Nullable + Chained planning variable (for unassigned tasks)

[Southclaws]

Is your feature request related to a problem? Please describe.

So this is something called out in the docs and code via exception, but I was wondering essentially what's the blocker for implementing this.

Currently chained planning variables are not compatible with nullable.

java.lang.IllegalArgumentException: The entityClass (class ...solver.productionschedule.AssemblyDomain$Task) has a @PlanningVariable annotated property (previousStandstill) with chained (true), which is not compatible with nullable (true).

Describe the solution you'd like

I'm unfamiliar with the inner workings of Optaplanner and how chained-variables work, or even why they're special compared to regular planning variables. So, it would be a "nice to have" if I could model "unassigned" (impossible to assign, due to line/machine failure that day etc) but maybe this was omitted due to complexity, which I'd completely understand!

Describe alternatives you've considered

For now I've just done a phantom "Workcentre" (fake production line) where unassigned tasks are "scheduled" and post-processed this to mark them as unassigned so the scheduler can manually push these next-day or deal with it some other way. This works fine so it's not a huge issue.

[triceo]

@Southclaws Hello and thanks for reaching out.

Chained variable is getting replaced by list variable, which presents an overall simpler and better programming model. Chained variable is no longer getting improved, and will be removed in the next major version of Timefold Solver.

If you're still using chained variables, you should start planning your migration away from them. List variable has full support for unassigned values.