fix case of single template

[sbailey]

This PR fixes #345. Instead of recasting the 1D coeff column back to 2D, it fixes the upstream code that was incorrectly causing the 1D coeff column in the first place. At the same time, if allzfit['coeff'].ndim == 1 (maybe possible with archetypes?), then it was a bug anyway to have ntarg = allzfit['coeff'].shape instead of ntarg = allzfit['coeff'].shape[0], so I fixed that here too.

Tested that this produces the same answer as main for default templates

source /global/common/software/desi/desi_environment.sh main
cd /pscratch/sd/s/sjbailey/desi/redrock/template1
salloc -N 1 -C cpu -t 04:00:00 -q interactive

# main branch, standard templates
srun -n 64 rrdesi_mpi \
    --zscan-galaxy=1.2,1.3,1e-4 \
    -i /global/cfs/cdirs/desi/spectro/redux/daily/tiles/cumulative/83582/20260116/coadd-4-83582-thru20260116.fits \
    -o redrock-main.fits \
    --templates /global/common/software/desi/perlmutter/desiconda/20240425-2.2.0/code/redrock-templates/main/

# this branch, standard templates
module swap redrock/sjb
srun -n 64 rrdesi_mpi \
    --zscan-galaxy=1.2,1.3,1e-4 \
    -i /global/cfs/cdirs/desi/spectro/redux/daily/tiles/cumulative/83582/20260116/coadd-4-83582-thru20260116.fits \
    -o redrock-branch.fits \
    --templates /global/common/software/desi/perlmutter/desiconda/20240425-2.2.0/code/redrock-templates/main/

# compare
fitsdiff redrock-main.fits redrock-branch.fits

--> bitwise identical for data, only differing in header timestamps and vesion tracking keywords

fixes problem reported in #345

srun -n 64 rrdesi_mpi \
    --zscan-galaxy=1.2,1.3,1e-4 \
    -i /global/cfs/cdirs/desi/spectro/redux/daily/tiles/cumulative/83582/20260116/coadd-4-83582-thru20260116.fits \
    -o redrock-lae.fits \
    --templates /global/cfs/cdirs/desicollab/users/rongpu/data/redrock/lae_templates_20260120

(note that compared to original report I restricted redshift range for faster testing of array dimensionality).

[coveralls]

coverage: 42.482%. remained the same
when pulling 4e42918 on template1
into 5705e67 on main.

[Copilot]

Pull request overview

This PR fixes issue #345 where redrock encounters errors when running with only a single template spectrum. The root cause was that the coeff column was being incorrectly flattened from 2D to 1D when it had only one coefficient, breaking downstream code that expects 2D arrays.

Changes:

• Prevents automatic flattening of the coeff array when it has only one coefficient (shape[1] == 1)
• Fixes a bug where a tuple was incorrectly assigned to ntarg instead of extracting the integer from shape[0]

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[rongpu]

Looks good. Thanks for identifying the source of this bug!