@xingwangzhe/stalux 1.22.7 → 1.23.0

package/src/internal/font-subset.ts

@@ -1,619 +0,0 @@
-/**
- * Per-route font subsetting engine.
- *
- * Scans each content file individually, extracts its unique character set,
- * and generates a minimal TTF subset per route. Also generates a "common"
- * subset for UI text (navs, i18n, layout).
- *
- * Output layout (under public/fonts/):
- *   fonts/
- *     common-{hash}.ttf       — UI text subset (ASCII, nav, i18n)
- *     subset-{hash}.ttf       — per-route subset (shared by routes with same chars)
- *     common.css              — @font-face for common subset
- *     subset-{route}.css      — per-route @font-face
- *     manifest.json           — route → subset file mapping
- */
-
-import { createHash } from "node:crypto";
-import {
-    existsSync,
-    mkdirSync,
-    readFileSync,
-    writeFileSync,
-    readdirSync,
-    unlinkSync,
-} from "node:fs";
-import { resolve, basename, extname, join, dirname } from "node:path";
-import { fileURLToPath } from "node:url";
-
-import type { AstroIntegrationLogger } from "astro";
-// 与 Astro 内容集合同款的 YAML 解析器（astro 依赖 js-yaml 解析 frontmatter，
-// 见 node_modules/astro/dist/content/loaders/file.js）。用它解析 frontmatter
-// 得到的结果与 getCollection 的 data 一致：引号剥离、类型还原、结构完整。
-import { load as parseYaml } from "js-yaml";
-import subsetFont from "subset-font";
-
-// ---------------------------------------------------------------------------
-// Configuration
-// ---------------------------------------------------------------------------
-
-/** Content root directory (relative to project root) */
-const CONTENT_ROOT = "stalux";
-
-/** Full font input path (relative to project root or package) */
-const FONT_INPUT = "src/assets/fonts/LXGWWenKai-Regular.ttf";
-
-/** Code font input path (Google Sans Code, deterministic static @font-face) */
-const CODE_FONT_INPUT = "src/assets/fonts/GoogleSansCode.woff2";
-
-/** Output directory for generated subsets */
-const FONT_OUT_DIR = "public/fonts";
-
-/** Characters always included in common subset */
-const ALWAYS_INCLUDE_CHARS = new Set(
-    // ASCII printable
-    [...Array(0x7e - 0x20 + 1)]
-        .map((_, i) => String.fromCodePoint(0x20 + i))
-        .join("")
-        // Basic CJK punctuation
-        .concat(
-            "\u3000\u3001\u3002\u3008\u3009\u300a\u300b\u300c\u300d\u3010\u3011" +
-                "\uff01\uff08\uff09\uff0c\uff0e\uff1a\uff1b\uff1f\uff5e\u2013\u2014" +
-                "\u2018\u2019\u201c\u201d\u2026\uff09\uff08\u00b7",
-        ),
-);
-
-// ---------------------------------------------------------------------------
-// Types
-// ---------------------------------------------------------------------------
-
-interface RouteCharSet {
-    route: string;
-    chars: string;
-    hash: string;
-}
-
-interface FontManifest {
-    common: string; // filename of common subset TTF
-    commonCss: string; // filename of common subset CSS (content-addressed)
-    routes: Record<string, string>; // route → subset filename
-    /** Per-post mapping: abbrlink → subset filename */
-    posts: Record<string, string>;
-}
-
-// ---------------------------------------------------------------------------
-// Character extraction
-// ---------------------------------------------------------------------------
-
-function extractChars(text: string): string {
-    const set = new Set<string>();
-    for (const ch of text) set.add(ch);
-    return [...set].sort().join("");
-}
-
-function hashChars(chars: string): string {
-    return createHash("md5").update(chars).digest("hex").slice(0, 12);
-}
-
-// ---------------------------------------------------------------------------
-// Content scanning
-// ---------------------------------------------------------------------------
-
-interface ContentFile {
-    route: string;
-    filePath: string;
-    abbrlink?: string;
-}
-
-/**
- * 用 js-yaml（内容集合同款解析器）解析 frontmatter。
- *
- * 结果与 getCollection("posts") 的 data 一致：
- * - abbrlink：number 转 string、引号剥离（YAML 解析天然处理）
- * - tags/categories：单行字符串 "Bing" 按 schema preprocess 转 ["Bing"]
- */
-function parsePostFrontmatter(content: string): {
-    abbrlink?: string;
-    tags: string[];
-    categories: string[];
-} {
-    const m = content.match(/^---\r?\n([\s\S]*?)\r?\n---/);
-    if (!m) return { abbrlink: undefined, tags: [], categories: [] };
-
-    let data: Record<string, unknown>;
-    try {
-        data = (parseYaml(m[1]) as Record<string, unknown>) ?? {};
-    } catch {
-        return { abbrlink: undefined, tags: [], categories: [] };
-    }
-
-    // 复刻 posts schema 的 preprocess：单行字符串转数组
-    const toList = (v: unknown): string[] => {
-        if (Array.isArray(v)) return v.map((x) => String(x));
-        if (typeof v === "string") return v.trim() ? [v.trim()] : [];
-        return [];
-    };
-
-    const abbr = data.abbrlink;
-    return {
-        abbrlink: abbr === undefined || abbr === null ? undefined : String(abbr),
-        tags: toList(data.tags),
-        categories: toList(data.categories),
-    };
-}
-
-function scanContent(projectRoot: string): ContentFile[] {
-    const files: ContentFile[] = [];
-    const contentRoot = resolve(projectRoot, CONTENT_ROOT);
-
-    // Posts
-    const postsDir = join(contentRoot, "posts");
-    const allPostContents: {
-        abbrlink: string;
-        tags: string[];
-        categories: string[];
-        content: string;
-    }[] = [];
-    if (existsSync(postsDir)) {
-        for (const f of readdirSync(postsDir)) {
-            if (!f.endsWith(".md") && !f.endsWith(".mdx")) continue;
-            const content = readFileSync(join(postsDir, f), "utf-8");
-            const fm = parsePostFrontmatter(content);
-
-            // Per-post routes：与内容集合 glob（*.{md,mdx}，含 _ 前缀文件，
-            // 见 schemas/collections.ts）保持一致——这些文件都会生成页面，
-            // 字体子集必须覆盖，否则页面 subset css 缺失 → 404。
-            const abbrlink = fm.abbrlink ?? basename(f, extname(f));
-            files.push({
-                route: `posts/${abbrlink}`,
-                filePath: join(postsDir, f),
-                abbrlink,
-            });
-            // Virtual routes (archives/tags/categories): include ALL posts including templates
-            // because content collections load them and they render on those pages
-            allPostContents.push({
-                abbrlink,
-                tags: fm.tags,
-                categories: fm.categories,
-                content,
-            });
-        }
-    }
-
-    // Build virtual routes for archives, tags, categories (aggregated pages)
-    // These pages render content from ALL posts
-    if (allPostContents.length > 0) {
-        const allContent = allPostContents.map((p) => p.content).join("\n");
-
-        // Archives: all posts' content (titles, dates, etc.)
-        virtualRouteCache.set("archives", extractChars(allContent));
-
-        // Tags index: all tags + all posts' content
-        const allTags = [...new Set(allPostContents.flatMap((p) => p.tags))].join(" ");
-        virtualRouteCache.set("tags", extractChars(allContent + allTags));
-
-        // Categories index: all categories + all posts' content
-        const allCategories = [...new Set(allPostContents.flatMap((p) => p.categories))].join(" ");
-        virtualRouteCache.set("categories", extractChars(allContent + allCategories));
-
-        // Per-tag routes: tag name + posts with that tag
-        const tagPosts = new Map<string, string[]>();
-        for (const p of allPostContents) {
-            for (const tag of p.tags) {
-                const list = tagPosts.get(tag) ?? [];
-                list.push(p.content);
-                tagPosts.set(tag, list);
-            }
-        }
-        for (const [tag, contents] of tagPosts) {
-            const tagContent = tag + " " + contents.join("\n");
-            virtualRouteCache.set(`tags-${tag}`, extractChars(tagContent));
-        }
-
-        // Per-category routes: category name + posts in that category
-        const catPosts = new Map<string, string[]>();
-        for (const p of allPostContents) {
-            for (const cat of p.categories) {
-                const list = catPosts.get(cat) ?? [];
-                list.push(p.content);
-                catPosts.set(cat, list);
-            }
-        }
-        for (const [cat, contents] of catPosts) {
-            const catContent = cat + " " + contents.join("\n");
-            virtualRouteCache.set(`categories-${cat}`, extractChars(catContent));
-        }
-    }
-
-    // Build virtual routes for config-driven pages (home, links)
-    // These pages render text from config YAML files (nav titles, site title, etc.)
-    const configDir = join(contentRoot, "config");
-    if (existsSync(configDir)) {
-        const allConfigContent: string[] = [];
-        for (const f of readdirSync(configDir)) {
-            if (f.endsWith(".yml") || f.endsWith(".yaml")) {
-                allConfigContent.push(readFileSync(join(configDir, f), "utf-8"));
-            }
-        }
-        if (allConfigContent.length > 0) {
-            const configText = allConfigContent.join("\n");
-            // Home page: site title, author, navs, typetexts, social media, footer
-            virtualRouteCache.set("home", extractChars(configText));
-            // Links page: links config + navs + footer
-            virtualRouteCache.set("links", extractChars(configText));
-        }
-    }
-
-    // About
-    const aboutDir = join(contentRoot, "about");
-    if (existsSync(aboutDir)) {
-        for (const f of readdirSync(aboutDir)) {
-            if (!f.endsWith(".md") && !f.endsWith(".mdx")) continue;
-            if (f.startsWith("_")) continue;
-            files.push({
-                route: "about",
-                filePath: join(aboutDir, f),
-            });
-        }
-    }
-
-    // Words - aggregate all words files' characters into one
-    const wordsDir = join(contentRoot, "words");
-    if (existsSync(wordsDir)) {
-        const allWordsContent: string[] = [];
-        for (const f of readdirSync(wordsDir)) {
-            if (!f.endsWith(".md")) continue;
-            if (f.startsWith("_")) continue;
-            try {
-                allWordsContent.push(readFileSync(join(wordsDir, f), "utf-8"));
-            } catch {
-                /* skip */
-            }
-        }
-        if (allWordsContent.length > 0) {
-            files.push({
-                route: "words",
-                filePath: join(wordsDir, "words-combined.md"),
-                // We'll handle this virtual file specially in char extraction
-            });
-            // Store combined content in a map for later extraction
-            wordCharsCache.set("words", extractChars(allWordsContent.join("\n")));
-        }
-    }
-
-    // Add virtual routes (archives, tags, categories) as content files
-    for (const [route] of virtualRouteCache) {
-        files.push({
-            route,
-            filePath: resolve(projectRoot, CONTENT_ROOT, `_virtual/${route}.md`),
-        });
-    }
-
-    return files;
-}
-
-// Cache for virtual route char sets
-const wordCharsCache = new Map<string, string>();
-const virtualRouteCache = new Map<string, string>();
-
-// ---------------------------------------------------------------------------
-// Common subset: extract chars from UI source files
-// ---------------------------------------------------------------------------
-
-function collectCommonChars(projectRoot: string): string {
-    const chars = new Set(ALWAYS_INCLUDE_CHARS);
-
-    // Scan all source directories for UI text characters
-    const srcDir = resolve(projectRoot, "src");
-
-    function scanDir(dir: string) {
-        if (!existsSync(dir)) return;
-        for (const entry of readdirSync(dir, { withFileTypes: true })) {
-            const fullPath = join(dir, entry.name);
-            if (entry.isDirectory()) {
-                scanDir(fullPath);
-            } else if (
-                entry.name.endsWith(".astro") ||
-                entry.name.endsWith(".ts") ||
-                entry.name.endsWith(".json")
-            ) {
-                try {
-                    const content = readFileSync(fullPath, "utf-8");
-                    // Only extract string literals and text content (skip code)
-                    for (const ch of content) chars.add(ch);
-                } catch {
-                    /* skip */
-                }
-            }
-        }
-    }
-
-    scanDir(srcDir);
-
-    // Also scan user config YAML files for nav titles, descriptions, etc.
-    const configDir = resolve(projectRoot, CONTENT_ROOT, "config");
-    if (existsSync(configDir)) {
-        for (const f of readdirSync(configDir)) {
-            if (!f.endsWith(".yml") && !f.endsWith(".yaml")) continue;
-            try {
-                const content = readFileSync(join(configDir, f), "utf-8");
-                for (const ch of content) chars.add(ch);
-            } catch {
-                /* skip */
-            }
-        }
-    }
-
-    // Also scan post frontmatter for tags, categories so those pages' chars are covered
-    const postsDir = resolve(projectRoot, CONTENT_ROOT, "posts");
-    if (existsSync(postsDir)) {
-        for (const f of readdirSync(postsDir)) {
-            if (!f.endsWith(".md") && !f.endsWith(".mdx")) continue;
-            if (f.startsWith("_")) continue;
-            try {
-                const content = readFileSync(join(postsDir, f), "utf-8");
-                for (const ch of content) chars.add(ch);
-            } catch {
-                /* skip */
-            }
-        }
-    }
-
-    return [...chars].sort().join("");
-}
-
-// ---------------------------------------------------------------------------
-// Deduplication: merge files with identical char sets
-// ---------------------------------------------------------------------------
-
-function deduplicate(routes: RouteCharSet[]): Map<string, string[]> {
-    // hash → [route1, route2, ...]
-    const groups = new Map<string, string[]>();
-    for (const r of routes) {
-        const list = groups.get(r.hash) ?? [];
-        list.push(r.route);
-        groups.set(r.hash, list);
-    }
-    return groups;
-}
-
-// ---------------------------------------------------------------------------
-// Main entry
-// ---------------------------------------------------------------------------
-
-/**
- * Run font subsetting during build.
- * Called from integration hooks.
- *
- * @param projectRoot - Project root directory
- * @param logger - Astro integration logger
- * @param skipPerRoute - Skip per-route subsets (for dev mode, only generate common)
- */
-export async function runFontSubsetting(
-    projectRoot: string,
-    logger: AstroIntegrationLogger,
-    skipPerRoute = false,
-): Promise<void> {
-    logger.info("Font subsetting: scanning content...");
-
-    // 1. Locate font file (check project root first, then stalux package dir)
-    const fontPaths = [
-        resolve(projectRoot, FONT_INPUT),
-        resolve(projectRoot, "node_modules", "@xingwangzhe", "stalux", FONT_INPUT),
-        resolve(dirname(fileURLToPath(import.meta.url)), "..", "..", FONT_INPUT),
-    ];
-
-    let fontPath: string | undefined;
-    for (const p of fontPaths) {
-        if (existsSync(p)) {
-            fontPath = p;
-            break;
-        }
-    }
-
-    if (!fontPath) {
-        logger.warn(`Font not found at ${FONT_INPUT}, skipping subsetting`);
-        return;
-    }
-    const fontBuffer = readFileSync(fontPath);
-
-    // Clean old generated font files to avoid stale cached references
-    const outDir = resolve(projectRoot, FONT_OUT_DIR);
-    if (existsSync(outDir)) {
-        for (const f of readdirSync(outDir)) {
-            if (f.startsWith("common-") || f.startsWith("subset-") || f === "common.css") {
-                try {
-                    unlinkSync(join(outDir, f));
-                } catch {
-                    /* skip */
-                }
-            }
-        }
-    }
-
-    // 1.5 代码字体（Google Sans Code）— 确定性静态 @font-face，不经过 Astro 字体管线
-    // Astro 的 <Font> 组件会在 prerender 阶段启动一个随机端口的 HTTP server，
-    // 端口号会嵌进 font-file-url-resolver 虚拟模块的编译产物，破坏增量构建依赖图 hash。
-    // 这里直接复制 woff2 到 public/fonts/ 并生成静态 code.css，完全绕开该管线。
-    try {
-        const codeFontPaths = [
-            resolve(projectRoot, CODE_FONT_INPUT),
-            resolve(projectRoot, "node_modules", "@xingwangzhe", "stalux", CODE_FONT_INPUT),
-            resolve(dirname(fileURLToPath(import.meta.url)), "..", "..", CODE_FONT_INPUT),
-        ];
-        const codeFontPath = codeFontPaths.find((p) => existsSync(p));
-        if (codeFontPath) {
-            const codeFontBuffer = readFileSync(codeFontPath);
-            const codeFontOut = join(outDir, "GoogleSansCode.woff2");
-            if (!existsSync(codeFontOut) || !readFileSync(codeFontOut).equals(codeFontBuffer)) {
-                writeFileSync(codeFontOut, codeFontBuffer);
-                logger.info(
-                    `  code font: ${codeFontBuffer.length / 1024} KB → GoogleSansCode.woff2`,
-                );
-            }
-            // 静态 CSS：@font-face + --font-code 变量（确定性内容，无 hash）
-            const codeCSS = `/* Auto-generated code font (deterministic, no Astro font pipeline) */
-:root {
-    --font-code: "Google Sans Code", "JetBrains Mono", "Fira Code", "Consolas", "Courier New", monospace;
-}
-@font-face {
-    font-family: "Google Sans Code";
-    src: url("/fonts/GoogleSansCode.woff2") format("woff2");
-    font-display: swap;
-}
-`;
-            writeFileSync(join(outDir, "code.css"), codeCSS);
-        } else {
-            logger.warn(
-                `Code font not found at ${CODE_FONT_INPUT}, --font-code falls back to system monospace`,
-            );
-        }
-    } catch (error) {
-        logger.warn(`Code font setup failed: ${String(error)}`);
-    }
-
-    // 2. Scan content files
-    const contentFiles = scanContent(projectRoot);
-    logger.info(`Font subsetting: found ${contentFiles.length} content files`);
-
-    // 3. Collect common chars (UI, nav, i18n)
-    const commonChars = collectCommonChars(projectRoot);
-    const commonHash = hashChars(commonChars);
-
-    // 4. Extract per-route char sets
-    const routeSets: RouteCharSet[] = [];
-    for (const file of contentFiles) {
-        let content: string;
-        if (file.route === "words" && wordCharsCache.has("words")) {
-            content = wordCharsCache.get("words")!;
-        } else if (virtualRouteCache.has(file.route)) {
-            content = virtualRouteCache.get(file.route)!;
-        } else {
-            content = readFileSync(file.filePath, "utf-8");
-        }
-        const chars = extractChars(content);
-        routeSets.push({
-            route: file.route,
-            chars,
-            hash: hashChars(chars),
-        });
-    }
-
-    // 5. Deduplicate and generate subsets using subset-font (WASM)
-    mkdirSync(outDir, { recursive: true });
-
-    // Helper: subset using subset-font (async WASM)
-    async function doSubset(chars: string): Promise<Uint8Array> {
-        if (!chars) return new Uint8Array(0);
-        try {
-            const result = await subsetFont(fontBuffer, chars, {
-                targetFormat: "sfnt", // "sfnt" = TTF output
-            });
-            return result;
-        } catch {
-            return new Uint8Array(0);
-        }
-    }
-
-    // Generate common subset
-    const commonFilename = `common-${commonHash}.ttf`;
-    const commonOutPath = join(outDir, commonFilename);
-    if (!existsSync(commonOutPath)) {
-        const data = await doSubset(commonChars);
-        if (data.length > 0) {
-            writeFileSync(commonOutPath, data);
-            logger.info(
-                `  common subset: ${(data.length / 1024).toFixed(1)} KB → ${commonFilename}`,
-            );
-        }
-    }
-
-    // Generate common CSS (content-addressed filename: 字符集变化 → hash 变化 →
-    // 文件名变化，浏览器必然拉新 CSS，避免缓存旧 CSS 引用已删除的旧 TTF 导致 404)
-    // dev 模式（skipPerRoute）用固定名 common.css，由 dev 中间件场景复用；
-    // 生产构建用 common-<hash>.css，HTML 从 manifest 读取引用。
-    const commonCssName = skipPerRoute ? "common.css" : `common-${commonHash}.css`;
-    const commonCSS = `/* Auto-generated common font subset */
-@font-face {
-    font-family: "LXGW WenKai Subset";
-    src: url("/fonts/${commonFilename}") format("truetype");
-    font-display: swap;
-}
-`;
-    writeFileSync(join(outDir, commonCssName), commonCSS);
-
-    const groups = deduplicate(routeSets);
-    const routeMap: Record<string, string> = {};
-    const postMap: Record<string, string> = {};
-
-    let subsetCount = 0;
-
-    if (!skipPerRoute) {
-        for (const [hash, routes] of groups) {
-            const filename = `subset-${hash}.ttf`;
-            const outPath = join(outDir, filename);
-            if (!existsSync(outPath)) {
-                const route = routeSets.find((r) => r.hash === hash);
-                const chars = route?.chars ?? "";
-                const uniqueChars = chars
-                    .split("")
-                    .filter((ch) => !commonChars.includes(ch))
-                    .join("");
-                if (uniqueChars.length > 0) {
-                    const data = await doSubset(uniqueChars);
-                    if (data.length > 0) {
-                        writeFileSync(outPath, data);
-                        subsetCount++;
-                        logger.info(
-                            `  subset ${subsetCount}: ${(data.length / 1024).toFixed(1)} KB → ${filename}`,
-                        );
-                    }
-                }
-            }
-
-            // Generate per-route CSS (each route gets its own CSS file)
-            // CSS 文件名内容寻址：内容只引用一个 TTF，文件名带上该 TTF 的 hash，
-            // 内容变化 → 文件名变化 → 浏览器必然拉新，避免缓存旧 CSS 引用已删除 TTF。
-            // If no unique chars, reference the common TTF as fallback
-            const refTtf = existsSync(outPath) ? filename : commonFilename;
-            const refHash = existsSync(outPath) ? hash : commonHash;
-            for (const route of routes) {
-                const cssContent = `/* Auto-generated font subset for route: ${route} */
-@font-face {
-    font-family: "LXGW WenKai Subset";
-    src: url("/fonts/${refTtf}") format("truetype");
-    font-display: swap;
-}
-`;
-                const cssName = route.replace("/", "-");
-                const cssFilename = `subset-${cssName}-${refHash}.css`;
-                writeFileSync(join(outDir, cssFilename), cssContent);
-                routeMap[route] = cssFilename;
-
-                if (route.startsWith("posts/")) {
-                    const abbrlink = route.replace("posts/", "");
-                    postMap[abbrlink] = cssFilename;
-                }
-            }
-        }
-    }
-
-    // 6. Generate manifest（dev 也生成：routes/posts 为空，供 Stalux.astro import）
-    const manifest: FontManifest = {
-        common: commonFilename,
-        commonCss: commonCssName,
-        routes: routeMap,
-        posts: postMap,
-    };
-    writeFileSync(join(outDir, "manifest.json"), JSON.stringify(manifest, null, 2));
-
-    // 副本写入包内 generated/：Stalux.astro import 它，让 manifest 进入 Astro
-    // 依赖图——manifest 变化时页面必然重建，增量构建 restored 的页面不会引用旧资源。
-    const genDir = resolve(dirname(fileURLToPath(import.meta.url)), "generated");
-    mkdirSync(genDir, { recursive: true });
-    writeFileSync(join(genDir, "fonts-manifest.json"), JSON.stringify(manifest));
-
-    logger.info(
-        `Font subsetting done: ${subsetCount} route subsets + common (${(fontBuffer.length / 1024 / 1024).toFixed(1)} MB → variable)`,
-    );
-}