Version control with Git

Introduction

Version control, also known as revision control or source control, is the management and tracking of changes to computer code and other certain other types of data in an automated way.

Any project (collections of files in directories) under version control has changes and additions/deletions to its files and directories recorded and archived over time so that you can recall specific versions later.

Version control is in fact the technology embedded in the versioning of various word processor and spreadsheet applications (e.g., Google Docs, Overleaf).

---

Fig. 6 An overview of how version control works. Some of the project mainline’s branches may eventually be merged back into the mainline if they are successful in achieving some objective that is worth integrating into the mainline.

---

Why Version Control?

With version control of biological computing projects, you can:

• record all changes made to a set of files and directories, including text (usually ASCII) data files, so that you can access any previous version of the files

• “roll back” data, code, documents that are in plain text format (other file formats can also be versioned; see section on binary files below).

• collaborate more easily with others on developing new code or writing documents – branch (and merge) projects

• back up your project (but git is not a backup software - see sections on binary and large files below).

---

Fig. 7 This will very likely happen if you don’t use Version Control.
(Source: maktoons.blogspot.com).

---

Why git ?

We will use git, developed by Linus Torvalds, the “Linu” in Linux. This is currently the most popular tool for version control. It is a distributed version control system designed to handle everything from small to very large projects with speed and efficiency. It allows multiple developers to work on a project simultaneously without overwriting each other’s changes, facilitating collaboration and maintaining a history of all modifications.

The man advantages of Git are:

• Distributed System: Every developer has a complete copy of the project on their computer, including its history.

• Efficiency: Fast performance for both local and remote operations.

• Data Integrity: Every file and project snapshot is checked for data integrity.

• Flexibility: Supports “non-linear” development workflows through branching and merging.

• Collaboration: Facilitates teamwork by allowing multiple people to work on different parts (typically on separate branches) simultaneously.

Let’s get started!

In git, each user stores a complete local copy of the project, including the history and all versions (the “repository”). So you do not rely as much on a centralized (“remote” - somewhere else, accessible through the internet) server.

The Git repository is a directory that contains your project’s files and the entire history of changes made to those files. Here is a graphical outline of the git repository, workflow and command structure:

---

Fig. 8 The git repository, workflow and command structure. - The “workspace” is your working directory - the actual files and folders you see and edit. The index” is your staging area (more on this below) - a place where you can group changes you want to commit together. The “local repository” contains the database of all objects (commits, files (“blobs”), directories (“trees”), etc.) in your repository along with their metadata. The index and the actual data (the objects) associated with each repo snapshot you commit are all all inside a hidden directory named .git that sits within your project directory.

There are several core git concepts and commands that you need to understand. But first things first - let’s first install get and set up your first git repository!

Installing git

Let’s install and configure git. On Ubuntu:

sudo apt-get install git

Then, some configurations:

git config --global user.name "Your Name"
git config --global user.email "your.login@imperial.ac.uk"
git config --global init.defaultBranch main

Now check your git configurations:

git config --list

This output displays the configuration variables and their corresponding values that are currently active in your Git environment. As you are just getting started, the output will consist of just a few lines.

Tip

Git configurations are stored in different files at three levels:

• System Level (e.g., /etc/gitconfig): Applies to all users on the system.

• Global Level (e.g., ~/.gitconfig or ~/.config/git/config): Specific to your user account.

• Local Level (.git/config in the repository): Specific to a single repository.

git config reads these config files in order, with local settings overriding global and system settings. If you want to know which of your configurations are local vs global, use this:

git config --list --show-origin

Your first repository

Time to bring your TMQB coursework under version control. First, let’s navigate to it:

Then initialize your local git repository:

git init

Initialised empty Git repository in /home/mhasoba/Documents/CMEECourseWork/.git/

Create your README file (as a markdown document, so with the .md extension):

echo "My CMEE Coursework Repository" > README.md

Check again git config --list to see if anything has changed.

Check the files and directories that have been created:

ll #same as "ls -al"

Note

Note the hidden .git directory! (More on this below).

Add the README (AKA “stage” it for for a git commit):

git add README.md

Check the current status of your repository:

git status

On branch main

No commits yet

Changes to be committed:
  (use "git rm --cached <file>..." to unstage)
	new file:   README.md

Untracked files:
  (use "git add <file>..." to include in what will be committed)
	week1/
	week2/

This tells you that there are local changes that need to be committed.

git commit -m "Added README file."

[main (root-commit) b2a73ef] Added README file.
 1 file changed, 1 insertion(+)
 create mode 100644 README.md

Tip

In git commands, you can also combine flags like in base UNIX commands; so you can both, add (-a flag) and commit (-m flag) at one go, with -am.

Check status again:

git status

On branch main
Untracked files:
  (use "git add <file>..." to include in what will be committed)
	week1/
	week2/

nothing added to commit but untracked files present (use "git add" to track)

What does it say now? - Compare with the above output of git status.

You can also add changes matching a particular pattern.

For example

git add *.txt

will add all text files that were changed since last commit.

Note that this adds ALL changes that match the pattern, recursively (through all sub-directories).

Next you can commit the rest of these files inside your current coursework (e.g., week 1) with another message, just like you did for readme above:

Warning

Before making your first commit, please read the section about “Ignoring files” below.

git commit -m "Full first commit of new project"

On branch main
Untracked files:
  (use "git add <file>..." to include in what will be committed)
	week1/
	week2/

nothing added to commit but untracked files present (use "git add" to track)

git status

On branch main
Untracked files:
  (use "git add <file>..." to include in what will be committed)
	week1/
	week2/

nothing added to commit but untracked files present (use "git add" to track)

---

Fig. 9 Make meaningful comments in your git commit. Please don’t neglect to make each commit message meaningful. And use this mantra: “commit often, comment always”. The correct meaning of “often” in this context comes with experience.
(Source: XKCD)

---

Commits

A commit is a snapshot of your repository at a specific point in time. Each commit records:

• The changes made (differences from the previous commit).

• Metadata such as the author, date, and a commit message.

• A unique hash that identifies the commit.

Note

The unique hash (SHA-1 checksum) in Git is a 40-character hexadecimal string. It looks like this:

e83c5163316f89bfbde7d9ab23ca2e25604af290

(You will soon see examples of this below)

It is generated by applying the SHA-1 (Secure Hash Algorithm 1) cryptographic function to the contents and metadata of a Git object. OK, that was some serious Gobbledygook! All you need to know is that this hash acts as a unique identifier for all commits, files, directories, and other objects within a Git repo so that you can track specific snapshots of the repo.

How Git Stores Data

Git doesn’t store data as a series of changes or differences, but as a series of snapshots of the entire repository. When you make a commit, Git stores a reference to the staged snapshot and a pointer to the commit that came immediately before it.

Git uses four types of Objects to store data:

• Blobs: Stores file data.

• Trees: Represents directories and contains pointers to blobs and other trees.

• Commits: Points to a tree object (snapshot of the working directory), parent commits, and includes metadata.

• Tags: Points to another object (usually a commit) and includes metadata.

These objects are stored in the .git/objects directory and are identified by their SHA-1 hash.

Basic git commands

Here are some fundamentally important git commands (please make sure you read more about them):

Command	What it does
git init	Initialize a new repository
git clone	Download a repository from a remote server
git status	Show the current status
git diff	Show differences between commits
git blame	Blame somebody for the changes!
git log	Show commit history
git commit	Commit changes to current branch
git branch	Show branches
git branch name	Create new branch
git checkout name	Switch to a different commit/branch called name
git fetch	Gather commits info from a remote branch to your current local branch without merging them
git merge	Merge two or more versions of one or more files
git pull	Upload from remote repository (this is effectively the same as fetch followed by merge)
git push	Send changes to remote repository

There are more git commands which you will also learn, but it is important that you familiarize yourself with these ones in particular first.

The most common sequence of your git work and commands will be:

• Edit Files: Modify files in your working directory.

• Stage Changes: Use git add <file> to add changes to the staging area.

• Committ Changes: Use git commit -m "commit message" to save changes to the repository.

But before you do that, you will always want to

• Use git status to see the status of your working directory and staging area.

• Use git log to view the commit history.

Your remote repository

Remote repositories are versions of your project hosted on the internet or another network. Common platforms include GitHub, GitLab, and Bitbucket.

Nothing has been sent to the remote server yet (more on the remote server below).

So let’s go to your online git service (e.g., github) and set up. Note that github and bitbucket both give you unlimited free private repositories if you register with an academic email. Not a big deal if you will not be writing private code, handy if you are (can you think of examples when you would need to write private code?).

So let’s proceed with connecting your local git repository to your remote server:

• Login to your github or bitbucket account

• Set up your ssh-based access. SSH (Secure Socket Shell) is a protocol that allows you to connect to and interact with remote servers. Here are two sets of guidelines (you can use whichever seems easier to you, irrespective of whether you are using githib or bitbucket):

  • github

  • bitbucket (most of you will want the “Set up SSH on macOS/Linux” option)

Note

When accessing remote Git repositories, you have the option to use either HTTPS or SSH protocols. Both methods have their own advantages and disadvantages, and the better choice depends on your specific needs and circumstances.

• In general, use SSH if:

  • You interact with Git repositories frequently.

  • You prefer a secure, key-based authentication method.

  • You’re setting up automated scripts or development workflows.

• Use HTTPS if:

  • You’re in a restrictive network environment where SSH is blocked.

  • You prefer a simpler setup without managing SSH keys.

  • You’re accessing repositories infrequently or from different machines.

We will use SSH beause it is generally better for regular code development due to its security and convenience after the initial setup. You may need to use HTTPS in situations where network restrictions apply.

• Next, create a new repository on your remote service with the same name as your local project (e.g., CMEECourseWork), and push your new project to this newly created remote git repository. Instructions for this step are here:

  • github

Note that you have already done the git init step, so no need to repeat those bits.

You are done. Now you can really start to use git!

The first step after having created your remote repository and added your ssh key to it, is to link the remote to your local repo (as the instructions in web pages linked above will already have told you).

The command will look like this:

git remote add origin git@github.com:YourGithubUsername/CMEECourseWork.git

Then list the URLs of the remote repositories associated with your local Git repository

git remote -v

The output would look like this

origin  git@github.com:YourGithubUsername/CMEECourseWork.git (fetch)
origin  git@github.com:YourGithubUsername/CMEECourseWork.git (push)

The output provides both fetch and push URLs for each remote, showing where Git pulls updates from and where it pushes updates to. This is helpful for understanding which remote repository is set as the target for various Git commands.

In the above example output:

• origin is the name of the remote (the default name for the main remote when cloning the repository).

• https://github.com/username/repo-name.git is the URL for the remote repository on GitHub.

• In this case fetch and push have that the same URL for both fetching updates and pushing commits.

No you can git push all your local commits:

git push origin master

This pushes the (committed) changes in your local repository up to the remote repository you specified as the origin. Note that master refers to the branch (you currently only have one). More on branching below.

Note

Only when you push or fetch do you need an internet connection, because before that you are only archiving in a local (hidden) repository (that sits in a hidden .git directory within your project).

General command sequence for connecting to Remotes

Here the a general sequence of commands for connecting a new remote

• Add the Remote: Use git remote add <name> <url> to connect your local repository to a remote one.

• Fetch Changes: Use git fetch <remote> to download commits, files, and references from a remote repository.

• Pull Changes: Use git pull to fetch and integrate changes from a remote repository into your current branch.

• Push Changes: Use git push <remote> <branch> to upload your commits to a remote repository.

Branching

Imagine you want to try something out, but you are not sure it will work well.

For example, say you want to rewrite the Introduction of your \(\LaTeX\) paper using a different angle, or you want to see whether switching to a new library or package for a piece of code improves speed. What you then need is branching, which creates a project copy in which you can experiment.

git branch anexperiment

git branch

  anexperiment
* main

git tells you that you have two branches.

Warning

If you are not seeing this specific output (with two branches) at this stage of your git chapter work, scroll back and re-check / rerun the above sequence of commands again! Note that we had renamed all master branches as main above (if you had not done that, your main branch will be called master instead).

git checkout anexperiment

Switched to branch 'anexperiment'

git branch

* anexperiment
  main

git tells you that you have successfully switched to the new branch

Or if you want to create and switch to a new branch immediately, you can use the -b flag:

git checkout -b anexperiment

Now make a change (update the Readme to record the new naughty thing you have started doing!):

echo "I am going to try this new naughty thing on this experimental branch!" >> README.md

cat README.md

My CMEE Coursework Repository
I am going to try this new naughty thing on this experimental branch!

git commit -am “Testing experimental branch”

The -am flag performs the commit with a message while also automatically staging certain changes before committing.

Specifically, the a part of the flag stages all changes to tracked files (files that have already been committed at least once). This flag does not stage new, untracked files; you must use git add to stage those separately.

Now if you decide to merge the new branch after modifying it:

git checkout main

M	README.md
Switched to branch 'main'

git merge anexperiment

Already up-to-date.

cat README.md

My CMEE Coursework Repository
I am going to try this new naughty thing on this experimental branch!

Unless there are conflicts, i.e., some other files that you changed locally had diverged from those files in the main branch in the meantime (say, due to new changes pushed by another collaborator), you are done, and you can delete the branch:

git branch -d anexperiment

Deleted branch anexperiment (was 3fbab16).

If instead you are not satisfied with the result, and you want to abandon the branch:

git branch -D anexperiment

Note

If there are conflicts between branches, Git will notify you. You’ll need to resolve these conflicts manually in the files and then stage the resolved files with git add, followed by a git commit to complete the merge.

Obvously, we have been just messing around, so you don’t really want the Readme of your main branch to say what it is currently saying!

To fix this, you now have two options: edit the Readme back to the state you wanted (and then add, commit and push), or use the git reset --hard command. However, use the latter with caution. Before you do anything to reverse the merge, read about resetting and revertinf below.

Tip

When you want to test something out, always branch! Reverting changes, especially in code, is typically painful. Merging can be tricky, especially if multiple people have simultaneously worked on a particular document. In the worst-case scenario, you may want to delete the local copy and re-clone the remote repository.

---

Fig. 10 Try not to do this. But most of us mortals will have, at some point!
(Source: XKCD)

---

Note

You can perform Git branching and merging entirely within a local repository without needing a remote.

Teamwork using git and common branching mistakes

As you can see, branchhing is the cornerstone of collaborative coding and software development using git.

Here are common mistakes to avoid with git branching and merging in team projects:

• Working Directly on the Main Branch

  • Issue: Making changes directly to the main or master branch without using feature branches.

  • Why It’s a Problem: This practice can lead to a cluttered commit history and increases the risk of introducing unstable code into the main codebase, affecting all team members.

• Not Keeping Branches Up-to-Date

  • Issue: Failing to regularly merge or rebase the main branch into feature branches.

  • Why It’s a Problem: Leads to large, complex merge conflicts when it’s time to integrate changes back into the main branch.

• Not Pulling Latest Changes Before Starting Work

  • Issue: Forgetting to fetch and merge the latest changes from the remote repository before starting new work.

  • Why It’s a Problem: Leads to working on outdated code and causes significant merge conflicts when pushing changes later.

• Improper Branching Practices

  • Issue: Not creating separate branches for new features or bug fixes, or using inconsistent branch naming conventions.

  • Why It’s a Problem: Makes it difficult to manage code changes, track progress, and collaborate effectively.

• Poor Commit Hygiene

  • Issue: Making large, infrequent commits with vague messages like “Fixed stuff” or “Changes.”

  • Why It’s a Problem: Hinders the ability to trace specific changes, debug issues, and understand the project’s evolution.

• Not Handling Merge Conflicts Properly

  • Issue: Overwriting code, deleting others’ work, or incorrectly resolving conflicts without understanding the implications.

  • Why It’s a Problem: Can introduce bugs, cause loss of important code, and disrupt the functionality of the application.

• Overwriting Team Members’ Changes

  • Issue: Using git push --force or not merging the latest remote changes before pushing local commits.

  • Why It’s a Problem: Leads to loss of others’ work and creates confusion in the team’s codebase.

• Ignoring the .gitignore File

  • Issue: Committing unnecessary files like binaries, temporary files, or sensitive information.

  • Why It’s a Problem: Clutters the repository, increases clone/download times, and poses security risks.

• Misusing Git Commands

  • Issue: Incorrectly using commands like git reset, git revert, or rebasing public branches.

  • Why It’s a Problem: Can rewrite commit history, cause loss of work, and confuse team members due to unexpected repository states.

• Lack of Communication and Coordination

  • Issue: Not informing team members about new branches, changes, or issues encountered.

  • Why It’s a Problem: Leads to duplicated work, merge conflicts, and a lack of cohesion in development efforts.

• Pushing Broken or Untested Code

  • Issue: Committing untested code.

  • Why It’s a Problem* Disrupts the workflow for others, especially if continuous integration pipelines fail due to the broken code.

• Confusion Between Local and Remote Branches

  • Issue: Forgetting to push local branches to the remote repository or misunderstanding the state of branches.

  • Why It’s a Problem: Causes discrepancies between local and remote repositories, leading to confusion and potential loss of work.

• Conflict Between Personal and Team Workflows

  • Issue: Using personal Git configurations or workflows that don’t align with the team’s agreed-upon practices.

  • Why It’s a Problem: Creates inconsistency and can cause issues with code integration and collaboration.

• Over-Reliance on GUI Tools Without Understanding Git Basics

  • Issue: Using graphical interfaces without grasping the underlying Git commands.

  • Why It’s a Problem: Limits the ability to troubleshoot issues and understand what’s happening behind the scenes, leading to mistakes.

• Ignoring Merge Warnings and Errors

  • Issue: Proceeding with merges without addressing warnings or errors.

  • Why It’s a Problem: Can result in incomplete merges, broken code, and unresolved conflicts that affect the entire team.

• Mismanaging Merge Strategies

  • Issue: Inappropriately using git merge vs. git rebase without understanding the consequences.

  • Why It’s a Problem: Can create a confusing commit history and complicate collaboration efforts.

• Deleting Branches Prematurely

  • Issue: Removing branches that others are still using or that contain unmerged changes.

  • Why It’s a Problem: Causes loss of work and disrupts team members who are relying on those branches.

Tips to Avoid These Mistakes:

• Educate Yourself: Take time to learn Git fundamentals, including branching, merging, and resolving conflicts.

• Use Feature Branches: Create a new branch for each feature or bug fix to isolate your work.

• Commit Often with Clear Messages: Make small, incremental commits with descriptive messages.

• Pull Frequently: Regularly fetch and merge changes from the remote repository to stay up-to-date.

• Communicate: Keep open lines of communication with your team about what you’re working on.

• Test Before Committing: Ensure your code compiles and passes tests before pushing.

• Understand Git Commands: Before using advanced commands like git reset or git rebase, make sure you understand their effects.

• Use .gitignore Wisely: Configure your .gitignore file to exclude unnecessary or sensitive files.

• Review Pull Requests: Participate in code reviews to catch issues early and share knowledge.

• Follow Team Conventions: Adhere to agreed-upon workflows, branch naming conventions, and commit message formats.

Tip

Cloning vs. Forking: Forking a git repository allows you to create a personal copy of someone else’s repository on your own GitHub account. This is ideal if you want to contribute to the original repository. You can make changes in your forked copy and then submit those changes back to the original repository through pull requests. Merely Cloning on the other hand, only creates a local copy on your machine. It does not provide a way to contribute changes back to the original repository without special permissions. Forking is a common approach in open-source projects because it doesn’t require permission from the original repository’s owner to make changes. You can work on your copy and submit a pull request without needing write access to the original repository.

The README file

A README (like the README.md that you created in your git repo above) is a text file that introduces and explains a project. It contains information that is required to understand what the project is about and how to use or run it.

While READMEs can be written in any text file format, Markdown (saved as an .md file) is most commonly used, as it allows you to add simple text formatting easily. Two other formats that you might most often see are plain text and reStructuredText (saved as an .rst file, common in Python projects).

You can find many README file suggestions (and templates) online. Essentially, it should ideally have the following features/content:

• Project name / title

• Brief description: what your project does and/or is for. Provide context and add links to any references to help new visitors.

• Languages: List language(s) and their versions used in the project

• Dependencies: What special packages (which are not part of standard libraries of the language(s) used) might be needed for a new user to run your project

• Installation: Guidelines for installing the project (if applicable), including dependencies.

• Project structure and Usage: How the project is structured and how to run/use it. Explain, if relevant, what specific files do. No need to list every file, such as data or experimental ones (like the ones in sandbox).

• Author name and contact

In addition, you may want to include(but not necessary for your current coursework), License, Acknowledgments, and instructions for Contributing.

Ignoring Files

You will have some files you don’t want to track (log files, temporary files, executables, etc). You can ignore entire classes of files with .gitignore.

\(\star\) Let’s try it (make sure you are in your coursework directory (e.g., CMEECourseWork)):

echo -e "*~ \n*.tmp" > .gitignore

cat .gitignore

*~ 
*.tmp

git add .gitignore

touch temporary.tmp

Then,

git add *

You can also create a global gitignore file that lists rules for files to be ignored in every Git repository on your computer.

Tip

Templates for .gitignore: You can find standard .gitignore templates online. For example, search for “.gitignore templates”.

Directory and file patterns for gitignoring

It’s important that you use the correct patterns for what to ignore in your .gitignore, to make sure that all the correct directories and files are being excluded. Note that you can put additonal .gitignore files in sub-directories if you need more fine-grained control over what is to be ignored, bt it is not necessary, because git-ignoring works recursively (including thorugh sub-directories) as long as you include the correct pattern.

Here are the basic patterns (rules) :

Pattern	gitignore result
#comment	This is a comment - this gitignore entry will be ignored (avoid having any file or directory with # in its name!)
target	Every file OR directory (because lack of trailing /) named target is ignored, recursively
target/	Every directory (due to the trailing /) named target is ignored, recursively
/target	Every file OR directory named target in the top-most directory (due to the leading /) is ignored only there (not recursively)
/target/	Every directory named target in the top-most directory (because of the leading and trailing /) is ignored only there (not recursively)
*.extension	Every file or directory ending with the extension .extension is ignored, recursively
*.py[co]	Every file or directory ending in .pyc or .pyo is ignored (but not .py)

There are additonal patterns - please see the Readings and Resources section.

.gitignore-ng after commit-ing

If you find that a file or directory belonging to a pattern that you incuded in your .gitignore fails to be ignored (still comes under version control), it most likely means that you gitignored it AFTER committing and pushing it. In this scenario, you need to use

git rm --cached <file>

for a file, and

git rm -r --cached <folder>

for a directory.

While these commands will not remove the physical file from your local repository, it will remove the files from other locals on their next git pull.

General good practice for gitignoring

When writing a gitignore it is best to ignore any file that does not contain plaintext code. This type of file is what git was designed for!

If you do need to include a small csv file, image, or output (such as when creating reports or including test data) it is a good idea to do this using exceptions (aka the ! symbol)

For example the following gitignore file would ignore ALL csv files except for pushabledata.csv

*.csv
!pushabledata.csv

Tip

As git reads gitignore files top-to-bottom and can overwrite contradicticting rules with others, it is important to make sure that these are in the correct order.

Whitelist principle

Gitignore files are usually used as a blacklist: where anything you add to your git repository is allowed unless you specifically outlaw it in the gitignore.

An alternative, safer technique (though one which is definitely more work to maintain) is a whitelist-based approach.

To do this, you firstly ignore every file with the pattern *.* and then explicitly allow files and file types that you know you want. For example:

*.*
!.gitignore
!*.R
!*.py
!*.sh
!*.tex
!*.bib
!Week1/results/myfirstlatexdoc.tex

This is not necessarily a better or worse way to do things, you are in the end trading your time and effort for increased safety and cleanliness of your repository.

Dealing with binary files

A binary file is computer-readable but not human-readable, that is, it cannot be read by opening them in a text viewer. Examples of binary files include compiled executables, zip files, images, word documents and videos. In contrast, text files are stored in a form (usually ASCII) that is human-readable by opening in a text reader (e.g., gedit). Without some git extensions and configurations (coming up next), binary files cannot be properly version-controlled because each version of the entire file is saved as is in a hidden directory in the repository (.git).

However, with some more effort, git can be made to work for binary formats like *.docx or image formats such as *.jpeg, but it is harder to compare versions; have a look at this and this¹, and also, this.

Dealing with large files

As such, git was designed for version control of workflows and software projects, not large files (say, >100mb) (which may be plain-text or binary). Binary files are particularly problematic because each version of the file is saved as is in .git, when you have a large number of versions it means that there are the same number of binary files in the hidden directory (for example 100 \(\times\) >100mb files!).

So please do not keep large files (especially binary files) under version control². For example, if you are doing GIS work, you may have to handle large raster image files. Do not bring such files under version control. We suggest that you include files larger than some size in your .gitignore. For example, you can use the following bash command:

find . -size +100M | cat >> .gitignore

The 100M means 100 mb – you can reset it to whatever you want.

Then what about code that needs large files? For this, the best approach is write code that scales up with data size. If it works on a 1 mb file, it should also work on a 1000 mb file! If you have written such code, then you can include a smaller file as a MWE (minimum working example).

Tip

And how do you back up your large data files? Remember, version control software like git are not meant for backing up data. The solution is to back up separately, either to an external hard drive or a cloud service. rsync is a great Linux utility for making such backups. Look it up on the ‘net!

You may also explore alternatives such as git-annex (e.g., see this), and git-lfs (e.g., see this).

Tip

Checking size of your git repo: You have two options in Linux/UNIX to check the size of your git repo. You can use (cd to your repo first) du -sh .git, or for more detailed information about what’s using the space, use git count-objects -vH (this will work across platforms as this is a git command).

Pre-Commit Git Hook

pre-commit is a framework that manages the installation and execution of pre-commit hooks across multiple programming languages (including Python and R). These hooks are scripts that run automatically every time a git commit is made, aborting the commit if the script fails. These can be used to enforce standards in code format and style.

So, how would this be useful in practice?

To avoid the occurrence of a bloated git history, i.e. by mistakenly committing a large file to your repository, a good option is to use the check-added-large-files hook from the pre-commit-hooks package.

This will abort a commit if the file size is too large (default size limit = 500kB).

Here’s how you can install and configure check-added-large-files on Linux:

1. pip install pre-commit

2. Create new configuration file .pre-commit-config.yaml for configuring the check-added-large-files hook; its contents should look like:

# .pre-commit-config.yaml
repos:
  - repo: https://github.com/pre-commit/pre-commit-hooks
    rev: v5.0.0  # or use the latest version
    hooks:
      - id: check-added-large-files
        args: ['--maxkb=500']  # optional: set a custom size limit in KB

Here,

• repo is the GitHub repository where the pre-commit-hooks package is hosted.

• rev is the version of the pre-commit-hooks repository to use. You can specify a tag (like v5.0.0) or use master for the latest version, though pinning to a specific version is best for stability.

• hooks are a list of hooks to configure:

  • id: Specifies the hook to use, in this case, check-added-large-files.

  • args: Optional argument to set the maximum file size limit (in KB). By default, it is 500 KB, but you can adjust this as needed with --maxkb=<size_in_kb>.

This configuration will ensure that no added files exceed the specified size limit before a commit is allowed.

Note: By default, only files that have been staged for addition to the repository via git add will be checked by the check-added-large-files hook. To allow the hook to check all files in the repo (both staged and unstaged), add --enforce-all as an argument by modifying the above configuration yaml file with

args: ['--maxkb=1000', '--enforce-all']

The .pre-commit-config.yaml file should be stored in the root directory of your repository. This allows pre-commit to detect the configuration automatically when you run the pre-commit install command, which sets up the hooks for the specific repository. Once set up, pre-commit will refer to this configuration file each time you attempt to make a commit.

To set up the hooks after creating this file, run the following command in the root of your repository:

pre-commit install

This installs the hooks locally, ensuring they run on every commit.

Removing files

To remove a file (i.e., stop version controlling it) use git rm:

echo "Text in a file to remove" > FileToRem.txt

git add FileToRem.txt

git commit -am "added a new file that we'll remove later"

[master 706f40d] added a new file that we'll remove later
 2 files changed, 3 insertions(+)
 create mode 100644 .gitignore
 create mode 100644 FileToRem.txt

git rm FileToRem.txt

rm 'FileToRem.txt'

git commit -am "removed the file"

[master 284c62e] removed the file
 1 file changed, 1 deletion(-)
 delete mode 100644 FileToRem.txt

I typically just make all my changes and then just use git add -A for the whole directory (and it’s subdirectories; -A is recursive).

Un-tracking files

.gitignore will prevent untracked files from being added to the set of files tracked by git. However, git will continue to track any files that are already being tracked. To stop tracking a file you need to remove it from the index. This can be achieved with this command.

git rm --cached <file>

The removal of the file from the head revision will happen on the next commit.

Accessing history of the repository

To see particular changes introduced, read the repo’s log :

git log

For a more detailed version, add -p at the end.

Reverting to a previous version

If things go horribly wrong with new changes, you can revert to the previous, “pristine” state:

git reset --hard
git commit -am "returned to previous state" #Note I used -am here

If instead you want to move back in time (temporarily), first find the “hash” for the commit you want to revert to, and then check-out:

git status

And then,

git log

Then, you can

git checkout *version number*

e.g, git checkout 95f7d0

Now you can play around. However, if you do want to commit changes, you create a “branch” (see below). To go back to the future, type

git checkout master

Running git commands on a different directory

Since git version 1.8.5, you can run git directly on a different directory than the current one using absolute or relative paths. For example, using a relative path, you can do:

git -C ../SomeDir/ status

Tip

Cloning the MulQuaBio repository: You can clone the master repository of this book on your computer and regularly git pull it to keep it updated. That way the these notes and the associated code and data files are easily and locally available on your computer. DO NOT clone the repository into your own coursework repository though!

Running git commands on multiple repositories at once

For git pulling in multiple subdirectories (each a separate repository), here is an example:

find . -mindepth 1 -maxdepth 1 -type d -print -exec git -C {} pull \;

Breaking down these commands one by one,

find . searches the current directory

-type d finds directories, not files

-mindepth 1 sets min search depth to one sub-directory

-maxdepth 1 sets max search depth to one sub-directory

-exec git -C {} pull \ runs a custom git command one on every git repo found

Using git through a GUI

There are many nice git GUI’s (Graphical User Interfaces) out there, such as gitKraken. Or if you are using a code editor like Visual Studio Code, there are nice extensions that will give you considerable GUI functionality.

Tip

Checking git status: Always run the git statuscommand on a repository before pulling/fetching from, or pushing to a remote repository!

A friendly concluding note on git

Git can feel hard and unintuitive. It is likely that at some point you will make a mistake and lose some work as a consequence.

THIS IS NORMAL

Don’t worry about it, this has happened to pretty much everyone who has ever used git.

When things like this do happen though, it is imperative that you work out what went wrong so you can make sure it doesn’t happen again.

Questions to ask yourself include:

• Did I use git status to check what I was adding?

• Did I commit regularly enough, whenever I had completed a logical section of work?

• Did I make sure to only commit code rather than committing data files and outputs too?

• Did I make sure to gitignore the things I needed to?

When you get used to working with git, it becomes as natural as juggling chainsaws: i.e. it will always feel a bit odd and strange but it won’t be as scary as it once was.

Next Steps:

• Learn Commands: Familiarize yourself with common Git commands.

• Explore Branching: Practice creating, switching, and merging branches.

• Use Hosting Services: Try using platforms like GitHub to collaborate with others.

• Advanced Features: Delve into rebasing, cherry-picking, and submodules for more complex workflows.

Practicals

Instructions

Don’t modify anything (or refer to any files) in your local copy of the git repository of these notes (on Github). Copy whatever you need from the master repository to your own repository.

Git commit and push every time you do some significant amount of coding work (after testing it), and then again before the given deadline (this will be announced in class). This includes UnixPrac1.txt from the Unix Chapter.

Push your git repository

• The only practical submission for git is pushing your coursework git repository, .gitignore and readme files included. Make sure your .gitignore has meaningful exclusions, and your readme has useful information (as explained in the section above). Also search online “readme good practices” or something like that to find additional tips/info.

• Also, invite your assessor to your coursework repository (e.g, CMEECourseWork) repository with write privileges (AKA push privilege in GitHub). The current assessor is s.pawar@imperial.ac.uk (or “mhasoba” on both bitbucket and github).

Readings & Resources

General

• Excellent book on Git: http://git-scm.com/book

• Look up the Bitbucket Git resources

• Really great git tutorials: https://try.github.io

Markdown

• A cheatsheet for github-flavored Markdown: adam-p/markdown-here

Committing practices

• Some tips for good commit practices: https://chris.beams.io/posts/git-commit/

.gitignore

• Guidelines for the .gitignore file: https://labs.consol.de/development/git/2017/02/22/gitignore.html, https://docs.github.com/en/free-pro-team@latest/github/using-git/ignoring-files

Branching

• Guidelines for brancing: https://gist.github.com/digitaljhelms/4287848

• Git Kraken’s tips for collaborating effectively on github

---

Footnotes

1: There you will find the following phrase: “…one of the most annoying problems known to humanity: version-controlling Microsoft Word documents.”. LOL!

2: None of the computing weeks assessments will require you to use such large files anyway