What tools have you used in other projects that you wish were supported?

Pan-STARRS: See http://czwa.org/pub/czw.20190618.pdf for actual screenshots of use cases with "kapa," an image viewer like ds9 controlled by python. It was used within DVO (the panstarrs desktop virtual observatory). Matplotlib does all of this. There was no technical reason that matplotlib wasn't used for this use case.  You could also click on peaks and do photometry on them. 

• Pull up images when debugging ISR problems
• Pull up the full focal plane at the summit. Bad amps marked. Some amps do video pointings. "Yes, we saw this horrible moon glint. Problem understood." Their equivalent of "processCcd" would produce images binned at a few different scales. The image would come in. Run script. Gets binned at every level, and saved. [We could generate these during AP?
• Summit: full focal plane defect overlays
• During data release processing: Images get automatically binned for tract-level display.  Allowed staff to quickly answer, "Did the algorithm work?" The stretch was set to something strict to show background. *If* they could zoom in and change the script they would have used that functionality.
• During coaddition, they also generated six maps for every coadd: variance,  masks, exposure time, N image
• Everyone complains that files were simultaneously too big and resolution was too low. 

HSC:

• HscMap https://hscmap.mtk.nao.ac.jp/hscMap4
• NAOJ is annoyed that ds9 cannot read our SIP (:jointcal: emoji) or mask plane compression.
• If you ask for a thumbnail width > the patch overlap,  know which patches to get next to it. hscMap does this well. (When looking at something large scale, like M51). Useful for visualizing bright star masks. Pan-STARRS did not do this, and everyone was annoyed.

What's functionality is missing

- Zoom out to the full focal plane, zoom into to individual pixels, and change stretch
- Callbacks
- Display footprints and heavy footprints

Use Cases:

• Situations: Debugging, Summit monitoring, DRP monitoring (visit selection), Does the camera still match the calibrations?
• Jim says: we only care that there are APIs for all these things.
• Thumbnails across time (lightcurves), surveys  (current: matplotlib

• Quick cutouts of warp, to  look at input images that went into a coadd, 
• Where the fields fall in the sky (galaxy, solar system, RA/dec)  (current: hscMap for HSC, doesn't  exist for LSST)
• Overlay sources from other surveys   (current: hscMap for HSC, ????  for LSST)
• Overlay source catalog with brushing and linking   (current: Firefly. Though not everyone knows how to use this.)

• API bells and whistles
  • API to specify a cutout in RA/Dec coordinates and arcseconds. (current: write a few lines of lsst.geom/wcs calls + matplotlib)
  • Be able to viz intrinsic model (not PSF convolved) (current: matplotlib)
  • Thumbnails of sources.  (current: matplotlib)
  •  Subtract models from images. Be able to flip from image, model, residuals.   (current: matplotlib)

* Something like gnome on lsst-dev for pngs/jpg display (lsst-dev). Recommend something like lsst-dev to always exist through commissioning. 
* Full sky viz for the area covered per night.  Overlay on full sky viz. Accumulated N of visits on the full sky. Wished for during ops rehearsal.

Tools/Wishes

* Catalog integration.
* overplotting circles/ellipses/footprints
* mask integration (Firefly does this well)
* Can load arbitrary numpy arrays
* Obvious nans (e.g. in another color)
* Want everything to be scriptable. 

hscp Map multi-scale thing + Firefly. 

Do we need firefly APIs for the use cases above? 
Or is matplotlib/Ginga backends just fine?
Why can't we use aladinlite?