Animated functions with matplotlib

[Steve Keller]

I am trying to plot a live signal using matplotlib and it really
drives me crazy.  I have tried dozens of variants, with and without
interactive (plt.ion()), with animation.FuncAnimation and doing things
by hand, calling plt.show() and/or plt.draw() but nothing works as
expected.  One problem is that running commands in the interactive
python shell behaves differently from the behavior when using the same
commands in a script, so trying out different versions is very
confusing.

I'd like to do something like this:

    fig, ax = plt.subplots()

    ival = .1
    now  = 0
    t = list(numpy.linspace(now, now, 500))
    y = [0] * len(t)
    while True:
        time.sleep(ival)
        now += ival
        val = get_sample_from_some_signal()
        t = t[1:] + [now]
        y = y[1:] + [val]
	ax.clear()
	ax.plot(t, y)
	plt.pause(1e-9)

This works but is *very* slow and it seems the plot window is closed
and opened again or raised for each new value to plot.

I have tried to call plt.show(block=False) only once and then
repeatedly only change data with ax.lines[0].set_xdata(),
ax.lines[0].set_ydata(), and ax.set_xlim() and then calling
plt.draw().  That works in the interactive python shell but not in a
script.  When calling set[xy]_data() and set_xlim() in an amimate()
function passed to animation.FuncAnimation, but I haven't managed to
redraw the x scale values (although ax.xlim() is called).

Also, with animation.FuncAnimation() I loose control of the main
thread of execution.  When calling plt.show() the function doesn't
return.

I may want to change the loop so that instead of

    time.sleep(ival)
    val = get_sample_from_some_signal()

I would have

    val = wait_for_next_same_from_some_signal()

and I think this doesn't work with FuncAnimation since that calls the
passed function with a fixed time interval.

I am completely lost, can someone help?

Steve