Writing multiple NumPy arrays (the ndarray objects) into a file

The functions savez() and savez_compressed() both write multiple numpy arrays into a file. The savez() function writes the contents in uncompressed format and the savez_compressed() writes in compressed format.

Read Write Files Updated 22 Aug 2024 2 min read

Overview

  • The savez() and savez_compressed() functions of the Python NumPy module write multiple ndarray objects into a disk file.
  • While savez() writes the ndarray objects in uncompressed form, the savez_compressed() writes the arrays in compressed form.
  • For both savez() and savez_compressed(), when no file name extension is given .npz is assumed as the extension of the file.
  • Unlike savez() and savez_compressed() functions the save() function writes only one ndarray per disk file.

Example

python
# Example Python program that uses numpy.savez()

            # function to save multiple arrays into a file.

            import numpy as np

# Create a two-dimensional array

            a1 = np.ones((2, 2), dtype = np.int8)

for x in range (0, 1):

                for y in range (0, 1):

                    a1[x][y] = x * y

# Create a three-dimensional array

            a2 = np.ones((3, 3, 3), dtype = np.int8)

for x in range (0, 3):

                for y in range (0, 3):

                    for z in range (0, 3):

                        a2[x][y][z] = x * y * z

# Write both the arrays into a .npz file - Uncompressed form

            np.savez("multi_array_uncompressed.npz", array1 = a1, array2 = a2)

# Write both the arrays into a .npz file - Compressed form

            np.savez_compressed("multi_array_compressed.npz", array1 = a1, array2 = a2)

Comparing the size of .npz files - compressed vs uncompressed

The compressed version saves the overhead of 129 bytes when compared to the uncompressed version.

python
-rw-r--r--  1 root  u1  412 Aug 22 14:29 multi_array_compressed.npz

-rw-r--r--  1 root  u1  541 Aug 22 14:29 multi_array_uncompressed.npz

Contents of the uncompressed file

python
PK^C^D-^@^@^@^@^@^@^@!^@ ¯F´ÿÿÿÿÿÿÿÿ

^@^T^@array1.npy^A^@^P^@<84>^@^@^@^@^@^@^@<84>^@^@^@^@^@^@^@<93>NUMPY^A^@v^@{'descr': '|i1', 'fortran_order': False, 'shape': (2, 2), }

^@^A^A^APK^C^D-^@^@^@^@^@^@^@!^@À¾<95>·ÿÿÿÿÿÿÿÿ

^@^T^@array2.npy^A^@^P^@<9b>^@^@^@^@^@^@^@<9b>^@^@^@^@^@^@^@<93>NUMPY^A^@v^@{'descr': '|i1', 'fortran_order': False, 'shape': (3, 3, 3), }

^@^@^@^@^@^@^@^@^@^@^@^@^@^A^B^@^B^D^@^@^@^@^B^D^@^D^HPK^A^B-^C-^@^@^@^@^@^@^@!^@ ¯F´<84>^@^@^@<84>^@^@^@

^@^@^@^@^@^@^@^@^@^@^@<80>^A^@^@^@^@array1.npyPK^A^B-^C-^@^@^@^@^@^@^@!^@À¾<95>·<9b>^@^@^@<9b>^@^@^@

^@^@^@^@^@^@^@^@^@^@^@<80>^AÀ^@^@^@array2.npyPK^E^F^@^@^@^@^B^@^B^@p^@^@^@<97>^A^@^@^@^@

Contents of the compressed file

The file is opened using the vi editor.

python
sudo vi multi_array_compressed.npz

The savez_compressed() function writes everything in compressed form including the header information.

python
PK^C^D-^@^@^@^H^@^@^@!^@ ¯F´ÿÿÿÿÿÿÿÿ

^@^T^@array1.npy^A^@^P^@<84>^@^@^@^@^@^@^@J^@^@^@^@^@^@^@<9b>ì^Wê^[^PÉÈPÆP­<9e><92>Z<9c>\¤n¥ ^<93>i¨®£ <9e><96>_TR<94><98>^W<9f>_<94><92>

^RwKÌ)N^E<8a>^Wg$^V¤^Bù^ZF:

F<9a>:

µ

d^C.^FFFF^@PK^C^D-^@^@^@^H^@^@^@!^@À¾<95>·ÿÿÿÿÿÿÿÿ

^@^T^@array2.npy^A^@^P^@<9b>^@^@^@^@^@^@^@T^@^@^@^@^@^@^@<9b>ì^Wê^[^PÉÈPÆP­<9e><92>Z<9c>\¤n¥ ^<93>i¨®£ <9e><96>_TR<94><98>^W<9f>_<94><92>

^RwKÌ)N^E<8a>^Wg$^V¤^Bù^ZÆ:

 ¤©£P«@^^àb@^F<8c>L^LL, ^F<90>dá^@^@PK^A^B-^C-^@^@^@^H^@^@^@!^@ ¯F´J^@^@^@<84>^@^@^@

^@^@^@^@^@^@^@^@^@^@^@<80>^A^@^@^@^@array1.npyPK^A^B-^C-^@^@^@^H^@^@^@!^@À¾<95>·T^@^@^@<9b>^@^@^@

^@^@^@^@^@^@^@^@^@^@^@<80>^A<86>^@^@^@array2.npyPK^E^F^@^@^@^@^B^@^B^@p^@^@^@^V^A^@^@^@^@