Re: [f2py] mixed f77/f90 code

On 11/24/2010 10:41 AM, Jasper Werhahn wrote:
> dear pearu,
>
> thanks a lot. i just don't understand one thing - i do of course only
> call fit_gm from python. but fit_gm itself continues to call the
> lmder-subroutines that actually do the fitting. how do i get this passed
> on then? greetings,

As I read the code, fit_gm uses gMorse as a callable argument to lmder 
subroutines. Since Morse_mod.f90 is linked to the generated extension 
module then all bits are there for fit_gm. May be I should clarify that
f2py builds wrappers to fortran subroutines, f2py does not convert 
fortran codes to python. A wrapper is a lightweight C function that
converts Python objects to C objects that can be passed on to Fortran 
subroutines.

Regards,
Pearu

> 2010/11/24 Pearu Peterson <[email protected]
> <mailto:[email protected]>>
>
>     Hi jaspeR,
>
>     The fcn issue that you saw is strictly speaking expected. When
>     generating wrappers to lmdif.f f2py needs to know the signature of
>     external fcn but it is neither provided nor available (not called within
>     the subroutine). Hence the failure. If you would provide the
>     signature then f2py would probably build the extension module
>     successfully.
>
>     However, it seems that you don't need to wrap all these fortran
>     subroutines to python. Assuming that only fit_gm is needed to be called
>     from python, I suggest the following workflow:
>
>     1) create pyf file containing signatures of only those functions that
>     will be used within python:
>
>        f2py  -m foo -h foo.pyf fit_gM.f90
>
>     1.2) optionally check foo.pyf to see if signatures needs adjustments
>     (specify input and output parameters using intent(in) or intent(out),
>     etc). See f2py users guide or f2py paper for more information.
>
>     2) build the extension module:
>     f2py -c foo.pyf *.f fit_gM.f90 Morse_mod.f90
>
>     3) check what was built:
>      >>> import foo
>      >>> print foo.__doc__
>     This module 'foo' is auto-generated with f2py (version:2_6882).
>     Functions:
>        x,chisqr,exit_status = fit_gm(file_name,data_length,alpha,gamma1)
>     .
>
>     HTH,
>     Pearu
>
>     On 11/24/2010 08:59 AM, jaspeR wrote:
>      > Pearu Peterson<pearu.peterson@...>  writes:
>      >
>      >>
>      >> Hi,
>      >>
>      >> Could you send a more complete example demonstrating the problem?
>      >> The failure may not be related to f77 and f90 cooperation...
>      >>
>      >> Pearu
>      >
>      > Hi Pearu!
>      >
>      > thanks for replying so fast once more... can i mail it to you? i
>     will include a
>      > makefile. the code dependencies are structured in the following way:
>      >
>      > Morse_mod.f90 contains the function subroutine required for the
>     fit, as well as
>      > all the functions necessary (that is, the function, and its
>     partial derivatives
>      > after the constants i am optimizing). The main program for the
>     standalone
>      > fortran compilation is called fit_scalable_pot.f90, for the
>     wrapping in f2py i
>      > reduced that to fit_gM.f90, excluding the terminal-output-based
>     command
>      > interface. the main program (fit_scalable_pot.f90) and the
>     reduced subroutine
>      > for f2py wrapping (fit_gM.f90) call the linpack subroutines. the
>     dependency
>      > tree is:
>      >
>      > lmder1.f
>      >    ◊--lmder.f
>      >        ◊--dpmpar.f,enorm.f,lmpar.f,qrfac.f
>      >                            ◊--dpmpar.f,enorm.f,qrsolv.f
>      >
>      > line 175ff. lists the dependencies of lmder.f in detail, same
>     goes for line
>      > 95ff. of lmpar.f. there might be some more cross-dependencies,
>     they can be
>      > found quickly with grep dpmpar *.f, for example. they are too
>     interwoven to
>      > give the whole dependency tree.
>      >
>      > thanks a lot, again. not only is f2py a great software, the
>     support in this
>      > mailing list is also superior! all the best,
>      >
>      > jaspeR
>      >
>      >
>      > ps: the original f90 main program fit_scalable_pot.f90 also
>     allows for fit to
>      > some other potentials. their module files (*_mod.f90) are
>     included for
>      > completeness to allow test compiling of the main program via the
>     makefile. for
>      > testing purposes, an example file 'li' is included.
>      >
>      >
>      >
>      >
>      > _______________________________________________
>      > f2py-users mailing list
>      > [email protected] <mailto:[email protected]>
>      > http://cens.ioc.ee/mailman/listinfo/f2py-users
>
>     _______________________________________________
>     f2py-users mailing list
>     [email protected] <mailto:[email protected]>
>     http://cens.ioc.ee/mailman/listinfo/f2py-users
>
>
>
>
> --
> Dipl.-Phys. Jasper Werhahn
> TU Muenchen, Physikdepartment E11
> James-Franck-Strasse
> 85748 Garching
> Germany
>
>
>
> _______________________________________________
> f2py-users mailing list
> [email protected]
> http://cens.ioc.ee/mailman/listinfo/f2py-users

_______________________________________________
f2py-users mailing list
[email protected]
http://cens.ioc.ee/mailman/listinfo/f2py-users