Peach: Local-first video metadata scraper and media manager.
feat(metadata): 补女优资料在 minnano-av 上找不到她时由 avbase 女优页补位
longmeidao627152be942f
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex caa7ae1ca..6f1b98a6b 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -21,6 +21,10 @@ Removed / Fixed / Security。## [未发布]+### 新增++- **采集**:补女优资料时,如果 minnano-av 上找不到她,就从 avbase 女优页补生日、身高、三围、罩杯,以及站上填了的出身地、血型和趣味,人物页页头不再空着。已有 minnano-av 资料或人工资料的只核对、不覆盖;她之后出现在 minnano-av 上,资料会换成那边的。+### 变更- **关注**:订阅源拉取改为每半小时看一次到期的订阅,各订阅错开时间拉取,每个仍是 6 小时一次;以前同时添加的订阅总在同一时刻一起拉,有时还会隔一轮才拉到。diff --git a/docs/REUSE.md b/docs/REUSE.mdindex 4e11dc0ad..4e0140bdb 100644--- a/docs/REUSE.md+++ b/docs/REUSE.md@@ -439,6 +439,8 @@ Babepedia 主图库和 Vixen 网络公开 GraphQL 的字段核对 [Stash Communi公司资料复用 `entity.metadata_json`、HTTPX、官网公司概要表和定义列表与标识采集器。`company_profiles` 只有通用判据(概要标签、服务条款订立方原句),按字段保存出处、原文和批次,只填空;一页多张公司表时按法人名对实体名与别名选表,公司成立与品牌启动分列,集团与分工关系不由代码推导,只作候选。`harvest_company_profiles.py` 默认产复核 JSON,`--input --apply --backup --batch auto:company-profile@…` 按缓存页重放判据后落库,清单里的别名、链接与第三方页面不写。`revert_auto_landing.py` 可撤回同批公司字段及标识,无新增依赖。+avbase 女优页补位(ADR-0136)参考 OpenAver `ae7fc235`(MIT)的来源选择,只借用「女优页地址加 `__NEXT_DATA__` 字段」这个做法,没有复制源码。取页复用 `MinnanoPages` 的缓存、冷却与请求上限,传输用 `SourceTransport`。实测 httpx 带 Chrome UA 就能取到页,所以没有照上游改走 curl_cffi。解析与 `sources/avbase.py` 的作品页共用同一个 `__NEXT_DATA__` 读法,名字锚定复用 `performer_alias_followup` 的 `search_keys`、`match_key`、`rejection`,资料写入复用 `performer_profiles.write_profile`。无新增依赖。+Babepedia 资料采集复用 `babepedia_page` 的主名与别名交叉核对、`performer_profiles` 的整行来源保护和 `performer_alias_followup.land` 的占用检查。英文艺名只在明确允许的来源中登记,单个罗马字词只在它就是页面主名时登记,JAV 后继保留自己的名字策略。真人账号与出演者共用资料表、别名分组与头像候选;保留账号角色及旧地址,与已登记账号说法不一的链接只进回执。出演者主图经 `avatar_followup.match_gallery` 与单人作品封面比上脸才补空头像,账号与本人档案图库仅作候选,推荐人物排除,单图失败继续其余有限候选;候选表只认判词「命中」的档案地址。`harvest_western_profiles.py` 每轮至多 16 位,计划、备份、回执与批次撤回齐备;既有资料和头像保留。真实只读复核取得 Ria Kurumi 等 16 份匹配档案,无新增依赖。身份分类复用 `entity`、来源关系、`field_owners`、SQLite 迁移与版本触发器。职业与账号角色使用独立多值断言,跨角色身份使用可撤回关系;[Schema.org Person](https://schema.org/Person) 的 `hasOccupation`、`sameAs` 与 [Role](https://schema.org/Role) 提供语义参照,2026-10-06 已核对官方文档。实现不复制外部解析器或新增依赖。目录推断和搜索命中只存候选,公开来源事实与用户复核分别保留状态。diff --git a/docs/SOURCING.md b/docs/SOURCING.mdindex 8c6f0f10c..caf4b504f 100644--- a/docs/SOURCING.md+++ b/docs/SOURCING.md@@ -356,8 +356,9 @@ minnano-av 也在 Cloudflare 后面,但对 HTTP 客户端是间歇拦截:202`resync_performer_agency.py` 都经 `SourceTransport` 取 minnano-av,共用同一份冷却与固定状态。2026-10-01 实测:临时 profile 的 Chrome 取检索页,跟到资料页 `actress695633.html`,资料表解析正常。-候选备援(未接入):OpenAver 0.16.13 拔掉 minnano-av 后按 xcity → 维基 → graphis 逐字段合并女优资料;只作 minnano-av-取不到时的补位,不替换它,因为它是目前唯一能从名字走到事务所名册的入口(调研见 ADR-0092)。+资料的补位已接入 avbase 女优页(ADR-0136,见下文「女优别名与资料的自动登记」)。它只在 minnano-av 上没有她时写,+不替换 minnano-av,因为后者是目前唯一能从名字走到事务所名册的入口。OpenAver 另有 xcity → 维基 → graphis 的逐字段合并,+仍未接入(调研见 ADR-0092)。### JavArchive@@ -604,12 +605,17 @@ Instagram 的独立用户登录会话未取得,自动适配器不进正式依- `サイズ` 形如 `T156 / B86( Eカップ ) / W58 / H85 / S`,数字就是厘米;末尾那个 `S` 含义未取得,只留原文。资料没填的页没有生年月日、血液型、出身地那几行。- `ブログ` 那格显示 `http://`、`href` 是 `https://`,取 `href`。站内头像 `/p_actress_125_125/` 只有 125×125。+ - avbase 女优页 `/talents/<名字>`(ADR-0136):直连 httpx 带 Chrome UA 就回 200,名字站上没有回 404。+ `__NEXT_DATA__` 的 `pageProps.talent.actors[]` 是站上归到她名下的名义,资料取 `meta.basic_info`;+ `actors[].meta.fanza` 是另一份可能过时的 FANZA 资料,只留原文。没有出道年、活动期、博客与官网。一页+ 650~750 KB,缓存只存 `compact_talent_page` 裁出的 `talent` 与同名列表。冷却与作品采集共用 `avbase` 那一份。+ 它写的行不算 minnano-av 的「未过期」,minnano-av 取到就整行换掉。- avwikidb.com 直连 httpx 带 Chrome UA 就回整页(登记在 `scraping_access.SOURCES`,拒绝访问停 6 小时)。作品页`/work/<番号>/` 头里的 JSON-LD `Movie.actor` 每项给 `name`、罗马字 `alternateName` 与 `/actor/<编号>/`,编号从这里取。女优页 JSON-LD `Person.alternateName` 前两项是读音与罗马字,后面混着站上归并的其他名义(含男优名),不收。三围只在正文「身長・スリーサイズ」一格(`T148 B83(B) W55 H85`)。女优图是 DMM `actjpgs` 的125×125,不作头像;站内搜索是 `⌘K` 弹层,地址形态未取得。- - 出生日期与身高和 minnano-av 不一致时,只记进外部编号的 `metadata_json.conflicts`。+ - 出生日期与身高和资料表那一行(minnano-av 或 avbase 写的)不一致时,只记进外部编号的 `metadata_json.conflicts`。## 番号目录、创作者与水印diff --git a/docs/adr/0136-avbase-talent-page-backs-up-performer-profiles.md b/docs/adr/0136-avbase-talent-page-backs-up-performer-profiles.mdnew file mode 100644index 000000000..9fa69414e--- /dev/null+++ b/docs/adr/0136-avbase-talent-page-backs-up-performer-profiles.md@@ -0,0 +1,65 @@+# ADR-0136:avbase 女优页给补女优资料后继补位++- 状态:Accepted+- 日期:2026-11-04+- 关系:扩展 [ADR-0067](0067-performer-profiles-land-in-their-own-table.md) 的来源链;落库口径照 [ADR-0052](0052-entity-results-land-when-evidence-is-sufficient.md)++## 背景++补女优资料后继(ADR-0067)的资料只来自 minnano-av。账本里 939 位有 JAV 目录站编号的女优中,有 104 位+带作品却没有资料行,其中多数在 minnano-av 上检索不到,页头的生日、身材那几项因此空着。++OpenAver 0.17.13([ae7fc235](https://github.com/slive777/OpenAver/commit/ae7fc235),MIT)把 avbase 的+女优页当成女优资料的备援。2026-11-04 实测:++- `https://www.avbase.net/talents/<名字>` 按名字进,是 Next.js 页。`__NEXT_DATA__` 的 `props.pageProps.talent`+ 里,`actors[]` 是站上归到这位名下的全部名义,`meta.basic_info` 是社区编辑的资料(生日、身高、三围、罩杯、+ 出身地、趣味、血型)。名字站上没有时回 HTTP 404。+- Peach 现有的 `SourceTransport`(Chrome UA、httpx)直接拿到 200;OpenAver 改用 curl_cffi 绕 403 的做法这里用不上。+ 每 4 秒一页、共 32 页,没有出现 403、429 或验证页。+- 一页 650~750 KB,大部分是嵌在里面的作品表。+- 从真实账本只读抽 30 位:没有资料的 20 位命中 13 位(3 位靠别名命中),已有 minnano-av 资料的 10 位命中 9 位。+ 命中的那 13 位,生日、身高、三围、罩杯都有值。已有资料的 9 位里,生日 8 位与 minnano-av 一致。++## 决策++**一、avbase 接在 minnano-av 后面、avwikidb 前面。** avbase 写下的资料,avwikidb 照样拿来交叉核对。++**二、身份按名字核对,同名不挑。** 名字链里收得下的写法(`performer_alias_followup.search_keys`)至多问三个,+404 就换下一个写法。页上 `actors[]` 的名义里要有账本里她的名字(`match_key` 折叠后相等,`rejection` 不收的写法+不作锚点),`pageProps.talents` 里只能有一位。满足这两条,就把 talent 编号绑成+`entity_external_ref(provider='avbase', external_kind='performer')`。编号已属另一条实体时只记「占用」,资料也不写。++**三、只填空。** 资料表里没有她,或者那一行原本就是 avbase 写的,才按这页整行写 `performer_profile`;+`source_url` 记女优页地址。minnano-av 写的行和人写的行都不动,生日或身高与 avbase 不一致时,两边的值记进这条编号的+`metadata_json.conflicts`。++**四、不挡 minnano-av。** avbase 写的那一行不算 minnano-av 的「未过期」:下一轮照问 minnano-av,取到就整行换掉。+avbase 自己按编号的 `last_synced_at` 计 30 天。++**五、只取 `basic_info`。** 同一页的 `actors[].meta.fanza` 另有一份官方资料,两份可能对不上(三上悠亜:`basic_info` 是+B90/H,FANZA 是 B84/F 的旧值)。那份不拿来补空格,原样留进 `raw`。站上没有出道年、活动期、出道作、博客和官网,+这几列留空。`primary.image_url` 是 DMM `actjpgs` 的 125×125,和 avwikidb 是同一张图,不当头像候选(ADR-0067 第六条)。++**六、缓存裁剪,限流与作品采集共用。** 这一站单独按 3 秒间隔,一条后继至多发三次请求。冷却记在 `avbase` 那一份里,+和作品采集共用:撞上 403、429 或验证页,两边在冷却期内都不再问。200 的页先经 `avbase.compact_talent_page` 裁成+`talent` 和同名列表两项再进缓存,单页从约 750 KB 降到约 2 KB。认不出结构的 200 页不进缓存,也不算问到了;+一个写法都没问到时记「未取得」,按 24 小时重试。++**七、判据版本加一。** `RULE` 升到 3,avbase 编号也进指纹,存量女优按每轮 16 条的名额各再问一次。落库与撤回照+ADR-0067:批次号仍是 `auto:performer-profile@<任务行 id>`,`revert_auto_landing.py --source auto:performer-profile`+会把资料行连同 avbase 编号一起撤回。++## 后果++- 绑了 avbase 编号的女优算 JAV 女优(`entry_links.JAV_DIRECTORIES` 本来就列了 `avbase`)。原先在哪个 JAV 目录站都+ 没有编号的,绑上之后人物页会多出 MISSAV 入口,同一轮后继也会去 javdb 搜她,搜到才多出 JavDB 入口。+- 每位存量女优在重派时多出一到三次 avbase 请求。+- 判词写进 `generated/performer-profile-landing.csv`,`site` 列为 `avbase`。++## 未决++- `actors[]` 名义和 `meta.sns`(X、Instagram、TikTok 帐号,命中的 13 位里有 12 位给了)都只留在 `raw` 与编号的+ `metadata_json` 里,还没接到别名(ADR-0055 要两站互证)和本人链接。+- `talent.frozen` 的含义未取得,现在不作判据。+- 女优页的作品表只给第一页的 30 部,没有用作身份证据。diff --git a/src/peach/performer_alias_followup.py b/src/peach/performer_alias_followup.pyindex 220c3c659..78719e461 100644--- a/src/peach/performer_alias_followup.py+++ b/src/peach/performer_alias_followup.py@@ -367,7 +367,8 @@ class MinnanoPages:self._hold(refused=True)raise Blocked("来源拒绝访问或要求机器人验证")if response.status != 200:- raise Unavailable(f"HTTP {response.status}")+ # avbase 女优页按名字拼地址,站上没有这个名字回 404:那是「问过、没有」,不是没取到。+ raise (Missing if response.status == 404 else Unavailable)(f"HTTP {response.status}")final_url = response.url or urlpath.parent.mkdir(parents=True, exist_ok=True)path.write_text(json.dumps({"url": url, "final_url": final_url, "body": body},diff --git a/src/peach/performer_profile_followup.py b/src/peach/performer_profile_followup.pyindex 841a3c072..b353f480c 100644--- a/src/peach/performer_profile_followup.py+++ b/src/peach/performer_profile_followup.py@@ -1,5 +1,5 @@"""补女优资料:minnano-av 的资料表整张落进 `performer_profile`,再绑 avwikidb 的女优编号(ADR-0067)-与 javdb 的演员 id(ADR-0081)。+与 javdb 的演员 id(ADR-0081);minnano-av 上没有她时由 avbase 女优页补位(ADR-0136)。判据是确定性的,按 ADR-0052 直接落库,不产生候选:@@ -9,6 +9,11 @@检索进来的页核对通过时,顺手把这个编号登记成她的 minnano-av 外部编号,下一轮直接进。3. **整行写入**:资料按 `minnano_av.profile` 规整后写 `performer_profile`,读不出的列留空。这一行是人写的(`source` 不以 `auto:` 开头)就不动。+3a. **avbase 只补位**:拿名字链里收得下的写法进 `/talents/<名字>`(至多 `alias.MAX_KEYS` 个,404 接着下一个),+ 页上 `actors[]` 的名义里有账本里她的名字、同名 talent 只有一位才算,绑上 talent 编号。资料表里没有她、+ 或那一行本来就是 avbase 写的,才按这页整行写;minnano-av 或人写的那一行在,只把出生日期与身高的不一致+ 记进这条编号的 `metadata_json.conflicts`。avbase 写的一行不算 minnano-av 的「未过期」,下一轮照问 minnano-av,+ 它取到了就整行换掉。4. **avwikidb 只从她自己的作品进**:取她非 FC2 作品的番号(出演人数少的在前,至多`MAX_WORKS` 部),读作品页的出演表;名字与账本里她的某个名字折叠后相等,或罗马字与minnano-av 上她的罗马字按词相等,恰好一位对得上才算。对不上、对上两位都不绑,写「未命中」,@@ -26,11 +31,12 @@`scripts/revert_auto_landing.py --source auto:performer-profile` 整批撤回。判词逐条写进`generated/performer-profile-landing.csv`。-取页复用补别名后继的 `MinnanoPages`:成功页落盘缓存,minnano-av 与 avwikidb 按 3 秒间隔,撞上+取页复用补别名后继的 `MinnanoPages`:成功页落盘缓存,minnano-av、avbase 与 avwikidb 按 3 秒间隔,撞上429、403 或机器人验证就记进 `scraping_access` 的冷却,本轮与之后在冷却期内都不再问,结论写「未取得」。三站的请求都走 `SourceTransport`,连接方式跟采集设置里那一站的设置走;minnano-av 的 HTTP 客户端被拦时先由本机浏览器验证一页,取到就固定走浏览器(`scraping_access.SOURCES` 的 `browser_fallback`)。javdb 按出口 IP计配额、一封 3~7 天,这一站单独 5 秒间隔(与目录采集同一档),冷却记录与作品采集共用 `javdb` 那一份。+avbase 的冷却也与作品采集共用 `avbase` 那一份;女优页缓存只存 `avbase.compact_talent_page` 裁过的那两项。多久再派一次由 `Attempts` 的记号决定:写成了的保 `REFRESH`(30 天),到期再派一次,资料页按取回时间过了 30 天才重取;有一站「未取得」的只保 `RETRY_UNFETCHED`;哪一站都没对上的等名字链、@@ -38,6 +44,7 @@"""from __future__ import annotations+import dataclassesimport hashlibimport jsonimport sqlite3@@ -52,14 +59,18 @@ from . import performer_alias_followup as aliasfrom .followups import Attempts, Followup, FollowupType, attempts_root, registerfrom .performer_profiles import fetched_at as profile_fetched_at, read_profile, write_profilefrom .scripting import HostLimiter+from .sources import avbaseTASK_KEY = "performer-profile"TASK_LABEL = "补女优资料"SOURCE = "auto:performer-profile"-MINNANO, AVWIKIDB, JAVDB = alias.MINNANO, "avwikidb", "javdb"+MINNANO, AVBASE, AVWIKIDB, JAVDB = alias.MINNANO, "avbase", "avwikidb", "javdb"#: javdb 的主机间隔,与 `harvest_directory_links.SOURCE_INTERVAL` 同一档(docs/SOURCING.md)。JAVDB_INTERVAL = 5.0_JAVDB_LIMITER = HostLimiter({"javdb.com": JAVDB_INTERVAL})+#: avbase 女优页与 minnano-av、avwikidb 同一档间隔(2026-11-04 实测 4 秒一页、32 页没撞墙)。+AVBASE_INTERVAL = 3.0+_AVBASE_LIMITER = HostLimiter({"avbase.net": AVBASE_INTERVAL})#: 每轮处理任务给存量的名额。一条最多问两站五六次、每次隔 3 秒,给多了会把头像与厂牌挤到很后面;#: 九百多位女优按这个名额要一个多月的处理轮次才轮遍,存量本来就不赶。@@ -68,8 +79,8 @@ STOCK_SHARE = 16REFRESH = 30 * 86400RETRY_UNFETCHED = alias.RETRY_UNFETCHED#: 入口判据的版本,进指纹。换了判据就加一,跑过的女优按新判据各再问一次。-#: 2 是 javdb 那一站(ADR-0081)。-RULE = 2+#: 2 是 javdb 那一站(ADR-0081),3 是 avbase 补位(ADR-0136)。+RULE = 3#: avwikidb 最多翻她几部作品的出演表。MAX_WORKS = 3@@ -121,11 +132,12 @@ def has_entry(connection: sqlite3.Connection, entity_id: int) -> bool:def fingerprint(connection: sqlite3.Connection, entity_id: int) -> str:- """会让结论变的量:名字链、三站编号、avwikidb 的入口番号与判据版本。"""+ """会让结论变的量:名字链、四站编号、avwikidb 的入口番号与判据版本。"""names = alias._names(connection, entity_id) or ("", [])keys = sorted({alias.match_key(name) for name in [names[0], *names[1]] if name})parts = [keys, _refs(connection, entity_id, MINNANO), _refs(connection, entity_id, AVWIKIDB),- work_codes(connection, entity_id), _refs(connection, entity_id, JAVDB), f"r{RULE}"]+ work_codes(connection, entity_id), _refs(connection, entity_id, JAVDB),+ _refs(connection, entity_id, AVBASE), f"r{RULE}"]raw = json.dumps(parts, ensure_ascii=False)return hashlib.sha1(raw.encode("utf-8")).hexdigest()[:16]@@ -175,18 +187,42 @@ def stock(connection: sqlite3.Connection, attempts, *, limit: int, skip=()) -> ldef open_sites(contract) -> dict:- """三站的取页器。缓存目录与补别名后继共用,同一页不问第二次。测试把这一步整个换掉。"""+ """四站的取页器。缓存目录与补别名后继共用,同一页不问第二次。测试把这一步整个换掉。"""from .scraping_access import SourceTransportcache = Path(contract.candidate_root) / "provider-cache"cooldown = alias._cooldown_root(contract)return {MINNANO: alias.MinnanoPages(cache / "minnano-av-pages", cooldown, max_age=REFRESH),+ AVBASE: alias.MinnanoPages(cache / "avbase-talent-pages", cooldown,+ TalentPages(SourceTransport(cooldown)),+ limiter=_AVBASE_LIMITER, max_requests=alias.MAX_KEYS,+ source=AVBASE, max_age=REFRESH),AVWIKIDB: avwikidb_pages(contract),JAVDB: alias.MinnanoPages(cache / "javdb-pages", cooldown, SourceTransport(cooldown),limiter=_JAVDB_LIMITER, max_requests=alias.MAX_KEYS,source=JAVDB, max_age=REFRESH)}+class TalentPages:+ """avbase 女优页的传输:200 的页先裁成 `avbase.compact_talent_page`,缓存里只落那两项。++ 别的属性(`hold`、`close`)照转给里面那个传输,冷却仍按它这次走的那条路记。+ """++ def __init__(self, transport):+ self.transport = transport++ def __call__(self, request, timeout: float, max_bytes: int):+ response = self.transport(request, timeout, max_bytes)+ if response.status != 200:+ return response+ page = avbase.compact_talent_page(response.body.decode("utf-8", "replace"))+ return dataclasses.replace(response, body=page.encode("utf-8"))++ def __getattr__(self, name: str):+ return getattr(self.transport, name)++def avwikidb_pages(contract):"""avwikidb 的取页器。补头像后继读单人作品页也用它:同一份缓存、同一套冷却。"""from .scraping_access import SourceTransport@@ -242,16 +278,29 @@ def bind(connection: sqlite3.Connection, entity_id: int, provider: str, externalreturn BIND-def conflicts(profile: dict | None, actor: dict) -> list[dict]:- """两站都有值却不一致的列:[{"field", "minnano-av", "avwikidb"}]。任一边没有值不算冲突。"""+def conflicts(profile: dict | None, actor: dict, *, held: str = MINNANO,+ site: str = AVWIKIDB) -> list[dict]:+ """两站都有值却不一致的列:[{"field", <资料表那一站>, <这一站>}]。任一边没有值不算冲突。"""found = []for column, key in CROSS_CHECKED:mine, theirs = (profile or {}).get(column), actor.get(key)if mine is not None and theirs is not None and mine != theirs:- found.append({"field": column, MINNANO: mine, AVWIKIDB: theirs})+ found.append({"field": column, held: mine, site: theirs})return found+def _conflict_rows(url: str, site: str, found: list[dict], batch: str, base: dict) -> list[dict]:+ return [{**base, "site": site, "page": url, "action": CONFLICT, "batch": batch,+ "detail": f"{item['field']}:" + ",".join(+ f"{name} {value}" for name, value in item.items() if name != "field")+ + ";资料表不改"} for item in found]+++def _avbase_row(row: dict | None) -> bool:+ """资料表这一行是不是 avbase 补位写的:它不算 minnano-av 的资料,下一轮照问 minnano-av。"""+ return str((row or {}).get("source_url") or "").startswith(avbase.talent_url(""))++def _alive(connection: sqlite3.Connection, entity_id: int, canonical: str) -> bool:current = alias._names(connection, entity_id)return current is not None and current[0] == canonical@@ -294,9 +343,12 @@ def _run(contract, key: str, handle) -> dict:with contract.database.read_connection() as connection:names = alias._names(connection, entity_id)state = {"minnano": _refs(connection, entity_id, MINNANO),+ "avbase": _refs(connection, entity_id, AVBASE),+ "avbase_synced": _synced(connection, entity_id, AVBASE),"avwikidb": _refs(connection, entity_id, AVWIKIDB),"codes": work_codes(connection, entity_id),- "fetched": profile_fetched_at(connection, entity_id)}+ "fetched": None if _avbase_row(read_profile(connection, entity_id))+ else profile_fetched_at(connection, entity_id)}if names is None:return {"outcome": "档案已不存在"}canonical = names[0]@@ -308,7 +360,9 @@ def _run(contract, key: str, handle) -> dict:rows: list[dict] = []try:_visit_minnano(contract, sites[MINNANO], entity_id, names, state, batch, base, reports, rows)- conflict_count = _visit_avwikidb(contract, sites[AVWIKIDB], entity_id, names, state, batch,+ conflict_count = _visit_avbase(contract, sites[AVBASE], entity_id, names, state, batch,+ base, reports, rows)+ conflict_count += _visit_avwikidb(contract, sites[AVWIKIDB], entity_id, names, state, batch,base, reports, rows)_visit_javdb(contract, sites[JAVDB], entity_id, names, batch, base, reports, rows)finally:@@ -388,6 +442,95 @@ def _visit_minnano(contract, pages, entity_id: int, names, state: dict, batch: s+ ("" if bound == BIND else " 已属另一条实体")})+def _visit_avbase(contract, pages, entity_id: int, names, state: dict, batch: str, base: dict,+ reports: dict, rows: list) -> int:+ """avbase:按名字进女优页,名义里有她就绑 talent 编号。返回记下的冲突条数。++ 资料表里没有她、或那一行是 avbase 写的,整行按这页写;别的来源那一行在,只记冲突。+ """+ canonical, aliases = names+ if state["avbase"] and _fresh(state["avbase_synced"]):+ reports[AVBASE] = "资料未过期"+ rows.append({**base, "site": AVBASE, "page": "", "action": FRESH,+ "detail": "编号已绑,女优页取回不到 30 天"})+ return 0+ keys = alias.search_keys(canonical, aliases)[:alias.MAX_KEYS]+ if not keys:+ reports[AVBASE] = f"{MISS}:没有能拿去检索的名字"+ return 0+ found, notes, asked = _find_talent(pages, keys, {+ alias.match_key(name) for name in [canonical, *aliases] if not alias.rejection(name)})+ if found is None:+ detail = ";".join(notes)+ if not asked:+ # 一个写法都没问成:站那天没让进或页面认不出,不是站上没有她,不结算指纹。+ reports[AVBASE] = f"{UNFETCHED}:{detail}"+ return 0+ reports[AVBASE] = f"{MISS}:{detail}"+ rows.append({**base, "site": AVBASE, "page": "", "action": MISS, "detail": detail})+ return 0+ url, parsed = found+ with contract.database.write_transaction() as connection:+ if not _alive(connection, entity_id, canonical):+ reports[AVBASE] = f"{MISS}:数据库里这条实体已经变了"+ return 0+ row = read_profile(connection, entity_id)+ other = row is not None and not _avbase_row(row)+ found_conflicts = conflicts(row, parsed, site=AVBASE) if other else []+ verdict = bind(connection, entity_id, AVBASE, parsed["talent_id"],+ {"source": SOURCE, "batch": batch, "names": parsed["names"],+ **({"conflicts": found_conflicts} if found_conflicts else {})})+ wrote = (verdict == BIND and not other+ and write_profile(connection, entity_id, parsed, source=batch, source_url=url))+ reports[AVBASE] = f"命中 {url}" if verdict == BIND else f"{TAKEN}:talent {parsed['talent_id']} 已属另一条实体"+ rows.append({**base, "site": AVBASE, "page": url, "action": verdict,+ "batch": batch if verdict == BIND else "",+ "detail": f"talent {parsed['talent_id']} 的名义里有她" if verdict == BIND+ else "编号已属另一条实体,资料不写"})+ if wrote:+ state["profile"], state["profile_site"] = parsed, AVBASE+ rows.append({**base, "site": AVBASE, "page": url, "action": WRITE, "batch": batch,+ "detail": "资料表里没有别的来源,按这页补写"})+ if verdict != BIND:+ return 0+ rows.extend(_conflict_rows(url, AVBASE, found_conflicts, batch, base))+ return len(found_conflicts)+++def _find_talent(pages, keys: list[str], mine: set[str]) -> tuple[tuple[str, dict] | None, list[str], int]:+ """((页面地址, 资料) 或 None, 说明, 问成了几个写法)。名义里有她、同名只有一位的第一页算数。++ 404 是站上没有这个写法,算问成了;认不出的 200 页丢掉缓存、不算问成,下一轮再问。+ """+ notes, asked = [], 0+ for key in keys:+ url = avbase.talent_url(key)+ try:+ _final, html = pages.get(url)+ except alias.Blocked as error:+ return None, [f"{key} {error}"], 0+ except alias.Missing:+ asked += 1+ notes.append(f"站上没有 {key}")+ continue+ except alias.Unavailable as error:+ notes.append(f"{key} {error}")+ continue+ parsed = avbase.talent_profile(html)+ if parsed is None:+ pages.forget(url)+ notes.append(f"{key} {avbase.UNRECOGNISED}")+ continue+ asked += 1+ if not alias._anchored(parsed["names"], mine):+ notes.append(f"{key} 的页上没有数据库里她的名字:{'、'.join(parsed['names'])}")+ elif parsed["talents"] > 1:+ notes.append(f"站上叫 {key} 的不止一位")+ else:+ return (url, parsed), notes, asked+ return None, notes, asked++def _visit_avwikidb(contract, pages, entity_id: int, names, state: dict, batch: str, base: dict,reports: dict, rows: list) -> int:"""avwikidb:从她的作品页定编号,再读女优页做交叉核对。返回冲突条数。"""@@ -417,7 +560,8 @@ def _visit_avwikidb(contract, pages, entity_id: int, names, state: dict, batch:actor = {}if actor and actor.get("id") != actor_ref:actor = {}- found = conflicts(state.get("profile") or profile, actor)+ held = state.get("profile_site") or (AVBASE if _avbase_row(profile) else MINNANO)+ found = conflicts(state.get("profile") or profile, actor, held=held)with contract.database.write_transaction() as connection:if not _alive(connection, entity_id, canonical):reports[AVWIKIDB] = f"{MISS}:数据库里这条实体已经变了"@@ -431,10 +575,15 @@ def _avwikidb_state(contract, entity_id: int, refs: list[str]) -> tuple[dict | N"""(minnano-av 资料, 已绑编号的资料是否未过期)。"""with contract.database.read_connection() as connection:profile = read_profile(connection, entity_id)- synced = connection.execute(- "SELECT max(last_synced_at) FROM entity_external_ref WHERE entity_id=? AND provider=?",- (entity_id, AVWIKIDB)).fetchone()[0] if refs else None- return profile, bool(refs) and _fresh(_parse_stamp(synced))+ synced = _synced(connection, entity_id, AVWIKIDB) if refs else None+ return profile, bool(refs) and _fresh(synced)+++def _synced(connection: sqlite3.Connection, entity_id: int, provider: str) -> datetime | None:+ """这一站在她名下的编号最近一次取回资料的时间。"""+ return _parse_stamp(connection.execute(+ "SELECT max(last_synced_at) FROM entity_external_ref WHERE entity_id=? AND provider=?",+ (int(entity_id), provider)).fetchone()[0])def _locate_actor(pages, names, state: dict, profile: dict | None, base: dict, reports: dict,@@ -476,10 +625,7 @@ def _record_bind(url: str, verdict: str, work: str, actor: dict, found: list[dic"detail": detail if verdict == BIND else "编号已属另一条实体"})if verdict != BIND:return 0- for item in found:- rows.append({**base, "site": AVWIKIDB, "page": url, "action": CONFLICT, "batch": batch,- "detail": f"{item['field']}:minnano-av {item[MINNANO]},"- f"avwikidb {item[AVWIKIDB]};资料表不改"})+ rows.extend(_conflict_rows(url, AVWIKIDB, found, batch, base))return len(found)diff --git a/src/peach/sources/avbase.py b/src/peach/sources/avbase.pyindex b67072986..6a3efa890 100644--- a/src/peach/sources/avbase.py+++ b/src/peach/sources/avbase.py@@ -9,6 +9,10 @@它不给时长,`runtime` 留空。拒绝访问时回的是 Cloudflare 验证页,状态码可能是 403,也可能是 200:403 由 `SourceTransport` 整站冷却,200 的那张由 `parse` 认成 `cloudflare_challenge`,经 `holding()`进同一份冷却记录(上限 `scraping_access.SOURCES['avbase']['blocked_pause']`)。++女优页 `/talents/<名字>`(2026-11-04 实测)也是 Next.js 页,`props.pageProps.talent` 是这位:`actors[]`+是站上归到她名下的全部名义,`meta.basic_info` 是社区编辑的资料。补女优资料后继拿它作 minnano-av 的+备援(ADR-0136),解析在 `talent_profile`;名字站上没有时回 404。"""from __future__ import annotations@@ -120,3 +124,98 @@ class AVBaseSource(SiteSource):director=first(info("director")),release_date=listing_date(products[0].get("date") if products else work.get("min_date")),cover_urls=_covers(code, products))+++# -- 女优页 ------------------------------------------------------------------++TALENT_PAGE = "https://www.avbase.net/talents/{name}"+_BLOOD_TYPES = frozenset({"A", "B", "O", "AB"})+_CUP = re.compile(r"[A-Z]{1,2}")+_FIRST_NUMBER = re.compile(r"\d+")+++def talent_url(name: str) -> str:+ return TALENT_PAGE.format(name=urllib.parse.quote(name, safe=""))+++def _page_props(html: str) -> dict | None:+ found = _NEXT_DATA.search(html or "")+ try:+ props = json.loads(found.group(1))["props"]["pageProps"] if found else None+ except (ValueError, KeyError, TypeError):+ return None+ return props if isinstance(props, dict) else None+++def compact_talent_page(html: str) -> str:+ """只留女优页里解析会读到的那两项,再包回 `__NEXT_DATA__`;认不出的页原样交回。++ 整页有 650~750 KB,多半是嵌在里面的作品表,补女优资料的页面缓存按人一页,整页存下来+ 一千位就是六七百 MB。验证页、404 页没有 `talent`,原样交回,撞墙判定照常看得到它们。+ """+ props = _page_props(html)+ if props is None or not isinstance(props.get("talent"), dict):+ return html+ kept = {"talent": props["talent"],+ "talents": [{"id": item.get("id")} for item in props.get("talents") or []+ if isinstance(item, dict)]}+ data = json.dumps({"props": {"pageProps": kept}}, ensure_ascii=False).replace("</", "<\\/")+ return f'<script id="__NEXT_DATA__" type="application/json">{data}</script>'+++def _date(value: object) -> str | None:+ from datetime import date++ text = str(value or "").strip()+ try:+ return date.fromisoformat(text).isoformat() if len(text) == 10 else None+ except ValueError:+ return None+++def _centimetres(value: object) -> int | None:+ """`159`、`159.5` 取整数部分;空串、`0` 与没有数字的都是没填。"""+ found = _FIRST_NUMBER.search(str(value or ""))+ return int(found.group()) or None if found else None+++def _text(value: object) -> str | None:+ return value.strip() or None if isinstance(value, str) else None+++def talent_profile(html: str) -> dict | None:+ """女优页上的资料,列与 `minnano_av.PROFILE_COLUMNS` 对齐;不是女优页返回 None。++ 另有 `talent_id`、`names`(`actors[]` 的名义,加上 `primary` 里不在其中的那个)、`talents`+ (同名 talent 有几位)与 `raw`(`basic_info`、各名义的 FANZA 资料、SNS 原样)。++ 资料只取 `meta.basic_info`。同一页 `actors[].meta.fanza` 另有一份官方资料,两份可以不一致+ (三上悠亜 B90/H 对 B84/F,FANZA 那份是旧值),不拿它补空格,只留在 `raw`。出道年、活动期、+ 出道作、博客与官网站上没有,留空。`primary.image_url` 是 DMM `actjpgs` 的 125×125,不作头像。+ """+ props = _page_props(html)+ talent = props.get("talent") if props else None+ if not isinstance(talent, dict) or talent.get("id") is None:+ return None+ meta = talent.get("meta") if isinstance(talent.get("meta"), dict) else {}+ basic = meta.get("basic_info") if isinstance(meta.get("basic_info"), dict) else {}+ actors = [actor for actor in talent.get("actors") or [] if isinstance(actor, dict)]+ primary = talent.get("primary") if isinstance(talent.get("primary"), dict) else {}+ names = list(dict.fromkeys(name for name in map(_named, [*actors, primary]) if name))+ cup = str(basic.get("cup") or "").strip().upper()+ blood = str(basic.get("blood_type") or "").strip().upper()+ return {+ "talent_id": str(talent["id"]), "names": names, "talents": len(props.get("talents") or []),+ "kana": _text(primary.get("ruby")), "romaji": None,+ "birth_date": _date(basic.get("birthday")),+ "height_cm": _centimetres(basic.get("height")), "bust_cm": _centimetres(basic.get("bust")),+ "cup": cup if _CUP.fullmatch(cup) else None,+ "waist_cm": _centimetres(basic.get("waist")), "hip_cm": _centimetres(basic.get("hip")),+ "blood_type": blood if blood in _BLOOD_TYPES else None,+ "birthplace": _text(basic.get("prefectures")), "hobbies": _text(basic.get("hobby")),+ "debut_year": None, "active_until": None, "debut_title": None, "debut_date": None,+ "blog_url": None, "site_url": None, "tags": [],+ "raw": {"basic_info": basic, "sns": meta.get("sns") or [],+ "fanza": {_named(actor): actor["meta"].get("fanza") for actor in actors+ if isinstance(actor.get("meta"), dict) and _named(actor)}},+ }diff --git a/tests/test_performer_profile_followup.py b/tests/test_performer_profile_followup.pyindex 4c03e350e..782302ca4 100644--- a/tests/test_performer_profile_followup.py+++ b/tests/test_performer_profile_followup.py@@ -27,6 +27,7 @@ from peach.performer_profiles import read_profile, write_profilefrom peach.repository import LedgerDatabasefrom peach.review_csv import read_rowsfrom peach.scraping_access import paused_until+from peach.sources import avbasefrom support.ledger import fresh_ledgerROOT = Path(__file__).resolve().parents[1]@@ -133,6 +134,34 @@ def avwikidb_movie(code: str, floor: str, cid: str, *cast: tuple[str, int],"actor": [{"name": name, "fanzaAvActressId": number} for name, number in cast]}+def avbase_talent(talent_id: int, *names: str, talents: int = 1, **basic) -> str:+ """avbase 女优页:按 2026-11-04 实测的小笠原菜乃(talent 69214)裁剪,作品表只留一部占位。++ `basic` 覆盖 `meta.basic_info` 的键;FANZA 那份资料故意写成另一组值,解析不该拿它。+ """+ info = {"birthday": "2005-04-12", "height": "159", "bust": "90", "waist": "53", "hip": "84",+ "cup": "I", "hobby": "サウナ・映画鑑賞", "prefectures": "岡山県", "blood_type": None, **basic}+ actors = [{"id": talent_id + index, "name": name, "ruby": "おがさわらなの",+ "image_url": "https://pics.dmm.co.jp/mono/actjpgs/ogasawara_nano.jpg",+ "meta": {"fanza": {"birthday": "2004-01-01", "height": "150", "bust": "80"}}}+ for index, name in enumerate(names)]+ talent = {"id": talent_id, "frozen": False, "actors": actors, "primary": actors[0],+ "profile": "", "meta": {"sns": [{"id": "nano_ogasawara", "sns": "twitter"}],+ "basic_info": info}}+ props = {"talent": talent, "talents": [{"id": talent_id + n} for n in range(talents)],+ "works": [{"work_id": "START-399", "title": "x" * 2000}], "total": 24}+ data = json.dumps({"props": {"pageProps": props}, "page": "/talents/[name]"}, ensure_ascii=False)+ data = data.replace("<", "\\u003c") # Next.js 写 `__NEXT_DATA__` 时把 `<` 转义成这样+ return (f'<!DOCTYPE html><html lang="ja"><body><div id="__next"></div>'+ f'<script id="__NEXT_DATA__" type="application/json">{data}</script></body></html>')+++# 站上没有这个名字:HTTP 404,Next.js 的错误页。+AVBASE_MISSING = ('<!DOCTYPE html><html><head><title>404: This page could not be found</title></head><body>'+ '<script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":'+ '{"statusCode":404}},"page":"/_error"}</script></body></html>')++AVWIKIDB_SINGLE = avwikidb_works([avwikidb_movie("DVAJ-495", "videoa", "dvaj00495", ("星野あかり", 11823), ("杉野綾子", 1025548)),avwikidb_movie("EEA-266", "videoc", "eesthe266", ("杉野綾子", 1025548), unknown=True),@@ -243,6 +272,39 @@ class ParsingTests(unittest.TestCase):page = AVWIKIDB_SINGLE.replace('"movies":', '"singleCount": 2, "movies":')self.assertEqual(avwikidb.single_count(page), 2)+ def test_the_talent_page_fills_the_profile_columns_from_basic_info_only(self):+ """同一页 FANZA 那份资料是另一组旧值,不拿它;出道年、博客这几项站上没有,留空。"""+ found = avbase.talent_profile(avbase_talent(69214, "小笠原菜乃", "旧名なの"))+ self.assertEqual({column: found[column] for column in minnano_av.PROFILE_COLUMNS}, {+ "kana": "おがさわらなの", "romaji": None, "birth_date": "2005-04-12", "height_cm": 159,+ "bust_cm": 90, "cup": "I", "waist_cm": 53, "hip_cm": 84, "blood_type": None,+ "birthplace": "岡山県", "hobbies": "サウナ・映画鑑賞", "debut_year": None,+ "active_until": None, "debut_title": None, "debut_date": None, "blog_url": None,+ "site_url": None})+ self.assertEqual((found["talent_id"], found["names"], found["talents"]),+ ("69214", ["小笠原菜乃", "旧名なの"], 1))+ self.assertEqual(found["raw"]["fanza"]["小笠原菜乃"]["height"], "150")++ def test_blank_and_malformed_talent_values_stay_empty(self):+ found = avbase.talent_profile(avbase_talent(+ 1, "甲", birthday="", height="158.5", bust="0", cup="f", hip="-", blood_type="AB",+ prefectures="", hobby=""))+ self.assertEqual((found["birth_date"], found["height_cm"], found["bust_cm"], found["cup"],+ found["hip_cm"], found["blood_type"], found["birthplace"], found["hobbies"]),+ (None, 158, None, "F", None, "AB", None, None))+ self.assertIsNone(avbase.talent_profile(avbase_talent(1, "甲", birthday="2005-02-30"))+ ["birth_date"])++ def test_a_missing_name_page_is_not_a_talent(self):+ self.assertIsNone(avbase.talent_profile(AVBASE_MISSING))++ def test_the_cached_talent_page_keeps_only_what_the_parser_reads(self):+ page = avbase_talent(69214, "小笠原菜乃", hobby="</script>")+ small = avbase.compact_talent_page(page)+ self.assertEqual(avbase.talent_profile(small), avbase.talent_profile(page))+ self.assertNotIn("START-399", small)+ self.assertEqual(avbase.compact_talent_page(AVBASE_MISSING), AVBASE_MISSING)+def test_romaji_compares_word_order_free(self):self.assertEqual(avwikidb.romaji_key("Hikaru Minazuki"), avwikidb.romaji_key("Minazuki Hikaru"))self.assertNotEqual(avwikidb.romaji_key("Hikaru Minazuki"), avwikidb.romaji_key("Hikari Minazuki"))@@ -266,6 +328,7 @@ class Case(unittest.TestCase):MINNANO_URL: (200, MINNANO_PROFILE)})self.avwikidb = Transport({WORK_URL: (200, AVWIKIDB_WORK), ACTOR_URL: (200, AVWIKIDB_ACTOR)})self.javdb = Transport({}, empty_search=javdb_search())+ self.avbase = Transport({})def entity(self, name: str, *aliases: str) -> int:with self.database.write_transaction(notify=False) as connection:@@ -310,6 +373,10 @@ class Case(unittest.TestCase):return {followup.MINNANO: alias.MinnanoPages(cache / "minnano", self.cooldown, self.minnano, limiter=NoWait(),max_age=followup.REFRESH),+ followup.AVBASE: alias.MinnanoPages(+ cache / "avbase", self.cooldown, followup.TalentPages(self.avbase),+ limiter=NoWait(), max_requests=alias.MAX_KEYS, source=followup.AVBASE,+ max_age=followup.REFRESH),followup.AVWIKIDB: alias.MinnanoPages(cache / "avwikidb", self.cooldown, self.avwikidb, limiter=NoWait(),source=followup.AVWIKIDB, max_age=followup.REFRESH),@@ -506,6 +573,112 @@ class AvwikidbTests(Case):self.assertEqual(followup.work_codes(connection, fc2), [])+class AvbaseTests(Case):+ """minnano-av 上没有她时,avbase 女优页补资料;有别的来源那一行时只记冲突(ADR-0136)。"""++ NANO_URL = avbase.talent_url("小笠原菜乃")+ MEGURI_URL = avbase.talent_url("美ノ嶋めぐり")++ def test_she_is_missing_on_minnano_so_avbase_writes_the_row_and_binds_the_talent(self):+ nano = self.entity("小笠原菜乃")+ self.work(1, nano)+ self.avbase.pages[self.NANO_URL] = (200, avbase_talent(69214, "小笠原菜乃"))+ summary = self.run_followup(nano, run_id=5)+ found = self.profile(nano)+ self.assertEqual((found["birth_date"], found["height_cm"], found["cup"]), ("2005-04-12", 159, "I"))+ self.assertEqual((found["source"], found["source_url"]), (f"{followup.SOURCE}@5", self.NANO_URL))+ number, metadata = self.refs(nano)[followup.AVBASE]+ self.assertEqual((number, metadata["batch"], metadata["names"]),+ ("69214", f"{followup.SOURCE}@5", ["小笠原菜乃"]))+ self.assertTrue(summary["landed"])+ self.assertEqual({row["action"] for row in self.review() if row["site"] == followup.AVBASE},+ {followup.BIND, followup.WRITE})+ cached = next((self.generated / "provider-cache" / "avbase").glob("*.json")).read_text(encoding="utf-8")+ self.assertNotIn("START-399", cached, "缓存里只留裁过的那两项")++ def test_an_alias_key_is_tried_after_a_404(self):+ nano = self.entity("小笠原菜乃", "旧名なの")+ self.work(1, nano)+ self.avbase.pages[self.NANO_URL] = (404, AVBASE_MISSING)+ self.avbase.pages[avbase.talent_url("旧名なの")] = (200, avbase_talent(69214, "旧名なの"))+ self.run_followup(nano)+ self.assertEqual(self.refs(nano)[followup.AVBASE][0], "69214")++ def test_a_minnano_row_is_kept_and_only_the_disagreement_is_recorded(self):+ meguri = self.entity("美ノ嶋めぐり")+ self.work(1, meguri)+ self.ref(meguri, alias.MINNANO, "387589")+ self.avbase.pages[self.MEGURI_URL] = (200, avbase_talent(+ 47000, "美ノ嶋めぐり", birthday="2001-12-09", height="156"))+ summary = self.run_followup(meguri)+ found = self.profile(meguri)+ self.assertEqual((found["birth_date"], found["source_url"]), ("2001-12-08", MINNANO_URL))+ metadata = self.refs(meguri)[followup.AVBASE][1]+ self.assertEqual(metadata["conflicts"], [+ {"field": "birth_date", alias.MINNANO: "2001-12-08", followup.AVBASE: "2001-12-09"}])+ self.assertEqual(summary["conflicts"], 1)+ self.assertIn("birth_date:minnano-av 2001-12-08,avbase 2001-12-09;资料表不改",+ [row["detail"] for row in self.review()])++ def test_an_avbase_row_does_not_keep_minnano_away(self):+ """avbase 补的一行不算 minnano-av 的「未过期」:她后来上了 minnano-av,下一轮整行换成那边的。"""+ meguri = self.entity("美ノ嶋めぐり")+ self.work(1, meguri)+ self.avbase.pages[self.MEGURI_URL] = (200, avbase_talent(47000, "美ノ嶋めぐり"))+ self.minnano.pages.pop(minnano_av.search_url("美ノ嶋めぐり"))+ self.run_followup(meguri, run_id=7)+ self.assertEqual(self.profile(meguri)["source_url"], self.MEGURI_URL)+ self.ref(meguri, alias.MINNANO, "387589")+ self.run_followup(meguri, run_id=8)+ found = self.profile(meguri)+ self.assertEqual((found["source_url"], found["birth_date"]), (MINNANO_URL, "2001-12-08"))+ self.assertEqual(self.avbase.calls, [self.MEGURI_URL], "avbase 编号已绑、未过期,第二轮不再问")++ def test_a_name_missing_everywhere_is_a_miss_not_unfetched(self):+ nano = self.entity("小笠原菜乃")+ self.work(1, nano)+ summary = self.run_followup(nano)+ self.assertTrue(summary["sites"][followup.AVBASE].startswith(followup.MISS))+ self.assertEqual(summary["outcome"], "哪一站都没对上她")+ self.assertIsNone(self.profile(nano))++ def assert_nothing_written(self, page: str) -> None:+ nano = self.entity("小笠原菜乃")+ self.work(1, nano)+ self.avbase.pages[self.NANO_URL] = (200, page)+ summary = self.run_followup(nano)+ self.assertTrue(summary["sites"][followup.AVBASE].startswith(followup.MISS))+ self.assertIsNone(self.profile(nano))+ self.assertNotIn(followup.AVBASE, self.refs(nano))++ def test_a_page_that_does_not_list_her_name_writes_nothing(self):+ self.assert_nothing_written(avbase_talent(1, "別人"))++ def test_a_name_shared_by_two_talents_writes_nothing(self):+ self.assert_nothing_written(avbase_talent(1, "小笠原菜乃", talents=2))++ def test_a_refusal_pauses_avbase_and_writes_nothing(self):+ nano = self.entity("小笠原菜乃")+ self.work(1, nano)+ self.avbase.pages[self.NANO_URL] = (403, "")+ summary = self.run_followup(nano)+ self.assertTrue(summary["sites"][followup.AVBASE].startswith(followup.UNFETCHED))+ self.assertTrue(paused_until(self.cooldown, followup.AVBASE))+ self.assertIsNone(self.profile(nano))++ def test_the_batch_reverts_the_row_and_the_talent_number(self):+ nano = self.entity("小笠原菜乃")+ self.work(1, nano)+ self.avbase.pages[self.NANO_URL] = (200, avbase_talent(69214, "小笠原菜乃"))+ self.run_followup(nano, run_id=5)+ base = ["--db", str(self.db), "--logo-root", str(self.root / "logos"),+ "--source", followup.SOURCE, "--apply", "--backup", str(self.root / "b.db")]+ with contextlib.redirect_stdout(io.StringIO()):+ self.assertEqual(load_revert().main(base), 0)+ self.assertIsNone(self.profile(nano))+ self.assertNotIn(followup.AVBASE, self.refs(nano))++class JavdbTests(Case):def setUp(self):super().setUp()