R help - Jan 2012 - Calling function in DLL using .C

I cannot understand why a function appears to be available to R 
(is.loaded('planckwR') returns TRUE) but the call with .C results in an 
error message:

C symbol name 'planckwR' not in DLL for package <path to DLL file
here>

This is what I do:

  Loading a homebrewed DLL, compiled with MS VS97. The functions are all 
void, the C calling sequence is used and all arguments are passed by 
reference. Done things much like this many times with both Fortran and C 
subroutines and functions and called those successfully from either VBA 
or Python.

R code like this:

++++++++++++++++

dllpath<-file.path('path to dill', paste('Planck', 
.Platform$dynlib.ext,sep=''))
dll<-dyn.load(dllpath)


dllinfo<-getLoadedDLLs()
is.loaded('planckwR')
getNativeSymbolInfo('planckwR')

+++++++++++++++++++++
is.loaded returns TRUE.

getNativeSymbolInfo returns the name and a pointer to the address of the 
function planckwR. The $package element of the list returned by 
getNativeSymbolInfo is NULL, which adds to my confusion as it should be 
a list of 3 elements.

What am I missing here? All help much appreciated.

Alex van der Spek

On 24/01/2012 6:49 AM, Alex van der Spek wrote:
> I cannot understand why a function appears to be available to R
> (is.loaded('planckwR') returns TRUE) but the call with .C results
in an
> error message:
>
> C symbol name 'planckwR' not in DLL for package<path to DLL file
here>
You posted this question a few days ago, and said you'd try using gcc 
instead of VS.  How did that go?

Duncan Murdoch

Uploaded the DLL here:

http://www.filefactory.com/file/c2c56ad/n/Planck.dll

Appreciate if you can check to see if anything is wrong with this DLL. I
use dumpbin to look at what is in the exports table of the DLL. No reason
for concerns there. I can work with this DLL from both VBA and Python.
Python's ctypes allow the use of both the C and stdcall versions. Both
work fine.


By the way, the DLL contains, for every function, one export using C
calling conventions for use in R (name appended with R) and one export
using stdcall for use with VBA/Python, the latter using decorated names.

Reading up on gcc in the mean time.

Thank you,
Alex van der Spek
> On 24/01/2012 10:35 AM, Alex van der Spek wrote:
>> Yes, sorry, I am learning how to do this. I am a simple and ageing
>> physicist who grew up with Fortran and Algol.
>>
>> Just wondering if what appears to be an internal R inconsistency
between
>> is.loaded() and .C() can be explained.
>>
>> Attempts to compile with gcc to a DLL unsuccessfull so far. That is
>> solely because I have never used gcc before and have a long way to go.
>> Years ago I taught myself how to use DLLs in VBA, it took 6 weeks to
>> learn how to make sense out of decorated names and Fortran string
>> passing. The use of DLLs in Python was easy. I am OK with investing
>> weeks again to learn how to do this in R but with the apparent
>> inconsistency I am uncertain where to begin. Drop VC97 and use gcc as a
>> cure all?
>
> That would be easiest, because then there would be lots of others
> available to help you.  You may be the only one using your particular
> compiler with R.
>
> You could also post the .dll online so others can take a look and see
> what's going on.  We might not be able to tell you how to fix the
> problem with your compiler, but we could likely tell you what it is.
>
> Duncan Murdoch
>
>> Thanks,
>> Alex van der Spek
>>
>>
>>
>> On 01/24/2012 04:11 PM, Duncan Murdoch wrote:
>> >  On 24/01/2012 6:49 AM, Alex van der Spek wrote:
>> >>  I cannot understand why a function appears to be available to
R
>> >>  (is.loaded('planckwR') returns TRUE) but the call
with .C results in
>> an
>> >>  error message:
>> >>
>> >>  C symbol name 'planckwR' not in DLL for
package<path to DLL file
>> here>
>> >
>> >  You posted this question a few days ago, and said you'd try
using gcc
>> >  instead of VS.  How did that go?
>> >
>> >  Duncan Murdoch
>> >
>>
>
>