seer-lab/MutationScorePredictor

Ignore overloaded methods

kevinjalbert opened this issue · 2 comments

We can overloaded methods within the data. This poses a problem mainly because the different tools will report the methods differently.

Emma will list the parameter and return types, so there is no problem there. The Eclipse Metric Plugin will not indicate the parameter/return types for each method. It simply lists them many times. Javalanche also does not distinguish overloaded methods. With Javalanche it is possible to acquire the line number of the class.method and by using this data you can find what overloaded method the mutant belongs too.

For example, the line numbers for a set of mutants for the load method are:

  • 565
  • 560
  • 598
  • 519
  • 598
  • 558

Everything but the mutant on line 519 belongs to the same method within the source code. To resolve this it would require a lot of parsing work to get unique methods from Javalanche.

The problem still resides in Eclipse Metrics Plugin... So the best choice is to simply ignore overloaded methods all together as we cannot guarantee the uniqueness.

Seems like we are currently taking the last occurrence of the overloaded method from the Eclipse Metric Plugin. Thus we will detect if there is a duplicate in the Eclipse Metric Plugin part and actually omit it there.

This issue has been solved with the following commit kevinjalbert/eclipse_metrics_xml_reader@a396005