I tried to use RcppArmadillo to compute the Frobenious norm across each slice of an array(arma::cube). The example code in file cxxFuns.cpp is presented below.
// [[Rcpp::depends(RcppArmadillo)]]
#include "RcppArmadillo.h"
// [[Rcpp::export]]
arma::vec array_norm (arma::cube & x) {
arma::vec out = x.each_slice([](arma::mat& x0) {return arma::norm(x0, "fro") ;}) ;
return out ;
}
After the cpp file is compiled successfully, the function array_norm throws an exception:
set.seed(2020)
Rcpp::sourceCpp('cxxFuns.cpp')
x <- array(rnorm(3*4*5), 3:5)
array_norm(x)
#> error: copy into matrix: can't interpret cube with dimensions 3x4x5 as a vector
#> Error in array_norm(x) :
#> copy into matrix: can't interpret cube with dimensions 3x4x5 as a vector
Created on 2020-12-02 by the reprex package (v0.3.0)
That actually an error message from Armadillo because something is not right with your logic yet -- the result is not a vector. You can see that by changing to this (self-contained, a very nice Rcpp trick) code:
Code
// [[Rcpp::depends(RcppArmadillo)]]
#include "RcppArmadillo.h"
// [[Rcpp::export]]
bool array_norm (arma::cube & x) {
auto out = x.each_slice([](arma::mat& x0) {return arma::norm(x0, "fro") ;}) ;
out.print("out");
return true;
}
/*** R
set.seed(2020)
x <- array(rnorm(3*4*5), 3:5)
array_norm2(x)
*/
Output
> Rcpp::sourceCpp("~/git/stackoverflow/65104769/answer.cpp")
> set.seed(2020)
> x <- array(rnorm(3*4*5), 3:5)
> array_norm2(x)
out
[cube slice 0]
0.3770 -1.1304 0.9391 0.1174
0.3015 -2.7965 -0.2294 -0.8531
-1.0980 0.7206 1.7591 0.9093
[cube slice 1]
1.1964 1.8000 -2.2890 1.0982
-0.3716 1.7040 0.0583 0.3182
-0.1233 -3.0388 2.1744 -0.0731
[cube slice 2]
0.8343 0.9367 -0.8125 2.4354
0.1988 -0.1474 -0.7437 0.3881
1.2978 0.1104 1.0953 0.2906
[cube slice 3]
-0.2856 0.4472 -0.3010 0.2531
0.0760 0.9085 -0.7260 -0.3707
-0.5603 -0.5051 -1.1801 0.0222
[cube slice 4]
0.6600 0.6014 0.1188 -1.3283
0.4888 -0.6738 0.1212 -0.5669
-0.1888 0.4761 -0.1860 0.5788
[1] TRUE
Related
I am facing a very similar issue to these questions:
convert Rcpp::NumericVector to Eigen::VectorXd
Converting between NumericVector/Matrix and VectorXd/MatrixXd in Rcpp(Eigen) to perform Cholesky solve
I am writing an R-package, that uses the RcppEigen library for Matrix arithmetic. My problem is that the project is not compiling because of an error in the conversion of the Rcpp::NumericMatrix input to an Eigen::MatrixXd.
The file looks like this:
#include <map>
#include <Rcpp.h>
...
using namespace Eigen;
...
// [[Rcpp::export]]
Rcpp::List my_function(Rcpp::NumericMatrix input_matrix)
{
...
Map<MatrixXd> GATE_matrixx(Rcpp::as<Map<MatrixXd> >(GATE_matrix));
...
}
This gives me the following error:
Myfile.cpp:40:65: required from here
C:/Users/User/AppData/Local/Programs/R/R-4.2.2/library/Rcpp/include/Rcpp/internal/Exporter.h:31:31:error:
matching function for call to 'Eigen::Map<Eigen::Matrix<double, -1,
-1> >::Map(SEXPREC*&) 31 | Exporter( SEXP x ) : t(x) | ^ In file included from
C:/Users/User/AppData/Local/Programs/R/R-4.2.2/library/RcppEigen/include/Eigen/Core:19
from
C:/Users/User/AppData/Local/Programs/R/R-4.2.2/library/RcppEigen/include/Eigen/SparseCore:11
from
C:/Users/User/AppData/Local/Programs/R/R-4.2.2/library/RcppEigen/include/Eigen/Sparse:26
from Myfile.cpp:8
I have also tried to change the line to:
MatrixXd input_matrix_eigen(Rcpp::as\<MatrixXd\>(input_matrix));
This gives me the equivalent error :
Myfile.cpp:40:54: required from here
C:/Users/User/AppData/Local/Programs/R/R
4.2.2/library/RcppEigen/include/Eigen/src/Core/Matrix.h:332:31:error: matching function for call to 'Eigen::Matrix<double, -1,
-1>::_init1<SEXPREC*>(SEXPREC* const&)
Do you have any ideas?
If more information is required to evaluate the issue, just let me know.
If you use the command
RcppEigen::RcppEigen.package.skeleton("demoPackage")
an example package demoPackage is created for you which you can install the usual way. It contains example functions to create amd return a matrix:
// [[Rcpp::export]]
Eigen::MatrixXd rcppeigen_hello_world() {
Eigen::MatrixXd m1 = Eigen::MatrixXd::Identity(3, 3);
// Eigen::MatrixXd m2 = Eigen::MatrixXd::Random(3, 3);
// Do not use Random() here to not promote use of a non-R RNG
Eigen::MatrixXd m2 = Eigen::MatrixXd::Zero(3, 3);
for (auto i=0; i<m2.rows(); i++)
for (auto j=0; j<m2.cols(); j++)
m2(i,j) = R::rnorm(0, 1);
return m1 + 3 * (m1 + m2);
}
If you set the same seed as I do you should get the same matrix
> set.seed(42)
> demoPackage::rcppeigen_hello_world()
[,1] [,2] [,3]
[1,] 8.11288 -1.694095 1.089385
[2,] 1.89859 5.212805 -0.318374
[3,] 4.53457 -0.283977 10.055271
>
I have a list of Numeric Vector and I need a List of unique elements. I tried Rcpp:unique fonction. It works very well when apply to a Numeric Vector but not to List. This is the code and the error I got.
List h(List x){
return Rcpp::unique(x);
}
Error in dyn.load("/tmp/RtmpDdKvcH/sourceCpp-x86_64-pc-linux-gnu-1.0.0/sourcecpp_272635d5289/sourceCpp_10.so") :
unable to load shared object '/tmp/RtmpDdKvcH/sourceCpp-x86_64-pc-linux-gnu-1.0.0/sourcecpp_272635d5289/sourceCpp_10.so':
/tmp/RtmpDdKvcH/sourceCpp-x86_64-pc-linux-gnu-1.0.0/sourcecpp_272635d5289/sourceCpp_10.so: undefined symbol: _ZNK4Rcpp5sugar9IndexHashILi19EE8get_addrEP7SEXPREC
It is unclear what you are doing wrong, and it is an incomplete / irreproducible question.
But there is a unit test that does just what you do, and we can do it by hand too:
R> Rcpp::cppFunction("NumericVector uq(NumericVector x) { return Rcpp::unique(x); }")
R> uq(c(1.1, 2.2, 2.2, 3.3, 27))
[1] 27.0 1.1 3.3 2.2
R>
Even if there isn't a matching Rcpp sugar function, you can call R functions from within C++. Example:
#include <Rcpp.h>
using namespace Rcpp;
Rcpp::Environment base("package:base");
Function do_unique = base["unique"];
// [[Rcpp::export]]
List myfunc(List x) {
return do_unique(x);
}
Thank you for being interested to this issue.
As I notified that, my List contains only NumericVector. I propose this code that works very well and faster than unique function in R. However its efficiency decreases when the list is large. Maybe this can help someone. Moreover, someone can also optimise this code.
List uniqueList(List& x) {
int xsize = x.size();
List xunique(x);
int s = 1;
for(int i(1); i<xsize; ++i){
NumericVector xi = x[i];
int l = 0;
for(int j(0); j<s; ++j){
NumericVector xj = x[j];
int xisize = xi.size();
int xjsize = xj.size();
if(xisize != xjsize){
++l;
}
else{
if((sum(xi == xj) == xisize)){
goto notkeep;
}
else{
++l;
}
}
}
if(l == s){
xunique[s] = xi;
++s;
}
notkeep: 0;
}
return head(xunique, s);
}
/***R
x <- list(1,42, 1, 1:3, 42)
uniqueList(x)
[[1]]
[1] 1
[[2]]
[1] 42
[[3]]
[1] 1 2 3
microbenchmark::microbenchmark(uniqueList(x), unique(x))
Unit: microseconds
expr min lq mean median uq max neval
uniqueList(x) 2.382 2.633 3.05103 2.720 2.8995 29.307 100
unique(x) 2.864 3.110 3.50900 3.254 3.4145 24.039 100
But R function becomes faster when the List is large. I am sure that someone can optimise this code.
It seems that the sugar Rcpp function setdiff orders the values, which is different from the standard R function setdiff.
As an example, consider the following code:
src <-
"IntegerVector setdiff_Rcpp(IntegerVector x, IntegerVector y){
IntegerVector d = setdiff(x,y);
return(d);
}"
Rcpp::cppFunction(src)
setdiff(15:11, c(13,12))
# [1] 15 14 11
setdiff_Rcpp(15:11, c(13,12))
# [1] 11 14 15
Is it possible to obtain a result as in the standard R function?
[Edit]
I was able to solve my problem. Here is the Rcpp function I used:
// [[Rcpp::export]]
IntegerVector setdiff_R(IntegerVector x, IntegerVector y)
{
// difference of sets x & y (without reordering)
x = x[duplicated(x) == 0]; x = na_omit(x);
y = y[duplicated(y) == 0]; y = na_omit(y);
IntegerVector out(0, NA_INTEGER);
for(int i=0; i < x.length(); i++)
{
if(is_false(any(x[i] == y)))
out.push_back(x[i]);
}
return(out);
}
It only works for IntegerVector type and probably it is not optimised but it gets the job done.
From R, I'm trying to run sourceCpp on this file:
#include <RcppArmadillo.h>
// [[Rcpp::depends(RcppArmadillo)]]
using namespace arma;
using namespace Rcpp;
// [[Rcpp::export]]
vec dnormLog(vec x, vec means, vec sds) {
int n = x.size();
vec res(n);
for(int i = 0; i < n; i++) {
res[i] = log(dnorm(x[i], means[i], sds[i]));
}
return res;
}
See this answer to see where I got the function from. This throws the error:
no matching function for call to 'dnorm4'
Which is the exact error I was hoping to prevent by using the loop, since the referenced answer mentions that dnorm is only vectorized with respect to its first argument. I fear the answer is obvious, but I've tried adding R:: before the dnorm, tried using NumericVector instead of vec, without using log() in front. No luck. However, adding R:: before dnorm does produce a separate error:
too few arguments to function call, expected 4, have 3; did you mean '::dnorm4'?
Which is not fixed by replacing dnorm above with R::dnorm4.
There are two nice teachable moments here:
Pay attention to namespaces. If in doubt, don't go global.
Check the headers for the actual definitions. You missed the fourth argument in the scalar version R::dnorm().
Here is a repaired version, included is a second variant you may find interesting:
#include <RcppArmadillo.h>
// [[Rcpp::depends(RcppArmadillo)]]
// [[Rcpp::export]]
arma::vec dnormLog(arma::vec x, arma::vec means, arma::vec sds) {
int n = x.size();
arma::vec res(n);
for(int i = 0; i < n; i++) {
res[i] = std::log(R::dnorm(x[i], means[i], sds[i], FALSE));
}
return res;
}
// [[Rcpp::export]]
arma::vec dnormLog2(arma::vec x, arma::vec means, arma::vec sds) {
int n = x.size();
arma::vec res(n);
for(int i = 0; i < n; i++) {
res[i] = R::dnorm(x[i], means[i], sds[i], TRUE);
}
return res;
}
/*** R
dnormLog( c(0.1,0.2,0.3), rep(0.0, 3), rep(1.0, 3))
dnormLog2(c(0.1,0.2,0.3), rep(0.0, 3), rep(1.0, 3))
*/
When we source this, both return the same result because the R API allows us to ask for logarithms to be taken.
R> sourceCpp("/tmp/dnorm.cpp")
R> dnormLog( c(0.1,0.2,0.3), rep(0.0, 3), rep(1.0, 3))
[,1]
[1,] -0.923939
[2,] -0.938939
[3,] -0.963939
R> dnormLog2(c(0.1,0.2,0.3), rep(0.0, 3), rep(1.0, 3))
[,1]
[1,] -0.923939
[2,] -0.938939
[3,] -0.963939
R>
A very simple question regarding RcppArmadillo. Trying to multiply a vector by a scalar, and getting different results depending on small changes in syntax.
Any ideas?
// [[Rcpp::depends("RcppArmadillo")]]
// [[Rcpp::export]]
arma::vec funtemp(arma::vec x)
{
// return(x/10); // this works
// return((1/10)*x); // this does not work
return(x*(1/10)); // this does not work
}
Ahh, the good ol' integer vs. double division problem in C++. Before we begin, note that: arma::vec is by default a double and 1, 10, 1/10 are all ints...
Let's take a look at your functions separately:
#include <RcppArmadillo.h>
// [[Rcpp::depends("RcppArmadillo")]]
// [[Rcpp::export]]
arma::vec funtemp_one(arma::vec x)
{
return(x/10); // this works
}
// [[Rcpp::export]]
arma::vec funtemp_two(arma::vec x)
{
return((1/10)*x); // this does not work
}
// [[Rcpp::export]]
arma::vec funtemp_three(arma::vec x)
{
return(x*(1/10)); // this does not work
}
Therefore, when we run through your problem we get:
> funtemp_one(1)
[,1]
[1,] 0.1
> funtemp_two(1)
[,1]
[1,] 0
> funtemp_three(1)
[,1]
[1,] 0
In the later functions (e.g. the 1/10), the operator/ that is being used is int based division. As a result, 2 ints enter and 1 int is returned. If the result is not divisible, then you end up with a zero being returned as it is outside of the integer scope.
In order to use the double version, which returns a double, at least one of the ints must be explicitly cast to a double. This happens by default in the first case as you have a double/int due to arma::vec's structure. The second and third cases have an int/int structure that can be dealt with in two ways: 1. use a .0 after the int or 2. explicitly cast the value as a double with double(int)
e.g.
// [[Rcpp::export]]
arma::vec funtemp_four(arma::vec x)
{
return(x*(1/10.0)); // this works
}
// [[Rcpp::export]]
arma::vec funtemp_five(arma::vec x)
{
return(x*(1/double(10))); // this works
}
Will give what you expect:
> funtemp_four(1)
[,1]
[1,] 0.1
> funtemp_five(1)
[,1]
[1,] 0.1