Skip to content

armgit / peach

Public Imported from GitHub
0

Peach: Local-first video metadata scraper and media manager.

fix(covers): 阻止 JavDB 水印封面进入图库

longmeidaob6b5032731e5
diff --git a/docs/REUSE.md b/docs/REUSE.md
index 57a9bde34..650436e40 100644
--- a/docs/REUSE.md
+++ b/docs/REUSE.md
@@ -213,11 +213,11 @@ JAV 默认封面(官方封面/预览图)与视频默认大小(大图/
详情人物头像复用 `card-art.entityFaceImg` 与 `avatarFrame`,图片直接挂在有尺寸的 `[data-id-face]` 圆框内。取景读取图片的父框几何,`display:contents` 包装层没有可用尺寸。首字与图片由同一段 HTML 承载,图片加载失败时共用回退链,浏览器回归覆盖带人脸框的桌面与手机详情。
-新作封面优先复用 `best_cover` 的官方候选,未取得时复用 `javdb_public` 的作品身份校验与 `RemoteMediaCache` 的持久字节缓存、失败冷却;来源图片沿用 ADR-0030、0032 的 240px 小图下限,按实际尺寸保存,证据区分小图,再经 `cover_artwork.install_cover` 安装本地封面、正封框、人脸坐标与来源证据。原图缓存与详情预览分键,JPEG 解包后保留原字节、原尺寸;其他格式按原尺寸转为 JPEG。缺图时 `/jav-cover` 提供同源显示回退。
+新作封面复用 `best_cover` 的官方候选与 `cover_artwork.install_cover`,保存本地封面、正封框、人脸坐标与来源证据。JavDB 图片仅作来源复核证据,不进入安装、封面索引、详情封面投影或缺图兜底;`scripts/quarantine_javdb_covers.py` 将有来源证据的历史图片及边车移入可恢复归档。未取得合格来源时显示未取得状态,详见 ADR-0107。
-取景由 `poster-crop-v12` 管理。Blu-ray 来源标题、明确的版次标签与图片比例共同决定 0.85 正封比例,标题随正封框保留,重算人脸与正封框均读取同一版次;有效宽正封框在没有版次标题的卡片上也完整显示。VR 标记、IENEE、IENFE、SDHS 数字发行拼图,以及可信主脸跨过 DVD 正封先验边界的大封保留整幅画面,由 `cover_frame.layout=still` 传给卡片。大封按主脸居中并铺满卡片,大小图均不使用模糊补边;JPEG 软边的书脊识别结合突出度、左忙右静的方向与正封侧基线。手工框保持用户取景。重试队列按最早尝试时间轮转,收起的合集沿用设置跳过;逐番号结果保存在 `feed-cover-attempts/`。复用现有 HTTP、Pillow、OpenCV 与文件锁,无新增依赖。
+取景由 `poster-crop-v12` 管理。Blu-ray 来源标题、明确的版次标签与图片比例共同决定 0.85 正封比例,标题随正封框保留,重算人脸与正封框均读取同一版次;有效宽正封框在没有版次标题的卡片上也完整显示。VR 标记、IENEE、IENFE、SDHS 数字发行拼图,以及可信主脸跨过 DVD 正封先验边界的大封保留整幅画面,由 `cover_frame.layout=still` 传给卡片。大封在竖版大图按主脸居中填满,横版小图完整显示,均不使用模糊补边;JPEG 软边的书脊识别结合突出度、左忙右静的方向与正封侧基线。手工框保持用户取景。重试队列按最早尝试时间轮转,收起的合集沿用设置跳过;逐番号结果保存在 `feed-cover-attempts/`。复用现有 HTTP、Pillow、OpenCV 与文件锁,无新增依赖。
-馆藏与新作取图共用 `cover_artwork.metadata_snapshots` 和 `best_cover`,片商官网、发行方、R18 与 DMM 的官方候选均参与尺寸比较。馆藏采集将长边 1600px 作为高清重探界限,社区转存图可由官方封面替换;同类官方图按像素面积升级,FC2 保留商品图的来源保护。`scripts/audit_cover_artwork.py` 逐张核对尺寸、来源、版次与取景框,输出 CSV;`scripts/upgrade_audited_covers.py` 按已有官方快照逐番号重探,保存原件备份与结果检查点。两者只处理可重建图片和边车,不写 ledger。
+馆藏与新作取图共用 `cover_artwork.metadata_snapshots` 和 `best_cover`,片商官网、发行方、R18 与 DMM 的官方候选均参与尺寸比较。馆藏采集将长边 1600px 作为高清重探界限,社区转存图可由官方封面替换;同类官方图按像素面积升级,FC2 保留商品图的来源保护。`scripts/audit_cover_artwork.py` 逐张核对尺寸、来源、版次与取景框,输出 CSV;`scripts/upgrade_audited_covers.py` 按已有官方快照逐番号重探,`--live` 经同一官方链在线补查,保存原件备份与结果检查点。两者只处理可重建图片和边车,不写 ledger。
## 复用决策门槛
@@ -499,7 +499,7 @@ A 盘垃圾隔离使用短名称 `peach-purge-<32 位 UUID>.peach-quarantine`,
| FANBOX 正文解析 | PixivUtil2 `v20251112` / `e537e96` 的公开正文模型(BSD-2-Clause,只复用数据模型,不引入整套下载器) | Peach 的独立规范化 DTO 覆盖 image/text/file/article/video/entry、`fileMap`、`embedMap`、`urlEmbedMap` 和旧 HTML 正文,保留正文顺序、稳定去重、可播放媒体与文件页边界;许可证依据写在实现头部,传输固定 `curl_cffi` |
| 云下载传输 | `grpcio==1.84.0` 与 `protobuf==7.36.2`(消息代码由 grpcio-tools 1.84.0 生成后提交),首个消费者 `downloads_clouddrive` | CloudDrive2 官方 proto 的七个方法子集、令牌存 CredentialStore、任务表与失败分类(`downloads.py`,ADR-0089) |
| 本地 NFO 导入 | Kodi/Jellyfin NFO 协议(2026-09-06 核对 [Kodi](https://kodi.wiki/view/NFO_files/Movies)、[Jellyfin](https://jellyfin.org/docs/general/server/metadata/nfo/)),标准库 ElementTree,图片走项目固定版 Pillow;实现在 `library_nfo.py` | 只适配影片、单集与音乐视频(整剧、专辑、播放记录与远端图片引用不套用),拒绝 DTD、超大输入与越目录图片引用。同名边车优先,`movie.nfo` 与通用海报只用于单影片目录;正片和图片同属一组连号(`(1).mp4` 配 `(1).jpg`…)时同名图是图集的一张。已知内容词投影为中文标签,画质、促销、发行属性与演员编成丢弃,未收录的自定义标签保留;远端补厂牌、导演、发行商、时长与图片出处,人物主名精确一致时补 DMM id、假名、罗马字与首次头像,不替换演员真值。网络复用 R18 JSON 入口、SourceTransport、头像缓存与封面解析器。候选按资产 ID 与路径定位,批准时复核目标,时长证据不覆盖探测时长(ADR-0029) |
-| JAV 高清封面 | Javinizer-Go `dd56998328d078c9baf68ff4fde2e6fcaa2a691a`(MIT)的 DMM modern `awsimgsrc.dmm.com/dig/...` 映射与尺寸门槛;Prestige 公开 API 的查询模型参考 MDCX `58e3f930f2e864fceb8a53ceef818716e2a6413d`(GPL-3.0,只作协议证据) | 候选、Range 量尺寸、面积最大者胜出与仅更大才原子升级见 `peach-jav-cover-workflow`。批量流程不请求社区来源,历史快照里的社区站只借厂牌;既有库采集在官方落空时经 `peach.community_catalog` 查 AVBase(取商品号认得出这个番号的店铺条目,名寄せ的标题可能来自合集)、JavBus 与 javdb,两个图源先按 dHash 求一致,取景不同时用 OpenCV ORB(相似变换内点 ≥60 算同一张),单一图源照用并留空 `verified_by`,遇验证页不绕过(ADR-0030、0032)。MDC-NG 只证明 Amazon 日本渠道存在,只留 POC 候选。不写 ledger |
+| JAV 高清封面 | Javinizer-Go `dd56998328d078c9baf68ff4fde2e6fcaa2a691a`(MIT)的 DMM modern `awsimgsrc.dmm.com/dig/...` 映射与尺寸门槛;Prestige 公开 API 的查询模型参考 MDCX `58e3f930f2e864fceb8a53ceef818716e2a6413d`(GPL-3.0,只作协议证据) | 候选、Range 量尺寸、面积最大者胜出与仅更大才原子升级见 `peach-jav-cover-workflow`。批量流程不请求社区来源,历史快照里的社区站只借厂牌;既有库采集在官方落空时经 `peach.community_catalog` 查 AVBase(取商品号认得出这个番号的店铺条目,名寄せ的标题可能来自合集)、JavBus 与 JavDB,两个图源先按 dHash 求一致,取景不同时用 OpenCV ORB(相似变换内点 ≥60 算同一张)。JavDB 只参与复核,不作为安装候选;其他单一图源留空 `verified_by`,遇验证页不绕过(ADR-0107)。MDC-NG 只证明 Amazon 日本渠道存在,只留 POC 候选。不写 ledger |
依赖的第一个消费者及其隔离测试必须在同一改动落地,否则不引入依赖。
表里没有单列的 Python 依赖及首个消费者:`numpy`(vision 组,`face_detect.decode` 直接 import)、`socksio`(PikPak 的 SOCKS 代理)、
diff --git a/docs/adr/0107-cover-quality-and-javdb-evidence.md b/docs/adr/0107-cover-quality-and-javdb-evidence.md
new file mode 100644
index 000000000..fab3fbbc8
--- /dev/null
+++ b/docs/adr/0107-cover-quality-and-javdb-evidence.md
@@ -0,0 +1,25 @@
+# ADR-0107:官方高清封面优先,JavDB 图片仅作复核证据
+
+状态:Accepted
+
+日期:2026-10-17
+
+关联:ADR-0030、ADR-0032、ADR-0031。
+
+## 背景
+
+封面来源、分辨率与取景分别决定图库显示质量。社区图片可能带有站点水印;可信官方快照中的大图需要进入候选比较。整幅宣传大封与包含封底、书脊、正封的封套具有相近比例,不能只依靠比例决定裁切。
+
+## 决策
+
+- 可信官方快照与在线官方候选共用身份校验、解码和尺寸比较。远程封面的高清重探界限为长边 1600px;本地自备图与 FC2 商品图沿用各自的来源保护。同类官方图只按像素面积升级,官方图可以替换社区转存图。
+- JavDB 图片只作来源复核与图像一致性证据,不作为正式封面、下载兜底或页面显示兜底。安装、封面索引与详情投影均执行此边界。官方图片未取得时显示未取得状态。
+- 有 JavDB 来源证据的历史封面及其边车移入 `review/` 下的可恢复归档,保留逐番号清单。来源未登记不等于 JavDB,不凭缺少来源批量移走文件。整个流程不写 ledger。
+- `poster-crop-v12` 结合主脸是否跨越正封先验边界识别整幅大封,结合梯度突出度与正封侧基线收紧书脊切点;手工取景保持原样。
+- 整幅大封在竖版大图卡片按主脸居中填满,在横版小图卡片完整放入容器;两种视图均不使用模糊补边。图片比例不同,横版完整显示时允许实色留边。
+
+## 验证与复核产物
+
+来源筛选、拒收安装、只读封面路径、详情投影、批次归档与取景计算分别运行行为回归。桌面及 390×844 手机浏览器核对取景切换、无模糊背景和横版完整显示。
+
+`scripts/audit_cover_artwork.py` 输出逐文件 CSV,`scripts/upgrade_audited_covers.py` 保存官方候选尺寸、来源和替换备份,`scripts/quarantine_javdb_covers.py` 保存归档清单;网络失败与确认未取得分别记录,不将低清放大视为高清。
diff --git a/frontend/e2e/fc2-artwork.test.ts b/frontend/e2e/fc2-artwork.test.ts
index c93cb6f7a..ac69ba4e7 100644
--- a/frontend/e2e/fc2-artwork.test.ts
+++ b/frontend/e2e/fc2-artwork.test.ts
@@ -92,7 +92,7 @@ describe('FC2 合集与封面取景', () => {
return { fit: getComputedStyle(img).objectFit, blur: pic.style.getPropertyValue('--cover-blur'),
widthGap: Math.abs(box.width - image.width), heightGap: Math.abs(box.height - image.height) };
}, shape);
- assert.equal(fill.fit, 'cover', JSON.stringify(fill));
+ assert.equal(fill.fit, shape === 'front' ? 'cover' : 'contain', JSON.stringify(fill));
assert.equal(fill.blur, '', JSON.stringify(fill));
assert.ok(fill.widthGap < 1 && fill.heightGap < 1, JSON.stringify(fill));
}
diff --git a/frontend/src/react/components/source-picture.tsx b/frontend/src/react/components/source-picture.tsx
index 0013de358..402c2229e 100644
--- a/frontend/src/react/components/source-picture.tsx
+++ b/frontend/src/react/components/source-picture.tsx
@@ -19,5 +19,5 @@ export function SourcePicture({ item, image }: {
const src = local ? coverUrl(item, true) : result.data?.image || (image === 'cover' && !item.source_image ? item.cover_url : '');
if (src) return <img data-jav-picture="" src={src} alt={image === 'cover' ? `${item.code} 封面` : `${item.code} 预览 ${Number(image) + 1}`} referrerPolicy="no-referrer" />;
if (remote && result.isPending) return <div data-jav-picture-wait="" data-skeleton="image" role="status" aria-label="正在读取图片"><span className="relative skeleton-sheen" /></div>;
- return <div data-jav-picture-wait=""><Note tone="neutral" action={remote ? <Button variant="secondary" {...busyProps(result.isFetching)} onClick={() => { void result.refetch() }}>重试</Button> : undefined}>{result.data?.message || (result.error ? '图片未取得' : '暂无封面')}</Note></div>;
+ return <div data-jav-picture-wait=""><Note tone="neutral" action={remote ? <Button variant="secondary" {...busyProps(result.isFetching)} onClick={() => { void result.refetch() }}>重试</Button> : undefined}>{result.data?.message || (result.error ? '图片未取得' : '封面未取得')}</Note></div>;
}
diff --git a/scripts/quarantine_javdb_covers.py b/scripts/quarantine_javdb_covers.py
new file mode 100644
index 000000000..a0e544f6e
--- /dev/null
+++ b/scripts/quarantine_javdb_covers.py
@@ -0,0 +1,66 @@
+"""将有 JavDB 来源证据的封面移入可恢复备份区,正式封面目录只保留合格图源。"""
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1] / 'src'))
+
+from peach.cover_artwork import has_javdb_source
+from peach.fsutil import atomic_write_text
+
+
+def main(argv=None):
+ parser = argparse.ArgumentParser(description=__doc__)
+ parser.add_argument('--data-root', type=Path, required=True)
+ parser.add_argument('--report', type=Path, required=True)
+ parser.add_argument('--apply', action='store_true')
+ args = parser.parse_args(argv)
+ root, report = args.data_root.resolve(), args.report.resolve()
+ if not report.is_relative_to(root / 'review'):
+ parser.error('备份区必须位于当前数据根的 review 内')
+ covers = root / 'generated/covers'
+ paths = [path for path in sorted(covers.glob('*.jpg')) if has_javdb_source(path)]
+ report.mkdir(parents=True, exist_ok=True)
+ manifest = report / 'manifest.json'
+ history = json.loads(manifest.read_text(encoding='utf-8')).get('items', []) if manifest.exists() else []
+ items = {row['code']: row for row in history}
+ def save():
+ result = dict(matched=len(paths), quarantined=sum(row['status'] == 'quarantined' for row in items.values()),
+ applied=args.apply, items=list(items.values()))
+ atomic_write_text(manifest, json.dumps(result, ensure_ascii=False, indent=2))
+ return result
+ for path in paths:
+ moves = []
+ for suffix in ('.jpg', '.face.json', '.poster.json', '.scraping.json'):
+ source = path.with_suffix(suffix)
+ if not source.exists():
+ continue
+ destination = report / 'originals' / source.name
+ moves.append((source, destination))
+ if args.apply:
+ for _, destination in moves:
+ if destination.exists():
+ raise FileExistsError(destination)
+ completed = []
+ try:
+ for source, destination in moves:
+ destination.parent.mkdir(parents=True, exist_ok=True)
+ source.rename(destination)
+ completed.append((source, destination))
+ except OSError:
+ for source, destination in reversed(completed):
+ destination.rename(source)
+ raise
+ items[path.stem] = dict(code=path.stem, status='quarantined' if args.apply else 'planned',
+ files=[str(destination) for _, destination in moves])
+ save()
+ result = save()
+ print(json.dumps({key: value for key, value in result.items() if key != 'items'}, ensure_ascii=False))
+ return 0
+
+
+if __name__ == '__main__':
+ raise SystemExit(main())
diff --git a/scripts/upgrade_audited_covers.py b/scripts/upgrade_audited_covers.py
index b9ad81d3c..d780b1acf 100644
--- a/scripts/upgrade_audited_covers.py
+++ b/scripts/upgrade_audited_covers.py
@@ -16,28 +16,32 @@ from peach.cover_artwork import install_cover
from peach.fsutil import atomic_write_text
from peach.images import measure_image_file
from peach.jav_cover_fetch import (HostLimitedTransport, NotFound, Unavailable, _download_best,
- _measure, _no_usable_official, _usable_candidates)
+ _measure, _no_usable_official, _usable_candidates, best_cover)
from peach.library_processing import _official_evidence, artwork_title
from peach.scraping_access import SourceTransport
from peach.community_catalog import origin_of
-def upgrade(path, known, transport, report):
+def upgrade(path, known, transport, report, *, live=False, metadata_root=None):
evidence = _official_evidence(known)
- if not evidence or not evidence.candidates:
+ if not live and (not evidence or not evidence.candidates):
return {'code': path.stem, 'status': 'no_official_snapshot'}
diagnostics = {}
def record(key):
diagnostics[key] = diagnostics.get(key, 0) + 1
deadline = time.monotonic() + 90
- candidates = _usable_candidates(path.stem, evidence.candidates, {}, 0, 0)
- measured = _measure(transport, candidates, record,
- minimum_width=240, delay=0, deadline=deadline)
- if not measured:
- kind = Unavailable if diagnostics.get('unanswered') else NotFound
- raise kind(_no_usable_official(diagnostics))
- candidate, size, data = _download_best(transport, measured, record,
- minimum_width=240, minimum_quality=0, minimum_pixels=0, deadline=deadline)
+ if live:
+ candidate, size, data = best_cover(transport, path.stem, 0, known=evidence,
+ metadata_root=metadata_root, minimum_width=240, deadline=deadline, diagnostics=diagnostics)
+ else:
+ candidates = _usable_candidates(path.stem, evidence.candidates, {}, 0, 0)
+ measured = _measure(transport, candidates, record,
+ minimum_width=240, delay=0, deadline=deadline)
+ if not measured:
+ kind = Unavailable if diagnostics.get('unanswered') else NotFound
+ raise kind(_no_usable_official(diagnostics))
+ candidate, size, data = _download_best(transport, measured, record,
+ minimum_width=240, minimum_quality=0, minimum_pixels=0, deadline=deadline)
previous = read(path.with_suffix('.scraping.json'))
old_size = measure_image_file(path)
community = origin_of(str(previous.get('source_url') or '')) in ('javdb', 'javbus')
@@ -54,9 +58,9 @@ def upgrade(path, known, transport, report):
shutil.copy2(original, saved)
digest = hashlib.sha256(data).hexdigest()
install_cover(path, path.stem, data, size, evidence=dict(source=candidate.source,
- source_url=candidate.url, title=artwork_title(known), width=size[0], height=size[1],
+ source_url=candidate.url, title=artwork_title(known) or previous.get('title', ''), width=size[0], height=size[1],
raw_sha256=digest, installed_sha256=digest, checked_at=time.time(),
- resolver='official-snapshot-upgrade-v1', verified_by=[]))
+ resolver='official-live-upgrade-v1' if live else 'official-snapshot-upgrade-v1', verified_by=[]))
return {**result, 'status': 'installed'}
@@ -67,6 +71,7 @@ def main():
parser.add_argument('--code', action='append', default=[])
parser.add_argument('--apply', action='store_true')
parser.add_argument('--retry-failures', action='store_true')
+ parser.add_argument('--live', action='store_true', help='通过现有官方链在线补查,无需已有快照')
args = parser.parse_args()
root = args.data_root / 'generated/covers'
paths = [root / (code + '.jpg') for code in args.code] if args.code else sorted(root.glob('*.jpg'))
@@ -84,13 +89,14 @@ def main():
evidence = read(path.with_suffix('.scraping.json'))
eligible = bool(args.code) or (size and (max(size) < 1600 or origin_of(
str(evidence.get('source_url') or '')) in ('javdb', 'javbus')))
- if not eligible or not _official_evidence(known):
+ if not eligible or (not args.live and not _official_evidence(known)):
continue
if not args.apply:
print(json.dumps({'code': path.stem, 'size': size, 'eligible': True}), flush=True)
continue
try:
- result = upgrade(path, known, transport, args.report)
+ result = upgrade(path, known, transport, args.report, live=args.live,
+ metadata_root=args.data_root / 'sources/metadata/javinizer-go')
except Exception as error:
from peach.library_processing import describe_failure
result = dict(code=path.stem, status='failed', error=describe_failure(error))
diff --git a/src/peach/community_catalog.py b/src/peach/community_catalog.py
index 9ee4a6176..c9051270c 100644
--- a/src/peach/community_catalog.py
+++ b/src/peach/community_catalog.py
@@ -22,7 +22,7 @@ import httpx
from PIL import Image
from .jav_cover_fetch import (SMALL_MIN_WIDTH, Candidate, NotFound, Unavailable, _fetch, candidate_for,
- is_cross_product_cover)
+ is_cross_product_cover, is_javdb_image)
from .scraping_access import SourcePaused
from .scripting import host_under, hostname_of
from .sources import AVBASE, JAVBUS, JAVDB
@@ -216,7 +216,10 @@ def _unverified(pool: list[Picture], usable: int,
if len(origins) > 1:
# 几家都答了、图都下到了,只是互相对不上:再问还是这几张,交 `NotFound` 让采集任务记住。
raise NotFound(f"{'、'.join(origins)} 给的封面不是同一张图,无法互相印证")
- largest = max(pool, key=lambda one: one.pixels)
+ allowed = [one for one in pool if not is_javdb_image(one.candidate.url)]
+ if not allowed:
+ raise NotFound('未取得无 JavDB 水印的封面')
+ largest = max(allowed, key=lambda one: one.pixels)
return largest.candidate, largest.size, largest.data, ()
@@ -241,6 +244,8 @@ def verified_cover(transport, code: str, works: list[tuple[str, dict]], *,
pool, fetched, usable = _pictures(transport, urls, reference, siblings, deadline=deadline)
best = None
for one in pool:
+ if is_javdb_image(one.candidate.url):
+ continue
origins = {other.origin for other in pool if same_picture(one, other)}
if len(origins) >= 2 and (best is None or one.pixels > best[0].pixels):
best = (one, tuple(sorted(origins)))
diff --git a/src/peach/cover_artwork.py b/src/peach/cover_artwork.py
index 8139dcaa1..6a8fbff19 100644
--- a/src/peach/cover_artwork.py
+++ b/src/peach/cover_artwork.py
@@ -12,12 +12,22 @@ from filelock import FileLock
from . import jav_poster_crop
from .face_detect import FaceDetector, FaceModelUnavailable, decode, main_face
+from .jav_cover_fetch import Unavailable, is_javdb_image
FRONT_START = 0.468
MIN_FACE_Y, MAX_FACE_Y = 0.05, 0.60
_INSTALL_LOCK = threading.RLock()
+def has_javdb_source(path: Path) -> bool:
+ """封面旁的来源证据是否指向 JavDB 转存图。"""
+ try:
+ evidence = json.loads(Path(path).with_suffix('.scraping.json').read_text(encoding='utf-8'))
+ except (OSError, ValueError):
+ return False
+ return isinstance(evidence, dict) and is_javdb_image(evidence.get('source_url'), evidence.get('source'))
+
+
@lru_cache(maxsize=1)
def _default_detector() -> FaceDetector:
return FaceDetector(allow_download=False)
@@ -93,6 +103,8 @@ def _sidecars(code: str, payload: bytes, size: tuple[int, int], detector=None, e
def install_cover(target: Path, code: str, payload: bytes, size: tuple[int, int], *,
evidence: dict | None = None, detector=None) -> None:
"""准备完成后安装整组文件;失败时还原旧组,避免图片与取景跨版本。"""
+ if evidence and is_javdb_image(evidence.get('source_url'), evidence.get('source')):
+ raise Unavailable('未取得无 JavDB 水印的封面')
target = Path(target)
target.parent.mkdir(parents=True, exist_ok=True)
prepared = _sidecars(code, payload, size, detector, evidence)
diff --git a/src/peach/feed_followup.py b/src/peach/feed_followup.py
index becc66f58..fd6043f12 100644
--- a/src/peach/feed_followup.py
+++ b/src/peach/feed_followup.py
@@ -22,8 +22,6 @@
from __future__ import annotations
import hashlib
-import base64
-import io
import json
import re
import time
@@ -163,16 +161,15 @@ def collect(provider, code: str) -> tuple[dict, str | None]:
def fetch_cover(contract, code: str,
cover_url: str | None = None) -> tuple[bytes, tuple[int, int], dict]:
- """优先官方封面,未取得时使用已绑定作品的来源封面;返回图片与来路。
+ """取得当前作品的官方封面与来路。
`cover_url` 是取资料那一步来源给的地址,它是官方图路径时一起参与择优。`DAZD-308`
这类番号 r18 与本机快照里都没有,官方渠道一条候选也列不出来,来源给的却正是 DMM 那张
封套;不收它,这部就只能在页面上拿远程图按固定比例切,书脊切不干净。
取不到就抛;原因交给调用方记账,一部取不到不影响同批其余几部。
"""
- from .jav_cover_fetch import HostLimitedTransport, Unavailable, best_cover, official_image_candidates
- from .scraping_access import SourcePaused, SourceTransport
- import httpx
+ from .jav_cover_fetch import HostLimitedTransport, best_cover, official_image_candidates
+ from .scraping_access import SourceTransport
raw = SourceTransport(contract.follow_secrets_root, max_requests=40,
max_bytes=32 * 1024 * 1024, max_seconds=COVER_SECONDS)
@@ -181,15 +178,12 @@ def fetch_cover(contract, code: str,
from .library_processing import _official_evidence, artwork_title
snapshots = metadata_snapshots(contract.follow_sources_root / 'library-metadata', code)
try:
- try:
- candidate, size, data = best_cover(
- transport, code, 0,
- metadata_root=contract.follow_sources_root / "metadata" / "javinizer-go",
- known=_official_evidence(snapshots),
- prior_candidates=tuple(official_image_candidates(cover_url or "")),
- deadline=time.monotonic() + COVER_SECONDS)
- except (Unavailable, SourcePaused, httpx.TransportError, OSError):
- return source_cover(contract, code)
+ candidate, size, data = best_cover(
+ transport, code, 0,
+ metadata_root=contract.follow_sources_root / "metadata" / "javinizer-go",
+ known=_official_evidence(snapshots),
+ prior_candidates=tuple(official_image_candidates(cover_url or "")),
+ deadline=time.monotonic() + COVER_SECONDS)
finally:
transport.close()
digest = hashlib.sha256(data).hexdigest()
@@ -201,37 +195,6 @@ def fetch_cover(contract, code: str,
return data, size, evidence
-def source_cover(contract, code: str) -> tuple[bytes, tuple[int, int], dict]:
- """已绑定作品的来源封面,复用详情的身份校验与持久图片缓存。"""
- from PIL import Image
- from .jav_cover_fetch import MIN_WIDTH, SMALL_MIN_WIDTH, Unavailable
- from .javdb_public import PUBLIC_API, public_cache_scope
- from .remote_media_cache import cache_for
-
- with contract.database.read_connection() as connection:
- row = connection.execute('SELECT link,title FROM feed_discovery WHERE code=?', (code,)).fetchone()
- if row is None:
- raise Unavailable('未取得已绑定的来源作品')
- link = re.fullmatch(r'https://javdb\.com/v/([A-Za-z0-9]+)/?', str(row['link'] or ''))
- if link is None:
- raise Unavailable('未取得已绑定的来源作品')
- with public_cache_scope(contract.cover_root.parent / 'jav-public'):
- PUBLIC_API.remember_movie_identity(code, {'number': code, 'id': link[1]})
- data = cache_for(contract).bytes(f'jav:{code}:cover:original', lambda:
- base64.b64decode(PUBLIC_API.image(code, 'cover', original=True).partition(',')[2], validate=True))
- with Image.open(io.BytesIO(data)) as image:
- size = image.size
- image.verify()
- if size[0] < SMALL_MIN_WIDTH:
- raise Unavailable('来源封面宽度不足 240px')
- digest = hashlib.sha256(data).hexdigest()
- return data, size, {'code': code, 'width': size[0], 'height': size[1],
- 'source': 'javdb-public', 'source_url': row['link'], 'title': row['title'],
- 'verified_by': [], 'raw_sha256': digest, 'installed_sha256': digest,
- 'quality': 'source-small' if size[0] < MIN_WIDTH else 'source',
- 'checked_at': time.time(), 'resolver': 'peach-jav-source-cover-v1'}
-
-
def record_cover_attempt(contract, code: str, *, ok: bool, error=None):
"""逐番号保存封面处理结果;不记录请求凭据或来源返回正文。"""
from .fsutil import atomic_write_text
diff --git a/src/peach/jav_cover_fetch.py b/src/peach/jav_cover_fetch.py
index 73af45918..11712b39f 100644
--- a/src/peach/jav_cover_fetch.py
+++ b/src/peach/jav_cover_fetch.py
@@ -23,7 +23,7 @@ r"""按番号抓官方封套:把所有候选量一遍,留像素最多的那
- 上轮成功日志里的原 URL,保住已经发现但当前无法重新检索的 DUGA 等官方图。
批量流程不请求社区来源;采集任务在官方渠道落空时经 `peach.community_catalog` 去问
-AVBase、JavBus 与 javdb,封面先求两个图源比对一致(ADR-0030、ADR-0032)。DUGA 批量搜索 API
+AVBase、JavBus 与 JavDB,封面先求两个图源比对一致;JavDB 只作比对证据(ADR-0107)。DUGA 批量搜索 API
需要代理店应用 ID,未配置前只复用成功日志中已经取得的精确图片 URL。
两条番号改写规则,都由实测得出:
@@ -111,6 +111,19 @@ FC2_HIRES = re.compile(r"https://contents-thumbnail\d*\.fc2\.com/w(?:7\d\d|[89]\
IMAGE_URL = re.compile(r"https?://[^\"'\\ )]+?\.(?:jpg|jpeg|png|webp)")
+def is_javdb_image(url: str = '', source: str = '') -> bool:
+ """JavDB 转存图片仅参与来源印证,不作为正式封面。"""
+ for value in (url, source):
+ value = str(value or '').strip().lower()
+ if value in {'javdb', 'javdb-public', 'javdb_public'}:
+ return True
+ host = urlparse(value).hostname if '://' in value else value
+ if host and any(host == domain or host.endswith('.' + domain)
+ for domain in ('jdbstatic.com', 'javdb.com')):
+ return True
+ return False
+
+
@dataclass(frozen=True)
class Candidate:
source: str
@@ -742,14 +755,12 @@ def best_cover(transport: HttpTransport, code: str, delay: float, *,
deadline: float | None = None,
diagnostics: dict[str, int] | None = None,
known: MetadataEvidence | None = None,
- sites_when_needed: bool = False,
) -> tuple[Candidate, tuple[int, int], bytes]:
"""官方封面择优:手上的候选、缓存快照里的候选,加上 r18.dev、MGS 或 Prestige 现问来的候选。
`known` 是调用方这一趟已经取到的官方快照证据(例如 `r18_payload_evidence`),与 `metadata_root`
- 那份缓存同等对待:`sources` 里有 `r18dev` 就不再问 r18.dev。`sites_when_needed` 时先量手上的候选,
- 其中有一张宽到 `MIN_WIDTH` 就直接用它,量不出这么宽的才去问站——资料那一步刚取回的快照
- 多半已经带着原图地址,再问一遍 r18.dev 只是白等一次主机间隔。
+ 那份缓存同等对待:`sources` 里有 `r18dev` 就不再问 r18.dev。所有可用官方候选一起量尺寸,
+ 并按像素面积选择。
"""
# 韩国 MIB 的编号在 JAV 目录站上要么不存在、要么撞上番号相同的日本作品:
# `HA-101`、`MY-102` 取回的都是那部日本片的封套,番号核验拦不住,因为番号本来
@@ -767,22 +778,11 @@ def best_cover(transport: HttpTransport, code: str, delay: float, *,
return _usable_candidates(code, found, known_sizes or {}, minimum_quality, minimum_pixels)
first = usable([*prior_candidates, *evidence.candidates])
- tried, measured = [], []
- if sites_when_needed and first:
- tried = first
- measured = _measure(transport, first, record, minimum_width=minimum_width,
- delay=delay, deadline=deadline)
- if any(size[0] >= MIN_WIDTH for _, _, size in measured):
- return _download_best(transport, measured, record, minimum_width=minimum_width,
- minimum_quality=minimum_quality, minimum_pixels=minimum_pixels,
- deadline=deadline)
asked = usable([*first, *_site_candidates(transport, code, evidence, delay=delay, deadline=deadline)])
- candidates = [candidate for candidate in asked if candidate not in tried]
- if not candidates and not tried:
+ if not asked:
raise NotFound("所有渠道都没有候选")
- if candidates:
- measured += _measure(transport, candidates, record, minimum_width=minimum_width,
- delay=delay, deadline=deadline)
+ measured = _measure(transport, asked, record, minimum_width=minimum_width,
+ delay=delay, deadline=deadline)
if not measured:
# 每张候选都回过话(404、太小、占位图、解不开)就是来源的确定答复,下次再问还是这几张:
# 交 `NotFound`,采集任务记下一周不问。有一张断在连接、5xx 或拒绝访问上才可能是这趟没问成。
diff --git a/src/peach/library_processing.py b/src/peach/library_processing.py
index af606f78f..e5ccd083c 100644
--- a/src/peach/library_processing.py
+++ b/src/peach/library_processing.py
@@ -583,8 +583,8 @@ class LibraryMetadataProvider:
`snapshots` 是资料那一步这一趟已经取到、或磁盘缓存里还新鲜的快照 `[(来源, 资料)]`,
`absent` 是一周内说过没有、或不在这个番号链上的站:两者都不再问。r18.dev 与 DMM 的快照
- 给出官方候选(`_official_evidence`),手上的候选里有够宽的就不再问 r18.dev 与 MGS
- (`best_cover` 的 `sites_when_needed`);社区那几家的快照直接交给图源印证。
+ 给出官方候选(`_official_evidence`),并与 `best_cover` 在线官方候选一起比较尺寸;
+ 社区那几家的快照直接交给图源印证,JavDB 图片仅用作比对证据。
几处没问成且全是冷却时报 `SourcePaused`(本趟没轮到),否则报 `Unavailable`。
"""
from .community_catalog import verified_cover
diff --git a/src/peach/routes_media.py b/src/peach/routes_media.py
index 5e80cafe7..b2225224a 100644
--- a/src/peach/routes_media.py
+++ b/src/peach/routes_media.py
@@ -588,21 +588,6 @@ def cover(request: Request, code: str = "", thumb: int = 0,
return _image_response(request, path, media_type="image/jpeg")
[email protected]_route("/jav-cover", methods=["GET", "HEAD"])
-def jav_cover(request: Request, code: str = "", args: dict[str, str] = Depends(require_auth)):
- """来源封面与详情共用图片缓存、身份校验和失败冷却。"""
- from .web_jav_detail import q_jav_image
- try:
- result = q_jav_image(request.app.state.web_contract, {'code': code, 'image': 'cover'})
- except ValueError as error:
- return JSONResponse({'error': str(error)}, status_code=400)
- if not result.get('image'):
- return _asset_response(request, None)
- body = base64.b64decode(result['image'].partition(',')[2], validate=True)
- return Response(b'' if request.method == 'HEAD' else body, media_type='image/jpeg',
- headers={'Cache-Control': 'private, max-age=3600', 'Content-Length': str(len(body))})
-
-
@router.post("/api/cover-crop")
async def cover_crop(request: Request, args: dict[str, str] = Depends(require_auth)):
"""人自己框的正封那一块,或者把它撤掉换回算出来的那个。
diff --git a/src/peach/web_jav_detail.py b/src/peach/web_jav_detail.py
index 9de1a1b02..23a3fd3c1 100644
--- a/src/peach/web_jav_detail.py
+++ b/src/peach/web_jav_detail.py
@@ -6,7 +6,6 @@ import json
import base64
from datetime import datetime, timezone
from functools import wraps
-from urllib.parse import quote
from . import entities, wants, web_catalog
from .catalog_rules import normalise_code_key, tag_cat
@@ -28,12 +27,12 @@ def detail_code(args) -> str:
def discovery_cover_url(item, source) -> str:
- """已知作品缺少封面地址时,由详情的来源缓存提供图片。"""
- if item.get('cover_url'):
- return item['cover_url']
- known = source.get('id') or re.fullmatch(r'https://javdb\.com/v/[A-Za-z0-9]+/?', str(item.get('link') or ''))
- code = normalise_code_key(item.get('code'))
- return '/jav-cover?code=' + quote(code) if known and code else ''
+ """卡片只显示合格图源的封面地址。"""
+ from .jav_cover_fetch import is_javdb_image
+ for value in (item.get('cover_url'), source.get('cover_url')):
+ if value and not is_javdb_image(value):
+ return value
+ return ''
def _refs(contract, connection, kind, names):
@@ -171,9 +170,11 @@ def source_performer_refs(contract, connection, actors):
def _artwork_fields(contract, code, item, source, *, refresh=False):
+ from .jav_cover_fetch import is_javdb_image
+ source_cover = (source or {}).get('cover_url') or (source or {}).get('thumb_url')
return {"has_cover": contract.has_cover(code), "cover_version": contract.cover_version(code),
- "cover_url": item.get("cover_url") or "",
- "source_image": bool(source and (source.get("cover_url") or source.get("thumb_url"))),
+ "cover_url": discovery_cover_url(item, source or {}),
+ "source_image": bool(source_cover and not is_javdb_image(source_cover)),
"preview_count": min(len(source.get("preview_images") or []), 30),
**_preview_fields(contract, code, source, refresh=refresh),
"source_url": "https://javdb.com/v/" + str(source["id"]) if source else item.get("link") or ""}
diff --git a/src/peach/web_state.py b/src/peach/web_state.py
index ba7fb64f3..26717eaa6 100644
--- a/src/peach/web_state.py
+++ b/src/peach/web_state.py
@@ -380,7 +380,8 @@ class WebContract:
if not key:
return None
path = self.cover_root / f"{key}.jpg"
- return path if path.is_file() else None
+ from .cover_artwork import has_javdb_source
+ return path if path.is_file() and not has_javdb_source(path) else None
def cover_index(self) -> dict[str, dict]:
"""封面目录扫一遍的索引:casefold(归一番号) → 两份取景提示加封面版本。
@@ -407,6 +408,9 @@ class WebContract:
for entry in entries:
if not entry.name.endswith(".jpg"):
continue
+ from .cover_artwork import has_javdb_source
+ if has_javdb_source(Path(entry.path)):
+ continue
stem = entry.path[:-len(".jpg")]
scanned[entry.name[:-len(".jpg")].casefold()] = {
"frame": self._cover_focus(stem + ".face.json"),
diff --git a/tests/test_community_catalog.py b/tests/test_community_catalog.py
index e52bdae9d..73fea643b 100644
--- a/tests/test_community_catalog.py
+++ b/tests/test_community_catalog.py
@@ -294,8 +294,8 @@ class VerifiedCoverTests(unittest.TestCase):
works = [("avbase", {"cover_urls": [DMM_COVER]}), ("javdb", {"cover_urls": [JAVDB_COVER]})]
pages = {JAVDB_COVER: gradient(800, 534), DMM_COVER: gradient(400, 267)}
candidate, size, _data, origins = verified_cover(serve(pages), "ABW-358", works)
- self.assertEqual((candidate.url, size, origins), (JAVDB_COVER, (800, 534), ("dmm", "javdb")))
- self.assertEqual(candidate.referer, "https://javdb.com/")
+ self.assertEqual((candidate.url, size, origins), (DMM_COVER, (400, 267), ("dmm", "javdb")))
+ self.assertEqual(candidate.referer, "https://www.dmm.co.jp/")
pages[DMM_COVER] = gradient(400, 267, rising=False)
with self.assertRaisesRegex(NotFound, "^dmm、javdb 给的封面不是同一张图"):
verified_cover(serve(pages), "ABW-358", works)
@@ -307,11 +307,10 @@ class VerifiedCoverTests(unittest.TestCase):
self.assertEqual((candidate.url, candidate.referer, origins),
(JAVBUS_COVER, "https://www.javbus.com/", ("javbus", "javdb")))
- def test_a_cover_only_one_origin_has_is_used_without_verification(self):
- """没有第二个图源可比时照样用最大那张,印证图源留空(ADR-0032):卡着没有封面更糟。"""
+ def test_a_javdb_only_cover_waits_for_an_eligible_image(self):
works = [("avbase", {"cover_urls": []}), ("javdb", {"cover_urls": [JAVDB_COVER]})]
- candidate, size, _data, origins = verified_cover(serve({JAVDB_COVER: gradient(800, 534)}), "IPX-060", works)
- self.assertEqual((candidate.url, size, origins), (JAVDB_COVER, (800, 534), ()))
+ with self.assertRaisesRegex(NotFound, '无 JavDB 水印'):
+ verified_cover(serve({JAVDB_COVER: gradient(800, 534)}), "IPX-060", works)
def test_a_small_official_cover_counts_as_the_second_origin(self):
works = [("javdb", {"cover_urls": [JAVDB_COVER]})]
@@ -319,7 +318,7 @@ class VerifiedCoverTests(unittest.TestCase):
reference = (small, (300, 200), gradient(300, 200))
candidate, _size, _data, origins = verified_cover(serve({JAVDB_COVER: gradient(900, 600)}), "ORETD-615",
works, reference=reference)
- self.assertEqual((candidate.url, origins), (JAVDB_COVER, ("javdb", "mgstage")))
+ self.assertEqual((candidate.url, origins), (small.url, ("javdb", "mgstage")))
with self.assertRaisesRegex(Unavailable, "下载失败"):
verified_cover(serve({}), "ORETD-615", works, reference=reference)
with self.assertRaisesRegex(NotFound, "^javdb、mgstage 给的封面不是同一张图"):
diff --git a/tests/test_feeds_web.py b/tests/test_feeds_web.py
index 395af010d..d98d4eb66 100644
--- a/tests/test_feeds_web.py
+++ b/tests/test_feeds_web.py
@@ -129,14 +129,29 @@ class FeedWebFixture(unittest.TestCase):
class FeedWebTest(FeedWebFixture):
- def test_known_discovery_without_cover_address_uses_source_cover_without_ledger_write(self):
+ def test_javdb_provenance_is_unavailable_to_cover_readers(self):
+ from PIL import Image
+ target = self.contract.cover_root / 'SSIS-950.jpg'
+ target.parent.mkdir(parents=True, exist_ok=True)
+ Image.new('RGB', (800, 538)).save(target)
+ evidence = target.with_suffix('.scraping.json')
+ evidence.write_text(json.dumps({'source_url': 'https://c0.jdbstatic.com/covers/a.jpg'}), encoding='utf-8')
+ self.contract.cache_bust()
+ self.assertFalse(self.contract.has_cover('SSIS-950'))
+ self.assertIsNone(self.contract.cover_path('SSIS-950'))
+ evidence.write_text(json.dumps({'source_url': 'https://pics.dmm.co.jp/a.jpg'}), encoding='utf-8')
+ self.contract.cache_bust()
+ self.assertTrue(self.contract.has_cover('SSIS-950'))
+ self.assertEqual(self.contract.cover_path('SSIS-950'), target)
+
+ def test_known_discovery_without_cover_address_has_no_javdb_image_fallback(self):
self._add()
self._check()
with self.contract.database.write_transaction() as connection:
connection.execute('UPDATE feed_discovery SET cover_url=NULL')
rows = dispatch_api_get(self.contract, '/api/feeds/discoveries', {})['items']
self.assertTrue(rows)
- self.assertTrue(all(row['cover_url'] == '/jav-cover?code=' + row['code'] for row in rows))
+ self.assertTrue(all('jdbstatic.com' not in row['cover_url'] and not row['cover_url'].startswith('/jav-cover') for row in rows))
with self.contract.database.read_connection() as connection:
self.assertTrue(all(row[0] is None for row in connection.execute('SELECT cover_url FROM feed_discovery')))
@@ -573,60 +588,19 @@ class FeedWebTest(FeedWebFixture):
self.assertFalse(attempt['ok'])
self.assertIn('error_type', attempt)
- def test_official_miss_uses_the_bound_source_image(self):
- import base64
+ def test_official_miss_does_not_install_a_bound_javdb_image(self):
from peach.jav_cover_fetch import Unavailable
self._add()
self._check()
with self.contract.database.write_transaction() as connection:
connection.execute("UPDATE feed_discovery SET link='https://javdb.com/v/Source01' WHERE code='SSIS-950'")
- body, size, _ = self._fake_cover(self.contract, 'SSIS-950')
with patch('peach.jav_cover_fetch.best_cover', side_effect=Unavailable('官方封面未取得')), \
- patch('peach.javdb_public.PUBLIC_API.image', return_value=
- 'data:image/jpeg;base64,' + base64.b64encode(body).decode()) as source_image, \
+ patch('peach.javdb_public.PUBLIC_API.image') as source_image, \
patch('peach.javdb_public.PUBLIC_API.remember_movie_identity') as identity:
- found, actual, evidence = self._fetch_cover(self.contract, 'SSIS-950')
- self.assertEqual((found, actual), (body, size))
- self.assertEqual(evidence['source'], 'javdb-public')
- identity.assert_called_once_with('SSIS-950', {'number': 'SSIS-950', 'id': 'Source01'})
- source_image.assert_called_once_with('SSIS-950', 'cover', original=True)
-
- def test_source_image_requires_a_bound_movie_and_full_size(self):
- import base64
- from peach.jav_cover_fetch import Unavailable
- self._add()
- self._check()
- with self.contract.database.write_transaction() as connection:
- connection.execute("UPDATE feed_discovery SET link='https://example.test/v/Source01' WHERE code='SSIS-950'")
- with patch('peach.javdb_public.PUBLIC_API.image') as image:
- with self.assertRaises(Unavailable):
- feed_followup.source_cover(self.contract, 'SSIS-950')
- image.assert_not_called()
- with self.contract.database.write_transaction() as connection:
- connection.execute("UPDATE feed_discovery SET link='https://javdb.com/v/Source01' WHERE code='SSIS-950'")
- from PIL import Image
- buffer = io.BytesIO()
- Image.new('RGB', (200, 200)).save(buffer, 'JPEG')
- with patch('peach.javdb_public.PUBLIC_API.image', return_value=
- 'data:image/jpeg;base64,' + base64.b64encode(buffer.getvalue()).decode()), \
- patch('peach.javdb_public.PUBLIC_API.remember_movie_identity'):
- with self.assertRaisesRegex(Unavailable, '240px'):
- feed_followup.source_cover(self.contract, 'SSIS-950')
-
- def test_bound_portrait_source_cover_keeps_its_native_size(self):
- import base64
- from PIL import Image
- self._add()
- self._check()
- buffer = io.BytesIO()
- Image.new('RGB', (560, 800)).save(buffer, 'JPEG')
- body = buffer.getvalue()
- with patch('peach.javdb_public.PUBLIC_API.image', return_value=
- 'data:image/jpeg;base64,' + base64.b64encode(body).decode()), \
- patch('peach.javdb_public.PUBLIC_API.remember_movie_identity'):
- actual, size, evidence = feed_followup.source_cover(self.contract, 'SSIS-950')
- self.assertEqual((actual, size), (body, (560, 800)))
- self.assertEqual(evidence['quality'], 'source-small')
+ with self.assertRaisesRegex(Unavailable, '官方封面未取得'):
+ self._fetch_cover(self.contract, 'SSIS-950')
+ identity.assert_not_called()
+ source_image.assert_not_called()
def test_vr_layout_survives_without_a_detected_face(self):
target = self.contract.cover_root / 'MDVR-276.face.json'
diff --git a/tests/test_jav_cover_artwork.py b/tests/test_jav_cover_artwork.py
index e7e176318..913db7dc1 100644
--- a/tests/test_jav_cover_artwork.py
+++ b/tests/test_jav_cover_artwork.py
@@ -29,6 +29,17 @@ class CoverArtworkTests(unittest.TestCase):
self.root = Path(tempfile.mkdtemp())
self.target = self.root / "259LUXU-1537.jpg"
+ def test_javdb_images_are_rejected_before_replacing_any_file(self):
+ from peach.jav_cover_fetch import Unavailable
+ data = jpeg((80, 90, 100))
+ self.target.write_bytes(data)
+ for evidence in ({'source_url': 'https://c0.jdbstatic.com/covers/x.jpg'},
+ {'source': 'javdb-public'}):
+ with self.subTest(evidence=evidence), self.assertRaisesRegex(Unavailable, '无 JavDB 水印'):
+ cover_artwork.install_cover(self.target, 'FC2-PPV-12345', data, (840, 472), evidence=evidence)
+ self.assertEqual(self.target.read_bytes(), data)
+ self.assertFalse(self.target.with_suffix('.face.json').exists())
+
def test_large_face_crossing_the_assumed_spine_uses_the_whole_cover(self):
detector = Detector()
with patch.object(detector, 'detect', return_value=[
diff --git a/tests/test_jav_poster_crop.py b/tests/test_jav_poster_crop.py
index db8b430d3..4ceee3a76 100644
--- a/tests/test_jav_poster_crop.py
+++ b/tests/test_jav_poster_crop.py
@@ -11,6 +11,7 @@ import io
import json
import tempfile
import unittest
+from unittest.mock import patch
from pathlib import Path
from PIL import Image
@@ -707,6 +708,60 @@ class BatchScriptTests(unittest.TestCase):
self.assertEqual(self.run_script(), 0)
self.assertFalse(jav_poster_crop.sidecar_path(self.cover).exists())
+ def test_online_upgrade_checks_official_sources_without_local_snapshots(self):
+ from peach.jav_cover_fetch import Candidate
+ script = batch_script('upgrade_audited_covers')
+ candidate = Candidate('pics.dmm.co.jp', 'https://pics.dmm.co.jp/a.jpg', '')
+ report = self.root / 'report'
+ with patch.object(script, 'best_cover', return_value=(candidate, (2184, 1468), self.before)) as fetch:
+ with patch.object(script, 'install_cover') as install:
+ result = script.upgrade(self.cover, [], object(), report, live=True,
+ metadata_root=self.root / 'metadata')
+ self.assertEqual(result['status'], 'installed')
+ self.assertEqual(fetch.call_args.args[1], 'ABW-232')
+ self.assertEqual(fetch.call_args.kwargs['metadata_root'], self.root / 'metadata')
+ self.assertEqual(install.call_args.kwargs['evidence']['resolver'], 'official-live-upgrade-v1')
+ self.assertEqual((report / 'originals/ABW-232.jpg').read_bytes(), self.before)
+
+ def test_javdb_quarantine_preserves_bytes_and_resumes_without_duplicates(self):
+ covers = self.root / 'generated/covers'
+ covers.mkdir(parents=True)
+ target = covers / 'MIDV-377.jpg'
+ target.write_bytes(self.before)
+ evidence = target.with_suffix('.scraping.json')
+ evidence.write_text(json.dumps({'source': 'javdb', 'source_url': 'https://c0.jdbstatic.com/a.jpg'}), encoding='utf-8')
+ official = covers / 'SPONE-001.jpg'
+ official.write_bytes(self.before)
+ report = self.root / 'review/quarantine'
+ args = ['--data-root', str(self.root), '--report', str(report)]
+ script = batch_script('quarantine_javdb_covers')
+ self.assertEqual(script.main(args), 0)
+ self.assertTrue(target.exists())
+ self.assertEqual(script.main([*args, '--apply']), 0)
+ self.assertFalse(target.exists())
+ self.assertFalse(evidence.exists())
+ self.assertEqual((report / 'originals/MIDV-377.jpg').read_bytes(), self.before)
+ self.assertEqual(official.read_bytes(), self.before)
+ self.assertEqual(script.main([*args, '--apply']), 0)
+ result = json.loads((report / 'manifest.json').read_text(encoding='utf-8'))
+ self.assertEqual((result['matched'], result['quarantined']), (0, 1))
+
+ def test_quarantine_collision_does_not_partially_move_an_item(self):
+ covers = self.root / 'generated/covers'
+ covers.mkdir(parents=True)
+ target = covers / 'MIDV-377.jpg'
+ target.write_bytes(self.before)
+ evidence = target.with_suffix('.scraping.json')
+ evidence.write_text(json.dumps({'source': 'javdb'}), encoding='utf-8')
+ report = self.root / 'review/quarantine'
+ (report / 'originals').mkdir(parents=True)
+ (report / 'originals/MIDV-377.scraping.json').write_text('{}', encoding='utf-8')
+ with self.assertRaises(FileExistsError):
+ batch_script('quarantine_javdb_covers').main([
+ '--data-root', str(self.root), '--report', str(report), '--apply'])
+ self.assertEqual(target.read_bytes(), self.before)
+ self.assertTrue(evidence.exists())
+
def test_audit_keeps_manual_frames_and_reports_unknown_provenance(self):
manual = jav_poster_crop.manual_record(800, 540, {'x0': 420, 'y0': 0, 'x1': 800, 'y1': 540})
jav_poster_crop.write_sidecar(self.cover, manual)
diff --git a/tests/test_javdb_public.py b/tests/test_javdb_public.py
index 4449ce92a..37f304452 100644
--- a/tests/test_javdb_public.py
+++ b/tests/test_javdb_public.py
@@ -275,38 +275,35 @@ class PublicProtocolTests(unittest.TestCase):
class DetailProjectionTests(unittest.TestCase):
- def test_source_cover_route_reuses_detail_image_bytes_and_checks_code(self):
+ def test_official_thumbnail_does_not_enable_a_javdb_cover(self):
+ from peach.web_jav_detail import _artwork_fields
+ source = {'id': 'Exact', 'cover_url': 'https://c0.jdbstatic.com/a.jpg',
+ 'thumb_url': 'https://pics.dmm.co.jp/a.jpg'}
+ with patch('peach.web_jav_detail._preview_fields', return_value={}):
+ result = _artwork_fields(self.contract, 'MIDA-085', {}, source)
+ self.assertFalse(result['source_image'])
+ self.assertEqual(result['cover_url'], '')
+
+ def test_explicit_source_evidence_reuses_detail_image_bytes(self):
import base64
- from fastapi import FastAPI
- from fastapi.testclient import TestClient
- from peach.routes_auth import require_auth
- from peach.routes_media import jav_cover
- app = FastAPI()
- app.state.web_contract = self.contract
- app.dependency_overrides[require_auth] = lambda: {}
- app.add_api_route('/jav-cover', jav_cover, methods=['GET', 'HEAD'])
original = io.BytesIO()
Image.new('RGB', (800, 438), '#777').save(original, 'JPEG')
data_url = 'data:image/jpeg;base64,' + base64.b64encode(original.getvalue()).decode('ascii')
with patch('peach.web_jav_detail.PUBLIC_API.image', return_value=data_url) as fetch:
detail = q_jav_image(self.contract, {'code': 'MIDA-085'})
- with TestClient(app) as client, patch('peach.web_jav_detail.PUBLIC_API.image', side_effect=AssertionError('缓存命中不请求来源')):
- response = client.get('/jav-cover?code=MIDA-085')
- self.assertEqual(response.status_code, 200)
- self.assertEqual(response.content, original.getvalue())
- self.assertEqual(response.headers['content-type'], 'image/jpeg')
- self.assertEqual(client.head('/jav-cover?code=MIDA-085').content, b'')
- self.assertEqual(client.get('/jav-cover?code=../secret').status_code, 400)
+ with patch('peach.web_jav_detail.PUBLIC_API.image', side_effect=AssertionError('缓存命中不请求来源')):
+ self.assertEqual(q_jav_image(self.contract, {'code': 'MIDA-085'})['image'], data_url)
self.assertTrue(detail['ok'])
fetch.assert_called_once()
def test_source_cover_url_projects_known_work_without_changing_source_fields(self):
from peach.web_jav_detail import discovery_cover_url
item = {'code': 'MIDA-085', 'cover_url': None, 'link': 'https://javdb.com/v/Exact'}
- self.assertEqual(discovery_cover_url(item, {}), '/jav-cover?code=MIDA-085')
+ self.assertEqual(discovery_cover_url(item, {}), '')
self.assertIsNone(item['cover_url'])
- self.assertEqual(discovery_cover_url({**item, 'cover_url': 'https://c0.jdbstatic.com/a.jpg'}, {}), 'https://c0.jdbstatic.com/a.jpg')
- self.assertEqual(discovery_cover_url({**item, 'link': ''}, {'id': 'Exact'}), '/jav-cover?code=MIDA-085')
+ self.assertEqual(discovery_cover_url({**item, 'cover_url': 'https://c0.jdbstatic.com/a.jpg'}, {}), '')
+ self.assertEqual(discovery_cover_url({**item, 'link': ''}, {'id': 'Exact'}), '')
+ self.assertEqual(discovery_cover_url({**item, 'cover_url': 'https://pics.dmm.co.jp/a.jpg'}, {}), 'https://pics.dmm.co.jp/a.jpg')
self.assertEqual(discovery_cover_url({**item, 'link': 'https://private.invalid/'}, {}), '')
def test_saved_identity_fetches_actor_portrait_without_code_search(self):
diff --git a/tests/test_metadata_library.py b/tests/test_metadata_library.py
index 1d4abb140..93098ed38 100644
--- a/tests/test_metadata_library.py
+++ b/tests/test_metadata_library.py
@@ -937,7 +937,7 @@ class LibraryNfoTests(unittest.TestCase):
provider.transport = Mock()
provider.community = Mock(return_value=[('javdb', {'cover_urls': ['https://c0.jdbstatic.com/covers/x.jpg']})])
small = (Candidate('image.mgstage.com', 'https://image.mgstage.com/pf_o1.jpg'), (300, 200), b'small')
- large = (Candidate('c0.jdbstatic.com', 'https://c0.jdbstatic.com/covers/x.jpg'), (1200, 800), b'large',
+ large = (Candidate('img.javbus.com', 'https://img.javbus.com/covers/x.jpg'), (1200, 800), b'large',
('javdb', 'mgstage'))
covers = self.root / 'covers'
with patch('peach.jav_cover_fetch.best_cover', return_value=small) as official, \
@@ -954,7 +954,7 @@ class LibraryNfoTests(unittest.TestCase):
self.assertTrue(provider.cover('ORETD-616', covers))
self.assertEqual((covers / 'ORETD-616.jpg').read_bytes(), b'small')
- lone = (Candidate('c0.jdbstatic.com', 'https://c0.jdbstatic.com/covers/y.jpg'), (800, 538), b'lone', ())
+ lone = (Candidate('img.javbus.com', 'https://img.javbus.com/covers/y.jpg'), (800, 538), b'lone', ())
with patch('peach.jav_cover_fetch.best_cover', side_effect=NotFound('所有渠道都没有候选')), \
patch('peach.community_catalog.verified_cover', return_value=lone):
self.assertTrue(provider.cover('IPX-060', covers))
diff --git a/web/css/19-immersive.css b/web/css/19-immersive.css
index 4c1d6f30b..ac2d5953f 100644
--- a/web/css/19-immersive.css
+++ b/web/css/19-immersive.css
@@ -112,5 +112,5 @@
/* 整张放进卡片:两侧留白透出 `.pic::before` 那层模糊底。 */
.poster.cover.whole{object-fit:contain;object-position:50% 50%}
/* 整幅画面按人脸取景并铺满卡片。 */
-.poster.cover.front[data-frame="still"],.poster.cover.whole[data-frame="still"]{
+.poster.cover.front[data-frame="still"]{
object-fit:cover;object-position:var(--cover-x,50%) var(--cover-y,50%)}
diff --git a/web/dist/peach-app.js b/web/dist/peach-app.js
index 86392a10c..546f0a04b 100644
--- a/web/dist/peach-app.js
+++ b/web/dist/peach-app.js
@@ -74424,7 +74424,7 @@ function gK({ item: e, image: t }) {
},
children: "重试"
}) : void 0,
- children: i.data?.message || (i.error ? "图片未取得" : "暂无封面")
+ children: i.data?.message || (i.error ? "图片未取得" : "封面未取得")
})
});
}