Premium
Are large C++ classes change‐prone? An empirical investigation
Author(s) -
Lindvall Mikael
Publication year - 1998
Publication title -
software: practice and experience
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.437
H-Index - 70
eISSN - 1097-024X
pISSN - 0038-0644
DOI - 10.1002/(sici)1097-024x(19981225)28:15<1551::aid-spe212>3.0.co;2-0
Subject(s) - set (abstract data type) , computer science , change impact analysis , interpretation (philosophy) , software , empirical research , object (grammar) , subject (documents) , software engineering , programming language , world wide web , artificial intelligence , statistics , mathematics
It is crucial, but hard, to predict the parts of a system that must be changed in the next release of a software system. Previous results showed that professional and experienced developers largely underpredict the number of software entities subject to change in the next release, even though a thorough impact analysis driven by requirements is conducted. Thus, there is a strong need for understanding what kind of properties of a system make it more or less vulnerable to change. This empirical study analyzes change of C++ source code classes that occurred in two releases of the industrial object‐oriented project PMR (Performance Management Traffic Recording). PMR is a part of the operation and support system of a cellular telecom system. The result, which is consistent over the two releases despite different kinds of new requirements, shows that the median size in the set of changed classes is significantly larger than in the set of unchanged classes. Our interpretation of the results is that large classes are more change‐prone than small classes, and that large classes therefore always should be regarded as candidates for change. © 1998 John Wiley & Sons, Ltd.