CODEWITHSUNDEEP
pythonnumpy
user3313834
user3313834

Reputation: 7827

numpy add dtype array

I want to add two dtype array and get the same dtype type of array:

>>> dtype=[('p', '<i8'), ('l', '<f8')]
>>> v1 = np.array([(22, 3.14), (4, 0.1)], dtype=dtype)
>>> v1
array([(22, 3.14), (4, 0.1)], dtype=[('p', '<i8'), ('l', '<f8')])

>>> v2 = np.array([(11, 3.14), (6, 0.2)], dtype=dtype)
>>> v2
array([(11, 3.14), (6, 0.2)], dtype=[('p', '<i8'), ('l', '<f8')])

I want to get:

>>> array([(33, 6.28), (10, 0.3)], dtype=[('p', '<i8'), ('l', '<f8')])

but I get:

>>> v = v1 + v2
TypeError: unsupported operand type(s) for +: 'numpy.ndarray' and 'numpy.ndarray'

or

>>> v = np.sum(v1, v2)
...
TypeError: cannot perform reduce with flexible type

Upvotes: 0

Views: 5323

Answers (2)

zhangxaochen
zhangxaochen

Reputation: 34017

If it's not that important to preserve the structured array:

In [703]: v=array(v1.tolist())+array(v2.tolist())

In [704]: v
Out[704]: 
array([[ 33.  ,   6.28],
       [ 10.  ,   0.3 ]])

otherwise, best way I could come up with is just add column by column as @unutbu mentioned.

Upvotes: 1

unutbu
unutbu

Reputation: 879421

Sadly, I don't know of an easier way than computing each column separately:

import numpy as np

dtype=[('p', '<i8'), ('l', '<f8')]
v1 = np.array([(22, 3.14), (4, 0.1)], dtype=dtype)
v2 = np.array([(11, 3.14), (6, 0.2)], dtype=dtype)

v = np.empty_like(v1)
for col in v1.dtype.names:
    v[col] = v1[col] + v2[col]
print(v)
# [(33L, 6.28) (10L, 0.30000000000000004)]

However, if you install pandas and make v1 and v2 DataFrames, then summing is simple:

import pandas as pd
v1 = pd.DataFrame.from_records(v1)
v2 = pd.DataFrame.from_records(v2)
v = v1 + v2
print(v)

yields

    p     l
0  33  6.28
1  10  0.30

Upvotes: 3

Related Questions