rebuild
fun rebuildCandidates(db: BenchmarkResultsDb, expId: Int, problemName: String, problemDefinition: ProblemDefinition, evaluationNumber: Int = 1): List<Solution>(source)
Rebuilds one problem's cell bests as Solution instances from stored rows alone.
Cells whose best was never valid — a failed or never-started member — are excluded, matching what confirmation does with them when a run happens live.
Parameters
db
the results database holding the source experiment
exp
the source experiment's id
problem
the problem whose cells to rebuild
problem
the problem the rebuilt solutions belong to; must be the definition the study ran, since solutions validate their problem identity
evaluation
the clock to stamp the rebuilt solutions with. Selection must be clock-independent, so this exists only so that anything recording a penalized value has a defensible number; it plays no part in the ranking.