识别并替换ndarray中的行

Question

This is similar to this question except I have a stricter condition: I need to replace a occurrences of an entire row and not just values that may exist elsewhere in the column. The line #368 should give an idea of how many fancy-indexed experiments have thus far failed:

In [368]:

a=np.array([[[ 42.30803907,   0.        ,   0.        ],
        [ 42.30803907,   3.0052592 ,   0.        ]],

       [[ 42.30803907,   3.0052592 ,   0.        ],
        [ 40.54907736,   5.44198582,   0.        ]],

       [[ 40.54907736,   5.44198582,   0.        ],
        [ 42.14713681,   8.84169967,   0.        ]]])
In [369]:

toBeReplaced=np.array([ 42.30803907,   3.0052592 ,   0.        ])
In [370]:

b=a==toBeReplaced
#b returns:
#array([[[ True, False,  True],
#        [ True,  True,  True]],
#
#       [[ True,  True,  True],
#        [False, False,  True]],
#
#       [[False, False,  True],
#        [False, False,  True]]], dtype=bool)

And here's where I can't seem to get the condition syntax right:

a=np.where(b==[True,True,True], [9,9,9], a)

Out[373]:
array([[[  9.        ,   0.        ,   9.        ],
        [  9.        ,   9.        ,   9.        ]],

       [[  9.        ,   9.        ,   9.        ],
        [ 40.54907736,   5.44198582,   9.        ]],

       [[ 40.54907736,   5.44198582,   9.        ],
        [ 42.14713681,   8.84169967,   9.        ]]])

You can see that the boolean mask replaces all occurrences on a column basis. Is there a way to alter the Boolean mask or the array on which it is used as a selector so that only rows whose value in all three columns matches the columns of the search array?

(NB I should also note that although my data is float64, all the "matching" occurrences I'm looking for return True on np.all() as they are the same underlying computed value.)

Answer 1

You can create a mask identifying where the condition is satisfied in all the columns:

mask = np.all(a==[42.30803907, 3.0052592, 0.], axis=2)
a[mask] = [9, 9, 9]

#array([[[ 42.30803907,   0.        ,   0.        ],
#        [  9.        ,   9.        ,   9.        ]],
# 
#       [[  9.        ,   9.        ,   9.        ],
#        [ 40.54907736,   5.44198582,   0.        ]],
# 
#       [[ 40.54907736,   5.44198582,   0.        ],
#        [ 42.14713681,   8.84169967,   0.        ]]])