RDataGet gets tabular R datasets from CRAN. It is an alternative to RDatasets.jl, working on demand, rather than bundling data.
The basic usage is similar to RDatasets.jl. You can install it as follows:
Pkg.add(url="https://github.com/frankier/RDataGet.jl.git")
After installing the RDataGet package, you can then load data sets using the
dataset() function, which takes the name of a package and a data set as
arguments:
using RDataGet
harman_political = dataset("psych", "Harman.political")
neuro = dataset("boot", "neuro")
This package currently just downloads source packages from CRAN and loads its dataset into memory in Julia. It does not depend on R itself.
The package has a few limitation, some of which are caused by this design, while others could be addressed in future:
- Does not support built-in R datasets, including the
datasetspackage, only ones which can be downloaded from CRAN - Can only load rda/RData/csv.gz files in the data directory
- As such it does not support packages which generate their data using a build script
- Cannot get any descriptions or further documentation related to the datasets from Julia (maybe TODO but needs .Rd parsing)
- Only supports getting the latest version of each package (TODO)
- Fixed, very-limited caching strategy
- The package index is re-downloaded every time we need to download any package (so as to find the latest version number) (TODO: should be by-default cached per session + longer caching allowed)
- Packages are downloaded exactly once per session, after which the same data is reused until Julia is restarted (TODO: should be customisable for longer caching)