On Tue, Sep 11, 2018 at 6:47 PM Junio C Hamano <gitster@xxxxxxxxx> wrote: > > Jonathan Tan <jonathantanmy@xxxxxxxxxx> writes: > > > By "without any objects" in your email subject, do you mean "without > > blob and tree objects"? If yes, there is some code in the > > md/filter-trees branch that can do that with a "--filter=tree:0" > > option. > > I too was wondering what the "without any objects" thing meant > myself. > Thanks for all replies, as you correctly deduced, I meant "without fetching any objects". The mentioned --filter=tree:0 would basically do what I want it seems, good to hear!. I wonder why not call it tree:none though to match blob:none. > What would it take on top of the following sequence to create such > an ultra-lazy clone? > > $ mkdir very-sparse && cd very-sparse > $ git init > $ git remote add origin $URL > Yes, this would be a good CLI API since the since the clone --no-checkout --filter --filter gets a bit long. Or maybe: git clone --lazy URL repo_local cd repo_local git checkout commit -- path/within/repo Or maybe even: git clene --lazy URL repo_local COMMITISH path/within/repo to do both in one go. People are also interested in commit-less directory / file "clones" BTW: https://stackoverflow.com/questions/2466735/how-to-checkout-only-one-file-from-git-repository-sparse-checkout > At this point, the repository does not have any object, but it > already knows whom to talk to to get the objects in the project. > The remote must be configured so that it is willing to feed you any > object you name, but would it be just some "git config" magic after > the above three steps to make it as if it was prepared with "git > clone --filter="? If so, what does that magic look like? > > Thanks.
