Peach: Local-first video metadata scraper and media manager.
"""要公开的文字发出去之前,对照本机账本找成串列出的馆藏细节。
PR、issue、评论与提交说明都公开在 GitHub 上,编辑后旧版仍留在编辑历史里。问题本身需要的
一两个番号、前缀或标签照常写;把账本里的作品番号、人物名、标题片段成串列出来就拦下。
判据只读本机账本(SQLite `mode=ro`),仓库里不带任何馆藏数据;没有账本的机器(CI、
其他贡献者)一律放行。确有必要多列时,在正文里写一行 `peach-public-text: <理由>` 放行。
用法:
python scripts/check_public_text.py FILE... # 检查文件,`-` 读 stdin
python scripts/check_public_text.py --hook-event # Claude PreToolUse 钩子:拦 gh 发布命令
"""
from __future__ import annotations
import argparse
import json
import re
import sqlite3
import sys
from dataclasses import dataclass, field
from pathlib import Path
ROOT = Path(__file__).resolve().parents[1]
CODE = re.compile(r"(?<![A-Za-z0-9])([A-Za-z]{2,7})-?(\d{2,6})(?![0-9])")
CJK = re.compile(r"[-ヿ㐀-鿿ヲ-゚]{2,}")
#: 标题片段按这么长的连续假名、汉字比对;短了会撞上普通用语。
GRAM = 7
#: 一段文字里达到这些数目才算「成串」;问题本身需要的一两个例子不拦。
LIMITS = {"番号": 3, "人物": 2, "标题片段": 2}
WAIVER = re.compile(r"peach-public-text:\s*\S")
#: 会把文字发到 GitHub 的 gh 子命令。
PUBLISH = re.compile(r"\bgh\s+(?:(?:pr|issue)\s+(?:create|edit|comment)\b|api\b[^\n]*\bcomments\b)")
BODY_FILE = re.compile(r"""(?:--body-file|-F)[=\s]+(?:body=@)?(?P<path>"[^"]+"|'[^']+'|[^\s"';|&]+)""")
def code_key(prefix: str, digits: str) -> str:
return f"{prefix.upper()}-{int(digits):03d}"
@dataclass
class Library:
"""账本里能认出馆藏的几样东西;只读载入,不写账本。"""
codes: set[str] = field(default_factory=set)
names: set[str] = field(default_factory=set)
grams: set[str] = field(default_factory=set)
@classmethod
def load(cls, ledger: Path) -> "Library":
library = cls()
connection = sqlite3.connect(f"file:{ledger.as_posix()}?mode=ro", uri=True)
try:
rows = connection.execute("SELECT code FROM asset WHERE code<>'' UNION SELECT code FROM feed_item WHERE code<>''")
library.codes = {code_key(*match.groups()) for (code,) in rows
for match in [CODE.fullmatch(code.strip())] if match}
names = connection.execute(
"SELECT canonical_name FROM entity WHERE kind IN ('performer','creator')"
" UNION SELECT label FROM follow_source WHERE label IS NOT NULL")
library.names = {name for (name,) in names if name and len(name) >= 3}
for row in connection.execute("SELECT catalog_title, original_title, name FROM asset"):
for text in row:
for run in CJK.findall(text or ""):
library.grams.update(run[i:i + GRAM] for i in range(len(run) - GRAM + 1))
finally:
connection.close()
return library
def findings(self, text: str) -> dict[str, list[str]]:
found: dict[str, set[str]] = {"番号": set(), "人物": set(), "标题片段": set()}
for match in CODE.finditer(text):
if code_key(*match.groups()) in self.codes:
found["番号"].add(match.group(0).upper())
for name in self.names:
if name.isascii():
if len(name) >= 5 and re.search(rf"(?<![\w-]){re.escape(name)}(?![\w-])", text):
found["人物"].add(name)
elif name in text:
found["人物"].add(name)
for run in CJK.findall(text):
if any(run[i:i + GRAM] in self.grams for i in range(len(run) - GRAM + 1)):
found["标题片段"].add(run)
return {kind: sorted(values) for kind, values in found.items() if values}
def problems(text: str, library: Library | None) -> list[str]:
"""达到「成串」的那几类,各一句;没有账本或写了放行理由时为空。"""
if library is None or WAIVER.search(text):
return []
found = library.findings(text)
return [f"{kind} {len(values)} 个:" + "、".join(values[:8]) + ("……" if len(values) > 8 else "")
for kind, values in found.items() if len(values) >= LIMITS[kind]]
def ledger_for(project: Path) -> Path | None:
"""这个检出所用 Peach 数据目录里的账本;找不到数据目录或账本时返回 None。"""
sys.path.insert(0, str(ROOT / "src"))
try:
from peach.settings_file import discover_data_root
root, found = discover_data_root(project)
except Exception: # noqa: BLE001 - 认不出数据目录就当这台机器没有账本
return None
ledger = root / "database" / "ledger.db"
return ledger if found and ledger.is_file() else None
def load_library(project: Path) -> Library | None:
ledger = ledger_for(project)
return Library.load(ledger) if ledger else None
def publish_text(command: str, cwd: Path) -> str | None:
"""gh 发布命令要发出去的全部文字:命令本身(含 --title、--body 与 heredoc)加上引用的正文文件。"""
if not PUBLISH.search(command):
return None
parts = [command]
for match in BODY_FILE.finditer(command):
raw = match.group("path").strip("\"'")
if raw == "-":
continue
path = Path(raw) if Path(raw).is_absolute() else cwd / raw
try:
parts.append(path.read_text(encoding="utf-8"))
except (OSError, UnicodeError):
continue
return "\n".join(parts)
def hook(payload: dict) -> int:
"""PreToolUse:gh 要发的文字成串列了馆藏就退出码 2,stderr 交给模型改。"""
command = str((payload.get("tool_input") or {}).get("command") or "")
if "gh" not in command:
return 0
cwd = Path(payload.get("cwd") or ROOT)
if not cwd.is_dir():
cwd = ROOT
text = publish_text(command, cwd)
if text is None:
return 0
found = problems(text, load_library(cwd))
if not found:
return 0
print("这次要发到 GitHub 的文字成串列出了本机账本里的馆藏细节(" + ";".join(found) + ")。"
"按 peach-issue-delivery:规则效果写类型和定性结论,只留问题本身需要的一两个例子;"
"确有必要多列时,正文加一行 `peach-public-text: <理由>`。", file=sys.stderr)
return 2
def main(argv: list[str] | None = None) -> int:
parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
parser.add_argument("files", nargs="*", help="要检查的文件,`-` 读 stdin")
parser.add_argument("--hook-event", action="store_true", help="从 stdin 收 PreToolUse 钩子 JSON")
args = parser.parse_args(argv)
if args.hook_event:
try:
payload = json.load(sys.stdin)
except (ValueError, OSError):
return 0
return hook(payload if isinstance(payload, dict) else {})
library = load_library(ROOT)
if library is None:
print("没有找到本机账本,跳过检查")
return 0
failed = False
for name in args.files or ["-"]:
text = sys.stdin.read() if name == "-" else Path(name).read_text(encoding="utf-8")
for problem in problems(text, library):
failed = True
print(f"{name}: {problem}")
return 1 if failed else 0
if __name__ == "__main__":
raise SystemExit(main())