Reading a mutation report

npm run test:mutation writes three files below reports/, and .gitignore excludes all of them. Read the report, not the cache:

• reports/mutation/mutation.json — the report: the result of the last run, and the file to query.
• reports/mutation/mutation.html — the same data, for a human to open in a browser.
• reports/stryker-incremental.json — the incremental cache. Same shape, but it carries results merged from earlier runs, so a mutant in it may be stale.

The survivors

Every run, start here:

jq -r '.files | to_entries[] | .key as $file | .value.mutants[]
       | select(.status == "Survived" or .status == "NoCoverage")
       | "\($file):\(.location.start.line):\(.location.start.column)  \(.mutatorName)  ->  \(.replacement)"' \
  reports/mutation/mutation.json

Each line is one change to the source that the suite stayed green through. replacement is the mutated code; location.start is where it goes. To see the original, read those lines of the file — or slice the copy the report keeps in .files[<path>].source.

The statuses

status decides whether a mutant is your problem:

Status	Meaning
Survived	A test ran the mutated code and passed. Act on it.
NoCoverage	No test ran the mutated code at all. Act on it.
Killed, Timeout	A test caught the change. statusReason holds the failure message.
CompileError, RuntimeError, Ignored	Stryker could not test the mutant. Not a signal about your tests.

Other fields on a mutant: id, killedBy (test ids), coveredBy, testsCompleted. Test ids resolve against .testFiles.

Acting on a survivor

A survivor is a review comment on the test, not on the implementation: the behaviour changed and the suite did not notice. For each one, either tighten the test until it fails against replacement, or state why that behaviour is unspecified and the mutant is accepted.

Done when every Survived and NoCoverage mutant in the report is accounted for.