Chained Variables / Anchor Variable and Error: Only a chained variable supports a singleton inverse - kotlin

Question

Preface: Google+SO+docu search did not appear to give relevant information.
Domain Model:
My domain model tries to picture an ProductionPlan, containing a List of Machines.
Each Machine has a list of chained jobs, thus a job which has a getNextEntry():Job Method, creating a list of jobs.
I tried to solve this problem with chained PlanningVariables, but apparently do not understand the concept of chainedVariables/shadowVariables/anchorVariables.
With my understanding, all jobs are getting chained and the anchorShadowVariable points to the beginning of the list, thus the machine.
To implement chaining, Job and Machine need to implement an interface or extend superclass, thus i created ChainSuperClass. Additionally i do not know if i have to overwrite getter/setter to set Annotations in Machine-Class, my guess was that because Machine extends ChainSuperClass, those Annotations carry over.
Edit: Kotlin specific improvements are also appreciated.
The full error log of my code execution is now:
Exception in thread "main" java.lang.IllegalArgumentException: The entityClass (class optaplanner.productionPlan.domain.ChainSuperClass) has a InverseRelationShadowVariable annotated property (nextEntry) which does not return a Collection with sourceVariableName (machine) which is not chained. Only a chained variable supports a singleton inverse.
ChainSuperClass:
#PlanningEntity
abstract class ChainSuperClass {
#PlanningId
open val id = Random.nextInt().toString()
#InverseRelationShadowVariable(sourceVariableName = "machine")
abstract fun getNextEntry(): Job?
abstract fun setNextEntry(job: Job)
}
Job:
#PlanningEntity
class Job(
val jobType: JobType,
val itemNumber: String,
val orderNumber: String,
val setupTime: Int,
val productionTime: Int
) : ChainSuperClass() {
#AnchorShadowVariable(sourceVariableName = "machine")
var machine: Machine? = null
private var nextEntry: Job? = null
#PlanningVariable(
valueRangeProviderRefs = ["jobList"],
graphType = PlanningVariableGraphType.CHAINED
)
override fun getNextEntry(): Job? {
return nextEntry
}
override fun setNextEntry(job: Job) {
this.nextEntry = nextEntry
}
}
Machine:
class Machine(override val id: String, val jobTypes: List<JobType>) : ChainSuperClass() {
private var nextEntry: Job? = null
override fun setNextEntry(job: Job) {
this.nextEntry = job
}
override fun getNextEntry(): Job? {
return nextEntry!!
}
}

Answer 1

I think the most important thing to realize with chained variables is this: When you have an entity, say Job A and solver assigns a value (job/machine) to its variable, it's not like the chain is being built forward starting from Job A. It's the other way around. By assigning a value to Job A's planning variable, Job A gets connected at the end of an existing chain.
Please take a look at the documentation to find more details about chaining and examples of valid chains.
By having understood this it should be clear that Job's planning variable name should be something like previousJobOrMachine (you'll probably want something simpler, for example previousStep) whereas the nextJob property is an inverse relation shadow variable derived from that (so when Job X gets connected to an existing chain ending with Job C by assigning Job X.previousStep=Job C, an inverse relation is established automatically: Job C.nextJob=Job X).
Based on that information, your model should look more like this:
#PlanningEntity
abstract class ChainSuperClass {
#PlanningId
open val id = Random.nextInt().toString()
// changed sourceVariableName to point to the planning variable
#InverseRelationShadowVariable(sourceVariableName = "previousStep")
abstract fun getNextEntry(): Job?
abstract fun setNextEntry(job: Job)
}
#PlanningEntity
class Job(
val jobType: JobType,
val itemNumber: String,
val orderNumber: String,
val setupTime: Int,
val productionTime: Int
) : ChainSuperClass() {
// changed sourceVariableName to point to the planning variable
#AnchorShadowVariable(sourceVariableName = "previousStep")
var machine: Machine? = null
// added planning variable
private var previousStep: ChainSuperClass? = null
private var nextEntry: Job? = null
#PlanningVariable(
// added machineList value range provider
valueRangeProviderRefs = ["jobList", "machineList"],
graphType = PlanningVariableGraphType.CHAINED
)
// getter for the new planning variable
fun getPreviousStep(): ChainSuperClass {
return previousStep
}
override fun getNextEntry(): Job? {
return nextEntry
}
override fun setNextEntry(job: Job) {
this.nextEntry = nextEntry
}
}
Notice that I added machineList as the source of possible values of the previousStep planning variable as previous step might be either a Job at the end of a non-empty chain or a Machine representing an empty chain.
Your domain is very similar to the vehicle routing domain, where you can see a working example of chaining. With good understanding of chaining principles you should be able to replicate it on your domain.