0

Lots of questions on how to keep a file in Github and not track its changes. I wonder if there is any way to update said file and not track the changes.

The use case is that I have some large files in a project repo that are constantly updated. I would like these to be kept part of the project and available on Github, but they are data, not codebase, and I do not need to preserve their historical states. It seems it would be in my and Github's interest not to track their changes.

Is there any way to delete previous commits from the git tree if, for example, they only edited the data files and no others?

geotheory
  • 22,624
  • 29
  • 119
  • 196
  • You cannot do this with *Git*, and to the extent that you're using GitHub as a Git hosting provider, you therefore cannot do it with GitHub either. What you *can* do is not store the large files in a Git repository at all: store, instead, instructions for downloading the large files via something other than Git, and have the software do that automatically. Then, whether and how you can store these files *on GitHub itself* becomes an independent question. – torek Oct 07 '21 at 21:56
  • I was starting to suspect as much. Thanks would you like to answer with this? – geotheory Oct 07 '21 at 23:57

2 Answers2

1

The general solution to this is to store larger files—often binary assets and the like—outside the repository. This can be anywhere, even in some other Git repository on GitHub if you like, or on some Git-LFS hosting site, or whatever, as long as there's some way for your users to retrieve the files. You then have your software store the URL (or URLs) by which it can retrieve the files; that's what goes into the Git repository, which makes for a small (or at least smaller) repository that fits within the usual GitHub limits.

You can either replace the binary files at the storage site in-place, keeping the URLs, or come up with new, slightly-different URLs so that it becomes possible to tell that some old checkout no longer has the binary assets available. The latter system is equivalent to using Git-LFS, which consists of install-able pieces of software that "wrap" your Git system and do this kind of file-shuffling for you semi-automatically, making whoever uses the Git repository "see" the large files as long as they have installed and use Git-LFS themselves. Those who haven't installed Git-LFS see instead the "pointer files" (which consist of URLs and hash-ID checksums), which is what really do end up in your repository.

(GitHub do support Git-LFS themselves, but you'll generally need to pay for it, as the initial storage limits are tiny, as these things go.)

torek
  • 448,244
  • 59
  • 642
  • 775
0

Did you try github lfs? In one of projects we had to store game assets in there.

Btw you will need to rewrite whole your history to exclude large files from it lfs migrate

Ayder Memetov
  • 11
  • 2