import FS from 'fs';
import path from 'path';
import JSZip from 'jszip';

const file = path.join(
  'C:/Users/Administrator/Desktop/IN2026A0418001_final_20260713 1.docx'
);
const xml = await (await JSZip.loadAsync(FS.readFileSync(file))).file('word/document.xml').async('string');
const paras = xml.split(/<\/w:p>/);

function plainOf(p) {
  return [...p.matchAll(/<w:t(?:\s[^>]*)?>([\s\S]*?)<\/w:t>/g)]
    .map((x) => x[1])
    .join('')
    .replace(/&amp;/g, '&')
    .replace(/\s+/g, ' ')
    .trim();
}

function blipCount(p) {
  return (p.match(/a:blip/g) || []).length;
}

function drawingCount(p) {
  return (p.match(/<w:drawing/g) || []).length;
}

for (let i = 0; i < paras.length; i++) {
  const plain = plainOf(paras[i]);
  if (!/^Figure\s+\d+/i.test(plain)) continue;
  // look back up to 8 paras for images / A B labels
  const back = [];
  for (let j = Math.max(0, i - 8); j < i; j++) {
    const t = plainOf(paras[j]);
    const b = blipCount(paras[j]);
    const d = drawingCount(paras[j]);
    if (b || d || t) {
      back.push({ j, blips: b, drawings: d, text: t.slice(0, 80) });
    }
  }
  const next = plainOf(paras[i + 1] || '').slice(0, 180);
  console.log('\nCAPTION', plain.slice(0, 100));
  console.log('BACK', JSON.stringify(back, null, 0));
  console.log('NOTE?', next);
}