From 3831171c06d154fa7d4c444302110ed787baeb32 Mon Sep 17 00:00:00 2001
From: seaHi <me@seahi.me>
Date: Wed, 16 Sep 2026 21:22:04 +0800
Subject: [PATCH] =?UTF-8?q?feat:=20=E6=94=AF=E6=8C=81=E5=9C=A8=E6=96=87?=
 =?UTF-8?q?=E7=AB=A0=E6=91=98=E8=A6=81=E4=B8=AD=E4=BF=9D=E7=95=99=E5=B9=B6?=
 =?UTF-8?q?=E6=AD=A3=E7=A1=AE=E6=B8=B2=E6=9F=93=E8=A1=8C=E5=86=85=E6=A0=B7?=
 =?UTF-8?q?=E5=BC=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 src/lib/markdown/excerpt.test.ts             |  63 +++++++++
 src/lib/markdown/excerpt.ts                  | 129 +++++++++++++++++++
 src/routes/+page.server.ts                   |  23 +---
 src/routes/+page.svelte                      |   3 +-
 src/routes/categories/[slug]/+page.server.ts |  16 +--
 src/routes/categories/[slug]/+page.svelte    |   3 +-
 src/routes/tags/[tag]/+page.server.ts        |  16 +--
 src/routes/tags/[tag]/+page.svelte           |   3 +-
 8 files changed, 207 insertions(+), 49 deletions(-)
 create mode 100644 src/lib/markdown/excerpt.test.ts
 create mode 100644 src/lib/markdown/excerpt.ts

diff --git a/src/lib/markdown/excerpt.test.ts b/src/lib/markdown/excerpt.test.ts
new file mode 100644
index 0000000..3edfedf
--- /dev/null
+++ b/src/lib/markdown/excerpt.test.ts
@@ -0,0 +1,63 @@
+import { describe, expect, it } from 'vitest';
+import { createExcerptHtml } from './excerpt';
+
+describe('createExcerptHtml', () => {
+	it('renders inline styles instead of stripping them', () => {
+		expect(createExcerptHtml('这是**粗体**与*斜体*，还有~~删除线~~。')).toBe(
+			'这是<strong>粗体</strong>与<em>斜体</em>，还有<del>删除线</del>。...'
+		);
+	});
+
+	it('keeps inline code and drops the backticks', () => {
+		expect(createExcerptHtml('先执行 `pnpm run build` 再发布')).toBe(
+			'先执行 <code>pnpm run build</code> 再发布...'
+		);
+	});
+
+	it('reduces links and images to readable text', () => {
+		expect(createExcerptHtml('参考[官方文档](https://example.com)与![配图](cover.png)说明')).toBe(
+			'参考官方文档与配图说明...'
+		);
+	});
+
+	it('removes block level markers so the excerpt stays a single line', () => {
+		expect(createExcerptHtml('# 标题\n\n- 第一项\n- 第二项\n\n> 引用内容')).toBe(
+			'标题 第一项 第二项 引用内容...'
+		);
+	});
+
+	it('drops Hugo shortcodes and raw HTML', () => {
+		expect(
+			createExcerptHtml('{{< admonition note "提示" >}}正文{{< /admonition >}}<span>片段</span>')
+		).toBe('正文片段...');
+	});
+
+	it('closes inline tags that are cut off by the length limit', () => {
+		const html = createExcerptHtml(`**${'a'.repeat(200)}**`, 10);
+
+		expect(html).toBe(`<strong>${'a'.repeat(10)}</strong>...`);
+	});
+
+	it('counts HTML entities as a single visible character', () => {
+		expect(createExcerptHtml('&'.repeat(30), 5)).toBe(`${'&amp;'.repeat(5)}...`);
+	});
+
+	it('never keeps a dangling inline tag when the cut lands on it', () => {
+		const html = createExcerptHtml(`${'a'.repeat(9)}**bc**`, 10);
+
+		expect(html).toBe(`aaaaaaaaa<strong>b</strong>...`);
+	});
+
+	it('escapes characters that would otherwise break the markup', () => {
+		const html = createExcerptHtml('对比 1 < 2 的结果 <script>alert(1)</script>');
+
+		expect(html).toContain('1 &lt; 2');
+		expect(html).not.toContain('<script');
+	});
+
+	it('returns an empty string when there is nothing to show', () => {
+		expect(createExcerptHtml('')).toBe('');
+		expect(createExcerptHtml('   \n\n  ')).toBe('');
+		expect(createExcerptHtml('{{< admonition note >}}{{< /admonition >}}')).toBe('');
+	});
+});
diff --git a/src/lib/markdown/excerpt.ts b/src/lib/markdown/excerpt.ts
new file mode 100644
index 0000000..3e98f1d
--- /dev/null
+++ b/src/lib/markdown/excerpt.ts
@@ -0,0 +1,129 @@
+import { Marked } from 'marked';
+import sanitizeHtml from 'sanitize-html';
+
+/** 摘要默认保留的可见字符数 */
+export const EXCERPT_LENGTH = 150;
+
+/** 摘要中允许保留的行内样式标签（粗体、斜体、删除线、行内代码） */
+export const EXCERPT_INLINE_TAGS = ['strong', 'em', 'del', 'code'];
+
+const excerptMarked = new Marked();
+
+/** 仅匹配摘要白名单内的行内标签，用于截断时配对 */
+const INLINE_TAG_PATTERN = new RegExp(
+	`^<\\s*(/?)\\s*(${EXCERPT_INLINE_TAGS.join('|')})\\s*>$`,
+	'i'
+);
+
+/** 匹配没有任何内容的行内标签，例如 `<strong></strong>` */
+const EMPTY_TAG_PATTERN = new RegExp(`<(${EXCERPT_INLINE_TAGS.join('|')})>\\s*</\\1>`, 'gi');
+
+/** 匹配 HTML 标签，要求以字母开头，避免误伤 `1 < 2` 这类正文 */
+const HTML_TAG_PATTERN = /<\/?[a-zA-Z][^>]*>/g;
+
+/** 匹配 HTML 实体，截断时按一个可见字符计算 */
+const HTML_ENTITY_PATTERN = /^&(?:#\d+|#x[0-9a-fA-F]+|[a-zA-Z]+);/;
+
+/**
+ * 把文章内容整理成适合做摘要的单行文本：
+ * 去掉 shortcode / HTML / 块级 Markdown 标记，保留粗体、斜体等行内标记。
+ */
+function normalizeExcerptSource(source: string): string {
+	return source
+		.replace(/\{\{[<%][\s\S]*?[>%]\}\}/g, '') // Hugo / Docsy shortcode
+		.replace(/<!--[\s\S]*?-->/g, '') // HTML 注释
+		.replace(HTML_TAG_PATTERN, '') // HTML 标签
+		.replace(/!\[([^\]]*)\]\([^)]*\)/g, '$1') // 图片 → alt 文本
+		.replace(/\[([^\]]+)\]\([^)]*\)/g, '$1') // 链接 → 链接文本
+		.replace(/^[ \t]{0,3}(?:`{3,}|~{3,})[^\n]*$/gm, '') // 代码块围栏
+		.replace(/^[ \t]{0,3}#{1,6}[ \t]+/gm, '') // ATX 标题
+		.replace(/^[ \t]{0,3}>[ \t]?/gm, '') // 引用
+		.replace(/^[ \t]{0,3}(?:[-*+]|\d{1,9}[.)])[ \t]+/gm, '') // 列表标记
+		.replace(/^[ \t]{0,3}\|?[ \t]*:?-{3,}:?[ \t]*(?:\|[ \t]*:?-{3,}:?[ \t]*)*\|?[ \t]*$/gm, '') // 表格分隔行
+		.replace(/^[ \t]{0,3}(?:[*_-][ \t]*){3,}$/gm, '') // 分隔线
+		.replace(/\|/g, ' ') // 表格竖线
+		.replace(/\s+/g, ' ')
+		.trim();
+}
+
+/**
+ * 按可见字符数截断行内 HTML，并补齐被截断的标签。
+ * 只接受白名单内的行内标签，遇到其它标签立即停止，保证输出安全。
+ */
+function truncateInlineHtml(html: string, maxLength: number): string {
+	let result = '';
+	let visible = 0;
+	const openTags: string[] = [];
+
+	for (let index = 0; index < html.length && visible < maxLength;) {
+		const char = html[index];
+
+		if (char === '<') {
+			const end = html.indexOf('>', index);
+			if (end === -1) break;
+
+			const tag = html.slice(index, end + 1);
+			const match = INLINE_TAG_PATTERN.exec(tag);
+			if (!match) break;
+
+			const name = match[2].toLowerCase();
+			if (match[1]) {
+				const position = openTags.lastIndexOf(name);
+				if (position !== -1) openTags.splice(position, 1);
+			} else {
+				openTags.push(name);
+			}
+
+			result += tag;
+			index = end + 1;
+			continue;
+		}
+
+		if (char === '&') {
+			const entity = HTML_ENTITY_PATTERN.exec(html.slice(index));
+			if (entity) {
+				result += entity[0];
+				visible += 1;
+				index += entity[0].length;
+				continue;
+			}
+		}
+
+		result += char;
+		visible += 1;
+		index += 1;
+	}
+
+	for (const name of openTags.reverse()) {
+		result += `</${name}>`;
+	}
+
+	let cleaned = result;
+	for (;;) {
+		const next = cleaned.replace(EMPTY_TAG_PATTERN, '').replace(/\s+$/, '');
+		if (next === cleaned) break;
+		cleaned = next;
+	}
+
+	return cleaned;
+}
+
+/**
+ * 由文章内容生成带行内样式的摘要 HTML（粗体、斜体、删除线、行内代码）。
+ * 返回值已做白名单净化，可安全地用 `{@html}` 渲染；无内容时返回空字符串。
+ */
+export function createExcerptHtml(source: string, maxLength: number = EXCERPT_LENGTH): string {
+	const normalized = normalizeExcerptSource(source || '');
+	if (!normalized) return '';
+
+	// 先解析再截断：源文本必须先完整解析，否则成对的行内标记会被切断而失效
+	const html = excerptMarked.parseInline(normalized) as string;
+	const sanitized = sanitizeHtml(html, {
+		allowedTags: EXCERPT_INLINE_TAGS,
+		allowedAttributes: {},
+		disallowedTagsMode: 'discard'
+	});
+
+	const truncated = truncateInlineHtml(sanitized, maxLength).trim();
+	return truncated ? `${truncated}...` : '';
+}
diff --git a/src/routes/+page.server.ts b/src/routes/+page.server.ts
index 301cfcb..092034c 100644
--- a/src/routes/+page.server.ts
+++ b/src/routes/+page.server.ts
@@ -1,5 +1,6 @@
 /* eslint-disable @typescript-eslint/no-explicit-any */
 
+import { createExcerptHtml } from '$lib/markdown/excerpt';
 import type { PageServerLoad } from './$types';
 
 export const load: PageServerLoad = async ({ platform }) => {
@@ -47,25 +48,11 @@ export const load: PageServerLoad = async ({ platform }) => {
 			const day = d.getDate();
 			const year = d.getFullYear();
 
-			let excerpt;
-			let contentStr = (post.content as string) || '';
+			// Prefer the manually marked summary, otherwise derive it from the beginning
+			const contentStr = ((post.content as string) || '').split('<!--more-->')[0];
 
-			// Filter out Hugo shortcodes like {{< admonition abstract "标题" >}}
-			contentStr = contentStr.replace(/\{\{<[\s\S]*?>\}\}/g, '');
-
-			const parts = contentStr.split('<!--more-->');
-			if (parts.length > 1) {
-				excerpt = parts[0].trim();
-			} else {
-				// Strip basic markdown syntax for the automatic excerpt
-				const plainText = contentStr.replace(/[#*`>~[\]()-]/g, '').trim();
-				excerpt = plainText.substring(0, 150).trim();
-			}
-
-			// Append ... to indicate there is more to read
-			if (excerpt && !excerpt.endsWith('...')) {
-				excerpt += '...';
-			}
+			// Inline styles (bold, italic, ...) are kept as safe HTML for the cards
+			const excerpt = createExcerptHtml(contentStr);
 
 			// Do not send full content to the list view to save bandwidth
 
diff --git a/src/routes/+page.svelte b/src/routes/+page.svelte
index 4bfc3f3..5049c9f 100644
--- a/src/routes/+page.svelte
+++ b/src/routes/+page.svelte
@@ -24,7 +24,8 @@
 					{/if}
 					<div class="content-wrap">
 						<h1 class="title">{post.title}</h1>
-						<p class="excerpt">{post.excerpt}</p>
+						<!-- eslint-disable-next-line svelte/no-at-html-tags -->
+						<p class="excerpt">{@html post.excerpt}</p>
 					</div>
 				</article>
 			</a>
diff --git a/src/routes/categories/[slug]/+page.server.ts b/src/routes/categories/[slug]/+page.server.ts
index 984bd42..a45a068 100644
--- a/src/routes/categories/[slug]/+page.server.ts
+++ b/src/routes/categories/[slug]/+page.server.ts
@@ -1,4 +1,5 @@
 import { error, isHttpError } from '@sveltejs/kit';
+import { createExcerptHtml } from '$lib/markdown/excerpt';
 import type { PageServerLoad } from './$types';
 
 const PAGE_SIZE = 20;
@@ -12,19 +13,6 @@ type CategoryPostRow = {
 	excerpt_source: string;
 };
 
-function createExcerpt(source: string): string {
-	const plainText = source
-		.replace(/\{\{<[\s\S]*?>\}\}/g, '')
-		.replace(/<[^>]+>/g, '')
-		.replace(/[#*`>~[\]()-]/g, '')
-		.replace(/\s+/g, ' ')
-		.trim()
-		.slice(0, 150)
-		.trim();
-
-	return plainText ? `${plainText}...` : '';
-}
-
 export const load: PageServerLoad = async ({ platform, params, url, parent }) => {
 	const db = platform?.env.DB;
 	const { aliases } = await parent();
@@ -84,7 +72,7 @@ export const load: PageServerLoad = async ({ platform, params, url, parent }) =>
 							year: 'numeric',
 							timeZone: 'UTC'
 						}),
-				excerpt: createExcerpt(post.excerpt_source || '')
+				excerpt: createExcerptHtml(post.excerpt_source || '')
 			};
 		});
 
diff --git a/src/routes/categories/[slug]/+page.svelte b/src/routes/categories/[slug]/+page.svelte
index 85f7be4..01427c7 100644
--- a/src/routes/categories/[slug]/+page.svelte
+++ b/src/routes/categories/[slug]/+page.svelte
@@ -38,7 +38,8 @@
 						{/if}
 						<div class="content-wrap">
 							<h1 class="title">{post.title}</h1>
-							<p class="excerpt">{post.excerpt}</p>
+							<!-- eslint-disable-next-line svelte/no-at-html-tags -->
+							<p class="excerpt">{@html post.excerpt}</p>
 						</div>
 					</article>
 				</a>
diff --git a/src/routes/tags/[tag]/+page.server.ts b/src/routes/tags/[tag]/+page.server.ts
index a981d02..38649a8 100644
--- a/src/routes/tags/[tag]/+page.server.ts
+++ b/src/routes/tags/[tag]/+page.server.ts
@@ -1,4 +1,5 @@
 import { error, isHttpError } from '@sveltejs/kit';
+import { createExcerptHtml } from '$lib/markdown/excerpt';
 import type { PageServerLoad } from './$types';
 
 const PAGE_SIZE = 20;
@@ -12,19 +13,6 @@ type TagPostRow = {
 	excerpt_source: string;
 };
 
-function createExcerpt(source: string): string {
-	const plainText = source
-		.replace(/\{\{<[\s\S]*?>\}\}/g, '')
-		.replace(/<[^>]+>/g, '')
-		.replace(/[#*`>~[\]()-]/g, '')
-		.replace(/\s+/g, ' ')
-		.trim()
-		.slice(0, 150)
-		.trim();
-
-	return plainText ? `${plainText}...` : '';
-}
-
 export const load: PageServerLoad = async ({ platform, params, url, parent }) => {
 	const db = platform?.env.DB;
 	const { aliases } = await parent();
@@ -85,7 +73,7 @@ export const load: PageServerLoad = async ({ platform, params, url, parent }) =>
 							year: 'numeric',
 							timeZone: 'UTC'
 						}),
-				excerpt: createExcerpt(post.excerpt_source || '')
+				excerpt: createExcerptHtml(post.excerpt_source || '')
 			};
 		});
 
diff --git a/src/routes/tags/[tag]/+page.svelte b/src/routes/tags/[tag]/+page.svelte
index f3686fc..4d41671 100644
--- a/src/routes/tags/[tag]/+page.svelte
+++ b/src/routes/tags/[tag]/+page.svelte
@@ -37,7 +37,8 @@
 						{/if}
 						<div class="content-wrap">
 							<h1 class="title">{post.title}</h1>
-							<p class="excerpt">{post.excerpt}</p>
+							<!-- eslint-disable-next-line svelte/no-at-html-tags -->
+							<p class="excerpt">{@html post.excerpt}</p>
 						</div>
 					</article>
 				</a>