create-pmos 0.15.0 → 0.16.0

package/workspace/repo-scan.test.js

@@ -1,327 +0,0 @@
-#!/usr/bin/env node
-/* Unit suite for lib/repo-scan.js — the product-ingest scanner.
- *
- * No framework, no mocks: every case builds a REAL temporary repository on disk and scans it, the
- * same way workspace/test.js builds real repos for the server suite. The filesystem behaviours that
- * matter here (a symlink, an oversized file, a directory that is not there) cannot be exercised
- * honestly against a stub.
- *
- * The suite is organised around the one invariant the module exists to hold:
- *
- *   A · proposal shape + the declared/guessed separation (a guess must never read as a declaration)
- *   B · the guess itself — single candidate, ambiguity, nothing, frontmatter-is-not-body
- *   C · absent inputs are CALM and STATED (missing planning/, missing okf/, empty directories)
- *   D · failed reads are RED and carry a reason — never silently skipped, never fatal to the scan
- *   E · ADRs
- *   F · purity + determinism (the scan writes nothing and repeats itself)
- *   G · the real repository, as a corpus check
- *
- * Exits non-zero if any check fails.
- */
-'use strict';
-const fs = require('fs');
-const os = require('os');
-const path = require('path');
-const assert = require('assert');
-
-const scan = require('./lib/repo-scan');
-
-const results = [];
-const ok = (name, cond, extra = '') => { results.push({ name, cond }); console.log(`${cond ? '✓' : '✗'} ${name}${extra ? ' — ' + extra : ''}`); };
-
-const TMP = fs.mkdtempSync(path.join(os.tmpdir(), 'pmos-repo-scan-'));
-const made = [];
-function makeRepo(name) {
-  const root = path.join(TMP, name);
-  fs.mkdirSync(root, { recursive: true });
-  made.push(root);
-  return root;
-}
-function write(root, rel, text) {
-  const full = path.join(root, rel);
-  fs.mkdirSync(path.dirname(full), { recursive: true });
-  fs.writeFileSync(full, text);
-  return full;
-}
-/** An initiative record with whatever frontmatter the case needs. */
-function record(fm, body = 'Body text with nothing anchor-shaped in it.\n') {
-  const keys = Object.entries(fm).map(([k, v]) => `${k}: ${v}`).join('\n');
-  return `---\n${keys}\n---\n\n${body}`;
-}
-const logHas = (r, needle) => r.log.some((l) => (l.message + ' ' + (l.path || '')).includes(needle));
-const byPath = (r, p) => r.initiatives.find((i) => i.path === p);
-
-try {
-  // ── A · proposal shape + the declared/guessed separation ────────────────────────────────────
-  const A = makeRepo('full');
-  write(A, 'planning/initiatives/declared.md', record({
-    type: 'Initiative', title: '"declared"', timestamp: '2026-07-29',
-    id: 'declared', lane: 'feature', parent_kr: 'q3-2026:o1:kr1',
-  }, 'This record also mentions health:tech in its prose, which must change nothing.\n'));
-  write(A, 'planning/initiatives/guessable.md', record({
-    type: 'Initiative', title: '"guessable"', timestamp: '2026-07-29', id: 'guessable', lane: 'chore',
-  }, 'Carved out of the work anchored to q3-2026:o1:kr2 last quarter.\n'));
-  write(A, 'planning/initiatives/ambiguous.md', record({
-    type: 'Initiative', title: '"ambiguous"', timestamp: '2026-07-29', id: 'ambiguous', lane: 'chore',
-  }, 'Touches q3-2026:o1:kr1 and also health:tech, so it is not obvious which one owns it.\n'));
-  write(A, 'planning/initiatives/bare.md', record({
-    type: 'Initiative', title: '"bare"', timestamp: '2026-07-29', id: 'bare', lane: 'chore',
-  }));
-  write(A, 'planning/initiatives/empty-anchor.md', record({
-    type: 'Initiative', title: '"empty anchor"', timestamp: '2026-07-29', id: 'empty-anchor',
-    lane: 'chore', parent_kr: '',
-  }));
-  write(A, 'planning/initiatives/README.md', record({ type: 'Reference', title: '"readme"', timestamp: '2026-07-29' }));
-  const a = scan.scanRepo(A);
-
-  ok('A1: a readable repo scans ok with the documented shape',
-    a.ok === true && a.error === null && Array.isArray(a.initiatives) && Array.isArray(a.adrs)
-      && Array.isArray(a.log) && Array.isArray(a.unreadable),
-    JSON.stringify({ ok: a.ok, n: a.initiatives.length }));
-
-  const dec = byPath(a, 'planning/initiatives/declared.md');
-  ok('A2: a declared parent_kr lands in declaredAnchor and NOTHING lands in guessedAnchor',
-    !!dec && dec.declaredAnchor === 'q3-2026:o1:kr1' && dec.guessedAnchor === null,
-    JSON.stringify({ declared: dec && dec.declaredAnchor, guessed: dec && dec.guessedAnchor }));
-
-  ok('A3: a declaration is never second-guessed — a body mention of a DIFFERENT anchor is ignored',
-    !!dec && dec.guessedAnchor === null && dec.why.includes('not an inference'),
-    dec ? dec.why.slice(0, 60) : 'missing');
-
-  ok('A4: the separation invariant — no row EVER carries both a declaration and a guess',
-    a.initiatives.every((r) => !(r.declaredAnchor && r.guessedAnchor)),
-    JSON.stringify(a.initiatives.map((r) => [r.id, !!r.declaredAnchor, !!r.guessedAnchor])));
-
-  ok('A5: every proposed row states its anchor reasoning in `why` (a field a PM can check)',
-    a.initiatives.every((r) => typeof r.why === 'string' && r.why.length > 40),
-    JSON.stringify(a.initiatives.map((r) => (r.why || '').length)));
-
-  ok('A6: a non-Initiative markdown file is NOT proposed and its absence is stated in the log',
-    !byPath(a, 'planning/initiatives/README.md') && logHas(a, 'declares `type: Reference`, not Initiative'),
-    JSON.stringify(a.initiatives.map((r) => r.id)));
-
-  // ── B · the guess ───────────────────────────────────────────────────────────────────────────
-  const g = byPath(a, 'planning/initiatives/guessable.md');
-  ok('B1: exactly one anchor-shaped token in the body becomes a GUESS, with its line number',
-    !!g && g.declaredAnchor === null && g.guessedAnchor === 'q3-2026:o1:kr2' && /line \d+/.test(g.why),
-    g ? g.guessedAnchor + ' / ' + g.why.slice(0, 70) : 'missing');
-
-  ok('B2: the guess `why` says out loud that a mention is not a declaration and D59 still bites',
-    !!g && g.why.includes('not a declaration') && g.why.includes('D59'),
-    g ? g.why.slice(-90) : 'missing');
-
-  const amb = byPath(a, 'planning/initiatives/ambiguous.md');
-  ok('B3: two candidates yields NO guess and names both — the scan does not pick a winner',
-    !!amb && amb.guessedAnchor === null && amb.why.includes('q3-2026:o1:kr1') && amb.why.includes('health:tech'),
-    amb ? amb.why.slice(0, 100) : 'missing');
-
-  const bare = byPath(a, 'planning/initiatives/bare.md');
-  ok('B4: nothing anchor-shaped anywhere → both fields null and the record is called unanchored',
-    !!bare && bare.declaredAnchor === null && bare.guessedAnchor === null && bare.why.includes('unanchored'),
-    bare ? bare.why.slice(0, 80) : 'missing');
-
-  const emptyA = byPath(a, 'planning/initiatives/empty-anchor.md');
-  ok('B5: `parent_kr:` present but empty is reported as present-and-empty, not as absent',
-    !!emptyA && emptyA.declaredAnchor === null && emptyA.why.includes('present but empty'),
-    emptyA ? emptyA.why.slice(0, 70) : 'missing');
-
-  // The frontmatter/body boundary. A status line quoting an anchor is prose about an anchor, not a
-  // second declaration site — if this leaked, every record with a chatty `status:` would acquire a
-  // guess nobody wrote.
-  const B = makeRepo('fm-boundary');
-  write(B, 'planning/initiatives/chatty.md', record({
-    type: 'Initiative', title: '"chatty"', timestamp: '2026-07-29', id: 'chatty', lane: 'chore',
-    status: '"BLOCKED — was going to hang off q3-2026:o1:kr1 until that KR was retired"',
-  }));
-  const b = scan.scanRepo(B);
-  ok('B6: an anchor quoted in ANOTHER frontmatter key is not mined — only `parent_kr` declares',
-    b.initiatives.length === 1 && b.initiatives[0].guessedAnchor === null && b.initiatives[0].declaredAnchor === null,
-    JSON.stringify(b.initiatives.map((r) => [r.declaredAnchor, r.guessedAnchor])));
-
-  ok('B7: the guesser is narrow by design — a bare legacy `O1.KR3` in prose is NOT guessed at',
-    scan.guessAnchorFromBody('body mentions O1.KR3 and **O1** only\n', 1).length === 0,
-    JSON.stringify(scan.guessAnchorFromBody('body mentions O1.KR3 and **O1**\n', 1)));
-
-  // ── C · absence is calm and stated ──────────────────────────────────────────────────────────
-  const C = makeRepo('bare-folder');
-  const c = scan.scanRepo(C);
-  ok('C1: a folder with no planning/ and no okf/ is a SUCCESSFUL read of nothing, not an error',
-    c.ok === true && c.error === null && c.initiatives.length === 0 && c.adrs.length === 0 && c.unreadable.length === 0,
-    JSON.stringify({ ok: c.ok, err: c.error }));
-
-  ok('C2: …and both absences are STATED in the log rather than left to be inferred from emptiness',
-    logHas(c, 'planning/initiatives') && logHas(c, 'does not exist') && logHas(c, 'okf'),
-    JSON.stringify(c.log.map((l) => l.path)));
-
-  const C2 = makeRepo('empty-dirs');
-  fs.mkdirSync(path.join(C2, 'planning', 'initiatives'), { recursive: true });
-  fs.mkdirSync(path.join(C2, 'okf', 'products', 'x', 'adr'), { recursive: true });
-  const c2 = scan.scanRepo(C2);
-  ok('C3: directories that exist but hold nothing are called empty reads, not failures',
-    c2.ok === true && c2.unreadable.length === 0 && logHas(c2, 'holds no markdown files')
-      && logHas(c2, 'exists but holds no markdown files'),
-    JSON.stringify(c2.log.filter((l) => l.level !== 'info').map((l) => l.message)));
-
-  // ── D · failed reads are red, carry a reason, and never end the scan ────────────────────────
-  const D = makeRepo('broken');
-  write(D, 'planning/initiatives/good.md', record({
-    type: 'Initiative', title: '"good"', timestamp: '2026-07-29', id: 'good', lane: 'chore', parent_kr: 'health:tech',
-  }));
-  write(D, 'planning/initiatives/no-type.md', '---\ntitle: "no type"\ntimestamp: 2026-07-29\nid: no-type\n---\n\nbody\n');
-  write(D, 'planning/initiatives/unterminated.md', '---\ntype: Initiative\ntitle: "oops"\nid: unterminated\n\nbody that was meant to be below a closing fence\n');
-  write(D, 'planning/initiatives/huge.md', '---\ntype: Initiative\ntitle: "huge"\nid: huge\n---\n' + 'x'.repeat(scan.MAX_BYTES + 10));
-  fs.symlinkSync(path.join(D, 'planning', 'initiatives', 'good.md'), path.join(D, 'planning', 'initiatives', 'linked.md'));
-  const d = scan.scanRepo(D);
-  const un = (p) => d.unreadable.find((u) => u.path === p);
-
-  ok('D1: a broken file does not end the scan — the readable record beside it is still proposed',
-    d.ok === true && d.initiatives.length === 1 && d.initiatives[0].id === 'good',
-    JSON.stringify({ ok: d.ok, ids: d.initiatives.map((r) => r.id) }));
-
-  ok('D2: a file with no `type:` is unreadable WITH ITS REASON, never silently skipped',
-    !!un('planning/initiatives/no-type.md') && un('planning/initiatives/no-type.md').error.includes('D63'),
-    un('planning/initiatives/no-type.md') ? un('planning/initiatives/no-type.md').error.slice(0, 70) : 'absent');
-
-  ok('D3: an unterminated frontmatter fence is reported as malformed, not parsed leniently',
-    !!un('planning/initiatives/unterminated.md') && un('planning/initiatives/unterminated.md').error.includes('never closes'),
-    un('planning/initiatives/unterminated.md') ? un('planning/initiatives/unterminated.md').error.slice(0, 70) : 'absent');
-
-  ok('D4: a file past the size cap is refused with its size — not read, nothing claimed about it',
-    !!un('planning/initiatives/huge.md') && /\d+ bytes/.test(un('planning/initiatives/huge.md').error),
-    un('planning/initiatives/huge.md') ? un('planning/initiatives/huge.md').error.slice(0, 80) : 'absent');
-
-  ok('D5: a symlinked record is refused rather than followed (provenance, not paranoia)',
-    !!un('planning/initiatives/linked.md') && un('planning/initiatives/linked.md').error.includes('symbolic link'),
-    un('planning/initiatives/linked.md') ? un('planning/initiatives/linked.md').error.slice(0, 60) : 'absent');
-
-  ok('D6: every unreadable file also appears in the human log as an error — neither list is partial',
-    d.unreadable.every((u) => d.log.some((l) => l.level === 'error' && l.path === u.path && l.message === u.error)),
-    JSON.stringify({ unreadable: d.unreadable.length, errorLines: d.log.filter((l) => l.level === 'error').length }));
-
-  ok('D7: the log warns that unreadable files are NOT represented in the proposal above',
-    logHas(d, 'NOT represented in the proposal'),
-    JSON.stringify(d.log.filter((l) => l.level === 'warn').map((l) => l.message.slice(0, 40))));
-
-  // A root that is not there, and a root that is a file. Both are "the scan could not run" — the one
-  // case where ok is false — and both must still return the full shape so no consumer crashes.
-  const missing = scan.scanRepo(path.join(TMP, 'no-such-repo'));
-  ok('D8: a missing root is ok:false with a reason, and still returns every array',
-    missing.ok === false && typeof missing.error === 'string' && missing.error.length > 20
-      && Array.isArray(missing.initiatives) && Array.isArray(missing.adrs)
-      && Array.isArray(missing.log) && Array.isArray(missing.unreadable),
-    (missing.error || '').slice(0, 60));
-
-  const filePath = write(makeRepo('is-a-file'), 'thing.md', 'x');
-  const notDir = scan.scanRepo(filePath);
-  ok('D9: a root that is a file is ok:false, not an empty proposal',
-    notDir.ok === false && notDir.error.includes('not a directory'), (notDir.error || '').slice(0, 50));
-
-  const noArg = scan.scanRepo(undefined);
-  ok('D10: no path at all is refused without throwing', noArg.ok === false && noArg.initiatives.length === 0,
-    (noArg.error || '').slice(0, 50));
-
-  // ── D11/D12 · identity edges ────────────────────────────────────────────────────────────────
-  const E0 = makeRepo('identity');
-  write(E0, 'planning/initiatives/stem-id.md', record({
-    type: 'Initiative', title: '"no id"', timestamp: '2026-07-29', lane: 'chore', parent_kr: 'health:tech',
-  }));
-  write(E0, 'planning/initiatives/dup-a.md', record({
-    type: 'Initiative', title: '"dup a"', timestamp: '2026-07-29', id: 'same', lane: 'chore', parent_kr: 'health:tech',
-  }));
-  write(E0, 'planning/initiatives/dup-b.md', record({
-    type: 'Initiative', title: '"dup b"', timestamp: '2026-07-29', id: 'same', lane: 'chore', parent_kr: 'health:tech',
-  }));
-  write(E0, 'planning/initiatives/index.md', '# initiatives\n\n- a listing, no frontmatter\n');
-  const e0 = scan.scanRepo(E0);
-  ok('D11: a record with no `id:` falls back to the filename stem and the fallback is logged as a guess',
-    !!byPath(e0, 'planning/initiatives/stem-id.md') && byPath(e0, 'planning/initiatives/stem-id.md').id === 'stem-id'
-      && logHas(e0, 'A filename is a guess at an identity'),
-    JSON.stringify(e0.initiatives.map((r) => r.id)));
-
-  ok('D12: two records claiming one id are BOTH kept and the collision is flagged for a human',
-    e0.initiatives.filter((r) => r.id === 'same').length === 2 && logHas(e0, 'propose the same id'),
-    JSON.stringify(e0.initiatives.map((r) => r.id)));
-
-  ok('D13: OKF-reserved index.md is not treated as a record, and the log says why',
-    !byPath(e0, 'planning/initiatives/index.md') && logHas(e0, 'OKF-reserved filename'),
-    JSON.stringify(e0.initiatives.map((r) => r.path)));
-
-  // ── E · ADRs ────────────────────────────────────────────────────────────────────────────────
-  const E = makeRepo('adrs');
-  write(E, 'okf/products/thing/adr/d01-first.md', '---\ntype: ADR\ntitle: D01 — the first decision\ntimestamp: 2026-07-29\n---\n\nbody\n');
-  write(E, 'okf/products/thing/adr/d02-second.md', '---\ntype: ADR\ntimestamp: 2026-07-29\n---\n\nno title here\n');
-  write(E, 'okf/products/thing/adr/notes.md', '---\ntype: Reference\ntitle: notes\ntimestamp: 2026-07-29\n---\n\nbody\n');
-  write(E, 'okf/core/concepts/thing.md', '---\ntype: Concept\ntitle: c\ntimestamp: 2026-07-29\n---\n\nbody\n');
-  write(E, 'okf/products/other/adr/d09-nested.md', '---\ntype: ADR\ntitle: D09 — nested elsewhere\ntimestamp: 2026-07-29\n---\n\nbody\n');
-  const e = scan.scanRepo(E);
-  ok('E1: ADRs are found in every okf/**/adr/ directory, id from the filename, title from frontmatter',
-    e.adrs.length === 3
-      && e.adrs.some((x) => x.id === 'd01-first' && x.title === 'D01 — the first decision' && x.path === 'okf/products/thing/adr/d01-first.md')
-      && e.adrs.some((x) => x.id === 'd09-nested'),
-    JSON.stringify(e.adrs.map((x) => x.id)));
-
-  ok('E2: an ADR with no title gets a NULL title — a filename is never printed as a title',
-    (e.adrs.find((x) => x.id === 'd02-second') || {}).title === null && logHas(e, 'null title rather than a filename'),
-    JSON.stringify(e.adrs.find((x) => x.id === 'd02-second')));
-
-  ok('E3: a non-ADR file inside adr/ is not imported on the strength of its location, and is logged',
-    !e.adrs.some((x) => x.id === 'notes') && logHas(e, 'declares `type: Reference`, not ADR'),
-    JSON.stringify(e.adrs.map((x) => x.id)));
-
-  ok('E4: a concept outside adr/ is not swept in',
-    !e.adrs.some((x) => x.path.includes('concepts')), JSON.stringify(e.adrs.map((x) => x.path)));
-
-  ok('E5: the ADR log line says the human still has to accept or supersede each one',
-    logHas(e, 'accepted or superseded by a human'),
-    JSON.stringify(e.log.filter((l) => (l.message || '').includes('ADR(s)')).map((l) => l.message)));
-
-  // ── F · purity + determinism ────────────────────────────────────────────────────────────────
-  function snapshot(dir) {
-    const out = [];
-    (function walk(d) {
-      for (const ent of fs.readdirSync(d, { withFileTypes: true }).sort((x, y) => (x.name < y.name ? -1 : 1))) {
-        const full = path.join(d, ent.name);
-        const st = fs.lstatSync(full);
-        out.push(`${path.relative(dir, full)}|${st.size}|${st.mtimeMs}|${st.isDirectory() ? 'd' : 'f'}`);
-        if (st.isDirectory()) walk(full);
-      }
-    })(dir);
-    return out.join('\n');
-  }
-  const before = snapshot(A);
-  const a2 = scan.scanRepo(A);
-  const after = snapshot(A);
-  ok('F1: the scan writes NOTHING — no file, size or mtime in the tree changes', before === after,
-    before === after ? '' : 'tree changed');
-
-  let same = true;
-  try { assert.deepStrictEqual(a2, a); } catch { same = false; }
-  ok('F2: two scans of the same tree return an identical proposal (stable order, no clock in it)', same);
-
-  ok('F3: proposals are path-sorted, so an import diff is signal rather than reordering noise',
-    a.initiatives.map((r) => r.path).join() === [...a.initiatives.map((r) => r.path)].sort().join(),
-    a.initiatives.map((r) => r.path).join(', '));
-
-  // ── G · the real corpus ─────────────────────────────────────────────────────────────────────
-  // The fixtures above are written by this file and therefore agree with it by construction. This
-  // one check runs against ~100 records nobody wrote for the scanner.
-  const self = scan.scanRepo(path.join(__dirname, '..'));
-  ok('G1: this repository scans ok and proposes its real initiative records',
-    self.ok === true && self.initiatives.length > 50,
-    `${self.initiatives.length} initiatives, ${self.adrs.length} ADRs, ${self.unreadable.length} unreadable`);
-
-  ok('G2: on a real corpus the declared/guessed separation still holds row for row',
-    self.initiatives.every((r) => !(r.declaredAnchor && r.guessedAnchor)),
-    `${self.initiatives.filter((r) => r.declaredAnchor).length} declared, ${self.initiatives.filter((r) => r.guessedAnchor).length} guessed`);
-
-  ok('G3: the real ADR set is found with titles',
-    self.adrs.length > 40 && self.adrs.every((x) => x.id && x.path) && self.adrs.filter((x) => x.title).length > 40,
-    `${self.adrs.length} ADRs, ${self.adrs.filter((x) => !x.title).length} untitled`);
-} catch (e) {
-  ok('no exceptions during suite', false, e && e.stack ? e.stack.split('\n').slice(0, 2).join(' @ ') : String(e));
-} finally {
-  try { fs.rmSync(TMP, { recursive: true, force: true }); } catch {}
-  const failed = results.filter((r) => !r.cond);
-  console.log(`\n${results.length - failed.length}/${results.length} passed${failed.length ? ' — FAILURES: ' + failed.map((f) => f.name).join('; ') : ''}`);
-  process.exit(failed.length ? 1 : 0);
-}