Data model reference¶

All persistent state lives in two places: a Neo4j graph (structure and metadata) and an object store (file bytes, addressed by SHA-1).

Graph nodes¶

Label	Module	Key properties	Purpose
`Branch`	`neogit.model.neo`	`name`	Named pointer to a commit
`Commit`	`neogit.model.neo`	`hash`, `sha1sum`, `name`, `date`, `description`	A snapshot
`Tree`	`neogit.model.merkle`	`hash`, `sha1sum`	A directory in the Merkle tree
`Blob`	`neogit.model.merkle`	`hash`, `sha1sum`	A file content reference
`PluginRun`	`neogit.model.neo`	(varies)	Attaches downstream analysis state to a commit

hash is the Merkle hash of the node; sha1sum is the SHA-1 of the content (for Blob) or of the canonical serialization (for Tree / Commit).

Relationships¶

Branch --[TRACKS_COMMIT]--> Commit
Commit --[HAS_PREVIOUS]--> Commit          (commit history)
Commit --[OWNS_FILESYSTEM]--> Tree         (the root directory)
Tree   --[HAS_CHILD_TREE {name}]--> Tree   (subdirectories)
Tree   --[HAS_CHILD_BLOB {name}]--> Blob   (files)

Edge properties (name) carry the path component, so the same Tree or Blob can appear under different names in different parents — that's the whole point of content addressing.

Object storage layout¶

The object store contains one entry per unique Blob SHA-1. Keys are the hex SHA-1; values are the raw file bytes. Identical files across commits, branches, or snapshots deduplicate automatically.

Example Cypher queries¶

Latest commit on master:

MATCH (b:Branch {name: "master"})-[:TRACKS_COMMIT]->(c:Commit)
RETURN c

Full file tree of a commit:

MATCH (c:Commit {name: "snap-1"})-[:OWNS_FILESYSTEM]->(root:Tree)
MATCH path = (root)-[:HAS_CHILD_TREE|HAS_CHILD_BLOB*]->(n)
RETURN path

History walk:

MATCH (b:Branch {name: "master"})-[:TRACKS_COMMIT]->(head:Commit)
MATCH path = (head)-[:HAS_PREVIOUS*]->(ancestor)
RETURN ancestor.name, ancestor.date
ORDER BY ancestor.date DESC