meeb 2e826b1c37
Merge pull request #1224 from tcely/patch-6
Avoid the signal games for `Media.save`
2025-07-17 17:44:36 +10:00

550 lines
22 KiB
Python

from functools import partial
from pathlib import Path
from tempfile import TemporaryDirectory
from django.conf import settings
from django.db import IntegrityError
from django.db.models.signals import pre_save, post_save, pre_delete, post_delete
from django.db.transaction import atomic, on_commit
from django.dispatch import receiver
from django.utils import timezone
from django.utils.translation import gettext_lazy as _
from background_task.signals import (
task_created, task_started, task_successful, task_rescheduled, task_failed,
)
from background_task.models import Task
from common.logger import log
from common.models import TaskHistory
from common.utils import glob_quote, mkdir_p
from .models import Source, Media, Metadata
from .tasks import (
delete_task_by_media,
get_media_download_task, get_media_metadata_task, get_media_thumbnail_task,
map_task_to_instance,
delete_all_media_for_source, rename_media, save_all_media_for_source,
check_source_directory_exists, download_source_images, index_source,
download_media_file, download_media_metadata, download_media_image,
)
from .utils import delete_file
from .filtering import filter_media
@receiver(pre_save, sender=Source)
def source_pre_save(sender, instance, **kwargs):
source = instance # noqa: F841
# Triggered before a source is saved, if the schedule has been updated recreate
# its indexing task
try:
existing_source = Source.objects.get(pk=instance.pk)
except Source.DoesNotExist:
log.debug(f'source_pre_save signal: no existing source: {sender} - {instance}')
return
args = ( str(instance.pk), )
check_source_directory_exists.call_local(*args)
existing_copy_channel_images = existing_source.copy_channel_images
new_copy_channel_images = instance.copy_channel_images
if new_copy_channel_images and not (existing_copy_channel_images or instance.is_playlist):
download_source_images(str(instance.pk))
existing_dirpath = existing_source.directory_path.resolve(strict=True)
new_dirpath = instance.directory_path.resolve(strict=False)
if existing_dirpath != new_dirpath:
path_name = lambda p: p.name
relative_dir = existing_source.directory
rd_parents = Path(relative_dir).parents
rd_parents_set = set(map(path_name, rd_parents))
ad_parents = existing_dirpath.parents
ad_parents_set = set(map(path_name, ad_parents))
# the names in the relative path are also in the absolute path
parents_count = len(ad_parents_set.intersection(rd_parents_set))
work_directory = existing_dirpath
for _count in range(parents_count, 0, -1):
work_directory = work_directory.parent
if not Path(work_directory).resolve(strict=True).is_relative_to(Path(settings.DOWNLOAD_ROOT)):
work_directory = Path(settings.DOWNLOAD_ROOT)
with TemporaryDirectory(suffix=('.'+new_dirpath.name), prefix='.tmp.', dir=work_directory) as tmp_dir:
tmp_dirpath = Path(tmp_dir)
existed = None
previous = existing_dirpath.rename(tmp_dirpath / 'previous')
try:
if new_dirpath.exists():
existed = new_dirpath.rename(tmp_dirpath / 'existed')
mkdir_p(new_dirpath.parent)
previous.rename(new_dirpath)
except Exception:
# try to preserve the directory, if anything went wrong
previous.rename(existing_dirpath)
raise
else:
existing_dirpath = previous = None
if existed and existed.is_dir():
existed = existed.rename(new_dirpath / '.existed')
for entry_path in existed.iterdir():
try:
target = new_dirpath / entry_path.name
if not target.exists():
entry_path = entry_path.rename(target)
except Exception as e:
log.exception(e)
try:
existed.rmdir()
except Exception as e:
log.exception(e)
elif existed:
try:
existed = existed.rename(new_dirpath / ('.existed-' + new_dirpath.name))
except Exception as e:
log.exception(e)
recreate_index_source_task = (
existing_source.name != instance.name or
existing_source.index_schedule != instance.index_schedule
)
if recreate_index_source_task:
# Indexing schedule has changed, recreate the indexing task
TaskHistory.schedule(
index_source,
str(instance.pk),
eta=instance.task_run_at_dt,
remove_duplicates=True,
vn_fmt=_('Index media from source "{}"'),
vn_args=(instance.name,),
)
@receiver(post_save, sender=Source)
def source_post_save(sender, instance, created, **kwargs):
source = instance
# Check directory exists and create an indexing task for newly created sources
if created:
check_source_directory_exists(str(source.pk))
if source.copy_channel_images and not source.is_playlist:
download_source_images(str(source.pk))
if source.is_active:
log.info(f'Scheduling first media indexing for source: {source.name}')
TaskHistory.schedule(
index_source,
str(source.pk),
delay=600,
vn_fmt=_('Index media from source "{}"'),
vn_args=(source.name,),
)
TaskHistory.schedule(
save_all_media_for_source,
str(source.pk),
remove_duplicates=True,
vn_fmt = _('Checking all media for "{}"'),
vn_args=(
source.name,
),
)
@receiver(pre_delete, sender=Source)
def source_pre_delete(sender, instance, **kwargs):
# Triggered before a source is deleted, delete all media objects to trigger
# the Media models post_delete signal
source = instance
log.info(f'Deactivating source: {instance.name}')
instance.deactivate()
# Fetch the media source
sqs = Source.objects.filter(filter_text=str(source.pk))
if sqs.count():
media_source = sqs[0]
# Schedule deletion of media
on_commit(partial(
TaskHistory.schedule,
delete_all_media_for_source,
str(media_source.pk),
str(media_source.name),
str(media_source.directory_path),
vn_fmt = _('Deleting all media for source "{}"'),
vn_args=(
media_source.name,
),
))
@receiver(task_created, dispatch_uid='sync.signals.task_task_created')
@atomic(durable=False)
def task_task_created(sender, task=None, **kwargs):
if task is None:
return
task_obj = task
th, created = TaskHistory.objects.get_or_create(
task_id=str(task_obj.pk),
name=task_obj.task_name,
queue=task_obj.queue,
)
th.scheduled_at = task_obj.run_at
th.priority = (100 - task_obj.priority)
th.repeat = task_obj.repeat
th.repeat_until = task_obj.repeat_until
th.task_params = list(task_obj.params())
th.verbose_name = task_obj.verbose_name
th.save()
if created:
log.debug(f'Created a new task history record: {th.pk}: {th.verbose_name}')
@receiver(task_started, dispatch_uid='sync.signals.task_task_started')
@atomic(durable=False)
def task_task_started(sender, **kwargs):
locked_tasks = Task.objects.locked(timezone.now())
for task_obj in locked_tasks:
th, created = TaskHistory.objects.get_or_create(
task_id=str(task_obj.pk),
name=task_obj.task_name,
queue=task_obj.queue,
)
th.attempts += 1
th.end_at = task_obj.locked_at
th.priority = (100 - task_obj.priority)
th.repeat = task_obj.repeat
th.repeat_until = task_obj.repeat_until
th.start_at = task_obj.locked_at
th.task_params = list(task_obj.params())
th.verbose_name = task_obj.verbose_name
th.save()
if created:
log.debug(f'Started a new task history record: {th.pk}: {th.verbose_name}')
@receiver(task_rescheduled, dispatch_uid='sync.signals.task_task_rescheduled')
@atomic(durable=False)
def task_task_rescheduled(sender, task=None, **kwargs):
if task is None:
return
now_dt = timezone.now()
task_obj = task
th, created = TaskHistory.objects.get_or_create(
task_id=str(task_obj.pk),
name=task_obj.task_name,
queue=task_obj.queue,
)
th.elapsed += (
now_dt - task_obj.locked_at
).total_seconds()
th.end_at = now_dt
th.scheduled_at = task_obj.run_at
th.start_at = task_obj.locked_at
th.save()
if created:
log.debug(f'Rescheduled a new task history record: {th.pk}: {th.verbose_name}')
def merge_completed_task_into_history(task_id, task_obj):
th, created = TaskHistory.objects.get_or_create(
task_id=str(task_id),
name=task_obj.task_name,
queue=task_obj.queue,
)
th.elapsed += (
(task_obj.failed_at or task_obj.run_at) - task_obj.locked_at
).total_seconds()
th.end_at = task_obj.run_at
th.failed_at = task_obj.failed_at
th.last_error = task_obj.last_error
th.repeat = task_obj.repeat
th.repeat_until = task_obj.repeat_until
th.start_at = task_obj.locked_at
th.verbose_name = task_obj.verbose_name
th.save()
@receiver(task_successful, dispatch_uid='sync.signals.task_task_successful')
@atomic(durable=False)
def task_task_successful(sender, task_id, completed_task, **kwargs):
merge_completed_task_into_history(task_id, completed_task)
@receiver(task_failed, dispatch_uid='sync.signals.task_task_failed')
@atomic(durable=False)
def task_task_failed(sender, task_id, completed_task, **kwargs):
merge_completed_task_into_history(task_id, completed_task)
# Triggered after a task fails by reaching its max retry attempts
obj, url = map_task_to_instance(completed_task, using_history=False)
if isinstance(obj, Source):
log.error(f'Permanent failure for source: {obj} task: {completed_task}')
obj.has_failed = True
obj.save()
if isinstance(obj, Media) and completed_task.task_name == "sync.tasks.download_media_metadata":
log.error(f'Permanent failure for media: {obj} task: {completed_task}')
obj.skip = True
obj.save()
@receiver(post_save, sender=Media)
def media_post_save(sender, instance, created, **kwargs):
media = instance
# If the media is skipped manually, bail.
if instance.manual_skip:
return
# Triggered after media is saved
skip_changed = False
can_download_changed = False
# Reset the skip flag if the download cap has changed if the media has not
# already been downloaded
downloaded = instance.downloaded
existing_media_thumbnail_task = get_media_thumbnail_task(str(instance.pk))
existing_media_metadata_task = get_media_metadata_task(str(instance.pk))
existing_media_download_task = get_media_download_task(str(instance.pk))
if not downloaded:
# the decision to download was already made if a download task exists
if not existing_media_download_task:
# Recalculate the "can_download" flag, this may
# need to change if the source specifications have been changed
if media.has_metadata:
if instance.get_format_str():
if not instance.can_download:
instance.can_download = True
can_download_changed = True
else:
if instance.can_download:
instance.can_download = False
can_download_changed = True
# Recalculate the "skip_changed" flag
skip_changed = filter_media(instance)
else:
# Downloaded media might need to be renamed
# Check settings before any rename tasks are scheduled
rename_sources_setting = getattr(settings, 'RENAME_SOURCES') or list()
create_rename_task = (
(
media.source.directory and
media.source.directory in rename_sources_setting
) or
settings.RENAME_ALL_SOURCES
)
if create_rename_task:
rename_media(str(media.pk))
# If the media is missing metadata schedule it to be downloaded
if not (media.skip or media.has_metadata or existing_media_metadata_task):
log.info(f'Scheduling task to download metadata for: {media.url}')
verbose_name = _('Downloading metadata for: {}: "{}"')
download_media_metadata(
str(media.pk),
verbose_name=verbose_name.format(media.key, media.name),
)
# If the media is missing a thumbnail schedule it to be downloaded (unless we are skipping this media)
if not media.thumb_file_exists:
media.thumb = None
if not (media.skip or media.thumb or existing_media_thumbnail_task):
thumbnail_url = media.thumbnail
if thumbnail_url:
log.info(
'Scheduling task to download thumbnail'
f' for: {media.name} from: {thumbnail_url}'
)
TaskHistory.schedule(
download_media_image,
str(media.pk),
thumbnail_url,
vn_fmt=_('Downloading thumbnail for "{}"'),
vn_args=(media.name,),
)
media_file_exists = False
try:
media_file_exists |= instance.media_file_exists
media_file_exists |= instance.filepath.exists()
except OSError as e:
log.exception(e)
pass
# If the media has not yet been downloaded schedule it to be downloaded
if not (media_file_exists or existing_media_download_task):
# The file was deleted after it was downloaded, skip this media.
if instance.can_download and instance.downloaded:
skip_changed = True if not instance.skip else False
instance.skip = True
downloaded = False
if (instance.source.download_media and instance.can_download) and not (
instance.skip or downloaded or existing_media_download_task):
TaskHistory.schedule(
download_media_file,
str(media.pk),
remove_duplicates=True,
vn_fmt=_('Downloading media for "{}"'),
vn_args=(media.name,),
)
# Save the instance if any changes were required
if skip_changed or can_download_changed:
Media.objects.filter(
pk=instance.pk,
).update(
can_download=instance.can_download,
skip=instance.skip,
)
@receiver(pre_delete, sender=Media)
def media_pre_delete(sender, instance, **kwargs):
# Triggered before media is deleted, delete any unlocked scheduled tasks
log.info(f'Deleting tasks for media: {instance.name}')
delete_task_by_media('sync.tasks.download_media_metadata', (str(instance.pk),))
# Remove thumbnail file for deleted media
if instance.thumb:
instance.thumb.delete(save=False)
# Save the metadata site & thumbnail URL to the metadata column
existing_metadata = instance.loaded_metadata
metadata_str = instance.metadata or '{}'
arg_dict = instance.metadata_loads(metadata_str)
site_field = instance.get_metadata_field('extractor_key')
thumbnail_field = instance.get_metadata_field('thumbnail')
arg_dict.update({
site_field: instance.get_metadata_first_value(
'extractor_key',
'Youtube',
arg_dict=existing_metadata,
),
thumbnail_field: instance.thumbnail,
})
instance.metadata = instance.metadata_dumps(arg_dict=arg_dict)
# Do not create more tasks before deleting
instance.manual_skip = True
instance.save()
@receiver(post_delete, sender=Media)
def media_post_delete(sender, instance, **kwargs):
# Remove the video file, when configured to do so
remove_files = (
instance.source and
instance.source.delete_files_on_disk and
instance.downloaded and
instance.media_file
)
if remove_files:
video_path = Path(str(instance.media_file.path)).resolve(strict=False)
instance.media_file.delete(save=False)
# the other files we created have these known suffixes
for suffix in frozenset(('nfo', 'jpg', 'webp', 'info.json',)):
other_path = video_path.with_suffix(f'.{suffix}').resolve(strict=False)
if other_path.is_file():
log.info(f'Deleting file for: {instance} path: {other_path!s}')
delete_file(other_path)
# subtitles include language code
subtitle_files = video_path.parent.glob(f'{glob_quote(video_path.with_suffix("").name)}*.vtt')
for file in subtitle_files:
log.info(f'Deleting file for: {instance} path: {file}')
delete_file(file)
# Jellyfin creates .trickplay directories and posters
for suffix in frozenset(('.trickplay', '-poster.jpg', '-poster.webp',)):
# with_suffix insists on suffix beginning with '.' for no good reason
other_path = Path(str(video_path.with_suffix('')) + suffix).resolve(strict=False)
if other_path.is_file():
log.info(f'Deleting file for: {instance} path: {other_path!s}')
delete_file(other_path)
elif other_path.is_dir():
# Delete the contents of the directory
paths = list(other_path.rglob('*'))
attempts = len(paths)
while paths and attempts > 0:
attempts -= 1
# delete files first
for p in list(filter(lambda x: x.is_file(), paths)):
log.info(f'Deleting file for: {instance} path: {p!s}')
delete_file(p)
# refresh the list
paths = list(other_path.rglob('*'))
# delete directories
# a directory with a subdirectory will fail
# we loop to try removing each of them
# a/b/c: c then b then a, 3 times around the loop
for p in list(filter(lambda x: x.is_dir(), paths)):
try:
p.rmdir()
log.info(f'Deleted directory for: {instance} path: {p!s}')
except OSError:
pass
# Delete the directory itself
try:
other_path.rmdir()
log.info(f'Deleted directory for: {instance} path: {other_path!s}')
except OSError:
pass
# Get all files that start with the bare file path
all_related_files = video_path.parent.glob(f'{glob_quote(video_path.with_suffix("").name)}*')
for file in all_related_files:
log.info(f'Deleting file for: {instance} path: {file}')
delete_file(file)
# Create a media entry for the indexing task to find
# Requirements:
# source, key, duration, title, published
created = False
create_for_indexing_task = (
not (
#not instance.downloaded and
instance.skip and
instance.manual_skip
)
)
if create_for_indexing_task:
skipped_media, created = Media.objects.get_or_create(
key=instance.key,
source=instance.source,
)
if created:
old_metadata = instance.loaded_metadata
site_field = instance.get_metadata_field('extractor_key')
thumbnail_url = instance.thumbnail
thumbnail_field = instance.get_metadata_field('thumbnail')
skipped_media.downloaded = False
skipped_media.duration = instance.duration
arg_dict=dict(
_media_instance_was_deleted=True,
)
arg_dict.update({
site_field: old_metadata.get(site_field),
thumbnail_field: thumbnail_url,
})
skipped_media.metadata = skipped_media.metadata_dumps(
arg_dict=arg_dict,
)
skipped_media.published = instance.published
skipped_media.title = instance.title
skipped_media.skip = True
skipped_media.manual_skip = True
skipped_media.save()
# Re-use the old metadata if it exists
instance_qs = Metadata.objects.filter(
media__isnull=True,
source__isnull=True,
site=old_metadata.get(site_field) or 'Youtube',
key=skipped_media.key,
)
try:
if instance_qs.count():
with atomic(durable=False):
# clear the link to a media instance
Metadata.objects.filter(media=skipped_media).update(media=None)
# choose the oldest metadata for our key
md = instance_qs.filter(
key=skipped_media.key,
).order_by(
'key',
'created',
).first()
# set the link to a media instance only on our selected metadata
log.info(f'Reusing old metadata for "{skipped_media.key}": {skipped_media.name}')
instance_qs.filter(uuid=md.uuid).update(media=skipped_media)
# delete any metadata that we are no longer using
instance_qs.exclude(uuid=md.uuid).delete()
except IntegrityError:
# this probably won't happen, but try it without a transaction
try:
# clear the link to a media instance
Metadata.objects.filter(media=skipped_media).update(media=None)
# keep one metadata
md = instance_qs.order_by('created').first()
instance_qs.filter(uuid=md.uuid).update(media=skipped_media)
except IntegrityError as e:
log.exception(f'media_post_delete: could not update selected metadata: {e}')
finally:
log.debug(f'Deleting metadata for "{skipped_media.key}": {skipped_media.pk}')
# delete the old metadata
instance_qs.delete()