In order to implement precision reduction, we had to push computing the value in buckets down to the backend. That logic exist in 2 places now.
2 solutions
This now live in the backend. That is my default assumption as it allows us flexibilty later if we find it was a bad idea
It live in hyper_register and we call it from the backend. This irk me architecturally as it generate a 2 way dependency but it would centralise the logic.
I would go for 1 but if someone has arguments for 2 i would take
In order to implement precision reduction, we had to push computing the value in buckets down to the backend. That logic exist in 2 places now.
2 solutions
I would go for 1 but if someone has arguments for 2 i would take