Closed esheldon closed 2 years ago
This is a mess. It might be worth it to make .exposure fully supported in Observation and make sure that it syncs the image data and mask data (with bmask not ormask)
I'd stick to only dm data structures here. Then in the dm task we use to run metacal, do the conversion to an ngmix obs there.
On Tue, Nov 9, 2021 at 9:29 AM Matthew R. Becker @.***> wrote:
I'd stick to only dm data structures here. Then in the dm task we use to run metacal, do the conversion to an ngmix obs there.
I need to process sims as well, so this needs to live outside of the task.
I'd argue that we are not very concerned about bugs in copying attributes and so more than one copy of this can live in various places.
This CoaddObs was adopted so it could be sent directly to the metacal code.
But it is keeping exps around so it can propagate the sheared images into deep copies of the exps for input to the lsst algorithms.
Make sure we have data consistency.