Git Internals

Quick refresher to git

$ git init
$ git add foo.txt
$ git commit
$ git push

What is git? A version control system

What is git? A content-addressable filesystem/object-store

What is git? A content-addressable filesystem/object-store

What is git? A content-addressable filesystem/object-store

Side note: hash functions

$$h(x) : \{0, 1\}^* \to \{0, 1\}^N$$

Side note: hash functions

$$h(x) : \{0, 1\}^* \to \{0, 1\}^N$$ $$SHA1(x) : \{0, 1\}^* \to \{0, 1\}^{160}$$

Side note: hash functions

Take CO487 to learn more :)

Blobs

• Stores data for individual files

blob 4\0 test

• Filename is SHA1 Hash of content

Trees

• Stores data for directories

tree N\0
100644 blob 30d258... test.txt

• Filename is SHA1 Hash of content

Commits

• Stores data for commits

commit N\0tree 095a05...
parent <parent sha>
[parent <parent sha> if several parents from merges]
author Neil Parikh <parikh.neil@me.com> 1658703268 -0400
committer Neil Parikh <parikh.neil@me.com> 1658703268 -0400

hello

• Filename is SHA1 Hash of content

Refs

• How does git know what the latest commit is (to use for the parent sha)?
• Refs

Refs

• Refs are pointers to commits
• Branches are just refs (see .git/refs/heads/BRANCH_NAME)

Refs

• How does git know which ref is "current/active"?
• Special ref called HEAD

Implications of this approach

• Git doesn't store diffs
• Checkout arbitrary commits in O(1) time
• Deduplication of identical files

Questions?