From 36969a4fa0a3ced2a6615c5424d4027682adecfe Mon Sep 17 00:00:00 2001 From: lychang Date: Sun, 19 Jul 2026 23:56:28 +0800 Subject: [PATCH] update --- Cargo.lock | 204 ++++++++++++++ assets/album.html | 27 +- assets/css/album.css | 419 ++++++++++++++++++++++------ assets/css/style.css | 162 ++++++++++- assets/index.html | 42 +-- assets/js/album.js | 12 +- assets/js/app.js | 136 +++++++-- docs/pipeline-architecture.md | 226 +++++++++++++++ docs/pipeline-spec.md | 124 ++++++++ src/db/metadata.rs | 15 + src/engine/person_merger.rs | 13 +- src/engine/pipeline.rs | 512 +++++++++++++++++++++++++++++----- src/engine/tag_extractor.rs | 110 ++++++-- src/engine/title_parser.rs | 130 +++++++-- src/main.rs | 119 ++++---- src/routes/pipeline.rs | 24 +- 16 files changed, 1925 insertions(+), 350 deletions(-) create mode 100644 docs/pipeline-architecture.md create mode 100644 docs/pipeline-spec.md diff --git a/Cargo.lock b/Cargo.lock index fa82d82..f4e04f9 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -214,6 +214,12 @@ version = "2.0.1" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "320119579fcad9c21884f5c4861d16174d0e06250625266f50fe6898340abefa" +[[package]] +name = "adler32" +version = "1.2.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "aae1277d39aeec15cb388266ecc24b11c80469deae6067e17a1a7aa9e5c1f234" + [[package]] name = "ahash" version = "0.7.8" @@ -426,6 +432,15 @@ dependencies = [ "shlex", ] +[[package]] +name = "cedarwood" +version = "0.4.6" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "6d910bedd62c24733263d0bed247460853c9d22e8956bd4cd964302095e04e90" +dependencies = [ + "smallvec", +] + [[package]] name = "cfg-if" version = "1.0.4" @@ -606,6 +621,12 @@ dependencies = [ "hybrid-array", ] +[[package]] +name = "dary_heap" +version = "0.3.9" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "8b1e3a325bc115f096c8b77bbf027a7c2592230e70be2d985be950d3d5e60ebe" + [[package]] name = "der" version = "0.7.10" @@ -877,6 +898,15 @@ dependencies = [ "slab", ] +[[package]] +name = "fxhash" +version = "0.2.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "c31b6d751ae2c7f11320402d34e41349dd1016f8d5d45e48c4312bc8625af50c" +dependencies = [ + "byteorder", +] + [[package]] name = "generic-array" version = "0.14.7" @@ -970,6 +1000,17 @@ dependencies = [ "allocator-api2", ] +[[package]] +name = "hashbrown" +version = "0.16.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "841d1cc9bed7f9236f321df977030373f4a4163ae1a7dbfe1a51a2c1a51d9100" +dependencies = [ + "allocator-api2", + "equivalent", + "foldhash", +] + [[package]] name = "hashbrown" version = "0.17.1" @@ -1265,6 +1306,7 @@ dependencies = [ "chrono", "config", "image", + "jieba-rs", "lazy_static", "mimalloc", "rand 0.8.7", @@ -1286,6 +1328,39 @@ version = "0.3.1" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "35a84fd5aa25fae5c0f4a33d9cac2ca017fc622cbd089be2229993514990f870" +[[package]] +name = "include-flate" +version = "0.3.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "48f173716febb1ad596c16ea5637b5f1790ea32de8e627493ff82bc73b0876ce" +dependencies = [ + "include-flate-codegen", + "include-flate-compress", +] + +[[package]] +name = "include-flate-codegen" +version = "0.3.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "4a7875b62a72ad3f3203cdd8950d4cf9947db036030b974b8b37ceae90c8d8c0" +dependencies = [ + "include-flate-compress", + "proc-macro-error3", + "proc-macro2", + "quote", + "syn 2.0.119", +] + +[[package]] +name = "include-flate-compress" +version = "0.3.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "44fbb9c5ccb9a5b67b4afa2974c27e5507ea1bf6d22828cef418e4dfaeca51dd" +dependencies = [ + "libflate", + "zstd", +] + [[package]] name = "indexmap" version = "2.14.0" @@ -1308,6 +1383,30 @@ version = "1.0.18" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "8f42a60cbdf9a97f5d2305f08a87dc4e09308d1276d28c869c684d7777685682" +[[package]] +name = "jieba-macros" +version = "0.7.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "7c676b32a471d3cfae8dac2ad2f8334cd52e53377733cca8c1fb0a5062fec192" +dependencies = [ + "phf_codegen", +] + +[[package]] +name = "jieba-rs" +version = "0.7.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "f5dd552bbb95d578520ee68403bf8aaf0dbbb2ce55b0854d019f9350ad61040a" +dependencies = [ + "cedarwood", + "fxhash", + "include-flate", + "jieba-macros", + "lazy_static", + "phf", + "regex", +] + [[package]] name = "jobserver" version = "0.1.35" @@ -1367,6 +1466,30 @@ version = "0.2.186" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "68ab91017fe16c622486840e4c83c9a37afeff978bd239b5293d61ece587de66" +[[package]] +name = "libflate" +version = "2.3.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "cd96e993e5f3368b0cb8497dae6c860c22af8ff18388c61c6c0b86c58d86b5df" +dependencies = [ + "adler32", + "crc32fast", + "dary_heap", + "libflate_lz77", + "no_std_io2", +] + +[[package]] +name = "libflate_lz77" +version = "2.3.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "ff7a10e427698aef6eef269482776debfef63384d30f13aad39a1a95e0e098fd" +dependencies = [ + "hashbrown 0.16.1", + "no_std_io2", + "rle-decode-fast", +] + [[package]] name = "libm" version = "0.2.16" @@ -1524,6 +1647,15 @@ dependencies = [ "windows-sys 0.61.2", ] +[[package]] +name = "no_std_io2" +version = "0.9.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "418abd1b6d34fbf6cae440dc874771b0525a604428704c76e48b29a5e67b8003" +dependencies = [ + "memchr", +] + [[package]] name = "nom" version = "7.1.3" @@ -1702,6 +1834,44 @@ dependencies = [ "pest", ] +[[package]] +name = "phf" +version = "0.11.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "1fd6780a80ae0c52cc120a26a1a42c1ae51b247a253e4e06113d23d2c2edd078" +dependencies = [ + "phf_shared", +] + +[[package]] +name = "phf_codegen" +version = "0.11.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "aef8048c789fa5e851558d709946d6d79a8ff88c0440c587967f8e94bfb1216a" +dependencies = [ + "phf_generator", + "phf_shared", +] + +[[package]] +name = "phf_generator" +version = "0.11.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "3c80231409c20246a13fddb31776fb942c38553c51e871f8cbd687a4cfb5843d" +dependencies = [ + "phf_shared", + "rand 0.8.7", +] + +[[package]] +name = "phf_shared" +version = "0.11.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "67eabc2ef2a60eb7faa00097bd1ffdb5bd28e62bf39990626a582201b7a754e5" +dependencies = [ + "siphasher", +] + [[package]] name = "pin-project-lite" version = "0.2.17" @@ -1778,6 +1948,28 @@ dependencies = [ "zerocopy", ] +[[package]] +name = "proc-macro-error-attr3" +version = "3.0.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "34e4dd828515431dd6c4a030d26f7eaed7dd4778226e9d2bb968d65ca4ec3d4d" +dependencies = [ + "proc-macro2", + "quote", +] + +[[package]] +name = "proc-macro-error3" +version = "3.0.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "5ee475e440453418ff1335189eddf7101ba502cd818ab7ae04209bc83aa925aa" +dependencies = [ + "proc-macro-error-attr3", + "proc-macro2", + "quote", + "syn 2.0.119", +] + [[package]] name = "proc-macro2" version = "1.0.106" @@ -1963,6 +2155,12 @@ dependencies = [ "windows-sys 0.52.0", ] +[[package]] +name = "rle-decode-fast" +version = "1.0.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "3582f63211428f83597b51b2ddb88e2a91a9d52d12831f9d08f5e624e8977422" + [[package]] name = "ron" version = "0.7.1" @@ -2220,6 +2418,12 @@ version = "0.3.10" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "3a219298ac11a56ea9a6d2120044824d6f01aeb034955e7af7bc16858527deea" +[[package]] +name = "siphasher" +version = "1.0.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "8ee5873ec9cce0195efcb7a4e9507a04cd49aec9c83d0389df45b1ef7ba2e649" + [[package]] name = "slab" version = "0.4.12" diff --git a/assets/album.html b/assets/album.html index 0ab87bd..6521e9f 100644 --- a/assets/album.html +++ b/assets/album.html @@ -15,10 +15,12 @@ 返回首页

加载中...

-
- 正在加载... +
+
+ 正在加载... +
+
-
@@ -38,13 +40,6 @@ 来自: 相册名称
- - 1 / 1 - @@ -70,7 +65,19 @@ 0 KB
+
+ + 1 / 1 + +
+ diff --git a/assets/css/album.css b/assets/css/album.css index ad7ed89..7672fb4 100644 --- a/assets/css/album.css +++ b/assets/css/album.css @@ -88,25 +88,103 @@ body { .album-info { display: inline-flex; align-items: center; - gap: 16px; - color: var(--text-light); - font-size: 0.95rem; - padding: 10px 20px; - background: var(--bg-surface); - border-radius: var(--radius-full); - border: 1px solid var(--border-gray); + gap: 6px; + color: #475569; + font-size: 0.85rem; + font-weight: 500; + line-height: 1.2; + padding: 6px 14px; + background: linear-gradient(135deg, #f1f5f9 0%, #e2e8f0 100%); + border-radius: 20px; + border: 1px solid #cbd5e1; + cursor: default; + user-select: none; + height: auto; + min-height: 0; } -.album-info i { color: var(--secondary-blue); } +.album-info i { color: #64748b; } -/* ============ 相册标签区（新：分类着色） ============ */ -.album-meta-tags { +/* 照片数量 + 标签 合并到一行 */ +.album-header-row { display: flex; flex-wrap: wrap; + align-items: center; justify-content: center; + gap: 12px 16px; + margin: 12px auto 24px; + max-width: 1100px; +} + +.album-header-row .album-info { + margin: 0; + flex: 0 0 auto; + width: auto; + max-width: none; +} + +/* ============ 相册标签区（简洁版：与照片数量同行） ============ */ +.album-header-row .album-meta-tags { + display: flex; + flex-wrap: wrap; + justify-content: flex-start; + gap: 8px; + margin: 0; + max-width: none; + min-width: 0; +} + +/* 标签分组 */ +.tag-group { + margin-bottom: 14px; + background: var(--bg-surface); + border-radius: var(--radius-card); + border: 1px solid var(--border-gray); + padding: 12px 16px; + transition: all 0.2s ease; + box-shadow: var(--shadow-sm); +} + +.tag-group:hover { + box-shadow: var(--shadow-md); + border-color: #cbd5e1; +} + +.tag-group-header { + display: flex; + align-items: center; + gap: 8px; + margin-bottom: 10px; + padding-bottom: 8px; + border-bottom: 1px dashed var(--border-gray); + font-size: 0.85rem; + font-weight: 600; + color: var(--text-light); +} + +.tag-group-header i { + font-size: 0.9rem; + color: var(--secondary-blue); +} + +.tag-group-header span { + color: var(--text-light); +} + +.tag-group-count { + margin-left: auto; + background: var(--bg-hover); + padding: 2px 8px; + border-radius: 10px; + font-size: 0.75rem; + font-weight: 600; + color: var(--text-muted); +} + +.tag-group-chips { + display: flex; + flex-wrap: wrap; gap: 8px; - margin: 20px auto 32px; - max-width: 900px; } .meta-chip { @@ -114,13 +192,31 @@ body { align-items: center; gap: 6px; padding: 6px 14px; - border-radius: var(--radius-full); + border-radius: 20px; font-size: 0.85rem; font-weight: 500; + line-height: 1.2; cursor: pointer; transition: all 0.2s ease; user-select: none; border: 1px solid transparent; + position: relative; + overflow: hidden; +} + +.meta-chip::before { + content: ''; + position: absolute; + top: 0; + left: -100%; + width: 100%; + height: 100%; + background: linear-gradient(90deg, transparent, rgba(255,255,255,0.3), transparent); + transition: left 0.5s ease; +} + +.meta-chip:hover::before { + left: 100%; } .meta-chip:hover { @@ -130,13 +226,14 @@ body { .meta-chip:active { transform: translateY(0); } -.meta-chip i { font-size: 0.75rem; } +.meta-chip i { font-size: 0.75rem; opacity: 0.9; } /* 人物标签 */ .meta-chip-person { - background: linear-gradient(135deg, rgba(139, 92, 246, 0.1) 0%, rgba(236, 72, 153, 0.1) 100%); + background: linear-gradient(135deg, rgba(139, 92, 246, 0.12) 0%, rgba(236, 72, 153, 0.12) 100%); color: var(--accent-purple); - border-color: rgba(139, 92, 246, 0.2); + border-color: rgba(139, 92, 246, 0.25); + font-weight: 600; } .meta-chip-person:hover { @@ -147,54 +244,120 @@ body { /* 分类标签（按类别着色） */ .meta-chip-tag-style { - background: #f0f9ff; + background: linear-gradient(135deg, #f0f9ff 0%, #e0f2fe 100%); color: #0284c7; border-color: #bae6fd; } -.meta-chip-tag-style:hover { background: #0284c7; color: #fff; border-color: #0284c7; } +.meta-chip-tag-style:hover { background: linear-gradient(135deg, #0284c7 0%, #0369a1 100%); color: #fff; border-color: transparent; } .meta-chip-tag-outfit { - background: #fdf2f8; + background: linear-gradient(135deg, #fdf2f8 0%, #fce7f3 100%); color: #db2777; border-color: #fbcfe8; } -.meta-chip-tag-outfit:hover { background: #db2777; color: #fff; border-color: #db2777; } +.meta-chip-tag-outfit:hover { background: linear-gradient(135deg, #db2777 0%, #be185d 100%); color: #fff; border-color: transparent; } .meta-chip-tag-scene { - background: #f0fdf4; + background: linear-gradient(135deg, #f0fdf4 0%, #dcfce7 100%); color: #15803d; border-color: #bbf7d0; } -.meta-chip-tag-scene:hover { background: #15803d; color: #fff; border-color: #15803d; } +.meta-chip-tag-scene:hover { background: linear-gradient(135deg, #15803d 0%, #166534 100%); color: #fff; border-color: transparent; } .meta-chip-tag-body { - background: #fef3c7; + background: linear-gradient(135deg, #fef3c7 0%, #fde68a 100%); color: #b45309; border-color: #fde68a; } -.meta-chip-tag-body:hover { background: #b45309; color: #fff; border-color: #b45309; } +.meta-chip-tag-body:hover { background: linear-gradient(135deg, #b45309 0%, #92400e 100%); color: #fff; border-color: transparent; } .meta-chip-tag-mood { - background: #faf5ff; + background: linear-gradient(135deg, #faf5ff 0%, #f3e8ff 100%); color: #7c3aed; border-color: #e9d5ff; } -.meta-chip-tag-mood:hover { background: #7c3aed; color: #fff; border-color: #7c3aed; } +.meta-chip-tag-mood:hover { background: linear-gradient(135deg, #7c3aed 0%, #6d28d9 100%); color: #fff; border-color: transparent; } .meta-chip-tag-topic { - background: #eff6ff; + background: linear-gradient(135deg, #eff6ff 0%, #dbeafe 100%); color: #0369a1; border-color: #bfdbfe; } -.meta-chip-tag-topic:hover { background: #0369a1; color: #fff; border-color: #0369a1; } +.meta-chip-tag-topic:hover { background: linear-gradient(135deg, #0369a1 0%, #075985 100%); color: #fff; border-color: transparent; } + +/* jieba 提取的新分类 */ +.meta-chip-tag-noun { + background: linear-gradient(135deg, #eff6ff 0%, #dbeafe 100%); + color: #1d4ed8; + border-color: #bfdbfe; +} +.meta-chip-tag-noun:hover { background: linear-gradient(135deg, #1d4ed8 0%, #1e40af 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-proper { + background: linear-gradient(135deg, #e0e7ff 0%, #c7d2fe 100%); + color: #4338ca; + border-color: #c7d2fe; +} +.meta-chip-tag-proper:hover { background: linear-gradient(135deg, #4338ca 0%, #3730a3 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-location { + background: linear-gradient(135deg, #ecfdf5 0%, #d1fae5 100%); + color: #047857; + border-color: #a7f3d0; +} +.meta-chip-tag-location:hover { background: linear-gradient(135deg, #047857 0%, #065f46 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-organization { + background: linear-gradient(135deg, #fef9c3 0%, #fef08a 100%); + color: #a16207; + border-color: #fde68a; +} +.meta-chip-tag-organization:hover { background: linear-gradient(135deg, #a16207 0%, #854d0e 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-noun_verb { + background: linear-gradient(135deg, #fce7f3 0%, #fbcfe8 100%); + color: #be185d; + border-color: #f9a8d4; +} +.meta-chip-tag-noun_verb:hover { background: linear-gradient(135deg, #be185d 0%, #9d174d 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-adj { + background: linear-gradient(135deg, #f3e8ff 0%, #e9d5ff 100%); + color: #6d28d9; + border-color: #ddd6fe; +} +.meta-chip-tag-adj:hover { background: linear-gradient(135deg, #6d28d9 0%, #5b21b6 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-verb { + background: linear-gradient(135deg, #fed7aa 0%, #fdba74 100%); + color: #c2410c; + border-color: #fed7aa; +} +.meta-chip-tag-verb:hover { background: linear-gradient(135deg, #c2410c 0%, #9a3412 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-chinese { + background: linear-gradient(135deg, #f1f5f9 0%, #e2e8f0 100%); + color: #475569; + border-color: #cbd5e1; +} +.meta-chip-tag-chinese:hover { background: linear-gradient(135deg, #475569 0%, #334155 100%); color: #fff; border-color: transparent; } + +.meta-chip-tag-english { + background: linear-gradient(135deg, #f1f5f9 0%, #e2e8f0 100%); + color: #64748b; + border-color: #cbd5e1; +} +.meta-chip-tag-english:hover { background: linear-gradient(135deg, #64748b 0%, #475569 100%); color: #fff; border-color: transparent; } /* 默认标签 */ +.meta-chip-tag-default, .meta-chip-tag { background: var(--bg-hover); color: var(--text-light); border-color: var(--border-gray); } -.meta-chip-tag:hover { background: var(--text-light); color: #fff; border-color: var(--text-light); } +.meta-chip-tag-default:hover, +.meta-chip-tag:hover { background: linear-gradient(135deg, #64748b 0%, #475569 100%); color: #fff; border-color: transparent; } /* ============ 瀑布流布局（核心） ============ */ .images-grid { @@ -214,7 +377,6 @@ body { box-shadow: var(--shadow-sm); background: var(--bg-surface); transition: all 0.3s ease; - cursor: pointer; border: 1px solid var(--border-gray); break-inside: avoid; page-break-inside: avoid; @@ -233,6 +395,7 @@ body { display: block; object-fit: contain; transition: transform 0.4s ease; + cursor: pointer; } .image-item:hover img { @@ -249,6 +412,7 @@ body { align-items: center; gap: 8px; border-top: 1px solid var(--border-gray); + cursor: default; } .image-caption small { @@ -368,7 +532,7 @@ body { justify-content: center; opacity: 0; visibility: hidden; - transition: all 0.3s ease; + transition: all 0.25s ease; } .preview-modal.active { opacity: 1; visibility: visible; } @@ -377,44 +541,48 @@ body { position: absolute; top: 0; left: 0; width: 100%; height: 100%; - background: rgba(15, 23, 42, 0.9); - backdrop-filter: blur(8px); + background: rgba(0, 0, 0, 0.92); + backdrop-filter: blur(4px); } .preview-container { position: relative; - background: var(--bg-surface); - border-radius: var(--radius-lg); - box-shadow: 0 20px 60px rgba(0, 0, 0, 0.3); - max-width: 92vw; - max-height: 90vh; - width: 1200px; - height: 800px; + width: 100vw; + height: 100vh; + max-width: 100vw; + max-height: 100vh; display: flex; flex-direction: column; - transform: scale(0.92); - transition: transform 0.3s ease; + background: transparent; + border-radius: 0; + box-shadow: none; + transform: scale(1); } .preview-modal.active .preview-container { transform: scale(1); } +/* 顶部栏 - 浮在图片上方 */ .preview-header { + position: absolute; + top: 0; left: 0; right: 0; + z-index: 10; display: flex; justify-content: space-between; align-items: center; - padding: 16px 20px; - border-bottom: 1px solid var(--border-gray); + padding: 16px 24px; + background: linear-gradient(to bottom, rgba(0,0,0,0.7) 0%, transparent 100%); flex-shrink: 0; + border: none; } .preview-title h3 { - font-size: 1.1rem; + font-size: 1rem; font-weight: 600; - color: var(--text-dark); + color: #f1f5f9; margin: 0 0 4px 0; } -.preview-info { font-size: 0.85rem; color: var(--text-light); } +.preview-info { font-size: 0.8rem; color: #94a3b8; } .preview-actions { display: flex; @@ -423,51 +591,60 @@ body { } .preview-nav { - width: 36px; height: 36px; - border: 1px solid var(--border-gray); - border-radius: var(--radius); - background: var(--bg-surface); - color: var(--text-dark); + width: 40px; height: 40px; + border: 1px solid rgba(255,255,255,0.15); + border-radius: 50%; + background: rgba(255,255,255,0.08); + color: #f1f5f9; cursor: pointer; display: flex; align-items: center; justify-content: center; transition: all 0.2s ease; + backdrop-filter: blur(10px); } .preview-nav:hover:not(:disabled) { - background: var(--bg-hover); - border-color: var(--secondary-blue); - color: var(--secondary-blue); + background: rgba(255,255,255,0.2); + border-color: rgba(255,255,255,0.3); + color: #fff; } -.preview-nav:disabled { opacity: 0.4; cursor: not-allowed; } +.preview-nav:disabled { opacity: 0.3; cursor: not-allowed; } -.preview-counter { font-size: 0.85rem; color: var(--text-light); font-weight: 500; } +.preview-counter { + font-size: 0.9rem; + color: #cbd5e1; + font-weight: 500; + padding: 0 4px; +} .preview-close { - width: 36px; height: 36px; - border: none; - border-radius: var(--radius); - background: var(--bg-hover); - color: var(--text-dark); + width: 40px; height: 40px; + border: 1px solid rgba(255,255,255,0.15); + border-radius: 50%; + background: rgba(255,255,255,0.08); + color: #f1f5f9; cursor: pointer; display: flex; align-items: center; justify-content: center; transition: all 0.2s ease; + backdrop-filter: blur(10px); } -.preview-close:hover { background: #ef4444; color: #fff; } +.preview-close:hover { background: rgba(239, 68, 68, 0.8); color: #fff; border-color: rgba(239, 68, 68, 0.6); } +/* 图片内容区 - 占满全屏 */ .preview-content { - flex: 1; - padding: 0; + position: absolute; + top: 0; left: 0; right: 0; bottom: 0; + padding: 60px 80px; overflow: hidden; display: flex; align-items: center; justify-content: center; - background: #0f172a; + background: transparent; } .preview-image-container { @@ -482,8 +659,12 @@ body { .preview-image { max-width: 100%; max-height: 100%; + width: auto; + height: auto; object-fit: contain; - transition: opacity 0.3s ease; + transition: opacity 0.3s ease, transform 0.3s ease; + box-shadow: 0 25px 80px rgba(0, 0, 0, 0.5); + border-radius: 4px; } .preview-loading { @@ -494,54 +675,107 @@ body { flex-direction: column; align-items: center; gap: 8px; - color: var(--text-light); + color: #94a3b8; } -.preview-loading i { font-size: 1.8rem; color: var(--secondary-blue); } +.preview-loading i { font-size: 2rem; color: #3b82f6; } +/* 底部栏 - 浮在图片上方 */ .preview-footer { + position: absolute; + bottom: 0; left: 0; right: 0; + z-index: 10; display: flex; justify-content: space-between; align-items: center; - padding: 12px 20px; - border-top: 1px solid var(--border-gray); + gap: 20px; + padding: 16px 24px; + background: linear-gradient(to top, rgba(0,0,0,0.75) 0%, transparent 100%); flex-shrink: 0; + border: none; } -.preview-meta { display: flex; gap: 16px; } +.preview-meta { display: flex; gap: 16px; flex: 1; } .preview-meta .meta-item { display: flex; align-items: center; gap: 6px; - font-size: 0.85rem; - color: var(--text-light); + font-size: 0.8rem; + color: #cbd5e1; } -.preview-controls { display: flex; gap: 10px; } +/* 底部居中的切换控制栏 */ +.preview-center-controls { + display: flex; + align-items: center; + gap: 16px; + flex: 1; + justify-content: center; +} + +.preview-center-controls .preview-nav { + width: 40px; + height: 40px; + border: 1px solid rgba(255,255,255,0.2); + border-radius: 50%; + background: rgba(255,255,255,0.1); + color: #f1f5f9; + cursor: pointer; + display: flex; + align-items: center; + justify-content: center; + transition: all 0.2s ease; + backdrop-filter: blur(10px); + font-size: 0.9rem; +} + +.preview-center-controls .preview-nav:hover:not(:disabled) { + background: rgba(255,255,255,0.25); + border-color: rgba(255,255,255,0.4); + transform: scale(1.08); +} + +.preview-center-controls .preview-nav:disabled { + opacity: 0.3; + cursor: not-allowed; +} + +.preview-center-controls .preview-counter { + font-size: 0.9rem; + color: #e2e8f0; + font-weight: 500; + min-width: 60px; + text-align: center; +} + +.preview-controls { display: flex; gap: 10px; flex: 1; justify-content: flex-end; } .preview-action { - padding: 7px 14px; - border: 1px solid var(--border-gray); - border-radius: var(--radius); - background: var(--bg-surface); - color: var(--text-dark); + padding: 8px 16px; + border: 1px solid rgba(255,255,255,0.15); + border-radius: 20px; + background: rgba(255,255,255,0.08); + color: #f1f5f9; cursor: pointer; - font-size: 0.85rem; + font-size: 0.8rem; font-weight: 500; transition: all 0.2s ease; display: flex; align-items: center; gap: 6px; + backdrop-filter: blur(10px); } -.preview-action:hover { background: var(--bg-hover); } +.preview-action:hover { background: rgba(255,255,255,0.2); } -.preview-remove { border-color: #fecaca; color: #dc2626; } -.preview-remove:hover { background: #fef2f2; border-color: #fca5a5; } +.preview-action-close:hover { + background: rgba(239, 68, 68, 0.7); + border-color: rgba(239, 68, 68, 0.5); +} -.preview-action.favorited { border-color: #fbcfe8; color: #ec4899; background: #fdf2f8; } -.preview-action.favorited:hover { background: #fce7f3; border-color: #f9a8d4; } +.preview-action.favorited { border-color: rgba(236, 72, 153, 0.5); color: #f9a8d4; background: rgba(236, 72, 153, 0.15); } +.preview-action.favorited:hover { background: rgba(236, 72, 153, 0.25); } /* 暗色模式 */ @media (prefers-color-scheme: dark) { @@ -554,6 +788,15 @@ body { --text-light: #cbd5e1; --text-muted: #94a3b8; } - .preview-container { background: var(--bg-surface); } - .preview-nav, .preview-close, .preview-action { background: var(--bg-hover); color: var(--text-dark); border-color: var(--border-gray); } +} + +/* 移动端适配 */ +@media (max-width: 768px) { + .preview-content { padding: 40px 10px 70px; } + .preview-header, .preview-footer { padding: 12px 16px; } + .preview-footer { flex-wrap: wrap; gap: 12px; } + .preview-meta { order: 2; justify-content: center; } + .preview-center-controls { order: 1; width: 100%; } + .preview-controls { order: 3; justify-content: center; } + .preview-center-controls .preview-nav { width: 36px; height: 36px; } } diff --git a/assets/css/style.css b/assets/css/style.css index df13b87..9d90491 100644 --- a/assets/css/style.css +++ b/assets/css/style.css @@ -107,30 +107,56 @@ input { font-family: inherit; } .sidebar.collapsed .brand-text, .sidebar.collapsed .nav-item span, .sidebar.collapsed .sidebar-section-title span, +.sidebar.collapsed .sidebar-section-title .section-arrow, .sidebar.collapsed .sidebar-tags, -.sidebar.collapsed .collapse-btn span { +.sidebar.collapsed .sidebar-people, +.sidebar.collapsed .sidebar-section-content, +.sidebar.collapsed .sidebar-divider { opacity: 0; pointer-events: none; + display: none; +} + +.sidebar.collapsed .sidebar-section { + padding: var(--spacer-md) 0; + gap: var(--spacer-xs); } .sidebar.collapsed .sidebar-section-title { justify-content: center; + padding: var(--spacer-sm) 0; + margin-bottom: 0; +} + +.sidebar.collapsed .sidebar-section-group { + margin-bottom: var(--spacer-sm); } .sidebar.collapsed .nav-item { justify-content: center; padding-left: 0; padding-right: 0; + margin: 0 var(--spacer-xs); +} + +.sidebar.collapsed .sidebar-brand { + padding: var(--spacer-lg) 0; + justify-content: center; } .sidebar.collapsed .collapse-btn { justify-content: center; + padding: var(--spacer-md) 0; } .sidebar.collapsed .collapse-btn i { transform: rotate(180deg); } +.sidebar.collapsed .sidebar-nav { + padding: var(--spacer-md) var(--spacer-xs); +} + /* 品牌 Logo */ .sidebar-brand { display: flex; @@ -228,6 +254,7 @@ input { font-family: inherit; } display: flex; flex-direction: column; min-height: 0; + gap: var(--spacer-sm); } .sidebar-section::-webkit-scrollbar { @@ -238,6 +265,12 @@ input { font-family: inherit; } border-radius: 2px; } +/* 可折叠区块组 */ +.sidebar-section-group { + display: flex; + flex-direction: column; +} + .sidebar-section-title { display: flex; align-items: center; @@ -251,7 +284,49 @@ input { font-family: inherit; } margin-bottom: var(--spacer-sm); } -.sidebar-section-title i { +.sidebar-section-title.collapsible { + cursor: pointer; + border-radius: var(--radius); + transition: background var(--transition-fast); + user-select: none; +} + +.sidebar-section-title.collapsible:hover { + background: var(--bg-hover); + color: var(--text-light); +} + +.section-arrow { + margin-left: auto; + font-size: 0.7rem; + transition: transform var(--transition-base); + opacity: 0.6; +} + +.sidebar-section-group.collapsed .section-arrow { + transform: rotate(-90deg); +} + +.sidebar-section-content { + overflow: hidden; + max-height: 600px; + transition: max-height 0.28s cubic-bezier(0.4, 0, 0.2, 1), + opacity 0.2s ease, + margin 0.2s ease; + opacity: 1; + margin-bottom: var(--spacer-sm); +} + +.sidebar-section-group.collapsed .sidebar-section-content { + max-height: 0; + opacity: 0; + margin-bottom: 0; + transition: max-height 0.2s cubic-bezier(0.4, 0, 0.2, 1), + opacity 0.15s ease, + margin 0.15s ease; +} + +.sidebar-section-title i:first-child { width: 20px; text-align: center; font-size: 0.85rem; @@ -309,24 +384,95 @@ input { font-family: inherit; } color: var(--white); } +/* 侧边栏人物列表 */ +.sidebar-people { + display: flex; + flex-direction: column; + gap: var(--spacer-xs); + transition: opacity var(--transition-base); +} + +.sidebar-person { + display: flex; + align-items: center; + gap: var(--spacer-sm); + padding: var(--spacer-sm) var(--spacer-lg); + border-radius: var(--radius); + cursor: pointer; + transition: all var(--transition-fast); + font-size: 0.9rem; + color: var(--text-light); +} + +.sidebar-person:hover { + background: var(--bg-hover); + color: var(--text-dark); +} + +.sidebar-person.active { + background: rgba(139, 92, 246, 0.1); + color: var(--accent-purple); + font-weight: 500; +} + +.sidebar-person-avatar { + width: 28px; + height: 28px; + border-radius: 50%; + background: #6366f1; + color: var(--white); + display: flex; + align-items: center; + justify-content: center; + font-size: 0.8rem; + font-weight: 600; + flex-shrink: 0; +} + +.sidebar-person-info { + display: flex; + flex-direction: column; + flex: 1; + min-width: 0; +} + +.sidebar-person-name { + white-space: nowrap; + overflow: hidden; + text-overflow: ellipsis; + font-size: 0.9rem; +} + +.sidebar-person-count { + font-size: 0.75rem; + color: var(--text-muted); +} + +.sidebar-person.active .sidebar-person-count { + color: var(--accent-purple); + opacity: 0.8; +} + /* 侧边栏底部 */ .sidebar-footer { padding: var(--spacer-md); border-top: 1px solid var(--border-gray); flex-shrink: 0; + margin-top: auto; } .collapse-btn { display: flex; align-items: center; - gap: var(--spacer-md); + justify-content: center; width: 100%; - padding: var(--spacer-md) var(--spacer-lg); + padding: var(--spacer-md); border-radius: var(--radius); color: var(--text-light); - font-size: 0.9rem; + font-size: 1rem; font-weight: 500; transition: all var(--transition-fast); + cursor: pointer; } .collapse-btn:hover { @@ -338,12 +484,6 @@ input { font-family: inherit; } width: 20px; text-align: center; transition: transform var(--transition-base); - flex-shrink: 0; -} - -.collapse-btn span { - transition: opacity var(--transition-base); - white-space: nowrap; } /* ============ 主内容区 ============ */ diff --git a/assets/index.html b/assets/index.html index f4f5421..c14d769 100644 --- a/assets/index.html +++ b/assets/index.html @@ -37,22 +37,38 @@
+ +
+
+ + 热门人物 + +
+
+ +
+
+ -
-
+
+
热门标签 +
-
-
@@ -124,20 +140,6 @@
- - -
diff --git a/assets/js/album.js b/assets/js/album.js index 6eaf549..0ffa665 100644 --- a/assets/js/album.js +++ b/assets/js/album.js @@ -262,7 +262,7 @@ async function loadAlbum(albumId) { function updateAlbumInfo() { document.getElementById('album-title').textContent = currentAlbum.title; - let infoHtml = ` 共 ${images.length} 张图片`; + let infoHtml = `${images.length} 张图片`; document.getElementById('album-info').innerHTML = infoHtml; // 显示相册元数据（人名 + 标签） @@ -316,7 +316,7 @@ function getTagCategory(tag) { return 'default'; } -// 更新相册元数据（人名 + 标签）点击可搜索 +// 更新相册元数据（人名 + 标签）简洁版：只用颜色区分，不写分类名 function updateAlbumMeta() { const metaContainer = document.getElementById('album-meta-tags'); if (!metaContainer) return; @@ -337,7 +337,6 @@ function updateAlbumMeta() { if (currentAlbum.tags) { let tagList = []; try { - // 格式1: JSON数组 [["标签名", "分类"], ...] const parsed = JSON.parse(currentAlbum.tags); if (Array.isArray(parsed)) { parsed.forEach(item => { @@ -353,7 +352,6 @@ function updateAlbumMeta() { }); } } catch (e) { - // 格式2: 逗号分隔字符串 currentAlbum.tags.split(',').forEach(tag => { const t = tag.trim(); if (t) tagList.push({ text: t, category: getTagCategory(t) }); @@ -378,9 +376,8 @@ function updateAlbumMeta() { metaContainer.style.display = 'flex'; metaContainer.innerHTML = chips.map(chip => { const icon = chip.type === 'person' ? 'fa-user' : 'fa-tag'; - // 生成 CSS 类名：人物用 person，标签用 tag-{category} const cssClass = chip.type === 'person' ? 'meta-chip-person' : `meta-chip-tag-${chip.category || 'default'}`; - return ` + return ` ${chip.text} `; }).join(''); @@ -446,7 +443,7 @@ function displayImages() { // 检查图片是否已被收藏 isImageFavorited(image.id || image.original_url); return ` -
+
${image.width ? `${image.width}×${image.height}` : ''} diff --git a/assets/js/app.js b/assets/js/app.js index c2bd661..5edca43 100644 --- a/assets/js/app.js +++ b/assets/js/app.js @@ -109,6 +109,7 @@ document.addEventListener('DOMContentLoaded', function () { initSearch(); initInfiniteScroll(); initLazyLoading(); // 初始化图片懒加载 + initSidebarSections(); // 初始化侧边栏区块折叠状态 // 检查URL中是否有search参数 const urlParams = new URLSearchParams(window.location.search); @@ -539,7 +540,10 @@ async function performServerSearch(searchTerm) { // 显示服务器搜索结果 displayAlbums(serverResults); - document.getElementById('stats').innerHTML = `在全站找到 ${serverResults.length} 个包含 "${searchTerm}" 的相册`; + const oldStats = document.getElementById('stats'); + if (oldStats) { + oldStats.innerHTML = `在全站找到 ${serverResults.length} 个包含 "${searchTerm}" 的相册`; + } // 初始化图片懒加载 initLazyLoading(); @@ -636,11 +640,11 @@ function updateStats(albums) { const oldStats = document.getElementById('stats'); if (oldStats) oldStats.innerHTML = `共发现 ${albums.length} 个相册，包含 ${totalImages} 张图片`; - // 获取人物数量 - fetch(`${API_BASE}/recommend/popular-persons?limit=1`).then(r => r.ok && r.json()).then(data => { - if (data && data.length > 0 && data[0].total) { + // 获取人物数量（用 metadata stats API） + fetch(`${API_BASE}/metadata/stats`).then(r => r.ok && r.json()).then(data => { + if (data && data.total_persons !== undefined) { const statPeople = document.getElementById('stat-people'); - if (statPeople) statPeople.textContent = data[0].total.toLocaleString(); + if (statPeople) statPeople.textContent = data.total_persons.toLocaleString(); } }).catch(() => {}); } @@ -799,14 +803,20 @@ function openAlbum(albumId) { // 随机打开相册 function openRandomAlbum() { - // 添加随机动画效果 - const randomButton = document.querySelector('.random-button'); - randomButton.style.transform = 'scale(0.95)'; - randomButton.disabled = true; - - // 显示加载状态 - const originalText = randomButton.textContent; - randomButton.textContent = '随机加载中...'; + // 添加随机动画效果 - 兼容侧边栏导航项 + const navItems = document.querySelectorAll('.nav-item'); + let randomButton = null; + navItems.forEach(item => { + if (item.textContent.includes('随机')) randomButton = item; + }); + // 也试试顶部操作按钮 + if (!randomButton) { + randomButton = document.querySelector('.action-btn[onclick*="openRandomAlbum"]'); + } + + if (randomButton) { + randomButton.style.transform = 'scale(0.95)'; + } // 调用新的随机相册API fetch(`${API_BASE}/album/random`) @@ -821,22 +831,22 @@ function openRandomAlbum() { }) .then(randomAlbum => { // 恢复按钮状态 - setTimeout(() => { - randomButton.style.transform = ''; - randomButton.disabled = false; - randomButton.textContent = originalText; - }, 200); + if (randomButton) { + setTimeout(() => { + randomButton.style.transform = ''; + }, 200); + } // 跳转到随机相册 window.location.href = `/album?id=${randomAlbum.id}`; }) .catch(error => { // 恢复按钮状态 - setTimeout(() => { - randomButton.style.transform = ''; - randomButton.disabled = false; - randomButton.textContent = originalText; - }, 200); + if (randomButton) { + setTimeout(() => { + randomButton.style.transform = ''; + }, 200); + } alert(error.message); }); @@ -847,6 +857,11 @@ function showFavorites() { // 设置当前在收藏页 isInFavoritesPage = true; + // 切换侧边栏选中状态 + document.querySelectorAll('.nav-item').forEach(item => item.classList.remove('active')); + const favNav = document.querySelector('.nav-item[data-view="favorites"]'); + if (favNav) favNav.classList.add('active'); + // 从后端API获取收藏列表 fetch(`${API_BASE}/favorites`) .then(response => { @@ -919,8 +934,10 @@ function displayFavorites(favoriteImages) { `} `; - // 更新统计信息 - 显示收藏数量 - document.getElementById('stats').innerHTML = `
+ // 更新统计信息 - 显示收藏数量（兼容新旧版布局） + const statsEl = document.getElementById('stats'); + if (statsEl) { + statsEl.innerHTML = `
@@ -940,6 +957,7 @@ function displayFavorites(favoriteImages) {
`; + } // 初始化图片懒加载，确保收藏的图片被加载 initLazyLoading(); @@ -959,8 +977,12 @@ function formatDate(dateString) { function backToHome() { // 重置收藏页标记 isInFavoritesPage = false; - // 重新加载首页相册 - loadAlbums(true); + // 切换导航高亮到首页 + document.querySelectorAll('.nav-item').forEach(item => { + item.classList.toggle('active', item.dataset.view === 'home'); + }); + // 清除筛选 + clearFilter(); } // 全局变量存储收藏列表和当前预览索引 @@ -1241,7 +1263,9 @@ async function loadDiscoverData() { ]); if (tagsResp.ok) { - const tags = await tagsResp.json(); + const tagsData = await tagsResp.json(); + // 兼容两种格式：直接数组 或 { total, tags } + const tags = Array.isArray(tagsData) ? tagsData : (tagsData.tags || tagsData.items || []); if (tags && tags.length > 0) { renderTagCloud(tags); } @@ -1300,7 +1324,24 @@ function renderTagCloud(tags) { // 渲染热门人物（横向滚动条） function renderPeopleCloud(people) { - // 新版：横向滚动条 + // 侧边栏人物列表 + const sidebarPeople = document.getElementById('sidebar-people'); + if (sidebarPeople && people && people.length) { + sidebarPeople.innerHTML = people.slice(0, 25).map(p => { + const name = p.name || p.display_name || ''; + const albumCount = p.total_albums || 0; + const isActive = currentFilter && currentFilter.type === 'person' && currentFilter.value === name; + return `
+
${name.charAt(0)}
+
+
${name}
+
${albumCount} 相册
+
+
`; + }).join(''); + } + + // 横向滚动条 const strip = document.getElementById('people-strip'); const scroll = document.getElementById('people-scroll'); if (strip && scroll && people && people.length) { @@ -1339,6 +1380,32 @@ function renderPeopleCloud(people) { } } +// 侧边栏区块折叠/展开 +function toggleSidebarSection(section) { + const group = document.querySelector(`.sidebar-section-group [data-section="${section}"]`)?.closest('.sidebar-section-group'); + if (!group) return; + group.classList.toggle('collapsed'); + // 保存状态到 localStorage + try { + const state = JSON.parse(localStorage.getItem('sidebarSections') || '{}'); + state[section] = group.classList.contains('collapsed'); + localStorage.setItem('sidebarSections', JSON.stringify(state)); + } catch(e) {} +} + +// 初始化侧边栏区块折叠状态 +function initSidebarSections() { + try { + const state = JSON.parse(localStorage.getItem('sidebarSections') || '{}'); + Object.keys(state).forEach(section => { + if (state[section]) { + const group = document.querySelector(`.sidebar-section-group [data-section="${section}"]`)?.closest('.sidebar-section-group'); + if (group) group.classList.add('collapsed'); + } + }); + } catch(e) {} +} + // 按标签筛选 async function filterByTag(tag) { currentFilter = { type: 'tag', value: tag }; @@ -1372,6 +1439,15 @@ async function filterByPerson(personName) { currentFilter = { type: 'person', value: personName }; showFilterStatus(`人物: ${personName}`); + // 刷新侧边栏人物高亮 + try { + const peopleResp = await fetch(`${API_BASE}/recommend/popular-persons?limit=30`); + if (peopleResp.ok) { + const peopleData = await peopleResp.json(); + renderPeopleCloud(peopleData); + } + } catch(e) {} + try { showSkeleton(6); const response = await fetch(`${API_BASE}/people/search?q=${encodeURIComponent(personName)}`); @@ -1489,6 +1565,8 @@ function switchView(view) { if (view === 'home') { backToHome(); + } else if (view === 'favorites') { + showFavorites(); } } diff --git a/docs/pipeline-architecture.md b/docs/pipeline-architecture.md new file mode 100644 index 0000000..df45331 --- /dev/null +++ b/docs/pipeline-architecture.md @@ -0,0 +1,226 @@ +# 流云图库 — 元数据管线架构文档 + +> 版本：v2.0 · 更新：2026-07-19 + +--- + +## 总览 + +流云图库是一个 Rust 编写的图片管理应用，核心功能是从相册标题中自动提取**人物名称**和**分类标签**，并建立人物-相册关联关系。所有提取和关联逻辑由一套三阶段的**处理管线（Pipeline）** 驱动。 + +``` +相册标题（如 "水淼Aqua cosplay Zani – Wuthering Waves 69 photos"） + │ + ▼ + ┌─ Pipeline ──────────────────────────────────────┐ + │ Phase 1 全量解析 提取人名 + 提取标签 │ + │ Phase 2 人物合并 按中文核心合并重复人物 │ + │ Phase 3 缓存刷新 重新加载内存缓存 │ + └──────────────────────────────────────────────────┘ + │ + ├──→ albums 表（person_name_original, tags 等） + ├──→ persons 表（人物主表） + ├──→ person_aliases 表（别名关联） + └──→ album_tags 表（标签关联） +``` + +--- + +## 启动流程与触发时机 + +应用启动后，管线任务按顺序串行执行三个阶段： + +``` +启动管线（单次，后台任务 A） + │ + ├── ① 扫描导入新相册（import） + ├── ② 全量解析 + 人物合并（run_pipeline） + └── ③ 刷新 AlbumsCache +``` + +- **启动管线**：import → run_pipeline → refreshCache，三件事**顺序串行**，不并发 +- **定时管线**：每 30 分钟触发一次，执行 run_pipeline → refreshCache（不重复 import） +- **API 触发**：POST /api/pipeline/run 可手动触发 run_pipeline + +--- + +## Phase 1：全量解析 + +**入口函数：** `phase_parse_all(db)` → `(person_updated, tag_updated)` + +对数据库中**每一个相册**独立执行以下流程： + +``` +for each album: + │ + ├── jieba 分词（词性标注） + │ ├──→ 提取推荐人名（词性 nr/nrfg/nrt/nz） + │ └──→ 提取推荐标签（词性 n/nz/ns/eng/x） + │ + ├── 确定最佳人名 + │ ├── 有人物库匹配？→ 用库匹配结果 + │ └── 否则 → 用 jieba 推荐 + │ + ├── 写入 DB（强制覆盖，不比对旧值） + │ + ├── 提取标签 + │ ├── 三步提取：分类词库 → jieba → 正则兜底 + │ ├── unique_tags() 去重+排序+坏标签过滤 + │ └── 写入 DB（强制覆盖，不比对旧值） + │ + └── 下一个相册 +``` + +### 人名提取策略 + +按优先级依次尝试： + +1. **人物库精确匹配** — `person_merger::match_title()`，在已有 persons 表中查询标题子串 +2. **jieba 词性标注** — `extract_person_from_tags()`，优先取 nr（人名）类词 +3. **标题模式匹配** — `extract_person_name_from_title()`，尝试 cosplay 格式、分隔符切分、正则扫描 + +### 人名处理规则 + +- **禁止中英混杂**：人名主体为纯中文或纯英文（中文优先级更高） +- 中英文混合（如"水淼Aqua"）由 `split_person_name_with_aliases` 拆分为：主名="水淼" + 别名="Aqua" +- 别名保存到 `person_aliases` 表，用于后续搜索匹配 + +### 标签分类体系 + +| 分类 | 标识 | 示例词 | +|------|------|--------| +| 风格 | style | 日系、韩系、复古、甜美、sexy | +| 服装 | outfit | 比基尼、旗袍、制服、丝袜 | +| 场景 | scene | 室内、户外、街拍、浴室、beach | +| 身体 | body | 美腿、翘臀、玉足、美背 | +| 情绪 | mood | 清纯、诱惑、唯美、冷艳 | +| 主题 | topic | 写真、私房、旅拍、cosplay | +| 名词 | noun | jieba 提取的普通名词 | +| 专名 | proper | jieba 提取的专有名词 | +| 地名 | location | jieba 提取的地名 | +| 其他 | other | 正则兜底提取的剩余关键词 | + +### 标签过滤链（从外到内共 5 层） + +``` +1. is_punctuation_only() → 检测不含任何字/字母/数字，则跳过 +2. 停用词表过滤 → photos、cosplay、美女、模特 等 50+ 词 +3. 人名排除 → 排除标签提取时确定的人名 +4. 词性排除 → 形容词(a)、动词(v)、名动词(vn)、机构(nt) 等不产出标签 +5. unique_tags() 最终过滤 → 调用 is_bad_tag() 二次确认 +``` + +### 强制覆盖写入 + +每次管线执行都是**完全重新解析**，提取到人名就写入覆盖、提取到标签就写入覆盖。不做"新旧好坏"比对： + +- **人名**：jieba/fuzzy 提取到 → 直接写入 `albums.person_name_*`，创建/更新 `persons` 记录及别名关联 +- **标签**：经过 5 层过滤链后仍有结果 → 直接写入 `albums.tags`，重建 `album_tags` 表 + +这样每次启动或定时触发时，标签和人名都会用最新的解析逻辑重新计算并强制覆盖，解析逻辑改进后老数据自动受益。 + +--- + +## Phase 2：人物合并 + +**入口函数：** `phase_merge_persons(db)` → `merged_count` + +按**中文核心**自动合并重复人物： + +``` +1. 扫描所有 persons，提取其中文核心（开头的连续中文字符） +2. 将中文核心相同的 person 分为一组 +3. 每组选相册数最多的作为目标（dst） +4. 其余合并到目标： + - 移动所有相册关联 + - 移动所有别名 + - 源人物原名也作为目标别名 + - 删除源人物 +5. 更新目标的统计计数 +``` + +示例：`杨晨晨Yome` 和 `杨晨晨Sugar` 的中文核心都是 "杨晨晨" → 合并为一个人物 + +--- + +## 核心数据表结构 + +### albums 表（关键字段） + +| 字段 | 类型 | 说明 | +|------|------|------| +| id | TEXT | UUID | +| title | TEXT | 相册标题，来源是文件夹名或 index.json | +| person_name_original | TEXT | 提取的原始人名（可能带英文后缀） | +| person_name_cleaned | TEXT | 清洗后的人名（仅中文主名） | +| tags | TEXT | JSON 数组，如 `["比基尼","室内","清纯"]` | +| all_words | TEXT | JSON 数组，如 `[["比基尼","outfit"],["室内","scene"]]` | +| processed_at | TEXT | 最后处理时间 | + +### persons 表 + +| 字段 | 类型 | 说明 | +|------|------|------| +| id | INTEGER | 自增主键 | +| name | TEXT | 人物名（通常是中文主名） | +| display_name | TEXT | 显示名 | +| total_albums | INTEGER | 关联相册数 | +| total_images | INTEGER | 关联图片数 | + +### person_aliases 表 + +| 字段 | 类型 | 说明 | +|------|------|------| +| id | INTEGER | 自增主键 | +| person_id | INTEGER | 关联 persons.id | +| alias_type | TEXT | 别名类型（english/nickname/chinese/original） | +| alias_name | TEXT | 别名文本 | + +### album_tags 表 + +| 字段 | 类型 | 说明 | +|------|------|------| +| id | INTEGER | 自增主键 | +| album_id | TEXT | 关联 albums.id | +| tag | TEXT | 标签文本 | + +--- + +## 定时任务 + +每 30 分钟触发一次，顺序执行： + +``` +1. run_pipeline() → Phase 1（全量解析）+ Phase 2（人物合并） +2. 刷新 AlbumsCache（内存缓存，为 API 提供快速查询） +``` + +--- + +## API 接口 + +| 方法 | 路径 | 说明 | +|------|------|------| +| POST | /api/pipeline/run | 手动触发一次完整管线 | + +返回： +```json +{ + "success": true, + "persons_updated": 12, + "tags_updated": 45, + "persons_merged": 3, + "message": "管线执行完毕：更新人名 12 个，更新标签 45 个，合并人物 3 人" +} +``` + +--- + +## 关键设计原则 + +1. **幂等性**：管线可反复执行，每次都是完整重新解析 + 强制覆盖写入 +2. **渐进改进**：解析逻辑改进后，老数据在下次管线执行时自动受益 +3. **强制覆盖**：提取到即写入，不做新旧比对，确保数据始终使用最新逻辑 +4. **中文优先**：人名提取优先中文，中英文混合被拆分为主名+别名 +5. **多层过滤**：标签经过 5 层过滤链，确保质量 +6. **纯中文/纯英文**：人名主体不允许中英混杂 diff --git a/docs/pipeline-spec.md b/docs/pipeline-spec.md new file mode 100644 index 0000000..b00064d --- /dev/null +++ b/docs/pipeline-spec.md @@ -0,0 +1,124 @@ +# 元数据管线 — 开发规格 + +> 2026-07-19 + +--- + +## 1. 需求概述 + +系统在启动和运行时，需要自动完成三个任务： + +| # | 任务 | 说明 | +|---|------|------| +| 1 | **新相册解析** | 扫描磁盘导入的新相册 → 提取人名+标签 → 写入 DB | +| 2 | **老数据重解析** | 已有数据的相册用最新逻辑重新提取，强制覆盖 | +| 3 | **人物合并** | 按中文核心合并同一人的不同记录 | + +这三件事通过**同一个管道**完成，管道在**启动时跑一次** + **定时每 30 分钟跑一次**。 + +--- + +## 2. 启动流程（时序） + +当前问题：import 和 pipeline 是两个并发的 `tokio::spawn`，pipeline 固定等 3s，import 可能耗时更长，导致刚导入的相册没被解析。 + +**目标时序：** + +``` +服务启动 + │ + ├── DB 初始化 + │ + ├─── 后台任务 A：启动管线（单次）─────┐ + │ ├── ① 扫描导入新相册 │ + │ ├── ② Phase 1：全量解析（强制覆盖） │ + │ ├── ③ Phase 2：人物合并 │ + │ └── ④ 刷新内存缓存 │ + │ │ + ├─── 后台任务 B：定时管线（每 30 分钟）──┤ + │ ├── ① Phase 1：全量解析（强制覆盖） │ + │ ├── ② Phase 2：人物合并 │ + │ └── ③ 刷新内存缓存 │ + │ │ + └─── HTTP 服务启动 │ + │ +注：后台任务 A 和 B 互斥，不会同时执行 +``` + +### 变化要点 + +- **启动管线 = import + parse + merge + refreshCache** 依次串行 +- **定时管线 = parse + merge + refreshCache**（不重复 import） +- 管线内部顺序固定不可并发 + +--- + +## 3. 管线内部结构 + +``` +run_pipeline(db, with_import: bool) + │ + ├── [with_import=true] 扫描导入新相册 + │ + ├── Phase 1：全量解析 + │ for each album: + │ ├── jieba 分词 → 候选人名 + 候选标签 + │ ├── 人物库匹配 / jieba → 最佳人名 + │ ├── 强制写入 albums.person_name_* + persons + aliases + │ ├── 标签三步提取（词库→jieba→正则） + │ ├── unique_tags(去重+排序+坏标签过滤) + │ └── 强制写入 albums.tags + album_tags + │ + ├── Phase 2：人物合并 + │ ├── 提取所有人名的中文核心 + │ ├── 分组 → 每组选最多相册的为目标 + │ └── 合并相册关联 / 别名，删除源 + │ + └── Phase 3：刷新缓存 + └── 重新加载 albums 到 AlbumsCache +``` + +--- + +## 4. 模块职责 + +| 模块 | 文件 | 职责 | +|------|------|------| +| Pipeline 引擎 | `engine/pipeline.rs` | `phase_parse_all()`、`phase_merge_persons()`、`run_pipeline()` | +| 标题解析 | `engine/title_parser.rs` | jieba 分词、人名提取、中英文拆分、别名生成 | +| 标签提取 | `engine/tag_extractor.rs` | 分类词库匹配、jieba 名词提取、标点/停用词/坏标签过滤 | +| 人物合并 | `engine/person_merger.rs` | 中文核心提取、分组分析、合并执行 | +| 导入 | `services/import.rs` | 扫描磁盘目录、创建相册和图片记录 | +| DB | `db/metadata.rs` | persons、aliases、album_tags 表的读写 | +| 缓存 | `cache/mod.rs` | AlbumsCache + TagCache + PersonCache | + +--- + +## 5. 关键设计约束 + +1. **强制覆盖**：每次管线执行都是完整重解，直接覆盖 DB 数据 +2. **幂等**：管线可反复执行，不影响正确性 +3. **串行**：管线内部各阶段顺序执行，不并发 +4. **多级过滤**：标签经过 5 层过滤才入库，确保质量 +5. **中文优先**：人名优先中文，中英混杂拆为"主名+别名" +6. **纯中/纯英**：人名主体不允许中英混杂 + +--- + +## 6. 实现清单 + +### 6.1 Pipeline 引擎 + +- [x] `phase_parse_all()` — 全量解析 +- [x] `phase_merge_persons()` — 人物合并 +- [x] `phase_refresh_cache()` — 刷新缓存（在 main.rs 实现） +- [x] `run_pipeline(db)` — 整合解析+合并 + +### 6.2 main.rs 启动逻辑 + +- [x] 后台任务 A：顺序执行 import → run_pipeline → cache refresh +- [x] 后台任务 B：定时每 30 分钟 run_pipeline → cache refresh + +### 6.3 API + +- [x] POST /api/pipeline/run → 全量解析 + 人物合并 diff --git a/src/db/metadata.rs b/src/db/metadata.rs index 989d366..dfaf31f 100644 --- a/src/db/metadata.rs +++ b/src/db/metadata.rs @@ -171,6 +171,13 @@ impl Database { sqlx::query("DELETE FROM person_albums WHERE person_id = ?").bind(person_id).execute(&self.pool).await.ok(); } + /// 彻底删除一个人物（包括别名、相册关联、自身） + pub async fn delete_person_completely(&self, person_id: i64) { + sqlx::query("DELETE FROM person_aliases WHERE person_id = ?").bind(person_id).execute(&self.pool).await.ok(); + sqlx::query("DELETE FROM person_albums WHERE person_id = ?").bind(person_id).execute(&self.pool).await.ok(); + sqlx::query("DELETE FROM persons WHERE id = ?").bind(person_id).execute(&self.pool).await.ok(); + } + /// 把 src_id 的所有相册关联、别名移动到 dst_id，然后删除 src_id pub async fn merge_person_into(&self, src_id: i64, dst_id: i64) { // 1. 移动相册关联 @@ -216,6 +223,14 @@ impl Database { .bind(person_name_original).bind(person_name_cleaned).bind(album_id).execute(&self.pool).await.ok(); } + /// 清空该人物所有关联相册的 person 字段（当该人物被确认为非人物后调用） + pub async fn update_album_person_by_person_id(&self, person_id: i64) { + sqlx::query( + "UPDATE albums SET person_name_original = NULL, person_name_cleaned = NULL, \ + processed_at = datetime('now') WHERE id IN (SELECT album_id FROM person_albums WHERE person_id = ?)" + ).bind(person_id).execute(&self.pool).await.ok(); + } + pub async fn update_album_tags(&self, album_id: &str, tags_json: &str, all_words_json: &str) { sqlx::query("UPDATE albums SET tags = ?, all_words = ?, processed_at = datetime('now') WHERE id = ?") .bind(tags_json).bind(all_words_json).bind(album_id).execute(&self.pool).await.ok(); diff --git a/src/engine/person_merger.rs b/src/engine/person_merger.rs index d66ef3a..8f7d31e 100644 --- a/src/engine/person_merger.rs +++ b/src/engine/person_merger.rs @@ -296,13 +296,24 @@ pub async fn execute_merge(db: &Database) -> MergeResult { // 其余的合并到目标 for (src_id, src_name, _, _) in &items[1..] { println!("[合并] {} (id={}) → {} (id={})", src_name, src_id, dst_name, dst_id); + + // 合并前统计 src 的别名数 + let src_aliases = db.get_person_aliases(*src_id).await; + let alias_count_before = src_aliases.len(); + db.merge_person_into(*src_id, dst_id).await; + + // 合并后统计新增的别名数 (= src 别名数 + src 原名) result.persons_merged += 1; + // src 原名 + src 的 display_name + src 原有的别名 + let moved = alias_count_before + 1; // +1 因为原名会成为 merged_original 别名 + result.aliases_moved += moved; } } if result.persons_merged > 0 { - println!("[合并] 完成：分析 {} 组，合并 {} 个人物", result.groups_analyzed, result.persons_merged); + println!("[合并] 完成：分析 {} 组，合并 {} 个人物，迁移 {} 个别名", + result.groups_analyzed, result.persons_merged, result.aliases_moved); } result diff --git a/src/engine/pipeline.rs b/src/engine/pipeline.rs index b7845d4..c7067d9 100644 --- a/src/engine/pipeline.rs +++ b/src/engine/pipeline.rs @@ -1,91 +1,473 @@ //! 处理管线 — 批量处理相册元数据 +//! +//! 核心架构：脏集追踪 + 增量解析 +//! +//! Phase 0 清洗标签标点 +//! Phase A1 检测标签中的人名并直接提取到 person 字段 → 加入脏集 +//! Phase A2 清理非人物 → 加入脏集 +//! Phase A3 收集新相册/半成品 → 加入脏集 +//! Phase 1 解析脏集相册（增量解析，不扫全库） +//! Phase 2 人物合并 +//! Phase B 后校验标签中的人名 +//! +//! 脏集（DirtySet）贯穿各阶段，每个阶段发现的"需要重新解析"的相册都会加入脏集， +//! 最终只处理脏集里的相册，而非全库扫描，大幅提升效率。 use crate::db::Database; use crate::engine::title_parser; use crate::engine::tag_extractor; use crate::engine::person_merger; +use std::collections::HashSet; -/// 处理所有相册的元数据：提取人名 + 标签（单次 jieba 分词） -/// 一阶段：一次调用 extract_person_and_tags 同时获取人名和标签 -pub async fn process_all_albums(db: &Database) -> (usize, usize) { +// ════════════════════════════════════════════════ +// 脏集 — 追踪需要重新解析的相册 ID +// ════════════════════════════════════════════════ + +/// 脏集：记录哪些相册需要重新解析 +pub struct DirtySet { + ids: HashSet, +} + +impl DirtySet { + pub fn new() -> Self { Self { ids: HashSet::new() } } + pub fn add(&mut self, id: &str) { self.ids.insert(id.to_string()); } + pub fn extend(&mut self, ids: impl IntoIterator) { self.ids.extend(ids); } + pub fn contains(&self, id: &str) -> bool { self.ids.contains(id) } + pub fn len(&self) -> usize { self.ids.len() } + pub fn is_empty(&self) -> bool { self.ids.is_empty() } + pub fn into_inner(self) -> HashSet { self.ids } +} + +// ════════════════════════════════════════════════ +// Phase 0：清洗已有标签中的头尾标点 +// ════════════════════════════════════════════════ + +pub async fn phase_clean_tags(db: &Database) -> usize { let albums = db.get_all_albums_full().await; - let mut person_count = 0; - let mut tag_count = 0; - - // 已有人物库时优先用匹配（更准确） - let patterns = person_merger::collect_person_patterns(db).await; - let use_patterns = !patterns.is_empty(); + let mut cleaned_count = 0usize; + let mut removed_tags = HashSet::new(); for album in &albums { - let title = album.title.as_deref().unwrap_or(""); - if title.is_empty() { continue; } + let tags_json = match album.tags.as_deref() { + Some(t) if !t.is_empty() && t != "[]" => t, + _ => continue, + }; + let old_tags: Vec = match serde_json::from_str(tags_json) { + Ok(t) => t, Err(_) => continue, + }; + if old_tags.is_empty() { continue; } - // 判断是否需要处理 - let need_person = album.person_name_original.is_none() - || album.person_name_original.as_deref() == Some(""); - let need_tags = album.tags.as_deref().unwrap_or("").is_empty() - || album.tags.as_deref() == Some("[]"); - - if !need_person && !need_tags { continue; } - - // 一次 jieba 分词，同时获取人名 + 标签 - let (person_from_jieba, tags_from_jieba) = title_parser::extract_person_and_tags(title); - - // ── 处理人名 ── - if need_person { - let person_name = if use_patterns { - // 已有人物库，优先用匹配；匹配不到再回退到 jieba - person_merger::match_title(title, &patterns) - .map(|(_, name, _)| name) - .or(person_from_jieba) - } else { - person_from_jieba + let mut changed = false; + let mut new_tags = Vec::new(); + for tag in old_tags { + let cleaned = match tag_extractor::clean_tag(&tag) { + Some(t) => t, + None => { removed_tags.insert(tag); changed = true; continue; } }; + if cleaned != tag { removed_tags.insert(tag); changed = true; } + new_tags.push(cleaned); + } + if !changed { continue; } + cleaned_count += 1; - if let Some(person_name) = person_name { - // 拆分中文主名 + 别名 - let (primary_name, aliases) = title_parser::split_person_name_with_aliases(&person_name); - let final_name = if primary_name.is_empty() { person_name.clone() } else { primary_name }; - db.update_album_person(&album.id, &person_name, &final_name).await; + let tags_json_new = serde_json::to_string(&new_tags).unwrap_or_else(|_| "[]".to_string()); + let all_words_json = album.all_words.as_deref().unwrap_or("[]"); + let all_words_new: Vec<[String; 2]> = serde_json::from_str::>(all_words_json) + .unwrap_or_default() + .into_iter() + .filter_map(|pair| { let cleaned = tag_extractor::clean_tag(&pair[0])?; Some([cleaned, pair[1].clone()]) }) + .collect(); + let all_words_json_new = serde_json::to_string(&all_words_new).unwrap_or_else(|_| "[]".to_string()); + db.update_album_tags(&album.id, &tags_json_new, &all_words_json_new).await; + } - // 创建人员（先查是否已有同名） - let pid = db.insert_person(&final_name, None).await; - db.insert_person_album(pid, &album.id).await; + if cleaned_count > 0 { + println!("[管线·Phase0] 标签清洗完成：清洗 {} 个相册，移除标签: {:?}", + cleaned_count, removed_tags.iter().collect::>()); + } else { + println!("[管线·Phase0] 标签清洗完成：无需要清洗的标签"); + } + cleaned_count +} - // 保存所有别名 - for (alias_type, alias_name) in &aliases { - if !alias_name.is_empty() && alias_name != &final_name { - db.insert_person_alias(pid, alias_type, alias_name).await; - } - } +// ════════════════════════════════════════════════ +// Phase A1：检测标签中的人名并直接提取到 person 字段 +// ════════════════════════════════════════════════ - db.update_person_counts(pid).await; - person_count += 1; +/// Phase A1：检测标签中可能的人名并直接提取到 person_name +/// +/// 核心变化：不再只标记脏集等待 Phase 1，而是直接从 tags 中提取人名写入 person 字段。 +/// 因为 tags 中的人名比从 title 重解析更直接、更准确。 +/// +/// 两层检测 + 最终写入： +/// 1. 逐个标签 — 标签是已知人物或像人名 → 收集为候选人名（中文加权） +/// 2. jieba 合并（空格分隔）— 发现跨标签拆分的人名（["杨晨","晨Yome"]） +/// 3. 最佳候选人名 → 写入 person_name，从标签中移除关联标签 +/// +/// 返回 (脏集, 从标签中移除的人名标签数) +pub async fn phase_detect_person_in_tags(db: &Database) -> (DirtySet, usize) { + let albums = db.get_all_albums_full().await; + let mut dirty = DirtySet::new(); + let mut person_like_removed = 0usize; + + let known_persons: HashSet = db.get_all_person_names().await + .into_iter().map(|(n, _)| n.to_lowercase()).collect(); + + for album in &albums { + let tags_json = match album.tags.as_deref() { + Some(t) if !t.is_empty() && t != "[]" => t, + _ => continue, + }; + let old_tags: Vec = match serde_json::from_str(tags_json) { + Ok(t) => t, Err(_) => continue, + }; + + let has_person = album.person_name_original.as_deref() + .map(|s| !s.is_empty()).unwrap_or(false); + + // 快速判断是否需要处理 + let has_known_tag = old_tags.iter().any(|t| known_persons.contains(&t.to_lowercase())); + let has_like_tag = !has_person && old_tags.iter().any(|t| + title_parser::is_likely_person_name(t) && !title_parser::is_excluded_name(t)); + if has_person && !has_known_tag && !has_like_tag { continue; } + + // ── 第一层：逐个标签收集候选人名 ── + let mut known_person_found: Option = None; + let mut person_like_tags: Vec = Vec::new(); + + for tag in &old_tags { + let tag_lower = tag.to_lowercase(); + if known_persons.contains(&tag_lower) { + known_person_found = Some(tag.clone()); + continue; + } + if !has_person && title_parser::is_likely_person_name(tag) && !title_parser::is_excluded_name(tag) { + person_like_tags.push(tag.clone()); + continue; } } - // ── 处理标签 ── - if need_tags { - // 用当前 album 的人名（优先已有的，否则用刚提取的） - let person = album.person_name_cleaned.as_deref() - .or_else(|| album.person_name_original.as_deref()); + // 从 person_like_tags 中选最优（中文加权） + let best_layer1 = person_like_tags.into_iter().max_by_key(|t| { + if let Some(cn) = title_parser::extract_cn(t) { + if cn.len() >= 2 { cn.len() * 10 + t.len() } else { t.len() } + } else { t.len() } + }); - // 结合分类词库匹配（高质量标签）+ jieba 名词提取 - let mut tags = tag_extractor::extract_tags_from_title(title, person); - tags.extend(tags_from_jieba); + // ── 第二层：空格连接所有标签，用 jieba 提取人名 ── + let combined = old_tags.join(" "); + let (jieba_name, _) = title_parser::extract_person_and_tags(&combined); - let unique = tag_extractor::unique_tags(tags); - if !unique.is_empty() { - let tag_names: Vec = unique.iter().map(|(t, _)| t.clone()).collect(); - let tags_json = serde_json::to_string(&tag_names).unwrap_or_else(|_| "[]".to_string()); - let all_words: Vec<[String; 2]> = unique.iter().map(|(t, tp)| [t.clone(), tp.clone()]).collect(); - let words_json = serde_json::to_string(&all_words).unwrap_or_else(|_| "[]".to_string()); - db.update_album_tags(&album.id, &tags_json, &words_json).await; - tag_count += 1; + // ── 确定最终人名 ── + let final_person = known_person_found.or_else(|| { + match (&jieba_name, &best_layer1) { + (Some(j), Some(b)) => { + // jieba 结果更完整 → 用 jieba + if j.len() > b.len() || title_parser::is_excluded_name(b) { Some(j.clone()) } else { Some(b.clone()) } + } + (Some(j), None) => Some(j.clone()), + (None, Some(b)) => Some(b.clone()), + (None, None) => None, + } + }); + + if let Some(ref person_name) = final_person { + // ── 从标签中移除所有人名相关标签 ── + let mut changed = false; + let mut clean_tags: Vec = Vec::new(); + let person_lower = person_name.to_lowercase(); + let (primary_cleaned, _) = title_parser::split_person_name_with_aliases(person_name); + let primary_lower = primary_cleaned.to_lowercase(); + + for tag in &old_tags { + let tag_lower = tag.to_lowercase(); + let should_remove = + tag_lower == person_lower // 完全匹配 + || tag_lower == primary_lower // 中文核心匹配 + || (tag.len() >= 2 && tag.chars().all(|c| c.is_ascii_alphabetic()) + && !tag_lower.starts_with("cos") + && person_lower.contains(&tag_lower)) // 英文后缀 + || (tag_lower.len() >= 2 && person_lower.contains(&tag_lower)); // 人名包含 + + if should_remove { + person_like_removed += 1; + changed = true; + continue; + } + clean_tags.push(tag.clone()); + } + + if changed || !has_person { + // ── 写入 person_name（直接提取，无需等待 Phase 1）── + let (primary, aliases) = title_parser::split_person_name_with_aliases(person_name); + let final_name = if primary.is_empty() { person_name.clone() } else { primary }; + db.update_album_person(&album.id, person_name, &final_name).await; + + if !title_parser::is_excluded_name(&final_name) { + let pid = db.insert_person(&final_name, None).await; + db.insert_person_album(pid, &album.id).await; + for (at, an) in &aliases { + if !an.is_empty() && an != &final_name { + db.insert_person_alias(pid, at, an).await; + } + } + db.update_person_counts(pid).await; + } + + // ── 写入清理后的标签 ── + let tags_json_new = serde_json::to_string(&clean_tags).unwrap_or_else(|_| "[]".to_string()); + let all_words_json = album.all_words.as_deref().unwrap_or("[]"); + let all_words_new: Vec<[String; 2]> = serde_json::from_str::>(all_words_json) + .unwrap_or_default() + .into_iter() + .filter(|pair| !known_persons.contains(&pair[0].to_lowercase())) + .collect(); + let words_json_new = serde_json::to_string(&all_words_new).unwrap_or_else(|_| "[]".to_string()); + db.update_album_tags(&album.id, &tags_json_new, &words_json_new).await; + + if has_person { dirty.add(&album.id); } } } } - println!("[管线] 处理完成：{} 个人名，{} 个标签", person_count, tag_count); - (person_count, tag_count) + if person_like_removed > 0 { + println!("[管线·PhaseA1] 从标签中检测并提取人名：移除 {} 个人名标签，写入 {} 个相册的 person 字段", + person_like_removed, dirty.len()); + } else { + println!("[管线·PhaseA1] 检测标签中的人名：无异常"); + } + (dirty, person_like_removed) +} + +// ════════════════════════════════════════════════ +// Phase A2：清理 persons 表中的非人物 +// ════════════════════════════════════════════════ + +pub async fn phase_clean_persons(db: &Database) -> (DirtySet, usize) { + let persons = db.get_all_persons_with_counts().await; + let mut dirty = DirtySet::new(); + let mut deleted = 0usize; + + for (pid, name, _albums, _images) in &persons { + if title_parser::is_excluded_name(name) || !title_parser::is_likely_person_name(name) { + let album_list = db.get_person_albums(*pid).await; + let album_ids: Vec = album_list.into_iter().map(|a| a.id.to_string()).collect(); + for aid in &album_ids { dirty.add(aid); } + println!("[管线·PhaseA2] 删除非人物：{} (id={})，释放 {} 个相册", name, pid, album_ids.len()); + db.update_album_person_by_person_id(*pid).await; + db.delete_person_completely(*pid).await; + deleted += 1; + } + } + + if deleted > 0 { + println!("[管线·PhaseA2] 清理完成：删除 {} 个非人物，标记 {} 个脏相册", deleted, dirty.len()); + } else { + println!("[管线·PhaseA2] 清理完成：无非人物需要删除"); + } + (dirty, deleted) +} + +// ════════════════════════════════════════════════ +// Phase A3：收集新相册 + 半成品 +// ════════════════════════════════════════════════ + +pub async fn phase_collect_new_albums(db: &Database) -> DirtySet { + let albums = db.get_all_albums_full().await; + let mut dirty = DirtySet::new(); + let mut new_count = 0usize; + let mut partial_count = 0usize; + + for album in &albums { + let tags_empty = album.tags.as_deref() + .map(|t| t.is_empty() || t == "[]" || t == "null").unwrap_or(true); + if tags_empty { dirty.add(&album.id); new_count += 1; continue; } + + let has_person = album.person_name_original.as_deref() + .map(|s| !s.is_empty()).unwrap_or(false); + if !has_person { dirty.add(&album.id); partial_count += 1; } + } + + println!("[管线·PhaseA3] 收集相册：{} 个全新，{} 个半成品，脏集合计 {} 个", new_count, partial_count, dirty.len()); + dirty +} + +// ════════════════════════════════════════════════ +// Phase 1：增量解析 — 只处理脏集中的相册 +// ════════════════════════════════════════════════ + +pub async fn phase_parse_dirty(db: &Database, dirty_set: &DirtySet) -> (usize, usize) { + if dirty_set.is_empty() { + println!("[管线·Phase1] 增量解析跳过：脏集为空"); + return (0, 0); + } + + let albums = db.get_all_albums_full().await; + let patterns = person_merger::collect_person_patterns(db).await; + let use_patterns = !patterns.is_empty(); + let mut person_updated = 0usize; + let mut tag_updated = 0usize; + let mut skipped = 0usize; + + for album in &albums { + if !dirty_set.contains(&album.id) { continue; } + let title = album.title.as_deref().unwrap_or(""); + if title.is_empty() { skipped += 1; continue; } + + let (jieba_person, jieba_tags) = title_parser::extract_person_and_tags(title); + + let best_person = if use_patterns { + person_merger::match_title(title, &patterns) + .map(|(_, name, _)| name).or(jieba_person) + } else { + jieba_person + }; + + if let Some(name) = &best_person { + let (primary, aliases) = title_parser::split_person_name_with_aliases(name); + let final_name = if primary.is_empty() { name.clone() } else { primary }; + db.update_album_person(&album.id, name, &final_name).await; + let pid = db.insert_person(&final_name, None).await; + db.insert_person_album(pid, &album.id).await; + for (at, an) in &aliases { + if !an.is_empty() && an != &final_name { + db.insert_person_alias(pid, at, an).await; + } + } + db.update_person_counts(pid).await; + person_updated += 1; + } + + // 确定标签排除用的人名 — 用最完整的名字 + // 优先级：person_name_original(完整含英文后缀) > best_person > person_name_cleaned(仅中文) + let exclude_person = album.person_name_original.as_deref() + .filter(|s| !s.is_empty()) + .or_else(|| best_person.as_deref()) + .or_else(|| album.person_name_cleaned.as_deref().filter(|s| !s.is_empty())); + + let mut raw_tags = tag_extractor::extract_tags_from_title(title, exclude_person); + raw_tags.extend(jieba_tags); + let unique = tag_extractor::unique_tags(raw_tags); + let tag_names: Vec = unique.iter().map(|(t, _)| t.clone()).collect(); + + if !tag_names.is_empty() { + let tags_json = serde_json::to_string(&tag_names).unwrap_or_else(|_| "[]".to_string()); + let all_words: Vec<[String; 2]> = unique.iter().map(|(t, tp)| [t.clone(), tp.clone()]).collect(); + let words_json = serde_json::to_string(&all_words).unwrap_or_else(|_| "[]".to_string()); + db.update_album_tags(&album.id, &tags_json, &words_json).await; + tag_updated += 1; + } + } + + println!("[管线·Phase1] 增量解析完成：脏集 {} 个，处理 {} 个，跳过 {} 个，更新人名 {} 个，更新标签 {} 个", + dirty_set.len(), person_updated.max(tag_updated), skipped, person_updated, tag_updated); + (person_updated, tag_updated) +} + +// ════════════════════════════════════════════════ +// Phase 2：人物合并 +// ════════════════════════════════════════════════ + +pub async fn phase_merge_persons(db: &Database) -> (usize, usize) { + let result = person_merger::execute_merge(db).await; + if result.persons_merged > 0 { + println!("[管线·Phase2] 人物合并完成：分析 {} 组，合并 {} 人，迁移 {} 个别名", + result.groups_analyzed, result.persons_merged, result.aliases_moved); + } else { + println!("[管线·Phase2] 人物合并完成：无需要合并的人物"); + } + (result.persons_merged, result.aliases_moved) +} + +// ════════════════════════════════════════════════ +// Phase B：后校验标签 — 用合并后的人名检查 +// ════════════════════════════════════════════════ + +pub async fn phase_clean_tags_after_parse(db: &Database) -> usize { + let albums = db.get_all_albums_full().await; + let all_person_names: HashSet = db.get_all_person_names().await + .into_iter().map(|(n, _)| n.to_lowercase()).collect(); + let all_aliases: HashSet = db.get_all_aliases().await + .into_iter().map(|(a, _)| a.to_lowercase()).collect(); + let person_or_alias: HashSet = all_person_names.into_iter().chain(all_aliases).collect(); + + if person_or_alias.is_empty() { + println!("[管线·PhaseB] 后校验标签完成：无人物数据，跳过"); + return 0; + } + + let mut tag_removed = 0usize; + for album in &albums { + let tags_json = match album.tags.as_deref() { + Some(t) if !t.is_empty() && t != "[]" => t, _ => continue, + }; + let old_tags: Vec = match serde_json::from_str(tags_json) { Ok(t) => t, Err(_) => continue, }; + + let mut album_removed = 0usize; + let good_tags: Vec = old_tags.into_iter().filter(|t| { + if person_or_alias.contains(&t.to_lowercase()) { album_removed += 1; false } else { true } + }).collect(); + + if album_removed == 0 { continue; } + tag_removed += album_removed; + + let all_words_json = album.all_words.as_deref().unwrap_or("[]"); + let all_words_new: Vec<[String; 2]> = serde_json::from_str::>(all_words_json) + .unwrap_or_default().into_iter().filter(|pair| !person_or_alias.contains(&pair[0].to_lowercase())).collect(); + + let tags_json_new = serde_json::to_string(&good_tags).unwrap_or_else(|_| "[]".to_string()); + let words_json_new = serde_json::to_string(&all_words_new).unwrap_or_else(|_| "[]".to_string()); + db.update_album_tags(&album.id, &tags_json_new, &words_json_new).await; + } + + if tag_removed > 0 { + println!("[管线·PhaseB] 后校验标签完成：移除 {} 个人名/别名", tag_removed); + } else { + println!("[管线·PhaseB] 后校验标签完成：无需要修复的标签"); + } + tag_removed +} + +// ════════════════════════════════════════════════ +// Pipeline 一键执行（7 阶段） +// ════════════════════════════════════════════════ + +#[derive(Debug, Clone, Default)] +pub struct PipelineResult { + pub tags_cleaned: usize, // Phase 0 + pub person_like_tags_removed: usize, // Phase A1 + pub dirty_albums_detected: usize, // A1+A2+A3 + pub non_persons_deleted: usize, // Phase A2 + pub new_albums_found: usize, // Phase A3 + pub persons_updated: usize, // Phase 1 + pub tags_updated: usize, // Phase 1 + pub persons_merged: usize, // Phase 2 + pub aliases_moved: usize, // Phase 2 + pub tags_removed_person: usize, // Phase B +} + +/// 一键执行完整管线（7 阶段顺序串行） +pub async fn run_pipeline(db: &Database) -> PipelineResult { + let tags_cleaned = phase_clean_tags(db).await; + let (dirty_a1, person_like_tags_removed) = phase_detect_person_in_tags(db).await; + let (dirty_a2, non_persons_deleted) = phase_clean_persons(db).await; + let dirty_a3 = phase_collect_new_albums(db).await; + let new_albums_found = dirty_a3.len(); + + let mut master_dirty = DirtySet::new(); + master_dirty.extend(dirty_a1.into_inner()); + master_dirty.extend(dirty_a2.into_inner()); + master_dirty.extend(dirty_a3.into_inner()); + let dirty_albums_detected = master_dirty.len(); + + let (persons_updated, tags_updated) = phase_parse_dirty(db, &master_dirty).await; + let (persons_merged, aliases_moved) = phase_merge_persons(db).await; + let tags_removed_person = phase_clean_tags_after_parse(db).await; + + PipelineResult { + tags_cleaned, person_like_tags_removed, dirty_albums_detected, + non_persons_deleted, new_albums_found, + persons_updated, tags_updated, persons_merged, + aliases_moved, tags_removed_person, + } } diff --git a/src/engine/tag_extractor.rs b/src/engine/tag_extractor.rs index 18b376a..33d7023 100644 --- a/src/engine/tag_extractor.rs +++ b/src/engine/tag_extractor.rs @@ -22,6 +22,35 @@ lazy_static! { lazy_static! { static ref WORD_RE: Regex = Regex::new(r"[a-zA-Z0-9\u{4e00}-\u{9fff}]+").unwrap(); static ref HAS_CHINESE: Regex = Regex::new(r"[\u{4e00}-\u{9fff}]").unwrap(); + /// 提取标签中的有效部分（核心字词，去头尾标点） + static ref TAG_CORE_RE: Regex = Regex::new( + r#"^[\s.,;:!?'\"‘’“”\-–—()\[\]{}【】《》〈〉、，。；：？！…～·•]+|[\s.,;:!?'\"‘’“”\-–—()\[\]{}【】《》〈〉、，。；：？！…～·•]+$"# + ).unwrap(); +} + +/// 判断字符串是否全为标点符号（不含任何字、字母或数字） +pub fn is_punctuation_only(s: &str) -> bool { + if s.is_empty() { return true; } + // 检查是否包含任意"有效字符"（中文字、字母、数字）， + // 一个都不包含 → 纯标点/符号 + !s.chars().any(|c| { + c.is_ascii_alphanumeric() // ASCII 字母/数字 + || ('\u{4e00}'..='\u{9fff}').contains(&c) // CJK 统一表意文字 + || ('\u{3400}'..='\u{4dbf}').contains(&c) // CJK 扩展 A + || ('\u{f900}'..='\u{faff}').contains(&c) // CJK 兼容 + }) +} + +/// 清洗单个标签：去除头尾标点符号 +/// 例如 "性感!" → "性感"，"...比基尼" → "比基尼", " – 室内" → "室内" +/// 清洗后如果只剩下标点或为空，则返回 None +pub fn clean_tag(tag: &str) -> Option { + let cleaned = TAG_CORE_RE.replace_all(tag.trim(), "").trim().to_string(); + if cleaned.is_empty() || is_punctuation_only(&cleaned) || cleaned.len() < 2 { + None + } else { + Some(cleaned) + } } // ── 标签词库（按分类） ── @@ -152,6 +181,13 @@ fn extract_tags_with_jieba( // 也提取中文核心名用于排除 if let Some(cn) = title_parser::extract_cn(n) { v.push(cn.to_lowercase()); + // 中英文混合名的英文后缀也要排除（如 "水淼Aqua" → "aqua"） + if let Some(rest) = n.strip_prefix(&cn) { + let suffix = rest.trim(); + if !suffix.is_empty() { + v.push(suffix.to_lowercase()); + } + } } v }) @@ -163,6 +199,9 @@ fn extract_tags_with_jieba( let word_lower = word.to_lowercase(); + // 纯标点/符号跳过 + if is_punctuation_only(word) { continue; } + // 排除纯数字 if word.chars().all(|c| c.is_ascii_digit()) { continue; } @@ -182,14 +221,14 @@ fn extract_tags_with_jieba( "n" | "ng" => "noun", // 普通名词 "nz" => "proper", // 其它专名 "ns" => "location", // 地名 - "nt" => "organization", // 机构团体名 - "vn" => "noun_verb", // 名动词 - "a" | "ad" | "an" => "adj", // 形容词 - "v" | "vd" | "vg" => "verb", // 动词 - "eng" => "english", // 英文词 + "eng" => "other", // 英文词（归入 other） "x" => "other", // 其它 - // 排除的词性 + // 排除的词性（这些不应该作为标签） "nr" | "nrfg" | "nrt" => continue, // 人名类跳过 + "nt" => continue, // 机构团体名（不是标签） + "vn" => continue, // 名动词（分类模糊） + "a" | "ad" | "an" => continue, // 形容词（不是标签） + "v" | "vd" | "vg" => continue, // 动词（不是标签） "m" | "mq" => continue, // 数词 "q" | "qv" | "qt" => continue, // 量词 "r" | "rr" | "rz" => continue, // 代词 @@ -207,9 +246,6 @@ fn extract_tags_with_jieba( _ => continue, }; - // 进一步过滤：太短的英文词或太通用的词 - if category == "english" && word.len() < 3 { continue; } - tags.push((word.to_string(), category.to_string())); } @@ -231,6 +267,15 @@ pub fn extract_tags_from_title( for part in name.split_whitespace() { title_lower = title_lower.replace(&part.to_lowercase(), " "); } + // 中英文混合名：人名的英文后缀也从 title 中移除（如 "水淼Aqua" → "aqua"） + if let Some(cn) = title_parser::extract_cn(name) { + if let Some(rest) = name.strip_prefix(&cn) { + let suffix = rest.trim().to_lowercase(); + if !suffix.is_empty() { + title_lower = title_lower.replace(&suffix, " "); + } + } + } } let stop_words = get_stop_words_set(); @@ -265,6 +310,7 @@ pub fn extract_tags_from_title( // 基础过滤 if word.len() < 2 { continue; } + if is_punctuation_only(word) { continue; } if word.chars().all(|c| c.is_ascii_digit()) { continue; } if stop_words.contains(word_lower.as_str()) { continue; } if let Some(ref pn) = person_name_lower { @@ -287,8 +333,7 @@ pub fn extract_tags_from_title( seen.insert(word_lower.clone()); - let tag_type = if HAS_CHINESE.is_match(word) { "chinese" } else { "english" }; - tags.push((word.to_string(), tag_type.to_string())); + tags.push((word.to_string(), "other".to_string())); } tags @@ -302,8 +347,7 @@ pub fn unique_tags(tags: Vec<(String, String)>) -> Vec<(String, String)> { // 分类排序权重：预设词库分类最优先，然后是 jieba 提取的名词类 let category_order: Vec<&str> = vec![ "style", "outfit", "scene", "body", "mood", "topic", - "noun", "proper", "location", "organization", "noun_verb", - "adj", "verb", "chinese", "english", "other" + "noun", "proper", "location", "other" ]; let mut sorted_tags = tags; @@ -314,14 +358,38 @@ pub fn unique_tags(tags: Vec<(String, String)>) -> Vec<(String, String)> { }); for (tag, tag_type) in sorted_tags { - let key = tag.to_lowercase(); + // 清洗标签：去除头尾标点 + let cleaned_tag = match clean_tag(&tag) { + Some(t) => t, + None => continue, // 清洗后为空 → 跳过 + }; + // 最终安全过滤：坏标签一律剔除（停用词等） + if is_bad_tag(&cleaned_tag) { continue; } + let key = cleaned_tag.to_lowercase(); if seen.insert(key) { - result.push((tag, tag_type)); + result.push((cleaned_tag, tag_type)); } } result } +/// 判断一个标签是否"坏标签"（应该在数据库中清理掉） +/// 坏标签定义：标点符号、含头尾标点需要清洗、太短、纯数字、停用词 +/// 如果标签需要清洗（如 "性感!"），也会被标记为坏标签 +pub fn is_bad_tag(tag: &str) -> bool { + let tag = tag.trim(); + if tag.is_empty() || tag.len() < 2 { return true; } + if tag.chars().all(|c| c.is_ascii_digit()) { return true; } + if is_punctuation_only(tag) { return true; } + // 如果标签清洗后变短了，说明含有头尾标点 → 坏标签 + if let Some(cleaned) = clean_tag(tag) { + if cleaned != tag { return true; } + } + let stop_words = get_stop_words_set(); + if stop_words.contains(&tag.to_lowercase().as_str()) { return true; } + false +} + /// 获取标签的中文分类名（用于前端展示） pub fn get_category_label(category: &str) -> &'static str { match category { @@ -334,12 +402,6 @@ pub fn get_category_label(category: &str) -> &'static str { "noun" => "名词", "proper" => "专名", "location" => "地名", - "organization" => "机构", - "noun_verb" => "名动", - "adj" => "形容词", - "verb" => "动词", - "chinese" => "中文", - "english" => "英文", _ => "其他", } } @@ -356,12 +418,6 @@ pub fn get_category_color(category: &str) -> (&'static str, &'static str) { "noun" => ("#eff6ff", "#1d4ed8"), // 蓝灰底 + 深蓝字 "proper" => ("#e0e7ff", "#4338ca"), // 靛蓝底 + 深靛字 "location"=> ("#ecfdf5", "#047857"), // 翠绿底 + 深绿字 - "organization" => ("#fef9c3", "#a16207"), // 浅黄底 + 暗黄字 - "noun_verb" => ("#fce7f3", "#be185d"), // 浅粉底 + 深粉字 - "adj" => ("#f3e8ff", "#6d28d9"), // 浅紫底 + 深紫字 - "verb" => ("#fed7aa", "#c2410c"), // 橙底 + 深橙字 - "chinese" => ("#f1f5f9", "#475569"), // 浅灰底 + 深灰字 - "english" => ("#f1f5f9", "#64748b"), // 浅灰底 + 灰字 _ => ("#f1f5f9", "#64748b"), // 默认 } } diff --git a/src/engine/title_parser.rs b/src/engine/title_parser.rs index 00b107f..4eaf987 100644 --- a/src/engine/title_parser.rs +++ b/src/engine/title_parser.rs @@ -96,9 +96,14 @@ lazy_static! { static ref RE_CHINESE: Regex = Regex::new(r"[\u{4e00}-\u{9fff}]").unwrap(); static ref RE_ENGLISH_SUFFIX: Regex = Regex::new(r"^([\u{4e00}-\u{9fff}]+)([A-Za-z].*)$").unwrap(); static ref RE_TRAILING_PUNCT: Regex = Regex::new(r"[\s.,;:，。；：\-–—]+$").unwrap(); - /// 人名模式：2-5个中文字符 + 可选英文后缀 + /// 人名模式：2-5个中文字符（纯中文，不混杂英文） + /// 中英文混杂的情况由 split_person_name_with_aliases 处理为 主名+别名 static ref RE_CN_NAME: Regex = Regex::new( - r"([\u{4e00}-\u{9fff}]{2,5}(?:[A-Za-z][A-Za-z0-9_]*)?)" + r"([\u{4e00}-\u{9fff}]{2,5})" + ).unwrap(); + /// 纯英文人名模式：2-15个英文字母，首字母可大写 + static ref RE_EN_NAME: Regex = Regex::new( + r"([A-Za-z][A-Za-z]{1,14})" ).unwrap(); /// 常见人名后缀（表明前面是人名） static ref RE_NAME_CONTEXT: Regex = Regex::new( @@ -201,6 +206,7 @@ pub fn extract_person_name_from_title(title: &str) -> Option { } // 策略C: 直接在整个标题中搜索人名模式 + // 先搜纯中文（优先级更高） for cap in RE_CN_NAME.captures_iter(&t) { let name = clean_name(&cap[1]); if is_likely_person_name(&name) { @@ -208,6 +214,16 @@ pub fn extract_person_name_from_title(title: &str) -> Option { } } + // 再搜纯英文 + for cap in RE_EN_NAME.captures_iter(&t) { + let name = clean_name(&cap[1]); + // 排除停用词 + if STOP_WORDS_SET.contains(&name.to_lowercase().as_str()) { continue; } + if is_likely_person_name(&name) { + return Some(name); + } + } + None } @@ -245,13 +261,25 @@ fn extract_person_with_jieba(title: &str) -> Option { } /// 从一个文本片段中提取人名 +/// 只支持纯中文（2-6字）或纯英文（2-15字），不混杂 +/// 中文优先级更高 fn extract_name_from_segment(segment: &str) -> Option { let seg = segment.trim(); if seg.is_empty() { return None; } - // 尝试直接匹配人名模式 + // 判断片段是否纯中文 + let is_pure_cn = seg.chars().all(|c| RE_CHINESE.is_match(&c.to_string())); + // 判断片段是否纯英文 + let is_pure_en = seg.chars().all(|c| c.is_ascii_alphabetic()); + + // 策略1: 纯中文片段（2-6字） + if is_pure_cn && seg.len() >= 2 && seg.len() <= 6 { + return Some(clean_name(seg)); + } + + // 策略2: 尝试匹配纯中文人名模式 if let Some(cap) = RE_CN_NAME.captures(seg) { let name = clean_name(&cap[1]); if name.len() >= 2 { @@ -259,10 +287,20 @@ fn extract_name_from_segment(segment: &str) -> Option { } } - // 如果整个片段是纯中文（2-5字），也认为是人名 - let is_pure_cn = seg.chars().all(|c| RE_CHINESE.is_match(&c.to_string())); - if is_pure_cn && seg.len() >= 2 && seg.len() <= 6 { - return Some(clean_name(seg)); + // 策略3: 纯英文片段（2-15字） + if is_pure_en && seg.len() >= 2 && seg.len() <= 15 { + // 过滤停用词 + if !STOP_WORDS_SET.contains(&seg.to_lowercase().as_str()) { + return Some(clean_name(seg)); + } + } + + // 策略4: 尝试匹配纯英文人名模式 + if let Some(cap) = RE_EN_NAME.captures(seg) { + let name = clean_name(&cap[1]); + if name.len() >= 2 && !STOP_WORDS_SET.contains(&name.to_lowercase().as_str()) { + return Some(name); + } } None @@ -482,13 +520,18 @@ pub fn extract_person_and_tags(title: &str) -> (Option, Vec<(String, Str } /// 从 jieba 分词结果中提取人名 +/// 返回完整的原始人名（可能含英文后缀），供 split_person_name_with_aliases 拆分为 主名+别名 +/// 例如 jieba 拆出 "水淼/nr Aqua/eng" → 返回 "水淼Aqua" fn extract_person_from_tags(tags: &[jieba_rs::Tag]) -> Option { let mut candidates: Vec<(String, usize)> = Vec::new(); - for tag in tags { + for (i, tag) in tags.iter().enumerate() { let word = tag.word.trim(); if word.len() < 2 { continue; } + // 过滤纯标点 + if is_punctuation_only(word) { continue; } + let priority = match tag.tag { "nr" => 0, // 人名（最高优先级） "nrfg" => 1, // 古体人名 @@ -499,7 +542,33 @@ fn extract_person_from_tags(tags: &[jieba_rs::Tag]) -> Option { }; if is_likely_person_name(word) { - candidates.push((word.to_string(), priority)); + // 如果是中文人名，检查后面是否紧跟英文词（如 水淼/nr Aqua/eng） + // 拼装成完整原始名，供 split_person_name_with_aliases 拆分为 主名+别名 + let mut full_name = word.to_string(); + let is_chinese = word.chars().any(|c| RE_CHINESE.is_match(&c.to_string())); + if is_chinese { + let mut j = i + 1; + while j < tags.len() { + let next_word = tags[j].word.trim(); + let next_tag = tags[j].tag; + if next_word.is_empty() { j += 1; continue; } + // 英文词（eng）或纯英文字母，且不是停用词 + if next_tag == "eng" || + (next_word.chars().all(|c| c.is_ascii_alphabetic()) && next_word.len() >= 2 && next_word.len() <= 15) { + let lower = next_word.to_lowercase(); + // 跳过停用词（如 photos、cosplay 等停在停用词之前） + let stop_words: std::collections::HashSet<&str> = STOP_WORDS.iter().copied().collect(); + if !stop_words.contains(lower.as_str()) && !EXCLUDED_LOWER.contains(&lower) { + full_name.push_str(next_word); + j += 1; + continue; + } + } + break; + } + } + + candidates.push((full_name, priority)); } } @@ -507,6 +576,18 @@ fn extract_person_from_tags(tags: &[jieba_rs::Tag]) -> Option { candidates.into_iter().next().map(|(name, _)| clean_name(&name)) } +/// 判断字符串是否全为标点符号（不含任何字、字母或数字） +fn is_punctuation_only(s: &str) -> bool { + if s.is_empty() { return true; } + // 检查是否包含任意"有效字符"（中文字、字母、数字） + !s.chars().any(|c| { + c.is_ascii_alphanumeric() + || ('\u{4e00}'..='\u{9fff}').contains(&c) + || ('\u{3400}'..='\u{4dbf}').contains(&c) + || ('\u{f900}'..='\u{faff}').contains(&c) + }) +} + /// 从 jieba 分词结果中提取标签（排除人名） fn extract_tags_from_jieba_result( tags: &[jieba_rs::Tag], @@ -529,6 +610,13 @@ fn extract_tags_from_jieba_result( } if let Some(cn) = extract_cn(n) { v.push(cn.to_lowercase()); + // 中英文混合名的英文后缀也要排除（如 "水淼Aqua" → "aqua"） + if let Some(rest) = n.strip_prefix(&cn) { + let suffix = rest.trim(); + if !suffix.is_empty() { + v.push(suffix.to_lowercase()); + } + } } v }) @@ -540,6 +628,9 @@ fn extract_tags_from_jieba_result( let word_lower = word.to_lowercase(); + // 纯标点/符号跳过 + if is_punctuation_only(word) { continue; } + // 纯数字跳过 if word.chars().all(|c| c.is_ascii_digit()) { continue; } @@ -558,14 +649,14 @@ fn extract_tags_from_jieba_result( "n" | "ng" => "noun", "nz" => "proper", "ns" => "location", - "nt" => "organization", - "vn" => "noun_verb", - "a" | "ad" | "an" => "adj", - "v" | "vd" | "vg" => "verb", - "eng" => "english", + "eng" => "other", "x" => "other", - // 排除的词性 + // 排除的词性（这些不应该作为标签） "nr" | "nrfg" | "nrt" => continue, // 人名 + "nt" => continue, // 机构团体名（不是标签） + "vn" => continue, // 名动词（分类模糊） + "a" | "ad" | "an" => continue, // 形容词（不是标签） + "v" | "vd" | "vg" => continue, // 动词（不是标签） "m" | "mq" => continue, "q" | "qv" | "qt" => continue, "r" | "rr" | "rz" => continue, @@ -583,8 +674,6 @@ fn extract_tags_from_jieba_result( _ => continue, }; - if category == "english" && word.len() < 3 { continue; } - result.push((word.to_string(), category.to_string())); } @@ -650,10 +739,13 @@ mod tests { #[test] fn test_extract_person_name() { - assert_eq!(extract_person_name_from_title("杨晨晨Yome 私房写真").unwrap(), "杨晨晨Yome"); - assert_eq!(extract_person_name_from_title("王馨瑶yanni 比基尼").unwrap(), "王馨瑶yanni"); + // 纯中文人名（中文优先级更高，英文后缀作为别名处理） + assert_eq!(extract_person_name_from_title("杨晨晨Yome 私房写真").unwrap(), "杨晨晨"); + assert_eq!(extract_person_name_from_title("王馨瑶yanni 比基尼").unwrap(), "王馨瑶"); assert_eq!(extract_person_name_from_title("[XiuRen] 杨晨晨 性感").unwrap(), "杨晨晨"); assert_eq!(extract_person_name_from_title("美女模特 杨晨晨 私房").unwrap(), "杨晨晨"); + // 纯英文人名也支持 + assert!(extract_person_name_from_title("Yome sexy photos").is_some()); } #[test] diff --git a/src/main.rs b/src/main.rs index d1e9865..2c63520 100644 --- a/src/main.rs +++ b/src/main.rs @@ -49,86 +49,56 @@ async fn main() -> std::io::Result<()> { let image_processor = Arc::new(image_processor::ImageProcessor::new(config.storage.clone())); - // 后台任务1: 扫描导入新相册 - let config_clone = config.clone(); - let db_import = Arc::clone(&db); - tokio::spawn(async move { - info!("[后台任务] 开始扫描图片目录并导入新相册..."); - match services::import::import_existing_data(&config_clone, db_import, None).await { - Ok(_) => info!("[后台任务] 相册导入完成"), - Err(e) => error!("[后台任务] 相册导入失败: {}", e), - } - }); - - // 后台任务2: 自动解析已有相册的 title（增量，针对老版本 db） - let db_parse = Arc::clone(&db); - tokio::spawn(async move { - // 等待 3 秒让导入任务先开始（导入后也会触发解析，避免重复） - tokio::time::sleep(tokio::time::Duration::from_secs(3)).await; - info!("[后台任务] 检查是否有未解析的相册（针对老版本 db）..."); - let (persons, tags) = engine::pipeline::process_all_albums(&db_parse).await; - if persons > 0 || tags > 0 { - info!("[后台任务] Title 解析完成：{} 个人名，{} 个标签", persons, tags); - } else { - info!("[后台任务] 所有相册已解析，无需处理"); - } - - // 解析完成后，执行一次人物合并 - tokio::time::sleep(tokio::time::Duration::from_secs(2)).await; - info!("[后台任务] 执行人物合并分析..."); - let result = engine::person_merger::execute_merge(&db_parse).await; - if result.persons_merged > 0 { - info!("[后台任务] 人物合并完成：分析 {} 组，合并 {} 人", result.groups_analyzed, result.persons_merged); - } else { - info!("[后台任务] 无需要合并的人物"); - } - }); - - // 相册内存缓存 + // 相册内存缓存（必须在管线任务之前创建） let albums_cache = Arc::new(tokio::sync::RwLock::new(AlbumsCache::new())); - // 初始加载缓存 - let initial_cache = Arc::clone(&albums_cache); - let initial_db = Arc::clone(&db); + // ── 启动管线（单次）：导入 → 解析 → 合并 → 缓存刷新 ── + let boot_config = config.clone(); + let boot_db = Arc::clone(&db); + let boot_cache = Arc::clone(&albums_cache); tokio::spawn(async move { - match initial_db.get_all_albums().await { - Ok(albums) => { - let mut cache = initial_cache.write().await; - cache.update(albums); - println!("✅ 初始相册缓存加载完成"); - } - Err(e) => { - eprintln!("❌ 初始缓存加载失败: {}", e); - } + // ① 扫描导入新相册 + info!("[启动管线] ① 扫描导入新相册..."); + match services::import::import_existing_data(&boot_config, Arc::clone(&boot_db), None).await { + Ok(_) => info!("[启动管线] 导入完成"), + Err(e) => error!("[启动管线] 导入失败: {}", e), } + + // ② 全量解析 + 人物合并 + info!("[启动管线] ② 全量解析 + 人物合并..."); + let result = engine::pipeline::run_pipeline(&boot_db).await; + info!("[启动管线] 完成：清洗标点{}，检测人名标签{}，脏集{}相册，删除非人物{}，发现新相册{}，更新人名{}，更新标签{}，合并人物{}，迁移别名{}，后移除人名标签{}", + result.tags_cleaned, result.person_like_tags_removed, result.dirty_albums_detected, + result.non_persons_deleted, result.new_albums_found, + result.persons_updated, result.tags_updated, + result.persons_merged, result.aliases_moved, result.tags_removed_person); + + // ③ 刷新内存缓存 + info!("[启动管线] ③ 刷新内存缓存..."); + refresh_albums_cache(&boot_db, &boot_cache).await; + info!("[启动管线] 启动管线执行完毕"); }); - // 定时任务（每30分钟）：刷新缓存 + 人物合并 - let refresh_cache = Arc::clone(&albums_cache); - let refresh_db = Arc::clone(&db); + // ── 定时管线（每 30 分钟）：解析 → 合并 → 缓存刷新 ── + let timer_cache = Arc::clone(&albums_cache); + let timer_db = Arc::clone(&db); tokio::spawn(async move { let mut interval = interval(Duration::from_secs(30 * 60)); loop { interval.tick().await; - // 1. 执行人物合并 - info!("[定时任务] 执行人物合并分析..."); - let result = engine::person_merger::execute_merge(&refresh_db).await; - if result.persons_merged > 0 { - info!("[定时任务] 人物合并完成：分析 {} 组，合并 {} 人", result.groups_analyzed, result.persons_merged); - } + // ① 全量解析 + 人物合并（不重复 import） + info!("[定时管线] ① 全量解析 + 人物合并..."); + let result = engine::pipeline::run_pipeline(&timer_db).await; + info!("[定时管线] 完成：清洗标点{}，检测人名标签{}，脏集{}相册，删除非人物{}，发现新相册{}，更新人名{}，更新标签{}，合并人物{}，迁移别名{}，后移除人名标签{}", + result.tags_cleaned, result.person_like_tags_removed, result.dirty_albums_detected, + result.non_persons_deleted, result.new_albums_found, + result.persons_updated, result.tags_updated, + result.persons_merged, result.aliases_moved, result.tags_removed_person); - // 2. 刷新缓存 - match refresh_db.get_all_albums().await { - Ok(albums) => { - let mut cache = refresh_cache.write().await; - cache.update(albums); - println!("🔄 相册缓存定时刷新完成"); - } - Err(e) => { - eprintln!("❌ 缓存刷新失败: {}", e); - } - } + // ② 刷新内存缓存 + info!("[定时管线] ② 刷新内存缓存..."); + refresh_albums_cache(&timer_db, &timer_cache).await; } }); @@ -241,4 +211,17 @@ async fn main() -> std::io::Result<()> { .bind(config.server_address())? .run() .await +} + +/// 从 DB 重新加载相册数据到内存缓存 +async fn refresh_albums_cache(db: &crate::db::Database, cache: &Arc>) { + match db.get_all_albums().await { + Ok(albums) => { + let mut c = cache.write().await; + c.update(albums); + } + Err(e) => { + eprintln!("❌ 缓存刷新失败: {}", e); + } + } } \ No newline at end of file diff --git a/src/routes/pipeline.rs b/src/routes/pipeline.rs index 2020f15..2d1b0fa 100644 --- a/src/routes/pipeline.rs +++ b/src/routes/pipeline.rs @@ -3,13 +3,27 @@ use crate::engine::pipeline; use actix_web::{HttpResponse, Responder, web}; use std::sync::Arc; -/// POST /api/pipeline/run — 处理所有相册元数据（提取人名+标签） +/// POST /api/pipeline/run — 执行完整管线：清洗→检测→清理→解析→合并→校验 pub async fn pipeline_run(db: web::Data>) -> impl Responder { - let (persons, tags) = pipeline::process_all_albums(db.get_ref()).await; + let result = pipeline::run_pipeline(db.get_ref()).await; HttpResponse::Ok().json(serde_json::json!({ "success": true, - "persons_updated": persons, - "tags_updated": tags, - "message": format!("处理完成：{} 个人名，{} 个标签", persons, tags) + "tags_cleaned": result.tags_cleaned, + "person_like_tags_removed": result.person_like_tags_removed, + "dirty_albums_detected": result.dirty_albums_detected, + "non_persons_deleted": result.non_persons_deleted, + "new_albums_found": result.new_albums_found, + "persons_updated": result.persons_updated, + "tags_updated": result.tags_updated, + "persons_merged": result.persons_merged, + "aliases_moved": result.aliases_moved, + "tags_removed_person": result.tags_removed_person, + "message": format!( + "管线执行完毕：清洗标点{}相册，检测人名标签{}，脏集{}相册，删除非人物{}，发现新相册{}，更新人名{}，更新标签{}，合并人物{}，迁移别名{}，后移除人名标签{}", + result.tags_cleaned, result.person_like_tags_removed, result.dirty_albums_detected, + result.non_persons_deleted, result.new_albums_found, + result.persons_updated, result.tags_updated, + result.persons_merged, result.aliases_moved, result.tags_removed_person + ) })) } \ No newline at end of file