This paper presents a largely algorithmic approach to identifying objects in COBOL and RPG procedural legacies, ie reverse engineering. They argue that objects
can be best found in persistent data sources such as files and datbases, rather than fom studying the code. Class methods can subsequently be found from the code
which acts on this data. However, this approach is tempered by OO design metrics. In this way, a higher level of object granularity is achieved than by purely
considering the code with all of its volatile data. This makes comprehension easier and classes more meaningfull.
Appears to be a fair bit of research in this area, witnessed by the relatively large background section. Not much on the implementation issues.
Related pages: Bibliography
|