Class FlowGraphEndPoint.Throwing

java.lang.Object
de.xima.fc.mdl.workflow.FlowGraphEndPoint
de.xima.fc.mdl.workflow.FlowGraphEndPoint.Throwing
All Implemented Interfaces:
INodeKeyProviding, Serializable
Enclosing class:
FlowGraphEndPoint
public static final class FlowGraphEndPoint.Throwing extends FlowGraphEndPoint
A special type of FlowGraphEndPoint, used when analyzing the control flow of a workflow processing chain.

A throwing end point is a node that can complete abruptly by throwing an exception and does not always transfer control to the node or a child of the node when it completes abruptly by throwing an exception. For example, most business logic actions can throw exceptions, in additional to also being able to complete normally. Special actions such as the throw statement always complete abruptly by throwing an exception.

More formally, given a syntax tree of nodes representing the parent-child relationship, the throwing end points of a node n are the set of all nodes {m} such that all the following conditions hold:

  • (a) m is in the subtree of n,
  • (b) m can potentially (but might not always) complete abruptly by throwing an exception; and
  • (c) m does not always transfer control to a node in the subtree of n when m completes abruptly by throwing an exception.

The subtree of a node n is set of all nodes {m} that are either equal to n or have n as one of their parents.

Condition (a) follows from the hierarchical execution principle of the workflow engine. A node, when asked to execute, may choose to execute some of its children in its subtree depending on the circumstances (such as an if-else statement that execute one of its two children depending on the outcome of a condition); but may never execute any other nodes outside its subtree. This implies that nodes can never transfer control to any node outside its subtree.

Condition (b) allows for other types of completion. A node is allowed to complete in other manners depending on the circumstances, such as a business logic action that usually completes normally unless there is an error.

Condition (c) is required for cases such as a try-catch statements, where no node within the try block can be a throwing end point with respect to the try-catch statement, as control is always transferred to the catch block when a node within the try block completes abruptly by throwing an exception.

Since:
8.2.0
Author:
XIMA MEDIA GmbH
See Also: