seer-lab/MutationScorePredictor

Refactor the way Number Of Tests (NOT) metric is calculated

kevinjalbert opened this issue · 1 comments

kevinjalbert commented

Currently we acquire the Number Of Tests (NOT) metric by using a by-product of Javalanche. When the mutation testing occurs with Javalanche it performs coverage analysis (we have it enabled so we can keep track of mutants), which also keeps track of the *NOT for each mutant.

This is not ideal as we want to decouple the features for prediction from the end result of mutation testing. This was done original to save time, but the correct approach would utilize the EMMA coverage tool.

Proposed Solution

  1. For each test in the test suite acquire the coverage it exhibits on the source code using EMMA
  2. For each source code unit (i.e., method or class) count the number of tests which cover at least one instruction/statement in the source code unit
kevinjalbert commented

I'm closing this issue as a great deal of time has passed. If there is renewed interest just reopen or make a new issue and link to this one.