How to write a multidimensional array to a text file?

File I/O can often be a bottleneck in codes. That's why it's important to know that ASCII I/O is always magnitudes slower that binary I/O. I've compared some of the suggested solutions with perfplot:

Code to reproduce the plot:

import json
import pickle

import numpy as np
import perfplot
import scipy.io


def numpy_save(data):
    np.save("test.dat", data)


def numpy_savetxt(data):
    np.savetxt("test.txt", data)


def numpy_savetxt_fmt(data):
    np.savetxt("test.txt", data, fmt="%-7.2f")


def pickle_dump(data):
    with open("data.pkl", "wb") as f:
        pickle.dump(data, f)


def scipy_savemat(data):
    scipy.io.savemat("test.dat", mdict={"out": data})


def numpy_tofile(data):
    data.tofile("test.txt", sep=" ", format="%s")


def json_dump(data):
    with open("test.json", "w") as f:
        json.dump(data.tolist(), f)


perfplot.save(
    "out.png",
    setup=np.random.rand,
    n_range=[2 ** k for k in range(20)],
    kernels=[
        numpy_save,
        numpy_savetxt,
        numpy_savetxt_fmt,
        pickle_dump,
        scipy_savemat,
        numpy_tofile,
        json_dump,
    ],
    equality_check=None,
)

You can also store NumPy multidimensional array data in .npy file type(it's a binary file).

1. Use numpy save() function to store data to a file:

import numpy as np
a = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) #shape (3x3)
np.save('filename.npy', a)

2. Unload using numpy load() function:

b = np.load('filename.npy')

Pickle is best for these cases. Suppose you have a ndarray named x_train. You can dump it into a file and revert it back using the following command:

import pickle

###Load into file
with open("myfile.pkl","wb") as f:
    pickle.dump(x_train,f)

###Extract from file
with open("myfile.pkl","rb") as f:
    x_temp = pickle.load(f)

Use JSON module for multidimensional arrays, e.g.

import json
with open(filename, 'w') as f:
   json.dump(myndarray.tolist(), f)

Write to a file with Python's print():

import numpy as np
import sys

stdout_sys = sys.stdout
np.set_printoptions(precision=8) # Sets number of digits of precision.
np.set_printoptions(suppress=True) # Suppress scientific notations.
np.set_printoptions(threshold=sys.maxsize) # Prints the whole arrays.
with open('myfile.txt', 'w') as f:
    sys.stdout = f
    print(nparr)
    sys.stdout = stdout_sys

Use set_printoptions() to customize how the objects are displayed.

If your array is numpy.array or torch.tensor and dimension is under 4.
Use this code.

# from util.npa2csv import Visualarr; Visualarr(x)
import numpy as np
import torch

def Visualarr(arr, out = 'array_out.txt'):
    dim = arr.ndim 
    if isinstance(arr, np.ndarray):
        # (#Images, #Chennels, #Row, #Column)
        if dim == 4:
            arr = arr.transpose(3,2,0,1)
        if dim == 3:
            arr = arr.transpose(2,0,1)

    if isinstance(arr, torch.Tensor):
        arr = arr.numpy()
    
    
    with open(out, 'w') as outfile:    
        outfile.write('# Array shape: {0}\n'.format(arr.shape))
        
        if dim == 1 or dim == 2:
            np.savetxt(outfile, arr, fmt='%-7.3f')

        elif dim == 3:
            for i, arr2d in enumerate(arr):
                outfile.write('# {0}-th channel\n'.format(i))
                np.savetxt(outfile, arr2d, fmt='%-7.3f')

        elif dim == 4:
            for j, arr3d in enumerate(arr):
                outfile.write('\n# {0}-th Image\n'.format(j))
                for i, arr2d in enumerate(arr3d):
                    outfile.write('# {0}-th channel\n'.format(i))
                    np.savetxt(outfile, arr2d, fmt='%-7.3f')

        else:
            print("Out of dimension!")

    

def test_va():
    arr = np.random.rand(4,2)
    tens = torch.rand(2,5,6,3)
    Visualarr(arr)

test_va()