Start by creating the recursive dependency tree for your target packages.
For example, imagine a scenario where you want to create a repository that consists of the package
foreach and its dependencies.
Start by creating the dependency list:
library("miniCRAN") # use Revolution Analytics CRAN mirror revolution <- c(CRAN = "http://cran.microsoft.com") # Specify list of packages to download pkgs <- c("foreach") pkgList <- pkgDep(pkgs, repos = revolution, type = "source", suggests = FALSE, availPkgs = cranJuly2014) pkgList
##  "foreach" "codetools" "iterators"
Next, create a repository with the function
makeRepo(). In this example, get the required files for
source packages as well as windows binaries:
# Create temporary folder for miniCRAN dir.create(pth <- file.path(tempdir(), "miniCRAN")) # Make repo for source and win.binary makeRepo(pkgList, path = pth, repos = revolution, type = c("source", "win.binary"))
Investigate the repository file structure:
# List all files in miniCRAN list.files(pth, recursive = TRUE, full.names = FALSE)
pkgAvail to list available packages in your repository:
To install packages from a local repository, you need to use the Uniform Resource Identifier (URI) convention
file:/// to point to your file location.
After creating a local miniCRAN repository, additional packages and their dependencies can easily be added. This mechanism can also be used to re-add an existing package to the miniCRAN repo.
# Add new packages (from CRAN) to the miniCRAN repo addPackage("Matrix", path = pth, repos = revolution, type = c("source", "win.binary")) pkgAvail(repos = pth, type = "win.binary")[, c(1:3, 5)]
The value that is returned (invisibly) via
addPackage is the number of packages written to the index file, i.e., the total number of packages in the repo of that type.
To add a specific version of a package from CRAN (or another CRAN-like repository), we can easily download the source packages from the CRAN archives. Dependencies for old package versions cannot be determined automatically and must be specified by the user.
Note: in order to to add binaries of older packages, you will need to download the source and build the binaries on the intended platform yourself. You will need the appropriate R development tools installed in order to build package binaries from source.
# create a data frame with the package and version info oldVers <- data.frame( package = c("foreach", "codetools", "iterators"), version = c("1.4.0", "0.2-7", "1.0.5"), stringsAsFactors = FALSE ) # download old source package version and create repo index addOldPackage(pkgList, path = pth, vers = oldVers$version, repos = revolution, type = "source")
You will get a warning whenever there are multiple versions of a package saved in the repository. Currently, you need to manually remove duplicate versions before rebuilding the repository’s package index.
Note: This last step is important, otherwise you may end up with a repo in an inconsistent state.
# List package versions in the miniCRAN repo (produces warning about duplicates) pkgVersionsSrc <- checkVersions(pkgList, path = pth, type = "source") pkgVersionsBin <- checkVersions(pkgList, path = pth, type = "win.binary") # After inspecting package versions, remove old versions basename(pkgVersionsSrc) # duplicate versions basename(pkgVersionsBin) file.remove(pkgVersionsSrc[c(2,4,6)]) # rebuild the package index after removing duplicate package versions updateRepoIndex(pth, type = c("source", "win.binary"))
To see the updated list of packages available in the miniCRAN repo:
Checking for updated versions of the packages currently stored in the miniCRAN repository:
# Check if updated packages are available oldPackages(path = pth, repos = revolution, type = "source")[, 1:3] # should need update oldPackages(path = pth, repos = revolution, type = "win.binary")[, 1:3] # should be current
Update the versions of the packages currently stored in the miniCRAN repository. By default, a prompt is given to confirm the update for each package. This prompt can be suppressed using
ask = FALSE, which will update all packages. Be careful using this option if you want to keep certain packages at an older version.