adhere to .Rbuildignore during downloading include_github()ed packages

I noticed that packages in packrat/src stemming from an include_github() are often a lot larger, and that, in fact, these *.tar.gz appear to include the entire upstream repos, often including lots of stuff that isn't required to build the package (say, some docs/ folder etc.).

With some of the packages I'm include_github()ing, this adds considerably to the already bloated packrat/src, making it hard to, in turn commit that to github.

I understand that, in an ideal world, package maintainers wouldn't have all sort of cruft on their repos, but that just isn't so.

I'm wondering whether it wouldn't help and be safe if packrat::snapshot() simple deleted everything from the github download listed in the respective .Rbuildignore, before tarring the whole thing up.

By definition, that should suffice to build the package from scratch, correct?

Start at packrat::snapshot() and the include_github() download path, focusing on how packages in packrat/src are assembled into *.tar.gz archives. Review the proposed use of each package's .Rbuildignore and verify that ignored content is excluded without preventing the resulting archive from building the package.