That’s really good information @tomljh , and very interesting to see the results.
Going back (briefly) to Column Expressions, you are correct re the flow variables, but it could generate a cumulative calculation from which the final row value would be taken. There is no way that this is going to match the performance of your java and python results though ![]()
Just as a comparison, I’ve attached examples (and also I looked at how Math Formula performs) onto your previous workflow,but it doesn’t provide any new ideas for better performance over what you have already done! I’d be interested to know on your system how these different methods perform.
With multiple uses of the same node type, in classic UI, if you turn on “Show Node IDs”, on the “Node” menu, it would make it easier to match the node number to the Timer Info results.
