how to read an outputted fortran binary NxNxN matrix into Python

Question

I wrote out a matrix in Fortran as follows:

real(kind=kind(0.0d0)), dimension(256,256,256) :: dense

[...CALCULATION...]

inquire(iolength=reclen)dense
open(unit=8,file=fname,&
form='unformatted',access='direct',recl=reclen)
write(unit=8,rec=1)dense(:,:,:) 
close(unit=8)

I want to read this back into Python. Everything I've seen is for 2D NxN arrays not 3D arrays. In Matlab I can read it as:

fid =    fopen(nfilename,'rb');
mesh_raw = fread(fid,ndim*ndim*ndim,'double');
fclose(fid);
mesh_reshape = reshape(mesh_raw,[ndim ndim ndim]);

I just need the equivalent in Python - presumably there is a similar load/reshape tool available. If there is a more friendly compact way to write it out for Python to understand, I am open to suggestions. It will presumably look something this: . I am just unfamiliar with the equivalent syntax for my case. A good reference would suffice. Thanks.

struct.unpack seems like the way to go but I'm not sure what to do for my case. — Griff
– Griff, Commented Dec 11, 2012 at 19:57
You will find the solution here: stackoverflow.com/questions/10475839/… — milancurcic
– milancurcic, Commented Dec 11, 2012 at 20:10

Griff · Accepted Answer · 2012-12-11 20:51:54Z

8

Using IRO-bot's link I modified/made this for my script (nothing but numpy magic):

def readslice(inputfilename,ndim):
    shape = (ndim,ndim,ndim)
    fd = open(fname, 'rb')
    data = np.fromfile(file=fd, dtype=np.double).reshape(shape)
    fd.close()
    return data

I did a mean,max,min & sum on the cube and it matches my fortran code. Thanks for your help.

answered Dec 11, 2012 at 20:51

Griff

2,1345 gold badges32 silver badges52 bronze badges

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

3 Comments

milancurcic Over a year ago

Just make sure you take into account array dimension ordering difference between Fortran (column major) and Python (row major).

Griff Over a year ago

Isn't there a command to tell it to reorder as fortran at the end of .reshape(shape)?

milancurcic Over a year ago

In the case of Fortran array being declared as dimension(im,jm,km), you would want to read it from Python as np.fromfile(file=fd, dtype=np.double).reshape((km,jm,im)). In the case of im=jm=km, you don't need any extra steps, but remember that last index varies fastest.

PearsonArtPhoto · Accepted Answer · 2012-12-11 20:40:41Z

0

I can't see anything but a direct read working here. Python doesn't do a great job of 2-D arrays, let alone 3-d, but this bit of code should work.

fin=open('filename.dat','rb')
output=[]
for x in range(0,ndim):
    xarr=[]
    for y in range(0,ndim):
        yarr=[]
        for z in range(0,ndim):
            yarr.append(struct.unpack('i', fin.read(4)))
        xarr.append(yarr)
   output.append(xarr)

answered Dec 11, 2012 at 20:40

PearsonArtPhoto

39.9k17 gold badges116 silver badges147 bronze badges

Collectives™ on Stack Overflow

how to read an outputted fortran binary NxNxN matrix into Python

2 Answers 2

3 Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related