Python, Matplotlib: Suptitle with arbitrary number of vertical figures

Question

I have a website that produces (depending on available data stations running) an arbitrary number of plots (as an image), that are vertically stacked over one another. An example is the following:

The problem is that depending on the number of vertical plots, the suptitle (top title) goes to a different position. Check the following examples of 5 and 10 plots:

5 plots:

And here's 10 plots:

So for every number of plots, I get a different result. Using fig.tight_layout() didn't help.

What I need is to have the bottom of my text at a certain distance from the top of the plots. Is there a general answer to this problem?

I created some minimal working code that has the number of plots parametrized. Please check it out if you would like to reproduce this problem.

import datetime
import random
import matplotlib
matplotlib.use('Agg') # Force matplotlib not to use any Xwindows backend.

import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import matplotlib.image as mpimg
import matplotlib.gridspec as gridspec
import numpy as np


random.seed(datetime.datetime.now())

#initial parameters
numOfPlots  = 2
dataLen     = 100
randomRange = 10*dataLen
dpiVal      = 180

#create data
xData = list(range(dataLen) for x in range(numOfPlots))
yData = list(random.sample(range(randomRange), dataLen) for x in range(numOfPlots))

#matplotlib initialize plot
gs = gridspec.GridSpec(numOfPlots,1)
plt.cla()
plt.clf()
fig = plt.figure()
ax = None

for i in list(range(numOfPlots)):
    if i == 0:
        ax = fig.add_subplot(gs[i])
    else:
        ax = fig.add_subplot(gs[i],sharex=ax)

    ax.plot(xData[i], yData[i])

    labelSize = 10
    ax.set_ylabel("Hi there",size=8)
    ax.get_yaxis().set_label_coords(-0.07,0.5)
    plt.yticks(size=8)
    plt.ticklabel_format(style='sci', axis='y', scilimits=(0,0),useOffset=True)
    plt.subplots_adjust(hspace = 0.3)
    if i == numOfPlots-1:
        plt.xticks(rotation=0,size=7)
        max_xticks = 10
        xloc = plt.MaxNLocator(max_xticks)
        ax.xaxis.set_major_locator(xloc)
        ax=plt.gca()
    else:
        plt.tick_params(
        axis='x',          # changes apply to the x-axis
        labelbottom='off') # labels along the bottom edge are off
    ax_right = ax.twinx()
    ax_right.yaxis.set_ticks_position('right')
    ax_right.set_ylabel("Nice to see you!",size=labelSize)
    ax_right.get_yaxis().set_ticks([])


#the following sets the size and the aspect ratio of the plot
fig.set_size_inches(10, 1.8*numOfPlots)

fig.suptitle("Hi there, this is the first line\nAnd this is the second!!!")
fig.savefig("img_"+str(numOfPlots)+".png",bbox_inches='tight',dpi=dpiVal)

These convenience functions are often useless if you need consistent output. Maybe you can try [something manual](http://stackoverflow.com/a/12958839/5067311). — Andras Deak -- Слава Україні, Jun 29 '16 at 18:29
@AndrasDeak I'm sorry I don't understand what's manual there. Could you please explain a little bit more? — The Quantum Physicist, Jun 29 '16 at 18:30
Check out the link, it demonstrates adding text annotation with axes-specific locations. I tried to suggest that you try to do that with your topmost axes, in order to get something consistent. Sorry for not being clear enough:) — Andras Deak -- Слава Україні, Jun 29 '16 at 18:33
@AndrasDeak I did look at the link. But my problem isn't with axis text. My problem is with suptitle vs number of plots. — The Quantum Physicist, Jun 29 '16 at 18:34
Yes, but can't you write a suptitle manually using text? Positioning it relative to your top axes, ensuring that it's places consistently? — Andras Deak -- Слава Україні, Jun 29 '16 at 18:36
@AndrasDeak I don't know how to do that, especially that the position is a function of the number of plots. I tried to create a linear interpolation/extrapolation for it and that didn't work either. It's not even linear!! — The Quantum Physicist, Jun 29 '16 at 18:48
I added an answer to show you exactly what I meant. You probably have to get your hands dirty to make it perfect, but it should be doable. — Andras Deak -- Слава Україні, Jun 29 '16 at 20:46

score 0 · Accepted Answer · answered Jun 29 '16 at 20:45

0

I suggest trying something manual: adding text annotation with position in units of the figure relative coordinates.

Consider these two dummy examples:

hf,ax = plt.subplots(nrows=3)
hf.text(0.5,0.92,
        "Hi there, this is the first line\nAnd this is the second!!!",
        horizontalalignment='center')
hf,ax = plt.subplots(nrows=7)
hf.text(0.5,0.92,
        "Hi there, this is the first line\nAnd this is the second!!!",
        horizontalalignment='center')

The result has the "suptitle" located in the exact same position:

answered Jun 29 '16 at 20:45

Andras Deak -- Слава Україні

33,737
11
83
111

This looks not bad! There's a problem, however. I did some tests and I saw that the vertical height is still a function of the number of plots (not as bad as before though)... What else other than the second parameter can I use to fix it? – The Quantum Physicist Jun 29 '16 at 21:14
@TheQuantumPhysicist I'm not sure I understand. If you open the two images I linked, and swap between them in the browser or an image viewer, you'll see that the titles are positioned exactly the same. Are you modifying something about the figures? You should omit `tight_layout()` for this; that performs dynamic modifications on the plot, which will depend on the number of subplots. – Andras Deak -- Слава Україні Jun 29 '16 at 21:16
@TheQuantumPhysicist I looked at your sample code, and I don't see `tight_layout` there. I suspect that the manual resizing of the figure messes with positioning. If your figure becomes longer, its `0.92` relative `y` coordinate will shift with respect to the top axes. – Andras Deak -- Слава Україні Jun 29 '16 at 21:20
Would you suggest a way to fix this? I removed the hspace part from my test code and the same issue still happens. – The Quantum Physicist Jun 29 '16 at 21:29
@TheQuantumPhysicist well at least as a proof of concept you should try setting the same figure size for every case. I know the window will be cluttered in case of many rows, but then we can determine if this is causing this effect. – Andras Deak -- Слава Україні Jun 29 '16 at 21:35
2

This is exactly what `suptitle` does (except at `0.98` rather than `0.92`). – tacaswell Jun 29 '16 at 23:32
1

See http://matplotlib.org/devdocs/users/annotations_guide.html for some more powerful tools for putting text exactly where you want (ex 15px below the top edge of the figure. – tacaswell Jun 30 '16 at 02:21

Python, Matplotlib: Suptitle with arbitrary number of vertical figures

1 Answers1