Reputation: 730
I'm currently working on an application which runs a model, and a common request is that we need to compare the current model to previous versions with the same inputs. As of now we keep a copy for each production version so it looks like this:
application_base
as their own folder, and that entire thing is pushed to the repository. This seems like ... not the best way to do this, and it causes pull requests for each version to make it look like hundreds of files had code changes, even if it was only a couple. I've seen a couple of approaches and all of them seem hacky, I can't be the only one who's ever had this issue, what's the best practice for this?
Thanks.
Upvotes: 1
Views: 1680
Reputation: 746
You can use git tags to mark versions/releases. This way, you can do all the work in one directory. Diff's instead of being huge are gonna be just the differences.
It is also easy to checkout the specific tag so you should be able "to compare the current model to previous versions with the same inputs".
Upvotes: 2