numpy 2d其中每行有约束

Question

I have a numpy array of shape (2,N) and three constraints to filter on.

1. No value in a column may be negative
2. Row 0 must be less than a constant, say xmax
3. Row 1 must be less than a constant, say ymax

Here is what I've tried:

xmax, ymax = 7, 8
dst = np.linspace(-2,10,20).reshape((2,-1)).astype(np.int32)
mask = np.where((0 <= dst) & (dst[1,:] < xmax) & (dst[0,:] < ymax))
mask = np.vstack(mask).T
for p in mask:
    print(p, dst[:,p[1]])

Which produces

(array([[-2, -1,  0,  0,  0,  1,  1,  2,  3,  3],
   [ 4,  4,  5,  6,  6,  7,  8,  8,  9, 10]]), (2L, 10L))
(array([0, 2]), array([0, 5]))
(array([0, 3]), array([0, 6]))
(array([0, 4]), array([0, 6]))
(array([1, 0]), array([-2,  4]))  #<-- Why do I get this ??
(array([1, 1]), array([-1,  4]))  #<-- Why do I get this ??
(array([1, 2]), array([0, 5]))
(array([1, 3]), array([0, 6]))
(array([1, 4]), array([0, 6]))

What am I doing wrong that is producing those two unwanted results (-2,4) and (-1,4)?

Answer 1

As mentioned by @hpaulj, the mistake was in the where clause. I did not realize that I could treat this as a continuous boolean array. What I learned is that if I keep the dst >= 0 check, I maintain my dimensions and then I can add two more constraints to check for less than zero explicitly on each row value.

xmax, ymax = 7, 8
dst = np.linspace(-2,10,20).reshape((2,-1)).astype(np.int32)
mask = np.argwhere((dst >= 0) & (dst[0,:] >= 0) & (dst[1,:] >= 0) & (dst[1,:] < xmax) & (dst[0,:] < ymax))
print(mask)
for p in mask:
    print(p, dst[:,p[1]])

Answer 2

In [5]: xmax, ymax = 7, 8
   ...: dst = np.linspace(-2,10,20).reshape((2,-1)).astype(np.int32)
   ...: mask = np.where((0 <= dst) & (dst[1,:] < xmax) & (dst[0,:] < ymax))
   ...: 
In [6]: mask
Out[6]: 
(array([0, 0, 0, 1, 1, 1, 1, 1], dtype=int32),
 array([2, 3, 4, 0, 1, 2, 3, 4], dtype=int32))

In [8]: dst[mask]
Out[8]: array([0, 0, 0, 4, 4, 5, 6, 6])

mask does select the right elements of dst , but from either row.

But this selects from both rows, using just part of the the mask info. So it has some of the masked values, but also the value from the other row.

In [9]: dst[:,mask[1]]
Out[9]: 
array([[ 0,  0,  0, -2, -1,  0,  0,  0],
       [ 5,  6,  6,  4,  4,  5,  6,  6]])
In [10]: dst[0,mask[1]]
Out[10]: array([ 0,  0,  0, -2, -1,  0,  0,  0])
In [11]: dst[1,mask[1]]
Out[11]: array([5, 6, 6, 4, 4, 5, 6, 6])