Static and dynamic distance metrics for feature-based code analysis

  • Authors:

  • W. Eric Wong;Swapna Gokhale

  • Affiliations:

  • Department of Computer Science, University of Texas at Dallas, 2601 N Floyd Road, Richardson, TX;Department of Computer Science and Engineering, University of Connecticut, Storrs, CT

  • Venue:
  • Journal of Systems and Software
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper we present metrics to determine the distance between the features of a software system. Such a measurement can elucidate how features of the system being examined are close to each other. We first use an execution slice-based technique to identify a set of code (basic blocks in our case) that is used to implement each feature. Then, depending on whether the execution frequency of each block is considered during the construction of such sets of code, a static as well as a dynamic distance is computed for each pair of features. These two types of distance differ in that the former computes the distance between two features only by how these features are implemented in the system, while the latter also takes into account how each feature is executed based on a user's operational profile. In other words, the static distance quantitatively gives the closeness of two features from the system implementation point of view, whereas the dynamic distance presents such closeness from the users' execution point of view. To illustrate the use of our metrics, we report a case study on a Symbolic Hierarchical Automated Reliability and Performance Evaluator (SHARPE). The results of our study suggest that the distance metrics discussed in this paper can provide a good measurement, in a quantitative way, of how close two program features are. Such information can also serve as a good start to understanding how a modification made to one feature is likely to affect other features.