Add "git blob hash" as available hash

[markrages]

Description of the desired feature:

In my project, I wanted to use pooch as a cache for getting files off of Github.

It is more convenient to use Git's "blob hash" for this, rather than a normal SHA1 or whatever. This is because I can "git clone" a zero-depth repository and see these hashes without ever downloading the files. This saves a lot of time and bandwidth for a large Git repository.

Are you willing to help implement and maintain this feature?

I have implemented this already, in a monkey-patched way.

See https://github.com/markrages/svd_gdb/blob/master/svd_gdb/github_dl/update_registry.py for a script to create a Pooch registry from a remote Git server.

See https://github.com/markrages/svd_gdb/blob/master/svd_gdb/github_dl/github_dl.py for an example of use. (and the monkeypatched hash calculation.)

Is there any interest in supporting this?

[leouieda]

Hi @markrages thanks for the suggestion and for showing an example! I hadn't seen this before and it's nice to know the use case exists.

This would definitely be worth adding. It won't add any dependencies from what I can tell (just a system call, right?). Do you think adding the git: hash prefix would be unique enough or are there multiple kinds of git blob hashes?

If you'd be willing to implement this, I'd be happy to merge. We'd need a test and an entry in the corresponding page in our docs (would be great to link to your project as an example).

[markrages]

This would definitely be worth adding. It won't add any dependencies from what I can tell (just a system call, right?). Do you think adding the git: hash prefix would be unique enough or are there multiple kinds of git blob hashes?

That is a matter of some controversy: https://lwn.net/Articles/898522/

I think it would be prudent to distinguish SHA1 and SHA256 hashes, just in case.

I will get a patch put together.

[leouieda]

👍🏽 thanks!