How to template over different numerics?

[thrope]

Hi,

I would like to template over different numerical types (ie normal real, double precision, 8,16,32,64 bit integers etc.)

How do I pass these as classes to a template? mymodule<real> works but mymodule<real*8> or mymodule<real(8)> doesn't.

Thanks

[thrope]

I guess I will post my follow up question here… what I really want to do is template over numeric types/kinds and array ranks… so I would like to be able to do something like
template <class data_type, class rank>
function myfunc_(x)
data_type :: x(:,:,:)
end function
end template

where the data_type can be real*4, real*8, integer*4, integer*8, integer*2 etc and rank is used to define the arrays of the right rank… ie
rank=3 =>  data_type :: x(:,:,:)
rank=5 => data_type :: x(:,:,:,:,:)

How can this best be achieved with PyF95++?

Thanks for any help

[thrope]

Sorry for the noise - of course I can pass an integer kind to the template and use that in the declaration.

So the only remaining question is how to template expressions like

data_type(kind=data_kind) :: x(:,:,:)
over ranks?

[thrope]

Never mind I think I solved this one as well… at least the following seems to work using a POD:
  template <class data_type, class data_kind, class data_rank>
  function fpGetPr_(mx) result (fp)
    pod colons from mypods
     n = data_rank
     src = "data_type(kind=data_kind), pointer :: fp($cols$)"
    end pod colons
…

with mypods.py:

def colons(blksrc, globalDict):
    n = None
    src = None
    code = compile(blksrc, '', 'exec')
    exec code

    cols = n*':,'
    cols = cols
    block = src.replace('$cols$', cols)

    return block

This is really a great tool!