Generating Diverse Test Suites for Gson Through Adaptive Fitness Function Selection

• Many fitness functions - such as those targeting test suite diversity—do not yield sufficient feedback to drive test generation. We propose that diversity can instead be improved through adaptive fitness function selection (AFFS), an approach that varies the fitness functions used throughout the generation process in order to strategically increase diversity. We have evaluated our AFFS framework, EvoSuiteFIT, on a set of 18 real faults from Gson, a JSON (de)serialization library. Ultimately, we find that AFFS creates test suites that are more diverse than those created using static fitness functions. We also observe that increased diversity may lead to small improvements in the likelihood of fault detection.

Subject headings

NATURVETENSKAP  -- Data- och informationsvetenskap -- Annan data- och informationsvetenskap (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Other Computer and Information Science (hsv//eng)

NATURVETENSKAP  -- Data- och informationsvetenskap -- Programvaruteknik (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Software Engineering (hsv//eng)

Keyword

Search-based test generation

fitness function

reinforcement learning

Fitness function

Reinforcement learning

Search-based test generation

Publication and Content Type

kon (subject category)

ref (subject category)