Logged In: YES 
user_id=1289249

I looked over the changes so far and they look reasonable. I
did have a few questions though:

1 Do we not need to Increment/Decrement references to
the code objects? We are using them as keys in the trees as
well as payload data later on, but I don't see how we are
guaranteed that they won't be reaped. (although it seems
unlikely)

2 Is it deliberate to use lsprof.YYY style names for some
of the objects (ie lsprof.Profiler) instead of _lsprof ? (is
this the normal convention?)

3 Do you have a feel for the performance differences
of using the tree instead of the lists? Doing the simple
benchmark test they seemed to be comparable, but
that test is pretty simple. I'm assuming tht in a large
program, the tree approach will be considerably faster?

4 In lsprof.py, is there a reason that the Stats class
does not derive from object?