Live Plotting without using a command line

[danielballan]

Ultimately this is the job of the replay project. @ericdill is working on a new incarnation that will be faster, more flexible, and better targeted to users' immediate needs.

But in the meantime, what's the "minimal viable product" for a live plotter that lets users set the x and y variables without using the command line?

[danielballan]

Relevant: https://github.com/mdebbar/jsonsocket

[jrmlhermitte]

Other than playback, I can only think of one other option as a quick GUI option: -play the average of consecutive sequences of images (in for high framerate/low signal time series): for example for 2: imgnew = (img[0::2,:,:] + img[1::2,:,:])/2. play back imgnew (is there terser notation or would someone have to write a function for this?)

[danielballan]

I'm talking bigger picture than that right now. Something along these lines:

1. Data collection serves up JSON documents (the same ones that go into metadatastore) to an HTTP socket.
2. On a separate process or even a separate machine, there is a graphical interface (notebook widgets? qt? matplotlib widgets?) that lets users input what to plot against what.
3. That separate process listen on the socket for JSON documents, caches them, and plots them live using the user input.
4. The same interface could be used to look at old (meaning non-live) data by going to the broker instead of a socket. But this would require adding more interface elements that let people search old scans.

[danielballan]

@ericdill is solving this properly, but I'm wondering two things:

1. What's the minimal functionality such a tool would need?
2. What's the easiest way to implement it -- notebooks, qt, matplotlib, or something else.

[ambarb]

RE item 1)

• Plot multiple scans on one graph. Scans need not be contiguous. (plot scans 3-10, 11,14)
• Based on the short list of General Users, need some minimal scan operations in order to align samples or compare data sets. --a) Normalization factor (i.e. ring current or some other Io monitor and/or user configurable equation or factor) --b) Subtract a background scan, minimum 1 background per all data scans, but multiple backgrounds would be nice. --c) numerical derivative of the scan

I can add more, but this goes beyond what is required to simply align a sample (e.g. single crystal) or compare spectroscopic data or look at an ROI counter evolution during a time scan. I omitted more advanced CCD data requirements because not all beamlines require these things.

RE item 2) I think it depends on how you implement.

[klauer]

Live plots are working reasonably well at the moment - in the same process as the collection environment. Plotting in a separate process will be implemented as part of https://github.com/NSLS-II/bluesky/pull/395