Dispatch brief
Evan, a dispatcher at Willow Bulk, needs an auditable scheduling recommendation from a manifest of whole-number load observations. Capacity is positive and arithmetic stays within int. Count every row, including duplicate observations. Reject negative loads and observations above capacity. Valid weight can be split across trips. Deliver a decision code plus accepted rows, rejected rows, accepted load and estimated trips. The last task checks both the recommendation and its evidence.
A dispatch decision needs the evidence that produced it. This case begins with a count of accepted load observations and gradually adds rejected rows, accepted weight and a trip estimate. Loads represent divisible bulk material, so valid rows can be repacked together across vehicle trips. A row exactly at capacity is valid. Negative loads and loads above capacity need review.
Keep one Evidence object and preserve its earlier output fields as the report grows. The final recommendation gives review precedence over the trip count: invalid source rows require attention even when the accepted weight fits in one trip. With no invalid rows, choose a single run for at most one trip and split runs otherwise. An empty manifest therefore needs zero trips and can use the single-run decision code without implying that a vehicle must depart.