Click here for full text:
Measuring the Entropy of Large Software Systems
Snider, Greg
HPL-2001-221
Keyword(s): entropy; software systems; structure; metric
Abstract: How does one measure a large software system to determine if it is "well-structured"? This report proposes a metric for doing just that, based on the concept of entropy from information theory. A tool that automatically extracts the metric from source code was built and used to compare two large software systems (each about 500,00 lines of source code): a commercial application that has existed and been heavily modified for several decades; and a recent rewrite of the same system that aimed at producing a well-structured system. The rewritten system was shown to have much lower entropy in each of its subsystems, compared to the legacy system, as well as much lower entropy overall.
12 Pages
Back to Index
| 
printable version
