Matplotlib - sequentially creating figures with the same size - python-3.x

Question

I need to create a sequence of .pdf files where each .pdf contains a figure with five plots.
As I am going to include them in a LaTeX article, I wanted them all to be the same width and height so that each figure's corners are vertically aligned on both left and right sides.
I thought this would be enough, but apparently not:
common_figsize=(6,5)
fig, ax = plt.subplots(figsize = common_figsize)
# five plots in a loop for the first figure.
# my_code()...
plt.savefig("Figure-1.pdf", transparent=True)
plt.close(fig)
fig, ax = plt.subplots(figsize = common_figsize)
# five plots in a loop for the new figure.
# my_code()...
plt.savefig("Figure-2.pdf", transparent=True)
plt.close(fig)
If I understand correctly, this does not do exactly what I want because of different scales originating from different yticks resolutions.
For both figures, pyplot is fed the same list for xticks.
In this case, it is a list of 50 values, from 1 to 50.
CHUNK_COUNT = 50
x_step = CHUNK_COUNT / 10
new_xticks = list(range(x_step, CHUNK_COUNT + x_step, x_step)) + [1]
plt.xticks(new_xticks)
ax.set_xlim(left=1, right=CHUNK_COUNT)
This creates both figures with an X-axis that goes from 1 to 50.
So far so good.
However, I haven't figured out how to deal with the problem of yticks resolution.
One of the figures had less yticks than the other, so I overrode it to have as many ticks as the other:
# Add yticks to Figure 1.
y_divisor = 6
y_step = (100 - min_y_tick) / y_divisor
new_yticks = [min_y_tick + y_step * i for i in range(0, y_divisor + 1)]
plt.yticks(new_yticks)
This resulted in the following images:
(click on each to open in new tab to see that in fact the bounding square of each figure is different)
Figure 1:
Figure 2:
In summary, I believe matplotlib is accepting the figsize parameter, but then rearranges plot elements to accommodate for different tick values and text lengths.
Is it possible for it to operate in reverse? To change label and text rotations automagically so that the squares are absolutely the same length and height?
Apologies if this is a duplicate and thanks for the help.
EDIT:
Finally able to provide a minimal, complete and verifiable example.
Among the tests, I removed the custom yticks code and the problem still persists:
from matplotlib.lines import Line2D
import matplotlib.ticker as mtick
import matplotlib.pyplot as plt
from matplotlib import rc
# activate latex text rendering
rc('text', usetex=True)
from matplotlib import rcParams
rcParams.update({'figure.autolayout': True})
CHUNK_COUNT = 50
common_figsize=(6,5)
plot_counter = 5
x_step = int(int(CHUNK_COUNT) / 10)
new_xticks = list(range(x_step, int(CHUNK_COUNT) + x_step, x_step)) + [1]
##### Plot Figure 1
fig, ax = plt.subplots(figsize = common_figsize)
plt.ylabel("Summary of a simple YY axis")
plt.yticks(rotation=45)
ax.yaxis.set_major_formatter(mtick.PercentFormatter(is_latex=False))
for i in range(0, plot_counter):
xvals = range(1, CHUNK_COUNT + 1)
yvals = []
for j in xvals:
yvals.append(j + i)
plt.plot(xvals, yvals)
plt.xticks(new_xticks)
ax.set_xlim(left=1, right=int(CHUNK_COUNT))
plt.savefig("Figure_1.png", transparent=True)
plt.close(fig)
##### Plot Figure 2
fig, ax = plt.subplots(figsize = common_figsize)
plt.ylabel("Summary of another YY axis")
plt.yticks(rotation=45)
ax.yaxis.set_major_formatter(mtick.PercentFormatter(is_latex=False))
for i in range(0, plot_counter):
xvals = range(1, CHUNK_COUNT + 1)
yvals = []
for j in xvals:
yvals.append((j + i) / 100)
plt.plot(xvals, yvals)
plt.xticks(new_xticks)
ax.set_xlim(left=1, right=int(CHUNK_COUNT))
plt.savefig("Figure_2.png", transparent=True)
plt.close(fig)

Answer 1

It turns out this was due to a mistake on my part.
I carried over code from another context where
autolayout
was active:
from matplotlib import rcParams
rcParams.update({'figure.autolayout': True})
After setting it to False, the figure squares all had the same dimensions:
from matplotlib import rcParams
rcParams.update({'figure.autolayout': False})
Despite the length differences in ytick elements, it is now respecting the dimensions specified in my original question.
These results were generated with the MWE example I added at the end of my question: