Decision Points 5/22

[rburghol]

• UCI changes (fine)
• Changes to state as object:
  • Much more concise, and future flexible proof
  • Performance issues in numba with classes, even though we currently are not susceptible, found in testing Reading from numpy array in @jitclass method slows down considerably beginning in numba 0.62 and persisting through numba 0.64 numba/numba#10451
• Moving toward IF-THEN in SPECL, will push that in concert with DataMatrix op-model as they will both share the same legacy HSPF testing problem.
• Moving toward integrated looping. Look at Some of the constructs in RQUAL_CLASS, and note their may be a slow-down due to same issue observed with numba 0.62+
• Dev code (newest pandas/numba) pathways: fine as long as it is super easy for us to maintain a dev build mode that works without extra effort.