Tree

A tree object describes the contents of one directory. Each line in a tree is one entry:

100644 blob 5f7e9c12...    README.md

100644 blob 8a3f2e91...    index.html

040000 tree e2b5a91f...    images

100644 blob b1d4a7e0...    style.css

Four columns:

• Mode - 100644 is a regular file, 100755 is executable, 040000 is a subdirectory, 120000 is a symlink, 160000 is a gitlink (submodule). Git stores the mode as a number; git ls-tree always prints it as six digits with a leading zero.
• Type - blob for a file, tree for a subdirectory.
• SHA - a pointer to the blob or tree object.
• Name - the filename or directory name within this directory.

The file content lives in a blob. The tree links a name to that content.

Recursive structure

A tree entry of type tree points to another tree object for a subdirectory. This builds a hierarchy of tree objects that mirrors the project's directory tree:

root tree

   ├── blob README.md

   ├── tree images/

   │      ├── blob logo.png

   │      └── blob banner.jpg

   └── blob style.css

When a tree is created

On git commit. Before the commit, no tree exists; there is only .git/index, which holds a flat list of paths. git write-tree converts the index into a hierarchy of tree objects and returns the SHA of the root.

Inspecting tree contents

git cat-file -p <tree-sha>

git ls-tree HEAD              # tree of the latest commit

git ls-tree -r HEAD           # recursive, flat list of all files

Deduplication at the tree level

If a subdirectory has not changed between two commits, its tree object is identical. The parent tree points to the same SHA. Git deduplicates not only individual files but entire subdirectories.