import { removeMediaFile } from "../lib/mediaTier-server";
import path from "node:path";
import fs from "fs-extra";
import type { Paths } from "../lib/paths";
import { isDoNotClean } from "../lib/doNotClean-server";
import { readVttProvenance } from "../lib/subtitleProvenance";
import {
  WHISPER_FILENAME,
  isEnglishVtt,
  listTranscriptVtts,
} from "../lib/videoStatus";

const { pathExists, readdir } = fs;

// Delete the YouTube auto-caption VTTs that our own transcript has superseded —
// the manual counterpart to the `supersededAutoSubs` snapshot bucket, and the
// one irreversible step in the replace-auto-captions lane. It is therefore
// never wired into the auto-queue: the backup only disappears on an explicit
// click, so an AI-vs-YouTube comparison stays possible until then.
//
// Deliberately narrow. A track is removed only when ALL of these hold:
//   - the dir has transcript.json (our transcript already won the index pick),
//   - the track is an ENGLISH one (the en / en-orig / en-US / en-en-* family) —
//     foreign-language tracks are separate content and are left alone,
//   - a fresh 4 KB provenance sniff still says "asr" — a manual or unclassifiable
//     track is never touched,
//   - the dir is not marked do-not-clean (same shield as the Clean-audio sweep).
//
// Recovering a purged track means re-running "Download missing subs".

export type PurgeSupersededAutoSubsOptions = {
  channelSlug: string;
  paths: Paths;
  // When set, restrict the sweep to these ids (the bucket the button showed).
  // Omitted = walk every video dir, like the Clean-audio sweep.
  ids?: string[];
  onLog?: (msg: string) => void;
  signal?: AbortSignal;
};

export type PurgeSupersededAutoSubsResult = {
  inspected: number;
  cleanedDirs: number;
  removedFiles: number;
  skipped: number;
};

export async function purgeSupersededAutoSubs({
  channelSlug,
  paths,
  ids,
  onLog,
  signal,
}: PurgeSupersededAutoSubsOptions): Promise<PurgeSupersededAutoSubsResult> {
  const log = onLog ?? ((m: string) => console.log(m));
  const dataDir = path.join(paths.channelsDir, channelSlug, "data");
  if (!(await pathExists(dataDir))) {
    log(`No data directory for ${channelSlug}`);
    return { inspected: 0, cleanedDirs: 0, removedFiles: 0, skipped: 0 };
  }
  const onDisk = await readdir(dataDir);
  const dirs = ids
    ? ids.filter((id) => onDisk.includes(id))
    : onDisk;

  let cleanedDirs = 0;
  let removedFiles = 0;
  let skipped = 0;

  for (const id of dirs) {
    if (signal?.aborted) break;
    const videoDir = path.join(dataDir, id);
    const entries = await readdir(videoDir).catch(() => [] as string[]);
    // No transcript of our own yet — nothing has been superseded, so the VTT is
    // still this video's only transcript. Never touch it.
    if (!entries.includes(WHISPER_FILENAME)) continue;
    const englishVtts = listTranscriptVtts(entries).filter(isEnglishVtt);
    if (englishVtts.length === 0) continue;
    if (await isDoNotClean(videoDir)) {
      log(`Skipped ${id} (marked do not clean)`);
      skipped++;
      continue;
    }
    let removedHere = 0;
    for (const name of englishVtts) {
      const provenance = await readVttProvenance(videoDir, name);
      if (provenance !== "asr") {
        log(`Kept ${id}/${name} (${provenance} captions — not auto-generated)`);
        skipped++;
        continue;
      }
      await removeMediaFile(videoDir, name); // the one rm of a video-dir entry (release 17)
      log(`Removed ${id}/${name}`);
      removedFiles++;
      removedHere++;
    }
    if (removedHere > 0) cleanedDirs++;
  }

  const skippedNote = skipped > 0 ? ` Skipped ${skipped}.` : "";
  log(
    `Purged ${removedFiles} superseded auto-caption file(s) from ${cleanedDirs} of ${dirs.length} video dir(s).${skippedNote}`,
  );

  return {
    inspected: dirs.length,
    cleanedDirs,
    removedFiles,
    skipped,
  };
}