@primmel/primmel 1.3.2 → 1.4.0

package/dist/src/text-coverage.js

@@ -0,0 +1,679 @@
+"use strict";
+// ─────────────────────────────────────────────────────────────────────
+// TODO.roadmap/26 — the normative-text coverage metric (concept doc
+// §11.6, layer 5 of the validation stack):
+//
+//   every normative sentence of the source maps to at least one model
+//   element (target 100 %), and no two elements are semantic duplicates
+//   (target 0).
+//
+// The inputs ship INSIDE the package (JSON — primmel-ts stays YAML-free),
+// generated from the YAML single source of truth by the OIML SMART build
+// (browser/scripts/build-prl-package.ts):
+//
+//   sources-prd/<part>.sentences.json — the sentence decomposition of one
+//     source document: every prose sentence with its address
+//     (<fragment-path>/s<N>), fragment, kind, fragment-level normativity,
+//     the modality classifier's verdict (only NORMATIVE sentences gate —
+//     shall/should/may/must + negatives, context rules first), and text.
+//     Table/figure fragments carry no sentences (tables are data, never
+//     text — their coverage is the clause-level congruence gate's).
+//   sources-prd/coverage.json — the per-package declarations:
+//     allowances (deliberate exclusions, sentence-PINNED so a new
+//     normative sentence in an allowed fragment is still a finding — the
+//     pair-pinned discipline of the congruence gate's order exceptions)
+//     and duplicate_adjudications (the human verdicts on flagged
+//     near-duplicate pairs; the metric REPORTS pairs, never auto-fails —
+//     acceptance is 0 UNRESOLVED pairs).
+//
+// The rules (check-rules.ts):
+//   C71 text-coverage-sentence-uncovered (AUDIT, warning) — a normative
+//       sentence bound by no model element and excluded by no allowance;
+//       budgeted by the package's text_coverage_budget (the C51/C52
+//       pattern — regressions fail the gate);
+//   C72 text-coverage-budget (AUDIT) — the budget caps C71 warnings
+//       (exceeded: error; slack: warning — the allowance only shrinks);
+//   C73 text-coverage-config — malformed/stale declarations: an
+//       allowance matching no normative sentence (drifted address, a pin
+//       on a sentence that stopped existing or being normative), a
+//       mis-sorted or silent adjudication, a STALE adjudication whose
+//       pair is no longer flagged. Runs whenever a manifest ships.
+//
+// Duplicate detection (the methodology's "0 semantic-equivalent elements"
+// rule, machine-assisted): over the elements carrying bound text
+// (requirement statements, term definitions) that bind ≥1 fragment of a
+// manifest document, flag pairs whose normalized token-multiset
+// containment similarity reaches DUPLICATE_SIMILARITY_THRESHOLD — two
+// requirements interpreting the same sentence differently is worse than
+// a gap, because both will compute verdicts (§11.6).
+// ─────────────────────────────────────────────────────────────────────
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.DUPLICATE_TEXT_KINDS = exports.DUPLICATE_MIN_TOKENS = exports.DUPLICATE_SIMILARITY_THRESHOLD = void 0;
+exports.loadTextCoverageData = loadTextCoverageData;
+exports.normalizeBinding = normalizeBinding;
+exports.collectProvenanceBindings = collectProvenanceBindings;
+exports.allowanceAddressMatches = allowanceAddressMatches;
+exports.allowanceMatches = allowanceMatches;
+exports.normalizeForSimilarity = normalizeForSimilarity;
+exports.containmentSimilarity = containmentSimilarity;
+exports.flagDuplicatePairs = flagDuplicatePairs;
+exports.computeTextCoverage = computeTextCoverage;
+exports.uncoveredSentenceMessage = uncoveredSentenceMessage;
+exports.packageTextCoverageReport = packageTextCoverageReport;
+exports.formatTextCoverageReport = formatTextCoverageReport;
+const fs_1 = require("fs");
+const path_1 = require("path");
+const SENTENCES_FORMAT = '0.1.0';
+const COVERAGE_FORMAT = '0.1.0';
+/** Load the text-coverage payloads of a package (<dir>/sources-prd/).
+ * Absent directory → empty (the metric is silent). Malformed payloads are
+ * C73 errors. */
+function loadTextCoverageData(dir) {
+    const issues = [];
+    const manifests = [];
+    let config = null;
+    const prdDir = (0, path_1.join)(dir, 'sources-prd');
+    if (!(0, fs_1.existsSync)(prdDir)) {
+        return { manifests, config, issues };
+    }
+    for (const f of (0, fs_1.readdirSync)(prdDir).sort()) {
+        if (!f.endsWith('.sentences.json')) {
+            continue;
+        }
+        const rel = `sources-prd/${f}`;
+        let raw;
+        try {
+            raw = JSON.parse((0, fs_1.readFileSync)((0, path_1.join)(prdDir, f), 'utf8'));
+        }
+        catch (e) {
+            issues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${rel}: not parseable JSON — ${e.message} (text-coverage-config)`,
+            });
+            continue;
+        }
+        if (raw.prd_sentences !== SENTENCES_FORMAT ||
+            typeof raw.document?.urn !== 'string' ||
+            !Array.isArray(raw.sentences) ||
+            raw.sentences.some(s => typeof s?.address !== 'string' ||
+                typeof s?.fragment !== 'string' ||
+                (s?.modality !== 'normative' && s?.modality !== 'informative') ||
+                typeof s?.text !== 'string')) {
+            issues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${rel}: not a prd_sentences ${SENTENCES_FORMAT} manifest (document.urn + sentences[{address, fragment, modality, text}]) — regenerate the package (text-coverage-config)`,
+            });
+            continue;
+        }
+        manifests.push(raw);
+    }
+    const configPath = (0, path_1.join)(prdDir, 'coverage.json');
+    if ((0, fs_1.existsSync)(configPath)) {
+        try {
+            const raw = JSON.parse((0, fs_1.readFileSync)(configPath, 'utf8'));
+            if (raw.prd_coverage !== COVERAGE_FORMAT) {
+                throw new Error(`prd_coverage must be "${COVERAGE_FORMAT}"`);
+            }
+            config = raw;
+        }
+        catch (e) {
+            issues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `sources-prd/coverage.json: ${e.message} — regenerate the package (text-coverage-config)`,
+            });
+        }
+    }
+    return { manifests, config, issues };
+}
+const URN_ADDRESS = /^(urn:oiml:pub:[a-z0-9:-]+)#([A-Za-z0-9./-]+)$/;
+const FRAGMENT_WORD = /^(clause|table|figure|front)-/;
+/** Normalize one binding to {doc, fragmentPath}: a doc may carry the
+ * glued `#fragment` form (doc "urn:…#table-4"), a bare clause number gets
+ * the clause- prefix, an existing address word passes through, and the
+ * optional sentence sub-address appends `/s<N>` (TODO.roadmap/26). */
+function normalizeBinding(doc, clause, fragment) {
+    let d = doc.trim();
+    let path = '';
+    const hash = d.indexOf('#');
+    if (hash >= 0) {
+        path = d.slice(hash + 1);
+        d = d.slice(0, hash);
+    }
+    if (path === '') {
+        const c = clause.trim();
+        if (c === '') {
+            return null;
+        }
+        path = FRAGMENT_WORD.test(c) ? c : `clause-${c}`;
+    }
+    if (fragment && fragment.trim() !== '') {
+        path += `/${fragment.trim()}`;
+    }
+    return { doc: d, fragmentPath: path };
+}
+function* urnStringBindings(value) {
+    for (const token of value.trim().split(/\s+/)) {
+        const m = URN_ADDRESS.exec(token);
+        if (m) {
+            yield { doc: m[1], fragmentPath: m[2] };
+        }
+    }
+}
+function isSourceRefLike(v) {
+    return (typeof v === 'object' &&
+        v !== null &&
+        (typeof v.doc === 'string' ||
+            typeof v.clause === 'string'));
+}
+/** Harvest every provenance binding of the composed package: sourceRefs/
+ * sourceRef blocks, `source { doc clause fragment? }` blocks, term-style
+ * `source "urn …"` strings, `reference "urn#fragment"` strings and
+ * references[].urn lists — on ANY id-bearing element (the generic walk is
+ * construct-agnostic; non-URN strings never match the address grammar). */
+function collectProvenanceBindings(standard) {
+    const out = [];
+    const seen = new Set();
+    const push = (element, kind, text, norm) => {
+        if (!norm || norm.fragmentPath === '') {
+            return;
+        }
+        const key = `${element}${norm.doc}#${norm.fragmentPath}`;
+        if (seen.has(key)) {
+            return;
+        }
+        seen.add(key);
+        out.push({
+            element,
+            kind,
+            doc: norm.doc,
+            fragmentPath: norm.fragmentPath,
+            ...(text !== undefined ? { text } : {}),
+        });
+    };
+    const harvestFields = (node, element, kind, text) => {
+        const fromRef = (s) => normalizeBinding(typeof s.doc === 'string' ? s.doc : '', typeof s.clause === 'string' ? s.clause : '', typeof s.fragment === 'string' ? s.fragment : undefined);
+        if (Array.isArray(node.sourceRefs)) {
+            for (const s of node.sourceRefs) {
+                if (isSourceRefLike(s)) {
+                    push(element, kind, text, fromRef(s));
+                }
+            }
+        }
+        // condition sets collect repeated source blocks into `sources` (plural)
+        if (Array.isArray(node.sources)) {
+            for (const s of node.sources) {
+                if (isSourceRefLike(s)) {
+                    push(element, kind, text, fromRef(s));
+                }
+            }
+        }
+        if (isSourceRefLike(node.sourceRef)) {
+            push(element, kind, text, fromRef(node.sourceRef));
+        }
+        if (isSourceRefLike(node.source)) {
+            push(element, kind, text, fromRef(node.source));
+        }
+        else if (typeof node.source === 'string') {
+            for (const u of urnStringBindings(node.source)) {
+                push(element, kind, text, u);
+            }
+        }
+        if (typeof node.reference === 'string') {
+            for (const u of urnStringBindings(node.reference)) {
+                push(element, kind, text, u);
+            }
+        }
+        if (Array.isArray(node.references)) {
+            for (const r of node.references) {
+                const urn = r?.urn;
+                if (typeof urn === 'string') {
+                    for (const u of urnStringBindings(urn)) {
+                        push(element, kind, text, u);
+                    }
+                }
+            }
+        }
+        // Form reference blocks (RoleReference { urn, role } — form-level
+        // formReferences, field-level fieldReferences)
+        for (const key of ['formReferences', 'fieldReferences']) {
+            if (Array.isArray(node[key])) {
+                for (const r of node[key]) {
+                    const urn = r?.urn;
+                    if (typeof urn === 'string') {
+                        for (const u of urnStringBindings(urn)) {
+                            push(element, kind, text, u);
+                        }
+                    }
+                }
+            }
+        }
+    };
+    const walk = (node, kind, element, text, depth) => {
+        if (depth > 12 || typeof node !== 'object' || node === null) {
+            return;
+        }
+        if (Array.isArray(node)) {
+            for (const item of node) {
+                walk(item, kind, element, text, depth + 1);
+            }
+            return;
+        }
+        const rec = node;
+        let el = element;
+        let tx = text;
+        if (typeof rec.id === 'string' && rec.id !== '') {
+            el = rec.id;
+            tx =
+                typeof rec.statement === 'string'
+                    ? rec.statement
+                    : typeof rec.definition === 'string'
+                        ? rec.definition
+                        : undefined;
+        }
+        if (el !== null) {
+            harvestFields(rec, el, kind, tx);
+        }
+        for (const v of Object.values(rec)) {
+            if (typeof v === 'object' && v !== null) {
+                walk(v, kind, el, tx, depth + 1);
+            }
+        }
+    };
+    for (const [field, value] of Object.entries(standard)) {
+        if (Array.isArray(value)) {
+            walk(value, field, null, undefined, 0);
+        }
+    }
+    return out;
+}
+// ── allowance matching (lockstep with the smart repo's prd-coverage.ts) ──
+/** True when sentence address A is covered by allowance address P: exact,
+ * or P a boundary prefix of A ('.' descends the clause tree, '/' a
+ * sub-address). */
+function allowanceAddressMatches(allowance, sentence) {
+    if (sentence === allowance) {
+        return true;
+    }
+    if (!sentence.startsWith(allowance)) {
+        return false;
+    }
+    const next = sentence.charAt(allowance.length);
+    return next === '.' || next === '/';
+}
+/** The sentences of one document an allowance discharges (pins first —
+ * exactly the pinned sentences of the addressed fragment; otherwise every
+ * sentence the address boundary-matches). */
+function allowanceMatches(allowance, sentences, urn) {
+    const hash = allowance.address.indexOf('#');
+    const fragment = allowance.address.slice(hash + 1);
+    if (allowance.sentences) {
+        const pins = new Set(allowance.sentences.map(p => `${urn}#${fragment}/${p}`));
+        return sentences.filter(s => pins.has(`${urn}#${s.address}`));
+    }
+    return sentences.filter(s => allowanceAddressMatches(allowance.address, `${urn}#${s.address}`));
+}
+// ── duplicate detection ──────────────────────────────────────────────
+/** The pinned similarity threshold (token-multiset containment:
+ * Σ min(cA, cB) / min(|A|, |B|) over normalized tokens). 0.8 flags
+ * near-duplicates while per-class specializations differing only in
+ * their class values still surface for adjudication; pairs sharing
+ * fewer than DUPLICATE_MIN_TOKENS tokens on either side never flag
+ * (short texts produce noise, not signal). */
+exports.DUPLICATE_SIMILARITY_THRESHOLD = 0.8;
+exports.DUPLICATE_MIN_TOKENS = 6;
+/** Normalize bound text for similarity: lowercase, punctuation folded to
+ * space, whitespace collapsed (the wording axis, not the symbols'). */
+function normalizeForSimilarity(text) {
+    return text
+        .toLowerCase()
+        .replace(/[‘’‚]/g, "'")
+        .replace(/[“”„]/g, '"')
+        .replace(/[‐‑‒–—]/g, '-')
+        .replace(/[^a-z0-9]+/g, ' ')
+        .trim()
+        .split(' ')
+        .filter(t => t !== '');
+}
+/** Multiset containment similarity of two token lists (1 = one side's
+ * wording is fully contained in the other's). */
+function containmentSimilarity(a, b) {
+    const count = (xs) => {
+        const m = new Map();
+        for (const x of xs) {
+            m.set(x, (m.get(x) ?? 0) + 1);
+        }
+        return m;
+    };
+    const ca = count(a);
+    const cb = count(b);
+    let inter = 0;
+    for (const [t, n] of ca) {
+        inter += Math.min(n, cb.get(t) ?? 0);
+    }
+    const denom = Math.min(a.length, b.length);
+    return denom === 0 ? 0 : inter / denom;
+}
+/** The element kinds whose text counts as BOUND text for duplicate
+ * detection (the methodology's "elements" are the model's provisions):
+ * requirement statements, term definitions, and instrument/variant
+ * definitions. Attribute/symbol/characteristic definitions are parameter
+ * semantics — formulaic by design ("maximum load that can be applied"),
+ * not source provisions; the congruence gate watches their provenance. */
+exports.DUPLICATE_TEXT_KINDS = new Set([
+    'requirements',
+    'terms',
+    'instruments',
+]);
+/** Flag near-duplicate pairs over the bound-text elements (see the module
+ * header). Deterministic: pairs sorted by (a, b). */
+function flagDuplicatePairs(bindings, manifestUrns, adjudications) {
+    const textByElement = new Map();
+    for (const b of bindings) {
+        if (b.text === undefined || !manifestUrns.has(b.doc)) {
+            continue;
+        }
+        if (!exports.DUPLICATE_TEXT_KINDS.has(b.kind)) {
+            continue;
+        }
+        if (!textByElement.has(b.element)) {
+            textByElement.set(b.element, b.text);
+        }
+    }
+    const elements = [...textByElement.keys()].sort();
+    const tokens = new Map(elements.map(e => [e, normalizeForSimilarity(textByElement.get(e))]));
+    const adjudicationByPair = new Map(adjudications.map(d => [[...d.elements].sort().join(''), d]));
+    const pairs = [];
+    for (let i = 0; i < elements.length; i++) {
+        for (let j = i + 1; j < elements.length; j++) {
+            const ta = tokens.get(elements[i]);
+            const tb = tokens.get(elements[j]);
+            if (Math.min(ta.length, tb.length) < exports.DUPLICATE_MIN_TOKENS) {
+                continue;
+            }
+            const sim = containmentSimilarity(ta, tb);
+            if (sim >= exports.DUPLICATE_SIMILARITY_THRESHOLD) {
+                const key = [elements[i], elements[j]].join('');
+                const adj = adjudicationByPair.get(key);
+                pairs.push({
+                    a: elements[i],
+                    b: elements[j],
+                    similarity: Math.round(sim * 1000) / 1000,
+                    ...(adj ? { adjudication: adj } : {}),
+                });
+            }
+        }
+    }
+    return pairs;
+}
+const SENTENCE_ADDRESS = /\/s[1-9][0-9]*$/;
+/** Compute the text-coverage metric over the composed package. */
+function computeTextCoverage(standard, manifests, config) {
+    const configIssues = [];
+    const bindings = collectProvenanceBindings(standard);
+    const manifestUrns = new Set(manifests.map(m => m.document.urn));
+    const allowances = config?.allowances ?? [];
+    const adjudications = config?.duplicate_adjudications ?? [];
+    const boundFragments = new Set();
+    const boundSentences = new Set();
+    for (const b of bindings) {
+        if (!manifestUrns.has(b.doc)) {
+            continue;
+        }
+        const key = `${b.doc}#${b.fragmentPath}`;
+        if (SENTENCE_ADDRESS.test(b.fragmentPath)) {
+            boundSentences.add(key);
+        }
+        else {
+            boundFragments.add(key);
+        }
+    }
+    const documents = [];
+    const matchedAllowanceSentences = new Map();
+    for (const manifest of manifests) {
+        const urn = manifest.document.urn;
+        const sentences = manifest.sentences;
+        const sentenceCovered = (s) => boundFragments.has(`${urn}#${s.fragment}`) ||
+            boundSentences.has(`${urn}#${s.address}`);
+        // First pass: allowance matches (per document).
+        const allowanceOf = new Map();
+        for (const a of allowances) {
+            const hash = a.address.indexOf('#');
+            if (a.address.slice(0, hash) !== urn) {
+                continue;
+            }
+            const matched = allowanceMatches(a, sentences, urn);
+            const arr = matchedAllowanceSentences.get(a) ?? [];
+            arr.push(...matched);
+            matchedAllowanceSentences.set(a, arr);
+            for (const s of matched) {
+                if (s.modality === 'normative' && !sentenceCovered(s)) {
+                    allowanceOf.set(s.address, a);
+                }
+            }
+        }
+        const uncoveredCounted = [];
+        const uncoveredAllowed = [];
+        let normative = 0;
+        let covered = 0;
+        let informativeUncovered = 0;
+        for (const s of sentences) {
+            if (s.modality !== 'normative') {
+                if (!sentenceCovered(s)) {
+                    informativeUncovered++;
+                }
+                continue;
+            }
+            normative++;
+            if (sentenceCovered(s)) {
+                covered++;
+                continue;
+            }
+            const allowance = allowanceOf.get(s.address);
+            if (allowance) {
+                uncoveredAllowed.push({ sentence: s, allowance });
+            }
+            else {
+                uncoveredCounted.push(s);
+            }
+        }
+        const allowed = uncoveredAllowed.length;
+        documents.push({
+            urn,
+            ...(manifest.document.short !== undefined
+                ? { short: manifest.document.short }
+                : {}),
+            ...(manifest.document.part !== undefined
+                ? { part: manifest.document.part }
+                : {}),
+            total: sentences.length,
+            normative,
+            covered,
+            allowed,
+            uncoveredCounted,
+            uncoveredAllowed,
+            informativeUncovered,
+            ratio: normative === 0 ? 1 : covered / normative,
+            ratioGated: normative - allowed === 0 ? 1 : covered / (normative - allowed),
+        });
+    }
+    // C73 — declaration hygiene (the KNOWN/STALE spirit): an allowance
+    // matching no sentence, or no NORMATIVE sentence, or pinning an
+    // informative-classified sentence is stale; silent declarations fail.
+    // So is a DEAD allowance — one whose matched normative sentences have
+    // all since been bound by model elements: it discharges nothing, so the
+    // declaration outlived the condition that justified it.
+    for (const a of allowances) {
+        const label = `coverage allowance '${a.address}'${a.sentences ? ` pins {${a.sentences.join(' ')}}` : ''}`;
+        const urn = a.address.slice(0, a.address.indexOf('#'));
+        if (!manifestUrns.has(urn)) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: not a document of the shipped manifests (stale or mistyped address) (text-coverage-config)`,
+            });
+            continue;
+        }
+        const matched = matchedAllowanceSentences.get(a) ?? [];
+        const normative = matched.filter(s => s.modality === 'normative');
+        if (matched.length === 0) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: matches no sentence of the decomposition (stale — remove the entry or fix the address/pins) (text-coverage-config)`,
+            });
+        }
+        else if (normative.length === 0) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: every matched sentence is informative — informative sentences need no allowance (stale — remove the entry) (text-coverage-config)`,
+            });
+        }
+        else if (a.sentences) {
+            const informativePins = matched
+                .filter(s => s.modality !== 'normative')
+                .map(s => s.address);
+            if (informativePins.length > 0) {
+                configIssues.push({
+                    check: 'C73',
+                    severity: 'error',
+                    message: `${label}: pinned sentence(s) ${informativePins.join(', ')} are informative-classified — the pin is stale (text-coverage-config)`,
+                });
+            }
+        }
+        if (normative.length > 0) {
+            const bound = (s) => boundFragments.has(`${urn}#${s.fragment}`) ||
+                boundSentences.has(`${urn}#${s.address}`);
+            if (normative.every(bound)) {
+                configIssues.push({
+                    check: 'C73',
+                    severity: 'error',
+                    message: `${label}: every matched normative sentence is now bound by a model element — the allowance discharges nothing (stale — remove the entry; the bindings carry the coverage) (text-coverage-config)`,
+                });
+            }
+        }
+        if (!a.reason) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: missing reason (allowances are deliberate, never silent) (text-coverage-config)`,
+            });
+        }
+    }
+    const flaggedPairs = flagDuplicatePairs(bindings, manifestUrns, adjudications);
+    const unresolvedPairs = flaggedPairs.filter(p => !p.adjudication);
+    // C73 — adjudication hygiene: sorted distinct ids, a reason, and the
+    // pair still flagged (a pair that dropped below the threshold makes
+    // its adjudication STALE — the declaration must die with the condition
+    // that justified it).
+    const flaggedKeys = new Set(flaggedPairs.map(p => [p.a, p.b].sort().join('')));
+    for (const d of adjudications) {
+        const label = `duplicate adjudication {${(d.elements ?? []).join(', ')}}`;
+        if (!Array.isArray(d.elements) || d.elements.length !== 2) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: exactly two element ids make a pair (text-coverage-config)`,
+            });
+            continue;
+        }
+        if (d.elements[0] === d.elements[1]) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: the pair names the same element twice (text-coverage-config)`,
+            });
+        }
+        if (d.elements[0] > d.elements[1]) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: element ids must be sorted (the pair key is order-free) (text-coverage-config)`,
+            });
+        }
+        if (!d.reason) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: missing reason (adjudications are deliberate, never silent) (text-coverage-config)`,
+            });
+        }
+        if (!flaggedKeys.has([...d.elements].sort().join(''))) {
+            configIssues.push({
+                check: 'C73',
+                severity: 'error',
+                message: `${label}: the pair is not currently flagged (similarity below ${exports.DUPLICATE_SIMILARITY_THRESHOLD} or an element unbound/renamed) — the adjudication is STALE (text-coverage-config)`,
+            });
+        }
+    }
+    return { documents, flaggedPairs, unresolvedPairs, configIssues };
+}
+/** The C71 finding text of one uncovered counted sentence. */
+function uncoveredSentenceMessage(urn, s) {
+    const text = s.text.length > 100 ? `${s.text.slice(0, 97)}…` : s.text;
+    return `sentence ${urn}#${s.address} (normative, ${s.kind}): bound by no model element and excluded by no allowance — "${text}" (text-coverage-sentence-uncovered)`;
+}
+/** Load the package at `dir` and compute its text-coverage report (the
+ * `primmel check --coverage` path — null when the package ships no
+ * sentence manifests). Package-load failures propagate as the loader's
+ * diagnostics (checkPackage reports them as issues in the same run). */
+function packageTextCoverageReport(dir, loadPackage) {
+    const { manifests, config, issues } = loadTextCoverageData(dir);
+    if (manifests.length === 0) {
+        return null;
+    }
+    const { standard } = loadPackage(dir);
+    const report = computeTextCoverage(standard, manifests, config);
+    report.configIssues.unshift(...issues);
+    return report;
+}
+/** The human-readable --coverage report (per-document tables with and
+ * without allowances, the uncovered lists, and the duplicate-pair
+ * adjudication status). */
+function formatTextCoverageReport(report) {
+    const lines = [];
+    lines.push('── normative-text coverage (TODO.roadmap/26, concept doc §11.6) ──');
+    lines.push('');
+    lines.push(`  ${'document'.padEnd(26)} ${'sentences'.padStart(9)} ${'normative'.padStart(9)} ${'covered'.padStart(8)} ${'allowed'.padStart(8)} ${'uncovered'.padStart(9)} ${'coverage'.padStart(10)} gated`);
+    for (const d of report.documents) {
+        const label = (d.short ?? d.urn).padEnd(26);
+        const pct = (r) => `${(r * 100).toFixed(1)}%`;
+        lines.push(`  ${label} ${String(d.total).padStart(9)} ${String(d.normative).padStart(9)} ${String(d.covered).padStart(8)} ${String(d.allowed).padStart(8)} ${String(d.uncoveredCounted.length).padStart(9)} ${pct(d.ratio).padStart(10)} ${pct(d.ratioGated)}`);
+    }
+    lines.push('');
+    lines.push('  coverage = covered / normative (without allowances); gated = covered / (normative − allowed) — the 100 % target.');
+    for (const d of report.documents) {
+        if (d.uncoveredCounted.length > 0) {
+            lines.push('');
+            lines.push(`  UNCOVERED normative sentences (${d.short ?? d.urn}) — each is a C71 finding:`);
+            for (const s of d.uncoveredCounted) {
+                lines.push(`    ✗ ${d.urn}#${s.address}  "${truncate(s.text, 90)}"`);
+            }
+        }
+        if (d.uncoveredAllowed.length > 0) {
+            lines.push('');
+            lines.push(`  Allowed exclusions (${d.short ?? d.urn}) — reported, not gated:`);
+            for (const { sentence, allowance } of d.uncoveredAllowed) {
+                lines.push(`    ◦ ${d.urn}#${sentence.address}  "${truncate(sentence.text, 70)}"`);
+                lines.push(`      allowance: ${allowance.reason ?? '(no reason)'}`);
+            }
+        }
+        lines.push('');
+        lines.push(`  informative uncovered (${d.short ?? d.urn}): ${d.informativeUncovered} (reported, never gated)`);
+    }
+    lines.push('');
+    const adjudicated = report.flaggedPairs.length - report.unresolvedPairs.length;
+    lines.push(`  duplicate pairs: ${report.flaggedPairs.length} flagged, ${adjudicated} adjudicated, ${report.unresolvedPairs.length} unresolved (target 0 unresolved — pairs are REPORTED, never auto-failed)`);
+    for (const p of report.flaggedPairs) {
+        const status = p.adjudication ? `${p.adjudication.verdict}` : 'UNRESOLVED';
+        lines.push(`    ${p.adjudication ? '◦' : '?'} ${p.a} × ${p.b} — similarity ${p.similarity} [${status}]${p.adjudication ? ` — ${p.adjudication.reason ?? ''}` : ''}`);
+    }
+    return lines.join('\n');
+}
+function truncate(text, n) {
+    return text.length > n ? `${text.slice(0, n - 1)}…` : text;
+}
+//# sourceMappingURL=text-coverage.js.map