conda package for dask-ms

[sebastian-luna-valero]

Hi,

I would like to create a conda-forge package for dask-ms.

I have an initial conda environment that seems to work:

name: dask-ms

channels:
 - conda-forge
 - defaults

dependencies:
 - python
 - bokeh
 - dask
 - distributed
 - mock
 - numpy
 - python-casacore
 - pytest
 - setuptools
 - xarray

Would you be happy to assist if I find issues?

Best regards,
Sebastian

[sjperkins]

I would like to create a conda-forge package for dask-ms.

Would you be happy to assist if I find issues?

Hi @sebastian-luna-valero. Thanks for reaching out. In principle yes, although my experience with conda and conda packaging is limited -- I've worked with pypi exclusively as a release mechanism up till this point. I'm keen to try.

I have an initial conda environment that seems to work:

name: dask-ms

channels:
 - conda-forge
 - defaults

dependencies:
 - python
 - bokeh
 - dask
 - distributed
 - mock
 - numpy
 - python-casacore
 - pytest
 - setuptools
 - xarray

I feel that these are the core dependencies:

• python
• dask
• numpy
• python-casacore
• setuptools

Adding the xarray optional extra

• xarray

Adding the testing optional extra expands it

• pytest
• mock

distributed is not yet a dependency, although I might try to detect for it in the code itself.

[sjperkins]

Also note that I'm aiming to release 0.2.4 in the next week or so.

[sebastian-luna-valero]

Great, thanks @sjperkins

Could you please let me know once version 0.2.4 has been released so I start working on the conda package?

Best regards,
Sebastian

[sebastian-luna-valero]

By the way, would you (or anyone else) like to be listed as a recipe maintainer?

[sebastian-luna-valero]

Initial recipe here:

conda-forge/staged-recipes#11321

We can easily update the recipe to include more maintainers.

Could you confirm whether I should include the following license title:

Copyright (c) 2019, National Research Foundation
(South African Radio Astronomy Observatory)
All rights reserved.

[sjperkins]

By the way, would you (or anyone else) like to be listed as a recipe maintainer?

Could you outline what this entails? I'd be happy to look at a URL.

[sjperkins]

Could you confirm whether I should include the following license title:

Copyright (c) 2019, National Research Foundation
(South African Radio Astronomy Observatory)
All rights reserved.

I've updated the year in #105 so this should now change to:

Copyright (c) 2019 - 2020, National Research Foundation
(South African Radio Astronomy Observatory)
All rights reserved.

[sebastian-luna-valero]

Could you outline what this entails? I'd be happy to look at a URL.

It will give you push access to the conda-forge feedstock for the dask-ms package so you have permissions to update it, and approve/reject PRs. See examples of conda-forge feedstocks here:

https://conda-forge.org/feedstocks/

As it is configured now, every new release of dask-ms on PyPI will trigger a build in conda-forge to update the package. So far I have just included a python -c "import daskms" test on the conda recipe, but we can add more tests so they will be executed automatically before we release a new version of the conda package.

More general info is available here: https://conda-forge.org/docs

[sjperkins]

Could you outline what this entails? I'd be happy to look at a URL.

It will give you push access to the conda-forge feedstock for the dask-ms package so you have permissions to update it, and approve/reject PRs. See examples of conda-forge feedstocks here:

https://conda-forge.org/feedstocks/

OK thanks, I'd be happy to be a maintainer here. @o-smirnov it might make sense to make you or someone else a maintainer too in order to reduce the bus factor.

[sjperkins]

So far I have just included a python -c "import daskms" test on the conda recipe, but we can add more tests so they will be executed automatically before we release a new version of the conda package.

dask-ms has a full test suite that is designed to be run with pytest. It's currently run on travis as follows:

https://github.com/ska-sa/dask-ms/blob/master/.travis/python36.docker
https://github.com/ska-sa/dask-ms/blob/master/.travis/python37.docker

Is this something that you would be prepared to adapt for whatever CI environment runs on conda so that I could learn how this is done, or would you prefer me to set this up?

[sebastian-luna-valero]

That's great, many thanks. Given that you already test it before PyPI releases it probably doesn't make sense to repeat the full suite in conda again. Could you please provide a few more python imports to add to this recipe just as a sanity check?

I will wait for @o-smirnov to confirm, thanks!

[sjperkins]

I think these are the most important imports:

python -c "from daskms import xds_from_ms"
python -c "from daskms import xds_from_table"
python -c "from daskms import xds_to_table"
python -c "from daskms import TableProxy"

[sjperkins]

Hi @sebastian-luna-valero

dask-ms 0.2.4 has been released on pypi: https://pypi.org/project/dask-ms/0.2.4/

and is tagged in this repo as 0.2.4

[sebastian-luna-valero]

Thanks for letting me know @sjperkins

Conda package has been created and uploaded to the conda-forge channel.