为什么这种证明功能不能正常工作？ 证明正确而不是贬低

如何解决为什么这种证明功能不能正常工作？ 证明正确而不是贬低

如果存在NaN值，则后面的justify function应该将数据fown移动到行中（这样，NaN将位于顶部。 但是不知何故，“向下”选项与“正确”功能相同。 如何解决？

以下是可再现的数据（3d数组）：

import numpy as np
def justify(a,invalid_val=0,axis=1,side='left'):    
    """  Justifies a 2D array 
    Parameters
    ----------
    A : ndarray
        Input array to be justified
    axis : int
        Axis along which justification is to be made
    side : str
        Direction of justification. It could be 'left','right','up','down'
        It should be 'left' or 'right' for axis=1 and 'up' or 'down' for axis=0. 
    """ 
    if invalid_val is np.nan:
        mask = ~np.isnan(a)
    else:
        mask = a!=invalid_val
    justified_mask = np.sort(mask,axis=axis)
    if (side=='up') | (side=='left'):
        justified_mask = np.flip(justified_mask,axis=axis)
    out = np.full(a.shape,invalid_val) 
    if axis==1:
        out[justified_mask] = a[mask]
    else:
        out.T[justified_mask.T] = a.T[mask.T]
    return out

a = np.asarray([
 np.asarray([np.asarray([ 1.21643707,0.9280912,5.20711915]),np.asarray([-2.01148217,2.72869681,2.54161257]),np.asarray([ 0.49170286,0.72304396,1.56706948]),np.asarray([-0.77553082,2.74300372,0.10107189]),np.asarray([ 2.54368976,0.53107898,0.09351025]),np.asarray([ 1.03176737,-0.54742843,np.nan])])])


b = justify(a,invalid_val=np.nan,side='down')
print(a) 
print(b)

实际输出：

#a
[[[ 1.21643707  0.9280912   5.20711915]
  [-2.01148217  2.72869681  2.54161257]
  [ 0.49170286  0.72304396  1.56706948]
  [-0.77553082  2.74300372  0.10107189]
  [ 2.54368976  0.53107898  0.09351025]
  [ 1.03176737 -0.54742843         nan]]]
_
#b
[[[ 1.21643707  0.9280912          nan]
  [ 5.20711915 -2.01148217  2.72869681]
  [ 2.54161257  0.49170286  0.72304396]
  [ 1.56706948 -0.77553082  2.74300372]
  [ 0.10107189  2.54368976  0.53107898]
  [ 0.09351025  1.03176737 -0.54742843]]]

实际预期输出：

#a
[[[ 1.21643707  0.9280912   5.20711915]
  [-2.01148217  2.72869681  2.54161257]
  [ 0.49170286  0.72304396  1.56706948]
  [-0.77553082  2.74300372  0.10107189]
  [ 2.54368976  0.53107898  0.09351025]
  [ 1.03176737 -0.54742843         nan]]]
_
# expected b
[[[ 1.21643707  0.9280912          nan]
  [-2.01148217  2.72869681  5.20711915]
  [ 0.49170286  0.72304396  2.54161257]
  [-0.77553082  2.74300372  1.56706948]
  [ 2.54368976  0.53107898  0.10107189]
  [ 1.03176737 -0.54742843  0.09351025]]]

因此，“真实”输出与输入side =“ right”相同。为什么会发生这种情况，或者有更好的解决办法？

编辑：不自觉地写了“我的”，我并不是说它是我的，对不起。 （来源：Divakar）

EDIT2：更详细的示例（3d）：

# in reality were dealing with about 1million samples (==2d matrices) of 50 'columns' and 10.000 'rows'

samples = np.asarray([np.asarray([
np.asarray([89.319787,1.329743,99.234670,52.329743,0.319787,2.319787]),np.asarray([84.319787,49.329743,0.319,np.asarray([12.319787,np.nan,33.329743,np.asarray([33.319787,23.329743,np.asarray([23.319787,45.234670,0.32721,np.asarray([89.319787,2.319787])
]),np.asarray([
np.asarray([89.319787,np.asarray([np.nan,2.319787])])])