No functionality of qbao parameter in DirectWiggleSplitPowerSpectrumTemplate

[erikzaborowski]

Hi, sometime between now and May 7 (when I was able to successfully train some emulators with qbao), I think the qbao parameter has become broken in DirectWiggleSplitPowerSpectrumTemplate. Here is some code that shows that the theory doesn't vary with qbao, but does vary e.g. with Omega_m:

import numpy as np
from desilike.theories.galaxy_clustering import LPTVelocileptorsTracerPowerSpectrumMultipoles
from desilike.theories.galaxy_clustering import DirectWiggleSplitPowerSpectrumTemplate
template_marg_test = DirectWiggleSplitPowerSpectrumTemplate(z=1.0)
theory_marg_noem_test = LPTVelocileptorsTracerPowerSpectrumMultipoles(template=template_marg_test, prior_basis='physical', tracer='LRG')
np.any(theory_marg_noem_test(qbao=1.0) - theory_marg_noem_test(qbao=1.15) != 0)
np.any(theory_marg_noem_test(Omega_m=0.3) - theory_marg_noem_test(Omega_m=0.35) != 0)

Thank you for any help!

[erikzaborowski]

For what it's worth, the latest commit on DirectWiggleSplitPowerSpectrumTemplate was 97e3b19, and if I checkout the last commit before that one, things work again.

[adematti]

Thanks Erik! --- sorry about this! the last commit should fix it again

[erikzaborowski]

Thanks Arnaud! I'm guessing maybe related to this change, but if I source the cosmodesi env and try import desilike I get a ModuleNotFoundError.

[johannesulf]

@erikzaborowski Can you let me know whether this been resolved? In any case, could you provide a short unit test so that this feature doesn't brake again?

(A good unit test is a short code snippet that gives an error, ideally an AssertionError, if the code doesn't work as expected. Code similar to what you have above with assert np.any(...) would be great. The more exact and demanding, the better. Happy to integrate the test into the codebase.)