Commits on Aug 13, 2021

1. Fix probabilistic subsampling for small values …

   Fixes a regression from the rewrite of augur filter where probabilistic
   subsampling for small values of `--subsample-max-sequences` could
   randomly select zero strains and randomly cause our CI tests to fail.
   Prior to the rewrite of augur filter and introduction of priority
   queues, we fixed this issue by repeatedly attempting to calculate
   sequences per group that summed to an integer value greater than zero.
   However, the way I implemented random queue sizes inside the `PriorityQueue`
   class in the rewrite prevented me from using a similar "multiple
   attempts" approach.
   
   This commit redesigns the way we create priority queues. In the case
   where we already know the number of sequences per group in the first
   pass, we create an appropriately-sized priority queue for each group as
   we encounter it. There is no possibility that the sum of these queue
   sizes could be zero.
   
   In the case where we need to calculate the number of sequences per group
   from the first pass, we already know all possible groups and can create
   their priority queues in bulk. The new `create_queues_by_group` function
   allows us to create fixed-sized or randomly-sized queues and also make
   multiple attempts when queue sizes sum to zero. As a result, the
   `PriorityQueue` class is much simpler (it requires no logic about random
   max sizes) and we can actually test the fixed and random behaviors more
   carefully with doctests for `create_queues_by_group`.

   

   huddlej committed Aug 13, 2021
   Configuration menu

   • View commit details
   • Copy full SHA for 5db04fc
   • Browse repository at this point

   Copy the full SHA

   5db04fc View commit details

   Browse the repository at this point in the history