I have been trying to get simple numpy masking working without having to resort to non-numpy functions, however, I have ran into what seems like a bug.
import numpy as np
origlist = np.array([[b'\x00'] * 100 ] * 128, dtype=np.object)
origlist[0][0] = b'\x00\x00'
newlist = origlist[0][origlist[0] != b'\x00\x00']
provides newlist as [b'\x00\00', b'\x00', b'\x00', ... ], where as it is supposed to provide it as [b'\x00', b'\x00', ... ].
Similarly,
import numpy as np
origlist = np.array([[b'\x00'] * 100 ] * 128, dtype=np.object)
origlist[0][0] = b'\x00\x00'
newlist = origlist[origlist != b'\x00']
also provides newlist as [b'\x00\00', b'\x00', b'\x00', ... ], where as it is supposed to provide it as [b'\x00\x00' ].
UPDATE
Comparing numpy array of dtype object
I have tried basically everything mentioned in the post above, and nothing helped, all comparisons always evaluated to True. I have also tried replacing the initialization with np.full instead, and same results.
The only way I was able to get it working with numpy version 1.19, is to not use numpy.
import numpy as np
origlist = np.full((128,100),b'\x00',dtype=object) #np.object is depreciated
origlist[0][0] = b'\x00\x00'
newlist = origlist[0].tolist()
newlist = list( filter( (b'\x00').__ne__, newlist))
now returns the correct [b'\x00\x00' ].
If there is any other numpy way to do this, or any numpy way to fix this, please let me know.
