An alternative to setRepositories() is use of (the CRAN package)
BiocManager::install("gRbase") instead of install.packages(). BiocManager
installs CRAN packages as well as Bioconductor packages.
Another, more transparent, solution is to use
install.packages("gRbase", repos = BiocManager::repositories())
where the key idea is to include Bioconductor repositories explicitly. These
approaches are preferred to setRepositories(), because of the details of the
twice-yearly Bioconductor release cycle, compared to the annual R release and
patch cycles.
The usual approach to your problem is to move the package to Suggests:. But
then the namespace commands like Imports, and the direct use of imported
package functions, is not possible; you'll need to litter your code with fully
resolved functions (graph::foo() instead of foo()). Also Suggests: is usually
home to packages that have a limited role to play, but that does not seem
likely for RBGL etc in your package.
Also, in implementing this approach one would normally check that the package
were installed, and fail with an error message telling the user how to fix the
problem (e.g., by installing the package). This doesn't really sound like
progress. If you instead try to automatically install the package (in
.onAttach(), I guess was your plan) you'll shortly run into users who need to
use arguments to install.packages() that you have not made available to them.
Your CRAN page took me quickly to your package web site and clear installation
instructions; I do not think use of Bioc packages is a particular barrier to
use.
Martin Morgan
On 10/11/20, 2:52 PM, "R-help on behalf of Søren Højsgaard"
<r-help-boun...@r-project.org on behalf of sor...@math.aau.dk> wrote:
Dear all,
My gRbase package imports functionality from the bioconductor packages
graph, Rgraphviz and RBGL.
To make installation of gRbase easy, I would like to have these
bioconductor packages installed in connection with installation of gRbase, but
to do so the user must use setRepositories() to make sure that R also installs
packages from bioconductor.
Having to call setRepositories causes what can perhaps be called an
(unnecessary?) obstacle. Therefore I have been experimenting with deferring
installation of these bioc-packages until gRbase is loaded the first time using
.onAttach; please see my attempt below.
However, if the bioc-packages are not installed I can not install gRbase so
that does not seem to be a viable approach. (The bioc-packages appear as
Imports: in DESCRIPTION).
Can anyone tell if it is a futile approach and / or perhaps suggest a
solution. (I would guess that there are many CRAN packages that use
bioc-packages, so other people must have faced this challenge before).
Thanks in advance.
Best regards
S�ren
.onAttach<-function(libname, pkgname) {
## package startup check
toinstall=c(
"graph",
"Rgraphviz",
"RBGL"
)
already_installed <- sapply(toinstall, function(pkg)
requireNamespace(pkg, quietly=TRUE))
if (any(!already_installed)){
packageStartupMessage("Need to install the following package(s): ",
toString(toinstall[!already_installed]), "\n")
}
## install if needed
if(!base::all(already_installed)){
if (!requireNamespace("BiocManager", quietly=TRUE))
install.packages("BiocManager")
BiocManager::install(toinstall[!already_installed],
dependencies=TRUE)
}
}
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
