Skip to main content
mehen metrics is the single-file analysis command. It takes one path, auto-detects the language, and emits a complete metrics report.

​
Examples

​
Exit codes

​
Configured thresholds

When a mehen.toml is present, every configured threshold that matches a metric this file publishes is evaluated against the report — the language’s override table wins over the global one. Crossed limits print a grouped violation report on stderr after the report lands on stdout, and the command exits 1. Files with blocking parse diagnostics exit 1 without threshold evaluation: partial metrics are never gated.

​
What gets reported

Source-code files yield the code metric family (cyclomatic, cognitive, Halstead, MI, ABC, LOC family, NOM, NPA, NPM, NARGS, NEXITS, WMC). Markdown files yield the Markdown metric suite — DMI, MRPC, MCC, Markdown Halstead, link debt, table burden, visual scaffold, artifact debt, repository grounding, evidence coverage, filler/lazy risk, review criticality, section balance, good scaffold. The opt-in prose layer ships behind Cargo features. When an analyzer supports source-resolved evidence, the default profile also emits a contributions array. Each entry identifies the metric, weighted amount, stable reason code, and byte/line span responsible for that amount. SQL change risk is the first supported metric: sql.change_risk_score explains destructive DDL/DML and distinct object-touch terms. The ci/strict profiles skip this optional evidence to keep automated payloads lean.

​
Profiles

--profile selects a built-in preset for what gets reported and how thresholds are interpreted: Explicit CLI flags always override profile defaults.

​
Inspecting contribution evidence

Maintainers can print only the contribution array for a file through the developer task:
The task runs the production analyzer through the real mehen metrics path, so language detection, profile behavior, reason codes, and spans match the public CLI report.

​
See also