Better seeding?

[eddelbuettel]

On a "empty" machine it is tedious to get the build dependencies installed. Something approximative:

pkg <- ...    # some package
IP <- installed.packages()
AP <- available.packages()
p <- tools::package_dependencies(package=pkg, recursive=TRUE, db=AP, reverse=TRUE)[[1]]

and p can then be fed to install.packages(), maybe after checking with IP first. At that point it
becomes a little 'empirical' to install (OS-level) build-dependencies as needed.

What would be really nice: pulling in the OS-level depends from JSON DBs that @gaborcsardi has.