datahike

whilo 2026-06-30T18:14:47.677149Z

There was a gap in the planner for function calls, it could not estimate cardinality and cost of them, this is now possible with https://github.com/replikativ/datahike/blob/main/doc/query-engine.md#function-cost-model-plancljc. Without the planner you could control this by manually ordering the clauses (you still can when you turn of the planner completely, but this puts all the burden of finding the optimal query plan on the programmer). I tested this with a set of graph algorithms that are now also https://github.com/replikativ/datahike/blob/main/doc/graph-algorithms.md. I also had graph based https://github.com/replikativ/datahike/blob/main/doc/oddball-anomaly-detection.md still lying around from the earlier work Stratum's anomaly detection. Feedback is highly appreciated!