refactor handling of weight maps

[beckermr]

refactor handling of weight maps

[beckermr]

From @esheldon:

under region=="weight" you have

wt_us = wt

seems a bit misleading, maybe that should be None?

[beckermr]

Note the previous is in simpsimmedsio.py

[esheldon]

Also monkey patching can be problematic

[esheldon]

In simpsimmedsio

_get_band_observation(..):
...
obs.weight_us = wt_us

The above is what I mean by monkey patching

[beckermr]

Yeah totally agree. Just hacking things up so it runs for now. Def need to clean up when the whole thing gets refactored.

[esheldon]

Design-wise, not sure if it makes sense to add set_usweight() etc. though. Maybe this belongs in the meta dictionary, which is explicitly a grab-bag?

[beckermr]

Right. So in terms of design, the reason that I am mucking with the weights in the first place is because the MOF expects two weights. The first is for the initial iteration of the fitting and the second is for the subsequent ones. This happens to be uberseg, but actually a more general interface to the MOF should be provided. Something like initial_weight and iter_weight or whatever.

[esheldon]

Maybe a new datastructure is needed, designed for MOF.

This could either take an Observation/Obslist/ etc. on construction or inherit from one of those.

With a new data type you can apply constraints/requirements to make sure you have what you want without resorting to monkey patching Observations.

[beckermr]

closing this in favor of #25