This example in the docs seems a bit odd, and a basic question

[JohnAtl]

(Tried this on the Discourse, but discussions seem to be happening here.)

I’m a bit of a noob, and I’m trying to understand how dvc install changes the steps introduced in Dr. O’Brien’s introductory videos.

Question 1

The final example on this page of the docs shows a source file stored in a dvc repository, thus when changed, dictates running repro to create the data, then everything is somehow up to date.

I would think that:

source would be saved to git
running modified source with repro would result in a changed dataset that would need to be added, committed, and pushed.

Question 2

When I’m working on my project and I run git commit, dvc status is run and (let’s assume) shows me that some of my dvc-managed files have changed. Am I correct in these commands (from memory):

# add all modified files in data folder
dvc add data
dvc push
# make sure files changed by dvc makes sense
git status
git add -a -m “dvc modified”
git push
This seems pretty convoluted, so if there’s a better way, would love to know. If not, not looking a gift horse in the mouth 🙂

Question 3

Okay, a bonus - what about merging, say, a branch into main? Say, the signatures in the *.dvc files don’t match. Is it just a matter of always selecting the branch data over main in the conflicted .dvc files (unless something has gone really wrong)?

Thanks!

[dberenbaum]

Hi @JohnAtl, it sounds like you have the right ideas here.

Question 1

This example is based on the example pipeline in our get started docs. src/featurization.py is tracked in git, but it's also a dependency of a stage in the pipeline, so dvc still checks whether the pipeline has been run with the changes to that file. The changed dataset is tracked in dvc by dvc.lock, which is what you see being added in that example.

Question 2

Yup, that looks about right. It is very explicit, but we prefer to make it that way so it's clear what's happening.

Question 3

Have you taken a look at https://dvc.org/doc/user-guide/how-to/resolve-merge-conflicts?

[JohnAtl]

Thanks for the information!
Sometimes it's good just to know what 'normal' is.