...is really cool, but I don't see how to modify it for the case where the domain of the permutation has a cardinality that is not a power of 2, as it is in my case (cardinality = 900000).
---
(In the crypto world there are "format preserving encryption" techniques that probably could do what I want to do, but their focus on cryptographic strength makes learning and implementing them tough going, plus, the performance will probably be poor, since high workloads are an asset for such crypto applications. Since cryptographic strength is not something I need, I'm trying to find non-crypt-grade alternatives.)