atomic-design-poc/docs/project/refactor-backlog-setup/refactor-backlog/tools/baseline-scan.mjs

#!/usr/bin/env node
// Baseline scanner for 00-baseline.md: duplication % and C# cyclomatic complexity.
// Both are the same crude text scan, so they share one file.
//
// ponytail: line-window hashing, not token-based like jscpd, and regex method
// detection that does not understand C# expression-bodied members or nested
// lambdas. Deterministic and zero-install, which is what a *baseline* needs.
// Upgrade path: jscpd for duplication, a Roslyn analyzer for C# complexity —
// only if a ticket's before/after needs more precision than "did it move".
//
// Usage: node baseline-scan.mjs [--dup] [--complexity]   (default: both)

import { readFileSync, readdirSync, statSync } from 'node:fs';
import { createHash } from 'node:crypto';
import { join, relative } from 'node:path';

const ROOT = new URL('../../../../..', import.meta.url).pathname.replace(/\/$/, '');
const SKIP_DIR =
  /(^|\/)(node_modules|dist|coverage|bin|obj|\.angular|\.git|storybook-static.*|TestResults|public|openzaak)$/;
const SKIP_FILE = /(api-client\.ts|\.Designer\.cs|AppDbContextModelSnapshot\.cs)$/;
const SCAN = ['apps', 'libs', 'backend/src', 'backend/tests', 'e2e'];
const WINDOW = 6; // duplicate = >= 6 consecutive normalized lines seen elsewhere

function walk(dir, out = []) {
  for (const e of readdirSync(dir)) {
    const p = join(dir, e);
    if (SKIP_DIR.test(p)) continue;
    if (statSync(p).isDirectory()) walk(p, out);
    else if (/\.(ts|cs)$/.test(p) && !SKIP_FILE.test(p)) out.push(p);
  }
  return out;
}

// Module attribution: the unit the baseline table reports on.
function moduleOf(rel) {
  let m;
  if ((m = rel.match(/^apps\/(ssp|behandelportal)\/src\/app\/([^/]+)\//)))
    return `${m[1] === 'ssp' ? 'ssp' : 'bhp'}/${m[2]}`;
  if (rel.startsWith('apps/ssp/')) return 'ssp/root';
  if (rel.startsWith('apps/behandelportal/')) return 'bhp/root';
  if ((m = rel.match(/^libs\/(shared|beheer)\/src\/([^/]+)\//))) return `libs/${m[1]}/${m[2]}`;
  if (rel.startsWith('libs/')) return rel.split('/').slice(0, 2).join('/');
  if (rel === 'backend/src/BigRegister.Api/Program.cs') return 'backend/Program.cs';
  if ((m = rel.match(/^backend\/src\/BigRegister\.Api\/([^/]+)\//))) return `backend/${m[1]}`;
  if (rel.startsWith('backend/tests/')) return 'backend/tests';
  if (rel.startsWith('e2e/')) return 'e2e';
  return 'other';
}

const files = SCAN.flatMap((d) => {
  try {
    return walk(join(ROOT, d));
  } catch {
    return [];
  }
}).map((p) => ({ abs: p, rel: relative(ROOT, p) }));

// --- Duplication ------------------------------------------------------------
// Normalize away formatting/comments, hash every WINDOW-line run, and mark a run
// duplicated when its hash occurs in more than one place.
function duplication() {
  const norm = new Map(); // rel -> [{line, text}]
  for (const f of files) {
    const kept = [];
    readFileSync(f.abs, 'utf8')
      .split('\n')
      .forEach((raw, i) => {
        const t = raw
          .replace(/\/\/.*$/, '')
          .replace(/\s+/g, ' ')
          .trim();
        if (t.length > 3) kept.push({ line: i + 1, text: t });
      });
    norm.set(f.rel, kept);
  }

  const seen = new Map(); // hash -> [{rel, idx}]
  for (const [rel, lines] of norm)
    for (let i = 0; i + WINDOW <= lines.length; i++) {
      const h = createHash('sha1')
        .update(
          lines
            .slice(i, i + WINDOW)
            .map((l) => l.text)
            .join('\n'),
        )
        .digest('hex');
      (seen.get(h) ?? seen.set(h, []).get(h)).push({ rel, idx: i });
    }

  const dupLines = new Map(); // rel -> Set(line)
  const clones = new Map(); // "relA|relB" -> count
  for (const occ of seen.values()) {
    if (occ.length < 2) continue;
    for (const { rel, idx } of occ) {
      const set = dupLines.get(rel) ?? dupLines.set(rel, new Set()).get(rel);
      for (let k = 0; k < WINDOW; k++) set.add(norm.get(rel)[idx + k].line);
    }
    const pair = [...new Set(occ.map((o) => o.rel))].sort();
    if (pair.length > 1) {
      const key = pair.slice(0, 2).join(' | ');
      clones.set(key, (clones.get(key) ?? 0) + 1);
    }
  }

  const per = new Map(); // module -> {total, dup}
  for (const [rel, lines] of norm) {
    const mod = moduleOf(rel);
    const e = per.get(mod) ?? per.set(mod, { total: 0, dup: 0 }).get(mod);
    e.total += lines.length;
    e.dup += dupLines.get(rel)?.size ?? 0;
  }

  console.log('## Duplication (normalized lines, window=' + WINDOW + ')\n');
  console.log('| Module | Sig. lines | Duplicated | % |');
  console.log('|---|---:|---:|---:|');
  let T = 0,
    D = 0;
  for (const [mod, e] of [...per].sort((a, b) => b[1].dup / b[1].total - a[1].dup / a[1].total)) {
    T += e.total;
    D += e.dup;
    console.log(`| ${mod} | ${e.total} | ${e.dup} | ${((100 * e.dup) / e.total).toFixed(1)}% |`);
  }
  console.log(`| **TOTAL** | **${T}** | **${D}** | **${((100 * D) / T).toFixed(1)}%** |`);

  console.log('\n### Top clone pairs (distinct duplicated windows)\n');
  for (const [pair, n] of [...clones].sort((a, b) => b[1] - a[1]).slice(0, 15))
    console.log(`- ${n} × — ${pair}`);
}

// --- C# cyclomatic complexity (approximate) ---------------------------------
// Two numbers per file. FILE CC (sum of decision points) is exact enough to trust.
// Per-METHOD CC uses depth-aware regex detection and is the approximate one.
const BRANCH =
  /\bif\s*\(|\bwhile\s*\(|\bfor\s*\(|\bforeach\s*\(|\bcase\s+|\bcatch\s*[({]|\?\?|&&|\|\||\?\.|\bwhen\s+/g;
const TYPE_DECL =
  /^\s*(?:\[[^\]]*\]\s*)*(?:public|private|internal|protected|static|sealed|abstract|partial|file|\s)*\b(?:class|record|struct|interface|enum|namespace)\b/;
const NOT_A_CALL =
  /^(if|for|foreach|while|switch|catch|using|lock|return|throw|new|await|yield|else|do|fixed|checked)$/;
// A member: optional attrs/modifiers, a return type, a name, then `(`.
const SIG =
  /^\s*(?:\[[^\]]*\]\s*)*(?:(?:public|private|internal|protected|static|async|override|virtual|sealed|partial|extern|new|unsafe)\s+)*[\w<>,\[\]?.]+\s+(\w+)\s*(?:<[^>()]*>)?\s*\(/;

function branchesIn(line) {
  return (line.replace(/\/\/.*$/, '').match(BRANCH) ?? []).length;
}

function csComplexity() {
  const rows = [];
  const fileCc = [];
  for (const f of files.filter((f) => f.rel.endsWith('.cs'))) {
    const lines = readFileSync(f.abs, 'utf8').split('\n');
    let depth = 0,
      cur = null,
      total = 1,
      code = 0;
    for (let i = 0; i < lines.length; i++) {
      const l = lines[i];
      const bare = l.replace(/\/\/.*$/, '');
      const b = branchesIn(l);
      total += b;
      if (bare.trim().length > 1 && !/^\s*(\/\/|\/\*|\*)/.test(l)) code++;

      const opens = (bare.match(/{/g) ?? []).length;
      const closes = (bare.match(/}/g) ?? []).length;

      if (cur) {
        cur.lines++;
        cur.cc += b;
        // Expression-bodied member: `=> expr;` with no block of its own.
        if (cur.depth === null && /=>/.test(bare)) {
          if (/;\s*$/.test(bare) && opens === closes) {
            rows.push(cur);
            cur = null;
          } else if (opens > closes) cur.depth = depth;
        } else if (cur.depth === null && opens > closes) cur.depth = depth;
        else if (cur.depth !== null && depth + opens - closes <= cur.depth) {
          rows.push(cur);
          cur = null;
        }
      } else if (!TYPE_DECL.test(l)) {
        const m = SIG.exec(bare);
        // `Name(` must not be a call/keyword, and the line must not be a statement.
        if (
          m &&
          !NOT_A_CALL.test(m[1]) &&
          !/^\s*(var|return|await)\b/.test(bare) &&
          !/;\s*$/.test(bare.replace(/=>.*/, ''))
        )
          cur = {
            file: f.rel,
            name: m[1],
            line: i + 1,
            cc: 1 + b,
            lines: 1,
            depth: opens > closes ? depth : null,
          };
      }
      depth += opens - closes;
    }
    if (cur) rows.push(cur);
    fileCc.push({ file: f.rel, cc: total, code });
  }

  const p = (arr, q) =>
    arr.slice().sort((a, b) => a - b)[Math.min(arr.length - 1, Math.floor(q * arr.length))] ?? 0;

  const perFile = new Map();
  for (const r of fileCc) {
    const mod = moduleOf(r.file);
    (perFile.get(mod) ?? perFile.set(mod, []).get(mod)).push(r);
  }
  console.log('\n\n## C# complexity — per module\n');
  console.log(
    '| Module | Files | Σ file CC | max file CC | Methods | max method CC | p90 method CC | CC>10 |',
  );
  console.log('|---|---:|---:|---:|---:|---:|---:|---:|');
  const perMethod = new Map();
  for (const r of rows) {
    const mod = moduleOf(r.file);
    (perMethod.get(mod) ?? perMethod.set(mod, []).get(mod)).push(r);
  }
  for (const [mod, fs] of [...perFile].sort(
    (a, b) => Math.max(...b[1].map((r) => r.cc)) - Math.max(...a[1].map((r) => r.cc)),
  )) {
    const ms = perMethod.get(mod) ?? [];
    const ccs = ms.map((r) => r.cc);
    console.log(
      `| ${mod} | ${fs.length} | ${fs.reduce((a, r) => a + r.cc, 0)} | ${Math.max(...fs.map((r) => r.cc))} | ` +
        `${ms.length} | ${ccs.length ? Math.max(...ccs) : 0} | ${p(ccs, 0.9)} | ${ms.filter((r) => r.cc > 10).length} |`,
    );
  }

  console.log('\n### C# files by CC (top 12)\n');
  console.log('| File CC | Code lines | File |');
  console.log('|---:|---:|---|');
  for (const r of fileCc.sort((a, b) => b.cc - a.cc).slice(0, 12))
    console.log(`| ${r.cc} | ${r.code} | ${r.file} |`);

  console.log('\n### C# methods over CC 10 (approximate detection)\n');
  console.log('| CC | Lines | Method | Location |');
  console.log('|---:|---:|---|---|');
  for (const r of rows.filter((r) => r.cc > 10).sort((a, b) => b.cc - a.cc))
    console.log(`| ${r.cc} | ${r.lines} | \`${r.name}\` | ${r.file}:${r.line} |`);

  const all = rows.map((r) => r.lines);
  console.log(
    `\nMethod-length distribution (n=${rows.length}): p50 ${p(all, 0.5)}, p90 ${p(all, 0.9)}, p99 ${p(all, 0.99)}, max ${Math.max(...all)}`,
  );
}

const args = process.argv.slice(2);
const all = args.length === 0;
if (all || args.includes('--dup')) duplication();
if (all || args.includes('--complexity')) csComplexity();