Equivalent of numpy.where in Fortran

Help
1

In python:

import numpy as np
a=np.array([0, 10, 2, 3, 4, 51, 6, 7, 8, 9])
print (np.where(a<=5)[0])
#output: array([0, 2, 3, 4])

Is there an equivalent of “where” in Fortran?

2

PACK should work fine for this. A quick google search turned up this example, which should help you out.

1 Like
3

Thanks for pointing out this.

Here is the example:

module mod_mask
    implicit none

    contains
    
    subroutine where(arr, condition_scalar, indices)        
        real, intent(in) :: arr(:)
        real, intent(in):: condition_scalar
        integer, allocatable, intent(out) :: indices(:)
    
        integer:: i, n
        n = size(arr, dim=1)
        ! print*, "arr:", arr
        ! print*, "condition_scalar", condition_scalar
        indices = PACK([(i, i=1,n)], arr > condition_scalar)
        ! print*, "where:", indices
    end subroutine where

end module mod_mask

program maskk
    use mod_mask, only: where
    implicit none

    integer, allocatable :: indices(:)
    integer :: i !, n
    real, allocatable :: r(:)

    r = ([(i*1.1,i=10,15)])
    call where(r, 13., indices)
    print*, 'input array:', r
    print*, 'conditional scalar:', 13.
    print*, "indices:", indices
end program maskk
!
! input array:   11.0000000       12.1000004       13.2000008       14.3000002       15.4000006       !16.5000000    
! conditional scalar:   13.0000000    
! indices:           3           4           5           6
!
4

According to the NumPy documentation (numpy.where — NumPy v1.25 Manual), NumPy’s where is meant for merging entries as in

c = np.where(mask,a,b)

where c get values assigned from a and b depending on mask.

In Fortran, merge (MERGE (The GNU Fortran Compiler)) or where (https://www.intel.com/content/www/us/en/docs/fortran-compiler/developer-guide-reference/2023-1/where.html) can be used to achieve this.

5

My aim is to look for the indices of the input array where index values is <=5. Which numpy.where can do easily.

import numpy as np
a=np.array([0, 10, 2, 3, 4, 51, 6, 7, 8, 9])
print (np.where(a<=5)[0])
#output: array([0, 2, 3, 4])

And the same is done by the pack function [as shown in the example above] in Fortran. It seems that merge may not do the same as it selects values from two arrays as shown here merge function.

6

yes, but the pythonic way of doing this would be

import numpy as np
a=np.array([0, 10, 2, 3, 4, 51, 6, 7, 8, 9])
print (np.asarray(a<=5).nonzero()[0])
#output: [0, 2, 3, 4]

so if people search for the equivalent of np.where, merge or where are relevant Fortran functions.

1 Like