execute embedded R script in specified env

[filipsch]

Fixes #851

As mentioned in the issue, I have no idea about how to include a unit test to verify the new behavior; I am open to suggestions or somebody else adding it. Thanks!

I checked locally and it works:

test.rcpp

#include <Rcpp.h>
using namespace Rcpp ;

// [[Rcpp::export]]
int answer(){
    return 42 ;
}

/*** R
    # call answer and check you get the right result
    x <- answer()
    x
*/

in R:

> library(Rcpp)
> testEnv <- new.env()
> sourceCpp("test.cpp", env = testEnv)

>     # call answer and check you get the right result
>     x <- answer()

>     x
[1] 42
> ls(testEnv)
[1] "answer" "x"  

[eddelbuettel]

For unit tests, please have a look at inst/unitTests/runit.*.R and inst/unitTests/cpp/*. Not all that complicated---R functions calling C++ code built via Rcpp Attributes. Just like you did in your reprex. Or am I missing something here while I barely started the first coffee cup of the day?

[filipsch]

@eddelbuettel Thanks for the pointers, I'll see how far I get with those and will report here!

[filipsch]

@eddelbuettel Added a test, that fails if the local = env portion is not there.

[eddelbuettel]

In that case could you actually have it fail and catch the failure?

We have plenty of other tests catching exceptions and errors; RUnit supports that.

[filipsch]

@eddelbuettel in this particular case, I think it is hard to do: It would involve me overriding the source() function to ignore the local argument, not taking in the env parameter as such. Feels like a lot of trickery.

[eddelbuettel]

Maybe. Methinks you could mod your example from #851 to provide the "wrong" answer() function and check for the wrong result. Or maybe not. Have not thought to hard about this.

Took a number of years for someone to come along and declare this a problem :)

[filipsch]

I'm not sure if I understand. This fix makes sure that embedded R code is executed in the R environment in which the RCpp-generated functions are made available, instead of the default globalenv().
To check that this errs, I would have to programmatically update the R source of sourceCpp() and not take in the env argument. I don't see another way. Doing this feels very contrived.
Not sure if there's anymore I can contribute :)

[eddelbuettel]

My thought was that you don't have to have the R code of your C++ snippet executed via the sourceCpp() functionality you so like (and which is great!). Use sourceCpp() to compile a new function in your new environment -- the whole point of #851 as I understand it -- and then in the normal RUnit R code execution look for it. It should fail as it should be tucked away in the new environment. No?

[filipsch]

It's actually the other way around. The point of #851 is to source the R code in my new environment. Making available the Rcpp functions works fine already.
I've added a spec accordingly that fails as expected.
Thanks for your pointers!

[eddelbuettel]

I went all-in and did a reverse depends check, with was (as expected) without issues, so folding this in.