Fix srcsets overriding rewritten image srcs

lib/frenzy/pipeline/extractor/birchtree.ex

@@ -0,0 +1,19 @@
+defmodule Frenzy.Pipeline.Extractor.Birchtree do
+  @moduledoc """
+  Extractor for https://birchtree.me
+  """
+
+  alias Frenzy.Pipeline.Extractor
+  @behaviour Extractor
+
+  @impl Extractor
+  def extract(html_tree) do
+    case Floki.find(html_tree, "section.post-content") do
+      [content_elem | _] ->
+        {:ok, content_elem}
+
+      _ ->
+        {:error, "no matching elements"}
+    end
+  end
+end

lib/frenzy/pipeline/scrape_stage.ex

@@ -158,6 +158,16 @@ defmodule Frenzy.Pipeline.ScrapeStage do
               attr
           end)
 
+        has_src = Enum.find(new_attrs, fn {name, _} -> name == "src" end)
+
+        # remove srcsets because our transformation only applies to the src attribute, so that should always be used
+        new_attrs =
+          if has_src do
+            Enum.reject(new_attrs, fn {name, _} -> name == "srcset" end)
+          else
+            new_attrs
+          end
+
         {"img", new_attrs}
 
       elem ->
@@ -191,5 +201,5 @@ defmodule Frenzy.Pipeline.ScrapeStage do
     end
   end
 
-  defp image_to_data_uri(src, _site_uri, false), do: src
+  defp image_to_data_uri(src, site_uri, false), do: to_string(URI.merge(site_uri, src))
 end

lib/frenzy_web/live/configure_stage/scrape_stage_live.ex

@@ -6,6 +6,7 @@ defmodule FrenzyWeb.ConfigureStage.ScrapeStageLive do
                 {"512 Pixels", Frenzy.Pipeline.Extractor.FiveTwelvePixels},
                 {"Ars Technica", Frenzy.Pipeline.Extractor.ArsTechnica},
                 {"beckyhansmeyer.com", Frenzy.Pipeline.Extractor.BeckyHansmeyer},
+                {"birchtree.me", Frenzy.Pipeline.Extractor.Birchtree},
                 {"daringfireball.net", Frenzy.Pipeline.Extractor.DaringFireball},
                 {"ericasadun.com", Frenzy.Pipeline.Extractor.EricaSadun},
                 {"finertech.com", Frenzy.Pipeline.Extractor.FinerTech},