import type { MetadataRoute } from "next";
import { bff } from "@/lib/bff";

/**
 * /sitemap.xml — one file for the whole public corpus (~33k URLs, well under
 * the 50,000-URL / 50 MB sitemap limits; the content is finite, so a single
 * file is correct for the foreseeable future).
 *
 * Why a single file (not generateSitemaps): Next's generateSitemaps() serves
 * shards at /sitemap/<id>.xml and does NOT expose an index at /sitemap.xml —
 * but /sitemap.xml is exactly where robots.txt + search engines look. A plain
 * default export is served at /sitemap.xml. (If the URL count ever nears 50k,
 * switch to a sitemap index referencing sharded children.)
 *
 * How this stays cheap on the BFF:
 *   • Two projection-only BFF calls (`/sitemap/content`, `/sitemap/users`)
 *     return every URL in a single sweep each — no per-document fan-out.
 *   • The fetches are cached 24h (revalidate) AND the rendered XML is
 *     ISR-cached, so crawlers hit Next's cache, not the BFF (~one query/day).
 *     Tag "sitemap" lets the BFF force-refresh on new signups / re-seeds.
 */

export const revalidate = 86400; // 24h

const SITE_URL =
  process.env.NEXT_PUBLIC_SITE_URL || "https://prabhupadasays.org";

// Top-level hubs that aren't content leaves but should be indexed.
const STATIC_PATHS = [
  "/",
  "/about",
  "/books",
  "/letters",
  "/transcriptions",
  "/disciples",
  "/glossary",
  "/essays",
  "/magazines",
  "/journals",
  "/other-documents",
];

async function fetchUrls(path: string): Promise<string[]> {
  try {
    const res = await bff<{ urls: string[] }>(path, {
      revalidate,
      tags: ["sitemap"],
    });
    return res?.urls ?? [];
  } catch {
    return [];
  }
}

/**
 * All indexable paths: static hubs + book/division landing pages (derived
 * from the content leaves) + every content leaf + public profiles. Resilient —
 * a failed BFF call still yields the static hubs, so /sitemap.xml never 500s.
 */
async function allPaths(): Promise<string[]> {
  const [content, users] = await Promise.all([
    fetchUrls("/sitemap/content"),
    fetchUrls("/sitemap/users"),
  ]);

  const hubs = new Set<string>();
  for (const p of content) {
    const book = p.match(/^\/books\/([^/]+)/);
    if (book) hubs.add(`/books/${book[1]}`);
    const division = p.match(/^\/books\/(sb|cc)\/([^/]+)\/\d+\/\d+$/);
    if (division) hubs.add(`/books/${division[1]}/${division[2]}`);
  }

  return [...new Set([...STATIC_PATHS, ...hubs, ...content, ...users])];
}

export default async function sitemap(): Promise<MetadataRoute.Sitemap> {
  const paths = await allPaths();
  return paths.map((path) => ({
    url: `${SITE_URL}${path}`,
    changeFrequency: "weekly",
    priority: path === "/" ? 1 : STATIC_PATHS.includes(path) ? 0.8 : 0.7,
  }));
}