Should the generic `instrument_view` be removed?

[nvaytet]

Should the instrument_view be removed altogether? I see it's use could be quite limited if we are adding instrument-specific versions in the other packages (e.g. dream.instrument_view).

There is almost nothing in the instrument view compared to using Plopp's scatter3d directly, apart from the additional beamline components that can be shown, but experience has shown that this is basically unused.

There is an additional maintenance burden of keeping both scatter3d and instrument_view. See #484 for example.

See also #452

[nvaytet]

For the record, a discussion on Slack about the usefulness of being able to display additional beamline components:

Neil Vaytet
Question (maybe mostly for the IDSs?): does anyone use the feature of the instrument view where you can display components other than detector pixels by passing a dict of components?
See an example here.
I have a feeling that it is not used much (if at all?), and maybe not so useful? It's not finding components from the Nexus files automatically, and even if it did I don't know if the visualization would be so useful?
You can use nexus-constructor to view the geometry and the components.
I feel like when one uses the instrument view, it's mostly to look at the counts on the detector pixels, not to look at the position of the sample or choppers.
Am I correct? If so, would you care if we removed that feature?
Thanks for any opinion on this 🙏

Gregory Tucker

1. It’s probably too early to discard use-cases since we’re not yet taking real data
2. It may not be useful to know where the source is relative to the detector, but I can imagine adding other markers to help orient the detector view in real space. One issue I’ve encountered during experiments using the Mantid instrument viewer is not knowing which way is up, for example.
3. I’m not sure that the recommendation to use nexus-constructor is valid long-term; if I understand correctly, there isn’t a desire for its view to be 100% accurate in all cases

Neil Vaytet
Thanks for feedback. About 2., aren't the RGB xyz axes indicators enough?

Gregory Tucker
Maybe they become more-useful with experience? Looking at it as-if a new user, I see three lines that probably represent Cartesian axes, but don’t know which line corresponds to which axis, or whether their intersection is the origin.

Neil Vaytet
So what could we add to make it more obvious? A semi-transparent plane (or a grid) below the instrument that represents the floor? (edited)

Gregory Tucker
🤷 I guess different ‘hints’ would help different people. I might like more-obvious axes — maybe labelling ‘0’, ‘x’, ‘y’, and ‘z’ — but then you need to know that z is vertical; so there might not be a good general answer

[nvaytet]

Finding this old issue: I think we should just replace the generic instrument view with the one from ESSdiffraction.
It's generic enough, and handles multiple banks in a DataGroup as input.

It would remove the plotting of additional components, which still seems to be unused.
We can always bring them back if there is a wish to have them.