Files
Evection/apps/articles/tasks.py
T
ksmithandClaude Sonnet 5.5 72bffbe97f Expire purge tombstones and time-box the retention purge (R2-T2, R2-T4; #38, #40)
Tombstones were kept forever, so enabling retention left a permanent row
per deleted article. They now expire after 365 days (cleaned up by the
same nightly task, even when retention is off); the documented price is
that an entry a feed still lists after a year, with no or a recent
publish date, can reappear.

The purge looped until every eligible row was gone with no run budget.
A run now works for at most four minutes, then queues a continuation, so
a huge first purge can't monopolise a worker or starve polling; hard
time limits back that up, and each chunk is its own transaction so an
interrupted run just resumes. Adds indexes on Article.fetched_at and
PurgedArticle.purged_at for the scans these do.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-01 12:00:52 -07:00

162 lines
6.1 KiB
Python

import logging
import time
from datetime import timedelta
from celery import shared_task
from django.db import transaction
from django.utils import timezone
from apps.core.models import SiteSettings
from apps.feeds.models import Subscription
from apps.rules.engine import evaluate_rules_for_article
from apps.rules.models import FilterRule
from .models import Article, ArticleTag, PurgedArticle, SubscriptionTag
logger = logging.getLogger(__name__)
@shared_task
def apply_filter_rules_for_feed(feed_id: int, article_ids: list[int]):
"""For every subscriber of ``feed_id``, evaluate their enabled
FilterRules (scoped to this feed/tag or unscoped) against the given
articles. Dispatched by the fetch pipeline after new/updated articles
are committed for a feed.
Per-rule scoping is left to evaluate_rules_for_article/rule_in_scope
rather than filtered into this query, so this and the backlog
re-apply task (apps.rules.tasks) can't scope rules differently.
Rules and subscription tags are fetched once for every subscriber of
this feed, not once per subscriber (C9) — a popular feed with many
subscribers previously issued 2 queries per subscriber here.
"""
if not article_ids:
return 0
articles = list(Article.objects.filter(pk__in=article_ids).select_related("feed"))
if not articles:
return 0
subscriptions = list(Subscription.objects.filter(feed_id=feed_id).select_related("user"))
if not subscriptions:
return 0
rules_by_user_id: dict[int, list[FilterRule]] = {}
for rule in FilterRule.objects.filter(
user_id__in={s.user_id for s in subscriptions}, is_enabled=True
).order_by("priority", "id"):
rules_by_user_id.setdefault(rule.user_id, []).append(rule)
tag_ids_by_subscription_id: dict[int, set[int]] = {}
for subscription_id, tag_id in SubscriptionTag.objects.filter(
subscription_id__in=[s.pk for s in subscriptions]
).values_list("subscription_id", "tag_id"):
tag_ids_by_subscription_id.setdefault(subscription_id, set()).add(tag_id)
evaluated = 0
for subscription in subscriptions:
rules = rules_by_user_id.get(subscription.user_id)
if not rules:
continue
subscription_tag_ids = tag_ids_by_subscription_id.get(subscription.pk, set())
for article in articles:
evaluate_rules_for_article(subscription.user, article, rules, subscription_tag_ids)
evaluated += 1
return evaluated
PURGE_CHUNK_SIZE = 500
# One run works for at most this long, then queues itself again (R2-T4) so
# a huge first purge never holds a worker for hours or starves feed
# polling; the hard limits below are the backstop for a single stuck chunk.
PURGE_RUN_BUDGET_SECONDS = 240
PURGE_CONTINUATION_DELAY_SECONDS = 30
PURGE_SOFT_TIME_LIMIT = 540
PURGE_TIME_LIMIT = 600
# How long a purged article's tombstone is kept (R2-T2). It only needs to
# outlive the time a feed keeps *listing* the entry; after this, an entry a
# feed still publishes (and that has no publish date, or a recent one) can
# be ingested as new again — the price of the tombstone table not growing
# forever. Documented in docs/install.md.
TOMBSTONE_TTL_DAYS = 365
def _expire_tombstones(out_of_time) -> int:
cutoff = timezone.now() - timedelta(days=TOMBSTONE_TTL_DAYS)
removed = 0
while not out_of_time():
chunk = list(
PurgedArticle.objects.filter(purged_at__lt=cutoff).values_list("pk", flat=True)[
:PURGE_CHUNK_SIZE
]
)
if not chunk:
break
PurgedArticle.objects.filter(pk__in=chunk).delete()
removed += len(chunk)
return removed
@shared_task(time_limit=PURGE_TIME_LIMIT, soft_time_limit=PURGE_SOFT_TIME_LIMIT)
def purge_old_articles():
"""Deletes articles older than SiteSettings.retention_days (C7), in
chunks so one run never holds a huge delete in a single transaction,
and within a time budget — if it runs out it queues a continuation
(R2-T4). Anything a user starred or saved for later (a system
ArticleTag) is kept — those are explicit "keep this" signals.
UserArticleState and ArticleTag rows go with their article via ON
DELETE CASCADE, and the SQLite FTS triggers drop the search entries.
0 days means keep everything forever (the default).
Each deleted article leaves a PurgedArticle tombstone first, so a feed
that still lists the entry can't bring it back as a new, unread
article on its next poll (see apps.feeds.services.fetcher). Tombstones
older than TOMBSTONE_TTL_DAYS are removed here too, whether or not
retention is currently on (R2-T2). Returns the number of articles
deleted by this run.
"""
started = time.monotonic()
def out_of_time() -> bool:
return time.monotonic() - started > PURGE_RUN_BUDGET_SECONDS
expired = _expire_tombstones(out_of_time)
if expired:
logger.info("Removed %d expired purge tombstone(s)", expired)
retention_days = SiteSettings.load().retention_days
if retention_days <= 0:
return 0
cutoff = timezone.now() - timedelta(days=retention_days)
keep = ArticleTag.objects.filter(tag__is_system=True).values("article_id")
deleted_total = 0
while True:
chunk = list(
Article.objects.filter(fetched_at__lt=cutoff)
.exclude(pk__in=keep)
.values_list("pk", "feed_id", "guid")[:PURGE_CHUNK_SIZE]
)
if not chunk:
return deleted_total
with transaction.atomic():
PurgedArticle.objects.bulk_create(
[PurgedArticle(feed_id=feed_id, guid=guid) for _pk, feed_id, guid in chunk],
ignore_conflicts=True,
)
Article.objects.filter(pk__in=[pk for pk, _feed_id, _guid in chunk]).delete()
deleted_total += len(chunk)
if out_of_time():
logger.info(
"Retention purge hit its %ds budget after %d article(s); continuing shortly",
PURGE_RUN_BUDGET_SECONDS,
deleted_total,
)
purge_old_articles.apply_async(countdown=PURGE_CONTINUATION_DELAY_SECONDS)
return deleted_total