Currently, I am using shell command line calls from my fortran program using non-standard SYSTEM intrinsic routine (similar to Fortran 2008 EXECUTE_COMMAND_LINE intrinsic):
CALL SYSTEM(commandStr)
where commandStr is a character string containing the shell command I want to execute. At the moment, I am not aware of a direct way to return the output of commandStr, but only its return status. So, what I am doing now is writing output into a file, and then reading the file from within the Fortran program. Example:
CALL SYSTEM('sed ''s/,//g'' myFile > dummyFile')
if I want to remove commas from myFile. I then use OPEN and READ to get contents of dummyFile.
This works just fine, however I am concerned about writing/reading files from disk, especially if I was doing this within a long loop, and if commandStr output was big. Is there a way to re-direct commandStr output into a memory buffer (not hard disk) which I could access from my Fortran program directly (maybe through a specific UNIT number)?
If this is in a POSIX environment, the library function
popen()
might also be available.Look at the documentation for your Fortran environment since I'm not sure of the semantics for connecting the i/o to Fortran. If it is like the C runtime library, the file connection also needs a special function to close it,
pclose()
.