Euclid requisites -- CAMB & CLASS

[JesusTorrado]

• Omega_b|cdm|nu_massive(z)
  • CAMB
  • CLASS
  • add test
• extrap_kmin for P(k,z) interpolator (common for CLASS and CAMB)
• angular_diameter_distance_2
  • pooling and requesting values
  • CAMB
  • CLASS
  • add test
• CLASS: sigma8(z) -- restore test
• CLASS: fsigma8(z) -- restore test
• try to incorporate add sigma_R to classy (WIP) #95
• CLASS: Pk_Weyl -- test
• CLASS: Remove Pk_max 1.5 factor
• CLASS: last neutrino case fails in GUI
• CLASS: update to 3.1.X (devel for now)
• Update CHANGELOG

Other:

• added --minimize flag to cobaya-run

[codecov-commenter]

Codecov Report

Merging #222 (7da498b) into master (36a9acc) will increase coverage by 0.31%.
The diff coverage is 91.34%.

@@            Coverage Diff             @@
##           master     #222      +/-   ##
==========================================
+ Coverage   87.54%   87.85%   +0.31%     
==========================================
  Files          92       92              
  Lines        8058     8333     +275     
==========================================
+ Hits         7054     7321     +267     
- Misses       1004     1012       +8     

Impacted Files	Coverage Δ
cobaya/input.py	86.03% <ø> (ø)
cobaya/likelihoods/base_classes/des.py	83.87% <ø> (ø)
cobaya/likelihoods/base_classes/sn.py	94.55% <ø> (ø)
cobaya/log.py	89.41% <ø> (ø)
cobaya/model.py	94.63% <ø> (+0.89%)	⬆️
cobaya/mpi.py	91.56% <ø> (ø)
cobaya/output.py	87.77% <ø> (ø)
cobaya/parameterization.py	91.13% <ø> (ø)
cobaya/prior.py	96.92% <ø> (+0.76%)	⬆️
cobaya/samplers/mcmc/proposal.py	97.19% <ø> (ø)
... and 14 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 36a9acc...7da498b. Read the comment docs.

[cmbant]

self.log

[JesusTorrado]

Sure, didn't notice that one, even though the linter complains. At some point, I'd like to change all the log attributes to logger, which is more precise. And maybe the raise LoggedError to a method?

[cmbant]

A method would be nice, but I think complicates the stack dump unless you hack it somehow

[cmbant]

Could do with some more tests of new code?

[JesusTorrado]

Almost done here (just a segfault in CLASS needing fixing).

• We are not using this global k_max anywhere at the moment, right? Could I remove it?

  cobaya/cobaya/theories/cosmo/boltzmannbase.py

  Line 133 in 36a9acc

  - ``k_max=[...]``: Fixes the maximum comoving wavenumber considered.

• When adding sigma_R tests, I've noticed that we return x, y, A[y,x], which is a bit weird. Do you think sigma_R is used little enough that we can still change that? (I mean, transpose A)

[cmbant]

Yes, looks like we are not using k_max. CAMB has "kmax" which can be set to force higher kmax. Sigma_R order doesn't bother me, but could make it y, x, A(y,x) I suppose; not sure if anyone uses it.

[JesusTorrado]

OK, I'll change it. Any preference for either z, R, sigma(z,R) or R, z, sigma(R,z)? I guess the second.

Feel free to review if/whenever you want. As soon as the new CLASS release (3.1.2) is out, I could merge.

[cmbant]

I think sigma(z,R) is the "natural" order from camb. I think I has a quick look at most things last review, anything specific new I should look at?

[cmbant]

Do a check for unused imports; minimize argument for run not in docstring.

[cmbant]

_get_z_dependent has different arguments in camb and base, should probbaly be consistent.

[JesusTorrado]

CLASS has been updated, so this can be merged. Some notes:

• _get_z_dependent indeed has different arguments. CAMB has 2 different pools of z values, and the way the code is written right now, CAMB's _get_z_dependent is in practice a wrapper over the general Boltzmann one that passes the right pool. I could give CAMB's re-implemented method a dummy pool argument that is never used if that's making your linter squeak (but it's a private method anyway).
• get_sigma_R now returns z, R, sigma(z,R).
• Unused imports removed (I think)
• minimize added to run docstring

I don't think there is anything left to check.