Repository navigation
Closed
Description
Bug report
Bug summary
fig.savefig(...) changes the size of the figure it is trying to save to file
minor note: problem doesn't occur when size_figure_inches<=5
Code for reproduction
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
FORCE_RESIZE_WINDOW = False
class Figure:
size_figure_inches = 10.0 #< size of full figure
size_world_half = 5 #< size of world (centered 0)
def __init__(self, fid):
self.fig = plt.figure(fid)
self.canvas = self.fig.canvas
#--- specify figure
sfi = self.size_figure_inches
self.fig.set_size_inches(sfi, sfi)
self.fig.set_dpi(100)
# --- specify axis
self.ax = self.fig.add_axes([0, 0, 1, 1]) # < (full image)
self.cla()
def cla(self):
if FORCE_RESIZE_WINDOW:
sfi = self.size_figure_inches
self.fig.set_size_inches(sfi, sfi)
self.fig.set_dpi(100)
# --- clear axis (resets limits)
plt.cla()
# --- set limits
swh = self.size_world_half
self.ax.set_ylim(-swh, +swh)
self.ax.set_xlim(-swh, +swh)
#--- setup grid
self.ax.set_xticks(range(-swh, +swh))
self.ax.set_yticks(range(-swh, +swh))
self.ax.grid(color='k')
self.ax.grid(True)
def plot2(self, P):
x = P[:, 0]
y = P[:, 1]
self.ax.plot(x, y)
#------------------------------------------------------------------------------------------
t = np.arange(0.0, 2.0, 0.01)
s = 1 + np.sin(2*np.pi*t)
x = np.cos(2*np.pi*t)
y = np.sin(2*np.pi*t)
M_restpose = np.vstack([x,y]).transpose()
print("backend: ", matplotlib.get_backend() )
fig = Figure(1)
offs = np.arange(-1, +1, .25)
for i, off in enumerate(offs):
fig.cla()
fig.plot2(M_restpose)
fig.fig.savefig("./test%d.png" % i)
plt.pause(.1)
print("Figure size for test%d " % i, fig.fig.get_size_inches())Actual outcome
backend: TkAgg
Figure size for test0 [10. 9.66]
Figure size for test1 [10. 9.32]
Figure size for test2 [10. 8.98]
Figure size for test3 [10. 8.64]
Figure size for test4 [10. 8.3]
Figure size for test5 [10. 7.96]
Figure size for test6 [10. 7.62]
Figure size for test7 [10. 7.28]
Below the images generated by test0 and test7


Expected outcome
Image size should not change.
Result can be achieved by setting FORCE_RESIZE_WINDOW=True in the code above.
The terminal output is then:
backend: TkAgg
Figure size for test0 [10. 9.66]
Figure size for test1 [10. 9.66]
Figure size for test2 [10. 9.66]
Figure size for test3 [10. 9.66]
Figure size for test4 [10. 9.66]
Figure size for test5 [10. 9.66]
Figure size for test6 [10. 9.66]
Figure size for test7 [10. 9.66]
While the script reports 9.66in in height, note the outputted files are correctly 1000x1000 as shown below:


Matplotlib version
- Operating system: debian
- Matplotlib version: 2.1.2 (installed w/ pip)
- Matplotlib backend (
print(matplotlib.get_backend())): TkAgg - Python version: 3.5
- Jupyter version (if applicable): N/A
- Other libraries:
Metadata
Metadata
Assignees
Labels
No labels
Activity
afvincent commentedon Feb 21, 2018
@ataiya Thanks for opening an issue ticket. This problem might have been fixed by #10292 , that is in Matplotlib 2.2.0rc1 (IIRC). Would it be possible for you to test the release candidate, to see if the issue is fixed or not for you? Not totally sure about the proper command line for
pip, but I guess that it would look likeIf somebody else could confirm ^^.
Edit: some French had made its way to the former version...
phobson commentedon Feb 21, 2018
This looks resolved at least since 78c484e
taiya commentedon Feb 21, 2018
Hi @afvincent
Indeed, upgrading to the pre-release seems to have fixed the problem!
And your command to upgrade the package was spot on :)
Too bad a google search didn't return anything :(
Took a while to assemble the bug report
@afvincent
What about the wrong size that the application is still reporting?
Should I file a separate bug for that (doesn't block me directly)
afvincent commentedon Feb 21, 2018
@ataiya You're welcome. And don't worry, the bug report was appreciated. Hopefully we'll see more of them if you notice other Matplotlib issues in the future. Be aware that if Google is not returning matches, there is also a search engine on the GitHub tracker that could help you to look for similar problems among the issues and PRs.