Push a new release to PyPI

[perimosocordiae]

There are a lot of good changes since v0.3, so I think we're almost ready to release v0.4.

Once most (hopefully all) of the remaining items on the milestone are finished, we should tag a commit on Github and push the new builds to PyPI.

@terrytangyuan: I'll update this issue when we're good to go.

[terrytangyuan]

Sounds good.

[mtngld]

Any update on this?

Thanks!

[perimosocordiae]

There's been a lot of movement recently with some new core contributors, and the APIs are going to be changing pretty significantly, so I'm not sure if now is a great time to cut a new release. I agree that we're overdue for one, though.

@wdevazelhes What do you think?

[wdevazelhes]

I'm thinking maybe the new APIs changes could be put in the next release (v.0.5.0), since a working version could be merged soon in master (there is not much left to do), but to have a more robust one it may still take some more time. What do you think ?

Also, do you want to have all milestones issues solved before releasing ? I am currently working on the new API, but I could look also at some issues in order to better fulfil the milestone. Maybe we could solve the easiest issues from the milestone of v0.4.0, do the release, and postpone the others to v0.5.0 ?

[perimosocordiae]

Yeah, I'm happy to push back any milestones that aren't quick to resolve. Feel free to move things around as you see fit.

[perimosocordiae]

And yes, releasing the new API as v0.5 sounds good.

[terrytangyuan]

There are quite some good changes. I agree that we should aim for v0.4.0 first and then target v0.5.0 later with new APIs.

[wdevazelhes]

Since all issues in the milestone are solved, I guess we can do the release ?

If I understand correctly, the way to go is the following:

• on Github:

  • In the "release" pages, click on "Draft a new release"
  • Write a small text summarizing the more important changes (see below)
  • Zip and tar gz my local master and add it as attachment in the appropriate area

• On the command line, run the following commands, as suggested by @nvauquie:

$ python3 setup.py sdist
$ python3 setup.py bdist_wheel
$ twine upload dist/*

I already ran these commands on Test PyPi (btw @perimosocordiae @bellet if you want to be included in the members of metric-learn on Test PyPi you should create an account on Test PyPi and give your credentials to @terrytangyuan)
The upload was successful but I just realized there's no project description (see https://test.pypi.org/project/metric_learn/)
I was thinking maybe it's because twine has a flag for config files that I didn't use ? (twine upload dist/* --config-file=setup.cfg)
I tried to redo the upload but it says the file "already exists" (apparently you cannot do an upload twice, whether on the real or the test PyPi)

@terrytangyuan, as you can manage the project, is there maybe an option to delete the upload ?

---

Here is a proposal for the text to write in github, tell me if it sounds good to you:

- Two newly introduced algorithms: 
    - MLKR (Metric Learning for Kernel Regression)
    - MMC (Mahalanobis Metric for Clustering)
- Improved documentation and examples

[bellet]

I would also add something like "Performance improvements" and "Minor bug fixes"

[wdevazelhes]

I would also add something like "Performance improvements" and "Minor bug fixes"

I agree, will do

[terrytangyuan]

@wdevazelhes I just granted you "owner" role (on Test PyPI only) so you can delete the release yourself now.

Also, when you make a release on Github, you don't have to "Zip and tar gz my local master and add it as attachment in the appropriate area" as Github will do that for you automatically once you create the release.

[wdevazelhes]

@terrytangyuan thanks !

I deleted the test v.0.4.0 repo but it was is indeed not possible to reupload it, so I uploaded tests versions by adding a figure at the end (ex: v.0.4.0.1). I couldn't manage to make the description file printed until I added these lines to setup.py to specify the path to the config file (and then it worked):

# Get the long description from README.md
with open('README.rst', encoding='utf-8') as f:
  long_description = f.read()

# below description='Python implementations of metric learning algorithms' in setup(...)
long_description=long_description,

@terrytangyuan did you use a specific command to take into account the setup.cfg file ?