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

blogex/test/blogex/blog_integration_test.exs

@@ -93,4 +93,38 @@ defmodule Blogex.BlogIntegrationTest do
       File.rm!(tmp_file)
     end
   end
+
+  describe "validation sees compiled post bodies" do
+    # `_validate_links/2` runs on `post.body` *after* NimblePublisher rendered
+    # the markdown, so validation only works if extraction reads the compiled
+    # HTML. This pins that contract against a real compiled body: if
+    # extraction ever goes back to scanning markdown syntax only, it fails.
+    test "extract_links finds internal links in a compiled post body" do
+      tmp_file = Path.join(System.tmp_dir!(), "test_blog_compiled_body.ex")
+
+      File.write!(tmp_file, """
+      defmodule TestBlogCompiledBody do
+        use Blogex.Blog,
+          blog_id: :test,
+          app: :blogex,
+          from: "priv/blog/test/2026/01-03-internal-link-post.md",
+          title: "Test Blog",
+          base_path: "/blog/test"
+      end
+      """)
+
+      [{TestBlogCompiledBody, _bytecode}] = Code.compile_file(tmp_file, __ENV__.file)
+
+      # `apply/3`: the module is compiled at runtime, so a static call would
+      # warn ("module is not available") on every test compile.
+      [post] = apply(TestBlogCompiledBody, :all_posts, [])
+
+      # The markdown source is `[hello world](/blog/engineering/hello-world)`.
+      assert post.body =~ "/blog/engineering/hello-world"
+
+      assert "/blog/engineering/hello-world" in Blogex.LinkValidator.extract_links(post.body)
+
+      File.rm!(tmp_file)
+    end
+  end
 end

blogex/test/blogex/link_validator_test.exs

@@ -29,6 +29,21 @@ defmodule Blogex.LinkValidatorTest do
       assert LinkValidator.extract_links(body) == ["/blog/engineering/tag/pi.dev"]
     end
 
+    # HTML anchors are not "the href right after <a (" - they are whatever the
+    # renderer emits. Pinned with the shapes a hand-rolled scan misses.
+    test "extracts internal blog links whatever the anchor's attributes look like" do
+      body =
+        ~s(<p><a class="post-link" href="/blog/engineering/first">first</a>) <>
+          ~s( <a href=/blog/releases/second>second</a>) <>
+          ~s( <a data-x="1" aria-label="third" href='/blog/engineering/third'>third</a></p>)
+
+      assert LinkValidator.extract_links(body) == [
+               "/blog/engineering/first",
+               "/blog/releases/second",
+               "/blog/engineering/third"
+             ]
+    end
+
     test "ignores external links" do
       body = "See [GitHub](https://github.com) and [internal](/blog/engineering/post)."
 
@@ -157,6 +172,12 @@ defmodule Blogex.LinkValidatorTest do
       assert LinkValidator.validate_link("/blog/engineering/tag/pi.dev#top") == :ok
     end
 
+    # Tags are user-defined: the documented rule restricts *slugs*, not tags.
+    test "accepts tag links with dots and capitals" do
+      assert LinkValidator.validate_link("/blog/engineering/tag/pi.dev") == :ok
+      assert LinkValidator.validate_link("/blog/engineering/tag/Pi.dev") == :ok
+    end
+
     test "rejects tag page link with empty tag" do
       assert LinkValidator.validate_link("/blog/engineering/tag/") ==
                {:error, "empty tag in tag link"}
@@ -267,5 +288,40 @@ defmodule Blogex.LinkValidatorTest do
                ]
              }
     end
+
+    # `_validate_links/2` hands this module `post.body`, which NimblePublisher
+    # already rendered to HTML, so failure reporting has to work on HTML bodies.
+    # These are the two reports the compile-time integration tests depend on.
+    test "reports an unknown blog ID in an HTML body" do
+      body = ~s(<p><a href="/blog/unknown/broken">bad</a></p>)
+
+      assert LinkValidator.validate_body(body, :engineering, post_id: "html-post") == {
+               :error,
+               [
+                 {
+                   1,
+                   "/blog/unknown/broken",
+                   "unknown blog ID: unknown",
+                   post_id: "html-post"
+                 }
+               ]
+             }
+    end
+
+    test "reports an invalid slug in an HTML body" do
+      body = ~s(<p><a href="/blog/engineering/My-Post">bad</a></p>)
+
+      assert LinkValidator.validate_body(body, :engineering, post_id: "html-post") == {
+               :error,
+               [
+                 {
+                   1,
+                   "/blog/engineering/My-Post",
+                   "slug must be lowercase alphanumeric with hyphens: My-Post",
+                   post_id: "html-post"
+                 }
+               ]
+             }
+    end
   end
 end