tiledbsoma 1.14.1 #200
tiledbsoma 1.14.1 #200
Conversation
johnkerl
requested review from
gspowley,
jdblischak,
mlin and
Shelnutt2
as code owners
| run: | ||
| - {{ pin_compatible('numpy', lower_bound='1.16', upper_bound='1.27') }} | ||
| - {{ pin_subpackage('libtiledbsoma', exact=True) }} | ||
| - pandas | ||
| - pyarrow | ||
| # Needed until https://github.com/single-cell-data/TileDB-SOMA/issues/1926 is resolved, | ||
| # when the repo's own setup.py won't depend on pyarrow-hotfix anymore | ||
| - pyarrow-hotfix |
There was a problem hiding this comment.
I'm not sure exactly how to handle this.
- In the pyarrow tracking Issue, it correctly states that we should add the lower bound
pyarrow >= 14.0.1when removing pyarrow-hotfix ([python] Update pyarrow dependency single-cell-data/TileDB-SOMA#1926 (comment)) - However, we explicitly build against pyarrow 11 for our TileDB Cloud builds
tiledbsoma-feedstock/recipe/conda_build_config.yaml
Lines 25 to 27 in 183d888
We obviously can't both pin >= 14.0.1 in the recipe and build a pyarrow 11 variant.
This is a judgement call. It's my understanding that SOMA doesn't use the problematic code paths anyways. And pyarrow-hotfix is explicitly installed in our cloud Dockerfiles anyways, so it'll still be installed if it's not a SOMA requirement.
But it would be possible to install this new SOMA build against pyarrow 11 into a conda env without pyarrow-hotfix installed.
There was a problem hiding this comment.
Logged into TileDB Cloud to confirm that both the vulnerable pyarrow 11 and pyarrow-hotfix are currently installed in the notebook image:
jovyan@jupyter-af59ce7f-50c8-4988-a167-2bf441d61631:~$ conda list arrow
# packages in environment at /opt/conda:
#
# Name Version Build Channel
arrow-cpp 11.0.0 ha770c72_5_cpu conda-forge
libarrow 11.0.0 h2ebd325_5_cpu conda-forge
pyarrow 11.0.0 py39hf0ef2fd_5_cpu conda-forge
pyarrow-hotfix 0.6 pyhd8ed1ab_0 conda-forge
There was a problem hiding this comment.
For me I think this suffices -- I've slacked @ihnorton as well to confirm ...
There was a problem hiding this comment.
One last thing. I confirmed that the upstream setup.py didn't add a lower bound on a pyarrow (which would have prevented building our pyarrow 11 variant for the cloud)
https://github.com/single-cell-data/TileDB-SOMA/blob/1.14.1/apis/python/setup.py#L335
There was a problem hiding this comment.
Thanks @ihnorton ! :)
Following our established procedure
See also #2999 and [sc-53002].