Peach: Local-first video metadata scraper and media manager.
"""把本机已取得的厂牌标识收进 `resources/marks/`,随仓库分发(ADR-0026)。
只搬运本机 `peach-data/generated/logos/` 里已经有的字节,不联网、不解析、不抓取。
收录判据是 ADR-0026 那五条,缺一条就留在本机:
1. provenance 里有 `source_url`;没有来源证据的不收,不靠文件名推断归属。
2. 归属是法人实体。本目录按定义只装厂牌标识,自然人头像在 `avatars/`,不在输入范围。
3. 用途是标识:`variant` 为 `icon`、`logo` 或没有变体的裸文件。
4. 单文件不超过 256 KB。
5. 总量不超过 8 MB。
输出形态和缓存目录不同,这是收录的重点:文件名带真实扩展名(`Attackers.png`),
扩展名按字节嗅探定而不是照抄源地址——源地址的后缀经常和内容对不上,`.jpg` 里装 PNG
在这批里就有。provenance 集中进一份 `manifest.json`,不为每个图放边车。
可重复执行:同一份输入两次运行产出逐字节一致的清单和文件,不产生无变化的改动。清单
按落盘名排序,时间戳取自源边车的 `imported_at` 而不是运行时刻——用 `now()` 的话每次
运行都会刷出一份全新的 diff。
默认只报告,`--apply` 才写盘。写盘会先清掉目标目录里清单不再登记的文件,否则厂牌改名
后旧名字那份会永远留在仓库里。
"""
from __future__ import annotations
import argparse
import hashlib
import json
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 import brand_marks
from peach.config import GENERATED_DIR
from peach.follow_assets import sniff
from peach.previews import LOGO_VARIANTS
#: 嗅探出的 MIME → 落盘扩展名。`brand_marks.CONTENT_TYPES` 是它的反查表,两边必须
#: 一一对应,否则收得进去、运行时认不出类型。
EXTENSIONS = {mime: ext for ext, mime in brand_marks.CONTENT_TYPES.items()}
def variant_of(stem: str) -> str:
"""`Attackers.icon` → `icon`;裸文件返回空串。"""
for variant in LOGO_VARIANTS:
if stem.endswith(f".{variant}"):
return variant
return ""
def collect(source: Path) -> tuple[list[dict], list[dict]]:
"""扫源目录,返回(收录条目, 落选条目)。两边都带原因,落选的要能解释。"""
taken: list[dict] = []
skipped: list[dict] = []
for path in sorted(source.glob("*.img")):
stem = path.name[:-len(".img")]
sidecar = path.with_name(f"{path.name}.provenance.json")
meta = {}
if sidecar.is_file():
try:
meta = json.loads(sidecar.read_text(encoding="utf-8"))
except (OSError, ValueError):
meta = {}
url = meta.get("source_url") or meta.get("url") or meta.get("resolved_url")
if not url:
skipped.append({"name": stem, "reason": "no_source_url"})
continue
data = path.read_bytes()
if len(data) > brand_marks.MAX_FILE_BYTES:
skipped.append({"name": stem, "reason": "too_large", "bytes": len(data)})
continue
mime = sniff(data)
extension = EXTENSIONS.get(mime or "")
if not extension:
skipped.append({"name": stem, "reason": "unknown_format", "sniffed": mime})
continue
taken.append({
"file": f"{stem}{extension}",
"studio": stem.rsplit(".", 1)[0] if variant_of(stem) else stem,
"variant": variant_of(stem),
"source": meta.get("source", ""),
"source_url": url,
"sha256": hashlib.sha256(data).hexdigest(),
"bytes": len(data),
"content_type": mime,
"imported_at": meta.get("imported_at") or meta.get("cached_at", ""),
"_data": data,
})
taken, same = _drop_copies_of_base(taken)
skipped.extend(same)
taken.sort(key=lambda row: row["file"])
return taken, skipped
def _drop_copies_of_base(taken: list[dict]) -> tuple[list[dict], list[dict]]:
"""和同名裸文件字节相同的 `icon`/`logo` 不收。
指定来源表一张图管两位,装的时候大位、小位和裸文件各落一份同样的字节;2026-09-23
量过,按内容去重前 9.9 MB,去重后 5.2 MB。`PreviewService.logo` 找不到变体就回落到
裸文件,所以少了这份副本,每个位置取到的仍是同一张图。
"""
base_sha = {row["studio"]: row["sha256"] for row in taken if not row["variant"]}
kept, same = [], []
for row in taken:
if row["variant"] and base_sha.get(row["studio"]) == row["sha256"]:
same.append({"name": row["file"], "reason": "same_as_base"})
else:
kept.append(row)
return kept, same
def write(taken: list[dict], target: Path) -> dict:
"""落盘并返回统计。清单不再登记的文件一并清掉。"""
studios = target / "studios"
studios.mkdir(parents=True, exist_ok=True)
wanted = {row["file"] for row in taken}
removed = []
for path in sorted(studios.iterdir()):
if path.is_file() and path.name not in wanted:
path.unlink()
removed.append(path.name)
written = 0
for row in taken:
path = studios / row["file"]
data = row["_data"]
# 内容相同就不重写:让重复执行不产生 mtime 抖动,也不在 diff 里留空条目。
if not path.is_file() or path.read_bytes() != data:
path.write_bytes(data)
written += 1
manifest = {
"version": 1,
"note": "ADR-0026:法人实体标识随仓库分发。由 scripts/sync_brand_marks.py 生成。",
"marks": [{key: row[key] for key in row if key != "_data"} for row in taken],
}
(target / "manifest.json").write_text(
json.dumps(manifest, ensure_ascii=False, indent=2, sort_keys=False) + "\n",
encoding="utf-8", newline="\n")
return {"written": written, "removed": removed}
def build_parser() -> argparse.ArgumentParser:
parser = argparse.ArgumentParser(description=__doc__.splitlines()[0])
parser.add_argument("--source", type=Path, default=GENERATED_DIR / "logos")
parser.add_argument("--target", type=Path, default=brand_marks.MARKS_ROOT)
parser.add_argument("--apply", action="store_true")
return parser
def main(argv: list[str] | None = None) -> int:
args = build_parser().parse_args(argv)
if not args.source.is_dir():
print(json.dumps({"ok": False, "error": "source missing",
"source": str(args.source)}, ensure_ascii=False))
return 1
taken, skipped = collect(args.source)
total = sum(row["bytes"] for row in taken)
result = {
"ok": True,
"applied": bool(args.apply),
"taken": len(taken),
"bytes": total,
"skipped": len(skipped),
"reasons": sorted({row["reason"] for row in skipped}),
}
if total > brand_marks.MAX_TOTAL_BYTES:
# 判据 5。越过上限说明混进了不该收的东西,调高上限不是修法。
result.update(ok=False, error="total exceeds budget",
budget=brand_marks.MAX_TOTAL_BYTES)
print(json.dumps(result, ensure_ascii=False))
return 1
if args.apply:
result.update(write(taken, args.target))
print(json.dumps(result, ensure_ascii=False))
return 0
if __name__ == "__main__":
raise SystemExit(main())