[PYTHON] Real-time drawing with matplotlib

motivation

There are times when you want to check the values of sensors, or plot the learning status of machine learning on a graph in real time. I want to use python because there is a convenient graph creation library called matplotlib.

Known problems

Normally when making a graph with matplotlib

import matplotlib.pyplot as plt

"""
Processing to calculate the data you want to plot
...
"""
plt.plot(data)
plt.show()

I will do it. However, if you call show (), the program will be blocked there and you will not be able to draw in real time. If you look it up, use articles like draw () instead of show (), or combine it with GUI libraries such as Tkinter (tkinter) and Qt. In the Windows environment, when I touched the graph even once, the GUI got stuck and I couldn't use it though it wasn't very good. (I think it could be used in a Unix environment such as Mac.)

Let's use pause () !!!

As in the example, there was an exchange related to it on stackoverflow. Reference: real-time plotting in while loop with matplotlib

The conclusion is to use plt.pause (interval) instead of plt.show (). Below is an example of plotting a simple sin function forever.

I wrote notes in the comments, so please read them.

matplotlib_realtime_plot_example.py

# -*- coding: utf-8 -*-
"""
Example of real-time plotting with matplotlib

Continue to plot the sin function indefinitely
"""
from __future__ import unicode_literals, print_function

import numpy as np
import matplotlib.pyplot as plt


def pause_plot():
    fig, ax = plt.subplots(1, 1)
    x = np.arange(-np.pi, np.pi, 0.1)
    y = np.sin(x)
    #Must be plot once for initialization
    #At that time, it is necessary to receive the plotted object.
    #Note that list will be returned.
    lines, = ax.plot(x, y)

    #Infinitely plot from here
    while True:
        #Update plot data
        x += 0.1
        y = np.sin(x)

        #If you use the plot function when updating the drawing data
        #Note that the line object will increase each time.
        #
        #The easiest is for the lines received above
        # set_data()How to update drawing data with a method.
        lines.set_data(x, y)

        # set_data()It seems that the axis is not set automatically when using, so
        #In this example, the sin curve disappears from the drawing range in a blink of an eye.
        #Therefore, it is necessary to modify the x-axis range as appropriate.
        ax.set_xlim((x.min(), x.max()))

        #The best point
        # - plt.show()Blocked and cannot be depicted in real time
        # - plt.ion() + plt.draw()Cannot be used because the graph window freezes and the program stops
        # ----> plt.pause(interval)Use this!!!The argument is sleep time
        plt.pause(.01)

if __name__ == "__main__":
    pause_plot()

This will make a lot of progress !!!

bonus

I will introduce the method using the package drawnow in the stackoverflow mentioned above. Since drawnow is entered with pip

$ pip install drawnow

OK.

How to use

matplotlib_drawnow_realtime_plot_example.py

# -*- coding: utf-8 -*-
"""
Example of real-time plotting with matplotlib

Continue to plot the sin function indefinitely
"""
from __future__ import unicode_literals, print_function

import numpy as np
import matplotlib.pyplot as plt
from drawnow import drawnow, figure

def drawnow_plot():
    #Using the figure function included in the drawnow package
    #Create a figure object
    fig = figure()
    x = np.arange(-np.pi, np.pi, .1)
    y = np.sin(x)

    def draw():
        """
Describe the processing required for plot to be passed to the drawnow function
Variables required for plot should be variables in the scope that this function can access
        """
        plt.plot(x, y)

    while True:
        x += .1
        y = np.sin(x)
        drawnow(draw)

... I looked at the source code after writing it, but it seems that drawnow is calling draw () in the end, and it didn't work on Windows (the phenomenon that the GUI freezes). ..

Conclusion

If you want to plot in real time, use plt.pause () !!!