How can I pass large arrays between numpy and R?

I'm using python and numpy/scipy to do regex and stemming for a text processing application. But I want to use some of R's statistical packages as well.

What's the best way to pass the data from python to R? (And back?)

Also, I need to backup the array to disk at some point, so I'm open to saving from python and loading th R if that's the best solution. The matrices are pretty big (e.g. 100,000 x 10,000), so using sparse matrices might also be nice.

Apologies if this is a repost. I haven't been able to find anything that puts all these pieces together.