src/api/iserv-parse.js

// Parsers that run INSIDE the browser page: puppeteer's page.evaluate serializes a function, so every function here
// is self-contained (its helpers sit inside it). They get the HTML as the server sent it and read it through
// DOMParser. Page scripts do not run in that document and cannot change what is read: in the live DOM, Bootstrap
// tooltips move the title attribute of the status icon away, so a done task would look open.
//
// Port of the parsers of the IServ filter project (iserv.py: parse_task_list, parse_description, parse_start),
// which are tested against fake pages of the older and the current IServ layout. Keep both in step.

/**
 * Task list: { "<task id>": { tags, start, due, done, expired } }
 * tags: text of the "Tags" column ('' if empty), start: 'yyyy-mm-dd' (older layouts only), due: 'yyyy-mm-ddThh:mm',
 * done: the status icon says done/submitted, expired: it says expired (guessed pattern, only "Erledigt" was seen
 * live). Columns are found by their header, never by position (the column after the due date is the teacher's
 * feedback).
 */
export function parseListHtml(html) {
  const doc = new DOMParser().parseFromString(html, 'text/html');
  // BeautifulSoup's get_text(sep, strip=True): the stripped text nodes, joined
  const text = (el, sep = ' ') => {
    const parts = [];
    const walk = (node) => {
      for (const child of node.childNodes) {
        if (child.nodeType === 3) {
          const t = child.textContent.trim();
          if (t) parts.push(t);
        } else if (child.nodeType === 1 && !['SCRIPT', 'STYLE'].includes(child.tagName)) {
          walk(child);
        }
      }
    };
    walk(el);
    return parts.join(sep);
  };
  const headerIndex = (row, names) => {
    const table = row.closest('table');
    const head = table ? [...table.querySelectorAll('tr')].find((tr) => tr.querySelector('th')) : null;
    const texts = head ? [...head.querySelectorAll('th, td')].map((c) => text(c).replace(/:+$/, '')) : [];
    return texts.findIndex((t) => names.includes(t));
  };
  const tagsOf = (cell) => {
    const all = text(cell);
    const italic = [...cell.querySelectorAll('i')].map((i) => text(i)).join(' ');
    return all === italic || ['keine', 'none'].includes(all.replace(/^[()]+|[()]+$/g, '').toLowerCase()) ? '' : all;
  };
  const doneRe = /erledigt|abgegeben|done|submitted|completed/i;
  const expiredRe = /abgelaufen|expired/i;
  const out = {};
  for (const a of doc.querySelectorAll('a[href*="/exercise/show/"]')) {
    const m = /\/exercise\/show\/(\d+)/.exec(a.getAttribute('href'));
    const row = a.closest('tr');
    if (!m || !row || m[1] in out) continue;
    const cells = [...row.querySelectorAll('td')];
    let start = null;
    let due = null;
    for (const td of cells) {
      const ds = td.getAttribute('data-sort') || '';
      if (/^\d{8}$/.test(ds) && start === null) {
        start = `${ds.slice(0, 4)}-${ds.slice(4, 6)}-${ds.slice(6)}`;
      } else if (/^\d{14}$/.test(ds) && due === null) {
        due = `${ds.slice(0, 4)}-${ds.slice(4, 6)}-${ds.slice(6, 8)}T${ds.slice(8, 10)}:${ds.slice(10, 12)}`;
      }
    }
    const col = headerIndex(row, ['Tags']);
    const titles = [...row.querySelectorAll('[title]')].map((e) => e.getAttribute('title')).join(' ');
    out[m[1]] = {
      tags: col !== -1 && col < cells.length ? tagsOf(cells[col]) : '',
      start,
      due,
      done: doneRe.test(titles),
      expired: expiredRe.test(titles),
    };
  }
  return out;
}

/**
 * Task page: { description, start }
 * description: only the task's own description (never own submission, teacher feedback, creator, participants),
 * as text with the paragraphs and line breaks of the page. start: 'yyyy-mm-dd[Thh:mm]' from the "Starttermin" column.
 */
export function parseTaskHtml(html) {
  const doc = new DOMParser().parseFromString(html, 'text/html');
  const text = (el, sep = ' ') => {
    const parts = [];
    const walk = (node) => {
      for (const child of node.childNodes) {
        if (child.nodeType === 3) {
          const t = child.textContent.trim();
          if (t) parts.push(t);
        } else if (child.nodeType === 1 && !['SCRIPT', 'STYLE'].includes(child.tagName)) {
          walk(child);
        }
      }
    };
    walk(el);
    return parts.join(sep);
  };
  const th = (names) => [...doc.querySelectorAll('th')].find((t) => names.includes(text(t).replace(/:+$/, '')));
  const headerIndex = (row, names) => {
    const table = row.closest('table');
    const head = table ? [...table.querySelectorAll('tr')].find((tr) => tr.querySelector('th')) : null;
    const texts = head ? [...head.querySelectorAll('th, td')].map((c) => text(c).replace(/:+$/, '')) : [];
    return texts.findIndex((t) => names.includes(t));
  };
  const htmlText = (el) => {
    const copy = el.cloneNode(true);
    copy.querySelectorAll('script, style').forEach((n) => n.remove());
    copy.querySelectorAll('br').forEach((br) => br.replaceWith('\n'));
    copy.querySelectorAll('p, div, li, ul, ol, table, tr, td, th, h1, h2, h3, h4, h5, h6, blockquote, pre, dt, dd, hr, section, article')
      .forEach((block) => { block.before('\n'); block.after('\n'); });
    return copy.textContent
      .split(/\r\n|[\n\r\v\f]/)
      .map((line) => line.split(/\s+/).filter(Boolean).join(' '))
      .filter(Boolean)
      .join('\n');
  };

  // Older versions: the element after a "Beschreibung:" label. Current versions: the text block in the task panel,
  // the .panel that holds the date table. A task without a description yields '' and never another block of the page.
  const label = [...doc.querySelectorAll('div, th, dt, h4, h5, strong, label')]
    .find((t) => ['Beschreibung', 'Description'].includes(text(t, '').replace(/:+$/, '')));
  let box = label ? label.nextElementSibling : null;
  if (!box && label && label.tagName === 'TH') {
    for (let s = label.nextElementSibling; s; s = s.nextElementSibling) {
      if (s.tagName === 'TD') { box = s; break; }
    }
  }
  const dates = !box ? th(['Starttermin', 'Beginn', 'Start date', 'Abgabetermin', 'Due date']) : null;
  if (dates) {
    const panel = dates.closest('.panel');
    box = panel ? [...panel.querySelectorAll('div.text-break-word')].find((d) => !d.closest('form')) || null : null;
  } else if (!box) {
    box = [...doc.querySelectorAll('div.text-break-word.p-3')].find((e) => !e.closest('form[name="submission"]')) || null;
  }

  let start = null;
  const startNames = ['Starttermin', 'Beginn', 'Start date'];
  const startTh = th(startNames);
  const table = startTh ? startTh.closest('table') : null;
  const row = table ? [...table.querySelectorAll('tr')].find((tr) => tr.querySelector('td')) : null;
  const col = row ? headerIndex(row, startNames) : -1;
  const cells = row ? [...row.querySelectorAll('td')] : [];
  const m = col !== -1 && col < cells.length
    ? /(\d{2})\.(\d{2})\.(\d{4})(?:\D+(\d{2}):(\d{2}))?/.exec(text(cells[col]))
    : null;
  if (m) start = `${m[3]}-${m[2]}-${m[1]}` + (m[4] ? `T${m[4]}:${m[5]}` : '');

  return { description: box ? htmlText(box) : '', start };
}

/**
 * Provided files of a task page: [{ href, filename }]. Only the "provided files" form, never own submission or
 * teacher feedback. Per row: the download link if there is one (current versions also put a view link and a menu
 * toggle in each row), the file name from the row's name link. Port of parse_attachment_links in iserv.py.
 */
export function parseAttachmentLinks(html) {
  const doc = new DOMParser().parseFromString(html, 'text/html');
  const form = doc.querySelector('form[name="iserv_exercise_attachment"]');
  const out = [];
  for (const row of form ? form.querySelectorAll('tr') : []) {
    const anchors = [...row.querySelectorAll('a[href]')].filter(
      (a) => !['/', '#'].includes(a.getAttribute('href')) && !a.hasAttribute('data-toggle'),
    );
    if (anchors.length === 0) continue;
    const pick = anchors.find((a) => a.getAttribute('href').includes('/download/')) || anchors[0];
    const named = row.querySelector('a.text-break-word') || pick;
    out.push({ href: pick.getAttribute('href'), filename: (named.textContent || '').replace(/\s+/g, ' ').trim() });
  }
  return out;
}

/** 'dd.mm.yyyy[ hh:mm]' somewhere in a text to 'yyyy-mm-dd[Thh:mm]', null if there is none. Plain Node, not for page.evaluate. */
export function isoFromText(text) {
  const m = /(\d{1,2})\.(\d{1,2})\.(\d{4})(?:\D+(\d{1,2}):(\d{2}))?/.exec(String(text || ''));
  if (!m) return null;
  const [, d, mo, y, h, mi] = m;
  const date = `${y}-${mo.padStart(2, '0')}-${d.padStart(2, '0')}`;
  return h ? `${date}T${h.padStart(2, '0')}:${mi}` : date;
}