The drat package makes it trivially easy to deploy package repositories. There are essentially just two ways to use a package repository:
The core motivation for drat comes from GitHub and its implied web server. As you may know, any GitHub user (or organization) can enable a website for a given repository. All it takes is to create a git branch named
gh-pages. After that, a website is automatically visible.
To make this more explicit, consider a hypothetical user John with account
johndoe. Once John creates a repo
foo and in it a branch
gh-pages, he will have a web address
http://johndoe.github.io/foo for this repo.
More formally, for a user
USER, and a given repo named
drat, we can always assume
So for you as a package author with a given GitHub account, all that is needed is a repository named
drat with a
gh-pages branch. If you are familiar with
git at the command-line, you can just create the branch (and the
src/contrib/ directory structure in it; see below).
If you are less familiar with git, a really easy shortcut is to simply fork the actual drat repo. It contains the drat source code which you could keep, or remove. The fork only serves to set up the required
gh-pages branch, and the
We can now assume that you have a local git repository named
drat with a subdirectory
src containing a further subdirectory
You are now ready to insert a package into it. For simplicit, let us assume the package is named
myPkg and is at version 0.5. So
R CMD build created a file
the source package will be copied into the default drat repo at location
~/git/drat. Should your git repository checkouts live in a different place on your machine, just specify this either via the
options() entry “dratRepo” or directly:
In either case, the package will be copied into the repo, and the PACKAGES file will be updated.
Lastly, if you have
git (the command-line tool) or the wicked git2r package installed, then you can also use the
commit=TRUE option to have the new files added and committed. Neither of these variant pushes, so that last step is left to the user (as it commonly requires authentication).
Colin Gillespie has provided a nice walk-through of how to have Travis CI automagically push packages into a drat repo. This is included as another (currently work-in-progress) vignette entitled Combining Drat And Travis which can be found in the drat package just like this vignette.
Use of drat is not limited to GitHub. Any server you can
is suitable. A common use case may be a local repository within a work group or deparment, meant to be locally accessible but not from an outside network.
This is similar to the usage described above. Suppose that you are part of groupABC which has access to directory on shared filesystem somewhere, say under
/nfs/groups/groupABC/ where you created a directory
drat within a directory
R. We once again require that the resulting directory
/nfs/groups/groupABC/R/drat contains a
src/contrib directory structure.
Hence, the following command would copy the package and update the index files:
This updates the PACKAGES file (and its compressed variant) after which the repository is ready to serve files. See the companion vignette for how to deploy it. Note that the location URL should begin with
drat permits package authors to add packages very easily to R package repositories. These repositories can be public, and GitHub provides a very natural option to serve a package repository via the web server based on the
gh-pages branch of a drat.
Repositories can also be local (and private) as well: all that drat requires to add packages is write access to a directory.
Lastly, serving that directory as a repository then requires a web server (easiest via the automatic GitHub repo option) or other file access. How to access packages from drat repository is described in the companion vignette.