As was mentioned in the READONBLY thread pointers in F2000 will
can have an INTENT() attribute. What are the proposed rules?

I can think of three possible versions:
INTENT(IN OUT) and INTENT(OUT) I think are not controversial at all.
But INTENT(IN) allows several rules:

a) pointer status cannot be changed
   i.e., nullify, deallocate, another target are not allowed
   but the value of the target can be changed (if it exists)

b) the value of the target cannot be changed
   but the pointer can be pointed elsewhere

c) neither can the pointer be pointed elsewhere
   nor can the value of the target be changed

I personally prefer c) because it is the easiest and clearest version
though maybe too restrictive for others, while a) and b) have a
non-intuitive flavour.

Maybe we need a more detailed INTENT version for pointers?
One can certainly start with c) and relax later if that is needed.

NOTE that rule c) is NOT equivalent to passing to a non-pointer dummy
as this example shows:

With c) rule:
   subroutine foo_c( x )
     real, pointer, intent(in) :: x
     ...
   end subroutine foo_c

With a non-pointer dummy:
   subroutine foo_d( x )
     real, intent(in) :: x
     ...
   end subroutine foo_d

because one cannot check the association status of x in foo_c.
A more useful example is a standard dynamic data structure as
this queue (I coded this recently, that's why I am keen to know):

  type, private :: queue_node
       type(data)                :: dat
       type(queue_node), pointer :: next
  end type queue_node

  type :: queue
      private
      type(queue_node), pointer :: front, rear
  end type queue
  ! object-based style version of a queue, if you like


  function isEmpty( q ) result( Empty )

      type(queue), pointer :: q
      logical              :: Empty

      if( .not.associated(q) ) then
	  Empty = .true.
      else
	  Empty = .not.associated(q%front)
      end if

      return
  end function isEmpty

Here I could use (for safety, clarity) an INTENT(IN) for q where rule c)
seems most appropriate though a) is possible. But a) would allow to
allocate q%front, for example.

Cheers,
WWS
-----------------------------------------------------------------------
| Werner W Schulz                                                     |
| Dept of Chemistry                  email:     [log in to unmask]       |
| University of Cambridge            Phone:     (+44) (0)1223 336 502 |
| Lensfield Road                     Secretary:          1223 336 338 |
| Cambridge CB2 1EW                  Fax:                1223 336 536 |
| United Kingdom                     WWW:                             |
-----------------------------------------------------------------------




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%