Coupling and Size Measures. Both OO and AO implementations present very
similar stability with respect to Coupling Between Components (CBC) and size
metrics. Figure 8 illustrates the curves results for coupling and size metrics of Health
Watcher. The CBC graphic shows that the AO implementations fared better, but in a
not very significant way. It happens mainly because, although many aspects reduce the
coupling of system classes by modularizing the design by contract crosscutting concern,
they still reference the classes in which they introduce some behavior. Despite the
small di®erence between OO andAO regarding coupling, theAO implementation tend
to be better while applying the changing scenarios. This happens because during the
maintenance activities performed by the changing scenarios, it was observed that no
new aspects were created and the existing ones provide few heterogenous DbC features.
As a result, even though the VS graphic shows that the OO implementations
performed better than AO, the di®erence between them remained constant (due to
zero new aspects). Hence, the increase in the VS metric is justified by the introduction
of aspects to modularize the DbC concern. The same reason justifies the worse results
of AO implementations in relation to the Number of Operations (NOO) metric. The
main di®erence is that this metric varies in each release due to some heterogenous
contracts and changes scenarios (e.g. adding/removing functionalities).
Coupling and Size Measures. Both OO and AO implementations present verysimilar stability with respect to Coupling Between Components (CBC) and sizemetrics. Figure 8 illustrates the curves results for coupling and size metrics of HealthWatcher. The CBC graphic shows that the AO implementations fared better, but in anot very significant way. It happens mainly because, although many aspects reduce thecoupling of system classes by modularizing the design by contract crosscutting concern,they still reference the classes in which they introduce some behavior. Despite thesmall di®erence between OO andAO regarding coupling, theAO implementation tendto be better while applying the changing scenarios. This happens because during themaintenance activities performed by the changing scenarios, it was observed that nonew aspects were created and the existing ones provide few heterogenous DbC features.As a result, even though the VS graphic shows that the OO implementationsperformed better than AO, the di®erence between them remained constant (due tozero new aspects). Hence, the increase in the VS metric is justified by the introductionof aspects to modularize the DbC concern. The same reason justifies the worse resultsof AO implementations in relation to the Number of Operations (NOO) metric. Themain di®erence is that this metric varies in each release due to some heterogenouscontracts and changes scenarios (e.g. adding/removing functionalities).
การแปล กรุณารอสักครู่..
