tomMoral / dicodile

Experiments for "Distributed Convolutional Dictionary Learning (DiCoDiLe): Pattern Discovery in Large Images and Signals"
https://tommoral.github.io/dicodile/
BSD 3-Clause "New" or "Revised" License
18 stars 10 forks source link

Automatic build and publish action #31

Closed tomMoral closed 3 years ago

tomMoral commented 3 years ago

I create the PR as @deepcharles might be able to help us with the automatic version update.

codecov[bot] commented 3 years ago

Codecov Report

Merging #31 (01a6d4e) into main (fa0be0a) will increase coverage by 0.34%. The diff coverage is 93.75%.

Impacted file tree graph

@@            Coverage Diff             @@
##             main      #31      +/-   ##
==========================================
+ Coverage   74.18%   74.53%   +0.34%     
==========================================
  Files          41       43       +2     
  Lines        2572     2611      +39     
==========================================
+ Hits         1908     1946      +38     
- Misses        664      665       +1     
Flag Coverage Δ
unittests 74.53% <93.75%> (+0.34%) :arrow_up:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files Coverage Δ
dicodile/config.py 100.00% <ø> (ø)
dicodile/data/_text.py 0.00% <0.00%> (ø)
dicodile/__init__.py 100.00% <100.00%> (ø)
dicodile/_dicodile.py 79.51% <100.00%> (+0.24%) :arrow_up:
dicodile/data/gait.py 100.00% <100.00%> (ø)
dicodile/data/images.py 76.92% <100.00%> (+12.21%) :arrow_up:
dicodile/data/tests/test_gait.py 100.00% <100.00%> (ø)
dicodile/data/tests/test_images.py 100.00% <100.00%> (ø)
dicodile/tests/test_config.py 100.00% <100.00%> (ø)
... and 2 more

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update fa0be0a...01a6d4e. Read the comment docs.

rprimet commented 3 years ago

We were on the brink of submitting it (although it needs restricting to pushes to main branch).

We were just concerned about making sure that an automated release does not happen if unit tests (or other critical steps in other workflows) fail.

But maybe having a manual trigger for release (i.e. a push-button action in the github UI) would let us sidestep this issue?

deepcharles commented 3 years ago

To avoid in-source versionning without changing the current github action, you could use setuptools_scm to automatically create a version.py file that contains the version number (in a __version__ variable).

The version number is computed from the latest tag (defaults to 0.0.0 I think), the distance from this tag, and the directory state (e.g. uncommitted changes since latest tag).

You could also add the following snippet to your github action to manually trigger the upload to Pypi, with a specific version number:

on:
  workflow_dispatch:
    inputs:
      overrideVersion:
        description: Manually force a version
env:
  SETUPTOOLS_SCM_PRETEND_VERSION: ${{ github.event.inputs.overrideVersion }}

This is particularly useful if you want to push a release candidate before an actual release.

deepcharles commented 3 years ago

You get this interface to trigger an upload.

image

rprimet commented 3 years ago

The version number is computed from the latest tag (defaults to 0.0.0 I think), the distance from this tag, and the directory state (e.g. uncommitted changes since latest tag).

So this is a bit similar to versioneer then?

deepcharles commented 3 years ago

Yes, this is very similar.

Two differences that make it a better tool in my opinion:

rprimet commented 3 years ago

@tomMoral @agramfort @deepcharles @hndgzkn since we're overhauling build files, should we attempt to move as much metadata as possible into setup.cfg?

tomMoral commented 3 years ago

@tomMoral @agramfort @deepcharles @hndgzkn since we're overhauling build files, should we attempt to move as much metadata as possible into setup.cfg?

That would be nice!

agramfort commented 3 years ago

+1 for setup.cfg

rprimet commented 3 years ago

Still needs to be discussed / solved:

For now there's no explicit dependency of the release workflow on the lint and unit tests workflows. I'm not sure what the best solution is:

Any thoughts?

tomMoral commented 3 years ago

Thx a lot @rprimet ! Merging this one, let's see how it goes and update if we find something that is not working.

rprimet commented 3 years ago

Looks like the merge did not trigger a release to testpypi, so there's probably something wrong in the testpypi if guard?... I'll have a look.

rprimet commented 3 years ago

Nevermind, the pipeline runs twice (once for event pull_request and one for push) and the push pipeline did trigger a testpypi release.