I want to execute two filtering and one replacing steps on a large array but I cannot figure a way to do it efficiently without making some copy or for loop.
Example, imagine two numpy arrays of different sizes. I want to apply two filters. The first one to select only elements of b that are inferior to 6 (b') and a second filters to replace the values of b' by the values of a if a is inferior to b':
a = np.array( [ 2,2,2,2,2,2] )
b = np.array( [ 0,1,2,3,4,5,6,7,8,9] )
I apply a first masking by selecting the elements of b inferior to 6 :
m = b < 6
Now, I want to replace the value of b[m] by the minimal value between a and b[m], with expected results in b :
[ 0,1,2,2,2,2,6,7,8,9]
Using :
n = a < b[m]
b[m][n] = a[n]
doesn't work. Probably because of some intermediate array. With
c = np.array( [ 0, 1, 2, 3, 4, 5 ] )
I can directly do :
c[ a < c ] = a [ a < c ]
and it works. Any cool slicing way to do it without making secondary array ? Thanks.
numpy.where()could be useful here, possibly nested... – MattDMo Dec 13 '13 at 15:19