:Author: Manuel Pichler :Copyright: All rights reserved :Description: This article describes the CSZ - Class Size metric as it is defined in the literature and how it is implemented in PHP_Depend. Later in this article we describe the impact of the class size on several quality-factors in a software system. :Keywords: CSZ, CS, Class Size, Khaled et al., Optimal Class Size, SRP, Single Responsibility Principle, testability, responsibility, maintainability
The Class Size or CSZ metric is a another measure for the complexity and
size of a class. In many publications and books this metric is frequently
referenced with the abbreviation CS. The definition and detailed description
of this metric can be found in The Optimal Class Size for Object-Oriented
Software paper by Khaled, Benlarbi, Nishith & Shesh [#ooocsz]_.
There are three different definitions of the Class Size metric, two of them are only a different name for already existing metrics. The first one takes the number of statements, respectively the Logical Lines Of Code (LLOC), to measure the size of class. The second variant uses the Lines Of Code (LOC) as an indicator for the class size.
The third variant uses two language constructs of object-oriented programming
languages to measure the size of class and calculates the sum of both values.
These to language constructs are methods and attributes, so that the existing
metrics Number Of Methods (NOM) and Number Of Attributes (VARS) can
be reused to calculate the Class Size.
CSZ = NOM + VARS
PHP_Depend implements this third variant of Class Size metric algorithm to measure the size of a class.
As several researches and studies on procedural software projects have shown, the optimal size of a software component, lies somewhere between too small and too large, to be less fault-prone. This model can be visualized with an u-curve, where the lower arc of the U represents the optimal size of a component. But this model does not apply to object-oriented software systems, as Khaled et al. have shown in their paper [#ooocsz]_. Object-oriented systems tend to be more fault-prone when they get bigger and less fault-prone when they use small classes and follow the Single Responsibility Principle (SRP) [#poodsrp]_.
For the CSZ metric this means that smaller values are always better, than greater ones, because a smaller size is an indicator for better testability, good maintainability and a well defined responsibility of the class.
Based on their research of several software systems and their reflection of other studies about the optimal size of a software component Khaled et al. suggest an upper limit 39 for a Class Size metric based on the number of a class' methods and attributes. But as always, this threshold can only be a clue for own limits, that fit to custom requirements.
IEEE Transactions on Software Engineering; The Optimal Class Size for Object-Oriented Software; Khaled, Benlarbi, Nishith & Shesh; 2002
ButUncleBob.com; The Single Responsibility Principle; Robert C. Martin