Fortran bug: rank problem gfortran

[billiards]

This is my own code, and it won't compile with gfortran. All I want to do is extract the location of the cell with the minimum value in an array. A seemingly simple task but one that does not work with the intrinsic function minloc, for reasons I do not understand.

The error message:

    iclosept=minloc(dist)
    1
Error: Incompatible ranks 0 and 1 in assignment at (1)

The code:

  subroutine distance_to_trench(pt,tr,mindist,iclosept)
  ! find (great circle) distance from point to each point on the trench
  ! distance = the minimum distance to the trench
  
    type(point) :: pt
    type(trench) :: tr
    real, parameter :: pi=3.14159265359
    real, parameter :: eradius=6371
    real :: rad,deg
    real :: x1,y1,z1,x2,y2,z2
    real,dimension(:),allocatable :: dist
    real :: mindist
    integer :: ii, length, iclosept, a
    
    rad=pi/180
    deg=180/pi
    
    length=size(tr%poly%points%x)
    allocate(dist(length))
    
  ! method:
  ! convert lat lons to cartesian vectors
  ! use angle between vectors to find the distance
  
    x1=cos(rad*pt%y)*cos(rad*pt%x)
    y1=cos(rad*pt%y)*sin(rad*pt%x)
    z1=sin(rad*pt%y)
  
    do ii=1, length
      
      x2=cos(rad*tr%poly%points(ii)%y)*cos(rad*tr%poly%points(ii)%x)
      y2=cos(rad*tr%poly%points(ii)%y)*sin(rad*tr%poly%points(ii)%x)
      z2=sin(rad*tr%poly%points(ii)%y)
  
      dist(ii)=eradius*acos(x1*x2+y1*y2+z1*z2)
      
      print *,dist(ii)
    
    enddo
    
    mindist=minval(dist)
    
    a=size(dist)
    print *,'size of dist', size(dist), a

    iclosept=minloc(dist)
    print *,'minloc', minloc(dist), iclosept
    
    deallocate(dist)
    
  end subroutine distance_to_trench

 

[billiards]

Fixed it.

Turns out I was confused about the output of the minloc function. It returns an array of integer(s). I was assuming it returned an integer. Because I was only thinking about a 1-d array. But in the case of a 1-d array, it actually returns an array of size 1, where the element contains the integer I was after.

To fix my above code I introduced a temporary array and assigned it by minloc. Then I extracted the single element from it into a pure integer.

Thus:

  subroutine distance_to_trench(pt,tr,mindist,iclosept)
  ! find (great circle) distance from point to each point on the trench
  ! distance = the minimum distance to the trench
  
    type(point) :: pt
    type(trench) :: tr
    real, parameter :: pi=3.14159265359
    real, parameter :: eradius=6371
    real :: rad,deg
    real :: x1,y1,z1,x2,y2,z2
    real,dimension(:),allocatable :: dist
    real :: mindist
    integer :: ii, length, iclosept, temp(1)
    
    rad=pi/180
    deg=180/pi
    
    length=size(tr%poly%points%x)
    allocate(dist(length))
    
  ! method:
  ! convert lat lons to cartesian vectors
  ! use angle between vectors to find the distance
  
    x1=cos(rad*pt%y)*cos(rad*pt%x)
    y1=cos(rad*pt%y)*sin(rad*pt%x)
    z1=sin(rad*pt%y)
  
    do ii=1, length
      
      x2=cos(rad*tr%poly%points(ii)%y)*cos(rad*tr%poly%points(ii)%x)
      y2=cos(rad*tr%poly%points(ii)%y)*sin(rad*tr%poly%points(ii)%x)
      z2=sin(rad*tr%poly%points(ii)%y)
  
      dist(ii)=eradius*acos(x1*x2+y1*y2+z1*z2)
      
      print *,dist(ii)
    
    enddo
    
    mindist=minval(dist)
    temp=minloc(dist)
    
    iclosept=temp(1)
    
    deallocate(dist)
    
  end subroutine distance_to_trench

 

[jedishrfu]

When you see rank errors it means there's something wrong with the dimensions of some variable.

In you example iclosept is defined as an integer value (rank 0) but minloc returns an array of rank 1 (1D array) and that may be why it fails.

http://h21007.www2.hp.com/portal/download/files/unprot/Fortran/docs/lrm/lrm0293.htm

 

[billiards]

When you see rank errors it means there's something wrong with the dimensions of some variable.

In you example iclosept is defined as an integer value (rank 0) but minloc returns an array of rank 1 (1D array) and that may be why it fails.

http://h21007.www2.hp.com/portal/download/files/unprot/Fortran/docs/lrm/lrm0293.htm

Yep. That's the bug. Cheers.