theseus/addon-build/docx-editor/test/corpus.mjs

// Runs the round-trip over a folder of real .docx files.
//
//   node test/corpus.mjs <dir-or-file> [more…]
//
// Reports structure only — block counts, features found, whether the saved
// package is well-formed and whether a second read matches the first. It
// never prints document text, and it writes its output to a temp folder
// rather than next to the input.
import { readFileSync, writeFileSync, readdirSync, statSync, mkdtempSync } from "node:fs";
import path from "node:path";
import os from "node:os";
import { loadAddonLibs, summarise, diff } from "./harness.mjs";

const DocxEditor = loadAddonLibs();
const schema = DocxEditor.schema.build();
const outDir = mkdtempSync(path.join(os.tmpdir(), "docx-corpus-"));

function collect(target, into) {
  let st;
  try { st = statSync(target); } catch { return into; }
  if (st.isDirectory()) {
    for (const name of readdirSync(target)) {
      if (name.startsWith("~$")) continue;            // Word lock files
      collect(path.join(target, name), into);
    }
  } else if (/\.docx$/i.test(target)) {
    into.push(target);
  }
  return into;
}

const targets = [];
for (const arg of process.argv.slice(2)) collect(arg, targets);
if (!targets.length) { console.error("no .docx files found"); process.exit(2); }

const counters = { ok: 0, drift: 0, invalid: 0, failed: 0 };
const featureTally = new Map();
const problems = [];

for (const file of targets) {
  const label = path.basename(file).replace(/[^\x20-\x7e]/g, "?");
  let bytes;
  try { bytes = new Uint8Array(readFileSync(file)); }
  catch (e) { problems.push(`${label}: unreadable (${e.message})`); counters.failed++; continue; }

  try {
    const first = await DocxEditor.read.docxToDoc(bytes, schema);
    for (const f of first.report.features) {
      featureTally.set(f.label, (featureTally.get(f.label) || 0) + 1);
    }
    const saved = await DocxEditor.write.docToDocx(first.doc, {
      originalBytes: bytes, setup: first.setup, meta: first.meta,
    });
    const second = await DocxEditor.read.docxToDoc(saved.bytes, schema);

    // Package sanity, same checks as the fixture round-trip.
    const zip = await DocxEditor.pkg.loadZip(saved.bytes);
    const names = Object.keys(zip.files);
    const bad = [];
    for (const name of names) {
      if (!name.endsWith(".xml") && !name.endsWith(".rels")) continue;
      try { DocxEditor.pkg.parseXml(await zip.file(name).async("string")); }
      catch (e) { bad.push(name); }
    }
    const relsDoc = DocxEditor.pkg.parseXml(await zip.file("word/_rels/document.xml.rels").async("string"));
    const declared = new Set(Array.from(relsDoc.getElementsByTagName("*"))
      .filter((e) => e.localName === "Relationship").map((e) => e.getAttribute("Id")));
    const docXml = await zip.file("word/document.xml").async("string");
    const dangling = [...new Set([...docXml.matchAll(/r:(?:id|embed|link)="([^"]+)"/g)].map((m) => m[1]))]
      .filter((id) => !declared.has(id));
    const missingParts = Array.from(relsDoc.getElementsByTagName("*"))
      .filter((e) => e.localName === "Relationship" && e.getAttribute("TargetMode") !== "External")
      .map((e) => "word/" + (e.getAttribute("Target") || "").replace(/^\.\//, ""))
      .filter((p) => !p.includes("://") && !names.includes(p));

    const d = diff(summarise(first.doc), summarise(second.doc));
    const sizeKb = (bytes.length / 1024).toFixed(0);

    if (bad.length || dangling.length || missingParts.length) {
      counters.invalid++;
      problems.push(`${label}: INVALID PACKAGE — ${[
        bad.length ? `malformed ${bad.join(",")}` : "",
        dangling.length ? `dangling ${dangling.join(",")}` : "",
        missingParts.length ? `missing ${missingParts.join(",")}` : "",
      ].filter(Boolean).join("; ")}`);
      writeFileSync(path.join(outDir, label + ".out.docx"), Buffer.from(saved.bytes));
    } else if (d.length) {
      counters.drift++;
      problems.push(`${label} (${sizeKb} KB, ${first.doc.childCount} blocks): ${d.length} drift — ` +
        d.slice(0, 3).map((x) => x.replace(/"[^"]{40,}"/g, '"…"')).join(" | "));
    } else {
      counters.ok++;
    }
  } catch (e) {
    counters.failed++;
    problems.push(`${label}: THREW — ${(e && e.message || e).toString().slice(0, 160)}`);
  }
}

console.log(`\ncorpus: ${targets.length} documents`);
console.log(`  clean round-trip : ${counters.ok}`);
console.log(`  content drift    : ${counters.drift}`);
console.log(`  invalid package  : ${counters.invalid}`);
console.log(`  threw            : ${counters.failed}`);
console.log(`\nfeatures across the corpus:`);
for (const [label, n] of [...featureTally].sort((a, b) => b[1] - a[1])) {
  console.log(`  ${String(n).padStart(3)}x ${label}`);
}
if (problems.length) {
  console.log(`\nproblems:`);
  for (const p of problems) console.log("  " + p);
}
console.log(`\noutput for failures: ${outDir}`);
process.exit(counters.invalid || counters.failed ? 2 : counters.drift ? 1 : 0);