Peach: Local-first video metadata scraper and media manager.
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
r"""按片长确认「名字像番号、里面的文件却不带番号」的目录。
`audit_code_creators.py` 只清有文件级证据的那批:目录名与目录内文件解析出同一个番号。
剩下的存疑行里两种东西长得一模一样——`Tokyo-Hot n0780-HD\Tokyo-Hot.mp4` 是发行目录,
`banbi_555` 是上传者账号——名字本身答不了。
本脚本去来源要这个番号的片长,和目录里最长那条视频比:对得上就是这部片,按番号处理
(补 code、摘掉假创作者);对不上或来源查无此片,留在人工队列(用户 2026-09-16 定的
判据,容差见 `code_creators.DURATION_TOLERANCE_SECONDS`)。
默认只联网比对并产出 CSV,`--apply` 才写 ledger 且必须给 `--backup`。
"""
from __future__ import annotations
import argparse
import sqlite3
import sys
from pathlib import Path
PROJECT_ROOT = Path(__file__).resolve().parent.parent
SRC_DIR = PROJECT_ROOT / "src"
if str(SRC_DIR) not in sys.path:
sys.path.insert(0, str(SRC_DIR))
from peach.code_creators import ( # noqa: E402
VERDICT_CODE,
VERDICT_UNCLEAR,
apply_rows,
collect,
duration_confirms_code,
main_video_seconds,
)
from peach.config import DATABASE_PATH, GENERATED_DIR, SECRETS_DIR # noqa: E402
from peach.metadata_policy import PREFERRED_COMMUNITY_SOURCE # noqa: E402
from peach.migrations import sqlite_backup # noqa: E402
from peach.review_csv import write_rows # noqa: E402
FIELDS = ("entity_id", "creator", "identity", "assets", "local_seconds", "source",
"runtime_minutes", "confirmed", "reason")
def source_runtime(provider, code: str) -> tuple[str, float | None, str]:
"""这个番号的片长:先问 r18.dev,它没有再问社区来源。返回 (来源, 分钟, 说明)。
社区来源之间按 `PREFERRED_COMMUNITY_SOURCE` 排:`n0780` 的片长 javbus 报 36 分、
javdb 报 96 分,盘里那条 98 分——按返回顺序取第一家就会把这个目录判成对不上。
"""
try:
payload = provider.query(code)
if payload.get("runtime"):
return "r18dev", float(payload["runtime"]), ""
except Exception as error: # noqa: BLE001
note = f"r18dev:{type(error).__name__}"
else:
note = "r18dev 没给片长"
try:
found = sorted(provider.community(code),
key=lambda item: item[0] != PREFERRED_COMMUNITY_SOURCE)
except Exception as error: # noqa: BLE001
return "", None, f"{note};社区来源:{type(error).__name__}"
for name, payload in found:
if payload.get("runtime"):
return name, float(payload["runtime"]), note
return "", None, f"{note};社区来源没给片长"
def assets_of(connection: sqlite3.Connection, entity_id: int) -> list:
cursor = connection.cursor()
cursor.row_factory = sqlite3.Row
return cursor.execute(
"SELECT a.id,a.name,a.path,a.code,a.medium,a.duration FROM asset_entity ae "
"JOIN asset a ON a.id=ae.asset_id WHERE ae.entity_id=? AND ae.role='creator' "
"ORDER BY a.id", (entity_id,)).fetchall()
def examine(connection: sqlite3.Connection, provider) -> list[dict[str, object]]:
"""对每个存疑目录取一次片长,判它是不是这个番号。"""
rows = []
for row in collect(connection):
if row["verdict"] != VERDICT_UNCLEAR:
continue
entity_id = int(row["entity_id"])
local = main_video_seconds(assets_of(connection, entity_id))
source, runtime, note = source_runtime(provider, str(row["identity"]))
confirmed, reason = duration_confirms_code(local, runtime)
rows.append({
"entity_id": entity_id, "creator": row["creator"], "identity": row["identity"],
"assets": row["assets"], "local_seconds": "" if local is None else round(local),
"source": source, "runtime_minutes": "" if runtime is None else runtime,
"confirmed": "是" if confirmed else "否",
"reason": reason if runtime else (reason + ";" + note),
# 判定成立的按番号处理,交给 `code_creators.apply_rows` 走既有那条写入路径。
"verdict": VERDICT_CODE if confirmed else VERDICT_UNCLEAR,
})
return rows
def build_parser() -> argparse.ArgumentParser:
parser = argparse.ArgumentParser(description="按片长确认存疑的番号目录")
parser.add_argument("--db", type=Path, default=DATABASE_PATH)
parser.add_argument("--secrets", type=Path, default=SECRETS_DIR)
parser.add_argument("--review-csv", type=Path,
default=GENERATED_DIR / "code-dir-duration-review.csv")
parser.add_argument("--apply", action="store_true", help="写 ledger;默认只出 CSV")
parser.add_argument("--backup", type=Path, help="--apply 必需:写库前的 SQLite 备份路径")
return parser
def run(args: argparse.Namespace) -> int:
if args.apply and not args.backup:
raise SystemExit("--apply 必须同时给 --backup")
from peach.library_processing import LibraryMetadataProvider
connection = sqlite3.connect(args.db)
provider = LibraryMetadataProvider(args.secrets)
try:
rows = examine(connection, provider)
finally:
provider.close()
confirmed = [row for row in rows if row["verdict"] == VERDICT_CODE]
print(f"存疑目录 {len(rows)} 个,片长对得上的 {len(confirmed)} 个")
for row in rows:
print(f" {row['creator']:<34} {row['identity']:<14} {row['confirmed']} {row['reason']}")
if args.apply and confirmed:
sqlite_backup(args.db, args.backup)
print(f"已备份 → {args.backup}")
connection.execute("BEGIN IMMEDIATE")
try:
counts = apply_rows(connection, confirmed)
connection.commit()
except Exception:
connection.rollback()
raise
print(f"已写入:删创作者关系 {counts['links']} 条、删实体 {counts['entities']} 个、"
f"补 code {counts['codes']} 条、清扁平字段 {counts['flat']} 条")
connection.close()
# 判定用的 `verdict` 不进 CSV:那一列是给 `apply_rows` 的开关,复核看的是 `confirmed`。
write_rows(args.review_csv, FIELDS,
[{key: row[key] for key in FIELDS} for row in rows])
print(f"复核 CSV → {args.review_csv}")
if not args.apply:
print("这是预览:未写 ledger。确认后再加 --apply --backup。")
return 0
def main(argv: list[str] | None = None) -> int:
return run(build_parser().parse_args(argv))
if __name__ == "__main__":
raise SystemExit(main())