Issue/592/ccl v3

[hsinfan1996]

• Update _ccl_supported_versions, README and INSTALL accordingly after merging main
• Remove warning filter in the example notebooks (CCL removed problematic warnings in v2.7.1.dev10, which is the minimum supported version proposed by this PR)
• Update tests/data/numcosmo/generate_einasto_hernquist_benchmarks_fromNC.ipynb
• Update demo_mock_ensemble_realistic.ipynb
• Add option to use _projected_quad for the CCL backend (both OO and functional interfaces)
• Add tests for CCL virial massdef.
• Drop CT virial massdef.
• Show how to enable projected_quad
• Partially address Update installation instructions #491

[coveralls]

coverage: 100.0%. remained the same when pulling 2efcd32 on issue/ccl_v3 into c6cf806 on main.

[hsinfan1996]

If we are not supporting both v2 and v3 at the same time, I will delete this PR as this is not necessary. However, we will need to update README to specify the maximum or minimum version of CCL supported.

[m-aguena]

I don't think it will be feasible to maintain support for both versions. We should just keep track which releases are currently supported.

[hsinfan1996]

Don't know if anyone is using CCL v2.8, but this PR is for them and is ready. We can still wait for the official release of CCL v3.

[m-aguena]

@hsinfan1996 do you know if there is an ETA for CCL v3?

[hsinfan1996]

@hsinfan1996 do you know if there is an ETA for CCL v3?

I think there is no exact date for the release of CCL v3. What I know is that before the official release of v3, there are still some tweaks needed on documentation and example notebooks, which can be found here https://github.com/LSSTDESC/CCL/milestone/5.

[hsinfan1996]

CCL v3 will be tagged on Monday.

[m-aguena]

There are a few notebooks that are failing due CCL v3.0 API changes, can you update them?

Paper_v1.0/mcmc.ipynb
Paper_v1.0/validation_tests.ipynb
Paper_v1.0/mass_bias_from_models.ipynb
other_compare_NFW_critical_massdef.ipynb

Thanks

[hsinfan1996]

There are a few notebooks that are failing due CCL v3.0 API changes, can you update them?

Paper_v1.0/mcmc.ipynb

Paper_v1.0/validation_tests.ipynb

Paper_v1.0/mass_bias_from_models.ipynb

other_compare_NFW_critical_massdef.ipynb

Thanks

Yes, will do. I thought they should be in the state as when the paper was published for archival purposes.

[m-aguena]

There are a few notebooks that are failing due CCL v3.0 API changes, can you update them?

Paper_v1.0/mcmc.ipynb

Paper_v1.0/validation_tests.ipynb

Paper_v1.0/mass_bias_from_models.ipynb

other_compare_NFW_critical_massdef.ipynb

Thanks

Yes, will do. I thought they should be in the state as when the paper was published for archival purposes.

That version will still be able to be accessed by going to CLMM v1.0

[hsinfan1996]

Don't know how to update "examples/Paper_v1.0/mcmc.ipynb". Maybe leave it to issue #606 since it is related to NumCosmo.

[m-aguena]

Don't know how to update "examples/Paper_v1.0/mcmc.ipynb". Maybe leave it to issue #606 since it is related to NumCosmo.

You are right, this is related to NumCosmo

[hsinfan1996]

Somehow cmake is installed from conda instead of conda-forge. Forcing cmake=3.27.4 fixes the GH Actions, but makes solving environment very slow. Can wait a little longer to see if everything fixes itself again.

[hsinfan1996]

Fixed GH Actions by adding --override-channels when installing cmake. I want to avoid any additional arguments, so let's wait and see if it fixes by itself.

[hsinfan1996]

Not sure why, maybe related to the release of cmake 3.27.5, conda cannot solve the environment consistently and get all packages from conda-forge if --override-channels is not issued. On my computer, issuing conda install -c conda-forge cmake swig can get different results. Most notably, swig may be installed from conda or conda-forge every time I issue the same command.

[hsinfan1996]

Only cmake from conda-forge works regardless of the versions.

[hsinfan1996]

On a side note, numpy 1.26 is compatible.

[m-aguena]

Not sure why, maybe related to the release of cmake 3.27.5, conda cannot solve the environment consistently and get all packages from conda-forge if --override-channels is not issued. On my computer, issuing conda install -c conda-forge cmake swig can get different results. Most notably, swig may be installed from conda or conda-forge every time I issue the same command.

This will do for now. We can keep an issue to see if it will be solved by a newer version of conda