yak(link-validator-never-sees-compiled-html-internal-link-validation-is-vacuous-ddd1): Link validator never sees compiled HTML — internal-link validation is vacuous

app/test/firehose_web/controllers/blog_test.exs

@@ -88,6 +88,49 @@ defmodule FirehoseWeb.BlogTest do
     end
   end
 
+  describe "internal links in compiled posts" do
+    # `Blogex.Blog._validate_links/2` validates `post.body` after
+    # NimblePublisher rendered it to HTML. If extraction cannot read that HTML,
+    # every post looks link-free and compile-time validation is vacuous - which
+    # is the failure the real posts below pin against.
+    defp posts_with_internal_links do
+      Enum.filter(Firehose.EngineeringBlog.all_posts(), fn post ->
+        Blogex.LinkValidator.extract_links(post.body) != []
+      end)
+    end
+
+    test "the validator sees internal links in real compiled posts" do
+      refute posts_with_internal_links() == []
+    end
+
+    test "the post that links to swim-the-test exposes that link" do
+      post =
+        Enum.find(posts_with_internal_links(), &(&1.id == "delayed-gratification-in-ai-summer"))
+
+      assert post, "expected the compiled post to be visible with internal links extracted"
+
+      assert "/blog/engineering/swim-the-test" in Blogex.LinkValidator.extract_links(post.body)
+    end
+
+    test "no internal link in a real compiled post is invisible to extraction" do
+      # Deliberately permissive scan (any <a> whose href appears in any
+      # position, quoted or not): every /blog/ link it finds must also come out
+      # of extract_links/1, or a bad one could never be reported.
+      links_in_body = fn body ->
+        ~r/<a\s[^>]*?href=["']?(\/blog\/[^"'\s>]+)/
+        |> Regex.scan(body)
+        |> Enum.map(fn [_, link] -> link end)
+        |> Enum.uniq()
+      end
+
+      for post <- Firehose.EngineeringBlog.all_posts(),
+          link <- links_in_body.(post.body) do
+        assert link in Blogex.LinkValidator.extract_links(post.body),
+               "#{post.id}: #{link} is in the compiled body but not extracted"
+      end
+    end
+  end
+
   describe "input validation" do
     test "GET /blog/nonexistent returns 404", %{conn: conn} do
       assert conn |> get("/blog/nonexistent") |> html_response(404)