Comparability of compound value sets
Closed this issue · 1 comments
How do you plan to make compound value sets comparable?
I think for the analysis to work and not result in immediate exploits it always has to choose the smallest value when there are multiple options. But "smallest value" is not really defined, when you have a vector of values and there's an infinite ways to define it.
E.g. compare weighted sum, distance to 0, compare compounds in a specific order, choose the minimum for each compound type independently, ...
I think at the moment this problem is avoided by just always using the first calculated value, but from my experience this will not always find the smallest number and makes the analyzer recipe-order dependent.
They are by definition comparable.
The Compoundtype behaviour determines how those comparisons happen.