I have a set of R-functions that I have developed into a local package. The problem is that it requires potentially 15Gb of data for the functions to interrogate. I can break the data into 85 175Mb files and have the user select which one they want when they run the package. No single user is likely to require more than 1 or 2 of the 85 data sets.
My question is how to publish a package of this nature. Should I have the 85 files sitting on a server and have the user download all of them prior to loading the package and then store them locally once and for all? This will make it hard to compile the package for CRAN. Or should I just have the single file download when the user specifies the particular file they need?
Are there other places where I could upload everything, including all the data, that users could easily access – for instance Github?
