Request to add the kernel used for resampling in the meta of the mosaiced image

[darioflute]

When reading the meta of a mosaiced image generated with MosaicPipeline, the meta has a resample field which contains the main parameters used for the resampling. Unfortunately, the information about the kernel used for resampling is not conserved.

A different kernel can be used for resampling, for instance, one can use the "Lanczos3" kernel as in this example: from romancal.pipeline import MosaicPipeline steps = {'resample': {'kernel': 'lanczos3','weight_type':'ivm'},'outlier_detection':{'skip':True}} result = MosaicPipeline.call('r4444401001001001001_wfi11_asn.json', save_results=True,steps=steps)

However, when reading the meta about the resampling:

import roman_datamodels as rdm path = '/grp/roman/SCIENCE_PLATFORM_DATA/roman-17.0.0/Mosaic_Lanczos3/' with rdm.open(path+'r0003201001001001004_wfi11_coadd.asdf') as f: meta = f.meta.copy() print(meta.resample)

the information about the kernel used for resampling is lost:

{'good_bits': '~DO_NOT_USE+NON_SCIENCE', 'members': ['r0003201001001001004_0001_wfi11_f106_cal.asdf', 'r0003201001001001004_0002_wfi11_f106_cal.asdf', 'r0003201001001001004_0003_wfi11_f106_cal.asdf', 'r0003201001001001004_0004_wfi11_f106_cal.asdf'], 'pixel_scale_ratio': 1.0, 'pixfrac': 1.0, 'pointings': 4, 'product_exposure_time': 182.55169677734375, 'weight_type': 'ivm'}

This is confusing when looking back to products since it is no more clear what has been used to generate them.

[schlafly]

Thanks Dario. @braingram , this reminds me that this is a special case of our discussion to add the 'spec' parameters generically to all of the outputs, e.g., roman.meta.kwargs.{step} = kwargs or something along those lines?

Dario, we don't intend to use more than one kernel in operations, so I'm reluctant to add this as a generic archive-searchable field, but we could capture this and any other keyword parameters via this kind of approach. Were we to do that, would that address your concerns? In this case these parameters would not make it to the archive but would be in the individual files for people using the pipeline.

[braingram]

@darioflute is the information in the cal_logs for the produced files? I'm not advocating for this as "the solution" but perhaps it is helpful in this case.

I do think recording step parameters would be useful. I can't immediately find a ticket for that work though.

[schlafly]

I'll make a ticket (not for you, but at least so we record the discussion)---see https://github.com/spacetelescope/romancal/issues/2019.