Why does plotting something with matplotlib change the first outcome of a numpy.linalg.inv inversion with complex entries and how to avoid it?

Question

I was coding for a project of mine and found some strange nan values appearing in my code the first time I ran it, which disappeared when I ran the same line of code a second time. Searching for the reason took me some hours and I nailed it down to the following unbelievable situation:

The first time one runs a matrix inversion (with complex entries) via numpy.linalg.inv after plotting some matplotlib figure changes the outcome of the matrix inversion (and makes some nan values appear). The correct matrix inversion result (with no nans) can then be recovered, when calculating the same inversion again (or a different one with complex values).

These simple lines of code should be enough to reproduce the problem:

import numpy as np
import matplotlib.pyplot as plt

print(np.linalg.inv([[1+0j,0],[0,1]]))

fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
plt.show()    

print(np.linalg.inv([[1+0j,0],[0,1]]))
print(np.linalg.inv([[1+0j,0],[0,1]]))

I get the output of:

[[1.+0.j 0.+0.j]
 [0.+0.j 1.+0.j]]

**<here sits the canavas of the empty plot>**

[[nan+nanj  0. +0.j]
 [nan +0.j  1. +0.j]]
[[1.+0.j 0.+0.j]
 [0.+0.j 1.+0.j]]

As one can see in the second inversion the first entries in the inverted matrix contain nan values.

If anyone can explain this behaviour to me and give me a way to work around it (other than calculating each inversion twice), I would be very happy!

Thanks a lot!

Frederix96

I use jupythernotebook with python over anaconda on Windows 10 (ipython 7.19.0, python 3.7.3, matplotlib 3.3.3, numpy 1.19.4).

Update:

I just reinstalled anaconda and set up a separate python environment in anaconda. I inside that environment I just installed python, ipython, numpy and matplotlib (and all given dependencies).

Also still in this environment the problem still appears. If anyone has any further ideas, I would be happy to hear them.

The only installed packages in this enviroment are now:

# Name                    Version                   Build  Channel
backcall                  0.2.0                      py_0
blas                      1.0                         mkl
ca-certificates           2020.12.8            haa95532_0
certifi                   2020.12.5        py39haa95532_0
colorama                  0.4.4              pyhd3eb1b0_0
cycler                    0.10.0           py39haa95532_0
decorator                 4.4.2                      py_0
freetype                  2.10.4               hd328e21_0
icu                       58.2                 ha925a31_3
intel-openmp              2020.2                      254
ipython                   7.19.0           py39haa95532_0
ipython_genutils          0.2.0              pyhd3eb1b0_1
jedi                      0.17.2           py39haa95532_1
jpeg                      9b                   hb83a4c4_2
kiwisolver                1.3.0            py39h604cdb4_0
libpng                    1.6.37               h2a8f88b_0
libtiff                   4.1.0                h56a325e_1
lz4-c                     1.9.2                hf4a77e7_3
matplotlib                3.3.2                haa95532_0
matplotlib-base           3.3.2            py39h35d3fe4_0
mkl                       2020.2                      256
mkl-service               2.3.0            py39h196d8e1_0
mkl_fft                   1.0.6            py39h054117c_0
mkl_random                1.0.2            py39h848d8c7_0
numpy                     1.19.2           py39h729668d_0
numpy-base                1.19.2           py39hbd0edd7_0
olefile                   0.46                       py_0
openssl                   1.1.1i               h2bbff1b_0
parso                     0.7.0                      py_0
pickleshare               0.7.5           pyhd3eb1b0_1003
pillow                    8.0.1            py39h4fa10fc_0
pip                       20.3.3           py39haa95532_0
prompt-toolkit            3.0.8                      py_0
pygments                  2.7.3              pyhd3eb1b0_0
pyparsing                 2.4.7                      py_0
pyqt                      5.9.2            py39hd77b12b_6
python                    3.9.1                h6244533_2
python-dateutil           2.8.1                      py_0
qt                        5.9.7            vc14h73c81de_0
setuptools                51.0.0           py39haa95532_2
sip                       4.19.13          py39hd77b12b_0
six                       1.15.0           py39haa95532_0
sqlite                    3.33.0               h2a8f88b_0
tk                        8.6.10               he774522_0
tornado                   6.1              py39h2bbff1b_0
traitlets                 5.0.5                      py_0
tzdata                    2020d                h14c3975_0
vc                        14.2                 h21ff451_1
vs2015_runtime            14.27.29016          h5e58377_2
wcwidth                   0.2.5                      py_0
wheel                     0.36.2             pyhd3eb1b0_0
wincertstore              0.2              py39h2bbff1b_0
xz                        5.2.5                h62dcd97_0
zlib                      1.2.11               h62dcd97_4
zstd                      1.4.5                h04227a9_0

For further information this is the MKL numpy is using (generated by np.show_config())

blas_mkl_info:
    libraries = ['mkl_rt']
    library_dirs = ['C:/Users/Frederic/anaconda3\\Library\\lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['C:/Users/Frederic/anaconda3\\Library\\include']
blas_opt_info:
    libraries = ['mkl_rt']
    library_dirs = ['C:/Users/Frederic/anaconda3\\Library\\lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['C:/Users/Frederic/anaconda3\\Library\\include']
lapack_mkl_info:
    libraries = ['mkl_rt']
    library_dirs = ['C:/Users/Frederic/anaconda3\\Library\\lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['C:/Users/Frederic/anaconda3\\Library\\include']
lapack_opt_info:
    libraries = ['mkl_rt']
    library_dirs = ['C:/Users/Frederic/anaconda3\\Library\\lib']
    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]
    include_dirs = ['C:/Users/Frederic/anaconda3\\Library\\include']