list reference to mpl.figure.Figure does not show plot

Question

I am trying to handle a list of figures with an object. Unfortunately there seems to be a problem with plotting from a list of figures.

Please comment out the line in the example below and you see how the plotting breaks:

import matplotlib as mpl
from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg, FigureManagerQT


class Test:
    def __init__(self):
        self.figs = [mpl.figure.Figure(),mpl.figure.Figure()]
        self.fig = mpl.figure.Figure()
        ax = self.fig.subplots()
        ax.plot([1,2],[3,4])

    def show(self):
        fig = self.fig  # works
#        fig = self.figs[0]  # does not work
        canvas = FigureCanvasQTAgg(fig)
        figManager = FigureManagerQT(canvas, 0)

a=Test()
a.show()

Result (this is what I want):

Result with line uncommented:

In some other tests I found it might be connected with destructing the object. As a list is a mutable object, this might be the connection. I also tried (unsuccessfully) several workarounds to copy the figure object for plotting: I used something like fig = myCopy(self.figs[0]) in combination with a pickle-copy.

Can you please give me some explanation of what is happening and what might be a workaround?

Answer 1

In __init__, you give axes to self.fig and plot to this Axes object:

class Test:
    def __init__(self):
        self.figs = [mpl.figure.Figure(),mpl.figure.Figure()]
        self.fig = mpl.figure.Figure()
        ax = self.fig.subplots()
        ax.plot([1,2],[3,4])

The figure objects in self.figs have no Axes object attached to them, so they're basically empty. As a result, what you see is an empty figure:

def show(self):
    fig = self.figs[0] # This is a figure with no axes
    canvas = FigureCanvasQTAgg(fig)
    figManager = FigureManagerQT(canvas, 0)

The problem with your logic is that it's not really meaningful to plot data in the __init__ method. Your workflow should be:

1. Initialization
2. Figure selection
3. Plot
4. Show

I suggest that you add two methods, select_figure and plot, so as to improve the overall usability of your figure manager:

class Test:
    def __init__(self):
        self.fig = None
        self.figures = [mpl.figure.Figure(), mpl.figure.Figure()]

    def select_figure(self, index):
        self.fig = self.figures[index]

    def plot(self, x, y):
        ax = self.fig.subplots()
        ax.plot(x, y)

    def show(self):
        canvas = FigureCanvasQTAgg(self.fig)
        figManager = FigureManagerQT(canvas, 0)

Then you can implement the workflow I described above:

test = Test()

test.select_figure(0)
test.plot([1, 2], [3, 4])
test.show()

test.select_figure(1)
test.plot([3, 4], [5, 6])
test.show()