Matplotlib video creation

Question

EDIT: ImportanceOfBeingErnest provided the answer, however I am still inviting you all to explain, why is savefig logic different from animation logic.

I want to make a video in matplotlib. I went through manuals and examples and I just don't get it. (regarding matplotlib, I always copy examples, because after five years of python and two years of mathplotlib I still understand 0.0% of matplotlib syntax)

After half a dozen hours here is what I came up to. Well, I get empty video. No idea why.

import os
import math
import matplotlib
matplotlib.use("Agg")
from matplotlib import pyplot as plt
import matplotlib.animation as animation

# Set up formatting for the movie files
Writer = animation.writers['ffmpeg']
writer = Writer(fps=15, metadata=dict(artist='Me'), bitrate=1800)

numb=100
temp=[0.0]*numb
cont=[0.0]*numb
for i in range(int(4*numb/10),int(6*numb/10)):
    temp[i]=2
    cont[i]=2

fig = plt.figure()
plts=fig.add_subplot(1,1,1)

plts.set_ylim([0,2.1])
plts.get_xaxis().set_visible(False)
plts.get_yaxis().set_visible(False)
ims = []

for i in range(1,10):
    line1, = plts.plot(range(0,numb),temp, linewidth=1, color='black')
    line2, = plts.plot(range(0,numb),cont, linewidth=1, color='red')
#    savefig is here for testing, works perfectly!
#    fig.savefig('test'+str(i)+'.png', bbox_inches='tight', dpi=300)
    ims.append([line1,line2])
    plts.lines.remove(line1)
    plts.lines.remove(line2)
    for j in range(1,10):
        tempa=0
        for k in range (1,numb-1):
            tempb=temp[k]+0.51*(temp[k-1]-2*temp[k]+temp[k+1])
            temp[k-1]=tempa
            tempa=tempb
        temp[numb-1]=0
    for j in range(1,20):
        conta=0
        for k in range (1,numb-1):
            contb=cont[k]+0.255*(cont[k-1]-2*cont[k]+cont[k+1])
            cont[k-1]=conta
            conta=contb
        cont[numb-1]=0

im_ani = animation.ArtistAnimation(fig, ims, interval=50, repeat_delay=3000,blit=True)
im_ani.save('im.mp4', writer=writer)

Can someone help me with this?

Answer 1

If you want to have a plot which is not empty, the main idea would be not to remove the lines from the plot.

That is, delete the two lines

plts.lines.remove(line1)
plts.lines.remove(line2)

If you delete these two lines the output will look something like this

[Link to orginial size animation]

Now one might ask, why do I not need to remove the artist in each iteration step, as otherwise all of the lines would populate the canvas at once?

The answer is that the ArtistAnimation takes care of this. It will only show those artists in the supplied list that correspond to the given time step. So while at the end of the for loop you end up with all the lines drawn to the canvas, once the animation starts they will all be removed and only one set of artists is shown at a time.

In such a case it is of course not a good idea to use the loop for saving the individual images as the final image would contain all of the drawn line at once,

The solution is then either to make two runs of the script, one for the animation, and one where the lines are removes in each timestep. Or, maybe better, use the animation istself to create the images.

im_ani.save('im.png', writer="imagemagick") 

will create the images as im-<nr>.png in the current folder. It will require to have imagemagick installed.

---

I'm trying here to answer the two questions from the comments:

1. I have appended line1 and line2 before deleting them. Still they disappeared in the final result. How come?

You have appended the lines to a list. After that you removed the lines from the axes. Now the lines are in the list but not part of the axes. When doing the animation, matplotlib finds the lines in the list and makes them visible. But they are not in the axes (because they have been removed) so the visibility of some Line2D object, which does not live in any axes but only somewhere in memory, is changed. But that isn't reflected in the plot because the plot doesn't know this line any more.

2. If I understand right, when you issue line1, = plts.plot... command then the line1 plot object is added to the plts graph object. However, if you change the line1 plot object by issuing line1, = plts.plot... command again, matplotlib does change line1 object but before that saves the old line1 to the plts graph object permanently. Is this what caused my problem?

No. The first step is correct, line1, = plts.plot(..) adds a Line2D object to the axes. However, in a later loop step line1, = plts.plot() creates another Line2D object and puts it to the canvas. The initial Line2D object is not changed and it doesn't know that there is now some other line next to it in the plot. Therefore, if you don't remove the lines they will all be visible in the static plot at the end.