I thought we had agreed on adding the (fiber#,y)->lambda map to the existing pfsFiberTrace file, which I can easily do if you give me the map. Is there any difference between this map and the file you created for SIM2D-68? Isn't the map part of SIM2D-68?

As much as I'd like to for neatness sake, putting the lambda map in ``pfsFiberTrace`` files would require two forms of the ``pfsFiberTrace`` file, because of the order in which new flats and arcs are measured, and that the file would need to be rewritten or appended to. Basically, the lambda map would be different before and after the arc exposure is processed.

If anyone can think of a clean way of dealing with this it'd be great. I assert that rewriting FITS files is evil, though. Hence some ``pfsFiberWaves`` file just to hold the new arc solution, which was measured using the traces from the ``pfsFiberTrace`` file.

The hack version I emailed you is essentially the same as the earlier version. I bumped the ``xc`` and ``pixelWave`` columns to doubles, and dropped the entirely redundant ``yc``.

I think I proposed that the command would take both the flat(s) and the arc(s) and solve both problems before performing any output. Does this solve your problem?

Is the proposal that we would then have:

• ``pfsFiberTrace`, which contains the position information extracted from the fiber flats
• no file for only the wavelength solution, because that would go directly into the:
• ``pfsFiberMap``/``pfsFiberSolution``/something which would contain both the fiber positions and wavelengths, and would thus be all that is needed to reduce a science file.

Is that what you have in mind? Is there any scenario where the arc extraction process would modify/tweak the fiber traces?

I think what RHL has in mind is to put everything into the pfsFiberTrace file which would be created by one command that extracts the FiberTraces from the Flats and the wavelength solution from the Arcs. In this case no re-writing or appending is required. I guess that would mean that we have one seed pfsFiberTrace file from the simulations which is read in as a 'guess' by the task which produces the output pfsFiberTrace file.

Since stella both generates and consumes the wavelength solution I think it would make more sense for Andreas/Stella to make the datamodel additions – I'd have to guess at some implementation details.

Following discussions with cloomis, the DetectorMap addresses this.