Skip to content

[Rcpp-devel] inline with classes

7 messages · Whit Armstrong, Dirk Eddelbuettel

Original
Whit Armstrong
# 
I just reviewed the Rcpp documentation.

I see plenty of cfunction/cxxfunction examples, but I'm curious
whether one can provide a class definition inline in an R script and
then initialize an instance of the class and call a method on the
class.  all inline in R.

Is this feature something you all would consider adding to Rcpp?

-Whit
Dirk Eddelbuettel
# 
Hi Whit,
On 5 October 2010 at 12:25, Whit Armstrong wrote:
| I just reviewed the Rcpp documentation.

Great. Now tell us how to make it sticky so that we get you to contribute :)
 
| I see plenty of cfunction/cxxfunction examples, but I'm curious
| whether one can provide a class definition inline in an R script and
| then initialize an instance of the class and call a method on the
| class.  all inline in R.
| 
| Is this feature something you all would consider adding to Rcpp?

a) A change would be an inline feature, not an Rcpp feature.

b) But we already do :) support this, and even use it in numerous cases in
   the unit tests; this uses the   include=   argument to cxxfunction et al.

So you can supply your templated logic in an R string to include= and then
use the code in an R string to to src= and pass it all through cxxfunction()
for a quick test.

See e.g. runit.macros.R or or runit.Module.R.  I think we also have blog
posts and / or vignette examples that use it.

Hth, Dirk
Whit Armstrong
# 
aha, thanks for correcting me. on the inline/Rcpp division of labor.

So, what you are suggesting should have been obvious to me.

I pass the class def in the include statement of the inline function?
Is that it?

I'll ping back after I try this.

If this really works, then I think WinBUGS can finally die.

-Whit


2010/10/5 Dirk Eddelbuettel <edd at debian.org>:
Dirk Eddelbuettel
#
On 5 October 2010 at 13:21, Whit Armstrong wrote:
| aha, thanks for correcting me. on the inline/Rcpp division of labor.
| 
| So, what you are suggesting should have been obvious to me.
| 
| I pass the class def in the include statement of the inline function?
| Is that it?
| 
| I'll ping back after I try this.
| 
| If this really works, then I think WinBUGS can finally die.

It works over here, chances are you may get lucky too. 

Here, I took a simple templated class square from one of the unit tests and
wrapped it in (fairly verbose) example:


edd at max:~$ cat /tmp/whit.r 
#!/usr/bin/r -ti

suppressMessages(library(inline))
suppressMessages(library(Rcpp))

inc <- 'template <typename T>
	class square : public std::unary_function<T,T> {
	public:
	    T operator()( T t) const { return t*t ;}
	};
       '

src <- '
       double x = Rcpp::as<double>(xs);
       int i = Rcpp::as<int>(is);
       square<double> sqdbl;
       square<int> sqint;
       return Rcpp::List::create(Rcpp::Named("x", sqdbl(x)),
                                 Rcpp::Named("i", sqint(i)));
       '
fun <- cxxfunction(signature(xs="numeric", is="integer"), body=src, include=inc, plugin="Rcpp")

print(fun(2.0, 3L))
edd at max:~$ chmod 0755 /tmp/whit.r 
edd at max:~$ /tmp/whit.r 
$x
[1] 4

$i
[1] 9

edd at max:~$ 


All good?  Send my condolences to the family of WinBUGS.

Dirk



 
| -Whit
| 
| 
| 2010/10/5 Dirk Eddelbuettel <edd at debian.org>:
| >
| > Hi Whit,
| >
| > On 5 October 2010 at 12:25, Whit Armstrong wrote:
| > | I just reviewed the Rcpp documentation.
| >
| > Great. Now tell us how to make it sticky so that we get you to contribute :)
| >
| > | I see plenty of cfunction/cxxfunction examples, but I'm curious
| > | whether one can provide a class definition inline in an R script and
| > | then initialize an instance of the class and call a method on the
| > | class. ?all inline in R.
| > |
| > | Is this feature something you all would consider adding to Rcpp?
| >
| > a) A change would be an inline feature, not an Rcpp feature.
| >
| > b) But we already do :) support this, and even use it in numerous cases in
| > ? the unit tests; this uses the ? include= ? argument to cxxfunction et al.
| >
| > So you can supply your templated logic in an R string to include= and then
| > use the code in an R string to to src= and pass it all through cxxfunction()
| > for a quick test.
| >
| > See e.g. runit.macros.R or or runit.Module.R. ?I think we also have blog
| > posts and / or vignette examples that use it.
| >
| > Hth, Dirk
| >
| > --
| > Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com
| >
Whit Armstrong
# 
Thanks for the example!  that makes everything clear.

-Whit


2010/10/5 Dirk Eddelbuettel <edd at debian.org>:
3 days later
Whit Armstrong
# 
Thanks again, Dirk.

full example (sorry for the verbose post):

library(inline)
library(Rcpp)

inc <- '
#include <iostream>
#include <armadillo>
#include <cppbugs/cppbugs.hpp>

using namespace arma;
using namespace cppbugs;

class TestModel: public MCModel {
public:
  const mat& y; // given
  const mat& X; // given

  Normal<vec> b;
  Uniform<double> tau_y;
  Deterministic<mat> y_hat;
  Normal<mat> likelihood;
  Deterministic<double> rsq;

  TestModel(const mat& y_,const mat& X_): y(y_), X(X_),
                                          b(randn<vec>(X_.n_cols)),
                                          tau_y(1),
                                          y_hat(X*b.value),
                                          likelihood(y_,true),
                                          rsq(0)
  {
    add(b);
    add(tau_y);
    add(y_hat);
    add(likelihood);
    add(rsq);
  }

  void update() {
    y_hat.value = X*b.value;
    rsq.value = as_scalar(1 - var(y - y_hat.value) / var(y));
  }
  double logp() const {
    return b.logp(0.0, 0.0001) + tau_y.logp(0,100) +
likelihood.logp(y_hat.value,tau_y.value);
  }
};
'

src <- '
mat X(REAL(XR),100,2);
mat y(REAL(yr),100,1);

TestModel m(y,X);
int iterations = 1e5;
m.sample(iterations, 1e4, 10);
return Rcpp::List::create(Rcpp::Named("b", m.b.mean()),
Rcpp::Named("ar", m.acceptance_ratio()));
'
fun <- cxxfunction(signature(XR="numeric", yr="numeric"), body=src,
include=inc, plugin="Rcpp")
NR <- 100
NC <- 2
x <- matrix(rnorm(NR*NC),NR,NC)
y <- rnorm(NR)
print(system.time(ans <- fun(x,y)))
print(ans)


actual run:
user  system elapsed
  0.420   0.020   0.435
$b
[1] -0.01127731 -0.11217234

$ar
[1] 0.29405




2010/10/5 Dirk Eddelbuettel <edd at debian.org>:
Dirk Eddelbuettel
#
On 8 October 2010 at 23:26, Whit Armstrong wrote:
| Thanks again, Dirk.
| 
| full example (sorry for the verbose post):

Thanks for posting the full example! 

Let us know how you find it and how it helps you with CppBugs.

Just briefly:

| mat X(REAL(XR),100,2);
| mat y(REAL(yr),100,1);

Personally, I like to avoid all UPPERCASE macros but that looks like an
almost reasonable shortcut as you're going straight to Armadillo.

Also in

| fun <- cxxfunction(signature(XR="numeric", yr="numeric"), body=src,
| include=inc, plugin="Rcpp")

I would suggest trying plugin="RcppArmadillo" so that you get the Rcpp
to/from Armadillo data type wrapping etc.  There have been a few
RcppArmadillo examples posted on the list in response to other questions.

Cheers, Dirk