We should be able to avoid that chunk of state, along with Values::Id, if we cache these in the optimizer instead of in the Factor. It means linearizing a factor outside an optimizer will never benefit from the cache unless the user caches the index entries themselves, but that's probably fine
We should be able to avoid that chunk of state, along with
Values::Id, if we cache these in the optimizer instead of in the Factor. It means linearizing a factor outside an optimizer will never benefit from the cache unless the user caches the index entries themselves, but that's probably fine