> Later the RunTimeStatistics class can be expanded to parse and return other information
> as needed for other tests. Does this strategy sound ok?
Sounds good to me. I was loosely considering doing something like this for a new test that I'm writing for
DERBY-47. In the end I just decided to do a naive string search on the result set stats string to get what I need. But the approach you mention here sounds better.
And when/if the work that Felix Beyer did is contributed, the RunTimeStatistics class could be modified appropriately to take advantage of the new functionality.
So yes, I like this idea