将 Rcpp 函数扩展到任何类型的输入向量 [英] Extending Rcpp function to input vector of any type
问题描述
我有以下函数,它在 NumericVector
上做一个简单的循环并返回 int
类型值.
I have following function which does a simple loop on NumericVector
and returns int
type value.
Rcpp::cppFunction({'
int calc_streak( NumericVector x, int i1, int i2){
int cur_streak=1;
if (NumericVector::is_na(x[0])){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
for(int j = i1; j <= i2 ; ++j) {
if( x[ j ] == x[ j-1 ]){
cur_streak += 1;
} else if(NumericVector::is_na( x[ j ] )){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
}
return cur_streak;
}
"})
calc_streak(c(1,1,1,1),i1=0,i2=3)
# [1] 4
功能对我来说很好,但真正的问题是我试图在其他输入类型上扩展此功能.我一直在搜索堆栈 here 和 此处,但这些示例在我的案例或我不知道如何正确使用示例.我尝试了几种处理未知输入类型的方法,但在我的情况下没有成功.下面三个例子
Function works fine for me but the real issue is when I'm trying to extend this functionality on other input-types. I've been searching on stack here and here, but those examples doesn't work in my case or I don't know how to use examples properly. I've tried few methods dealing with unknown input type, and none was successful in my case. Three examples below
最简单的灵感来自这个 -创建了 main 函数,它根据参数
TYPEOF(x)
的类型运行先前定义的函数之一.此函数返回integer
和numeric
的预期值.对于character
会话崩溃
The simplest one inspired by this - created main function which run one of previously defined functions depending on type of argument
TYPEOF(x)
. This function returns expected value forinteger
andnumeric
. Forcharacter
session crashes
Rcpp::cppFunction('
#include <Rcpp.h>
using namespace Rcpp;
int streak_run_int(IntegerVector x, int i1, int i2){
int cur_streak=1;
if (IntegerVector::is_na(x[0])){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
for(int j = i1; j <= i2 ; ++j) {
if( x[ j ] == x[ j-1 ]){
cur_streak += 1;
} else if(IntegerVector::is_na( x[ j ] )){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
}
return cur_streak;
}
int streak_run_char(CharacterVector x, int i1, int i2){
int cur_streak=1;
if (CharacterVector::is_na(x[0])){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
for(int j = i1; j <= i2 ; ++j) {
if( x[ j ] == x[ j-1 ]){
cur_streak += 1;
} else if(CharacterVector::is_na( x[ j ] )){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
}
return cur_streak;
}
// [[Rcpp::export]]
int streak_run4(SEXP x, int i1, int i2) {
switch (TYPEOF(x)) {
case INTSXP: {
return streak_run_int(as<IntegerVector>(x), i1, i2);
}
case STRSXP: {
return streak_run_char(as<CharacterVector>(x), i1, i2);
}
default: { return 0; }
}
}
')
# expected results for int and real - for character session crashes
streak_run4( c(1,1,1,1),i1=0, i2=3)
streak_run4( as.integer(c(1,1,1,1)),i1=0, i2=3)
streak_run4( as.character(c(1,1,1,1)),i1=0, i2=3)
<小时>
第二个函数有完全相同的想法,但使用模板而不是定义多个函数.结果与上述相同 - 会话在
character
输入时崩溃
Second function has exactly the same idea, but using template instead of defining multiple functions. Same results as above - session crash on
character
input
Rcpp::cppFunction('
#include <Rcpp.h>
using namespace Rcpp;
namespace impl {
template <int RTYPE>
int streak_run_impl(const Vector<RTYPE>& x, int i1, int i2)
{
int cur_streak=1;
if (Vector<RTYPE>::is_na(x[0])){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
for(int j = i1; j <= i2 ; ++j) {
if( x[ j ] == x[ j-1 ]){
cur_streak += 1;
} else if(Vector<RTYPE>::is_na( x[ j ] )){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
}
return cur_streak;
}
}
// [[Rcpp::export]]
int streak_run3(SEXP x, int i1, int i2) {
switch (TYPEOF(x)) {
case INTSXP: {
return impl::streak_run_impl(as<IntegerVector>(x), i1, i2);
}
case REALSXP: {
return impl::streak_run_impl(as<NumericVector>(x), i1, i2);
}
case STRSXP: {
return impl::streak_run_impl(as<CharacterVector>(x), i1, i2);
}
case LGLSXP: {
return impl::streak_run_impl(as<LogicalVector>(x), i1, i2);
}
case CPLXSXP: {
return impl::streak_run_impl(as<ComplexVector>(x), i1, i2);
}
default: {
return 0;
}
}
}
')
streak_run3( c(1,1,1,1),i1=0, i2=3)
streak_run3( as.integer(c(1,1,1,1)),i1=0, i2=3)
streak_run3( as.character(c(1,1,1,1)),i1=0, i2=3)
<小时>
另一个灵感来自这篇文章,并且这次我什至无法编译 C++ 函数,同时有一个错误
使用重载运算符 '==' 是模棱两可的
.无论如何,在检查了上述两个示例之后,我不期望有任何其他结果.
Another one is inspired by this article, and this time I wasn't even able to compile C++ function, while having an error
use of overloaded operator '==' is ambiguous
. Anyway, after examining two above examples, I don't expect any other result.
Rcpp::cppFunction('
#include <Rcpp.h>
using namespace Rcpp;
class streak_run2_impl {
private:
int i1;
int i2;
public:
streak_run2_impl(int i1, int i2) : i1(i1), i2(i2) {}
template <int RTYPE>
IntegerVector operator()(const Vector<RTYPE>& x)
{
int cur_streak=1;
if (Vector<RTYPE>::is_na(x[0])){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
for(int j = i1; j <= i2 ; ++j) {
if( x[ j ] == x[ j-1 ] ){
cur_streak += 1;
} else if(Vector<RTYPE>::is_na( x[ j ] )){
cur_streak = NumericVector::get_na();
} else {
cur_streak = 1;
}
}
return cur_streak;
}
};
// [[Rcpp::export]]
RObject streak_run2(RObject x, int i1 = 0, int i2=6){
RCPP_RETURN_VECTOR(streak_run2_impl(i1, i2), x);
}
')
<小时>
所以我的问题是:如何正确定义此函数以获得任何 R 类的输入向量的结果?
我很乐意提供任何建议.
So my question is:
How to properly define this function to obtain results for input vector of any R class?
I would be obliged for any advices.
推荐答案
我认为示例中的主要错误是您在 j = 0
处开始循环,因此您调用 operator[](-1)
.以下对我有用.制作以下 func.cpp
I think the main error in examples are that you start your loop at j = 0
so you call operator[](-1)
. The following works for me. Make the following func.cpp
#include <Rcpp.h>
#include <algorithm>
using namespace Rcpp;
template <int RTYPE>
int streak_run_impl(const Vector<RTYPE>& x, int i1, int i2)
{
int cur_streak = 1;
if (Vector<RTYPE>::is_na(x[0])){
cur_streak = NA_INTEGER;
} else {
cur_streak = 1;
}
for(int j = std::max(i1, 1) /* have to start at one at least */;
j < std::min(i2 + 1, (int)x.size()) /* check size of x */; ++j){
if(x[j] == x[j - 1]){
cur_streak += 1;
} else if(Vector<RTYPE>::is_na(x[j])){
cur_streak = NA_INTEGER;
} else {
cur_streak = 1;
}
}
return cur_streak;
}
// [[Rcpp::export]]
int streak_run3(SEXP x, int i1, int i2) {
switch (TYPEOF(x)) {
case INTSXP: {
return streak_run_impl(as<IntegerVector>(x), i1, i2);
}
case REALSXP: {
return streak_run_impl(as<NumericVector>(x), i1, i2);
}
case STRSXP: {
return streak_run_impl(as<CharacterVector>(x), i1, i2);
}
case LGLSXP: {
return streak_run_impl(as<LogicalVector>(x), i1, i2);
}
case CPLXSXP: {
return streak_run_impl(as<ComplexVector>(x), i1, i2);
}
default: {
return 0;
}
}
}
然后运行此 R 脚本,并将工作目录设置为 .cpp
文件的目录
Then run this R script with the working directory set to that of the .cpp
file
Rcpp::sourceCpp("func.cpp")
streak_run3(c(1,1,1,1), i1=0, i2=3)
streak_run3(as.integer(c(1,1,1,1)), i1=0, i2=3)
streak_run3(as.character(c(1,1,1,1)), i1=0, i2=3)
这篇关于将 Rcpp 函数扩展到任何类型的输入向量的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!