Building Nit: My Own Git in Rust

I wrote a small Git clone in Rust. I called it Nit. It was one of the best ways to understand how Git really works.

Why I Built It

You use Git every day, but most people only know a few commands. I wanted to know what happens inside when I run git commit. The best way was to build it myself.

How Git Works Inside

Git stores everything as objects in a folder called .git/objects. There are three main types:

• blobs: the file content
• trees: folders that point to blobs and other trees
• commits: a snapshot with a message, author, and parent

Each object has a name based on its content, using SHA-1. Git also compresses objects with zlib. My Rust code did the same with the flate2 crate.

The Index

The index (or staging area) is a binary file. It lists the files you added with git add. I had to implement its exact format, so that real Git tools like git status could still read my repositories.

Commands I Implemented

• nit init
• nit add
• nit status
• nit commit
• nit clone (local repositories)
• and more plumbing commands

Repositories I create with Nit can be opened with normal Git. That was a great test.

What I Learned

• hashing and content-based naming
• binary formats and parsing
• compression with zlib
• how commits link together as a graph

Hard Parts

• packfiles with delta objects are still not complete
• .gitignore is not implemented yet
• remote clone over HTTP is incomplete

These are the parts I want to do next.

Final Thoughts

Building Nit made me a better Git user and a better programmer. If you want to understand a tool deeply, try to build a small version of it.

Repository: GitHub - m-elhabib-dev/nit