Juts

Jupyter widgets for scheduling processes and visualizing the resulting (live) data. While it is designed to use custom data-specific visualizations (plot widgets) based on ipywidgets, visualizations for time series data (bqplot), animations (matplotlib) and 3D data (ipyvolume) are included.

Quickstart

To start a process a so-called job has to be set up. A job consists of a Python function and a configuration. The function contains the code (machine learning, simulation, ....) to be executed, for example:

import time as tim
def wait_n_times_x_ms(config, process_queue=None, return_dict=None):
    n = config["parameter"]["n"]
    x = config["parameter"]["x"]
    
    time = list()
    time_series = list()
    for i in range(n):
        time.append(i)
        time_series.append(i / x)
        
        process_queue.put(dict(progress=int(i/n * 100),
                               time=i,
                               time_series=i/x))
        tim.sleep(x)

The configuration(s) can be read from a yaml file like this:

wait_40_times_100_ms:
    parameter:
        n: 40
        x: 0.1
        
wait_10_times_300_ms:
    parameter:
        n: 10
        x: 0.3

or can be defined as python dictionary.

Screencast min example.webm

Also live data can be monitored during script execution:

Screencast min example live.webm

A simulation example

A more practical relevant example is the following simulation of a reaction wheel pendulum under state feedback control.

Screencast pendulum.webm

Widgets can be easily connected if required:

Screencast pendulum link.webm

Since the simulation runs quite fast, a loop was put after it which sends the data with a delay. This shows that a job could also record live data from an experiment, so one could compare the simulation data with the measurements from a real pendulum.

Screencast pendulum live.webm

Visualizing 3D data

An example plot widget to view 3D data is also included. This one relies on ipyvolume.

Screencast 3d.webm