[Nix-dev] Fetching only a small part of a huge git repository

[Michiel Leenaars]

Hi all,

I was wondering if there is a recommended way to retrieve only a certain
path (or a set of paths) from a git repository as a source for a Nix
closure? The use case is working with a repository that is huge but only a single
small folder is needed (a practical example is Google fonts on Github - hundreds 
of Mb which is a bit much to store if all that is needed is a single font such
as the Roboto font which is the default font of Android). fetchFromGitHub and 
fetchgit seem to always copy the entire repository.

git has an option for this (git archive, [1]) which seems to do exactly what
I want - a checkout of only a single folder of a larger git repository -
and there are other options like a filter branch or a sparse-checkout.
I've grepped nixpkgs to see if there was any usage of git archive but could 
not find it - as a user I would probably like to see something in 
fetchFromGitHub with something like:

  url = fetchFromGitHub {
    owner = "name";
    repo = "reallyhugerepo";
    rev = "revisionnumber";
    paths = [ "tinypath-a" "path-b" ];
    sha256 = "22bfd184608ca00e38515c8d5d01d38460a66a503c75a35d0fea3e33559bb3b6" 
    };

That seems like a reasonable and declarative way to specify what can be
discarded (a path in git archive can be a file or a folder with all the files 
in it btw).

Probably I'm not the first to think he needs this, and perhaps such a thing 
is currently already possible? If so, I'd love any pointers ... If this
is a bad or redundant idea, please also let me know ...

Best,
Michiel

[1] https://www.kernel.org/pub/software/scm/git/docs/git-archive.html