I want to read from the file where git stores commit history to store each commit information in my project's DB and display all histories in my project view.
Add a comment |
2 Answers
There's no single file you can interrogate to get the commit history. There are plenty of good explanations of git's object model (e.g. git for computer scientists, Pro Git, the git community book), but it might be useful to have a quick explanation here:
There are various types of objects in git, most importantly:
- blobs (files) - just binary files
- trees (directories) - a tree is a list of other objects (usually blobs and trees) with their name, a hash and a limited set of permissions
- commits (versions) - each commit includes the hashes of its parent commits, the author, the commit message and other metadata
Each of these is identified by a hash of its contents, and this hash is known as the object name - these are the 40 digit hex strings you've probably seen in the course of using git. Each object is stored in the .git/objects/ directory, either as a loose object (one per file) or as one of many objects stored efficiently in a pack file. The file .git/HEAD represents the version that your repository is currently at, and usually contains a reference to a particular branch, represented by a file under .git/refs/heads or a reference stored in pack file. (HEAD may also point directly to a particular commit's object name.) One of these files representing a branch, such as .git/refs/heads/master, just contains an object name.
In order to traverse the history back from this branch tip, git will find the object named in that file in the object database, and recursively follow the pointers to its parents.
However, for the use case you describe (i.e. traversing the history to export it), I would strongly suggest that you do one of the following:
- Invoke git commands to find the history. If you stick to using the so-called "plumbing" commands, their output should be stable across git versions.
- Use the libgit2 library to interrogate the repository. libgit2 is a fully re-entrant library for interrogating git repositories, which now has bindings for many languages.
2 Comments
wilhelmtell
You're missing the 4th object: tags
Mark Longair
@wilhelmtell: That's precisely why I said "There are various types of objects in git, most importantly [...]".
You can get an idea of how git stores objects here: How git stores objects. Read the other chapters in that book in the "Internals and Plumbing" section to understand how it all works.
Essentially, there is no "file where the commit history is stored", it's more complex than that. You should use one of the existing APIs for your language of choice (search for "git API" and your language, there are a lot out there).
