA but NOT change the files that were committed. The changes will not be staged for commiting
    • If used with --hard - If used with --hard it will delete all the commits after SHA AND delete all changes after SHA. The changes might be recoverable with git reflog but don’t count on it. Be very careful with --hard.

Git Revert

git revert SHA undoes a specific commit. It doesn’t delete that commit, just reverse all the changes in the original commit and makes a new commit with those undone changes. Deleting and editing old commits can be done but requires a force push and so should not be done on shared branches; probably shouldn’t be done on unshared branches either.

Git Stash

Have something you’re working on and need to go somewhere else? Note that stashes are local ONLY.

1
2
3
4
5
6
7
8
9

git stash

#do other stuff

# See all the different stashes
git stash list

# Extract and reapply the current stash
git stash pop

Stashes can be given names, include untracked or ignored files (--include-untracked and --all) and you can create a branch directly from a stash (git stash branch <optional branch name>). Delete all stashes with git stash clear

Git Cherry-pick

git cherry-pick SHA grabs the single commit with the given SHA and copies it onto the currently active branch. Useful it you want to grab the changes from an older commit or move just one commit between branches

Git Add -p

git add -p <optional filename> lets you commit single lines/chunks of code without commiting all the changes in a file. See this guide for more info

Git Clean

git clean is used for operating on untracked files. Useful for deleting many untracked files if you need to. Go read the documentation if you think you might need this

Git -C path

Runs git as if it’s in the directory specified by path. Fantastic for scripts. I use it in my gitall.sh script to check multiple repos at once for updates and in Repo-check-updates.sh to check if there are any updates for a given branch/repo that I need to apply.

Note: lowercase -c is totally different

8. Other

GitHub: Issues, Pull Requests, Actions, Wiki

GitHub, and other similar services, do many other things besides just host git repositories, here I’ll give you a quick run down of some of the highlights.

  • Issues: A place to raise issues or ask questions about a specific repository. This is how you would notify a repository owner about an issue. They can also be used to keep track of to-do items in a repo. They can be referenced in a commit message by writing #Issue-Number.
  • Pull Requests (PRs): A request that the owner of a repository “pulls” your code into the repo. It takes a bit to get used to how to use these but it makes managing large repos much easier. They can be referenced in a commit message by writing #PR-Number.
  • Actions: Actions are tasks that you can have GitHub automatically or manually perform on certain triggers. Common use cases are compiling and publishing a website, building and testing code, build and publishing documentation, etc. They have TONS of uses and aren’t too hard to start using.
  • Wiki: A wiki for the repo. Exactly what it sounds like, all pages are written in markdown.

Dotfiles Organization

You might find that your settings files, such as .bashrc, .gitconfig, .pythonrc, etc. can become difficult to manage or keep synced between computers; these files are also called “dotfiles” since their names usually start with a period. To help manage these more effectively I highly recommend you put them all in a git repo and then link those files to the appropriate location. There are many repos out there that will help you do exactly that but I’ve found that many of them are too complex to easily use and don’t always work well when you might have to work on multiple different systems/clusters. With the help of my friend Seth Cook I wrote a system for managing my dotfiles which you can find in my dotfiles-skeleton repo. That repo contains all the documentation required to get up and running with managing your dotfiles, simply fork the repo, add your dotfiles, and start working. I periodically update the skeleton repo with any major changes in my private dotfiles repo.

Git, Guides
This post is licensed under CC BY 4.0 by the author.