Fixing requisites for bao generic with different kinds of observables

[Kushallodha]

Fixing computation of requisites for generic BAO when supplied with measurements_file that contains different kinds of observables at several redshifts. To validate the logic, I tested it on the dummy file in the attachment. The older version incorrectly returns {'angular_diameter_distance': {'z': array([2.34])}, 'Hubble': {'z': array([2.34])}, 'rdrag': None, 'fsigma8': {'z': array([2.34])}} .

z	value	observable
0.29	1	DV_over_rs
1.31	1	DV_over_rs
1.31	1.5	F_AP
0.5	1.8	DV_over_rs
0.5	1.2	F_AP
0.7	1.1	DV_over_rs
0.7	1.6	F_AP
0.91	1.5	DV_over_rs
0.91	1	F_AP
1.49	1.5	DV_over_rs
2.34	2	DM_over_rs
2.34	5	DH_over_rs
2.34	2	f_sigma8

bao_mean_check.txt

[cmbant]

Thanks, the tests are failing, but probably just a trivial whitespace issue.

[cmbant]

Shouldn't it be something like this?

        if self.has_type:
            for observable in self.data["observable"].unique():
                for req, req_values in theory_reqs[observable].items():
                    if req not in requisites:
                        requisites[req] = req_values
                    else:
                        for k, v in req_values.items():
                            if v is not None:
                                requisites[req][k] = np.unique(
                                    np.concatenate((requisites[req][k], v)))

[codecov-commenter]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (ea587ce) 81.78% compared to head (1e32d78) 81.75%.

Files	Patch %	Lines
cobaya/likelihoods/base_classes/bao.py	62.50%	3 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #344      +/-   ##
==========================================
- Coverage   81.78%   81.75%   -0.03%     
==========================================
  Files         133      133              
  Lines       11033    11039       +6     
==========================================
+ Hits         9023     9025       +2     
- Misses       2010     2014       +4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Kushallodha]

Yes, thank you @cmbant for pointing this out. It needs to be for observable in self.data["observable"].unique() but we still need to check (if isinstance(req_values, dict)) req_values is dict before for loop. Let me know if I missed a better/simplified way to do the same.

        if self.has_type:
            for observable in self.data["observable"].unique():
                for req, req_values in theory_reqs[observable].items():
                    if req not in requisites.keys():
                        requisites[req] = req_values
                    else:
                        if isinstance(req_values, dict):
                            for k, v in req_values.items():
                                if v is not None:
                                    requisites[req][k] = np.unique(np.concatenate((requisites[req][k], v)))

For the time being, all the checks have passed. Please let me know if you prefer all commit history squashed into one. If not, then I will submit the pull request.

[cmbant]

Thanks, why do you need to check it is dict - aren't they all dict? Don't worry about squash, we do that when merging.

[Kushallodha]

I think rdrag's value ( {"rdrag": None}) is not the dict.

[cmbant]

None is caught by the "if v is not None", the whole of {"rdrag":None} is a dict insance?

[cmbant]

(if you have a simple test, it would be good to add it to the unit tests along with the other bao tests)

[Kushallodha]

In case of {"rdrag":None}, req_values is None and for loop crashes before if condition as None can't be iterated over with .items().

I've created a basic test using mock data for mixed observables, similar to 'test_generic_camb`, which requires a mock mean and data. Please let me know if this is okay or if you have something different in mind.

[cmbant]

Thanks, probably just need to update BAO's github_release": to "v2.3" for new data tests to pass.

[Kushallodha]

I am not sure why tests are failing. I did not touch any camb-related files.

[cmbant]

Yeah, weird - I'm sure it's nothing to do with this PR.

[cmbant]

I think it was just a temporary github issue, seems to pass now. Thanks for the PR!