Making CLMM conda instalable

[eduardojsbarroso]

Changed the pyproject.toml file so we can publish CLMM to conda and pipy.

[eduardojsbarroso]

Hi @m-aguena and @marina-ricci . As we discussed in the last pipeline meeting, indeed pip install git+https://github.com/LSSTDESC/[email protected] works for CLMM. However I created a new pyproject.toml file and I was able to publish clmm to the PypiTest. I can also publish to Pypi and Conda now. However I would like your aproval. Also, should I publish or one of you guys? Since it will be associated with my account?

[marina-ricci]

@eduardojsbarroso Michel and I are not familiar with what you proposed. Could you ellaborate?

[eduardojsbarroso]

@eduardojsbarroso Michel and I are not familiar with what you proposed. Could you ellaborate?

What I am proposing is that we publish the clmm git repository to Pipy and conda, so one can simply do pip install clmm or conda install clmm. To so, we need a pyproject.toml file with all the requirements, dependencies and licenses of the repository. After making this file, we can use setuptools I guess or another program to build the files that will be posted in Pypi and conda (I used a program called hatch). Then, after having all the files, we just need to publish them to Pipy and conda. However, you need to have an account at Pipy to publish. I created one, and published clmm in PipyTest, which is just to see if everything is working. Now I can publish to Pipy and conda, but then the package will be associated to my account, so I will be the one to update it when something changes. I can do that no problem, but if you guys prefer to do it since the repository is not mine, feel free to do it. Otherwise I can do it this week. I just noted that there was some problem with the new pyproject.toml file in the test

[coveralls]

coverage: 100.0%. remained the same
when pulling 3d6b413 on conda_instal
into ad087d9 on main.

[m-aguena]

@eduardojsbarroso I talked to the CCL people, and they said an account for the code was created. So, I will make the same for CLMM and give you access.

[m-aguena]

@eduardojsbarroso can you add instructions on how to create these installation to the Creating installation via pip and conda section of CONTRIBUTING.md? Thanks

[hsinfan1996]

Since this PR touched upon the dependencies, can we check if astropy v7.0.0 is compatible, which was just released.

[hsinfan1996]

Since this PR touched upon the dependencies, can we check if astropy v7.0.0 is compatible, which was just released.

Never mind. It is compatible.

[eduardojsbarroso]

Since this PR touched upon the dependencies, can we check if astropy v7.0.0 is compatible, which was just released.

Never mind. It is compatible.

[caioolivv]

For some reason, when building the documentation, sphinx sees CCL's version as 3.0.0, instead of the one actually installed in the environment. This ends up breaking the build, since there's a CCL version check in clmm/theory/ccl.py.

[caioolivv]

Hey @m-aguena @hsinfan1996, is the hard check on CCL's version absolutely necessary? At the moment there's no problem running current CCL with current CLMM and the hard check breaks documentation building... We also don't make a hard check on any other package, so I think we could drop this.

[caioolivv]

Documentation breaks because CCL's mock version in docs/conf.py is set to 3:

sys.modules["pyccl"].__version__ = "3"

We're forced to do this because of the hard version check on CCL. Now we can either update this line every time we change version checks, or we remove the hard version check. I think the hard version check is unnecessary.