numpy：如何便宜地在numpy矩阵切片上应用布尔索引数组

Question

I'm trying to apply a boolean mask array over a slice of my numpy matrix. I only figure out the following to implement this, by first copy this slice out to a single array, then apply boolean mask over this temporary array and copy it back to the matrix.

I'm wondering if there exists (sure it exists! but I can't find it) a cheaper way to implement it? aka, without the cost of copying.

#(data is a numpy 2d matrix)

tmp = data[i,:]
tmp[tmp==0] = mean
data[i,:] = tmp

Answer 1

What about this?

mask = data[i,:] == 0
data[i,mask] = mean