| Summary: | Many tools have been built to analyze the source code of software systems. Most
of these tools do not adequately support software reengineering activities because
they do not allow a software engineer to analyze both existing and desired software
structures.
This thesis describes the conceptual modules approach and supporting tool
that aids the engineer in the investigation and analysis of desired structure with
relation to the existing structure of source code. This approach allows a selected
subset of lines of source to be treated as a logical unit. This subset is referred to as
a conceptual module. The lines of code that comprise a conceptual module need not
be contiguous, nor must they be related in any way in the source. Using variable
dependence and control transfer information extracted from the source, the tool
analyzes the conceptual module's constituent lines of code to determine its interface.
Additionally, the data- or control-flow between two or more conceptual modules can
be examined as a means of eliciting the relationships between the modules and
between conceptual modules and the source. To allow the necessary flexibility in
analysis, the functionality of the tool can be tailored through a programmatic query
language component.
The usefulness of the tool has been investigated in two different ways. First,
the tool was applied to several different reengineering scenarios: restructuring from
procedural to object-oriented program design, re-modularizing code in an existing
program with little structure, and extracting a portion of source for reuse. For each
scenario, several existing program understanding tools were also applied to provide
a basis of comparison between existing approaches and the conceptual modules
approach. Second, the tool was successfully applied to actual reengineering tasks by
two different groups of users. One group eliminated unnecessary parts of a system's
source to improve efficiency and to enable parallelization of a 47,000 line, 56-file
software package. The other group performed analysis on a procedural program so
as to better understand how to transform the existing source into an object-oriented
version.
|
|---|
