search.f90

!! 
!! f2py -c -m search search.f90
!! 


subroutine find_first(needle, haystack, haystack_length, index)
    !!
    !! Find the first index of `needle` in `haystack`.
    !!
    implicit none
    integer, intent(in) :: needle
    integer, intent(in) :: haystack_length
    integer, intent(in), dimension(haystack_length) :: haystack
!f2py intent(inplace) haystack
    integer, intent(out) :: index
    integer :: k

    index = -1
    do k = 1, haystack_length
        if (haystack(k)==needle) then
            index = k - 1
            exit
        endif
    enddo
end

Python code: https://gist.github.com/mverleg/3b63d70c806ea4eb73168870297d0bcb

The !f2py is important; without it the array will be copied and performance will drop to being similar to where.

Hi,
can you tell me why haystack_length is an optional parameter in python? The docs which are printed in python (via print(search.find_first.doc) ) say that that parameter is optional and with default = len(haystack), but I don't know where or how that is defined.
I'm not a fortran programmer but would like to start tinkering with algorithms to speed up my python applications.

Thanks in advance!

@hech92 It means that you should leave it empty in Python, and then in Fortran it will have as value the number of elements in the array "haystack". In Fortran the number of elements needs to be passed separately, it's not part of the array (or lets say it's complicated).

@mverleg Yes, that's what is confusing me. I'm calling the function via search.find_first(needle, array) but nowhere am I giving the length as a parameter. Is this handled internally in the python-fortran interface?

@hech92 Yeah Python knows this information, just Fortran doesn't, so Python can fill it in automatically before calling Fortran. And because it's always the same logic, it can be in the wrapper code so you don't have to do it.