Add example of using mpl-interactions as an engine for updating without sliders

[ianhi]

Problem

This package has put a lot of work into figuring out how to update various matplotlib artists. However sometimes people want to update a plot in a loop instead of wiht sliders, for example see matplotlib/ipympl#425. They should still be able to do this using mpl-interactions, but it's not obvious how to do this.

attn: @henrypinkard for inspiring this

Suggested Improvement

Include an example of something like this:

%matplotlib ipympl
import matplotlib.pyplot as plt
import mpl_interactions.ipyplot as iplt
import numpy as np
from matplotlib.widgets import AxesWidget
from matplotlib.cbook import CallbackRegistry
import time

def display_immediately(fig):
    """
    Force immediate display of a plot in ipympl. Not necessary in desktop backends
    """
    canvas = fig.canvas
    display(canvas)
    canvas._handle_message(canvas, {'type': 'send_image_mode'}, [])
    canvas._handle_message(canvas, {'type':'refresh'}, [])
    canvas._handle_message(canvas,{'type': 'initialized'},[])
    canvas._handle_message(canvas,{'type': 'draw'},[])

# currently needs to inherit form axeswidget rather than
# Widget due to limitations in mpl-interactions
class progress(AxesWidget):
    """A small widget to pass as a kwarg to mpl-interactions"""
    def __init__(self):
        self._val = 0
        self._observers = CallbackRegistry()
    @property
    def val(self):
        return self._val
    @val.setter
    def val(self, value):
        # could add validation here
        self._val = value
        self._observers.process("changed", value)
    def on_changed(self, func):
        return self._observers.connect("changed", lambda val: func(val))

def f(t):
    if t==0:
        # otherwise line doesn't get made and we will
        # get weird errors later on
        return [0,0]
    # make some fake data
    steps = np.arange(t)
    return [steps, np.sin(steps)]

# make our widget for updating the plot
prog = progress()

with plt.ioff():
    fig, ax = plt.subplots()
display_immediately(fig) # only necessary with ipympl backend

iplt.plot(f, t=prog, parametric=True)

   


for i in range(1, 10):
    # do some work that will update our data sources
    # make the updates plot
    prog.val = i
    # mpl-interactions should be handling this draw for us
    # but it only ever does draw_idle
    # should add an option to draw immediately.
    fig.canvas.draw() 
    time.sleep(.5)

[ianhi]

Should probably come up with a better name for that widget and then incorporate it into the library. Would also be sweet if this worked with ipywidgets progressbars

[ianhi]

oop turns out it just does work immediately - pretty sweet:

prog = IntProgress()
with plt.ioff():
    fig, ax = plt.subplots()
display_immediately(fig) # only necessary with ipympl backend

iplt.plot(f, t=prog, parametric=True)

for i in range(1, 10):
    # do some work that will update our data sources
    # make the updates plot
    prog.value = i
    # mpl-interactions should be handling this draw for us
    # but it only ever does draw_idle
    # should add an option to draw immediately.
    fig.canvas.draw() 
    time.sleep(.5)