library

On Wed, 7 Apr 1999, Heberto Ghezzo wrote:

Now I have another question, I would like to implement the 
'CutCorner" mothod of random variable generation, I have the C code 
which calls a function which returns the ordinates of the 
distribution to be sampled. Since I do not want to generate normals 
or betas but distributions estimated from lowess or splines the said 
function have to be written in R. Say it is MyDist()
So I need a 'Setup" function that calls the C and set up parameters 
and passes the name of the R function to compute the distribution and 
a "Rand" function that just calls a C function that returns the 
random variate.
Now in R I put .C("Setup_system",as.double(par1),MyDist,....
How do I pass the name of the R function that evaluates the 
distribution? i.e. MyDist()
and in Setup_system.c() how do I call the R function?
S_Call("MyDist",par1,..

Sounds to me as if you need to use call_R and pass the function (not its
name) as an argument to .C("Setup_system", ....

Now, there are not many examples of call_R/call_S about.  Read the
appropriate section in the Blue Book rather carefully.

On Wed, Apr 07, 1999 at 05:16:31PM +0100, Prof Brian D Ripley wrote:

On Wed, 7 Apr 1999, Heberto Ghezzo wrote:

Sounds to me as if you need to use call_R and pass the function (not its
name) as an argument to .C("Setup_system", ....

Now, there are not many examples of call_R/call_S about.  Read the
appropriate section in the Blue Book rather carefully.

  Not many, but one small which shows exactly this sort of things
  is the R dynload demo. Look to demos/dynload/zero.[cR]

On Wed, 7 Apr 1999, Guido Masarotto wrote:

On Wed, Apr 07, 1999 at 05:16:31PM +0100, Prof Brian D Ripley wrote:

On Wed, 7 Apr 1999, Heberto Ghezzo wrote:

Sounds to me as if you need to use call_R and pass the function (not its
name) as an argument to .C("Setup_system", ....

Now, there are not many examples of call_R/call_S about.  Read the
appropriate section in the Blue Book rather carefully.

  Not many, but one small which shows exactly this sort of things
  is the R dynload demo. Look to demos/dynload/zero.[cR]

That _is_ the example from the Blue book: you will need to read the book
(very, very carefully in my experience) to understand the incantations in
the C source.

You might want to try:
http://www.mathsoft.co.uk/splus/download/cfortran.pdf

There is an example in there which is reasonibly clear. 

The c/fortran bit of glm code in Splus, uses some special magic stuff with
frames and memcpys.  I've got some example code of that here is that might
be a better type of solution.  Although I have try this stuff in R, and it
might not work.