[Rcpp-devel] Strange behavior with boost::program_options

[Watal M. Iwasaki]

Thank you for the quick and kind reply, Kevin! You are right.

I confirmed that the program option object is defined as a local variable
in mylib::Simulation constructor, and is destructed correctly. But I found
that the object is initialized with other objects from several source
files, and they were defined as static variables. Changing them into local
variables has solved the problem. What I didn't know was, therefore, that
static variables do survive after Rcpp function calling. Now I will have to
take a careful look at other variables too.

Thanks a lot!!
Watal



On Wed, Jul 6, 2016 at 3:13 PM, Kevin Ushey <kevinushey at gmail.com> wrote:

> If I had to guess, maybe there is some initialization that occurs when
> your mylib::Simulation object is constructed, and you need to ensure
> that code is only run once? (Perhaps the code that defines the program
> options needs to be hidden behind an initialization flag, or something
> like that)
>
> Nonetheless, this isn't really an Rcpp question, and it's hard to know
> for sure without seeing how your Simulation object is defined.
>
> On Tue, Jul 5, 2016 at 10:56 PM, Watal M. Iwasaki <heavy.watal at gmail.com>
> wrote:
> > Hi all,
> >
> > I have been trying to make an R function to call my C++ simulation
> program
> > in this way:
> > ```c++
> > // [[Rcpp::plugins(cpp14)]]
> > #include <Rcpp.h>
> > #include <mylib.hpp>
> > // [[Rcpp::export]]
> > std::string run_simulation(Rcpp::CharacterVector
> > args=Rcpp::CharacterVector::create()) {
> >     auto vs_args = Rcpp::as<std::vector<std::string>>(args);
> >     mylib::Simulation sim(vs_args);
> >     sim.run();
> >     return sim.results();
> > }
> > ```
> >
> > The class `Simulation` parses arguments with `boost::program_options`
> and is
> > implemented in `libmylib.dylib`, which can be linked with `PKG_LIBS =
> > -lmylib`. Both `devtools::check()` and independent `Rcpp::sourceCpp()`
> can
> > successfully build the function `run_simulation()`. It runs perfectly as
> I
> > intended **only if it is the first call**. The second time I call
> > `run_simulation('-v')`, I get the following error:
> > ```
> > Error: option '-v' is ambiguous and matches different versions of
> > '--verbose'
> > ```
> >
> > Other options like `-h` or `--mutation=0` raise the same error. It
> indicates
> > that every option is re-defined after the first call (maybe?). Calling
> > without any argument is always OK, and an executable `a.out` built with
> > `main.cpp -lmylib` has no problem. I guess the key is how C++ objects are
> > cached in R environments. Is this a bug in Rcpp? Or am I missing
> something?
> >
> > Thanks,
> >
> > --
> > Watal M. Iwasaki / 岩嵜 航
> > SOKENDAI, Graduate University for Advanced Studies,
> > Hayama, Kanagawa 240-0193, Japan
> > +81-46-858-1576
> > https://heavywatal.github.io/
> >
> > _______________________________________________
> > 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
>



-- 
Watal M. Iwasaki / 岩嵜 航
SOKENDAI, Graduate University for Advanced Studies,
Hayama, Kanagawa 240-0193, Japan
+81-46-858-1576
https://heavywatal.github.io/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20160706/424bcc63/attachment.html>