import: Handle non-DVC Git repositories

[Baranowski]

After #2889, dvc import can also import files that are tracked by Git but not DVC. DVC still requires that they come from a DVC repository rather than any Git repository, although there is no longer need for that.

[chatcannon]

Hello all, @shcheklein asked me to work on this as a recruitment test. I have tracked down the reason why this is not yet working: the import process tries to create a Repo instance from a temporary local clone of the url argument (i.e. the source repo from which the file should be imported). Since this directory does not contain a '.dvc' subdirectory, the Repo.find_root function fails.

[chatcannon]

Here are some ideas for resolving this; I'm not sure which fits better with the overall design:

• Add code to Repo.find_root to allow creating a Repo instance from a git repo with no '.dvc' directory
• Run the equivalent of dvc init in the temporary clone before calling external_repo
• Have a completely separate code path for importing non-DVC repos

[efiop]

@chatcannon Sorry for the delay! The first two ideas are pretty hacky, the third one would ideally be the way to go. Unless there is a good reason to go with the hacks, of course 🙂But I haven't seen any, it looks like you could gracefully catch if the repo is not a dvc repo and fall back to git logic. Or you could clone git repo first and then check if it is a dvc repo or not. Maybe there is a better way to do it too, your call 🙂

[Suor]

@chatcannon I would say we can integrate this into existing code path. Otherwise will need to duplicate lots of things. Just don't create a Repo if there is no .dvc dir in there.

P.S. Some refactoring might be needed to both not make it messy and avoid code duplication.