Skip to content

R CMD check with non-standard .libPaths

2 messages · Ben Bolker, Dirk Eddelbuettel

Original
Ben Bolker
# 
Does anyone have advice on how to instruct R CMD check to use a
non-standard set of libraries?  Here's the situation:

  I'm trying to do some automated checking on package dependencies of a
package I maintain.  In order to do that I've written code that takes
the list of the dependent packages and for each package (1) downloads
the most recent/available .tar.gz file; (2) installs the "Suggests:" and
"Depends:" packages for the package that are not already installed; (3)
runs R CMD check and stores the output.

  I wanted to do this in a way that would not necessarily bloat my base
installation, so I wanted to do step #2 into a new library.  Once I've
figured out what the missing dependencies are (depMiss), I

 install.packages(depMiss,lib=libdir)

Then I run R CMD check as follows.

   ss <- suppressWarnings(system(
   paste("export R_LIBS=./library; R CMD check",
               file.path(tarballdir,tn)),
                 intern=TRUE))

However, this only seems to work partially.  It does prevent the check
from failing with an error that the package doesn't exist: *BUT* when
the examples are actually run, I get results like this (this is from the
'agridat' package, which "Suggests:" the hglm package, which has hence
been installed in the 'libdir' directory).

[168] "168: Loading required package: hglm"

[169] "169: Warning in library(package, lib.loc = lib.loc,
character.only = TRUE, logical.return = TRUE,  :"
[170] "170:   there is no package called ?hglm?"

  Checking installed.packages() shows that hglm is indeed installed in
the appropriate directory.  agridat "Suggests:" hglm.  The
"crowder.germination" example in agridat tries require(hglm) and fails;
it therefore doesn't fit the relevant HGLM model -- when the example
tries to reference this model a few lines later, the example fails.
Admittedly this could be seen as a bug in the example (it shouldn't try
to access a model it knows it can't fit), but I wonder if there's a way
I can get the examples run to see the non-standard package location.

   I could (I guess) modify my .Rprofile temporarily ... ?  But I'm
curious if there's a right way to do this ...

  thanks
    Ben Bolker
Dirk Eddelbuettel
#
On 14 April 2012 at 20:45, Ben Bolker wrote:
| 
|   Does anyone have advice on how to instruct R CMD check to use a
| non-standard set of libraries?  Here's the situation:

One way around is something like this:

edd at max:~$ tail -5 .R/check.Renviron

# edd Apr 2003  Allow local install in /usr/local, also add a directory for
#               Debian packaged CRAN packages, and finally the default dir 
# edd Jul 2007  Now use R_LIBS_SITE, not R_LIBS
R_LIBS_SITE=${R_LIBS_SITE-'/usr/local/lib/R/site-library:/usr/lib/R/site-library:/usr/lib/R/library'}
edd at max:~$ 

The file check.Renviron sets up R environment variables just for R CMD
check.  [ And IIRC there is also build.Renviron ]

|   I'm trying to do some automated checking on package dependencies of a
| package I maintain.  In order to do that I've written code that takes
| the list of the dependent packages and for each package (1) downloads
| the most recent/available .tar.gz file; (2) installs the "Suggests:" and
| "Depends:" packages for the package that are not already installed; (3)
| runs R CMD check and stores the output.
| 
|   I wanted to do this in a way that would not necessarily bloat my base
| installation, so I wanted to do step #2 into a new library.  Once I've
| figured out what the missing dependencies are (depMiss), I

So you could have a script setting the check.Renviron up alongside the
build-up and later tear-down of you test setup.

Would be good to publish such a set of scripts. I hacked up something much
smaller than that and much more ad-hoc too to test the reverse depends of one
of my package before uploading it.

Dirk
 


| 
|  install.packages(depMiss,lib=libdir)
| 
| Then I run R CMD check as follows.
| 
|    ss <- suppressWarnings(system(
|    paste("export R_LIBS=./library; R CMD check",
|                file.path(tarballdir,tn)),
|                  intern=TRUE))
| 
| However, this only seems to work partially.  It does prevent the check
| from failing with an error that the package doesn't exist: *BUT* when
| the examples are actually run, I get results like this (this is from the
| 'agridat' package, which "Suggests:" the hglm package, which has hence
| been installed in the 'libdir' directory).
| 
| [168] "168: Loading required package: hglm"
| 
| [169] "169: Warning in library(package, lib.loc = lib.loc,
| character.only = TRUE, logical.return = TRUE,  :"
| [170] "170:   there is no package called ?hglm?"
| 
|   Checking installed.packages() shows that hglm is indeed installed in
| the appropriate directory.  agridat "Suggests:" hglm.  The
| "crowder.germination" example in agridat tries require(hglm) and fails;
| it therefore doesn't fit the relevant HGLM model -- when the example
| tries to reference this model a few lines later, the example fails.
| Admittedly this could be seen as a bug in the example (it shouldn't try
| to access a model it knows it can't fit), but I wonder if there's a way
| I can get the examples run to see the non-standard package location.
| 
|    I could (I guess) modify my .Rprofile temporarily ... ?  But I'm
| curious if there's a right way to do this ...
| 
|   thanks
|     Ben Bolker
| 
| ______________________________________________
| R-devel at r-project.org mailing list
| https://stat.ethz.ch/mailman/listinfo/r-devel