crap-score

Run dotnet test with coverage collection to produce a Cobertura XML report, then compute cyclomatic complexity per method by counting decision points like if, case, and loops. Apply the formula comp^2 x (1 - cov)^3 + comp using each method's line-rate from the coverage data.

A CRAP score below 5 indicates simple, well-tested code, while 5-15 is acceptable for most methods. Scores of 15-30 are high risk needing more tests or simplification, and anything above 30 is critical and should be refactored and covered urgently.

Yes, it works with Cobertura XML produced by coverlet.collector via dotnet test --collect:"XPlat Code Coverage", and with Microsoft.Testing.Extensions.CodeCoverage on .NET 9 and later using the --coverage-output-format cobertura option.

The compiler generates mangled names for async methods, lambdas, and local functions, so Cobertura entries may not align with source names. Match methods by line ranges instead of names when this happens.

Do not use it when you only want to run tests, write new tests, or get a plain coverage percentage without complexity context. CRAP analysis is specifically for prioritizing risky code that is both complex and undertested.

No. When cyclomatic complexity is 15 or higher, the minimum possible CRAP score equals the complexity itself, so coverage alone cannot bring it below the threshold. The method must first be refactored, for example by extracting sub-methods.