CODEWITHSUNDEEP
arraysnumpy
Yan Chen
Yan Chen

Reputation: 21

Efficient way to combine pairs of arrays?

Lets say I have several x- y- pairs of arrays. I want to create a new pair of arrays that combines the data in the following way:

The new x array contains all the x-values in the x- arrays (only once) The new y array contains the sum of all the y- values in the y- arrays where the corresponding x- values are the same. For instance.

x1 = np.array([0, 1, 2, 3])
y1 = np.array([3, 2, 5, 7])

x2 = np.array([0, 2, 3, 4, 5])
y2 = np.array([4, 5, 4, 1, 6])

x3 = np.array([-2, -1, 0])
y3 = np.array([1, 0, 1])

a = np.array([x1, x2, x3])
b = np.array([y1, y2, y3])

x, y = array_combiner(a, b)

>>> x
array([-2, -1, 0, 1, 2,  3,  4, 5])
>>> y
array([ 1,  0, 8, 2, 10, 11, 1, 6])   #sum from previous arrays from corresponding x-values

Any ideas?

Upvotes: 2

Views: 147

Answers (1)

Divakar
Divakar

Reputation: 221534

We can use np.unique and np.bincount, like so -

# Collect all x-arrays and y-arrays into 1D arrays each
X = np.concatenate((x1,x2,x3))
Y = np.concatenate((y1,y2,y3))

# Get unique X elems & tag each X element based on its uniqueness among others
Xout,tag = np.unique(X,return_inverse=True)

# Use tags to perform tagged summation of Y elements
Yout = np.bincount(tag,Y)

Sample output -

In [64]: Xout
Out[64]: array([-2, -1,  0,  1,  2,  3,  4,  5])

In [65]: Yout
Out[65]: array([  1.,   0.,   8.,   2.,  10.,  11.,   1.,   6.])

Upvotes: 2

Related Questions