Great! I was hoping you could take a look at this thread.

I first thought of a CLR aggregate because I was thinking about the second part of the problem (aggregating the split tokens).

Very nicely done, Paul.