santifer/career-ops · critical · Error

senjob: still contains posting links but none could be…

Error message

senjob: ${url} still contains posting links but none could be parsed — the listing markup changed

What it means

assertParsedSomething is the senjob provider's silent-zero guard: if a fetched listing page still contains the recognizable posting-link shape (/jobseekers/..._e_<id>.html) but parseListingPage extracted no jobs, the parser — not the board — is broken, so the provider throws instead of returning an empty array. An empty result would falsely report a country board as having no openings, so markup drift must surface loudly.

Solutions

Fetch the live page at https://senjob.com/offres-d-emploi.php and diff its markup against the regexes in providers/senjob.mjs (POSTING_LINK_RE, POSTING_ANCHOR_RE)
Update POSTING_LINK_RE / POSTING_ANCHOR_RE / the <tr> split to match the new markup, keeping the /jobseekers/..._e_{id}.html contract
Check whether links are now HTML-escaped (&) or attribute order changed and relax the regex accordingly
Verify you are not feeding preprocessed HTML (entity-decoded or tag-stripped) into parseListingPage

Example fix

// before — parser no longer matches new markup
const POSTING_LINK_RE = /href="(https:\/\/senjob\.com\/jobseekers\/^"]*?_e_(\d+)\.html)"/i;
// after — tolerate single-quoted or unquoted hrefs introduced by a redesign
const POSTING_LINK_RE = /href=["'](https:\/\/senjob\.com\/jobseekers\/[^"']*?_e_(\d+)\.html)["']/i;

Defensive patterns

Strategy: try-catch

Validate before calling

function looksLikeSenjobListing(html) {
  return /senjob\.com\/jobseekers\/[^"']*?_e_\d+\.html/i.test(String(html ?? ''));
}
// After fetch: if it throws this error, the page had posting links the parser missed —
// save the raw HTML for a fixture before reporting.

Try / catch

try {
  const jobs = await senjobProvider.fetch(entry, ctx);
} catch (e) {
  if (String(e.message).includes('listing markup changed')) {
    console.error('Senjob markup drifted; pin a fixture and update the parser regexes');
    // alert/monitor: this is a parser bug, not an empty board
  } else throw e;
}

Prevention

Treat this error as a parser regression alert, never as 'no jobs'
Keep saved HTML fixtures from live senjob pages in unit tests so markup drift is caught early
Only anchor the parser to the stable URL shape (/jobseekers/..._e_{id}.html), not CSS classes
When the error fires, diff the live HTML against the regexes before patching

When it happens

Trigger: Senjob changes its listing HTML so rows no longer match POSTING_LINK_RE / POSTING_ANCHOR_RE (e.g. new attribute order, escaped hrefs, links built by JS) while the raw href shape still appears in the source; page 1 of the listing fetches OK but parses to zero jobs.

Common situations: A senjob.com site redesign or CMS update; a CDN or caching layer rewriting markup (e.g. minified/attribute-quoted differently); testing against a saved/stripped HTML fixture that contains raw links but no <tr> row structure the parser splits on.

Related errors

AI-assisted analysis of santifer/career-ops@aac998c7ed (2026-09-16). Data as JSON: /api/errors/739cf9a0e204bed1. Report an issue: GitHub.

Appendix: source

Thrown at providers/senjob.mjs:178

  return [...byId.values()].filter((job) => job.title && job.url);
}

/**
 * A listing page that parses to nothing is either a markup change or a block —
 * both are failures, and both must be reported. Returning [] would show up as a
 * board with no openings, which is indistinguishable from a healthy quiet board
 * and is the failure mode that makes scrapers untrustworthy.
 *
 * The emptiness test is the posting-link SHAPE rather than a marker word: if the
 * page still contains posting links and the parser found none, the parser is
 * what broke.
 * @param {string} html
 * @param {string} url
 */
export function assertParsedSomething(html, url) {
  if (!/\/jobseekers\/[^"]*?_e_\d+\.html/i.test(String(html ?? ''))) return;
  throw new Error(
    `senjob: ${url} still contains posting links but none could be parsed — the listing markup changed`,
  );
}

/** @type {Provider} */
export default {
  id: 'senjob',

  detect(entry) {
    return entry?.provider === 'senjob' ? { url: LIST_URL } : null;
  },

  async fetch(entry, ctx) {
    // `max_pages` on the portals entry is the user's setting; `ctx.maxPages` is a
    // caller-side bound — verify-portals' health probe passes 1. Reading only the
    // latter ignored the configuration entirely. Same shape as alibaba.mjs.
    const entryMaxPages = Number.isInteger(entry?.max_pages) && entry.max_pages > 0
      ? Math.min(entry.max_pages, MAX_PAGES_CAP)

View on GitHub (pinned to aac998c7ed)