Kyoo/scanner/scanner/fsscan.py

from contextlib import asynccontextmanager
import os
import re
from logging import getLogger
from mimetypes import guess_file_type
from os.path import dirname, exists, isdir, join

from watchfiles import Change, awatch

from .client import KyooClient
from .database import get_db
from .identifiers.identify import identify
from .models.metadataid import EpisodeId, MetadataId
from .models.videos import For, Video, VideoInfo
from .requests import Request, RequestCreator

logger = getLogger(__name__)


@asynccontextmanager
async def create_scanner():
	async with get_db() as db:
		yield FsScanner(KyooClient(), RequestCreator(db))


class FsScanner:
	def __init__(self, client: KyooClient, requests: RequestCreator):
		self._client = client
		self._requests = requests
		self._info: VideoInfo = None  # type: ignore
		self._root_path = os.environ.get("SCANNER_LIBRARY_ROOT", "/video")
		try:
			pattern = os.environ.get("LIBRARY_IGNORE_PATTERN")
			self._ignore_pattern = re.compile(pattern) if pattern else None
		except re.error as e:
			logger.error(f"Invalid ignore pattern. Ignoring. Error: {e}")

	async def scan(self, path: str | None = None, remove_deleted=False):
		if path is None:
			path = self._root_path
			logger.info("Starting scan at %s. This may take some time...", path)
			if self._ignore_pattern:
				logger.info(f"Applying ignore pattern: {self._ignore_pattern}")
			await self._requests.clear_failed()

		try:
			videos = self.walk_fs(path)

			self._info = await self._client.get_videos_info()

			# TODO: handle unmatched
			to_register = videos - self._info.paths
			to_delete = self._info.paths - videos if remove_deleted else set()

			if (
				not any(to_register)
				and any(to_delete)
				and len(to_delete) == len(self._info.paths)
			):
				logger.warning("All video files are unavailable. Check your disks.")
				return

			# delete stale files before creating new ones to prevent potential conflicts
			if to_delete:
				logger.info("Removing %d stale files.", len(to_delete))
				await self._client.delete_videos(to_delete)

			if to_register:
				logger.info("Found %d new files to register.", len(to_register))
				await self._register(to_register)

			logger.info("Scan finished for %s.", path)
		except Exception as e:
			logger.error("Unexpected error while running scan.", exc_info=e)

	async def monitor(self):
		logger.info(f"Watching for new files in {self._root_path}")
		async for changes in awatch(self._root_path, ignore_permission_denied=True):
			try:
				for event, file in changes:
					if not isdir(file) and not is_video(file):
						continue
					if (
						self._ignore_pattern and self._ignore_pattern.match(file)
					) or is_ignored_path(file):
						logger.info("Ignoring event %s for file %s", event, file)
						continue

					match event:
						case Change.added if isdir(file):
							logger.info("New dir found: %s", file)
							await self.scan(file)
						case Change.added:
							logger.info("New video found: %s", file)
							await self._register([file])
						case Change.deleted:
							logger.info("Delete video at: %s", file)
							await self._client.delete_videos([file])
						case Change.modified:
							pass
			except Exception as e:
				logger.error("Unexpected error while monitoring files.", exc_info=e)

	async def _register(self, videos: list[str] | set[str]):
		# TODO: we should probably chunk those
		vids: list[Video] = []
		for path in list(videos)[:1]:
			try:
				vid = await identify(path)
				vid = self._match(vid)
				vids.append(vid)
			except Exception as e:
				logger.error("Couldn't identify %s.", path, exc_info=e)
		created = await self._client.create_videos(vids)

		await self._requests.enqueue(
			[
				Request(
					kind=x.guess.kind,
					title=x.guess.title,
					year=next(iter(x.guess.years), None),
					external_id=x.guess.external_id,
					videos=[Request.Video(id=x.id, episodes=x.guess.episodes)],
				)
				for x in created
				if not any(x.entries) and x.guess.kind != "extra"
			]
		)

	def _match(self, video: Video) -> Video:
		video.for_ = []

		year_info = (
			self._info.guesses[video.guess.title]
			if video.guess.title in self._info.guesses
			else {}
		)
		slugs = set(
			x
			for x in (
				[
					year_info[str(y)].slug if str(y) in year_info else None
					for y in video.guess.years
				]
				+ ([year_info["unknown"].slug] if "unknown" in year_info else [])
			)
			if x is not None
		)

		if video.guess.kind == "movie":
			for slug in slugs:
				video.for_.append(For.Movie(movie=slug))

			for k, v in video.guess.external_id.items():
				video.for_.append(
					For.ExternalId(external_id={k: MetadataId(data_id=v)})
				)
		else:
			for ep in video.guess.episodes:
				if ep.season is not None:
					for slug in slugs:
						video.for_.append(
							For.Episode(
								serie=slug, season=ep.season, episode=ep.episode
							)
						)

				for k, v in video.guess.external_id.items():
					video.for_.append(
						For.ExternalId(
							external_id={
								k: EpisodeId(
									serie_id=v, season=ep.season, episode=ep.episode
								)
							}
						)
					)

		# TODO: handle specials & movie as episodes (needs animelist or thexem)
		return video

	def walk_fs(self, root_path: str) -> set[str]:
		videos: set[str] = set()
		for dirpath, dirnames, files in os.walk(root_path):
			# Skip directories with a `.ignore` file
			if ".ignore" in files:
				# Prevents os.walk from descending into this directory
				dirnames.clear()
				continue

			for file in files:
				file_path = os.path.join(dirpath, file)
				# Apply ignore pattern, if any
				if self._ignore_pattern and self._ignore_pattern.match(file_path):
					continue
				if is_video(file_path):
					videos.add(file_path)
		return videos


def is_ignored_path(path: str) -> bool:
	current_path = path
	# Traverse up to the root directory
	while current_path != "/":
		if exists(join(current_path, ".ignore")):
			return True
		current_path = dirname(current_path)
	return False


def is_video(path: str) -> bool:
	(mime, _) = guess_file_type(path, strict=False)
	return mime is not None and mime.startswith("video/")