problem exporting Rcpp functions

[lukas-rokka]

I want to make a C++ function aviliable to R with the // [[Rcpp::export]] command. This does not work when using the rstan.package.skeleton. I get the error "object '[name of the rcpp function]' not found". It appears to be some problem with registering the routine.

However, if I remove lines 16:20 from src\init.cpp the registration works. Here is the code from those lines:

void attribute_visible R_init_sem5r1c(DllInfo *dll) {
  // next line is necessary to avoid a NOTE from R CMD check
  R_registerRoutines(dll, NULL, CallEntries, NULL, NULL);
  R_useDynamicSymbols(dll, TRUE); // necessary for .onLoad() to work
}

What do these lines of code actually do and are they safe to remove?

[jtipton25]

I'm having the same issue but deleting this block of code doesn't fix it. I used rstanstools::rstan_package_skeleton to setup a package and added the default timesTwo.cpp to the package. I'm getting the error

Error in timesTwo(rnorm(10)) : object '_test_timesTwo' not found

I'd like to have a package to fit a model using RStan and then use an RcppArmadillo function to post-process the data. I can't find any documentation referring to implementing stand-alone Rcpp functions within an RStan package. Any help would be appreciated.

[bgoodri]

There isn't documentation for that in rstantools because no one else (that I know of) has wanted to try, it is usually a better idea to put stuff like that into the generated quantities block of your Stan programs, and doing so presumably just involves adding the usual stuff for a package that uses RcppArmadillo on top of what is generated by `rstantools::rstan_package_skeleton`.

…

On Wed, May 30, 2018 at 1:26 PM, jtipton25 ***@***.***> wrote: I'm having the same issue but deleting this block of code doesn't fix it. I used rstanstools::rstan_package_skeleton to setup a package and added the default timesTwo.cpp to the package. I'm getting the error Error in timesTwo(rnorm(10)) : object '_test_timesTwo' not found I'd like to have a package to fit a model using RStan and then use an RcppArmadillo function to post-process the data. I can't find any documentation referring to implementing stand-alone Rcpp functions within an RStan package. Any help would be appreciated. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#13 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADOrqpLfco802dlh8kwj6KtWYD-HmgL7ks5t3tZDgaJpZM4PlwYA> .

[jsilve24]

This is a problem that should probably be fixed. Putting code in generated quantities block is very limiting.

[lucasnell]

I agree with above. I'm trying to make a package that includes simulations using Rcpp, and having those functions in generated quantities blocks instead of their own .cpp files is a real problem.

[jtimonen]

I would also need RcppArmadillo to work in a package created using Rstantools. I have the same problem: object [name of function] not found

[helske]

I have successfully combined external C++ and Stan models in the same package in the past (https://github.com/helske/walker), but I just started to update some stuff and something has changed so that I new I get an infinite recompilation loops... And if I switch to newer package skeleton structure, I also face the issue that either Stan models or other C++ functions are not getting exported (depending on how I tweak Makevars etc).

[bgoodri]

I think roxygen2 changed its behavior, so you should not have your package rebuild your manpages inside the cleanup function.

…

On Fri, Oct 12, 2018 at 4:52 PM Jouni Helske ***@***.***> wrote: I have successfully combined external C++ and Stan models in the same package in the past (https://github.com/helske/walker), but I just started to update some stuff and something has changed so that I new I get an infinite recompilation loops... And if I switch to newer package skeleton structure, I also face the issue that either Stan models or other C++ functions are not getting exported (depending on how I tweak Makevars etc). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#13 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADOrqgG2DxnLGbjZNaZ4kq6QNoWPk9xsks5ukQD6gaJpZM4PlwYA> .

[helske]

Yes I just realized that as well, removing the roxygen line fixed the infinite loop issue and the package compiles. But the Stan models are not getting exported. Investigation continues...

edit: Oops, I had a wrong version of the stanmodels.R file. Everything works now: https://github.com/helske/walker.

[bgoodri]

Do you have something like this in your zzz.R? https://github.com/stan-dev/rstanarm/blob/master/R/zzz.R#L18

…

On Fri, Oct 12, 2018 at 5:11 PM Jouni Helske ***@***.***> wrote: Yes I just realized that as well, removing the roxygen line fixed the infinite loop issue and the package compiles. But the Stan models are not getting exported. Investigation continues... — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#13 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADOrqoAlaNeg901_KZg5_HaQqbw2ddefks5ukQVlgaJpZM4PlwYA> .

[const-ae]

I encountered the same problem today, that I couldn't use Rcpp functions in a package generated with rstanstools::rstan_package_skeleton.

I found that removing the R_init_mypackage function fixes the problem of calling the Rcpp functions, because the function is then generated by Rcpp in RcppExports.cpp. But I got another error when I loaded the package that "object 'm' not found" as described in issue #43.

I solved this issue by adding

// [[Rcpp::init]]
void rstan_additional_init(DllInfo *dll){
  R_useDynamicSymbols(dll, TRUE); // necessary for .onLoad() to work
}

which makes sure that the dynamic symbol is accessible in the .onLoad function in zzz.R.

So in conclusion, replacing the R_init_mypackage function with the code above fixed my problem.

[bgoodri]

Nice. There may be other solutions but I am glad you found one. Be sure to read https://discourse.mc-stan.org/t/how-to-update-a-package-that-has-stanheaders-and-rstan-in-its-linkingto/6041?u=bgoodri for how to get it to work with rstan 2.18.1.

…

On Tue, Oct 23, 2018 at 12:56 PM Constantin ***@***.***> wrote: I encountered the same problem today, that I couldn't use Rcpp functions in a package generated with rstanstools::rstan_package_skeleton. I found that removing the R_init_mypackage function fixes the problem of calling the Rcpp functions, because the function is then generated by Rcpp in RcppExports.cpp. But I got another error when I loaded the package that "object 'm' not found" as described in issue #43 <#43>. I solved this issue by adding // [[Rcpp::init]] void rstan_additional_init(DllInfo *dll){ R_useDynamicSymbols(dll, TRUE); // necessary for .onLoad() to work } which makes sure that the dynamic symbol is accessible in the .onLoad function in zzz.R. So in conclusion, replacing the R_init_mypackage function with the code above fixed my problem. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#13 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADOrqjXMmZWvWiv1UtjUaPz54wn-ESpXks5un0osgaJpZM4PlwYA> .

[bart1]

Just a quick comment to add to @const-ae, encountered the same issue and had some trouble getting it to work, in addition to @const-ae suggestion that were required. A few quick notes that would have helped me. You also need to include RcppExports.o and the other Rcpp source files in the sources variable of Makevars (and I assume Makevars.win but have not tested that)
I was also having troubles with including RcppArmadillo, it turns out I needed to import it before importing RcppEigen in the DESCRIPTION file.