How do I read a FORTRAN unformatted binary data file into MATLAB?

Below is a data file generated by a FORTRAN program. The following is the hexadecimal dump of the FORTRAN binary data file:

 0000000  0000 0048 4920 616d 2061 2067 656e 6975

 0000020  7320 6772 6f75 6e64 686f 6721 2020 5275

 0000040  6e21 2020 5275 6e21 2020 5275 6e21 2020

 0000060  2020 2020 2020 2020 2020 2020 2020 2020

 0000100  2020 2020 2020 2020 2020 2020 0000 0048

 0000120  0000 000c 3f8c cccd 400c cccd 4053 3333

 0000140  0000 000c 0000 0008 0000 0004 0000 0005

 0000160  0000 0008

A similar data file from C (flat binary file) is attached to this solution. Below is the code that was used to generate the binary data file.

program wrt_bin

character*72 name

real*4 xsource, zsource, dt

integer*4    ntrace, nt

name='I am a genius groundhog!  Run!  Run!  Run!'

xsource=1.1

zsource=2.2

dt=3.3

ntrace=4

nt=5

open(unit=11, file='wrt_bin.bin', form='unformatted',

       +     status='new')

    WRITE(11) NAME

    WRITE(11) XSOURCE,ZSOURCE,DT

    WRITE(11) NTRACE,NT

close(11)

end

Here is the C code that was used to generate the flat binary data file:

#include<stdio.h>

#include<stdlib.h>

void main()

{

   char name[72];

   int  ntrace, nt;

   double xsource, zsource, dt;

   FILE *fp;

   sprintf(name," I am a genius groundhog!  Run!  Run!  Run!");

   xsource=1.1;

   zsource=2.2;

   dt=3.3;

   ntrace=4;

   nt=5;

fp=fopen("wrt_binc.bin", "wb");

   fwrite(name, sizeof(char), 72, fp);

   fwrite(&xsource, sizeof(double), 1, fp);

   fwrite(&zsource, sizeof(double), 1, fp);

   fwrite(&dt, sizeof(double), 1, fp);

   fwrite(&ntrace, sizeof(int), 1, fp);

   fwrite(&nt, sizeof(int), 1, fp);

    fclose(fp);

}

Answer 1

[MathWorks Support Team]

Binary files that are generated by FORTRAN are quite different than ones that are expected by MATLAB. MATLAB expects a flat binary file. FORTRAN binary files are not flat.

UNIX: An unformatted FORTRAN binary file contains record length information along with each record. A record is a WRITE statement. Thus, the code above has 3 records. The record length information is tagged on at the beginning and end of each record. At the beginning of the FORTRAN binary data file is the record length "0000 0048". This is the record length in hex, i.e. 72 bytes. Counting 72 bytes from that point is another "0000 0048" which is followed by "0000 000c" (this is the length of the next record, i.e. 12 bytes or 3 times 4 bytes). The extraneous characters must either be filtered out or stripped off. The easiest thing to do would be to filter them out when you are reading the data into MATLAB. Stripping them off would require writing C code.

Here is the sequence of MATLAB commands that illustrates what you need to do:

fid=fopen('wrt_bin.bin', 'rb');        % Open the file.

hr1=fread(fid, 1, 'int32')             % Read the first record start tag. Returns hr1 = 72

name=setstr(fread(fid, 72, 'uchar'))'  % Read the first record.  Returns name = I am a genius                                        % groundhog!  Run!  Run!  Run!                              

hr1=fread(fid, 1, 'int32')             % Read the first record end tag. Returns hr1 = 72

hr2=fread(fid, 1, 'int32')             % Read the second record start tag. Returns hr2 = 12

xsource=fread(fid, 1, 'float32')       % Read data. Returns xsource = 1.1000

zsource=fread(fid, 1, 'float32')       % Read data. Returns zsource = 2.2000

dt=fread(fid, 1, 'float32')            % Read data. Returns dt = 3.3000

hr2=fread(fid, 1, 'int32')             % Read the second record end tag. Returns hr2 = 12

hr3=fread(fid, 1, 'int32')             % Read the third record start tag. Returns hr3 = 8

ntrace=fread(fid, 1, 'int32')          % Read data. Returns ntrace = 4

nt=fread(fid, 1, 'int32')              % Read data. Returns nt = 5

hr3=fread(fid, 1, 'int32')             % Read the third record end tag. Returns hr3 = 8

fclose(fid);                           % Close file.

You will have to modify your MATLAb file to read out the record length information and access the data in which you are interested. An alternative would be to use the FSEEK() command in MATLAB to jump over the record length information and access the data directly. Here is an example file that does so:

fid=fopen('wrt_bin.bin', 'rb'); 

fseek(fid, 4, 'cof'); 

name=setstr(fread(fid, 72, 'uchar'))';

fseek(fid, 4, 'cof'); 

fseek(fid, 4, 'cof'); 

xsource=fread(fid, 1, 'float32');

zsource=fread(fid, 1, 'float32'); 

dt=fread(fid, 1, 'float32'); 

fseek(fid, 4, 'cof');

fseek(fid, 4, 'cof'); 

ntrace=fread(fid, 1, 'int32'); 

nt=fread(fid, 1, 'int32');  

fclose(fid);

xsource, zsource, dt, ntrace, nt

Here is the output:

 xsource = 1.1000  zsource = 2.2000  dt =  3.3000  ntrace =  4    nt = 5

To apply the MATLAB code to a binary file created with Digital Visual Fortran and Windows requires a few modifications. Since Digitial Fortran writes two unsigned characters as the start tag and has no end tag, you need code such as that below in order to read a binary file produced by Digital Fortran code:

tag=setstr(fread(fid,2,'uchar')); %read the first record start tag

capf=fread(fid,1,'float');  %capture fraction

rot=fread(fid,1,'float');  %grid rotation, radians

tag=setstr(fread(fid,2,'uchar')); %read the second record start tag

nx=fread(fid,1,'int'); %number cells alongwind

xgrid=fread(fid,nx,'float'); %alongwind grid points, m

tag=setstr(fread(fid,2,'uchar')); %read the third record start tag

ny=fread(fid,1,'int'); %number cells crosswind

ygrid=fread(fid,ny,'float'); %crosswind grid points, m

tag=setstr(fread(fid,2,'uchar')); %read the fourth record start tag

dose=fread(fid,nx*ny,'float32'); %read the dose matrix

tag=setstr(fread(fid,2,'uchar')); %read the fifth record start tag

dep=fread(fid,nx*ny,'float32'); %read the deposition matrix

NOTE: The Digital Visual Fortran information was submitted by one of our customers and has not been verified by MathWorks.