Skip to content

[Rcpp-devel] Access package installed into user library

4 messages · Thomas Lin Pedersen, Dirk Eddelbuettel

Original
Thomas Lin Pedersen
# 
Hi

While testing my package on an Amazon cloud instance using the Bioconductor AMI (https://www.bioconductor.org/help/bioconductor-cloud-ami/ <https://www.bioconductor.org/help/bioconductor-cloud-ami/>) I ran into the following problem I had not encountered on my work computer (A Mac OS X).

I have a C++ function that makes use of functionality in the igraph package, so to access this I have the following in my function:

Environment igraph("package:igraph");
Function neighbors = igraph["neighbors?];

This worked fine on my work machine but gave the following error when tried on the Amazon instance:

Error: cannot convert to environment

I was able to reduce the error into trying to get access to a package environment that is installed outside the core package library. The instances created using the Bioconductor AMI have the following package storage structure:
[1] "/home/ubuntu/R-libs"      "/usr/local/lib/R/library"

Any package residing in usr/local/lib/R/library will be loaded into the environment fine, while any package in /home/ubuntu/R-libs will throw the error.

Minimal example:

library(Rcpp)

cppFunction('
  void testFail() {
    Environment testEnv("package:boot");
  }
')

cppFunction('
  void testSucces() {
    Environment testEnv("package:stats");
  }
')

testFail() # Will give error
testSucces() # Will work

The session:
R version 3.2.2 (2015-08-14)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 14.04.2 LTS

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=en_US.UTF-8       
 [4] LC_COLLATE=en_US.UTF-8     LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                  LC_ADDRESS=C              
[10] LC_TELEPHONE=C             LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

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

other attached packages:
[1] Rcpp_0.12.1          BiocInstaller_1.20.0

loaded via a namespace (and not attached):
[1] tools_3.2.2

Best

Thomas Lin Pedersen

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20151204/45197ecc/attachment.html>
Dirk Eddelbuettel
#
On 4 December 2015 at 12:31, Thomas Lin Pedersen wrote:
| Hi
| 
| While testing my package on an Amazon cloud instance using the Bioconductor AMI
| (https://www.bioconductor.org/help/bioconductor-cloud-ami/) I ran into the
| following problem I had not encountered on my work computer (A Mac OS X).
| 
| I have a C++ function that makes use of functionality in the igraph package, so
| to access this I have the following in my function:
| 
| Environment igraph("package:igraph");
| Function neighbors = igraph["neighbors?];
| 
| This worked fine on my work machine but gave the following error when tried on
| the Amazon instance:
| 
| 
| Error: cannot convert to environment
| 
| 
| I was able to reduce the error into trying to get access to a package
| environment that is installed outside the core package library. The instances
| created using the Bioconductor AMI have the following package storage
| structure:
| 
| 
| > .libPaths()
| [1] "/home/ubuntu/R-libs"      "/usr/local/lib/R/library"
| 
| 
| Any package residing in usr/local/lib/R/library will be loaded into the
| environment fine, while any package in /home/ubuntu/R-libs will throw the
| error.

So maybe don't install into /home/ubuntu/R-libs ?

There are no hardcoded path in Rcpp. This strikes me as a run-time config
(and hence not an Rcpp question for this list).

Dirk
 
| Minimal example:
| 
| library(Rcpp)
| 
| cppFunction('
|   void testFail() {
|     Environment testEnv("package:boot");
|   }
| ')
| 
| cppFunction('
|   void testSucces() {
|     Environment testEnv("package:stats");
|   }
| ')
| 
| testFail() # Will give error
| testSucces() # Will work
| 
| The session:
| 
| R version 3.2.2 (2015-08-14)
| Platform: x86_64-pc-linux-gnu (64-bit)
| Running under: Ubuntu 14.04.2 LTS
| 
| locale:
|  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=en_US.UTF-8
|  [4] LC_COLLATE=en_US.UTF-8     LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
|  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                  LC_ADDRESS=C
| [10] LC_TELEPHONE=C             LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
| 
| attached base packages:
| [1] stats     graphics  grDevices utils     datasets  methods   base
| 
| other attached packages:
| [1] Rcpp_0.12.1          BiocInstaller_1.20.0
| 
| loaded via a namespace (and not attached):
| [1] tools_3.2.2
| 
| 
| Best
| 
| Thomas Lin Pedersen
| 
| _______________________________________________
| 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
Thomas Lin Pedersen
# 
Ok - (slight embarrassment to follow) I looked at the Environment constructor and found that it basically just called as.environment, which led my to try this out in R. This also failed, but for the reason that the requested package was only loaded, not attached (It?s part of the Import field of my package). calling library() to attach the package solved the issue?

The obvious follow up question is: Is there no way to use R functions from Rcpp if they are not part of the users search path (but part of the package namespace). I generally tend to avoid putting packages in Depends in order to not pollute the users search path.

Thomas
Dirk Eddelbuettel
#
On 4 December 2015 at 14:01, Thomas Lin Pedersen wrote:
| > On 04 Dec 2015, at 13:16, Dirk Eddelbuettel <edd at debian.org> wrote:
| > On 4 December 2015 at 12:31, Thomas Lin Pedersen wrote:
| > | Any package residing in usr/local/lib/R/library will be loaded into the
| > | environment fine, while any package in /home/ubuntu/R-libs will throw the
| > | error.
| > 
| > So maybe don't install into /home/ubuntu/R-libs ?
| > 
| > There are no hardcoded path in Rcpp. This strikes me as a run-time config
| > (and hence not an Rcpp question for this list).
| > 
| > Dirk
| > 
| 
| Ok - (slight embarrassment to follow) I looked at the Environment constructor and found that it basically just called as.environment, which led my to try this out in R. This also failed, but for the reason that the requested package was only loaded, not attached (It?s part of the Import field of my package). calling library() to attach the package solved the issue?
| 
| The obvious follow up question is: Is there no way to use R functions from Rcpp if they are not part of the users search path (but part of the package namespace). I generally tend to avoid putting packages in Depends in order to not pollute the users search path.

Join the r-package-devel list and discuss there.  This is now an R discussion
beyond Rcpp because Rcpp just talks the R session it was called from.

As for current best practices: use Import: with appropriate global (or better
still if you can) per-symbol importFrom(package, funA, funB, funC) in NAMESPACE.

That gets you the symbols and leaves your search path alone. You do not need
a package as the mechanism for this just as you do for Depends:.

Dirk