Rabo-rules

Rabo-Rules is a forward chaining rule engine, built in Scala. The engine comes with a DSL with which derivations can be modelled as small steps that combine into large derived networks of facts.

The project started with a Dutch DSL to facilitate bankers writing mortgage-related calculations. The development of an English DSL is currently in progress (see Future Work).

Getting Started

The minimal requirement to get started is to add the artifact containing the rule engine to your project. Below are snippets for SBT and Maven. Be sure to check if any newer versions are available and adjust the snippets accordingly.

SBT - build.sbt

libraryDependencies += "nl.rabobank.oss.rules" %% "rule-engine" % "0.6.0"

Maven - pom.xml

<dependency>
	<groupId>nl.rabobank.oss.rules</groupId>
	<artifactId>rule-engine_2.11</artifactId>
	<version>0.6.0</version>
</dependency>

Concepts

The rule engine uses a Fact as the base for all of its derivations. A Fact defines its name and type so it can be referenced in evaluations down the road. Facts are best defined inside a Glossary:

object MyGlossary extends Glossary {
	val factA = defineFact[Int]
	val factB = defineFact[Int]
	val factC = defineFact[Int]
}

Note: the name of the fact is automatically deduced from the variable it is assigned to. defineFact has an optional parameter for a description of the fact.

Using this glossary, it is now possible to define derivations. The Scala Rules DSL provides an easy way to express how facts interact and come together to form your logic. To enable the DSL, create a class that extends Berekening:

import nl.rabobank.oss.rules.dsl.nl.grammar._
import MyGlossary._

class MyArithmetics extends Berekening (
	Gegeven (factA > 0) Bereken factC is factB - factA 
)

Note the parenthesis behind Berekening, using braces won’t work. The Berekening constructor requires a series of DslDerivations, using braces causes this argument to be an empty list and yields you no executable derivations.

Note 2: for more information about the possibilities of the DSL, see the Wiki page about it

The two listings above are actually all you need to define your calculations, validations or evaluations. The engine will have enough information to start working for you. Only one thing is still missing for the scenario to make sense to you: values.

The engine requires you to construct a Context mapping a set of initial Facts to their values. When you have that, you can let the engine do the rest:

val initialContext: Context = Map(
  factA -> 4,
  factB -> 10
)
val derivations: List[Derivation] = new MyArithmetics().berekeningen

val resultContext: Context = FactEngine.runNormalDerivations(initialContext, derivations)

println(PrettyPrinter.printContext(resultContext))

Executing the above code will yield the following:

Values in context:
  factA = 4
  factB = 10
  factC = 6
Done.

Debugging

If you want to see exactly what the engine is doing, you can replace the runNormalDerivations with runDebugDerivations. The return type of that function is a tuple containing the resulting Context and a list of Step objects. The latter describe exactly what actions the engine performed and why:

val initialContext: Context = Context(
  factA -> 4,
  factB -> 10
)
val derivations: List[Derivation] = new MyArithmetics().berekeningen

val (resultContext, steps) = FactEngine.runDebugDerivations(initialContext, derivations)

println(PrettyPrinter.printSteps(steps))

Executing this snippet will yield an overview of the steps taken by the engine:

Steps taken:
 * Evaluate: factC
   * Result: Evaluated
   * Change: Map(factC -> 6)
Done.

For each step there is the Fact that was about to be evaluated (factC). Next is the status of the evaluation. The engine might indicate that the condition was false and thus the evaluation was skipped. Finally the change field shows what is added to the Context as a result of this evaluation.

Going Hardcore

If you have complicated tasks to perform, or our DSL simply does not fit your needs, you can write custom evaluations.

The DSL in the Berekening class yields a list of Derivation-objects. Constructing one yourself is not difficult, but requires some explanation:

case class DefaultDerivation(
  input: List[Fact[Any]], 
  output: Fact[Any], 
  condition: Context => Boolean, 
  operation: Evaluation[Any]) extends Derivation

For a derivation to work, the engine must know all the Facts it uses as input. The input parameter requires you to provide these Facts as a list.

The output allows the engine to store the result of the derivation in the Context. The output fact is the key for this.

The condition is a function to decide whether the operation should be executed. The function is provided with the current Context containing all values known at the point of execution. Returning true causes the operation to be executed, false will cause it to be skipped.

Finally, the operation is the function which will result in the value assigned to the output fact. You need to instantiate or extend any of the available Evaluation classes.