[Rcpp-devel] devtools::load_all() working differently for Rcpp >= 0.11.0

[Kevin Ushey]

Hi Finlay,

I'm unable to reproduce this on Mac OS X, although I am running the
development version of devtools. Can you try that?

That said, if you still have this problem after updating I would
recommend posting it at https://github.com/hadley/devtools/issues as
even if there is a problem, I don't think it's an Rcpp one.

Cheers,
Kevin

> sessionInfo()
R Under development (unstable) (2014-07-01 r66058)
Platform: x86_64-apple-darwin13.2.0 (64-bit)

locale:
[1] en_CA.UTF-8/en_CA.UTF-8/en_CA.UTF-8/C/en_CA.UTF-8/en_CA.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] NewPackage_1.0       Rcpp_0.11.2          knitr_1.6
[4] devtools_1.5.0.99    BiocInstaller_1.15.3

On Thu, Jul 3, 2014 at 2:27 AM, Finlay Scott <drfinlayscott at gmail.com> wrote:
> Hi,
>
> First of all, thanks for all the hard work that has been put into Rcpp. It
> really is a great package.
>
> I'm redeveloping a package I was working on last year that uses Rcpp,
> particularly attributes. I used the devtools package to help with the
> development process. My general development approach was to add or change a
> function with attributes in the C++ source code, and then reload the package
> using the devtools function load_all(). This would recompile the source
> code, and update the RcppExports.R and .cpp files as necessary. The new
> exported function would then be immediately available to the current R
> session.
>
> I have now updated all my R packages and am using Rcpp 0.11.2.
> Unfortunately, the behaviour of load_all() is now different. For example, if
> I start a new package from scratch:
>
> library(Rcpp)
> Rcpp.package.skeleton("NewPackage", attributes = TRUE)
>
> # Then load devtools and use load_all()
>
> library(devtools)
> load_all("NewPackage")
>
> # The demo function works just fine
>
> rcpp_hello_world()
>
> # I make a change to the source code of the rcpp_hello_world() function
> # For example, changing one of the values in the returned list
> # I rerun load_all()
>
> load_all("NewPackage")
>
> # The package recompiles, and I can see that the resulting *.o and *.so have
> been updated.
> # However, calling rcpp_hello_world() still returns the old output, i.e. the
> function called from R does not seem to have been updated.
> # Similarly, if I add a new simple function with attributes to
> rcpp_hello_world.cpp
> # For example:
>
> // [[Rcpp::export]]
> int do_nothing(){
>     return 0;
> }
>
> # If I call load_all() again, the package compiles and the RcppExports.R and
> .cpp files have been updated to include this new function (i.e. it looks
> like compileAttributes() has been called by load_all()).
> # However, if I try to call do_nothing() from R I get:
>
> Error in .Call("NewPackage_do_nothing", PACKAGE = "NewPackage") :
>   "NewPackage_do_nothing" not available for .Call() for package "NewPackage"
>
> So it looks like load_all() is updating and recompiling the package
> accordingly, but it is not being 'reloaded' in the current R session.
>
> I ran a test using Rcpp 0.10.6 and the latest devtools (1.5) and the above
> tests passed (i.e. after running load_all() the new function was available
> to the current R session), so the problem seems to be with Rcpp > 0.11.0
>
> Does anyone else have experience of this, and a workaround?
>
> Apologies if this is more of a devtools question than an Rcpp one. I will
> repost at the devtools mailing list if necessary.
>
> I am using Ubuntu 14.04 , R 3.1.0; x86_64, devtools 1.5
>
> Thanks
>
> Finlay
>
>
> _______________________________________________
> 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