Release 0.1.0

[jrbourbeau]

Following #182 we now have things set up for us to have coiled-runtime on PyPI, but will require a new release. I'm not aware of any blockers for pushing out a new coiled-runtime release, so I'm going to propose we update package versions and do a new release this Friday.

cc @fjetter @hayesgb @scharlottej13 @ntabris for visibility

xref #231

[ncclementi]

... so I'm going to propose we update package versions and do a new release this Friday.

Do you mean update dask and distributed versions and do a new release? It looks like there was a bug introduced in dask = 2022.8.0 dask/dask#9359 (comment) I do not know if this count as a regression, I saw that Pavithra tagged it as regression and then removed it.

To which version should we bump? Could we just do a pypi release of the 0.0.4 or this is not possible?

[jrbourbeau]

It looks like there was a bug introduced in dask = 2022.8.0 dask/dask#9359 (comment) I do not know if this count as a regression, I saw that Pavithra tagged it as regression and then removed it.

It's not a regression, but is a bug. This bug also occurs in dask < 2022.8.0 but was slightly hidden by another bug in the divisions logic in dd.from_pandas. In dask=2022.8.0 the divisions logic issue was fixed (xref dask/dask#9221) which made the issue reported in dask/dask#9359 slightly more likely to occur. Hopefully that helps clarify

Could we just do a pypi release of the 0.0.4 or this is not possible?

Not possible due to not all packages in 0.0.4 being available on PyPI. We changed packages slightly in #182 to handle this.

Do you mean update dask and distributed versions and do a new release?

Yes, unless there's a known issue where we wouldn't want to update. To my knowledge bumping from 2022.6.0 --> 2022.8.0 is okay. I'll push up a PR later to confirm this against the test suite here

[ncclementi]

Ohh that makes sense, then I'm ok to release! 🚀

[ncclementi]

This conda-forge/xgboost-feedstock#89 (comment) was fixed so I wonder if we should try to include xgboost on the next release or if it's too soon.

I'm trying things out over #131 to see what happens.
cc: @jrbourbeau

[jrbourbeau]

Closing as coiled-runtime=0.1.0 is out on PyPI and conda-forge