Fortran [Fortran] Function implicit type error

[avikarto]

I am having an issue with the declared type of the return value of a function not being recognized in its call.

This function clearly has its return value declared as complex*16:

!----------4-vector dot product under Minkowski metric----------
function dot4(v1,v2) result(res)
    implicit none
    complex*16 v1(4),v2(4)
    complex*16, dimension(0):: res
   
    res=v1(1)*v2(1)-v1(2)*v2(2)-v1(3)*v2(3)-v1(4)*v2(4)
    return
end

But then when I try to call dot4 from another function as such:

!----------laser-dressed 4-vec----------
function bar(v,k,mu) result(res)
    implicit none
    complex*16 v(4),k(4)
    complex*16, dimension(4):: res
    real*8 me,c, mu
   
    me=9.10938291E-31 !electron rest mass, in kg
    c=137.0 !speed of light, in au
   
    res=v+((mu*me*c)**2)/(4*dot4(k,v))*k
    return
end

it throws the following error on compiling (g95):

res=v+((mu*me*c)**2)/(4*dot4(k,v))*k
.........1
Error: function 'dot4' at (1) has no implicit type

I can't declare dot4 in the "bar" function because it would then be considered a local variable instead of a function being called. This creates the error of "dot4(k,v)" trying to look for a matrix element in dot4, causing a type mismatch within "bar".

I was searching for information about this issue and saw that using modules and public statements may help, but I am confused as to how these work. Is this the right direction to be looking in, or is it some other syntax issue? Thanks.

 

[SteamKing]

In your program, the function DOT4 should be declared as REAL FUNCTION DOT4 (v1, v2). I would ignore the RESULT (res) and add a new last executable statement in the function routine which reads DOT4 = RES.

In FUNCTION BAR, you can add a new declaration statement REAL DOT4. I would modify FUNCTION BAR in a similar manner as suggested for DOT4.

For more information, see this article:

http://www.cs.mtu.edu/~shene/COURSES/cs201/NOTES/F90-Subprograms.pdf

 

[Svein]

I do not like FORTRAN, but I think I can see it:

function dot4(v1,v2) result(res)
implicit none
complex*16 v1(4),v2(4)
complex*16 dot4 ← I think you must add this.

 

[avikarto]

I do not like FORTRAN, but I think I can see it:

function dot4(v1,v2) result(res)
implicit none
complex*16 v1(4),v2(4)
complex*16 dot4 ← I think you must add this.

The name of the function is just that, the function's name. There is no variable with this name used. I first tried it that way without using the result(res), but had no success. Making only that change you listed results in an error for trying to declare the function as a local variable.

In your program, the function DOT4 should be declared as REAL FUNCTION DOT4 (v1, v2). I would ignore the RESULT (res) and add a new last executable statement in the function routine which reads DOT4 = RES.

In FUNCTION BAR, you can add a new declaration statement REAL DOT4. I would modify FUNCTION BAR in a similar manner as suggested for DOT4.

For more information, see this article:

http://www.cs.mtu.edu/~shene/COURSES/cs201/NOTES/F90-Subprograms.pdf

The issue with doing that with the function bar is that the returned value is a complex 4-vector. The declaration

complex*16,  dimension(4) function bar...

also throws a syntax error in the function definition.

 

[Svein]

As I said, I do not like FORTRAN. I used Google (https://en.wikibooks.org/wiki/Fortran/Fortran_procedures_and_functions) and found that I had missed a double colon.

function dot4(v1,v2) result(res)
implicit none
complex*16 v1(4),v2(4)
complex*16 :: dot4 ← I think you must add this.

 

[avikarto]

As I said, I do not like FORTRAN. I used Google (https://en.wikibooks.org/wiki/Fortran/Fortran_procedures_and_functions) and found that I had missed a double colon.

function dot4(v1,v2) result(res)
implicit none
complex*16 v1(4),v2(4)
complex*16 :: dot4 ← I think you must add this.

The double colon is just a formality as far as I know. Making that change,

!----------4-vector dot product under Minkowski metric----------
function dot4(v1,v2)! result(res)
    implicit none
    complex*16 v1(4),v2(4)
!    complex*16, dimension(0):: res
    complex*16:: dot4
   
!    res=v1(1)*v2(1)-v1(2)*v2(2)-v1(3)*v2(3)-v1(4)*v2(4)
    dot4=v1(1)*v2(1)-v1(2)*v2(2)-v1(3)*v2(3)-v1(4)*v2(4)
    return
end

results in the same error that was initially present.

 

[avikarto]

http://stackoverflow.com/questions/...when-declaring-a-variable-with-a-double-colon

This shows how the double colon must be used for parameter declarations, but not just for a basic variable.

 

[avikarto]

Thanks for trying to help, all. Putting all my my constants, functions, and subroutines into a "contains" block in a separate module file, and "use"ing that module in my main program file cleared up these errors. I have some really weird module related messages now though, so i'll make a new thread for that.

 

[gsal]

avikatro:
It would be best if you posted your entire program to get an idea of how it is structured...things can be in a single file and single program unit or could be over several files. The main program unit also allows for a 'contains' block, this fixes a lot of interface issues. Lastly, there is nothing wrong with function prototypes (function type declaration inside other subroutines/functions, Fortran also allows for that.

gsal

 

[avikarto]

I got it worked out, thanks gsal. To anyone looking at this thread in the future taking a similar approach, the module errors were being generated by neglecting to include the module output file in my command line execution of my main program as such:

g95 mainProg.f95 myModule.o -L/libs/dfftpack ...

Note also that the module needs to be compiled separately as

g95 myModule.f95 -c