Loading arrays from numpy npz files in python

Question

I usually save data in npz files in python. How to write a function which loads the npz file and automatically creates arrays which are present in the .npz file. For example, say there are three arrays A, B, and C in a file named some_data.npz.

What I want the function to do is load the npz file as

data1 = np.load('some_data.npz')

and then automatically create three arrays named data1A, data1B and data1C which stores the arrays A, B, and C from the original .npz file. How to do this?

Doesn't the numpy.load function already give you a dictionary of arrays? If you just want to bind each individual array to some fancy names, just assign them. Possibly, write a class and make the three names class attributes. — Cong Ma
– Cong Ma, Commented Sep 20, 2015 at 18:49
@CongMa exactly. The second example shows how this is accomplished. — MattDMo
– MattDMo, Commented Sep 20, 2015 at 18:52
You can read this as well stackoverflow.com/a/71183327/16733101 — Hamzah Al-Qadasi
– Hamzah Al-Qadasi, Commented Feb 19, 2022 at 8:34

Padraic Cunningham · Accepted Answer · 2015-09-20 19:08:22Z

8

If you want to create names store the arrays in a dict:

a1 = np.array([1,2,3])
a2 = np.array([4,5,6])
a3 = np.array([7,8,9])

np.savez("test", A=a1,B=a2,C=a3)
a = np.load("test.npz")
d = dict(zip(("data1A","data1B","data1C"), (a[k] for k in a)))
print(d)
{'data1A': array([4, 5, 6]), 'data1C': array([7, 8, 9]), 'data1B': array([1, 2, 3])}

If you want to create the keys without passing the names explicitly:

a1 = np.array([1, 2, 3])
a2 = np.array([4, 5, 6])
a3 = np.array([7, 8, 9])

np.savez("test", A=a1,B=a2,C=a3)
a = np.load("test.npz")

d = dict(zip(("data1{}".format(k) for k in a), (a[k] for k in a)))
print(d)

edited Sep 20, 2015 at 19:08

answered Sep 20, 2015 at 18:59

Padraic Cunningham

181k30 gold badges264 silver badges327 bronze badges

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

3 Comments

areuexperienced Over a year ago

Is this not essentially the same as the object returned by np.load? Do correct me if I'm wrong.

Padraic Cunningham Over a year ago

@areuexperienced, load does not let you pick the name so if you want to store all the arrays with certain names then this is how to do it

areuexperienced Over a year ago

Acknowledged. It's just that one would (or should) usually specify names at savetime, hence making it redundant. Your comment still stands though.

Warren Weckesser · Accepted Answer · 2015-09-21 05:19:52Z

You can almost do that already, via the f attribute of the object returned by numpy.load. For example, in the following, foo.npz contains three arrays, A, B and C:

In [1367]: foo = np.load('foo.npz')

In [1368]: foo.keys()
Out[1368]: ['A', 'C', 'B']

In [1369]: foo.f.A
Out[1369]: array([ 0.,  1.,  2.,  3.,  4.,  5.,  6.,  7.])

In [1370]: foo.f.B
Out[1370]: 
array([[ 0,  1],
       [-1,  0]])

In [1371]: foo.f.C
Out[1371]: array([ 3.14159265,  2.71828183,  0.57721566])

Note: The f attribute is not documented in the docstring of load. When load reads an npz file, it returns an instance of the class NpzFile. This class is available as numpy.lib.npyio.NpzFile. The docstring of the NpzFile class describes the f attribute. (As of this writing, the source code of the class can be found here: https://github.com/numpy/numpy/blob/master/numpy/lib/npyio.py#L95.)

Exa · Accepted Answer · 2019-07-18 10:32:35Z

1

# Assuming that you saved the original data with labels 'A', 'B', and 'C'
import numpy as np

a1 = np.array([1,2,3])
a2 = np.array([4,5,6])
a3 = np.array([7,8,9])
# Save the arrays:
np.savez_compressed('some_data.npz', A=a1,B=a2,C=a3)

# Now Load Using,
data1 = np.load('some_data.npz', 'r')
data1A = data1['A']
data1B = data1['B']
data1C = data1['C']

Hope this helps !!

edited Jul 18, 2019 at 10:32

Exa

4,1168 gold badges46 silver badges63 bronze badges

answered Feb 28, 2019 at 9:24

senior_mle

8591 gold badge11 silver badges20 bronze badges

Collectives™ on Stack Overflow

Loading arrays from numpy npz files in python

3 Answers 3

3 Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

3 Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related