DVC diff tool support (like git difftool)

[Voxelworld]

Hi all

I had a discussion on discord regarding a common usecase which may be of general interest.

I would like to use DVC for versioning a larger ML repo (~30K folders, 200K files). A typical workflow would be that a script or user is adding and improving data in the workspace, and afterwards I would like to check the changes for some samples, before adding them.
dvc diff lists the changed files and hashes, but currently there is no support to pass two file versions to a diff tool directly.

Ideally I would like to configure different tools for diffs on binary files (like images and masks) and text files (like json files) by its extension.
Is this a usecase worth supporting in future DVC versions?

'skshetry' suggested to extend it using dvc diff --json data and create an own script that calls specific difftools. Has someone in the community already implemented this workaround?

[casperdcl]

Duplicate of #770?

[dberenbaum]

Closing as duplicate

[casperdcl]

Thanks @Voxelworld for the request though, please do upvote/track #770 :)