specviz2d: plot spectrum in wavelength/frequency with non-linear axis

[kecnry]

#1514 forced the 1d spectrum to be plotted in pixel-space to maintain limit-syncing with the 2d spectrum until support for a non-linear x-axis on the 1d spectrum viewer could be supported.

Related:

• model fitting does not work for spectral axis in pixels #1616
• Issues with updating Axis.tick_labels bqplot/bqplot#1536
• markers plugin #1953 (comment)

- [ ] bump bqplot minversion to 0.12.39
- [ ] https://github.com/glue-viz/glue/pull/2378
- [ ] https://github.com/glue-viz/glue-jupyter/pull/346

🐱

[astrofrog]

Now that a new version of bqplot has been released with fixes for the custom axis labeling (thanks @mbreddels!) I've been able to wrap up a first prototype of glue-jupyter's profile viewer which can show the profile in pixel space while having the labeling be world coordinates, which was requested for Specviz2D. To try this out, you will need:

• latest stable bqplot version
• Improvements to faciliate showing world coordinates in profile viewer in glue-jupyter glue-viz/glue#2378
• Implement support for showing world labels in profile viewer when in pixel space glue-viz/glue-jupyter#346

With these in place, you can then set x_att on the viewer state for the profile viewer to be a pixel component, and then toggle display of world vs pixel coordinates using the x_display_world boolean property on the viewer state. This should respect x_display_units. Let me know if this is looking promising, and I can try and wrap up the pull requests by adding tests etc and getting them into a released version.

[astrofrog]

The glue-jupyter PR has a demo notebook in case you want to test this out easily.

[astrofrog]

Just to check, where do we stand on this?

[kecnry]

I think we can probably consider this closed by #2219 (abandoning the actual idea to plot with non-linear axis because of the consequences of needing to translate everything after the case)