Peach: Local-first video metadata scraper and media manager.
"""驳回某位女优名下的一家事务所:记进她的元数据,撤掉这条归属,这家没人了就连实体一起删。
minnano-av 的「所属事務所」只记现在签在谁名下。退役后转进个人经纪公司的人,那一格写的
是她自己的公司(三上悠亜 → 株式会社Miss),照它建出来的是一家只有一个人、却不是 AV 事务所
的「事务所」。只删实体不够:`repair_link_labels.py`、`resync_performer_agency.py` 会把站上
那一格重新写进 `metadata.agency`,`install_agencies.py` 再照它把实体建回来。所以驳回记在
人身上(`peach.entities.AGENCY_REJECTED`),这几条路径读到它就跳过。
输入是复核 CSV,一行一对:`performer,agency,reason[,replacement]`,名字按 canonical_name 给。
每一行:
1. 她的 `metadata.agency_rejected` 追加 `{name, reason, source, decided_at}`,同名不重复记;
2. `metadata.agency` 写的正是这一家就移除——那是证据,但它已经被人判定为不成立;
3. 她的归属指向这一家就删掉;
4. 这家再没有成员就删实体,别名、链接、外部编号跟着外键级联删;
5. 给了 `replacement`(她 AV 时期真正的事务所,写法同站上那一格,如 `ONE'S DOUBLE(ワンズダブル)`)
就改挂过去:按 `split_name` 拆出现用名与别名,缺实体就建,`metadata.agency` 也写成它,
这样 `install_agencies.py` 重跑得出的是同一个结论。
默认 dry-run。`--apply` 必须同时给 `--backup`:这是真实账本写入。
"""
from __future__ import annotations
import argparse
import json
import sys
import time
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
from peach.entities import ( # noqa: E402
AGENCY_REJECTED, agency_key, normalize_entity_name, rejected_agencies, split_name,
)
from peach.review_csv import read_rows # noqa: E402
from peach.scripting import ( # noqa: E402
add_ledger_write_args, counts_of, open_for_write, verify_after_write,
)
from peach.social_links import classify, host_owners # noqa: E402
#: 驳回是人的判断,写入者按 `field_owners` 的归属串署名。
SOURCE = "user:manual"
EXTRA_COUNTS = {
"agency": "SELECT count(*) FROM entity WHERE kind='agency'",
"entity_membership": "SELECT count(*) FROM entity_membership",
"entity_link": "SELECT count(*) FROM entity_link",
"entity_external_ref": "SELECT count(*) FROM entity_external_ref",
}
def plan(connection, rows: list[dict[str, str]]) -> list[dict[str, object]]:
"""每一行对回账本:她是谁、这家是哪个实体、她现在归哪家、这家还剩几个人。"""
found: list[dict[str, object]] = []
for raw in rows:
row = {key: (value or "").strip() for key, value in raw.items()}
person = connection.execute(
"SELECT id, metadata_json FROM entity WHERE kind='performer' AND canonical_name=?",
(row["performer"],)).fetchone()
wanted = agency_key(row["agency"])
agency = next((item for item in connection.execute(
"SELECT id, canonical_name FROM entity WHERE kind='agency'")
if agency_key(item["canonical_name"]) == wanted), None)
item = {**row, "entity_id": person["id"] if person else None,
"agency_id": agency["id"] if agency else None, "note": ""}
if not person or not wanted:
item["note"] = "未取得:账本里没有这位 performer" if not person else "未取得:事务所名为空"
found.append(item)
continue
held = connection.execute(
"SELECT agency_id FROM entity_membership WHERE member_id=?", (person["id"],)).fetchone()
item["holds"] = bool(agency and held and held["agency_id"] == agency["id"])
others = connection.execute(
"SELECT count(*) FROM entity_membership WHERE agency_id=? AND member_id<>?",
(item["agency_id"], person["id"])).fetchone()[0] if agency else 0
item["drop_entity"] = bool(agency) and others == 0
item["note"] = (f"撤归属{',删实体' if item['drop_entity'] else f',这家还有 {others} 人'}"
if item["holds"] else "她不在这家名下,只记驳回")
if row.get("replacement"):
item["note"] += f",改挂 {row['replacement']}"
found.append(item)
return found
def attach(connection, entity_id: int, raw: str, stamp: str) -> int:
"""把她挂到 AV 时期的事务所:缺实体就建,别名按 `split_name` 补上。返回新建的实体数。"""
canonical, aliases = split_name(raw)
key = normalize_entity_name(canonical)
connection.execute(
"INSERT OR IGNORE INTO entity(kind,canonical_name,normalized_name,"
"metadata_json,created_at,updated_at) VALUES('agency',?,?,'{}',?,?)",
(canonical, key, stamp, stamp))
created = connection.execute("SELECT changes()").fetchone()[0]
agency_id = connection.execute(
"SELECT id FROM entity WHERE kind='agency' AND normalized_name=?", (key,)).fetchone()[0]
for alias in aliases:
connection.execute(
"INSERT OR IGNORE INTO entity_alias"
"(entity_id,alias,normalized_alias,source,confidence) VALUES(?,?,?,?,1.0)",
(agency_id, alias, normalize_entity_name(alias), SOURCE))
connection.execute(
"INSERT OR REPLACE INTO entity_membership"
"(member_id,agency_id,source,confidence,checked_at) VALUES(?,?,?,1.0,?)",
(entity_id, agency_id, SOURCE, stamp))
return created
def relabel_links(connection, entity_id: int, rejected: str, replacement: str, stamp: str) -> int:
"""她名下标签写着这家事务所的官方链接,按域名归属重写标签。返回改了几条。
采集时 minnano-av 的「所属事務所」被贴在同一页每条 official 链接上:三上悠亜那条
ONE'S DOUBLE 的个人页写着「株式会社Miss」。事务所驳回了,标签还替它说话。重写走
`classify`,与 `repair_link_labels.py` 同一套:域名归谁写谁,不知道就写她 AV 时期的
事务所,再不知道写「官方网站」。
"""
owners = host_owners(connection)
agency = split_name(replacement)[0] if replacement else ""
changed = 0
for link_id, label, url in connection.execute(
"SELECT id,label,url FROM entity_link WHERE entity_id=? AND link_kind='official'",
(entity_id,)).fetchall():
if agency_key(label) != agency_key(rejected):
continue
connection.execute("UPDATE entity_link SET label=?,updated_at=? WHERE id=?",
(classify(url, agency, owners.get)[1], stamp, link_id))
changed += 1
return changed
def apply_rows(connection, rows: list[dict[str, object]], stamp: str) -> dict[str, int]:
done = {"驳回": 0, "归属": 0, "事务所": 0, "改挂": 0, "新建事务所": 0, "链接标签": 0}
# 外键默认关着,删实体时要它把别名、链接、外部编号一并级联删掉;PRAGMA 只能在事务外设。
connection.commit()
connection.execute("PRAGMA foreign_keys=ON")
for row in rows:
if row["entity_id"] is None or not row["agency"]:
continue
entity_id = int(row["entity_id"])
current = connection.execute(
"SELECT metadata_json FROM entity WHERE id=?", (entity_id,)).fetchone()
metadata = json.loads(current[0] or "{}")
key = agency_key(str(row["agency"]))
if key not in rejected_agencies(metadata):
metadata.setdefault(AGENCY_REJECTED, []).append(
{"name": row["agency"], "reason": row.get("reason", ""),
"source": SOURCE, "decided_at": stamp})
done["驳回"] += 1
if agency_key(str((metadata.get("agency") or {}).get("name") or "")) == key:
metadata.pop("agency")
replacement = str(row.get("replacement") or "")
if replacement:
metadata["agency"] = {"name": replacement, "source": f"{SOURCE}:{row.get('reason', '')}",
"checked_at": stamp}
connection.execute("UPDATE entity SET metadata_json=?,updated_at=? WHERE id=?",
(json.dumps(metadata, ensure_ascii=False), stamp, entity_id))
done["链接标签"] += relabel_links(connection, entity_id, str(row["agency"]), replacement, stamp)
if row.get("holds"):
connection.execute("DELETE FROM entity_membership WHERE member_id=? AND agency_id=?",
(entity_id, int(row["agency_id"])))
done["归属"] += connection.execute("SELECT changes()").fetchone()[0]
if row.get("drop_entity"):
connection.execute(
"DELETE FROM entity WHERE id=? AND kind='agency' AND NOT EXISTS"
" (SELECT 1 FROM entity_membership WHERE agency_id=?)",
(int(row["agency_id"]), int(row["agency_id"])))
done["事务所"] += connection.execute("SELECT changes()").fetchone()[0]
if replacement:
done["新建事务所"] += attach(connection, entity_id, replacement, stamp)
done["改挂"] += 1
return done
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__)
add_ledger_write_args(parser)
parser.add_argument("--input", type=Path, required=True,
help="复核 CSV:performer,agency,reason[,replacement]")
args = parser.parse_args()
connection = open_for_write(args)
try:
rows = plan(connection, read_rows(args.input))
for row in rows:
print(f"{row['performer']} × {row['agency']}:{row['note']}")
if not args.apply:
print("dry-run:没有写入。加 --apply --backup <路径> 才真的写。")
return 0
before = counts_of(connection, EXTRA_COUNTS)
done = apply_rows(connection, rows, time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()))
connection.commit()
after = counts_of(connection, EXTRA_COUNTS)
integrity, violations = verify_after_write(connection)
print("写入:" + ",".join(f"{key} {value}" for key, value in done.items()))
for key in sorted(before):
print(f" {key}: {before[key]} -> {after[key]}")
print(f"integrity_check={integrity} foreign_key_check={violations}")
return 0
finally:
connection.close()
if __name__ == "__main__":
raise SystemExit(main())