This blocks Kiyoto Yabe from his throughput analysis work.

pfsMerged files now have flux units of electrons per nm, as demonstrated in this plot from visit=68351 (a quartz with PFI):

reduceExposure.py ~/pfs/Subaru --calib ~/pfs/Subaru/CALIB-PFI-20211220 --rerun pipe2d-1024 --id visit=68351 -j 2 -c isr.doFlat=False doAdjustDetectorMap=False doMeasureLines=False repair.doCosmicRay=False --no-versions
mergeArms.py ~/pfs/Subaru --calib ~/pfs/Subaru/CALIB-PFI-20211220 --rerun pipe2d-1024 --id visit=68351 -c doSubtractSky1d=False

import matplotlib.pyplot as plt
from lsst.daf.persistence import Butler
from pfs.drp.stella.interpolate import calculateDispersion
butler = Butler("/home/pfs/pfs/Subaru/rerun/pipe2d-1024")
red = butler.get("pfsArm", visit=68351, arm="r")
blue = butler.get("pfsArm", visit=68351, arm="b")
merged = butler.get("pfsMerged", visit=68351)
fiberIndex = 309  # fiberId = 339
fig, axes = plt.subplots()
axes.plot(red.wavelength[fiberIndex], red.flux[fiberIndex], "r-")
axes.plot(blue.wavelength[fiberIndex], blue.flux[fiberIndex], "b-")
plt.xlabel("Wavelength (nm)")
plt.ylabel("Arm flux (counts)")
plt.xlim((350, 1000))
plt.twinx()
plt.plot(merged.wavelength[fiberIndex], merged.flux[fiberIndex], "k-")
plt.ylabel("Merged flux (counts/nm)")
plt.show()

Oh, the isr.doApplyGains config parameter should now be left alone. Attempts to switch it on (as we have been doing) will be met with an error.

arnaud.lefur, RHL asked me to check with you that these changes don't invalidate your audit of the extractions. I think the important changes are:
1. The gains are now automatically applied; you should not (and cannot) set isr.doApplyGains=True.
2. The pfsMerged file now contains fluxes in units of electrons/nm. The normalisation has also changed to be in the same units, so that the normalised flux is consistent with what it was previously.

My opinion is that these changes should not change any of your conclusions, but please correct me if I am wrong.

Code review complete. Only minor comment about units for pfsMerged files being in electron/nm rather than counts/nm.

Given that arnaud.lefur will not have time to perform his audit for the next week or so with the June Engineering Run looming, and that this issue is blocking for Kiyoto Yabe, we will proceed to merge the proposed changes to master. Arnaud can perform his audit when he is next available, as part of a separate ticket. Thus any problems he may find will be addressed as part of that separate ticket.

Merged.