Git Internals

Working Tree, Index, and Object Store

Separating the working tree, index, and object database is the clearest way to understand add, commit, restore, and reset.

Many Git problems become easier once these three layers are separated:

  • working tree: files you are editing now
  • index: the next snapshot being prepared
  • object store: durable Git objects already written

Why this matters

The real difference between many commands is simply which layer they touch:

  • git add updates the index
  • git commit writes index state into durable objects
  • git restore can restore the working tree or the index
  • git reset can move refs and may also realign the index and working tree

If you treat the index as “the candidate snapshot for the next commit,” Git behavior becomes much more predictable.

Related Reads

Git Object Database

Understand blob, tree, commit, and tag objects, and why Git describes itself as a content-addressable object database.

Git References and HEAD

Put branches, tags, remote-tracking refs, and HEAD into one model so Git's pointer behavior becomes easier to reason about.

Remote-Tracking References

Understand the difference between `main` and `origin/main` so fetch, pull, push, and branch sync become easier to reason about.

PreviousGit Object Database

Understand blob, tree, commit, and tag objects, and why Git describes itself as a content-addressable object database.

NextGit References and HEAD

Put branches, tags, remote-tracking refs, and HEAD into one model so Git's pointer behavior becomes easier to reason about.