Mini Git

Data structures and binary serialization, with a result you'll actually use: a working .git implementation that can hash, store, and version real files. The core idea — content-addressed storage — reappears later as the conceptual root of the B-Tree pages in the database engine, just keyed by offset instead of hash.

Planned notes

• The object model: blobs, trees, and commits as a content-addressed graph
• Hashing objects: SHA-1 (Git's legacy default) and the SHA-256 migration
• The blob: how Git stores file contents, zlib-compressed, keyed by hash
• The tree object: directories as ordered lists of (mode, name, hash)
• The commit object: parent pointers, author/committer, message — and how history becomes a DAG
• Writing objects to .git/objects/<aa>/<bb...> and reading them back
• A minimal init, hash-object, and cat-file
• Building a tree from the working directory (write-tree)
• Creating commits and walking parent pointers (log)
• Branches and refs: just files containing a hash, and why that's enough
• The index/staging area: what add actually does before commit
• Diffing two trees — the first real algorithmic piece of the project

Core sources

• "Write yourself a Git" (wyag) — the most complete from-scratch walkthrough of the object model.
• "Build Your Own Git" (codecrafters) — staged milestones with test coverage.
• Pro Git (Chacon & Straub), "Git Internals" chapter — free, canonical, and short.

Connects to: KV Store & Durability · Craftsmanship Low-Level II