Understanding Git as a graph of immutable objects
Summary
Charles Duan explains blobs, trees, commits and references as the connected data model behind Git's everyday commands. Git uses four central object types. Object names are derived from cryptographic hash values.
Ideas
- Blobs store file contents without path names.
- Trees link names with blobs and further trees.
- Commits point to a tree and previous commits.
- References give immutable commit objects movable names.
- The index prepares the tree of the next commit.
- Merge commits connect several parents in a shared graph.
Insights
- Commands become predictable once their changes to objects are visible.
- Content addressing automatically makes identical data shareable.
- History is a graph and not necessarily a linear timeline.
- Movable references provide convenience on top of immutable states.
Facts
- A commit does not contain a complete branch name.
- Tags can provide fixed names for particular objects.
Recommendations
- Examine a test repository with git cat-file and git ls-tree.
- Check references and the reflog before risky operations.
- Think in common ancestors and graph edges when merge problems arise.
References
Links to the original source and the Web Archive open in a new tab.