Git clone bare repo without blobs
Asked Answered
B

2

8

On my git repository, I use an algorithm to assign every commit one or more unique version numbers, based on branch names and tags. I want to use this mechanism with another large repository, that I would like to clone without transferring any files.

A bare clone helps me to get rid of the working copy of the blobs, but it still downloads them from the server. A shallow clone with --depth 1 skips most blobs, but also skips downloading the metadata for all commits except one.

Is there something like git fast-export --no-data which I can use on the client-side to get the graph information containing commit metadata and maybe filenames without cloning the repository from my server first? Ideally I would be able access the metadata like any other (bare, shallow) repo via git log|show|rev-parse|show-ref.

(I know git LFS and git Annex exist and can help reduce the size of some repos, but I can't use them on an existing repository without changing it.)

Bonni answered 19/12, 2016 at 22:30 Comment(0)
C
3

Is there something like git fast-export --no-data which I can use on the client-side?

No: beside git ls-remote (which gets metadata only for the heads of the remote repo), anything else would get the full repo history.

You would need your repo managed by a Git hosting service, like GitHub, providing an API (like the commits API), in order to query metadata without data.

Cason answered 20/12, 2016 at 5:22 Comment(0)
B
6

Another Idea for this some time later: As of 2017, the pack-protocol now allows partial clones, and there is a --filter=blob:none available that omits all blobs -- which should be sufficient on the server-side.

Given the current server-side implementation, sadly, this doesn't work as well as one would hope:

C:\Users\phi1010>git clone https://github.com/torvalds/linux.git --filter=blob:none
Cloning into 'linux'...
warning: filtering not recognized by server, ignoring
remote: Enumerating objects: 6876195, done.
[...]

Github even announced the support of the v2 protocol and its filtering capabilities, but this neither works with -c protocol.version=2, as also stated on Do GitHub and GitLab support git clone's --filter parameter?

Bonni answered 11/9, 2019 at 20:51 Comment(1)
It looks like this is now supported by Gitlab.Servais
C
3

Is there something like git fast-export --no-data which I can use on the client-side?

No: beside git ls-remote (which gets metadata only for the heads of the remote repo), anything else would get the full repo history.

You would need your repo managed by a Git hosting service, like GitHub, providing an API (like the commits API), in order to query metadata without data.

Cason answered 20/12, 2016 at 5:22 Comment(0)

© 2022 - 2024 — McMap. All rights reserved.