@blamejs/blamejs-shop 0.3.22 → 0.3.23

package/CHANGELOG.md

@@ -8,6 +8,8 @@ upgrading across more than a few patches at a time.
 
 ## v0.3.x
 
+- v0.3.23 (2026-05-30) — **Fix page corruption from a dollar sign in content, keep search pages out of the index, and add a working unsubscribe link.** Three storefront fixes. First, a content-corruption bug: a dollar sign in any page's dynamic content — a product description, a blog post, a content page, a cart or admin page — could mangle the rendered HTML, and a dollar immediately followed by a backtick could splice the page's head markup into the body. Every page assembled its body in a way that gave dollar sequences special meaning; all of them now insert content verbatim. Second, internal search result pages are now marked noindex so the many query and filter URLs are not indexed as thin, duplicate pages (their product links are still followed). Third, the unsubscribe link that marketing emails point to now has a route behind it, so a recipient can actually unsubscribe (and resubscribe) instead of hitting a missing page. **Added:** *Newsletter unsubscribe link works* — A /unsubscribe route now backs the one-click unsubscribe link in marketing emails: opening the link shows a confirmation page, confirming removes the address, and a resubscribe option is offered. Invalid, expired, and already-used links each show a clear page rather than an error. The backend for this already existed; it simply had no route. **Fixed:** *A dollar sign in content no longer corrupts the page* — Pages were assembled by substituting the body into a template in a way that treated dollar sequences (such as a dollar followed by a backtick) as special, so content containing them could corrupt the output or pull head markup into the body. Every render path now inserts dynamic content literally, so any text — prices, code samples, anything with a dollar sign — renders exactly as written. · *Search result pages are no longer indexed* — Internal search result pages now carry a noindex robots tag so the large number of query and facet URL combinations aren't indexed as thin or duplicate content; the product links on those pages are still followed.
+
 - v0.3.22 (2026-05-30) — **Create and publish content pages (About, Shipping, Returns, and the like) from the admin console.** The shop now has editable content pages. An admin Pages screen lets an operator create a page with a slug, title, Markdown body, layout, and meta fields, then publish, unpublish, archive, or restore it — and the storefront serves each published page at /pages/<slug>. Previously the content-page backend existed but had no admin screen and the storefront did not serve it, so it could not be used; a page is created as a draft and is not visible on the storefront until it is published. **Added:** *Content pages: author in the console, serve on the storefront* — A new Pages admin screen manages content pages through their full lifecycle (draft, published, archived, with restore). Published pages are served at /pages/<slug> with the page's chosen layout; a draft or archived slug returns not-found. The Markdown body is rendered with the same escaping and safe-link handling as the rest of the storefront. This makes a standard set of pages — about, shipping, returns, and similar — editable instead of requiring a code change.
 
 - v0.3.21 (2026-05-30) — **Customers can cancel an order from their account before it ships.** A signed-in customer can now cancel one of their own orders from the order page, as long as it has not started fulfillment. The Cancel control appears only while the order is still cancellable — a paid or pending order — and is absent once the order is fulfilling, shipped, delivered, already cancelled, or refunded. The action is tied to the signed-in customer, so a customer can only cancel their own order. Note: cancelling a paid order changes its status to cancelled but does not by itself refund the charge — issue the refund from the admin console as usual; a customer-initiated cancellation of a paid order is your signal to do so. **Added:** *Cancel an unfulfilled order from the account* — The order page now shows a Cancel button while an order is still in a pre-fulfillment state (paid or pending). Cancelling moves the order to cancelled and shows a confirmation; the control does not appear once an order is fulfilling, shipped, delivered, cancelled, or refunded, and a customer can only cancel their own order. A paid-order cancellation updates the status but does not automatically refund the captured payment — refund it from the console; treat a customer cancellation of a paid order as the prompt to do so.

package/lib/admin.js

@@ -6924,6 +6924,21 @@ function _renderTemplate(template, vars) {
   return out;
 }
 
+// Splice a fully-rendered HTML fragment into `html` at the first
+// occurrence of a literal `RAW_*` token, inserting the fragment LITERALLY.
+//
+// `String.prototype.replace(token, replacementString)` gives the
+// replacement string special meaning to `$` sequences — `$$`, `$&`,
+// `` $` `` (the text before the match), `$'` (the text after the match),
+// `$N`. An admin page body that contains a dollar followed by a backtick
+// would otherwise splice the page chrome into the body, and any other
+// dollar sequence corrupts the output. Passing a REPLACER FUNCTION makes
+// `String.replace` insert the return value verbatim, with no dollar
+// interpretation. Mirrors the storefront's `_spliceRaw`.
+function _spliceRaw(html, token, fragment) {
+  return html.replace(token, function () { return fragment == null ? "" : String(fragment); });
+}
+
 function _sparkSvg(byDay, currency) {
   // SVG sparkline rendered server-side from revenue-by-day rows of
   // the dashboard's primary currency. Returns an empty placeholder
@@ -7106,8 +7121,10 @@ function _renderAdminShell(shopName, subtitle, bodyHtml, active, available) {
     nav:          "RAW_NAV",
     body:         "RAW_BODY",
   }).replace("RAW_ADMIN_CSS", _adminStylesheetLink())
-    .replace("RAW_NAV", _adminNav(active, available))
-    .replace("RAW_BODY", bodyHtml);
+    .replace("RAW_NAV", _adminNav(active, available));
+  // Splice the admin body literally so a `$`-bearing fragment can't trip
+  // `String.replace`'s dollar substitution. See `_spliceRaw`.
+  html = _spliceRaw(html, "RAW_BODY", bodyHtml);
   // Token every admin POST form with the per-request double-submit CSRF value
   // (seeded on the ALS by mount()'s sync middleware). Single funnel — every
   // authenticated admin page assembles here — so this is the one place the

package/lib/asset-manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "0.3.22",
+  "version": "0.3.23",
   "assets": {
     "css/admin.css": {
       "integrity": "sha384-1SIn6oAf1DjECbRfKENZasdKHJiywGdXR58wn0hsFGcdVHzUmvfgMkEz5ANIAZJ3",

package/lib/security-middleware.js

@@ -105,6 +105,13 @@ var EDGE_POST_PATHS = [
   "/consent",
   "/currency",
   "/newsletter",
+  // One-click unsubscribe (RFC 8058). The unsubscribe token in the URL IS
+  // the bearer; the POST a mail client fires from List-Unsubscribe-Post
+  // carries no cookies and no session CSRF token, so the path must be
+  // exempt from the double-submit check (it keeps its SameSite +
+  // fetch-metadata defense and is single-use + timing-safe at the token
+  // layer). Container-only — there's no edge copy to keep parity with.
+  "/unsubscribe",
   "/announcements/",
 ];
 

package/lib/storefront.js

@@ -106,6 +106,23 @@ function webhookRawBodyCapture(paths) {
 // same XSS guard, same unknown / unused refusal).
 var _render = emailModule._render;
 
+// Splice a fully-rendered HTML fragment into `html` at the first
+// occurrence of a literal `RAW_*` token, inserting the fragment LITERALLY.
+//
+// `String.prototype.replace(token, replacementString)` gives the
+// replacement string special meaning to `$` sequences — `$$`, `$&`,
+// `` $` `` (the text before the match), `$'` (the text after the match),
+// `$1`. A page/blog body that contains a dollar followed by a backtick
+// would otherwise splice the entire document HEAD into the body, and any
+// other dollar sequence corrupts the output. The fragment is already
+// escaped/rendered at its own build site, so this is purely about the
+// replace mechanics: passing a REPLACER FUNCTION makes `String.replace`
+// insert the return value verbatim, with no dollar interpretation. Use
+// this for every dynamic (operator- or customer-supplied) fragment swap.
+function _spliceRaw(html, token, fragment) {
+  return html.replace(token, function () { return fragment == null ? "" : String(fragment); });
+}
+
 // ---- double-submit CSRF token injection ---------------------------------
 
 // The container's authenticated state-changing forms (account /
@@ -758,15 +775,23 @@ function _wrap(opts) {
 
   // Per-page robots directive. Indexable pages (the default) emit no
   // robots meta — the absence means "index, follow". Pages the storefront
-  // keeps out of search (cart / account / any session-scoped surface)
-  // pass `robots: "noindex"`, which stamps a noindex,nofollow meta in the
-  // <head> — a belt-and-suspenders pairing with the robots.txt Disallow +
-  // the edge `x-robots-tag` header, so a directly-fetched page is also
+  // keeps out of search pass a directive:
+  //   - `noindex` → `noindex,nofollow`. Session-scoped surfaces (cart /
+  //     account) whose links are all session-scoped too, so there's
+  //     nothing for a crawler to follow.
+  //   - `noindex,follow` → `noindex,follow`. Internal search result pages
+  //     are thin/duplicate indexable URLs (one per query + facet combo) —
+  //     keep the query URL out of the index but let crawlers follow the
+  //     product links the page lists.
+  // It's a belt-and-suspenders pairing with the robots.txt Disallow + the
+  // edge `x-robots-tag` header, so a directly-fetched page is also
   // self-describing. A noindex page needs no canonical (a crawler won't
   // index it to dedupe), so these pages don't thread a canonical_url.
-  var robotsMeta = (opts.robots === "noindex")
-    ? "  <meta name=\"robots\" content=\"noindex,nofollow\">\n"
-    : "";
+  var robotsMeta = (opts.robots === "noindex,follow")
+    ? "  <meta name=\"robots\" content=\"noindex,follow\">\n"
+    : (opts.robots === "noindex")
+      ? "  <meta name=\"robots\" content=\"noindex,nofollow\">\n"
+      : "";
 
   var assembled = _render(LAYOUT, vars)
     .replace("RAW_CSS_INTEGRITY", themeCssIntegrity)
@@ -776,15 +801,18 @@ function _wrap(opts) {
     .replace("RAW_CART_COUNT_SCRIPT", _islandScript("cart-count.js", { id: "cart-count-island" }))
     .replace("RAW_ANNOUNCEMENT_SCRIPT", announcementScript)
     .replace("RAW_CURRENCY_SWITCHER", switcherHtml)
-    .replace("RAW_LOCALE_SWITCHER", localeCtx.switcher_html || "")
-    .replace("RAW_BODY_PLACEHOLDER", opts.body);
+    .replace("RAW_LOCALE_SWITCHER", localeCtx.switcher_html || "");
   // The body is RAW HTML (already rendered + escaped at the
   // per-fragment level). The placeholder swap is post-render so the
   // outer renderer's HTML-escape doesn't double-escape the inner
-  // markup. `search_q` is HTML-escaped by the renderer like any
+  // markup, and routes through `_spliceRaw` so a body carrying a `$`
+  // sequence (a blog/CMS post, a reflected query) is inserted literally
+  // rather than triggering `String.replace`'s dollar-substitution.
+  // `search_q` is HTML-escaped by the renderer like any
   // other placeholder, so a customer-supplied query like
   // `"><script>` lands as escaped text inside the input's `value`.
   //
+  assembled = _spliceRaw(assembled, "RAW_BODY_PLACEHOLDER", opts.body);
   // Final pass: token the container's authenticated POST forms with the
   // per-request double-submit CSRF value (stashed on the locale ALS by
   // `localeMiddleware`). EDGE_POST_PATHS forms are skipped to preserve
@@ -1557,6 +1585,10 @@ function renderSearch(opts) {
     og_description: searchMetaDescription,
     canonical_url: opts.canonical_url,
     og_url:        opts.og_url,
+    // Internal search results are thin/duplicate indexable URLs (one per
+    // query + facet combination) — keep the query URL out of the index
+    // but let crawlers follow the product links the page lists.
+    robots:     "noindex,follow",
     body:       body,
   }, _currencyWrapOpts(opts)));
 }
@@ -5226,6 +5258,110 @@ function renderNewsletterError(opts) {
   });
 }
 
+// ---- newsletter unsubscribe (CAN-SPAM / RFC 8058) ----------------------
+//
+// A marketing email's unsubscribe link points at GET /unsubscribe?token=…
+// (a friendly confirm page) and List-Unsubscribe-Post fires a one-click
+// POST /unsubscribe with the same token. Both are unauthenticated — the
+// opaque, single-use, timing-safe token is the bearer — and CSRF-exempt
+// (the mail-client POST carries no session). The token is never echoed
+// back to the page; only its structured outcome is rendered.
+
+// The confirm page (GET). A no-JS form POSTs the token back to
+// /unsubscribe to perform the unsubscribe, so a visitor who followed the
+// link sees a deliberate "yes, unsubscribe me" step rather than being
+// unsubscribed by a link-prefetcher. The token rides in a hidden field
+// (HTML-escaped) and is the only handle — no email address is shown.
+function renderUnsubscribeConfirm(opts) {
+  opts = opts || {};
+  var token = typeof opts.token === "string" ? opts.token : "";
+  var body =
+    "<section class=\"newsletter-thanks\">" +
+      "<div class=\"newsletter-thanks__card\">" +
+        "<p class=\"eyebrow\">Newsletter</p>" +
+        "<h1 class=\"newsletter-thanks__title\">Unsubscribe from the list?</h1>" +
+        "<p class=\"newsletter-thanks__lede\">Confirm below and we'll stop emailing this address. You can re-subscribe any time from the footer of the site.</p>" +
+        "<form class=\"newsletter-unsub__form\" method=\"post\" action=\"/unsubscribe\">" +
+          "<input type=\"hidden\" name=\"token\" value=\"" + _escAttr(token) + "\">" +
+          "<div class=\"newsletter-thanks__cta\">" +
+            "<button type=\"submit\" class=\"btn-primary\">Unsubscribe</button>" +
+            "<a href=\"/\" class=\"btn-ghost\">Keep me subscribed</a>" +
+          "</div>" +
+        "</form>" +
+      "</div>" +
+    "</section>";
+  return _wrap({
+    title:      "Unsubscribe",
+    shop_name:  opts.shop_name || "blamejs.shop",
+    cart_count: opts.cart_count,
+    theme_css:  opts.theme_css,
+    // The confirm + result pages are token-bearer surfaces with no
+    // canonical/indexable identity — keep them out of search (`noindex`
+    // maps to noindex,nofollow in _wrap).
+    robots:     "noindex",
+    body:       body,
+  });
+}
+
+// The POST outcome page. `opts.outcome` is one of the structured results
+// the unsubscribe flow produces: "unsubscribed" (success — offers a
+// re-subscribe form), "not-found" (invalid / unknown token), "already"
+// (the token was already consumed — the address is already off the list),
+// "expired" (the token's TTL lapsed). Every branch renders a clean,
+// server-rendered page; no raw error is ever surfaced.
+function renderUnsubscribeResult(opts) {
+  opts = opts || {};
+  var outcome = opts.outcome;
+  var heading, lede, extra = "";
+  if (outcome === "unsubscribed") {
+    heading = "You're unsubscribed.";
+    lede    = "This address won't receive any more newsletter email. Changed your mind? Re-subscribe below.";
+    // Re-subscribe affordance — the storefront newsletter signup form,
+    // pre-pointed at the footer band's POST /newsletter so a single click
+    // (after typing the address) puts the visitor back on the list.
+    extra =
+      "<form class=\"newsletter-resub__form\" method=\"post\" action=\"/newsletter\">" +
+        "<label class=\"skip-link\" for=\"resub-email\">Email address</label>" +
+        "<input id=\"resub-email\" type=\"email\" name=\"email\" required placeholder=\"you@example.com\" autocomplete=\"email\">" +
+        "<button type=\"submit\" class=\"btn-primary\">Re-subscribe</button>" +
+      "</form>";
+  } else if (outcome === "already") {
+    heading = "Already unsubscribed.";
+    lede    = "This link was already used — the address is off the list, so there's nothing more to do.";
+  } else if (outcome === "expired") {
+    heading = "That link has expired.";
+    lede    = "Unsubscribe links are valid for a limited time. You're still on the list; use the unsubscribe link in a more recent email, or contact us and we'll remove you.";
+  } else {
+    // "not-found" + any unexpected code → the same generic, non-leaking
+    // copy. A bad / unknown / malformed token can't be distinguished from
+    // outside, by design.
+    heading = "This link isn't valid.";
+    lede    = "We couldn't match this unsubscribe link to a subscription. It may have already been used, or the link may be incomplete.";
+  }
+  var body =
+    "<section class=\"newsletter-thanks\">" +
+      "<div class=\"newsletter-thanks__card\">" +
+        "<p class=\"eyebrow\">Newsletter</p>" +
+        "<h1 class=\"newsletter-thanks__title\">" + heading + "</h1>" +
+        "<p class=\"newsletter-thanks__lede\">" + lede + "</p>" +
+        extra +
+        "<div class=\"newsletter-thanks__cta\">" +
+          "<a href=\"/\" class=\"btn-primary\">Back to the shop <span aria-hidden=\"true\">→</span></a>" +
+        "</div>" +
+      "</div>" +
+    "</section>";
+  return _wrap({
+    title:      "Unsubscribe",
+    shop_name:  opts.shop_name || "blamejs.shop",
+    cart_count: opts.cart_count,
+    theme_css:  opts.theme_css,
+    // Token-bearer outcome page — keep it out of search (`noindex` maps to
+    // noindex,nofollow in _wrap).
+    robots:     "noindex",
+    body:       body,
+  });
+}
+
 // ---- cookie preference center ------------------------------------------
 
 // The four toggleable categories + their operator-facing copy, mirroring
@@ -11263,6 +11399,71 @@ function mount(router, deps) {
         }));
       }
     });
+
+    // Newsletter unsubscribe (CAN-SPAM one-click / RFC 8058).
+    //
+    // GET /unsubscribe?token=… renders a friendly confirm page; the
+    // visitor clicks "Unsubscribe", which POSTs the token back. POST
+    // /unsubscribe is also the List-Unsubscribe-Post one-click target a
+    // mail client fires directly. The opaque, single-use, timing-safe
+    // token is the bearer — no session, no auth — so the route is
+    // CSRF-exempt (see EDGE_POST_PATHS) and reads each structured outcome
+    // `consumeUnsubscribeToken` returns onto a clean server-rendered page.
+    // The token is never echoed back beyond the confirm form's hidden
+    // field; the outcome page reveals nothing about whether a token
+    // existed, was already used, or expired beyond its own friendly copy.
+
+    // Map a `consumeUnsubscribeToken` result code onto the outcome the
+    // result page renders. The structured codes are stable; an unknown
+    // code degrades to the generic "not valid" page rather than a 500.
+    function _unsubscribeOutcome(result) {
+      if (result && result.ok === true) return "unsubscribed";
+      var code = result && result.error;
+      if (code === "already-consumed") return "already";
+      if (code === "expired")          return "expired";
+      return "not-found";   // "not-found" + anything unexpected
+    }
+
+    router.get("/unsubscribe", async function (req, res) {
+      var token = (req.query && typeof req.query.token === "string") ? req.query.token : "";
+      var cartCount = 0;
+      try { cartCount = await _cartCountForReq(req); } catch (_e) { /* drop-silent — empty cart fallback */ }
+      return _send(res, 200, renderUnsubscribeConfirm({
+        shop_name:  shopName,
+        cart_count: cartCount,
+        token:      token,
+      }));
+    });
+
+    router.post("/unsubscribe", async function (req, res) {
+      var body = req.body || {};
+      var token = typeof body.token === "string" ? body.token : "";
+      var cartCount = 0;
+      try { cartCount = await _cartCountForReq(req); } catch (_e) { /* drop-silent — empty cart fallback */ }
+      var outcome;
+      try {
+        // `consumeUnsubscribeToken` returns a structured result (it does
+        // not throw on a bad/missing token — it returns `{ ok:false,
+        // error:"not-found" }`). An empty token is handled the same way.
+        var result = await deps.newsletter.consumeUnsubscribeToken(token);
+        outcome = _unsubscribeOutcome(result);
+      } catch (e) {
+        // A real infrastructure fault (D1 unreachable) — record it server-
+        // side and render the generic non-leaking page rather than a 500
+        // that exposes internals on a public, unauthenticated route.
+        _log.error("storefront unsubscribe failed", {
+          route:      "/unsubscribe",
+          request_id: (req && req.requestId) || null,
+          err:        (e && e.message) || String(e),
+        });
+        outcome = "not-found";
+      }
+      return _send(res, 200, renderUnsubscribeResult({
+        shop_name:  shopName,
+        cart_count: cartCount,
+        outcome:    outcome,
+      }));
+    });
   }
 
   // ---- cookie consent -----------------------------------------------------

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blamejs/blamejs-shop",
-  "version": "0.3.22",
+  "version": "0.3.23",
   "description": "Open-source framework built on blamejs. Vendored stack, zero npm runtime deps, PQC-first crypto, security-on by default.",
   "main": "lib/index.js",
   "scripts": {