/**
 * HTML link-extraction helpers.
 *
 * The browser-side extraction lives inline in `crawler.extractLinks` so it
 * can run via `page.evaluate`. This module holds the bits that are either
 * pure (serializable without a DOM) or shared with other extractors.
 */

/**
 * Parse a `<meta http-equiv="refresh" content="...">` content attribute and
 * return the redirect URL, or null if the value doesn't carry one.
 *
 * Accepts any of these real-world shapes:
 *   "0;url=/next"
 *   "0; URL=/next"
 *   "5; url='/next'"
 *   "3;URL=https://example.com/"
 *
 * A plain `<meta http-equiv="refresh" content="5">` (delay only, no URL)
 * returns null.
 */
export function parseMetaRefreshUrl(content: string | null | undefined): string | null {
  if (!content) return null;
  // The first segment before the semicolon is the delay. Anything after is
  // parameter=value pairs, though in practice only `url=` is used.
  const semi = content.indexOf(";");
  if (semi === -1) return null;
  const rest = content.slice(semi + 1).trim();
  const m = rest.match(/^url\s*=\s*(.*)$/i);
  if (!m) return null;
  let url = m[1]!.trim();
  if (url.length === 0) return null;

  // Quoted URLs delimit with the matching quote — the URL may legitimately
  // contain `;`, so we can't just split on it. Unterminated quotes are a
  // malformed directive; return null rather than a truncated URL.
  if (url.startsWith('"') || url.startsWith("'")) {
    const quote = url[0]!;
    const end = url.indexOf(quote, 1);
    if (end === -1) return null;
    url = url.slice(1, end);
  } else {
    // Unquoted: terminate at the next parameter separator. Without this,
    // `0;url=/next;foo=bar` would captured as `/next;foo=bar` and queue
    // the wrong URL.
    const sep = url.indexOf(";");
    if (sep !== -1) url = url.slice(0, sep).trim();
  }

  return url.length > 0 ? url : null;
}