What is Git?

Marie-Hélène Burle






First, we need to answer the question:

What is version control?




Whenever we work on important documents, we know that we should keep key versions

Example:

  • The version of a manuscript that we sent to our supervisor
  • The revised version after we addressed their comments
  • The revised version after we addressed reviewer comments
  • Etc.


Home-made versioning:

It is quite messy…


From PhD


And inevitably, it leads to this:


From Geek&Poke







Version control systems (VCS) are software that handle versioning effectively







Which VCS should I use?


Several systems have been developed over the years with various functioning

Then came Git…


Git

Git is an open source distributed VCS created in 2005 by Linus Torvalds for the versioning of the Linux kernel during its development

In distributed VCS, the full history of projects lives on everybody’s machine—as opposed to being only stored on a central server as was the case with centralized VCS. This allows for offline work and multiple backups

Git also introduced an extremely powerful and light-weight branching system

Git is extremely powerful and almost universally adopted







How does Git work?



Git saves the history of a project as a series of snapshots

noshadow

The data is stored as blobs, doesn’t create unnecessary copies (unchanged files are referenced from old blobs), and uses excellent compression



These snapshots are identified by commits

noshadow




Each commit has a unique hash and contains the following metadata:

  • Author
  • Date and time
  • The hash of parent commit(s)
  • A descriptive message



When you create the 1st commit, a pointer called a branch is created and points to it

noshadow

By default, that first branch is called main

Another pointer (HEAD) points to the branch main. HEAD indicates where you are in the project history



As you create more commits the pointers HEAD and main move automatically

noshadow



As you create more commits the pointers HEAD and main move automatically

noshadow



For simplicity, the diagrams can be simplified this way

noshadow







How can these commits be used?



You can revisit old commits by moving HEAD to them

noshadow

This will uncompress the corresponding snapshot and you can look at the state of your files at that commit before going back to your branch

You can also print the differences between various commits



You can create multiple branches to explore freely and safely

noshadow

HEAD can be moved back and forth between branches



You can merge branches to bring your experiments into your main branch

noshadow

From xkcd.com








Time to get started!