fix(photos): 入库压缩保留 EXIF 并补未来上传的压缩缺口

- 迁移器与新增 photos_ingest.py 压缩时显式 exif= 传参(PIL 重存默认
  剥离——此前 21 张已全部丢 EXIF,图库参数条构建期数据源被抽空)
- exif=None 会使 PIL JPEG 保存 len(None) 崩,改条件传参
- photos_ingest.py:幂等入库处理器(长边 2400/q86,>1.5M 才动),
  堵住 gitea/本地上传原图直入仓的缺口(196M 事故成因)
- 重跑迁移:19/21 带 EXIF(2 张原图本无),体积 9.7M

验证: pnpm astro build 11 页 exit 0;EXIF 保留计数 19/21

Spec: none (Phase 2 归档后的独立小修:EXIF 修复 + ingest 补缺)

Co-authored-by: GLM-5.3
This commit is contained in:
wait4xandGLM-5.3 committed 2026-10-25 09:01:07 +08:00
1 parent aea21c8694
commit 382e56431f
21 files changed
+52 -1

No files matched your search

+47
View File
@@ -0,0 +1,47 @@
#!/usr/bin/env python3
"""照片入库处理器:把 content/photos/ 下超限图片压缩到站内规范。
为什么存在:gitea web/本地上传的是原图(曾发生 196M/21 张入仓事故,
见 git log aea21c8 的重写说明);sharp 只派生 dist 缩略图不管仓内体积。
工作流:上传照片到 content/photos/(顺带跑一次 sidecar 生成或手写)后、
git commit 前执行本脚本;幂等,已在规范内的文件原样跳过(零 diff)。
规范与迁移器一致:长边 ≤2400px、JPEG q86(scripts/migrate/photos_migrate.py)。
"""
import sys
from pathlib import Path
PHOTOS = Path("content/photos")
MAX_EDGE = 2400
QUALITY = 86
# 跳过阈值:小于此字节数的 jpg 视为已处理过(2400px q86 成片通常 200KB-1MB)
SKIP_BELOW_BYTES = 1_500_000
def main() -> int:
from PIL import Image
changed, skipped = [], 0
for jpg in sorted(PHOTOS.glob("*.jpg")):
size = jpg.stat().st_size
if size < SKIP_BELOW_BYTES:
skipped += 1
continue
img = Image.open(jpg)
# EXIF 显式保留:重存默认剥离会抽走图库参数条的构建期数据源
exif = img.info.get("exif")
if img.mode not in ("RGB", "L"):
img = img.convert("RGB")
img.thumbnail((MAX_EDGE, MAX_EDGE))
kwargs = {"exif": exif} if exif else {}
img.save(jpg, "JPEG", quality=QUALITY, optimize=True, **kwargs)
changed.append((jpg.name, size, jpg.stat().st_size))
for name, before, after in changed:
print(f" compressed: {name} {before/1e6:.1f}M -> {after/1e6:.2f}M")
print(f"photos ingest: {len(changed)} compressed, {skipped} skipped (within spec)")
return 0
if __name__ == "__main__":
sys.exit(main())