Entropyx computes per-file badness using weighted, min-max normalized metrics, including log-scaled source lines of code, complexity density, smells per KLOC, coupling proxies, and maintainability indices. Files with negligible badness (≤ 10⁻⁹) are excluded to reduce floating-point noise. Shannon entropy (base-2) is computed over the distribution of per-file badness values and normalized by log₂(N), where N is the number of non-negligible files. The final Entropyx Score is defined as normalized entropy multiplied by mean badness, capturing both diffusion and magnitude of structural degradation. The framework is language-agnostic and designed for cross-platform CLI implementation.
Espen Sande-Larsen (Thu,) studied this question.