Dynamic analysis of Ada programs for comprehension and quality measurement

Abstract

During maintenance and particularly during corrective and perfective tasks, systems tend to exhibit a weight gain. As a result, their quality tends to degrade. Software comprehension is vital in order to assess system quality. In this paper, we aim at deploying dynamic analysis of Ada programs for obtaining comprehension, and applying measurements to assess their quality. Program instrumentation is performed non-intrusively by AspectAda, an aspect-oriented extension to Ada which we discussed in earlier work. Events which are required for this analysis are captured as execution traces. We have defined a relational database schema to save execution traces, and a set of queries to obtain measures of quality metrics. New Ada-specific metrics are introduced and existing metrics have been adopted from the literature. Automation is also provided as a proof of concept through a prototypical tool which provides information on the runtime behavior of the system, performs measurements and provides visualization of the run-time behavior of the system through a call graph. An open source Ada program is used as a case study to demonstrate our approach.

A systematic mapping study on dynamic metrics and software quality

“…Empirical studies show that static measurement is -unsurprisingly -not sufficient for capturing dynamic dependencies among system modules such as those related to polymorphism, dynamic binding and inheritance [7]. It is also noted that static analysis may result in a huge amount of data [6] that can be difficult to understand and summarize [14].…”

“…This instrumentation can help a developer to collect precisely all the information needed to address a particular problem in the software [5]. Dynamic analysis has been used largely in the fields of software measurement [7,14,[17][18], software maintenance and reengineering [19][20], in clustering [21], and in program understanding and comprehension [12,22] .…”

“…The incomplete coverage 'problem' leads to another cited disadvantage: limited generalization of the results obtained. Dynamic metrics' results may not generalize for future executions, given that the gathered data pertain solely to the scenario that was executed at a given point in time [13][14]. There is no way of assuring that the scenario(s) in which the program was run is (are) representative of all possible program scenarios and executions [6].…”

“…This was an informal review as there were no defined research questions, search process or data extraction process. As a result the review did not cover some important works in the area such as [9,14,17,23,[25][26][27], and the latest article considered in the review had been published in 2006. We have found a number of studies (31 articles) on the topic that were published between the latest article cited in Chhabra and Gupta's study and our study.…”