Git Object Database
A git repository is essentially a key-value database, where the value is a git object and the key is a SHA-1 value which is is assigned to each value when firstly added into the database.
4 Kinds Of Git Objects
TL;DR
A tree (like directories in file systems) as a directory save metadata and address (the SHA-1 keys) to contained blob objects as well as sub-trees.
A commit references to a root tree object as well as metadata the commit. It represents a backed history of the working tree. That is why git is called by someone as a file system.
Reference can be seen as the human friendly version of hash keys that git provides for user to reference to objects (e.g. commits) with special meanings. Using references is like declaring a constant variable in programming to store a fixed magic number (the hash values).
Blob
All files (no matter what format it has) you submit to git are stored as blobs (binary large object 数据对象).
They all reside under <your repo>/.git/objects/. Take the path
<yourrepo>/.git/objects/fe/287a3e58d3d65a47e82f521b11df80ed2e540f for
example, the /fe/ directory represents the first 2 hash digits of all blobs
stored under this folder, the 287a3e58d3d65a47e82f521b11df80ed2e540f is the
blob hash value (with the prefixing 2 digits fe stripped off).
Use the git hash-object plumbing command to pack files into blobs as well as
assigning a SHA-1 value as the key, then add the key-value pairs into database.
The blobs can be inspected by the git cat-file plumbing command.