CODEWITHSUNDEEP
pythonnumpy
whitebear
whitebear

Reputation: 12433

one-dimesion numpy.sum returns 1

I have numpy one-dimension array like this:

array([0.6441961 , 0.36957273, 1.        , 0.4898495 , 0.24318133,
   0.3721704 , 0.3205053 , 0.16859561, 0.26045567, 0.5081331 ,
   0.66135716, 0.63181865])

My code is below, simply shows the variable.

print(a)
print(np.sum(a,dtype='int16'))

It shows like this below

[0.6441961  0.36957273 1.         0.4898495  0.24318133 0.3721704
 0.3205053  0.16859561 0.26045567 0.5081331  0.66135716 0.63181865]
1

Why does it return 1?

Upvotes: 1

Views: 44

Answers (2)

Daniel F
Daniel F

Reputation: 14399

You want:

int(a.sum())

or

a.sum.astype('int16')  # totally didn't see @DaniMesejo's answer before I edited this in

What's wrong:

np.sum(a, dtype='int16') casts a to int16 before summing. And casting to int does a .floor operation implicitly. So everything in your array that's not 1. becomes 0 - and then the sum is 1

Upvotes: 4

Dani Mesejo
Dani Mesejo

Reputation: 61910

When you do:

res = np.sum(a, dtype='int16')

it says transform every element of a to int16 and then sum, so as every element is between 0 and 1 but 1.0, it transforms them to 0. Let's change an element of the array to verify this claim:

a = np.array([0.6441961, 0.36957273, 1., 2.4898495, 0.24318133, 0.3721704,
              0.3205053, 0.16859561, 0.26045567, 0.5081331, 0.66135716, 0.63181865])

res = np.sum(a, dtype='int16')
print(res)

Output

3

The output is 3 because we now have 1. and 2.4898495. One solution to your problem is to do:

res = np.sum(a).astype('int16')
print(res)

Output

5

Upvotes: 4

Related Questions