BUG: matlab files with deeply lists of arrays with different sizes are no longer created

[Rumbelstilzchen]

Describe your issue.

After updating numpy to a version >=1.24 an error is thrown as the behavior of numpy asanyarry has changed from a deprecation warning to an error. For reference:
https://numpy.org/doc/stable/release/1.24.0-notes.html
and
numpy/numpy#22004

Possible workaround:
Wrapping the line 493 of _mio5.py within a try catch:

scipy/scipy/io/matlab/_mio5.py

Line 493 in dde5059

narr = np.asanyarray(source)

Replace:

narr = np.asanyarray(source)

with

try:
    narr = np.asanyarray(source)
except ValueError:
    narr = np.asanyarray(source, dtype=object)

Reproducing Code Example

import scipy.io as sio
import numpy as np

array_one = np.random.random((5,3))
array_two = np.random.random((6,3))

list_of_arrays = [array_one, array_two]

sio.savemat("D:\\lists.mat", {'data': list_of_arrays}, long_field_names=True, do_compression=True)

Error message

Traceback (most recent call last):
  File "C:\Program Files\JetBrains\PyCharm Community Edition 2021.2.3\plugins\python-ce\helpers\pydev\pydevconsole.py", line 364, in runcode
    coro = func()
  File "<input>", line 10, in <module>
  File "...\venv310\lib\site-packages\scipy\io\matlab\_mio.py", line 300, in savemat
    MW.put_variables(mdict)
  File "...\venv310\lib\site-packages\scipy\io\matlab\_mio5.py", line 884, in put_variables
    self._matrix_writer.write_top(var, name.encode('latin1'), is_global)
  File "...\venv310\lib\site-packages\scipy\io\matlab\_mio5.py", line 633, in write_top
    self.write(arr)
  File "...\venv310\lib\site-packages\scipy\io\matlab\_mio5.py", line 651, in write
    narr = to_writeable(arr)
  File "...\venv310\lib\site-packages\scipy\io\matlab\_mio5.py", line 493, in to_writeable
    narr = np.asanyarray(source)
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (2,) + inhomogeneous part.

SciPy/NumPy/Python version information

Scipy: 1.10.0; Numpy: 1.24.2

[matthew-brett]

Thanks for the report. Do you happen to have an example that triggers the error that we can use for a test?

[Rumbelstilzchen]

Sorry, I do not understand. What in addition to the example under "Reproducing Code Example" from my initial post is needed?
On my setup this raises the error and aborts writing the mat-file.