Identifying Inspectors to Mine Models of Object Behavior

In object oriented programming languages, classes are used to incorporate state (fields) and behavior (methods) that modifies the state. Typically, only a subset of a classes’ methods actually modifies the state. Methods that don’t modify the state are called side– effect free or pure methods. We propose to use purity information to classify methods as inspectors (methods that reveal information about an object’s state) or mutators (methods that change state). An inspector is a pure method that takes no parameters and has a return type other than void. Inspectors can often provide useful abstractions over the internal state of an object. For example, method isEmpty() in class Vector provides an abstraction over the internal size attribute (namely that size=0). The benefit of using inspectors is that they provide an abstract characterization of an object’s state that does not rely on internal implementation details like fields. We use inspector and mutator methods to dynamically mine models of object behavior. These models show the effect of mutator invocations on an object’s externally visible state (captured by calling all inspectors). Thus, the models are not only meaningful, but also aligned with the view of the user. In the remainder of this paper we show how common inspector methods are in the wild, and illustrate object behavior models using a real example mined from the execution of the JAVA Vector class.