Re: [rules-users] [planner] How to initialize LocalSearchSolverScope ?

Op 16-05-11 01:46, felipebr schreef: Yes, a starting solution is kinda like setting the problem [1]: you load the facts (rooms, timeslots, ... or airplanes, airports, ... or nurses, shifts, ...) from a dataset (coming from a database, xml file, webservice, ...) and give those facts to planner as a "starting solution". But that doesn't mean that starting solution is initialized: For example, just because it now knows the nurses and the shifts, it doesn't have the Assignments yet: which nurse will cover which shifts? The Assignments change during planning, to find the optimal state. The StartingSolutionInitializer's job is to initialize those Assignment as fast as possible, yet still get an ok solution. Usually this uses the simple First-Fit Decreasing algorithm: for example, sort all shifts and then assign them one by one to the best remaining nurse. [2] [1] Separate discussion: Maybe I should rename setStartingSolution(Solution) to setProblem(Solution))? But to do that cleanly, I would have to force the distinction of Problem and Solution on the domain model of my users, so it becomes setProblem(Problem). But as far as I can see, forcing that distinction of Problem and Solution on my users isn't a benefit for them, but instead an annoyance. Feed-back welcome :) [2] I am working on automating some types of StartingSolutionInitializer, so it supports that First-Fit Decreasing algorithm out-of-the-box. -- With kind regards, Geoffrey De Smet