IDE integration or online service

Do we need to integrate DVC to IDEs?

Open questions:

  1. Which IDE to choose\prioritize? Our research shows that PhCharm and Jupyter are the most popular ones. Also, some people use RStudio, Python in Visual Studio. A big portion of data scientists use vim\emacs of course. It is not easy to integrate with vi\emacs.
  2. Which features have to be integrated into IDE:
  • Graphs: like show pipeline and workflow. With current state probably.
  • Data files versions history. Pipeline versions.
  • Experiments/branches/commits and metrics visualization.
  1. Should we ignore IDEs and implement all these features as an online service instead? The service will “cover” all IDEs at once. Which of the features we still should cover through command line if the online service will be available?

We need your ideas and your opinion. Please feel free to comment.

2 Likes

I use a combination of RStudio and emacs for my research workflow, so it would be great to have integration in emacs with a full-on dvc-mode.

With R/RStudio it should be quite straightforward, even just a package, like git2r for git.

What I do now in emacs is to run a shell command from a project with projectile-run-shell-command-in-root, which I’m planning to map to a keybinding or hydra. But something like magit would be awesome!

The features that I would find most useful are adding files with dvc-add, pulling and pushing, and dvc-repro.