MeasureMultiTask implementation

I would like to use MeasureMultiTask from meas_modelfit for measuring moving sources. It is mentioned in the docs but the implementation seems to be missing. Can someone tell me if it was renamed or moved somewhere?


That task never made it out of the proof-of-concept phase, and was removed before ever being used in production (and everything it did then is now sufficiently out of date that I don’t think it’d be of any use). I wasn’t aware there was still a reference to it in the docs, and it sounds like that’s a problem we need to fix.

If you’re trying to write a new PipelineTaskto do multifit-style measurements, first be aware that this is a pretty significant undertaking, but here are a few starting points:

  • First you’ll want to read up on PipelineTask, the thing you’ll want to inherit from, itself. The class reference docs are here, but there’s also a big tutorial that’s in review and close to completion. I’ll be back with a link to that once it lands.

  • I think you’ll want you task to run on each patch independently (maybe patch+band combination, if you want to process each band independently), read one of our patch-level coadd-measurement catalogs as inputs, and load cutouts from the overlapping calexp datasets as you iterate over the objects in that catalog. You’ll also want to read FGCM and jointcal output datasets to update the photometric calibration and WCS (respectively) to our best estimates. There are examples of how to do all of these things scattered across many of our other tasks (in pipe_tasks, mostly) but there are none that do all of them, and all of those potential examples have a lot of other complicating features you don’t care about.

I’m happy to help more with the details, but I think it’s probably best to stop here instead of writing a novel that wouldn’t necessarily make sense to you yet anyway. And if you can find a way to arrange it, doing some pair-coding with a DM team member would definitely be the fastest way to get this up and running quickly.

Great. Thanks for a great introduction. I will follow those breadcrumbs and hopefully not get lost along the way. I will also be waiting for that tutorial.