component-identification-sizing

Map the directory or namespace structure and treat leaf nodes containing source files as components. If a namespace is extended by child directories, the parent becomes a subdomain rather than a component.

Count executable statements rather than lines of code, then compute each component's percentage of total statements and its distance from the mean in standard deviations. Components beyond 2 standard deviations or above the size threshold are oversized.

Thresholds depend on application size: over 30% of the codebase for small apps with fewer than 10 components, 15% for medium apps, and 10% for large apps with more than 20 components. The standard deviation rule also flags outliers.

Yes, the analysis is language agnostic. It uses package structures like com.company.domain.service for Java, module paths like app/billing/payment for Python, and directory patterns for Node.js, with language-specific statement counting rules.

Do not use it for runtime performance sizing or infrastructure capacity planning, since it measures static code volume only. It also should not mix infrastructure and domain components in the same analysis or count test files.