import { test } from "node:test";
import assert from "node:assert/strict";
import { canonicalScanId, resolveScanErrorId } from "./metadataScan";

// Run with:
//   pnpm --filter yt-dlp-transcript-common exec tsx --test common/ytdlp/metadataScan.test.ts

// yt-dlp's `id` is the NATIVE extractor id. Every consumer of the scan store
// keys by the CANONICAL id (extractVideoId of the URL — the name data/<id>/
// carries). They coincide on YouTube and diverge everywhere else, and keyed
// natively nothing on those platforms would ever settle: the settled set would
// be full of ids no playlist entry and no directory has ever been called.

test("YouTube: native and canonical are the same id", () => {
  assert.equal(
    canonicalScanId({
      id: "dQw4w9WgXcQ",
      webpage_url: "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    }),
    "dQw4w9WgXcQ",
  );
});

test("Rumble: the URL slug wins over the embed id", () => {
  // The case that motivated this: the archive keys by the embed id, a local cue
  // dir is named for the URL slug (see AGENTS.md).
  assert.equal(
    canonicalScanId({
      id: "v2apmfn",
      webpage_url: "https://rumble.com/v2846lb-some-title.html",
    }),
    "v2846lb",
  );
});

test("Twitch: the canonical id keeps the leading v the native id drops", () => {
  assert.equal(
    canonicalScanId({
      id: "2735747903",
      webpage_url: "https://www.twitch.tv/videos/v2735747903",
    }),
    "v2735747903",
  );
});

test("no usable URL falls back to the native id rather than dropping the record", () => {
  assert.equal(canonicalScanId({ id: "abc123" }), "abc123");
  assert.equal(canonicalScanId({ id: "abc123", webpage_url: "not a url" }), "abc123");
  assert.equal(canonicalScanId({}), "");
});

// --- error lines -------------------------------------------------------------
// yt-dlp names the NATIVE id on an ERROR: line and gives no URL to canonicalize.

const URLS = new Map([
  ["v2846lb", "https://rumble.com/v2846lb-some-title.html"],
  ["dQw4w9WgXcQ", "https://www.youtube.com/watch?v=dQw4w9WgXcQ"],
]);

test("an error id resolves through a record already read this run", () => {
  assert.equal(
    resolveScanErrorId("v2apmfn", new Map([["v2apmfn", "v2846lb"]]), URLS),
    "v2846lb",
  );
});

test("an error id that IS a canonical id is taken as-is (the YouTube case)", () => {
  assert.equal(resolveScanErrorId("dQw4w9WgXcQ", new Map(), URLS), "dQw4w9WgXcQ");
});

test("an error id found inside a target URL resolves to that target", () => {
  // Nothing was read for it this run — the very first video failing is exactly
  // when this matters — so the only link left is the URL itself.
  assert.equal(resolveScanErrorId("v2846lb-some", new Map(), URLS), "v2846lb");
});

test("an unrecognizable error id is kept, not dropped", () => {
  // A wrong key for an ERROR row costs one re-read next run. Dropping it would
  // make the backlog never reach zero.
  assert.equal(resolveScanErrorId("whoknows", new Map(), URLS), "whoknows");
});