hd-garden 0.10.0 → 0.10.2

package/scripts/link-meta-core.mjs

@@ -184,6 +184,17 @@ const absolute = (value, base) => {
   }
 };
 
+
+/** Site brand OG path `/og.png` (any host) — not an article cover for link cards. */
+export function isSiteBrandOgImage(url) {
+  if (!url) return false;
+  try {
+    return new URL(url).pathname === "/og.png";
+  } catch {
+    return false;
+  }
+}
+
 /** Title, description, image, site name, author, date and icon candidates from a page's head. */
 export function parseHead(html, baseUrl) {
   const head = (/<head[\s>][\s\S]*?(<\/head>|$)/iu.exec(html)?.[0]) ?? html.slice(0, 200_000);
@@ -210,10 +221,11 @@ export function parseHead(html, baseUrl) {
   const published = meta["article:published_time"] || meta["og:published_time"] || meta.date || meta.datepublished || meta["parsely-pub-date"] || "";
   const publishedTime = published && !Number.isNaN(Date.parse(published)) ? new Date(Date.parse(published)).toISOString().slice(0, 10) : "";
   const author = meta["article:author"] || meta.author || meta["twitter:creator"] || "";
+  const imageCandidate = absolute(meta["og:image:secure_url"] || meta["og:image"] || meta["og:image:url"] || meta["twitter:image"], baseUrl);
   return {
     title: clean(meta["og:title"] || meta["twitter:title"] || title, 200),
     description: clean(meta["og:description"] || meta["twitter:description"] || meta.description, 300),
-    image: absolute(meta["og:image:secure_url"] || meta["og:image"] || meta["og:image:url"] || meta["twitter:image"], baseUrl),
+    image: isSiteBrandOgImage(imageCandidate) ? "" : imageCandidate,
     siteName: clean(meta["og:site_name"] || meta["application-name"], 80),
     // article:author is sometimes a profile URL; only keep readable names.
     author: /^https?:\/\//iu.test(author) ? "" : clean(author, 80),
@@ -222,6 +234,18 @@ export function parseHead(html, baseUrl) {
   };
 }
 
+/** Target of an HTML meta refresh, when the page asks the browser to move. */
+export function metaRefreshTarget(html, baseUrl) {
+  for (const tag of html.match(/<meta\b[^>]*>/giu) ?? []) {
+    const attrs = attributes(tag);
+    if ((attrs["http-equiv"] || "").toLowerCase() !== "refresh") continue;
+    const match = /(?:^|;)\s*url\s*=\s*["']?([^"'>\s]+)/iu.exec(attrs.content || "");
+    const target = absolute(match?.[1] ?? "", baseUrl);
+    if (target) return target;
+  }
+  return "";
+}
+
 /** A page's head, fetched safely: HTML only, at most maxBytes, stopping at </head>. */
 export async function fetchHead(url, { fetch: fetcher, timeoutMs = DEFAULT_TIMEOUT_MS, maxBytes = DEFAULT_MAX_HTML_BYTES } = {}) {
   const { response, url: finalUrl } = await safeFetch(url, { fetch: fetcher, timeoutMs });
@@ -232,9 +256,22 @@ export async function fetchHead(url, { fetch: fetcher, timeoutMs = DEFAULT_TIMEO
   return { html: new TextDecoder().decode(bytes), url: finalUrl };
 }
 
-/** Open Graph fields for a page. */
+/** Open Graph fields for a page. Follows HTML meta refresh the way a browser would. */
 export async function fetchPageMeta(url, options = {}) {
-  const { html, url: finalUrl } = await fetchHead(url, options);
+  let current = url;
+  const seen = new Set();
+  for (let hop = 0; hop <= MAX_REDIRECTS; hop += 1) {
+    if (seen.has(current)) break;
+    seen.add(current);
+    const { html, url: finalUrl } = await fetchHead(current, options);
+    const refresh = metaRefreshTarget(html, finalUrl);
+    if (refresh && refresh !== finalUrl && !seen.has(refresh)) {
+      current = refresh;
+      continue;
+    }
+    return { ...parseHead(html, finalUrl), url: finalUrl };
+  }
+  const { html, url: finalUrl } = await fetchHead(current, options);
   return { ...parseHead(html, finalUrl), url: finalUrl };
 }
 

package/ui/garden/markdown-links.mjs

@@ -50,32 +50,68 @@ function mapInline(text, visit) {
 
 // Inline links read as CommonMark reads them (ui/garden/markdown-destination.mjs):
 // `[t](<https://e.com/a (1)>)` and `[t](https://e.com/Trie_(x))` are whole addresses.
-// Groups: 1 `!`, 3 or 4 the address, 6 `{card}`.
+// Groups: 1 `!`, 2 label, 3 or 4 the address, 6 `{card}`.
 const LINK = new RegExp(`${INLINE_LINK_SOURCE}(\\{card\\}|\\{source\\})?`, "giu");
 const linkUrl = (match) => match[3] ?? match[4];
 const isWebUrl = (url) => /^https?:\/\//iu.test(url);
 const AUTOLINK = /<(https?:\/\/[^\s>]+)>/giu;
 const BARE = /(^|[\s(（：:，,。])(https?:\/\/[^\s<>()（）\]，。、]+[^\s<>()（）\].,;:!?，。、；：！？'"])/giu;
 
+/** The marker that asks for a link card: `[text](url){card}`, nothing in between. */
+export const CARD_MARKER = "{card}";
+/** The marker that draws a link as an inline source chip: `[site](url){source}`. */
+export const SOURCE_MARKER = "{source}";
+
+
+/**
+ * A single-line paragraph whose only content is a web URL: bare, <autolink>,
+ * or `[url](url)` (optional `{card}`). Those render as link cards and share
+ * the OG fetch path with `{card}` markers.
+ */
+function standaloneUrl(line) {
+  const trimmed = String(line ?? "").trim();
+  if (!trimmed || /\n/u.test(trimmed)) return "";
+  // Fresh regex: a shared /g lastIndex would skip later matchAll calls.
+  const link = new RegExp(LINK.source, LINK.flags).exec(trimmed);
+  if (link && link.index === 0 && link[0].length === trimmed.length && link[1] !== "!") {
+    const url = linkUrl(link);
+    const label = link[2] ?? "";
+    if (isWebUrl(url) && (link[6] === CARD_MARKER || label === url)) return url;
+    return "";
+  }
+  const auto = /^<(https?:\/\/[^\s>]+)>$/iu.exec(trimmed);
+  if (auto) return auto[1];
+  const bare = /^(https?:\/\/[^\s<>()（）]+)$/iu.exec(trimmed);
+  return bare && isWebUrl(bare[1]) ? bare[1] : "";
+}
+
 /** Every external link in the body, in order: `{ url, card }`. Images are left out. */
 export function markdownLinks(markdown) {
   const links = [];
+  const standalone = new Set();
   mapOutsideCode(markdown, (text) => {
+    for (const block of text.split(/\n{2,}/u)) {
+      const url = standaloneUrl(block);
+      if (url) standalone.add(url);
+    }
     for (const match of text.matchAll(LINK)) {
-      if (match[1] !== "!" && isWebUrl(linkUrl(match))) links.push({ url: linkUrl(match), card: match[6] === CARD_MARKER });
+      if (match[1] !== "!" && isWebUrl(linkUrl(match))) {
+        const url = linkUrl(match);
+        links.push({ url, card: match[6] === CARD_MARKER || standalone.has(url) });
+      }
     }
     const rest = text.replace(LINK, " ");
-    for (const match of rest.matchAll(AUTOLINK)) links.push({ url: match[1], card: false });
-    for (const match of rest.replace(AUTOLINK, " ").matchAll(BARE)) links.push({ url: match[2], card: false });
+    for (const match of rest.matchAll(AUTOLINK)) {
+      links.push({ url: match[1], card: standalone.has(match[1]) });
+    }
+    for (const match of rest.replace(AUTOLINK, " ").matchAll(BARE)) {
+      links.push({ url: match[2], card: standalone.has(match[2]) });
+    }
     return text;
   });
   return links;
 }
 
-/** The marker that asks for a link card: `[text](url){card}`, nothing in between. */
-export const CARD_MARKER = "{card}";
-/** The marker that draws a link as an inline source chip: `[site](url){source}`. */
-export const SOURCE_MARKER = "{source}";
 
 /**
  * The same Markdown without `{card}` markers, for excerpts, RSS and the search

package/ui/garden/site-nav.mjs

@@ -2,12 +2,14 @@
  * The header's site navigation. Hrefs carry the trailing slash: a static
  * export serves these pages at /blog/ and so on, and the host answers /blog
  * with a 301 first (new tab, no JavaScript, crawlers).
+ *
+ * /tags/ stays as a deep route (article tag chips still link there) but is
+ * not listed in the primary nav.
  */
 export const SITE_NAV = [
   { href: "/blog/", label: "文章" },
   { href: "/thoughts/", label: "随想" },
   { href: "/columns/", label: "专栏" },
-  { href: "/tags/", label: "标签" },
   { href: "/about/", label: "关于" },
 ];