M_CLI2(3f) is a Fortran module that will crack the command line when given a prototype string that looks very much like an invocation of the program. Calls are then made for each parameter name to set the variables appropriately in the program. One approach is isolate all the parsing to the beginning of the program, which is generally just a few lines:
program compartmentalized
use M_CLI2, only : set_args, get_args, s=>sget, r=>rget, i=>iget, l=>lget
implicit none
!
! define command options and default values and parse command line
call set_args('-x 1 -y 2.0 -i 11 --title:T "my title" -l F -L F')
! just like calling the command accept --title:T means to give it the long
! name "title" and the short name "T" and that to define a boolean you give
! it the unquoted value of F.
!
! convert arguments to the desired types and call main program
call main( x=r('x'), y=r('y'), title=s('title'), i=i('i'), l=l('l'), lbig=l('L'))
contains
subroutine main(x,y,title,i,l,lbig)
! do something with the values, all the parsing is done
real :: x,y ;namelist /args/x,y
logical :: l,lbig ;namelist /args/l,lbig
integer :: i ;namelist /args/i
character(len=:),allocatable :: title ;namelist /args/title
write(*,nml=args)
end subroutine main
end program compartmentalized
This short program defines a command that can be called using conventional Unix-style syntax for short and long parameters:
# arrays can be allowed like for "-p":
./show -x 10 -y -20 -p 10,20,30 --title "plot of stuff" -L
./show -lL # in strict mode booleans may be concatenated
./show --title="my new title" # --name=value or --name value is OK
./show -T "my new title" # a short name for "title"
program show
use M_CLI2, only : set_args, get_args, set_mode
use M_CLI2, only : sget, rget, iget, lget
use M_CLI2, only : sgets, rgets, igets, lgets
implicit none
real :: x,y,z
logical :: l, lbig
integer,allocatable :: p(:)
character(len=:),allocatable :: title
namelist /args/x,y,z,l,lbig,p,title ! just for printing
call set_mode('strict')
!
! Define command and default values and parse supplied command line options
call set_args('-x 1 -y 2.0 -z 3.5e0 -p 11,-22,33 --title:T "my title" -l F -L F')
! use convenience functions for scalar values or allocatable arrays.
! The functions are particularly useful in expressions and as arguments on
! procedure calls:
x=rget('x') ! float
y=rget('y')
z=rget('z')
title=sget('title') ! string
p=igets('p') ! integer array
l=lget('l') ! logical
lbig=lget('L')
! All ready to go, print it as a namelist so everything is labeled
write(*,args)
!
! Alternatively: use get_args directly instead of via the convenience routines:
!
! multiple scalar non-allocatable values can be done in one call if desired
call get_args('x',x,'y',y,'z',z,'l',l,'L',lbig)
!
! allocatable arrays and allocatable string lengths need called by themselves
call get_args('title',title)
call get_args('p',p)
!
! All ready to go, print it as a namelist so everything is labeled
write(*,args)
end program show
running with no options shows the defaults
&ARGS
X= 1.00000000 ,
Y= 2.00000000 ,
Z= 3.50000000 ,
L=F,
LBIG=F,
P=11 ,-22 ,33 ,
TITLE="my title",
/
An arbitrary number of strings such as filenames may be passed in on the end of commands; you can query whether an option was supplied; and get_args(3f)-related routines can be used for refining options such as requiring lists of a specified size.
These parameters are defined automatically
--help
--usage
--verbose
--version
You must supply text for the optional "--help" and "--version" keywords, as described under SET_ARGS(3f).
The 3.2.0 release of the command-line parser module M_CLI2 has a standalone program available that will display the help text for the procedures as a substitute for the man(1) pages.
If the program is placed in your search path you can enter
fpm-m_cli2 --help
# if an fpm user
fpm m_cli2 --help
for a description of usage. An example to build it on a typical Linux platform would be
# create a scratch directory for the build
mkdir temp
cd temp
# get the documentation program
curl https://raw.githubusercontent.com/urbanjost/index/main/bootstrap/fpm-m_cli2.f90
# compile the program
gfortran fpm-m_cli2.f90 -o fpm-m_cli2
# copy it to somewhere in your path
mv fpm-m_cli2 $HOME/.local/bin/
Compile the M_CLI2 module and build all the example programs.
git clone https://github.com/urbanjost/M_CLI2.git
cd M_CLI2/src
# change Makefile if not using one of the listed compilers
# for gfortran
make clean
make gfortran
# for ifort
make clean
make ifort
# for nvfortran
make clean
make nvfortran
# display other options (test, run, doxygen, ford, ...)
make help
To install you then generally copy the .mod file and .a file to an appropriate directory. Unfortunately, the specifics vary but in general if you have a directory $HOME/.local/lib and copy those files there then you can generally enter something like
gfortran -L$HOME/.local/lib -lM_CLI2 myprogram.f90 -o myprogram
There are different methods for adding the directory to your default load path, but frequently you can append the directory you have placed the files in into the colon-separated list of directories in the $LD_LIBRARY_PATH or $LIBRARY_PATH environment variable, and then the -L option will not be required (or it's equivalent in your programming environment).
export LD_LIBRARY_PATH=$HOME/.local/lib:$LD_LIBRARY_PATH
NOTE: If you use multiple Fortran compilers you may need to create a different directory for each compiler. I would recommend it, such as $HOME/.local/lib/gfortran/.
If you desire a shared library as well, for gfortran you may enter
make clean gfortran gfortran_install
and everything needed by gfortran will be placed in libgfortran/ that you may add to an appropriate area, such as $HOME/.local/lib/gfortran/.
make clean ifort ifort_install # same for ifort
does the same for the ifort compiler and places the output in libifort/.
NOTE: The build instructions above are specific to a ULS (Unix-Like System) and may differ, especially for those wishing to generate shared libraries (which varies significantly depending on the programming environment). For some builds it is simpler to make a Makefile for each compiler, which might be required for a more comprehensive build unless you are very familiar with gmake(1).
If you always use one compiler it is relatively simple, otherwise make sure you know what your system requires and change the Makefile as appropriate.
Alternatively, fpm(1) users may download the github repository and build it with fpm ( as described at Fortran Package Manager )
git clone https://github.com/urbanjost/M_CLI2.git
cd M_CLI2
fpm test # build and test the module
fpm install # install the module (in the default location)
or just list it as a dependency in your fpm.toml project file.
[dependencies]
M_CLI2 = { git = "https://github.com/urbanjost/M_CLI2.git" }
Alternatively, meson(1) users may download the github repository and build it with meson ( as described at Meson Build System )
git clone https://github.com/urbanjost/M_CLI2.git
cd M_CLI2
meson setup _build
meson test -C _build # build and test the module
# install the module (in the <DIR> location)
# --destdir is only on newer versions of meson
meson install -C _build --destdir <DIR>
# older method if --destdir is not available
env DESTDIR=<DIR> meson install -C _build
or just list it as a subproject dependency in your meson.build project file.
M_CLI2_dep = subproject('M_CLI2').get_variable('M_CLI2_dep')
This is how the interface works --
Pass in a string to set_args(3f) that looks almost like the command you would use to execute the program except with all keywords and default values specified.
you add calls to the get_args(3f) procedure or one of its variants. The alternative convenience procedures (rget(3f),sget(3f),iget(3f) ...) allow you to use a simple function-based interface model. There are special routines for when you want to use fixed length. CHARACTER variables or fixed-size arrays instead of the allocatable variables best used with get_args(3f)).
Now when you call the program all the values in the prototype should be updated using values from the command line and queried and ready to use in your program.
These demo programs provide templates for the most common usage:
Response files are supported as described in the documentation for set_args. They are a system-independent way to create short abbreviations for long complex commands. This option is generally not needed by programs with just a few options, but can be particularly useful for programs with dozens of options where various values are frequently reused.
commit 598e44164eee383b8a0775aa75b7d1bb100481c3
was tested on 2020-11-22 with
commit 8fe841d8c0c1867f88847e24009a76a98484b31a
was tested on 2021-09-29 with
commit 732bcadf95e753ccdf025cec2c08d776ea2534c2
was tested on 2023-02-10 with
Last update: Saturday, February 4th, 2023 1:12:54 AM UTC-05:00