How Git stores data

Git is a content-addressable filesystem: at its core is a simple key-value data store ¹². Store some content and the key you get back is its SHA-1 hash, a checksum of the content plus a header ³, usually written as 40 hexadecimal characters ⁴. The object goes under .git/objects, in a folder named for the first 2 characters of the hash and a file named for the other 38 ⁵, compressed with zlib ⁶. A blob holds only the content, not the file name ⁷.

Names live in trees: each tree entry is the SHA-1 of a blob or subtree with its mode, type and filename ¹. A commit names the top-level tree of a snapshot and its parent commits, if any, plus author and committer ². Because every commit points at its parents, the commits form a directed acyclic graph ³, a graph of the same kind PageRank walks over links. Blob, tree, commit and tag are the four object types ⁴.

A branch is a simple pointer or reference to the head of a line of work ¹. On disk it is a file holding the 40-character hash of a commit, so creating one means writing 41 bytes: 40 characters and a newline ². HEAD is usually a symbolic reference to the branch you are on ³, and the branch head moves forward as you commit ⁴.