|
forums.silverfrost.com Welcome to the Silverfrost forums
|
View previous topic :: View next topic |
Author |
Message |
Anonymous Guest
|
Posted: Wed Apr 06, 2005 11:20 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
Hi,
I'm sorry, but I have also posted this on the general forum.
I hope this is the correct place to post it.
I'm using the ft95 personal edition compiler for one month or so.
I don't have access to the compiler error reference manual ( I think this do not comes with the distribution),
so I must guess why I receive some errors at compilation.
The compiler is working fine comparing to g95 or gfortran or intel, but there is some errors I get like this one
In a module for rational arithmetic I created interfaces like
type rational
sequence
integer :: n
integer :: d
end type rational
interface min
module procedure min_Q
end interface
....
interface abs
module procedure abs_Q
end interface
....
contains
...
function abs_Q(x) result (z)
type(rational), intent(in) :: x
type(rational) :: z
z=...
end function abs_Q
The compiler returns the following error:
error 435 - Specific procedure ABS of type REAL(KIND=1) FUNCTION is too similar to ABS
of type REAL(KIND=1) FUNCTION for overload ABS in module INTEGER_ARITHMETIC_CLASS. The arguments are too similar.
I should say that, in general, overloadings of operators and standard functions are working fine (for example, min, max, ...)
In fact I have a module for integer arithmetic which is used by the rational arithmetic module and where I have made a lot of overloadings without any problem, including overloadings of min, max, abs, ...
With respect with the abs function I did noticed that there is no abs function for Integer( arguments. At least I get
the following compilarion errors
for a code like this
.....
integer( u,v
....
u=abs(v)
....
error 137 - Operand incompatible with opcode
fatal 123 - Invalid machine instruction generated
Therefore I overloaded abs for integer( arguments with a function of my own in the integer-artithmetic_module, and this works
Finally one last error I got:
In a module for vector operations I created the type
type Vector
private
integer :: size ! vector length
real, pointer, dimension( :: value ! component values
end type Vector
I have a constructor which receives an array of values and should
return a variale of my type Vector
function as_Vector (length, values) result(z)
integer, intent(in) :: length
real, target, intent(in) :: values(length)
real, pointer :: pointer_to_vaues(
type (Vector) :: z
integer :: flag ! allocate flag
allocate ( pointer_to_val (length), stat = flag )
if (flag /= 0 ) stop 'allocate error'
pointer_to_values => values
z = Vector (length, pointer_to_values) ! ******
end function as_Vector
I obtain the following compilation error pointing to the line with ******
error 621 - The type constructor for TYPE(VECTOR) has not enough arguments
I fact it seems that the intrinsic constructor Vector(.,.) does not work.
I obtain the same compilation error when trying to call it from the main program.
Can any one help me on these problems?
Thank you in advance.
Carlos
cmmenez
cmmenez |
|
Back to top |
|
|
PaulLaidler Site Admin
Joined: 21 Feb 2005 Posts: 7932 Location: Salford, UK
|
Posted: Fri Apr 08, 2005 1:54 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
The first problem you mention (error 435) may well be due to a bug in the compiler.
The second (error 137) is certainly a bug.
The third (error 621) again looks like a potential bug.
You have supplied sufficient information for me the check how FTN95 responds to passing a pointer in a constructor (error 621). For the other two I need two very short programs that illustrate each problem. If you can post these I will investigate them as soon as possible. |
|
Back to top |
|
|
PaulLaidler Site Admin
Joined: 21 Feb 2005 Posts: 7932 Location: Salford, UK
|
Posted: Fri Apr 08, 2005 2:21 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
I can now confirm that error 621 represents a bug in FTN95.
There is also an inconsistency in your code.
You should have either
allocate(pointer_to_values(length))
pointer_to_values = values
if you want to copy the values or just
pointer_to_values => values
without ALLOCATE, if you want to point to the original values.
|
|
Back to top |
|
|
Anonymous Guest
|
Posted: Fri Apr 08, 2005 8:48 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
Here is some code where the other two errors occurs. Sorry if this is not sufficientrly short.
I have not modified substantially the code, but this is only a small fraction of the whole code.
For me Integer(JKIND) and integer(LKIND) is just, as you can see,
integer(SELECTED_INT_KIND(1)
The errorrs (for the ABS overloading) occurs with this code.
The error for the inexistence of ABS function for integer(JKIND)
occurs with this code if you "UNINTERFACE" function longinteger_abs_J
Thank you for your support.
Carlos
module standard_numbers
implicit none
integer, parameter :: IKIND=KIND(1)
integer, parameter :: JKIND=SELECTED_INT_KIND(1
integer, parameter :: LKIND=SELECTED_INT_KIND(1
integer, parameter :: DKIND=KIND(1.d0)
integer, parameter :: RKIND=KIND(1.e0)
logical, parameter :: TRUE=.TRUE.
logical, parameter :: FALSE=.FALSE.
end module standard_numbers
module Integer_Arithmetic_Class
!---------------------------------------------------------------------------------------------------
use standard_numbers
implicit none
!-----------------------------------------------------------------------------------------------------
type longinteger
sequence
! integer(IKIND),public :: nmax !maximum words allocated
integer(IKIND),public :: n !number of words used
integer(IKIND),public :: sign ! sign
integer(IKIND),public :: digit(0:40) !number of words
end type longinteger
interface abs
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
module procedure longinteger_abs_J
!!!! UNINTERFACE THIS TO GET ERROR 137
end interface
interface gcd
module procedure integer_gcd_JJ
end interface
contains
function longinteger_abs_J(x) result(z)
integer(JKIND),intent(in) :: x
integer(JKIND) :: z
z=x
if (z < 0) z= -z
end function longinteger_abs_J
function integer_gcd_JJ(x,y) result(d)
integer(JKIND) x,y,d
integer(JKIND) u,v,t
u=abs(x)
v=abs(y)
if(u>v) then
t=u
u=v
v=t
end if
if (v == 0)then
v=1
else
while (u>0) do
t=u
u=mod(v,t)
v=t
end do
end if
d=v
end function integer_gcd_JJ
end module Integer_Arithmetic_Class
module Rational_Arithmetic_Class
use Integer_Arithmetic_Class
implicit none
!---------------------------------------------------------------------------
!---------------------------------------------------------------------------
! Define a derived type called rational that contains two integer(LKIND)
! components, n and d.
! n is the numerator and d is the denominator
type rational
sequence
integer(LKIND) :: n
integer(LKIND) :: d
end type rational
interface abs
module procedure rational_abs_Q
end interface
interface min
module procedure rational_min_QQ
end interface
contains
function rational_abs_Q (x) result (z)
type(rational),intent(in) :: x
type(rational) ::z
z%n=abs(x%n)
z%d=abs(x%d)
end function rational_abs_Q
function rational_min_QQ(x,y) result (z)
type(rational),intent(in) :: x
type(rational),intent(in) :: y
type(rational) :: z
if(x%n * y%d - x%d*y%n < 0) then
z=x
else
z=y
end if
end function rational_min_QQ
end module Rational_Arithmetic_Class
Cmmenez |
|
Back to top |
|
|
PaulLaidler Site Admin
Joined: 21 Feb 2005 Posts: 7932 Location: Salford, UK
|
Posted: Mon Apr 11, 2005 5:25 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
There is only limited support for 64 bit integers in FTN95.
If you can manage with SELECTED_INT_KIND(9) then two of your problems will be solved.
As a work-around for the constructor bug you could use simple assignments such as:
v%size = 4
v%value => array |
|
Back to top |
|
|
PaulLaidler Site Admin
Joined: 21 Feb 2005 Posts: 7932 Location: Salford, UK
|
Posted: Mon Apr 11, 2005 9:38 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
This bug has now been fixed.
The current release is FTN95 version 4.80.
The fix will be available in the next release or service pack after 4.80. |
|
Back to top |
|
|
Anonymous Guest
|
Posted: Mon Apr 11, 2005 11:56 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
Thank you.
I suppose I can download the new version of ftn95.
As for the problem with the abs overloading, I will try to use only integer(4) variables instead of integer(,
to see if the overloading of abs for rational numbers (pairs of integer variables) will work.
May be tomorow.
Carlos.
cmmenez |
|
Back to top |
|
|
Anonymous Guest
|
Posted: Tue Apr 12, 2005 1:33 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
If the answer to the following question is YES, may be you don't want
to read what follows it. Otherwise please read what I have found.
IS IT A RULE (FROM THE FORTRAN 95 STANDARDS OR OF FTN95 DESIGN)
THAT AN INTRINSIC FUNCTION LIKE "ABS" CANNOT BE OVERLOADED IN TWO DIFFERENT MODULES?
!
! The following code do not produce any compilation error, although
! the integer variables n and d that define the type rational in module Rational_Arithmetic_Class
! are of selected_int_kind(1.
!
! However if you move the functions longinteger_abs_J or longinteger_sign_J or exp1 (which for now
! are in the module Rational_Arithmetic_Class ) to the
! module Integer_Arithmetic_Class, uncomment the interface code in this last module, and
! comment the interface to those routines in the Rational_Arithmetic_Class module
! you obtain the error I have reported before:
!
!
!error 435 - Specific procedure SIGN of type REAL(KIND=1) FUNCTION is too similar to SIGN of type REAL(KIND=1) FUNCTION for overload SIGN in module INTEGER_ARITHMETIC_CLASS. The arguments are too similar.
!error 435 - Specific procedure ABS of type REAL(KIND=1) FUNCTION is too similar to ABS of type REAL(KIND=1) FUNCTION for overload ABS in module INTEGER_ARITHMETIC_CLASS. The arguments are too similar.
!error 1044 - This generic call to ABS is ambiguous as both of the specifics ABS and LONGINTEGER_ABS_J could match the arguments as supplied
!error 1044 - This generic call to ABS is ambiguous as both of the specifics ABS and LONGINTEGER_ABS_J could match the arguments as supplied
!
! Please notice that there are two overloads of ABS (resp. SIGN, and EXP)
! One for longinteger_abs_J, and another for rational_abs_Q (resp. ....).
! The second overload must be in the module RATIONAL_ARITHMETIC_CLASS, and the proper place for the first
! overload is in the module INTEGER_ARITHMETIC_CLASS.
! The code as I present makes the two overloadings in the module
! RATIONAL_ARITHMETIC_CLASS and do not produce a compilation error.
! However if you make the changes I referred to above (one overload in each module).
! the error occurs.
!
! So I believe that the error 425 I'm being reporting is not caused by the integer(selected_int_kind(1),
! because with the above modifications the same error occurs
! --- if you change the fields n and d
! in the definition of type(rational) to integers of selected_int_kind(9),
! --- or if you modify longinteger_abs_J to have an argument and result
! of some other type you have created,
! for instance my
! type longinteger
! which is defined in module INTEGER_ARITHMETIC_CLASS
!
! In fact I believe that the problem is related to some names clash for ABS and SIGN and EXP,...
! when these functions are overloaded in different modules (perhaps only in the case where one uses
! one module within another and overloads on both modules the function ABS or SIGN or EXP,
! which is my case).
!
! In order to test if my conjecture is correct I created two new types, type T1 in module
! INTEGER_ARITHMETIC_CLASS and type T2 in module RATIONAL_ARITHMETIC_CLASS and decided to
! overload intrinsic function EXP with functions EXP1 with argument and result of type T1
! function EXP2 with argument and result of type T2 with some "nonsensical" code.
! When both functions overload EXP in the module RATIONAL_ARITHMETIC_CLASS (as is the present code)
! there is no compilation error. If I overload EXP with EXP1 in the INTEGER_ARITHMETIC_CLASS module
! and EXP with EXP2 in the RATIONAL_ARITHETIC_CLASS module the following error occurs
!
! error 435 - Specific procedure EXP of type REAL(KIND=1) FUNCTION is too similar to EXP of type REAL(KIND=1) FUNCTION for overload EXP in module INTEGER_ARITHMETIC_CLASS. The arguments are too similar.
!
! Therefore it seems that most intrinsic functions cannot be overloaded in two different modules.
!
! However notice that I have overloaded MIN and MAX (may be with |
|
Back to top |
|
|
PaulLaidler Site Admin
Joined: 21 Feb 2005 Posts: 7932 Location: Salford, UK
|
Posted: Tue Apr 12, 2005 2:42 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
I don't know of any constraint in the Standard that restricts the various overloads to one module.
It should work. However, it is probably simpler and tidier to put them all in the same module. |
|
Back to top |
|
|
Anonymous Guest
|
Posted: Wed Apr 13, 2005 5:13 am Post subject: operator overloading, function overloading, intrinsic constr |
|
|
I disagree with you. Sometimes it can be simpler and tidier. Most of times it is not, and forbids the construction of
"sub" modules for stand alone usage. For instance if module B uses module A, and the overloads are declared only on module B
one cannot use only module A with its own overloads unless one recodify them in module A. So one
must have two versions of module A. One for stand alone usage and one to be used by module B.
I think this is not good practice and it is error and version-problem prone.
Thank you for your attention.
Carlos
cmmenez |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|