Initialize data version control for managing test images

[maxrjones]

Description of proposed changes

This PR implements the solution for storing baseline test images proposed in #5724. Only a file describing the contents of a directory with baseline images are stored in the git repository; the actual PS file will be stored in the DAGsHub repository (e.g., https://dagshub.com/GenericMappingTools/gmt/src/manage-tests-dvc-refactor/test/baseline/api). While by default DAGsHub will store all versions of the images that are pushed to the remote storage, these can be easily garbage collected from either the user's local cache and the remote storage (if we reach the 10 GB limit) without breaking the git history.

These are the specific changes in this PR:

• Initialize dvc in the repository
• Add dvc as a dependency for developers
• Add dvc to the CI test workflow
• Migrate test/api and test/blockmean .PS files to dvc
• Add instructions for migrating baseline images to dvc to the contributing guide

Addresses #5724

Reminders

• Make sure that your code follows our style. Use the other functions/files as a basis.
• Add tests for new features or tests that would have caught the bug that you're fixing.
• Describe changes to function behavior and arguments in a comment below the function declaration.
• If adding new functionality, add a detailed description to the documentation and/or an example.

[PaulWessel]

I will let others do the initial approval on this...

[weiji14]

Sorry for the long delay! I only have a few minor comments for now, the dvc setup and instructions seem mostly ok for me, but I'll need to run a proper git clone of this branch and do some proper tests in the coming days. Disk space on my 10 year old laptop isn't great for big repos, so I'll need to use some university resources 🙂

[maxrjones]

Sorry for the long delay! I only have a few minor comments for now, the dvc setup and instructions seem mostly ok for me, but I'll need to run a proper git clone of this branch and do some proper tests in the coming days. Disk space on my 10 year old laptop isn't great for big repos, so I'll need to use some university resources 🙂

No worries on the timing - I really appreciate your willingness to give this a review!

[maxrjones]

Would anyone be willing to give this a review shortly? I think it would be great to make a decision about whether to go with dvc before we update the baseline images for the ~25 failing tests.

Here are some generally instructions for testing out the changes:

1. Install dvc (https://dvc.org/doc/install)
2. Checkout branch: git checkout manage-tests-dvc-refactor
3. Pull baseline images stored in dvc: dvc pull
4. Run the tests in your regular way (e.g., https://docs.generic-mapping-tools.org/dev/devdocs/contributing.html#running-tests).

Afterwards, you would need to do some cleanup (this wouldn't be necessary when switching branches if this PR gets merged in):

1. Checkout the other branch (e.g., git checkout master)
2. Navigate to the base of the repository (e.g., gtop if using aliases)
3. Delete dvc related files (rm -rf .dvc and rm -rf test/baseline)

One note: migrating all the tests files to dvc would initially add ~100 MB to the repository size for people who run GMT tests locally (see #5724 for details) if we don't also delete git history (which I would prefer avoiding). For anyone who doesn't run dvc pull, this change would have a negligible impact on repository size. To avoid this step increase in size for those who run the tests, I think we can only migrate baseline images when tests fail due to code updates or when new tests are added.

[PaulWessel]

I will give it a go following the instructions - will slack if running into troubles!

[maxrjones]

Thanks Paul!

[PaulWessel]

Was able to successfully do all 7 steps.

[weiji14]

Haven't found the time to run the full tests, but the documented dvc steps look ok to me, so won't hold this back any longer. 🚀

[maxrjones]

@PaulWessel do you approve of merging this in now? If so, I could use dvc for the postscript files in #6199 and I could update the reference images for the failing tests using dvc.

[PaulWessel]

Yes, I think that would be great. Go ahead.