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/lib/blogex/link_validator.ex

@@ -1,6 +1,12 @@
 defmodule Blogex.LinkValidator do
   @moduledoc """
-  Validates internal blog link semantics in markdown post bodies.
+  Validates internal blog link semantics in compiled post bodies.
+
+  Post bodies reach this module as HTML: `Blogex.Blog` validates them in
+  `@before_compile`, after NimblePublisher has rendered the markdown. Markdown
+  link syntax is still understood so bodies and snippets can be validated
+  before rendering, but the HTML `<a href="…">` anchors are the ones that
+  matter in the real pipeline.
 
   Checks that internal links follow the pattern `/blog/{blog_id}/{slug}`
   where `blog_id` is a valid blog identifier and `slug` matches the
@@ -50,21 +56,25 @@ defmodule Blogex.LinkValidator do
   @slug_regex ~r/^[a-z0-9]+(-[a-z0-9]+)*$/
 
   @doc """
-  Extracts internal blog links from a markdown body.
+  Extracts internal blog links from a post body.
 
-  Returns a list of link paths (strings) that match the pattern
-  `/blog/{engineering|releases}/{slug}`. External links and non-blog
-  internal links are ignored.
+  Returns a list of link paths (strings) that look like `/blog/…` links.
+  External links and non-blog internal links are ignored. Blog IDs and slugs
+  are *not* judged here — `validate_link/1` classifies those.
 
-  Handles both markdown link syntax `[text](url)` and HTML `<a href="url">`.
-  (Post bodies are HTML, rendered by NimblePublisher at compile time.)
+  HTML anchors are read with Floki, so attribute order, quoting and extra
+  attributes do not matter. Markdown link syntax `[text](url)` is extracted
+  too (raw markdown bodies and snippets), and the two scans are merged.
 
   ## Examples
 
       iex> extract_links("[link](/blog/engineering/post)")
       ["/blog/engineering/post"]
 
-      iex> extract_links("<p><a href=\"/blog/engineering/post\">link</a></p>")
+      iex> extract_links("<p><a href=\\"/blog/engineering/post\\">link</a></p>")
+      ["/blog/engineering/post"]
+
+      iex> extract_links("<p><a class=\\"x\\" href='/blog/engineering/post'>link</a></p>")
       ["/blog/engineering/post"]
 
       iex> extract_links("See [GitHub](https://github.com)")
@@ -72,21 +82,24 @@ defmodule Blogex.LinkValidator do
   """
   @spec extract_links(String.t()) :: [String.t()]
   def extract_links(body) when is_binary(body) do
-    markdown_links =
-      ~r/\[([^\]]+)\]\(([^)]+)\)/
-      |> Regex.scan(body)
-      |> Enum.map(fn [_, _, path] -> path end)
-
-    html_links =
-      ~r/<a\s+href=["']([^"']*)["']/i
-      |> Regex.scan(body)
-      |> Enum.map(fn [_, path] -> path end)
-
-    (markdown_links ++ html_links)
+    (markdown_links(body) ++ html_links(body))
     |> Enum.uniq()
     |> Enum.filter(&internal_blog_link?/1)
   end
 
+  defp markdown_links(body) do
+    ~r/\[([^\]]+)\]\(([^)]+)\)/
+    |> Regex.scan(body)
+    |> Enum.map(fn [_, _, path] -> path end)
+  end
+
+  defp html_links(body) do
+    body
+    |> Floki.parse_document!()
+    |> Floki.find("a[href]")
+    |> Enum.flat_map(&Floki.attribute(&1, "href"))
+  end
+
   defp internal_blog_link?(path) do
     case parse_blog_link(path) do
       {_, _} -> true