Metrics-Based Detection of Micro Patterns

Micro patterns are class-level patterns which aim to identify and formalize common programming techniques. A type (either a class or an interface) is an instance of a micro pattern if and only if all of its methods and/or attributes satisfy the constraints specified by the micro pattern. We suggest a novel approach to the detection of micro patterns which is aimed to identify types that are very close and similar to a correct micro pattern implementation, even if some of the methods and/or attributes of the type do not comply with the constraints defined by the micro pattern. The new interpretation is based on the number of attributes (NOA) and the number of methods (NOM) of a type. The identification of types similar to micro patterns allows the analysis of software systems along various releases, as well as the identification of possible critical classes that can't be detected with a precise matching approach.