Strings use a much simpler tree, and it only rebalances under extreme conditions. It's not ideal, but strings churn a lot more than appearances so even the small red-black overhead probably wouldn't be worth it. I think a slow sweep through the table looking for isolated cases that could be improved might be more beneficial.

GinjaNinja32 wrote:

Is there any way we can print debug information about the string tree in 510.x? If not, can you add something to 511 (maybe in the SIGUSR2 diagnostics?) so that when that goes stable and we update, we can get some actual data on this, even if it's just average/max height of the tree?

You really ahould update now, because the beta typically includes major fixes to bugs in the stable build, especially late in the beta run.

I don't think it's at all feasible to give useful info about the string tree structure, though.