At runtime, there may be many potential paths through a BPEL process, controlled by conditional statements such as switch or while activities. Typically, the business rules that govern which path to take at any given point are written as XPath expressions embedded within the appropriate activity.
Although this is an acceptable approach, we often find that while the process itself may be relatively static, the business rules embedded within the activities may change on a more frequent basis. This will require us to update the BPEL process and redeploy it, even though the process flow itself hasn’t changed.
In addition, by embedding the rule directly within the decision point, we often end up having to reimplement the same rule every time it is used, either within the same process or across multiple processes. Apart from being inefficient, this can lead to inconsistent implementations of the rules, as well as requiring us to update the rules in multiple places every time it changes.
The Oracle Business Rules engine that comes as part of the SOA Suite provides a declarative mechanism for defining business rules externally to our application. This not only ensures that each rule is used in a consistent fashion, but in addition, it makes it simpler and quicker to modify. We only have to modify a rule once and can do this with almost immediate effect, thus increasing the agility of our solution.
For those of you familiar with 10gR3, you will notice that JDeveloper comes with a new rules editor which is a lot more intuitive and simpler to use than the old browser-based editor. In addition, 11gR1 introduces decision tables, which provide a spreadsheet-like format for defining rules. While still very much a developer-oriented tool, these improvements make the tool a lot friendlier for business analysts, allowing them to better understand the rules that have been written as well as make simple changes.
In this article, we will introduce the new rules editor and look at how we can use it to define a decisions service to automate the approval of leave requests. Then, once we’ve done this, we’ll see how to invoke the rule from the leave approval BPEL process. We will first implement these as a standard set of rules and then examine how we can simplify these rules by using a decision table.
Business rule concepts
Before we implement our first rule, let’s briefly introduce the key components which make up a business rule. These are:
- Facts: Represent the data or business objects that rules are applied to.
- Rules: A rule consists of two parts, namely, an IF part that consists of one or more tests to be applied to a fact(s), and a THEN part that lists the actions to be carried out, should the test evaluate to true.
- Rule Set: As the name implies, it is just a set of one or more related rules that are designed to work together.
- Dictionary: A dictionary is the container of all components that make up a business rule. It holds all the Facts, Rule Sets, and Rules for a business rule.
In addition, a dictionary may also contain decision tables, functions, variables, and constraints. We will introduce these in more detail later in this article.
To execute a business rule, you assert (submit) one or more facts to the rules engine. It will apply the rules to the facts, that is, each fact will be tested against the IF part of the rule, and if it evaluates to true, then it will perform the specified actions for that fact. This may result in the creation of new facts or the modification of existing facts (which may result in further rule evaluation).
XML facts
The rule engine supports four types of facts: Java Facts, XML Facts, RL Facts, and ADF Facts. The type of fact that you want to use typically depends on the context in which you will be using the rules engine.
For example, if you are calling the rule engine from Java, then you would work with Java Facts as this provides a more integrated way of combining the two components. As we are using the rule engine within a composite, it makes sense to use XML facts. The rule editor uses XML schemas to generate JAXB 2.0 classes, which are then imported to implement the corresponding XML facts. Using JAXB, particularly when used in conjunction with BPEL, places a number of constraints on how we define our XML schemas, including:
- Within BPEL, you can only define variables based on globally defined elements. Thus all input and output facts passed to the decision service must be defined as global elements within our XML schemas.
- When defining the input and output facts for any complexType (for example, tLeaveRequest), there can only be one global element of that type (for example, leaveRequest).
- The element naming convention for JAXB means that elements or types with underscores in their names can cause compilation errors.
Decision services
To invoke a business rule within a composite, we need to go through a number of steps. First, we must create a session with the rules engine, then we can assert one or more facts, before executing the ruleset and finally we can retrieve the results.
We do this via a decision service (or function). This is essentially a web-service wrapper around a rules dictionary, which takes care of managing the session with the rules engine as well as governing which ruleset we wish to apply.
The wrapper allows a composite to assert one or more facts, execute a ruleset(s) against the asserted facts, retrieve the results, and then reset the session. This can be done within a single invocation of an operation or over multiple operations.
Leave approval business rule
For our first rule, we are going to build on Adding in Human Workflow. It’s a simple process requiring every leave request to go to an individual’s manager for approval. However, what we would like is a rule that automatically approves a request as long as it meets certain company guidelines.
To begin with, we will write a simple rule to automatically approve a leave request that is of the type Vacation and only for one day’s duration. This is a pretty trivial example, but once we’ve done this, we will look at how to extend this rule to handle more complex examples.
