model attributes of `WIMPEnergySpectrum` not loaded in correctly using configuration files

[plt109]

Context: I was trying to simulate some fd.SR1WIMPSource events while loading the model parameters in via the configuration file functionality (PR 140).

df = fd.SR1WIMPSource(config=config_path).simulate(100)

Specifically, I wanted to simulate some fd.SR1WIMPSource events between 30 keV - 40 keV instead of the default 0.7 - 50 keV so I had the following line in the config file

energy_edges                            = np.geomspace(30., 40, 100)

However, I got some simulated events with energy outside 30 keV - 40 keV.

Problem: The model attributes of WIMPEnergySpectrum were not loaded in correctly from configuration files.

The configurations are loaded in in source.py which is called after the energy spectrum has been computed in energy_spectrum.py.

Solution:
It is possible to create a specific class that inherits from fd.SR1WIMPSource and this somehow manages to overwrite the model attributes like mw and energy_edges.

class my_wimp(fd.SR1WIMPSource):
    mw = 50
    energy_edges = np.geomspace(30., 40., 100)

However, being able to load in all model attributes of the sources via a single configuration file is still a better solution in terms of keeping track of the configurations used for fits.

[JelleAalbers]

Good catch! The problem is that the WIMPsource computes and sets new attributes (energies and energy_hist) on initialization/setup, and does not recompute them when the original attributes (energy_edges, mw, etc.) are changed.

I think the fix would be to have BlockModelSource override set_defaults to call each Block.setup() again after Source.set_defaults. The one complication would be that the length of array columns like energy_edges might change as well, so we'd either have to disallow that (so you probably still have to do subclassing) or make the handling of the array_columns field more complex.

I was working on a fix when I realized the NEST sources currently use copy-paste-modified energy spectra -- so any changes would have to be applied there as well, and there is a big PR for the NEST sources open already. If it is not a big roadblock, I would wait until the big NEST PR is merged and we've had a change to reduce some of the duplication.

[plt109]

Nice that you have started working on it!

Before the NEST sources get merged in, I think it is okay to carry on with the patch of inheriting from fd.SR1WIMPSource and overwriting the model attributes from the inherited class.

class my_wimp(fd.SR1WIMPSource):
    mw = 50
    energy_edges = np.geomspace(30., 40., 100)

Yes, it might be better to keep this issue on hold and resolve it after the NEST sources get merged in so we can harmonize any differences (if any) in the way the current sources and the NEST sources read in the configuration files as well.