dsh-plugin-lookatstudy 0.11.0 → 0.12.0

package/lib/html-article-Da8ksU0i.mjs

@@ -0,0 +1,460 @@
+//#region src/vendor/html-article.ts
+const ENTITIES = {
+	amp: "&",
+	lt: "<",
+	gt: ">",
+	quot: "\"",
+	apos: "'",
+	nbsp: "\xA0",
+	hellip: "…",
+	mdash: "—",
+	ndash: "–",
+	rsquo: "'",
+	lsquo: "'",
+	rdquo: "\"",
+	ldquo: "\"",
+	middot: "·",
+	bull: "•",
+	copy: "©",
+	times: "×",
+	divide: "÷",
+	deg: "°"
+};
+function decodeEntities(s) {
+	return s.replace(/&(#x?[0-9a-fA-F]+|[a-zA-Z]+);/g, (_m, ent) => {
+		if (ent.startsWith("#x") || ent.startsWith("#X")) return String.fromCodePoint(parseInt(ent.slice(2), 16));
+		if (ent.startsWith("#")) return String.fromCodePoint(parseInt(ent.slice(1), 10));
+		return ENTITIES[ent.toLowerCase()] ?? _m;
+	});
+}
+const VOID_TAGS = /* @__PURE__ */ new Set([
+	"br",
+	"hr",
+	"img",
+	"input",
+	"meta",
+	"link",
+	"area",
+	"base",
+	"col",
+	"embed",
+	"source",
+	"track",
+	"wbr"
+]);
+/** Stack-parse well-formed-enough HTML/XHTML into a forest of nodes. */
+function parseHtml(html) {
+	const roots = [];
+	const stack = [];
+	const push = (n) => {
+		(stack.length ? stack[stack.length - 1].children : roots).push(n);
+	};
+	const tokenRe = /<!--[\s\S]*?-->|<!\[CDATA\[([\s\S]*?)\]\]>|<\/([a-zA-Z][-\w:]*)\s*>|<([a-zA-Z][-\w:]*)((?:"[^"]*"|'[^']*'|[^>"'])*?)(\/?)>|([^<]+)/g;
+	let m;
+	while ((m = tokenRe.exec(html)) !== null) if (m[2] !== void 0) {
+		const close = m[2].toLowerCase();
+		for (let i = stack.length - 1; i >= 0; i--) if (stack[i].tag === close) {
+			stack.length = i;
+			break;
+		}
+	} else if (m[3] !== void 0) {
+		const tag = m[3].toLowerCase();
+		const attrs = {};
+		const attrRe = /([-\w:]+)\s*=\s*("([^"]*)"|'([^']*)')/g;
+		let a;
+		while ((a = attrRe.exec(m[4] ?? "")) !== null) attrs[a[1].toLowerCase()] = decodeEntities(a[3] ?? a[4] ?? "");
+		const node = {
+			tag,
+			attrs,
+			children: [],
+			text: ""
+		};
+		push(node);
+		if (!VOID_TAGS.has(tag) && m[5] !== "/") stack.push(node);
+	} else if (m[1] !== void 0) push({
+		tag: "",
+		attrs: {},
+		children: [],
+		text: m[1]
+	});
+	else if (m[6] !== void 0) {
+		const text = decodeEntities(m[6]);
+		if (text.trim()) push({
+			tag: "",
+			attrs: {},
+			children: [],
+			text
+		});
+	}
+	return roots;
+}
+const BLOCK_TAGS = /* @__PURE__ */ new Set([
+	"p",
+	"div",
+	"section",
+	"article",
+	"header",
+	"footer",
+	"main",
+	"aside",
+	"nav",
+	"figure",
+	"figcaption",
+	"h1",
+	"h2",
+	"h3",
+	"h4",
+	"h5",
+	"h6",
+	"ul",
+	"ol",
+	"li",
+	"blockquote",
+	"pre",
+	"table",
+	"tr",
+	"thead",
+	"tbody",
+	"hr",
+	"br",
+	"address",
+	"details",
+	"summary"
+]);
+function nodeText(node) {
+	if (node.tag === "") return node.text;
+	return node.children.map(nodeText).join("");
+}
+function findDescendant(node, tag, cls) {
+	for (const c of node.children) {
+		if (c.tag === tag && (!cls || (c.attrs["class"] ?? "").split(/\s+/).includes(cls))) return c;
+		const hit = findDescendant(c, tag, cls);
+		if (hit) return hit;
+	}
+	return null;
+}
+function inlineRuns(nodes, stripImages) {
+	let out = "";
+	for (const n of nodes) switch (n.tag) {
+		case "":
+			out += n.text.replace(/\s+/g, " ");
+			break;
+		case "br":
+			out += "\n";
+			break;
+		case "strong":
+		case "b": {
+			const inner = inlineRuns(n.children, stripImages).trim();
+			if (inner) out += `**${inner}**`;
+			break;
+		}
+		case "em":
+		case "i": {
+			const inner = inlineRuns(n.children, stripImages).trim();
+			if (inner) out += `*${inner}*`;
+			break;
+		}
+		case "code":
+			out += "`" + nodeText(n).trim() + "`";
+			break;
+		case "a": {
+			const inner = inlineRuns(n.children, stripImages).trim();
+			const href = n.attrs["href"] ?? "";
+			out += inner && href && !href.startsWith("#") ? `[${inner}](${href})` : inner;
+			break;
+		}
+		case "img":
+		case "picture":
+		case "svg":
+		case "figure":
+			if (!stripImages && n.tag === "img") {
+				const alt = n.attrs["alt"] ?? "";
+				const src = n.attrs["src"] ?? "";
+				if (src) out += `![${alt}](${src})`;
+			}
+			break;
+		case "script":
+			if ((n.attrs["type"] ?? "").includes("math/tex")) {
+				const tex = nodeText(n).trim();
+				if (tex) out += tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`;
+			}
+			break;
+		case "style":
+		case "head": break;
+		case "span":
+			if ((n.attrs["class"] ?? "").split(/\s+/).includes("katex")) {
+				const ann = findDescendant(n, "annotation");
+				const tex = ann ? nodeText(ann).trim() : "";
+				if (tex) out += tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`;
+				break;
+			}
+			out += inlineRuns(n.children, stripImages);
+			break;
+		default: out += inlineRuns(n.children, stripImages);
+	}
+	return out.replace(/[ \t]+\n/g, "\n");
+}
+function emitBlocks(nodes, stripImages, lines, listDepth = 0, quotePrefix = "") {
+	const indent = "  ".repeat(listDepth);
+	for (const n of nodes) {
+		const prefix = quotePrefix;
+		switch (n.tag) {
+			case "": {
+				const t = n.text.trim();
+				if (t) lines.push(prefix + t);
+				break;
+			}
+			case "h1":
+			case "h2":
+			case "h3":
+			case "h4":
+			case "h5":
+			case "h6": {
+				const level = Number(n.tag[1]);
+				const t = inlineRuns(n.children, stripImages).trim();
+				if (t) lines.push(prefix + "#".repeat(level) + " " + t);
+				break;
+			}
+			case "p":
+			case "div":
+			case "section":
+			case "article":
+			case "main":
+			case "figcaption":
+			case "summary":
+			case "details":
+			case "address": {
+				if (n.children.some((c) => BLOCK_TAGS.has(c.tag) && c.tag !== "br")) {
+					emitBlocks(n.children, stripImages, lines, listDepth, prefix);
+					break;
+				}
+				const t = inlineRuns(n.children, stripImages).trim();
+				if (t) lines.push(prefix + indent + t);
+				break;
+			}
+			case "ul":
+			case "ol": {
+				let idx = 1;
+				for (const c of n.children) if (c.tag === "li") {
+					const marker = n.tag === "ol" ? `${idx++}. ` : "- ";
+					if (!c.children.some((cc) => BLOCK_TAGS.has(cc.tag) && cc.tag !== "br")) {
+						const t = inlineRuns(c.children, stripImages).trim();
+						if (t) lines.push(prefix + indent + marker + t);
+					} else {
+						const nested = [];
+						emitBlocks(c.children, stripImages, nested, listDepth + 1, prefix);
+						if (nested.length) lines.push(prefix + indent + marker + nested[0].trimStart());
+						lines.push(...nested.slice(1));
+					}
+				} else emitBlocks([c], stripImages, lines, listDepth, prefix);
+				lines.push("");
+				break;
+			}
+			case "blockquote": {
+				const inner = [];
+				emitBlocks(n.children, stripImages, inner, 0, "");
+				for (const l of inner) if (l.trim()) lines.push(prefix + "> " + l);
+				lines.push("");
+				break;
+			}
+			case "pre": {
+				const code = n.children.map((c) => nodeText(c)).join("").replace(/\n+$/, "");
+				if (code.trim()) {
+					lines.push(prefix + "```", code.split("\n").map((l) => prefix + l).join("\n"), prefix + "```");
+					lines.push("");
+				}
+				break;
+			}
+			case "table":
+				for (const c of n.children) if (c.tag === "tr") {
+					const cells = c.children.filter((cc) => cc.tag === "td" || cc.tag === "th").map((cc) => inlineRuns(cc.children, stripImages).trim().replace(/\|/g, "\\|"));
+					if (cells.length) lines.push(prefix + "| " + cells.join(" | ") + " |");
+				} else emitBlocks([c], stripImages, lines, listDepth, prefix);
+				lines.push("");
+				break;
+			case "hr":
+				lines.push(prefix + "---");
+				lines.push("");
+				break;
+			case "br":
+				lines.push("");
+				break;
+			case "script":
+				if ((n.attrs["type"] ?? "").includes("math/tex")) {
+					const tex = nodeText(n).trim();
+					if (tex) lines.push(prefix + (tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`));
+				}
+				break;
+			case "span": {
+				if ((n.attrs["class"] ?? "").split(/\s+/).includes("katex")) {
+					const ann = findDescendant(n, "annotation");
+					const tex = ann ? nodeText(ann).trim() : "";
+					if (tex) lines.push(prefix + (tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`));
+					break;
+				}
+				const t = inlineRuns([n], stripImages).trim();
+				if (t) lines.push(prefix + t);
+				break;
+			}
+			case "img":
+			case "picture":
+			case "svg":
+			case "figure":
+				if (!stripImages && n.tag === "img") {
+					const src = n.attrs["src"] ?? "";
+					if (src) lines.push(prefix + `![${n.attrs["alt"] ?? ""}](${src})`);
+				}
+				break;
+			case "style":
+			case "head":
+			case "nav":
+			case "button":
+			case "form":
+			case "noscript": break;
+			default: emitBlocks(n.children, stripImages, lines, listDepth, prefix);
+		}
+	}
+}
+/** 任意 HTML/XHTML → markdown（epub 章节用；数学源回收在 script 剥除之前，同上游）。 */
+function htmlToMarkdown(html, opts = {}) {
+	const stripImages = opts.stripImages ?? false;
+	let roots = parseHtml(html);
+	if (roots.length === 0) roots = parseHtml(`<html><body>${html}</body></html>`);
+	const lines = [];
+	emitBlocks(roots, stripImages, lines);
+	return lines.join("\n").replace(/\n{3,}/g, "\n\n").trim();
+}
+const STRIP_FOR_ARTICLE = /* @__PURE__ */ new Set([
+	"script",
+	"style",
+	"nav",
+	"header",
+	"footer",
+	"aside",
+	"form",
+	"noscript",
+	"button",
+	"iframe",
+	"svg"
+]);
+function textDensity(nodes) {
+	let chars = 0, tags = 0;
+	const walk = (ns) => {
+		for (const n of ns) {
+			if (n.tag === "") {
+				chars += n.text.trim().length;
+				continue;
+			}
+			if (STRIP_FOR_ARTICLE.has(n.tag) || n.tag === "div" && (n.attrs["id"] ?? n.attrs["class"] ?? "").match(/comment|sidebar|related|share|footer|nav/i)) continue;
+			tags++;
+			walk(n.children);
+		}
+	};
+	walk(nodes);
+	return {
+		chars,
+		tags
+	};
+}
+/** Pick the densest content root: <article> > <main> > <body> > whole forest. */
+function pickContentRoot(roots) {
+	const findFirst = (tag) => {
+		const walk = (ns) => {
+			for (const n of ns) {
+				if (n.tag === tag) return n;
+				const hit = walk(n.children);
+				if (hit) return hit;
+			}
+			return null;
+		};
+		return walk(roots);
+	};
+	return [findFirst("article") ?? findFirst("main") ?? findFirst("body") ?? {
+		tag: "",
+		attrs: {},
+		children: roots,
+		text: ""
+	}];
+}
+/** 从完整 HTML 抽取文章正文并转 markdown。非文章页返回 null（诚实失败，同上游契约）。 */
+function extractArticle(html, baseUrl = "") {
+	let roots = parseHtml(html);
+	if (roots.length === 0) return null;
+	const titleNode = findDescendant({
+		tag: "root",
+		attrs: {},
+		children: roots,
+		text: ""
+	}, "title");
+	const h1 = findDescendant({
+		tag: "root",
+		attrs: {},
+		children: roots,
+		text: ""
+	}, "h1");
+	const title = (titleNode ? nodeText(titleNode) : h1 ? inlineRuns(h1.children, false).trim() : "").trim() || "无标题文章";
+	const contentRoots = pickContentRoot(roots);
+	if (baseUrl) {
+		const walk = (ns) => {
+			for (const n of ns) {
+				if (n.tag === "img") {
+					const raw = n.attrs["src"] ?? "";
+					if (raw && !raw.startsWith("data:") && !/^https?:/i.test(raw)) try {
+						n.attrs["src"] = new URL(raw, baseUrl).toString();
+					} catch {}
+				}
+				walk(n.children);
+			}
+		};
+		walk(contentRoots);
+	}
+	const { chars } = textDensity(contentRoots);
+	if (chars < 120) return null;
+	const body = htmlToMarkdownOf(contentRoots, false);
+	if (!body || body.replace(/[#\s>*|-]/g, "").length < 40) return null;
+	return {
+		title,
+		markdown: stripTailNavigation(body.split("\n")[0]?.replace(/^#\s+/, "").trim() === title && body.startsWith("# ") ? body : `# ${title}\n\n${body}`)
+	};
+}
+/**
+* 尾部**站点模板指纹**清理(upstream v0.23.1, 2026-08-23 真实站点采样驱动; vendored verbatim)。
+* 原则:规则只管高置信度的**机器生成模板**(跨文章稳定、作为正文出现概率≈0);
+* 除此之外的不确定判断(作者自己写的推广段/水印/相关阅读算不算正文)一律不猜——
+* 那是课程设计层(tutor/Step4)的职责,解析层动了就是误删正文
+* (实测:按"欢迎关注公众号"删规则,把 CSDN 作者自己写的推广段删了)。
+*/
+const NAV_TAIL_PATTERNS = [
+	/返回\S{0,6}[，,]?\s*查看更多/,
+	/点击进入\S{0,8}首页/,
+	/^热门文章$/,
+	/^最新文章$/,
+	/^目录\s*$/,
+	/^END\b.*版权/
+];
+/** 行内导航后缀:正文与模板被并成一行时剥掉(精确短语,全文安全) */
+const INLINE_NAV_SUFFIXES = ["目录 热门文章 最新文章"];
+function stripTailNavigation(md) {
+	if (!md) return md;
+	for (const suffix of INLINE_NAV_SUFFIXES) md = md.split(suffix).join("");
+	const lines = md.split("\n");
+	let end = lines.length;
+	for (let i = lines.length - 1; i >= Math.max(0, lines.length - 25); i--) {
+		const ln = lines[i].trim();
+		if (!ln) continue;
+		const bare = ln.replace(/^#{1,6}\s*/, "").replace(/^[>\-*]\s*/, "");
+		const isNav = NAV_TAIL_PATTERNS.some((re) => re.test(bare) || re.test(ln));
+		const isBareImage = /^!\[[^\]]*\]\([^)]*\)$/.test(ln) || /^\S+\.(jpeg|jpg|png|gif|webp)\)?$/i.test(ln);
+		if (isNav || isBareImage) continue;
+		end = i + 1;
+		break;
+	}
+	return lines.slice(0, end).join("\n").trimEnd();
+}
+function htmlToMarkdownOf(roots, stripImages) {
+	const lines = [];
+	emitBlocks(roots, stripImages, lines);
+	return lines.join("\n").replace(/\n{3,}/g, "\n\n").trim();
+}
+//#endregion
+export { htmlToMarkdown as n, extractArticle as t };

package/lib/index.d.mts

@@ -32,10 +32,11 @@ declare const Config: z<Config>;
 declare const name = "lookatstudy-plugin";
 declare const inject: string[];
 /**
- * Register the activation-gated study surface: the 20 `study_*` tools (kept
- * unregistered while dormant), the tutor persona (stable core + soul), and
- * the dynamic learner-snapshot context — every prompt text renders empty
- * while inactive, and empty sections are dropped at assembly.
+ * Register the activation-gated study surface: the 25 `study_*` tools (kept
+ * unregistered while dormant), the tutor persona (stable core + soul), the
+ * dynamic learner-snapshot context, and the `/study` command — every prompt
+ * text renders empty while inactive, and empty sections are dropped at
+ * assembly.
  * @param ctx - plugin context carrying the tool registry and system prompt.
  * @param config - validated plugin configuration.
  */