Difference between array.copy() vs numpy.copy(array)

Question

I was just looking at the following numpy infographic.

https://s3.amazonaws.com/assets.datacamp.com/blog_assets/Numpy_Python_Cheat_Sheet.pdf

I am wondering if there is any difference between np.copy(a) and a.copy() - or are they just synonyms for the same operation?

aparpara · Accepted Answer · 2019-05-07 19:20:19Z

7

If a is a numpy.array, the result will be the same. But if a is something else, a.copy() will return the same type as a or fail depending on its type, and np.copy(a) will always return numpy.array. Try, e.g. the following:

import pandas as pd

for x in (list(range(3)), np.array(range(3)), pd.Series(range(3))):
    print()
    print(repr(x.copy()))
    print(repr(np.copy(x)))

UPD: There is another difference. Both methods have an additional order argument defining the memory order in the copy with different default values. In np.copy it is 'K', which means "Use the order as close to the original as possible", and in ndarray.copy it is 'C' (Use C order). E.g.

x = np.array([[1,2,3],[4,5,6]], order='F')
for y in [x, np.copy(x), x.copy()]:
    print(y.flags['C_CONTIGUOUS'], y.flags['F_CONTIGUOUS'])

Will print

False True
False True
True False

And in both cases the copies are deep in the sense that the array data itself are copied, but shallow in the sense that in case of object arrays the objects themselves are not copied. Which can be demonstrated by

x = np.array([1, [1,2,3]])
y = x.copy()
z = np.copy(x)
y[1][1] = -2
z[1][2] = -3
print(x)
print(y)
print(z)

All the three printed lines are

[1 list([1, -2, -3])]

edited May 7, 2019 at 19:20

answered May 7, 2019 at 18:37

aparpara

2,20110 silver badges25 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

AlanSTACK Over a year ago

Assuming that a is of type np.array, do both methods np.copy(a) and a.copy() make deep copies?

hpaulj Over a year ago

deep copy only applies to object dtype arrays. For others all copies are the same.

aparpara Over a year ago

@AlanSTACK, they both make deep copy in the sense that the array data are copied. But if you have an object array, the objects are not copied. If you need such a deep-deep copy, use copy.deepcopy.

Collectives™ on Stack Overflow

Difference between array.copy() vs numpy.copy(array)

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related