ndimage.maximum_filter undocumented behaviour using footprint and mode sequence #9632
Description
My issue is about the function scipy.ndimage.maximum_filter and its behaviour while using both footprints and a sequence of modes. The current documentation on the site* doesn't mention it, but unless the footprint is seperable scipy silently assumes that the "mode" argument is a string and then raises an error because it isn't.
*link: https://docs.scipy.org/doc/scipy/reference/generated/scipy.ndimage.maximum_filter.html
Reproducing code example:
import numpy as np
from scipy.ndimage.filters import maximum_filter
arr = np.zeros([5, 5])
arr[2, 0] = 1
#use a small diamond - shaped filter
footprint = np.ones([3, 3])
#""" #uncomment this line to make footprint seperable
footprint[0, 0] = 0
footprint[0, 2] = 0
footprint[2, 0] = 0
footprint[2, 0] = 0
#"""
#no sequence - works
filt = maximum_filter(arr,
footprint=footprint,
mode='constant')
#sequence (even one whih should do the same) - doesn't work
filt = maximum_filter(arr,
footprint=footprint,
mode=('constant', 'constant'))
Error message:
Traceback (most recent call last):
File "C:\Users\Sonicrs\Desktop\scipy_bug_tester.py", line 31, in <module>
mode=('constant', 'constant'))
File "C:\Users\Sonicrs\AppData\Local\Programs\Python\Python37\lib\site-packages\scipy\ndimage\filters.py", line 1097, in maximum_filter
cval, origin, 0)
File "C:\Users\Sonicrs\AppData\Local\Programs\Python\Python37\lib\site-packages\scipy\ndimage\filters.py", line 1020, in _min_or_max_filter
mode = _ni_support._extend_mode_to_code(mode)
File "C:\Users\Sonicrs\AppData\Local\Programs\Python\Python37\lib\site-packages\scipy\ndimage\_ni_support.py", line 52, in _extend_mode_to_code
raise RuntimeError('boundary mode not supported')
RuntimeError: boundary mode not supported
Scipy/Numpy/Python version information:
1.2.0 1.15.4 sys.version_info(major=3, minor=7, micro=0, releaselevel='final', serial=0)