Save numpy array as binary to read from FORTRAN

Question

I have a series of numpy array, i need to save these numpy array in a loop as a raw binary float32 (without any header information) which need to be read from FORTRAN.

import numpy as np
f=open('test.bin','wb+')
for i in range(0,10):
    np_data=np.random.rand(10,5)
    fortran_data=np.asfortranarray(np_data,'float32')
  fortran_data.tofile(f)

f.close()

Is this the correct way so that I can read this binary file created in python from FORTRAN correctly. Your suggestions will be highly apprecitaed

netCDF library might be worth a look, unidata.ucar.edu/software/netcdf/examples/programs — J.J. Hakala
– J.J. Hakala, Commented Jun 19, 2016 at 10:53
does this code work? i think tofile is a method of the array, not the file. In any case you can certainly read raw binary in fortran, open with access=stream and form=unformatted — agentp
– agentp, Commented Jun 19, 2016 at 15:00
now i'm not sure what the question is. There are a few drawbacks to this simple approach, notably the data type/dimensions are not encoded in the file, so the reader program will need to be hard coded to read 10 10x5 arrays of floats, and so on. (not a fortran issue, same if you want to read back into python) Also be aware of portablity issues moving binary files to different systems. — agentp
– agentp, Commented Jun 19, 2016 at 20:19

victortxa · Accepted Answer · 2018-03-08 17:41:03Z

The code you wrote is almost right, but the .tofile method always write the vector in C order. I don't know why the np.asfortranarray() avoids this when writing in the binary file, but I tested and unfortunately we need to transpose the matrix before writing to correct read in Fortran without any other concern (this means in Fortran you can give the actual matrix dimension without needing any transpose). The code below is to illustrate with a 3D matrix (which I ussually need to use) what I am saying:

a = np.arange(1,10*3*4+1)
b = a.reshape(10,12,order='F')
array([[  1,  11,  21,  31,  41,  51,  61,  71,  81,  91, 101, 111],
   [  2,  12,  22,  32,  42,  52,  62,  72,  82,  92, 102, 112],
   [  3,  13,  23,  33,  43,  53,  63,  73,  83,  93, 103, 113],
   [  4,  14,  24,  34,  44,  54,  64,  74,  84,  94, 104, 114],
   [  5,  15,  25,  35,  45,  55,  65,  75,  85,  95, 105, 115],
   [  6,  16,  26,  36,  46,  56,  66,  76,  86,  96, 106, 116],
   [  7,  17,  27,  37,  47,  57,  67,  77,  87,  97, 107, 117],
   [  8,  18,  28,  38,  48,  58,  68,  78,  88,  98, 108, 118],
   [  9,  19,  29,  39,  49,  59,  69,  79,  89,  99, 109, 119],
   [ 10,  20,  30,  40,  50,  60,  70,  80,  90, 100, 110, 120]])

b is already in Fortran order

c=b.reshape(10,3,4, order='F')
print(c[:,:,0])
[[ 1 11 21]
 [ 2 12 22]
 [ 3 13 23]
 [ 4 14 24]
 [ 5 15 25]
 [ 6 16 26]
 [ 7 17 27]
 [ 8 18 28]
 [ 9 19 29]
 [10 20 30]]

Then I save the matrix c in a binary file:

c.T.tofile('test_c.bin')

So, using this Fortran code I am able to read the binary data in the correct order I created the c matrix in Python:

PROGRAM read_saved_python
IMPLICIT NONE

INTEGER(KIND=8),ALLOCATABLE :: matrix(:,:,:)
INTEGER :: Nx, Ny, Nz

Nx = 10
Ny = 3
Nz = 4


ALLOCATE(matrix(Nx, Ny, Nz))


OPEN(33, FILE="/home/victor/test_c.bin",&
     FORM="UNFORMATTED", STATUS="UNKNOWN", ACTION="READ", ACCESS='STREAM')

READ(33) matrix


write(*,*) matrix(:,1,1)


CLOSE(33)
DEALLOCATE(matrix)

END PROGRAM read_saved_python

Notice in Fortran the indexes start in 1 and the print shows in column order (in this case: print the first column, the second and then the third). If you don't transpose the matrix here c.T.tofile('test_c.bin') when reading in Fortran you'll notice that the matrix is not as you wanted, even if you use function np.asfortranarray as you did ( I even tried np.asfortranarray(c).T.tofile('/home/victor/teste_d.bin') (just to make sure) but the matrix is written in c order in the binary file.

BChow · Accepted Answer · 2016-06-20 14:19:40Z

1

You will need the meta data of the array to read it in FORTRAN. This website (https://scipy.github.io/old-wiki/pages/Cookbook/InputOutput.html) has some information on using libnpy to write and an example code fex.f95 to read the binary file.

answered Jun 20, 2016 at 14:19

BChow

4712 silver badges7 bronze badges

Collectives™ on Stack Overflow

Save numpy array as binary to read from FORTRAN

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related