Good Fortran90 programming practices
We present some proper Fortran90 programming practices. By adopting them by programmers one could be working with readable and clean code.
Module names and file names
Use the same name for the module and the file. For instance, if your module is called foobar, call the file foobar.F90. This makes it easier to find the source code linked by “use foobar”.
It is possible to group several modules into one file. Do it when your modules share some common characteritics. Modules grouping in files prevents flood of individual module source files slowing down the repository connections.
Example module
module nice_module
! it is good to list only what you will use
! this keeps things nicely separated
use some_module, only: some_routine
! always use this
! it will be valid module-wide
implicit none
public nice_module_init
public nice_module_finalize
public nice_module_do_something
! everything else is private
private
! if false the module will refuse to be accessed
logical :: is_initialized = .false.
! this array is visible to everything below "contains"
! but not visible outside
! it is not a good idea to make it visible to outside
! rather access through get/set functions that first check whether
! the array is allocated
real(8), allocatable :: some_array(:)
contains
subroutine nice_module_init(n)
integer, intent(in) :: n
allocate(some_array(n))
is_initialized = .true.
end subroutine
subroutine nice_module_finalize()
! this routine should reset everything set by this module
! and deallocate all allocated memory
deallocate(some_array)
is_initialized = .false.
end subroutine
subroutine check_if_module_is_initialized()
if (.not. is_initialized) then
print *, 'error: you try to access nice_module'
print *, ' but this module is not initialized'
stop 1
end if
end subroutine
subroutine nice_module_do_something()
! first check whether the module is initialized
call check_if_module_is_initialized()
! only now continue with real work
! ...
end subroutine
end module
Include files and common blocks in the F90 era
Please do not introduce new common blocks. When you write new code and use F90, there is no need to introduce new common blocks (and include files) - use modules instead, they are much safer. Common blocks make the code harder to maintain.
- Sometimes you have to use already present include files in F90. To make this work:
only use “!” to start comments in include files (all F77 compilers understand this)
continuation use a & in column 73 of the first line and a & (nothing else!) in column 6 of the following line.
- This will work with both free form and fixed form when the file is included)
declare all variables explicitly, so that they can be used with the IMPLICIT NONE declaration.
There is a Python-script reformat_includes.py in the utils directory, that reformats include files according to these rules.
I/O control statements
To ease the localization of input/output-type bugs during the code run please provide main I/O statements (OPEN, READ) with the IOSTAT parameter evaluation or with other error-trapping programing scheme.
Examples:
!... control reading if gaunt=true
READ(LUCMD,*,IOSTAT=IOS) ILLINT,ISLINT,ISSINT,IGTINT
IF (IOS.NE.0) THEN
WRITE(LUPRI,'(/,2X,A)') 'Error in SCF INTFLG reading !'
& //'4 parameters needed for Gaunt term !'
CALL FLSHFO(LUPRI)
CALL QUIT( 'DHFINP: Error in INTFLG reading for GAUNT=true!')
ENDIF
or (assuming that LU is correct unit number:
WRITE(LU,desired_format,IOSTAT=IOS) SOMETHING
IF (IOS.NE.0) THEN
! Yell, that some variables in "SOMETHING" are out of desired format-bounds ;
! this could indicate run-time error (getting variable(s) hurt before)
! might activate program stop (CALL QUIT)....
ENDIF
F90 INCLUDE keyword
Do not use it. Use #include preprocessor statements only.
SIXLTR SBRTNE NAMING OBSOLT
Because it is so hard to read and understand. There is no need to to stick to 6 letter names anymore. The limit is 31 characters. Use meaningful names so that other people (and you in a couple of months) will understand your code.
Dynamical strings
The programmer may utilize short dynamical strings carrying descriptive labels, and attach this data type to his own data structures.
The module is
use mystring
Methods at hand are:
creating the string
type(string) :: this_string
call make_string(this_string,'some description text')
printing the string on the LUPRI channel
call print_string(this_string)
copying one string into another
call copy_string(this_string, another_string)
canceling (deallocating) the string
call cancel_string(this_string)
For more details about the dynamical string functionalities check directly
the dynamic_string.F90
source file.
Code debugging with LINEFILE
C-debugging symbols are defined by the preprocessor on each line. Therefore it is useful to insert printouts of file name and line number:
print *, 'I am in file ',__FILE__, ' on line ', __LINE__
Now, of course we don’t want to write that everywhere. What one can do is to use a macro:
#define MYFUNCTION(x) myfunction(x,__FILE__,__LINE__)
This allows ‘’’myfunction’’’ to print good error messages, because it knows the value of __FILE__ and __LINE__. But the macro has to be defined somewhere, so you need an ‘’’#include “mymem.h”’’’ or something.
Here is an example that works, just for using __FILE__ and __LINE__ statements:
subroutine mysub(someargument,file,line)
character(*) file
integer someargument, line
print *, 'mysub(',someargument,')'
print *, 'I was called in file "',file,'" at line ',line
end subroutine mysub
#define mysub_wrapped(arg) mysub(arg,__FILE__,__LINE__)
program test
integer x
x = 12
call mysub_wrapped(x)
end program test
For large scale employment one has to put the ‘’’#define’’’ in an include file, or better, in a Fortran90 module.