Update the version on anaconda

[davidwagner]

The version on Anaconda is outdated (https://anaconda.org/conda-forge/datascience).

I am not sure how to update the version on Anaconda to have the latest version of the library, so I don't know how to fix this. If anyone knows, I would appreciate detailed instructions.

Notes to self: https://docs.conda.io/projects/conda-build/en/latest/user-guide/tutorials/build-pkgs.html, https://www.anaconda.com/docs/tools/anaconda-org/user-guide/packages/conda-packages, https://www.anaconda.com/docs/tools/anaconda-org/user-guide/packages/standard-python-packages.

[whyvineet]

Hey @davidwagner, To update the version on Anaconda (via the conda-forge channel), it goes through the conda-forge feedstock:

• The datascience-feedstock repo controls builds for Anaconda/conda-forge.
• When a new release is published on PyPI, a bot called regro-cf-autotick-bot usually opens a PR to bump the version in meta.yaml.
• Once that PR is merged, conda-forge’s CI builds the package and publishes it to Anaconda.org automatically.

In this case, there are already open PRs to update the package:

• Bot’s PR: #3
• My manual PR: #4

As soon as one of these PRs is reviewed and merged, the latest version will be built and appear on Anaconda.org under the conda-forge channel.

I think @yuvipanda might be able to help merge either of these PRs.

[pancakereport]

Hey @whyvineet I'm working with @davidwagner and @yuvipanda to get this moving, but I haven't worked with conda forge before. It looks like some checks are failing on both your PR and the bot's. I don't really understand why this happening and what's up with the logging for it. Do you have any bright ideas?

[whyvineet]

Hey @pancakereport, I'm not completely sure yet, but from what I gathered, the CI seems to be failing because of a ModuleNotFoundError: importlib.metadata in the Python 3.7 environment. It looks like the issue comes from xyzservices, which tries to import importlib.metadata, not included in Python 3.7 by default.

The root cause might be that xyzservices >=2023 requires the backport importlib-metadata for Python <3.8, but it’s missing in the conda recipe. This causes folium imports (and consequently the datascience tests) to fail in the 3.7 environment.

Possible fixes can be:

• Adding importlib-metadata to the host and run requirements for Python <3.8, or
• Pinning xyzservices to an older version compatible with Python 3.7.

If you want, I can update my PR and try these changes to see if that resolves the CI issue.

[pancakereport]

Sure, would love to see if that fixes things! Thanks

[whyvineet]

Hey @pancakereport, my PR is finally passing all the CI Checks... Have a look!