@tschad Would you mind having a look at this and giving us your thoughts? It's a very early draft, but any feedback would be appreciated. I was trying to keep it very simple for now, hopefully we can add more examples later.

Thanks @Cadair for asking for feedback on this. The general concept of what to show looks good to me. I do have a number of comments though.

First for a movie of these datasets I'd recommend having a look here: https://share.nso.edu/shared/dkist/tschad/cn_daily_movies/cn_daily_movie_20250322.mp4

For me, the line sp_subtracted = sp - (sp_mean.data * sp_mean.unit)[..., None]
results in an error: TypeError: Dataset.init() got an unexpected keyword argument 'psf'

I could get around the above issue with:
with ProgressBar():
sp_subtracted = sp
sp_subtracted.data = (sp.data - (sp_mean.data)[..., None]).compute() ## into memory

The aspect ratio of the "mean subtracted line peak" and "Mean counts" plots is off, but I'm not sure how to fix it. The FOV of this raster is roughly 160 x 230 arcsec.

Just a note that in the future the units of the data will not be 'counts'. As I recall it may not have a unit, but perhaps the DC remembers. Currently the actual data units are in relative units compared to disk center flats.

Regarding the slit positions on the context imager...this plot is a bit misleading as the context imager FOV follows the slit-location. So for any moment in time, the slit location relative to the context image is fixed. The current plot makes it seem as if the slit scans across the context image, but this is not true.

Cheers!

Thanks!

Also, I would recommend being somewhat specific as to when data is put in memory and when it is not. I very often use dask.diagnostics.ProgressBar() for when I call a compute calculation on the dataset.

If one doesn't do this, or are not aware of it, then some commands take a long time. For example, in the example, the
sp_sum_wave.plot() call is where the dask array grabs the data, leading to a long plot time.

Alternatively, putting it in memory first and then plotting makes the plotting go fast...
with ProgressBar():
sp_sum_wave = sp.rebin((-1,-1,1), function=np.sum).squeeze()
sp_sum_wave.data = sp_sum_wave.data.compute() ## puts the summed wavelength into memory

Thanks so much for your feedback @tschad. Here some responses:

For me, the line sp_subtracted = sp - (sp_mean.data * sp_mean.unit)[..., None]
results in an error: TypeError: Dataset.__init__() got an unexpected keyword argument 'psf'

This bug is fixed in this PR.

The aspect ratio of the "mean subtracted line peak" and "Mean counts" plots is off, but I'm not sure how to fix it. The FOV of this raster is roughly 160 x 230 arcsec.

Do you mean the axis labels or the aspect ratio of the pixels? If it's the tick labels then we need to fix the WCSes 🙈 (If it's the aspect ratio, we also might need to fix the WCS or I've messed up the calculation).

Also, I would recommend being somewhat specific as to when data is put in memory and when it is not.

I definitely think we should document this, but I am hesitant to put it in this example, as I'd rather not take the diversion into when and how to use dask.compute() in a tutorial on Cryo-NIRSP plotting.

Regarding the slit positions on the context imager...this plot is a bit misleading as the context imager FOV follows the slit-location

The limitation here is that we have only included the first FITS file in the sample data download so as to not bloat things too much. I've updated this whole section quite a lot, including probably a rather too complicated last plot.

I'll probably not merge it until @tschad has had another look or just before we cut a release.

Thanks @Cadair . Looks better, though here are a couple more comments:

1. The aspect ratio issue looks related to how you calculate its value. The CryoNIRSP scan is not aligned with the WCS axes, so you need to calculate it along the stepping and slit axes, as follows:

dlon = np.abs(space_2[1,0].Tx - space_2[0,0].Tx)
dlat = np.abs(space_2[0,1].Ty - space_2[0,0].Ty)
aspect = dlon / dlat

I get dlon,dlat,aspect == (<Angle 0.44393963 arcsec>, <Angle 0.1126569 arcsec>, <Quantity 3.940634>)

With that change, you get a much better plot..

2. ) Wrt the slit orientation on the CI image, unfortunately we do have a bug in the CI coordinates wherein we need to add a flip of the image axes to get something close to what is correct.

Here's an example starting with ci (i.e. the defined CI dataset from the example)

files = sorted(glob.glob(str(ci.files.basepath) + '/*fits'))
ci_map = sunpy.map.Map(files[0])
ci_map = sunpy.map.Map(ci_map.data[::-1,::],ci_map.meta).rotate()  ## FLIP DATA 

fig = plt.figure(layout="constrained")
ax = plt.subplot(projection = ci_map)
vmin, vmax = np.nanpercentile(ci_map.data, [40,99.5])
ci_map.plot(norm = PowerNorm(0.3,vmin = vmin,vmax = vmax),cmap = 'Reds_r')
slit_coords = sp[0,:,0].axis_world_coords()[0]
with SphericalScreen(slit_coords[0].observer):
    ax.plot_coord(slit_coords, color="green")
fig.tight_layout()

Thanks again @tschad, I've incorporate the aspect and data flip changes, although not the rotate with sunpy map.

On the WCS issue, is the actual problem that the CDELT for that axis has the wrong sign?

Thanks @Cadair . Looks good to me.

The CI WCS issue could be remedied with a CDELT sign flip; though, we had originally decided to flip the data in processing so that the image renders in a non-WCS-aware image viewer with the expected on-sky perspective. We already flip the SP spectral axis in this way during processing to make the wavelengths increase to the right (on the camera, it is the opposite).

There are other WCS related issues with the CI that will need to be addressed in the near future...mostly correcting FOV offsets and camera ROI related offsets.

Thanks a lot for all your feedback on this @tschad hopefully get those wcs bits fixed soon 😁