Notes on live raw detector views

[SimonHeybrock]

• Instrument view seems pretty useless (to many pixels, layers, ...). I suggest we ignore this option for now and focus on regular 2-D visualizations.
• Instead, go for projection of real pixel positions, e.g., for Loki onto a plane in X-Y.
• There will be Moiré-like effects. This can mostly be avoided by randomizing pixel positions within their voxel (cylinders at Loki). There are two options:
  1. Randomize the new data's positions, histogram, then accumulate.
  2. Accumulate regular data, replicate pixels, then randomize, histogram.
• Running Beamlime from a notebook seems to have some problems. Can we avoid that? How can we get interactive plots in the case?

[SimonHeybrock]

Notes from discussion with @nvaytet on plotting:

• Try Matplotlib Qt window (plt.show() from within Python script).
• Setup plots once, update only data.
  • Could use plain Matplotlib or Plopp nodes.
  • After fixing this, saving to file is taking most of the time.
• If we need simple widgets, once could consider doing this in a tiny Qt application, instead of ipywidgets, which would require Jupyter.
• Measure how long (lower-level) Plopp needs to update data in multiple subfigures, and how this scales with the resolution.