search

simonsobs / scheduler

Scheduler for SO
0 stars 1 forks source link

Geometries seem too hardcoded #71

Closed skhrg closed 1 month ago

skhrg commented 4 months ago

Right now the way the UFM center's and radii are specified (ie: https://github.com/simonsobs/scheduler/blob/026bfc77498c598b8c02e15cc314b15a29332336/src/schedlib/policies/satp1.py#L16) seems a bit too hardcoded and "code as config" too me.

I know these values don't change very often but I think it would be cleaner to either take in a geometries.yaml or get these values directly from the output of finalize_focal_plane. If we do the former I can have finalize_focal_plane output the yaml file by default. Either change would be easy to do and also would reduce the need for each telescope to have its own make_geometry function.

kmharrington commented 3 months ago

I agree a config file would work equally well going into a make_geometry argument and is probably better long-term. In a "it's not broken while the sun avoidance issues are actually broken" world, I don't think this is be a high priority issue.

mhasself commented 2 months ago

I propose we close this as wontfix. I don't think we benefit by tying this code too tightly to pipeline outputs or configs. Code as config is inevitable here, as there are so many differences between platforms (from layouts, to schedule syntax, to actual algorithms and stuff).

guanyilun commented 1 month ago

closing as of now