What database does Git use for metadata and how can I access it?

Question

git log is a powerful Git command which allows to make queries on a lot of things: file authors, commit description, commit dates etc. Git itself says that metadata is stored in some database. I wasn't able to find what kind of database anywhere.

So I'm wondering what is this database? Is it relational and perhaps I can somehow use SQL in order to perform the queries directly from the database instead of using git log?

EDIT: this question is different from How does git store files? because that question mainly asks about how Git stores committed files while my question asks about metadata storage (metadata != data).

Why would you like to do so ? There are plenty of libraries to work with git in your favourited programming language — gogaz, Aug 20 '18 at 12:22
@gogaz because it will be easier for me to perform complicated queries in SQL rather than using multiple git flags. Also I'm genuinely interested how Git saves the metadata — Yos, Aug 20 '18 at 12:23
Git uses compressed files basically, the repository is a combination of metadata-objects and blobs (the actual file versions). The metadata objects are just compressed text files. So there is no SQL database at all. You're best off using a library for it, like libgit2 which is available for many runtimes and languages. — Lasse V. Karlsen, Aug 20 '18 at 12:25
Just to take an example, if you create a new repository, create a file, add the file to the index, then commit it, your "database" consists of the commit object (compressed text), a tree object (directory structure, compressed text) and the file blob, as 3 separate files, all under .git\objects — Lasse V. Karlsen, Aug 20 '18 at 12:27
Possible duplicate of [How does git store files?](https://stackoverflow.com/questions/8198105/how-does-git-store-files) — phd, Aug 20 '18 at 13:04
Specifically, it is a simple key-value store with the keys being hash IDs. The values stored under these keys are one of four value types: *annotated tag*, *commit*, *tree*, and *blob*. The first three have in their values fields that are defined to contain more keys and the entire set of values always forms a Directed Acyclic Graph or DAG. — torek, Aug 20 '18 at 15:40

score 2 · Accepted Answer · answered Aug 20 '18 at 12:44

In that description, Git describes its internal state as a "database" in the most literal sense of the word. From Wikipedia:

A database is an organized collection of data, stored and accessed electronically.

It is not a relational database, or one you can easily query using a query language such as SQL.

You can however access the database programatically. For example using LibGit2Sharp and Linq, if you're on .NET:

using (var repo = new Repository(@"path\to\repo.git"))
{
    // Rev walking
    var commits = repo.Commits
                      .StartingAt("sha")
                      .Where(c => c.Author.Email.Contains("@example.com"))
                      .ToList();
}

score 2 · Answer 2 · answered Aug 20 '18 at 12:52

From the git glossary (at https://git-scm.com/docs/gitglossary)

object database - Stores a set of "objects", and an individual object is identified by its object name. The objects usually live in $GIT_DIR/objects/.

The database is the collection of files typically stored under .git/objects/ (or, for a bare repo, just objects/).

This is not an RDBMS. It has no general-purpose query language, but rather it's designed to work with various git tools (and vice versa) to enable fast access to the information git needs.

What database does Git use for metadata and how can I access it?

2 Answers2