feat(up): capture follower snapshots at ingest and backfill history

This commit is contained in:
kairee committed 2026-11-01 11:18:06 +08:00
1 parent 4942d1d161
commit 519fc39fca
8 files changed
+393 -6

No files matched your search

+36
View File
@@ -39,6 +39,7 @@ from .services.favorites import FavoriteActionService
from .services.folder_sync import FolderSyncService from .services.folder_sync import FolderSyncService
from .services.snapshot import SnapshotService from .services.snapshot import SnapshotService
from .services.telegram_poller import TelegramPoller from .services.telegram_poller import TelegramPoller
from .services.up_profile import UpProfileService
from .timeutil import next_daily_run, parse_hhmm, shanghai_date from .timeutil import next_daily_run, parse_hhmm, shanghai_date
ROOT = Path(__file__).resolve().parent.parent ROOT = Path(__file__).resolve().parent.parent
@@ -120,6 +121,12 @@ def _build_parser() -> argparse.ArgumentParser:
"--preview", action="store_true", "--preview", action="store_true",
help="以预览模式启动(要求数据库含 preview:seeded=1 标记)", help="以预览模式启动(要求数据库含 preview:seeded=1 标记)",
) )
backfill = sub.add_parser(
"backfill-followers", parents=[common],
help="为历史视频回填 UP 粉丝数快照(只补空值)",
)
backfill.add_argument("--limit", type=int, default=None, help="最多处理多少个 UP")
backfill.add_argument("--dry-run", action="store_true", help="只统计待回填数量,不发请求")
sub.add_parser("status", parents=[common], help="查看数据库状态(离线)") sub.add_parser("status", parents=[common], help="查看数据库状态(离线)")
return parser return parser
@@ -226,6 +233,33 @@ def cmd_sync(args, logger: JsonLogger) -> int:
db.close() db.close()
def cmd_backfill_followers(args, logger: JsonLogger) -> int:
store = ConfigStore(args.config, require_auth=False)
cfg = store.get()
db = _open_db(cfg, logger)
try:
repo = Repository(db)
client = _client(
cfg.bilibili.read, logger, pacer=Pacer(1.0, 1.6), retries=2
)
service = UpProfileService(repo, lambda: client, logger=logger)
started = int(time.time())
result = service.backfill(limit=args.limit, dry_run=args.dry_run)
payload = result.as_dict()
payload["dry_run"] = bool(args.dry_run)
payload["coverage"] = repo.up_follower_coverage()
payload["database"] = str(cfg.storage.database_path)
_record_task(
repo, "backfill_followers", ok=result.failed == 0,
started_at=started, finished_at=int(time.time()), payload=payload,
error=";".join(result.errors) if result.errors else None,
)
_print(payload)
return 0 if result.failed == 0 else 1
finally:
db.close()
def cmd_snapshot(args, logger: JsonLogger) -> int: def cmd_snapshot(args, logger: JsonLogger) -> int:
store = ConfigStore(args.config, require_auth=False) store = ConfigStore(args.config, require_auth=False)
cfg = store.get() cfg = store.get()
@@ -673,6 +707,8 @@ def main(argv: Sequence[str] | None = None) -> int:
return cmd_sync(args, logger) return cmd_sync(args, logger)
if args.command == "snapshot": if args.command == "snapshot":
return cmd_snapshot(args, logger) return cmd_snapshot(args, logger)
if args.command == "backfill-followers":
return cmd_backfill_followers(args, logger)
if args.command == "discover": if args.command == "discover":
return cmd_discover(args, logger) return cmd_discover(args, logger)
if args.command == "candidates": if args.command == "candidates":
+19 -2
View File
@@ -410,12 +410,14 @@ class Repository:
following: int | None = None, following: int | None = None,
now: int | None = None, now: int | None = None,
backfilled: bool = False, backfilled: bool = False,
live_since: int | None = None,
) -> int: ) -> int:
"""写入 UP 粉丝快照,并补齐该 UP 尚未记录粉丝数的视频。 """写入 UP 粉丝快照,并补齐该 UP 尚未记录粉丝数的视频。
``follower_first`` 只在首次插入时写入;``follower_latest`` 每次刷新都会 ``follower_first`` 只在首次插入时写入;``follower_latest`` 每次刷新都会
更新。视频上的 ``up_follower`` 是入库时快照,已有值不会被覆盖。返回本次 更新。视频上的 ``up_follower`` 是入库时快照,已有值不会被覆盖。传入
补齐的视频数。 ``live_since`` 时,只有 ``first_seen >= live_since`` 的视频视为实时采集,
更早的视频标记为回填。返回本次补齐的视频数。
""" """
now = self._now() if now is None else int(now) now = self._now() if now is None else int(now)
mid = int(mid) mid = int(mid)
@@ -436,14 +438,29 @@ class Repository:
""", """,
(mid, name, follower, following_value, now, follower, now), (mid, name, follower, following_value, now, follower, now),
) )
if live_since is None:
cursor = conn.execute( cursor = conn.execute(
""" """
UPDATE video UPDATE video
SET up_follower = ?, up_follower_at = ?, up_follower_backfilled = ? SET up_follower = ?, up_follower_at = ?, up_follower_backfilled = ?
WHERE up_mid = ? AND up_follower IS NULL WHERE up_mid = ? AND up_follower IS NULL
AND status IN ('active', 'unfaved')
""", """,
(follower, now, backfill_flag, mid), (follower, now, backfill_flag, mid),
) )
else:
cursor = conn.execute(
"""
UPDATE video
SET up_follower = ?, up_follower_at = ?,
up_follower_backfilled = CASE
WHEN first_seen >= ? THEN 0 ELSE 1
END
WHERE up_mid = ? AND up_follower IS NULL
AND status IN ('active', 'unfaved')
""",
(follower, now, int(live_since), mid),
)
return max(cursor.rowcount, 0) return max(cursor.rowcount, 0)
return self.db.write(run) return self.db.write(run)
+32 -1
View File
@@ -21,6 +21,7 @@ from typing import Any
from ..config import AppConfig, ConfigStore, FolderConfig from ..config import AppConfig, ConfigStore, FolderConfig
from ..repository import Repository from ..repository import Repository
from .up_profile import UpProfileService
CAPACITY_WARN = 950 CAPACITY_WARN = 950
@@ -61,6 +62,8 @@ class SyncSummary:
finished_at: int = 0 finished_at: int = 0
folders: list[FolderSyncResult] = field(default_factory=list) folders: list[FolderSyncResult] = field(default_factory=list)
blacklist_seeded: int = 0 blacklist_seeded: int = 0
followers_updated: int = 0
followers_failed: int = 0
@property @property
def added(self) -> int: def added(self) -> int:
@@ -100,6 +103,8 @@ class SyncSummary:
"refreshed": self.refreshed, "refreshed": self.refreshed,
"missing": self.missing, "missing": self.missing,
"blacklist_seeded": self.blacklist_seeded, "blacklist_seeded": self.blacklist_seeded,
"followers_updated": self.followers_updated,
"followers_failed": self.followers_failed,
"folders": [f.as_dict() for f in self.folders], "folders": [f.as_dict() for f in self.folders],
"errors": self.errors, "errors": self.errors,
"warnings": self.warnings, "warnings": self.warnings,
@@ -119,6 +124,8 @@ class FolderSyncService:
logger=None, logger=None,
capacity_warn: int = CAPACITY_WARN, capacity_warn: int = CAPACITY_WARN,
allow_empty_folders: bool = False, allow_empty_folders: bool = False,
up_profile_service: UpProfileService | None = None,
capture_followers: bool = True,
) -> None: ) -> None:
if client is None and client_provider is None: if client is None and client_provider is None:
raise ValueError("必须提供 client 或 client_provider") raise ValueError("必须提供 client 或 client_provider")
@@ -130,6 +137,10 @@ class FolderSyncService:
self.logger = logger self.logger = logger
self.capacity_warn = capacity_warn self.capacity_warn = capacity_warn
self.allow_empty_folders = allow_empty_folders self.allow_empty_folders = allow_empty_folders
self.capture_followers = capture_followers
self.up_profile = up_profile_service or UpProfileService(
repository, self._client, clock=clock, logger=logger
)
def _client(self): def _client(self):
if self.client_provider is not None: if self.client_provider is not None:
@@ -189,6 +200,26 @@ class FolderSyncService:
) )
summary.folders.append(result) summary.folders.append(result)
if self.capture_followers:
try:
missing = self.repo.up_mids_missing_follower()
if missing:
capture = self.up_profile.capture(
[int(row["up_mid"]) for row in missing],
names={
int(row["up_mid"]): str(row.get("up_name") or "")
for row in missing
},
live_since=summary.started_at - 3600,
)
summary.followers_updated = capture.updated_videos
summary.followers_failed = capture.failed
except Exception as exc: # noqa: BLE001 粉丝采集失败不影响收藏夹同步
summary.followers_failed += 1
self._log(
"warning", "folder_sync.followers_failed", error=str(exc)
)
summary.finished_at = self._now() summary.finished_at = self._now()
self._log( self._log(
"info" if summary.ok else "error", "info" if summary.ok else "error",
@@ -200,7 +231,7 @@ class FolderSyncService:
errors=len(summary.errors), errors=len(summary.errors),
elapsed=max(0, summary.finished_at - summary.started_at), elapsed=max(0, summary.finished_at - summary.started_at),
) )
if summary.added or summary.removed or summary.refreshed: if summary.added or summary.removed or summary.refreshed or summary.followers_updated:
self.repo.bump_cache_version("videos") self.repo.bump_cache_version("videos")
return summary return summary
+132
View File
@@ -0,0 +1,132 @@
# -*- coding: utf-8 -*-
"""UP 粉丝快照采集。
视频的播放、互动记在 bvid 上,而 UP 的粉丝数是视频之外的受众基数。为了让以后
能计算“播放 / 入库时粉丝数”,本服务在视频入库时记录一次粉丝快照,并支持对历史
数据做一次性回填。
``up_profile.follower_first`` 永远保留首次采集值;``video.up_follower`` 保留每个
视频入库时的值,不会被后续刷新覆盖。
"""
from __future__ import annotations
import time
from dataclasses import dataclass, field
from typing import Any, Callable, Iterable, Mapping
from ..repository import Repository
@dataclass
class CaptureResult:
requested: int = 0
updated_videos: int = 0
failed: int = 0
errors: list[str] = field(default_factory=list)
def as_dict(self) -> dict[str, Any]:
return {
"requested": self.requested,
"updated_videos": self.updated_videos,
"failed": self.failed,
"errors": self.errors,
}
class UpProfileService:
def __init__(
self,
repository: Repository,
client_provider: Callable[[], Any],
*,
clock=time.time,
logger=None,
) -> None:
self.repo = repository
self.client_provider = client_provider
self._clock = clock
self.logger = logger
def _now(self) -> int:
return int(self._clock())
def _log(self, level: str, event: str, **fields) -> None:
if self.logger is not None:
getattr(self.logger, level)(event, **fields)
def capture(
self,
mids: Iterable[int],
*,
names: Mapping[int, str] | None = None,
live_since: int | None = None,
backfilled: bool = False,
) -> CaptureResult:
"""采集一组 UP 的当前粉丝数,并补齐其视频的空快照。
同一个 mid 在一次调用里只请求一次;单个失败不会中断其余 UP。
"""
ordered: list[int] = []
seen: set[int] = set()
for mid in mids:
if mid is None or isinstance(mid, bool):
continue
try:
value = int(mid)
except (TypeError, ValueError):
continue
if value <= 0 or value in seen:
continue
seen.add(value)
ordered.append(value)
result = CaptureResult(requested=len(ordered))
if not ordered:
return result
now = self._now()
names = names or {}
for mid in ordered:
try:
fan = self.client_provider().relation_stat(mid)
updated = self.repo.apply_up_follower(
mid=mid,
name=names.get(mid) or getattr(fan, "uname", None),
follower=int(fan.follower),
following=int(getattr(fan, "following", 0) or 0),
now=now,
backfilled=backfilled,
live_since=live_since,
)
except Exception as exc: # noqa: BLE001 单个 UP 失败不阻塞同步
result.failed += 1
result.errors.append(f"{mid}: {exc}")
self._log(
"warning", "up_profile.capture_failed", mid=mid, error=str(exc)
)
else:
result.updated_videos += updated
self._log(
"info", "up_profile.captured",
mid=mid, follower=int(fan.follower), updated=updated,
)
return result
def backfill(
self, *, limit: int | None = None, dry_run: bool = False
) -> CaptureResult:
"""为历史视频回填粉丝快照;已有值的视频不会被覆盖。"""
rows = self.repo.up_mids_missing_follower()
if limit is not None:
rows = rows[: max(0, int(limit))]
if dry_run:
return CaptureResult(requested=len(rows))
names = {
int(row["up_mid"]): str(row.get("up_name") or "")
for row in rows
}
return self.capture(
[int(row["up_mid"]) for row in rows],
names=names,
backfilled=True,
)
+16 -2
View File
@@ -51,7 +51,7 @@ def test_init_and_status():
path = write_config(tmp) path = write_config(tmp)
code, output = run_cli(["init", "--config", str(path)]) code, output = run_cli(["init", "--config", str(path)])
assert code == 0 assert code == 0
assert '"schema_version": 3' in output assert '"schema_version": 4' in output
db = Database(Path(tmp) / "app.db") db = Database(Path(tmp) / "app.db")
db.initialize() db.initialize()
@@ -319,6 +319,19 @@ def test_candidates_and_discover_commands():
assert '"notified": 1' in output assert '"notified": 1' in output
def test_backfill_followers_dry_run():
with tempfile.TemporaryDirectory() as tmp:
path = write_config(tmp, cookie="", csrf="")
code, output = run_cli([
"backfill-followers", "--dry-run", "--config", str(path),
])
assert code == 0
assert '"requested": 0' in output
assert '"failed": 0' in output
assert '"dry_run": true' in output
assert '"total": 0' in output
def main_test(): def main_test():
test_init_and_status() test_init_and_status()
test_missing_config() test_missing_config()
@@ -331,7 +344,8 @@ def main_test():
test_default_command_starts_web_and_scheduler() test_default_command_starts_web_and_scheduler()
test_snapshot_backup_failure_returns_error_and_retries() test_snapshot_backup_failure_returns_error_and_retries()
test_candidates_and_discover_commands() test_candidates_and_discover_commands()
print("OK: CLI init/sync/snapshot/auth-check/run/status/错误路径测试全部通过") test_backfill_followers_dry_run()
print("OK: CLI init/sync/snapshot/auth-check/run/status/backfill/错误路径测试全部通过")
if __name__ == "__main__": if __name__ == "__main__":
+17
View File
@@ -45,9 +45,17 @@ class FakeFavTransport:
self.omit_ids_data = omit_ids_data self.omit_ids_data = omit_ids_data
self.ids_calls = [] self.ids_calls = []
self.info_calls = [] self.info_calls = []
self.relation_calls = []
def request(self, method, url, headers, data=None, timeout=30): def request(self, method, url, headers, data=None, timeout=30):
query = urllib.parse.parse_qs(urllib.parse.urlparse(url).query) query = urllib.parse.parse_qs(urllib.parse.urlparse(url).query)
if "/x/relation/stat" in url:
mid = int(query["vmid"][0])
self.relation_calls.append(mid)
return http_json({
"code": 0, "message": "0", "ttl": 1,
"data": {"mid": mid, "following": 5, "follower": 1000 + mid},
})
if "/x/v3/fav/resource/ids" in url: if "/x/v3/fav/resource/ids" in url:
media_id = int(query["media_id"][0]) media_id = int(query["media_id"][0])
self.ids_calls.append(media_id) self.ids_calls.append(media_id)
@@ -165,6 +173,13 @@ def test_initial_import_and_incremental_skip():
assert repo.get_video(1104)["bvid"] == "BV1104" assert repo.get_video(1104)["bvid"] == "BV1104"
assert repo.folder_active_aids(F1) == {1101, 1102, 1103, 1104} assert repo.folder_active_aids(F1) == {1101, 1102, 1103, 1104}
assert repo.is_blacklisted(999) assert repo.is_blacklisted(999)
# 新入库视频记录入库时粉丝数;失效视频不再采集
assert summary.followers_updated == 2
assert summary.followers_failed == 0
assert repo.get_video(1101)["up_follower"] == 1100
assert repo.get_video(1101)["up_follower_backfilled"] == 0
assert repo.get_video(1102)["up_follower"] == 1999
assert sorted(transport.relation_calls) == [100, 999]
folder = repo.get_folder(F1) folder = repo.get_folder(F1)
assert folder.media_count == 4 assert folder.media_count == 4
@@ -173,10 +188,12 @@ def test_initial_import_and_incremental_skip():
# 6 小时内再次运行:没有新增,不请求 infos # 6 小时内再次运行:没有新增,不请求 infos
clock[0] += 60 clock[0] += 60
info_before = len(transport.info_calls) info_before = len(transport.info_calls)
relation_before = len(transport.relation_calls)
summary2 = service.run() summary2 = service.run()
assert summary2.added == 0 assert summary2.added == 0
assert summary2.refreshed == 0 assert summary2.refreshed == 0
assert len(transport.info_calls) == info_before assert len(transport.info_calls) == info_before
assert len(transport.relation_calls) == relation_before
assert summary2.folders[0].full_refresh is False assert summary2.folders[0].full_refresh is False
finally: finally:
db.close() db.close()
+9 -1
View File
@@ -339,9 +339,17 @@ def test_up_follower_snapshot_and_backfill():
assert repo.get_video(3)["up_follower"] == 88 assert repo.get_video(3)["up_follower"] == 88
assert repo.get_video(3)["up_follower_backfilled"] == 1 assert repo.get_video(3)["up_follower_backfilled"] == 1
# live_since 之前的视频标为回填,之后的视为实时采集
repo.upsert_from_fav(resource(5, up_mid=300), F1.media_id, was_added=True)
repo.apply_up_follower(
mid=300, name="UP3", follower=50, now=now + 180,
live_since=now + 1,
)
assert repo.get_video(5)["up_follower_backfilled"] == 1
coverage = repo.up_follower_coverage() coverage = repo.up_follower_coverage()
assert coverage == { assert coverage == {
"total": 4, "covered": 4, "missing": 0, "ups_covered": 2, "total": 5, "covered": 5, "missing": 0, "ups_covered": 3,
} }
finally: finally:
db.close() db.close()
+132
View File
@@ -0,0 +1,132 @@
# -*- coding: utf-8 -*-
"""UP 粉丝快照采集测试(不联网): uv run python tests/test_up_profile.py"""
import os
import sys
import tempfile
from pathlib import Path
from types import SimpleNamespace
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
sys.stdout.reconfigure(encoding="utf-8")
from reaction_board.bilibili.models import FavResource # noqa: E402
from reaction_board.config import FolderConfig # noqa: E402
from reaction_board.database import Database # noqa: E402
from reaction_board.repository import Repository # noqa: E402
from reaction_board.services.up_profile import UpProfileService # noqa: E402
FOLDER = FolderConfig(year=2026, media_id=3700023351, title="2026")
NOW = 1_800_000_000
def resource(aid, up_mid):
return FavResource(
aid=aid,
bvid=f"BV{aid}",
title=f"视频{aid}",
up_mid=up_mid,
up_name=f"UP{up_mid}",
pubdate=NOW - 86400,
duration=7200,
cover=None,
attr=0,
fav_time=None,
pages=2,
)
class FakeClient:
def __init__(self, *, failing=()):
self.calls = []
self.failing = set(failing)
def relation_stat(self, mid):
self.calls.append(int(mid))
if int(mid) in self.failing:
raise RuntimeError("simulated network error")
return SimpleNamespace(mid=int(mid), follower=1000 + int(mid), following=5)
def setup(tmp):
db = Database(Path(tmp) / "app.db")
db.initialize()
repo = Repository(db, clock=lambda: NOW)
repo.ensure_folder(FOLDER)
for aid, mid in ((1, 100), (2, 100), (3, 200)):
repo.upsert_from_fav(resource(aid, mid), FOLDER.media_id, was_added=True)
return db, repo
def test_capture_dedupes_and_continues_after_failure():
with tempfile.TemporaryDirectory() as tmp:
db, repo = setup(tmp)
try:
client = FakeClient(failing={200})
service = UpProfileService(repo, lambda: client, clock=lambda: NOW)
result = service.capture(
[100, 100, 200, None, 0], live_since=NOW - 3600
)
assert result.requested == 2
assert client.calls == [100, 200]
assert result.updated_videos == 2
assert result.failed == 1
assert "200" in result.errors[0]
assert repo.get_video(1)["up_follower"] == 1100
assert repo.get_video(2)["up_follower"] == 1100
assert repo.get_video(3)["up_follower"] is None
assert repo.get_up_profile(100)["follower_first"] == 1100
assert repo.get_up_profile(200) is None
finally:
db.close()
def test_capture_marks_old_videos_as_backfilled():
with tempfile.TemporaryDirectory() as tmp:
db, repo = setup(tmp)
try:
client = FakeClient()
service = UpProfileService(repo, lambda: client, clock=lambda: NOW)
service.capture([100], live_since=NOW + 1)
assert repo.get_video(1)["up_follower_backfilled"] == 1
assert repo.get_video(2)["up_follower_backfilled"] == 1
finally:
db.close()
def test_backfill_dry_run_and_limit():
with tempfile.TemporaryDirectory() as tmp:
db, repo = setup(tmp)
try:
client = FakeClient()
service = UpProfileService(repo, lambda: client, clock=lambda: NOW)
dry = service.backfill(dry_run=True)
assert dry.requested == 2
assert client.calls == []
assert repo.up_follower_coverage()["missing"] == 3
limited = service.backfill(limit=1)
assert limited.requested == 1
assert client.calls == [100]
assert repo.get_video(1)["up_follower_backfilled"] == 1
assert repo.get_video(3)["up_follower"] is None
rest = service.backfill()
assert rest.requested == 1
assert client.calls == [100, 200]
assert repo.up_follower_coverage()["missing"] == 0
finally:
db.close()
def main():
test_capture_dedupes_and_continues_after_failure()
test_capture_marks_old_videos_as_backfilled()
test_backfill_dry_run_and_limit()
print("OK: UP 粉丝快照采集与回填测试全部通过")
if __name__ == "__main__":
main()