A Metrics Suite for Variable CategorizationtTo Support Program Invariants
Radhika Raju P, Ananda Rao A · International Journal of Software Engineering & Applications · 2014
Invariants are generally implicit.Explicitly stating program invariants, help programmers to identify program properties that must be preserved while modifying the code.Existing dynamic techniques detect invariants which includes both relevant and irrelevant/unused variables and thereby relevant and irrelevant invariants involved in the program.Due to the presence of irrelevant variables and irrelevant invariants, speed and efficiency of techniques are affected.Also, displaying properties about irrelevant variables and irrelevant invariants distracts the user from concentrating on properties of relevant variables.To overcome these deficiencies only relevant variables are considered by ignoring irrelevant variables.Further, relevant variables are categorized as design variables and non-design variables.For this purpose a metrics suite is proposed.These metrics are validated against Weyuker's principles and applied on RFV and JLex open source software.Similarly, relevant invariants are categorized as design invariants, non-design invariants and hybrid invariants.For this purpose a set of rules are proposed.This entire process enormously improves the speed and efficiency of dynamic invariant detection techniques.