Peach: Local-first video metadata scraper and media manager.
fix(catalog): 补齐文件名番号提取与 KIRARI 采集入口
longmeidao9c1799250f33
diff --git a/docs/REUSE.md b/docs/REUSE.mdindex ef11c4ef8..b276f91f6 100644--- a/docs/REUSE.md+++ b/docs/REUSE.md@@ -587,6 +587,7 @@ A 盘垃圾隔离使用短名称 `peach-purge-<32 位 UUID>.peach-quarantine`,- 高亮、竖屏密度、索引骨架已验桌面/390×844,HTTPS 生效;手机命中区 44 px。- 主题三选一(跟随系统/浅色/深色),只存本机,首帧前由内联脚本写进 `<html>` 的 `data-theme`。- 同番号的分卷派生(A/B、1/2、CD/Disc/DVD/Part/Vol、「首卷裸名 + 后续卷 `-2`/`-3`」)折叠成一张卡并按时长排除完整版,首页、搜索、资料页网格、版次队列与角标计数共用这套判定。卷号后挂着尾缀时先剥掉组内每个文件名都带、且从分隔符起头的那一段,不按版次词表拆(`1080p` 会被拆成 `10` 加 `80p`)。+- 番号提取复用 `catalog_rules` 的视频容器后缀、日期式番号和 KIRARI 形态:扫描支持的媒体与图片后缀均可剥除,明确卷标后的两位序号保留,日期式序号接受 2~4 位并保留原分隔符。`MKBD-S`/`MKD-S` 共用文本、文件名、自定义规则、无码来源路由和标题与版次投影判据;目录预检、配套资料和资料采集均调用同一解析器。运行时只用标准库 `re`,不引入外部解析器;[NeoAVDC](https://github.com/haoo999/NeoAVDC) 0.0.4(MIT)的 TypeScript 解析器只作边界对照,其全串宽松提取不适用 Peach 的创作者账号与相机编号保护。- 分卷卡与版次卡(有码/中字/无码)不翻卡(各卷共用同一封套,翻了像卡住),悬浮走分段视频预览;封面只有一个计数(「N 卷」或「N 个版本」),叠层封面格垫 `--page` 挡住模糊衬底的半透明边。分卷详情标题带卷号,因为各卷标题、女优、厂牌逐字相同。- 本地播放版本复用 `catalog_rules` 的番号、版次、卷号与重复文件判断。`catalog_versions` 先区分版次、探测规格与修复标记,再识别各版分卷;完整文件与分卷同时存在时核对总时长,重复副本保留文件并共用一个版本身份。计数、首页、资料页、详情队列和相关推荐读取同一查询投影,ledger 不变。`scripts/audit_catalog_versions.py` 只读输出全库复核 CSV;没有时长、规格或明确番号证据的关系保留待复核。- 文件身份复用 `metadata.identifies_code` 核对 DMM 数字前缀与补零写法;通用文件名须同时有带归属的刮削标题和匹配的番号目录,显式文件号冲突优先拒绝。同名副本的已知时长相差超过 10% 时保持独立,缺测量项不附到有时长冲突的同名组。来源目录时长保留为证据,不代填媒体实测时长;特典、广告与缺内容的取舍记录在逐文件复核产物里。diff --git a/src/peach/catalog_rules.py b/src/peach/catalog_rules.pyindex 0492c1945..020f8a2c9 100644--- a/src/peach/catalog_rules.py+++ b/src/peach/catalog_rules.py@@ -194,7 +194,8 @@ _CODE_DATE = re.compile(rf"^{_MMDDYY}[-_]\d{{2,4}}$")#: `MKBD-S` 是 Blu-ray 版。javbus 上 `MKBD-S118`/`MKD-S118`/`MKBD-S89`/`MKD-S89`#: 四页都认这个写法、都给封面(2026-09-25 实测)。通用的「字母-S数字」不放开:#: 没有实证的形态放进来,素材编号和文件计数都会被当成番号。-_KIRARI_CODE = re.compile(r"^MKB?D-S\d{2,4}$")+_KIRARI_BODY = r"MKB?D-S\d{2,4}"+_KIRARI_CODE = re.compile(rf"^{_KIRARI_BODY}$", re.I)#: Tokyo-Hot 的编号没有厂牌字母段:`n1234`/`k1234` 是本编,`red-123` 是 Red Hot 支线。#: 规范写法取小写,不跟着别的番号大写——javbus 的作品页地址就是 `/n1234`,NeoAVDC@@ -242,7 +243,7 @@ _RELEASE_PREFIX_ALIASES = {}_DMM_LABEL_PREFIX = re.compile(r"^H_(?=\d)")_MEDIA_EXTENSION = re.compile(- r"\.(?:mp4|mkv|avi|wmv|mov|m4v|webm|ts|m2ts|mts|mpg|mpeg|flv|rm|rmvb|iso)$",+ r"\.(?:mp4|mkv|avi|wmv|mov|m4v|webm|ts|m2ts|mts|mpg|mpeg|flv|rm|rmvb|iso|strm)$",re.I,)_PROMO_DOMAIN = re.compile(@@ -289,13 +290,13 @@ _QUALITY_HEAD = re.compile(r"^(?:hd|fhd|sd|uhd|4k|2160p?|1080p?|720p?)[-_. ]+",VERSION_TAIL_TOKENS = ("ch", "sub", "uc", "fhd", "4k", "hd", "c", "u")VERSION_TAIL = re.compile(r"[-_. ]?(?:" + "|".join(VERSION_TAIL_TOKENS) + r")$", re.I)#: 一本道、加勒比是「日期+序号」体系,没有字母番号主体。分隔符进捕获组,原样带出去。-_DATE_CODE = re.compile(rf"(?<!\d)({_MMDDYY})([-_])(\d{{3}})(?!\d)")+_DATE_CODE = re.compile(rf"(?<!\d)({_MMDDYY})([-_])(\d{{2,4}})(?!\d)")#: UUID 首段长得像番号(`DCE7230C-730E-…` 会被拆成 `DCE`+`7230`),按整串形态排除。_UUID_LIKE = re.compile(r"[0-9A-F]{8}-[0-9A-F]{4}-[0-9A-F]{4}-[0-9A-F]{4}-[0-9A-F]{12}", re.I)-#: 目录名和文件名都可能带扩展名,图片也算:缩略图 `BNST033(2).jpg` 与正片同番号。+#: 媒体、串流入口和图片共用番号提取;视频容器沿用展示层的后缀判据。_ASSET_EXTENSION = re.compile(- r"\.(?:mp4|mkv|avi|wmv|ts|mov|m4v|jpg|jpeg|png|webp)$", re.I)+ rf"(?:{_MEDIA_EXTENSION.pattern}|\.(?:jpg|jpeg|png|webp|gif|bmp)$)", re.I)#: 文件名尾部的分卷、画质和重复计数;剥掉才能和目录名对齐。_FILE_NOISE = re.compile(r"(?:[-_. ]?(?:1080p?|720p?|2160p?|4k|fhd|hd|uc|sub|ch|c|u)"@@ -402,7 +403,10 @@ _QUALITY_TOKEN = re.compile(r"^(?:hd|fhd|sd|uhd|4k|2160p?|1080p?|720p?)$", re.I)#: 序号至少三位:两位的(`ShinoAoi-15`、`miru-10`)是创作者的分集号;后面紧跟月日的#: (`TG-2025-01-27`)是日期。只取词元开头这一段,紧跟的版次和分卷(`-C`、`-CD1`、#: `_part1`)不影响番号。-_HYPHENATED_CODE = re.compile(r"^(?:\d{3})?[A-Za-z]{2,8}-\d{3,5}(?![0-9]|-\d{2}-\d)")+_HYPHENATED_CODE = re.compile(+ rf"^(?:{_KIRARI_BODY}(?![0-9])|(?:\d{{3}})?[A-Z]{{2,8}}-\d{{3,5}}(?![0-9]|-\d{{2}}-\d))",+ re.I,+)#: 用户番号识别规则的上限。每条规则要对库里每个文件名跑一遍,条数和长度都设个顶。CODE_PATTERN_LIMIT = 20CODE_PATTERN_MAX_LENGTH = 200@@ -789,9 +793,9 @@ def release_code_from_text(value: str | None) -> str | None:if date:return f"{date.group(1)}{date.group(2)}{date.group(3)}"body = VERSION_TAIL.sub("", _QUALITY_HEAD.sub("", text))- if not _CODE_BODY.match(body) or is_code_body_stopword(body):- return Nonecanonical = normalise_code_key(body)+ if not (_CODE_BODY.fullmatch(body) or _KIRARI_CODE.fullmatch(canonical)) or is_code_body_stopword(body):+ return Nonereturn canonical if is_jav_code(canonical) else None@@ -864,7 +868,9 @@ def release_code_from_filename(name: str | None, *, patterns: Iterable[re.Patterreturn Nonepartless = _ASSET_EXTENSION.sub("", _PART_MARKER.sub("", str(name or "").strip()))stripped = _FILE_NOISE.sub("", strip_promo_markers(partless))+ # 明确卷标摘掉后先核对番号主体,两位序号也属于完整番号。code = release_code_from_text(stem) or (+ release_code_from_text(partless) if _STANDS_ALONE.search(partless) else None) or (release_code_from_text(stripped) if _STANDS_ALONE.search(stripped) else None)if code:return code@@ -901,6 +907,9 @@ def _titled_release_code(stem: str) -> str | None:def _jav_code_pattern(code: str | None) -> str:"""Return one regex fragment matching compact and separated forms of a canonical code."""canonical = normalise_code_key(code)+ if _KIRARI_CODE.fullmatch(canonical):+ letters, serial = canonical.split('-', 1)+ return rf"{re.escape(letters)}[-_ ]{re.escape(serial)}"fc2 = re.fullmatch(r"FC2-PPV-(\d+)", canonical)if fc2:return rf"FC2(?:[-_ ]?PPV)?[-_ ]*0*{re.escape(fc2.group(1))}"@@ -927,9 +936,10 @@ def _jav_code_pattern(code: str | None) -> str:#: 无码片商自己的编号法:Caribbeancom/1Pondo/10musume/Pacopacomama 走日期式#: (形状与分隔符含义见 `_CODE_DATE`,同一份实现),HEYZO 用 `HEYZO-1380`,-#: Tokyo-Hot 用 `n1234`(见 `_TOKYO_HOT_BODY`)。有码片商不用这三种形状。+#: Tokyo-Hot 用 `n1234`(见 `_TOKYO_HOT_BODY`),KIRARI 用 `MKBD-S`/`MKD-S`。UNCENSORED_CODE_SHAPES = (_CODE_DATE,+ _KIRARI_CODE,re.compile(r"^HEYZO-\d{2,5}$", re.I),_TOKYO_HOT_BODY,)@@ -1096,12 +1106,12 @@ def jav_fallback_title(name: str | None, code: str | None) -> str:code_pattern = _jav_code_pattern(code)if code_pattern:repeated = re.compile(- rf"^[\s._\-—]*(?:{code_pattern})(?:{_GLUED_EDITION})?(?=$|[\s._\-—\[])", re.I)+ rf"^[\s._\-—]*(?:{code_pattern})(?:[-_.]?{_GLUED_EDITION})?(?=$|[\s._\-—\[])", re.I)while repeated.search(text):text = repeated.sub("", text, count=1)# 番号不总在开头:`1pon-092415-001-fhd1_(new).mp4` 把发行站放在了前面。# 只认前缀,整个番号就会留在「标题」里显示出来。- text = re.sub(rf"(?<![A-Z0-9])(?:{code_pattern})(?:{_GLUED_EDITION})?(?![A-Z0-9])",+ text = re.sub(rf"(?<![A-Z0-9])(?:{code_pattern})(?:[-_.]?{_GLUED_EDITION})?(?![A-Z0-9])"," ", text, flags=re.I)text = _EDITION_TAIL.sub("", text)text = _RELEASE_NOISE.sub(" ", text)diff --git a/tests/test_jav_code_domain.py b/tests/test_jav_code_domain.pyindex eff59b4ef..3f06c2246 100644--- a/tests/test_jav_code_domain.py+++ b/tests/test_jav_code_domain.py@@ -41,6 +41,7 @@ from peach.catalog_rules import (is_jav_code,is_repost_site_label,is_uncensored_code,+ jav_display_metadata,jav_fallback_title,normalise_code_key,release_code_from_filename,@@ -154,6 +155,25 @@ class ScrapeGateTests(unittest.TestCase):class ExtractionTests(unittest.TestCase):+ def test_scannable_media_suffixes_preserve_short_release_codes(self):+ from peach.scan import IMAGE, VIDEO+ for suffix in sorted(VIDEO | IMAGE):+ for name, code in ((f'LAF-41{suffix}', 'LAF-041'),+ (f'MEYD911{suffix.upper()}', 'MEYD-911')):+ with self.subTest(name=name):+ self.assertEqual(release_code_from_filename(name), code)+ if suffix in VIDEO:+ self.assertEqual(jav_fallback_title(name, code), '')++ def test_media_suffixes_preserve_volumes_and_exclude_camera_names(self):+ for suffix in ('.mpg', '.flv', '.rmvb', '.m2ts', '.iso', '.strm'):+ with self.subTest(suffix=suffix):+ self.assertEqual(release_code_from_filename(f'LAF-41-CD2{suffix}'), 'LAF-041')+ self.assertEqual(release_code_from_filename(f'LAF-41-HD{suffix}'), 'LAF-041')+ self.assertEqual(release_code_from_filename(f'ABW-358-C{suffix}'), 'ABW-358')+ for stem in ('IMG_3092', 'VIDEO-2022', 'no0037_01', '[mtfdz.club]WX17.3'):+ self.assertIsNone(release_code_from_filename(stem + suffix))+def test_named_volume_suffix_preserves_the_release_code(self):for suffix in ('CD1', 'disc02', 'Part-3', 'DVD4', 'Vol_5'):with self.subTest(suffix=suffix):@@ -204,6 +224,17 @@ class DatedCodeSeparatorTests(unittest.TestCase):self.assertEqual(normalise_code_key("092415-001"), "092415-001")self.assertEqual(normalise_code_key(" 092415_001 "), "092415_001")+ def test_each_supported_serial_length_reaches_filename_extraction(self):+ for separator in ('-', '_'):+ for number in ('01', '001', '0012'):+ code = f'040221{separator}{number}'+ with self.subTest(code=code):+ self.assertEqual(release_code_from_text(code), code)+ self.assertEqual(release_code_from_filename(f'{code}.mpg'), code)+ self.assertEqual(release_code_from_filename(f'{code}-CD2.mkv'), code)+ for code in ('040221-1', '040221-00123', '135735_0012'):+ self.assertIsNone(release_code_from_filename(f'{code}.mpg'), code)+def test_the_two_separators_are_two_different_releases(self):self.assertFalse(same_release_code("092415_001", "092415-001"))self.assertFalse(same_release_code("092416-001", "092415-001"))@@ -275,6 +306,32 @@ class TokyoHotShapeTests(unittest.TestCase):for value in ("ABC-S12", "MKBD-S1", "MKBD-89S"):self.assertFalse(is_jav_code(normalise_code_key(value)), value)+ def test_kirari_codes_reach_text_filename_and_custom_rule_entrypoints(self):+ from peach.metadata_routes import classify+ rules = compile_code_patterns([r'番号\{(?P<code>[^}]+)\}'])+ for code in ('MKBD-S94', 'MKD-S94', 'MKBD-S118', 'MKD-S1234'):+ self.assertTrue(is_uncensored_code(code.lower()), code)+ self.assertEqual(classify(code), 'uncensored')+ for suffix in ('', '-C', '-CD2', '_part1'):+ name = f'{code.lower()}{suffix}.mkv'+ with self.subTest(name=name):+ self.assertEqual(release_code_from_filename(name), code)+ self.assertEqual(release_code_from_text(code.lower()), code)+ self.assertEqual(release_code_from_text(code.replace('-', '_')), code)+ self.assertEqual(release_code_from_filename(f'{code} 片名.mpg'), code)+ self.assertEqual(release_code_from_filename(f'番号{{{code}}}.mp4', patterns=rules), code)+ for code in ('ABC-S12', 'MKBD-S1', 'MKBD-S12345', 'MKBD-89S'):+ for suffix in ('.mkv', ' 片名.mp4'):+ self.assertIsNone(release_code_from_filename(code + suffix), code)++ def test_kirari_display_projects_titles_and_edition_badges(self):+ for separator in ('-', '_'):+ name = f'MKBD{separator}S94-C 作品标题.mkv'+ self.assertEqual(jav_display_metadata(name, 'MKBD-S94'), {+ 'display_code': 'MKBD-S94', 'display_title': '作品标题',+ 'edition_badges': ['中字', '无码']})+ self.assertEqual(jav_fallback_title('MKD-S118.mp4', 'MKD-S118'), '')+def test_a_bare_number_is_read_only_at_the_head_of_the_name(self):self.assertEqual(release_code_from_filename("n1032.mkv"), "n1032")self.assertEqual(release_code_from_filename("n0762.mkv"), "n0762")@@ -350,6 +407,12 @@ class TitledFilenameTests(unittest.TestCase):("300MIUM-1239 片名.mp4", "300MIUM-1239")):self.assertEqual(release_code_from_filename(name), code, name)+ def test_an_edition_marker_between_code_and_title_is_not_title_text(self):+ for code in ('ABW-358', 'MKBD-S94'):+ for edition in ('C', 'CH', 'SUB', 'UC', 'U'):+ self.assertEqual(jav_fallback_title(f'{code}-{edition} 作品标题.mkv', code), '作品标题')+ self.assertEqual(jav_fallback_title(f'{code} Club 标题.mkv', code), 'Club 标题')+def test_an_fc2_number_is_read_anywhere_in_the_name(self):for name in ("+++ FC2-PPV-1234567 片名 前編.mp4","+++ FC2-PPV-1234567_part1.mp4",diff --git a/tests/test_metadata_library.py b/tests/test_metadata_library.pyindex 3f96c8723..4ae18555c 100644--- a/tests/test_metadata_library.py+++ b/tests/test_metadata_library.py@@ -420,6 +420,37 @@ class LibraryNfoTests(unittest.TestCase):self.addCleanup(temporary.cleanup)self.root = Path(temporary.name).resolve()+ @windows_ledger_roots+ def test_filename_code_families_enter_metadata_collection(self):+ media = self.root / 'media'+ media.mkdir()+ codes = {'LAF-41.mpg': 'LAF-041', 'MKBD-S94.mkv': 'MKBD-S94',+ 'MKD-S118.iso': 'MKD-S118', '040221_0012.flv': '040221_0012'}+ for name in codes:+ (media / name).write_bytes(b'video')+ db = fresh_ledger(self.root)+ config = PeachConfig(self.root, self.root / 'config.toml', present=True,+ locations={'local': (str(media),)})+ process_library(config, db, self.root / 'generated', self.root / 'covers', stage='scan')+ provider = stub_provider()+ provider.query.side_effect = lambda code, *args, **kwargs: {+ 'id': code, 'title': f'测试作品 {code}', 'maker': '测试厂牌',+ 'release_date': '2026-01-01', 'actresses': [{'japanese_name': '测试演员'}]}+ with patch('peach.library_processing._asks_cover', return_value=False), \+ patch('peach.library_processing._enrich_finished_performers',+ return_value={'aliases': 0, 'avatars': 0, 'conflicts': 0, 'failed': 0}):+ result = process_library(config, db, self.root / 'generated', self.root / 'covers',+ stage='metadata', provider_factory=lambda: provider, database=LedgerDatabase(db),+ route_overrides='censored=r18dev;uncensored=r18dev')+ self.assertEqual(result['status'], 'complete')+ self.assertEqual(result['identified'], len(codes))+ self.assertEqual({call.args[0] for call in provider.query.call_args_list}, set(codes.values()))+ with closing(sqlite3.connect(db)) as connection:+ rows = connection.execute('SELECT name,code,catalog_title FROM asset').fetchall()+ self.assertEqual({name: code for name, code, _ in rows}, codes)+ self.assertEqual({code: title for _, code, title in rows},+ {code: f'测试作品 {code}' for code in codes.values()})+def test_javboss_fields_and_original_tags(self):path = self.root / 'sample.nfo'path.write_text('<movie><generator>JavBoss</generator><uniqueid type="javboss">ABW-358</uniqueid>'diff --git a/tests/test_processing_inventory.py b/tests/test_processing_inventory.pyindex ad82b70cc..514cf6b49 100644--- a/tests/test_processing_inventory.py+++ b/tests/test_processing_inventory.py@@ -50,6 +50,17 @@ class InventoryTests(unittest.TestCase):self.assertGreater(result['estimated_bytes'], 0)self.assertEqual(self.db.read_bytes(), before)+ def test_release_code_families_count_as_jav_during_inventory(self):+ for name in ('LAF-41.mpg', 'MKBD-S94.mkv', 'MKD-S118.iso', '040221_0012.flv',+ 'IMG_3092.mpg', 'VIDEO-2022.rmvb'):+ (self.media / name).touch()+ before = self.db.read_bytes()+ result = preview(self.config, self.db, 'all', None)+ self.assertEqual(result['workload']['video_files'], 6)+ self.assertEqual(result['workload']['jav_files'], 4)+ self.assertEqual(result['workload']['jav_titles'], 4)+ self.assertEqual(self.db.read_bytes(), before)+def test_each_query_advances_a_batch_and_preserves_deduplication(self):for number in range(1100):(self.media / f'ABW-001-CD{number}.mp4').touch()