Implementation log

How does the existing implementation in supar use the buckets of lengths? Does this logic live in the batch sampler?