[Rcpp-devel] function modifying it self its argument

[Matt D.]

On 2/26/2015 16:37, Pierre GLOAGUEN wrote:
> Thanks, it works!
> I'm not familiar with C++, is it necessary to always use such a 
> function in C++ or is it because of R/C++ interface?
Hi Pierre!

You're witnessing yet another fascinating difference between (usual in 
the world of C++) value semantics and (not that usual in the world of 
C++) reference semantics:
Compare the difference in behavior of `vector_reference` (using 
`Rcpp::NumericVector`) and `vector_value` (using `std::vector<double>`) 
in the code attached below.

To make things clear: no, in C++ you're not (not by default and not in 
the standard world) supposed to care about (nor use) a function such as 
`clone` -- this is idiomatic in programming languages with reference 
semantics, like Java: https://en.wikipedia.org/wiki/Clone_%28Java_method%29

If this makes you feel better: If you're coming from C++ your 
expectations are perfectly reasonable and this is how any well-designed 
(as in: not violating the POLS) C++ library (including the C++ standard 
library) will behave by default.
Which incidentally brings me to the advice I usually give in these 
situations: unless you're absolutely dependent on the "features" of 
`Rcpp::NumericVector` just forget about it and replace all uses with the 
standard container `std::vector<double>`.
// Arguably, NumericVector's POLS violation -- as illustrated by the 
numerous posts raising similar questions -- perhaps isn't necessarily 
the optimal design choice?

Everything will behave just as you expect and as well-documented in the 
usual places -- e.g., http://en.cppreference.com/w/cpp/container/vector
In contrast, `help.search("Rcpp::NumericVector")` happily announces "No 
results found" at the moment -- and, as mentioned in another reply, 
you're apparently expected to Google around to find methods for solving 
problems you wouldn't even encounter otherwise.
As fun as it can be to learn-via-a-search-engine, if you don't find the 
curiosities like the aforementioned reference-vs-value-semantics 
differences _all_ _that_ fascinating, I think you may find yourself more 
productive and better served by sticking to the standard containers :-)

HTH!

#include <Rcpp.h>

// [[Rcpp::export]]
Rcpp::NumericVector vector_reference(Rcpp::NumericVector input)
{
     Rcpp::NumericVector output(input);
     if (output.size() >= 1) output[0] = 123;
     return output;
}

// [[Rcpp::export]]
std::vector<double> vector_value(const std::vector<double> & input)
{
     std::vector<double> output(input);
     if (output.size() >= 1) output[0] = 123;
     return output;
}

/*** R
v = rep(1, 6)
vector_reference(v)
show(v)

v = rep(1, 6)
vector_value(v)
show(v)

#Output:

#> v = rep(1, 6)
#> vector_reference(v)
#[1] 123   1   1   1   1   1
#> show(v)
#[1] 123   1   1   1   1   1

#> v = rep(1, 6)
#> vector_value(v)
#[1] 123   1   1   1   1   1
#> show(v)
#[1] 1 1 1 1 1 1
*/

Best,

Matt
> Thanks again for your help,
>
> Pierre
>
> Le 26/02/2015 16:30, Jeffrey Pollock a écrit :
>> Perhaps use the clone() function?
>>
>> library(Rcpp)
>>
>> cppFunction("
>> NumericVector par_CMAtR(NumericVector vec_CMA) {
>>     NumericVector out = clone(vec_CMA);
>>     out[5] = exp(out[5]);
>>     return out;
>> }
>> ")
>>
>> vec_C <- rep(1, 6)
>> par_CMAtR(vec_C)
>> print(vec_C)
>>
>> On Thu, Feb 26, 2015 at 3:16 PM, Pierre GLOAGUEN 
>> <Pierre.Gloaguen at ifremer.fr <mailto:Pierre.Gloaguen at ifremer.fr>> wrote:
>>
>>
>>     Hello everybody,
>>
>>     I have a very simple example
>>     I have a vector vec_CMA which length is a multiple of 6.
>>     I want to get the exact same vector, except the last element
>>     which is the exponential of the last element of vec_CMA
>>     The code is the following
>>
>>     //myfun.cpp
>>     #include <Rcpp.h>
>>     using namespace Rcpp;
>>
>>     // [[Rcpp::export]]
>>     NumericVector par_CMAtR(NumericVector vec_CMA){
>>       int K = (vec_CMA.size())/6;
>>       NumericVector out(6*K);
>>       out = vec_CMA;
>>       out[6*K-1] = exp(vec_CMA[6*K-1]);
>>       return out;
>>     }
>>
>>     I apply the function with the R code
>>     sourceCpp("myfun.cpp")
>>     vec_C <- rep(1,6)
>>     par_CMAtR(vec_C)
>>     [1] 1 1 1 1 1 2.718282
>>
>>     8
>>
>>     works fine. Except the vec_C is modified too!
>>     vec_C
>>     [1] 1 1 1 1 1 2.718282
>>
>>     It's the first time I have this kind of problem. What is wrong in
>>     my code?
>>     Thanks for your help,
>>
>>     Pierre Gloaguen
>>
>>
>>     _______________________________________________
>>     Rcpp-devel mailing list
>>     Rcpp-devel at lists.r-forge.r-project.org
>>     <mailto:Rcpp-devel at lists.r-forge.r-project.org>
>>     https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
>>
>>
>
>
> _______________________________________________
> Rcpp-devel mailing list
> Rcpp-devel at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20150226/817d6af0/attachment.html>