Matlab incorrectly reading a binary file

I need MatLab to read a binary file that has a specific format. The file was generated with Fortran and I know the exact format that the file takes. It has 12 integers in the header that describes the length of the rest of the file, which is double precision numbers. I can read the file with a Fortran code so I know the file has the correct values. I am having two issues with the code. The first is when I try to read the file there are 2 extraneous integer values that should not be there and do not show up when I read the file with Fortran. The second issue is when I try to read the double precision numbers none of the values are correct (ranging between the limits of the exponent +-308). The following is the bit of code I am trying to get working:

_ _fid=fopen('history.hs','r');

header_pe=fread(fid,14,'int32','ieee-be');

ct_i=abs(header_pe(8)-header_pe(7))+1;

ct_j=abs(header_pe(10)-header_pe(9))+1;

ct_k=abs(header_pe(12)-header_pe(11))+1;

ct_time=header_pe(3)*header_pe(2)-1;

fseek(fid,14*4,'bof');

x=fread(fid,[ct_i ct_j],'real*8','ieee-be');__

Any insight into why this would be happening are greatly appreciated.

Answer 1

[James Tursa]

Please show the Fortran code that was used to write the file, and also the Fortran code that correctly reads the file. Depending on the Fortran options you use, there can be extra bytes inserted in the binary file that MATLAB would have to account for. Those may be the "2 extraneous integer values" that you are referring to ... they really are in the file and the Fortran write statement put them there silently.

The exponent +-308 stuff often indicates that either a shift in where you need to start the reading, or that the bytes are swapped around (e.g. the file came from a different machine). The latter can often be fixed by using the swapbytes() function.

[Greg Heinlein]

This is the basic fortran code used to read the file. The write code is basically the same except that since it writes in big endian the only change is that the reads are changed to writes and the convert in the open statement is excluded. The constants used in the read statements below were set in the code (i.e. nj, nk, tmax, etc.) I just excluded them for brevity.

integer :: j,k,nj,nk

integer :: ct_pln,blkn,pmi,max_plns

integer :: istr,ien,jstr,jen,kstr,ken

integer :: id,nprt,ntc,cti,ctj,ctk,t,tmax

real*8,allocatable,dimension(:,:,:) :: x,y,z,ax,ay,az

real*8,allocatable,dimension(:,:,:,:) :: p,force

open(unit=id,file=name1,form='unformatted',convert='big_endian')

read(id)nprt,ntc,ct_pln,max_plns,blkn,istr,ien,jstr,jen,kstr,ken,pmi

read(id) ((x(1,j,k),j=1,nj),k=1,nk), &

((y(1,j,k),j=1,nj),k=1,nk), &

((z(1,j,k),j=1,nj),k=1,nk), &

((ax(1,j,k),j=1,nj-1),k=1,nk-1), &

((ay(1,j,k),j=1,nj-1),k=1,nk-1), &

((az(1,j,k),j=1,nj-1),k=1,nk-1)

do t=1,tmax

read(id) ((p(1,j,k,t),j=1,nj-1),k=1,nk-1)

enddo

close(id)