Markdown metrics overview - mehen

mehen treats Markdown as a first-class artifact. Instead of stripping code fences, diagrams, tables, images, links, and math before counting words, every construct contributes to a dedicated metric suite. The suite is split into two layers:

Structural layer — language-opaque, AST-driven: LOC family, section tree, reading-path complexity, link/table/visual scaffolds, repository grounding, filler/lazy risk, review criticality.
Prose layer — language-aware (English and Japanese): readability ensemble, lexical diversity, wording quality, JTF rules, textlint subset.

Code-style metrics like cyclomatic complexity, NOM, NPA, NPM, and WMC do not apply to Markdown — prose has no functions, classes, or interfaces to score. mehen emits a Markdown-specific suite tailored to prose, code blocks, and references.

Structural layer

Page	Purpose
LOC family	Markdown LOC variants by construct: prose / code / tables / math / blank / artifact.
Section tree	Heading-derived structure with quality flags.
Effective Content Units	Review-mass normalization across prose and artifacts.
MRPC	Reading Path Complexity — cyclomatic-complexity analogue for navigation.
MCC	Cognitive Complexity — local reading burden.
Markdown Halstead	Markdown-native operators / operands.
DMI	Documentation Maintainability Index — composite 0–100.
Link Debt	Link health and reference burden.
Table Burden + Scaffold	Table cost and cognitive scaffolding.
Visual Scaffold	Diagrams and images: helping or hurting?
Artifact Debt	Unlabelled, unparsable, oversized, unexplained artifacts.
Repository Grounding	Resolved repo links, paths, identifiers, version facts.
Evidence Coverage	Per-section structural support density.
Filler / Lazy Structure Risk	AI-era flag for “big but vacuous”.
Review Criticality Index	”Should I review this carefully?” score.
Section Balance	Are sections sized in a maintainable way?
Good Scaffold	Bonus credit for helpful technical structure.

Prose layer (opt-in, feature-gated)

Page	Purpose
Overview	Architectural constraints and tier model.
Block-level language detection	Per-block language tagging.
English readability	FKGL, Fog, SMOG, ARI, Coleman-Liau, Dale-Chall, FORCAST, LIX/RIX.
Lexical diversity	MATTR, hapax, density.
Wording quality	Passive, hedges, weasels, wordy, adverbs, nominalizations, cliches.
Inclusive language	alex / retext-equality flags.
Japanese script composition	Kanji/hiragana/katakana ratios, registers.
Tateishi RS + Jōyō grade	Japanese readability scores.
JTF rules	Japan Translation Federation 12 rules.
textlint-ja subset	Selected rules from `textlint-rule-preset-ja-technical-writing`.

See also

PR comment design — how Markdown metric deltas surface on pull requests.
Concepts → Output formats — Markdown vs. JSON shape.