fix qproc non-empty fibermaps

[sybenzvi]

Two updates needed for using non-empty fibermaps in qproc:

1. Pass fibermap to read_raw to prevent preproc from creating an empty default fibermap.
2. Select fibermap rows corresponding to one camera to prevent qproc_boxcar_extraction from running past the end of the input traceset.

These changes address #1315. I tried not to touch qproc to fix the issue but the changes are necessary.

[coveralls]

Coverage decreased (-0.006%) to 25.198% when pulling 351eb94 on fibermap_fix_jan22 into 8a8bd59 on master.

[sbailey]

This looks fine to me, but for a cross check could you provide a cut-and-pasteable test case that fails (or does the incorrect thing) with current master and works with this branch?

[sybenzvi]

This looks fine to me, but for a cross check could you provide a cut-and-pasteable test case that fails (or does the incorrect thing) with current master and works with this branch?

Yes, I'll add a unit test that demonstrates that not including the fibermap just reproduces the behavior in the main branch.

[sbailey]

I don't want to turn down the offer for adding unit tests, but I was asking for something even more minimal than that: what nightwatch or qproc command should someone run and how can they tell whether the right fibermap is being used, even if done as a one-off test of the PR?

[sybenzvi]

OK, in the course of writing instructions on how to test this, I discovered a problem with the branch, so I'm converting the PR to a draft until I fix it.

Meanwhile, here are instructions for a one-off test at NERSC. This default mode of creating an empty fibermap in preproc and using that empty fibermap in qproc is tested by running:

desi_qproc -i /global/project/projectdirs/desi/spectro/data/20220111/00118039/desi-00118039.fits.fz --camera z1

To test using a fibermap, create one and use it as follows:

assemble_fibermap -n 20220111 -e 118039 -o fibermap-00118039.fits
desi_qproc -i /global/project/projectdirs/desi/spectro/data/20220111/00118039/desi-00118039.fits.fz --fibermap fibermap-00118039.fits --camera z1 --auto

Then check the FIBERMAP HDU in the preproc and qproc output FITS files using desispec.io.read_fibermap. Here is a script check_fibermap.py to do that:

from argparse import ArgumentParser
from desispec.io import read_fibermap

p = ArgumentParser(description='Read and dump fibermap to stdout')
p.add_argument('ifiles', nargs='+', help='Input fits file with a FIBERMAP HDU')

args = p.parse_args()

for ifile in args.ifiles:
    print(f'\n{ifile}')
    try:
        fibermap = read_fibermap(ifile)
        print(fibermap)
    except Exception as e:
        print(e)

Please let me know if you want this documented in a README or if this comment is good enough.

[sbailey]

This comment is good enough; no need for adding README. Thanks for specific example.

[sybenzvi]

I'm converting this PR back from draft mode. The issue was I noticed differences in the Nightwatch QA and wanted to track down the origin. I confirmed the PR does not affect any of the qframe or qcframe output. Only qsky is different, because if a good fibermap with SKY fibers isn't available, qproc sorts the fibers by flux and uses the 50% faintest ones to compute the background.