Fix Rich Media Previews for updated activities

The Rich Media Previews were not regenerated when a post was updated due to a cache invalidation issue. They are now cached by the activity id so they can be evicted with the other activity cache objects in the :scrubber_cache.

changelog.d/rich_media.fix

@@ -0,0 +1 @@
+Rich Media Preview cache eviction when the activity is updated.

lib/pleroma/activity/html.ex

@@ -28,7 +28,7 @@ defp get_cache_keys_for(activity_id) do
     end
   end
 
-  defp add_cache_key_for(activity_id, additional_key) do
+  def add_cache_key_for(activity_id, additional_key) do
     current = get_cache_keys_for(activity_id)
 
     unless additional_key in current do

lib/pleroma/html.ex

@@ -6,8 +6,6 @@ defmodule Pleroma.HTML do
   # Scrubbers are compiled on boot so they can be configured in OTP releases
   #  @on_load :compile_scrubbers
 
-  @cachex Pleroma.Config.get([:cachex, :provider], Cachex)
-
   def compile_scrubbers do
     dir = Path.join(:code.priv_dir(:pleroma), "scrubbers")
 
@@ -67,27 +65,20 @@ def ensure_scrubbed_html(
     end
   end
 
-  def extract_first_external_url_from_object(%{data: %{"content" => content}} = object)
+  @spec extract_first_external_url_from_object(Pleroma.Object.t()) ::
+          {:ok, String.t()} | {:error, :no_content}
+  def extract_first_external_url_from_object(%{data: %{"content" => content}})
       when is_binary(content) do
-    unless object.data["fake"] do
-      key = "URL|#{object.id}"
-
-      @cachex.fetch!(:scrubber_cache, key, fn _key ->
-        {:commit, {:ok, extract_first_external_url(content)}}
-      end)
-    else
-      {:ok, extract_first_external_url(content)}
-    end
-  end
-
-  def extract_first_external_url_from_object(_), do: {:error, :no_content}
-
-  def extract_first_external_url(content) do
+    url =
       content
       |> Floki.parse_fragment!()
       |> Floki.find("a:not(.mention,.hashtag,.attachment,[rel~=\"tag\"])")
       |> Enum.take(1)
       |> Floki.attribute("href")
       |> Enum.at(0)
+
+    {:ok, url}
   end
+
+  def extract_first_external_url_from_object(_), do: {:error, :no_content}
 end

lib/pleroma/web/rich_media/helpers.ex

@@ -8,6 +8,8 @@ defmodule Pleroma.Web.RichMedia.Helpers do
   alias Pleroma.Object
   alias Pleroma.Web.RichMedia.Parser
 
+  @cachex Pleroma.Config.get([:cachex, :provider], Cachex)
+
   @config_impl Application.compile_env(:pleroma, [__MODULE__, :config_impl], Pleroma.Config)
 
   @options [
@@ -71,7 +73,24 @@ def fetch_data_for_object(object) do
   def fetch_data_for_activity(%Activity{data: %{"type" => "Create"}} = activity) do
     with true <- @config_impl.get([:rich_media, :enabled]),
          %Object{} = object <- Object.normalize(activity, fetch: false) do
+      if object.data["fake"] do
         fetch_data_for_object(object)
+      else
+        key = "URL|#{activity.id}"
+
+        @cachex.fetch!(:scrubber_cache, key, fn _ ->
+          result = fetch_data_for_object(object)
+
+          cond do
+            match?(%{page_url: _, rich_media: _}, result) ->
+              Activity.HTML.add_cache_key_for(activity.id, key)
+              {:commit, result}
+
+            true ->
+              {:ignore, %{}}
+          end
+        end)
+      end
     else
       _ -> %{}
     end

test/fixtures/rich_media/google.html

@@ -0,0 +1,12 @@
+<meta property="og:url" content="https://google.com">
+<meta property="og:type" content="website">
+<meta property="og:title" content="Google">
+<meta property="og:description" content="Search the world's information, including webpages, images, videos and more. Google has many special features to help you find exactly what you're looking for.">
+<meta property="og:image" content="">
+
+<meta name="twitter:card" content="summary_large_image">
+<meta property="twitter:domain" content="google.com">
+<meta property="twitter:url" content="https://google.com">
+<meta name="twitter:title" content="Google">
+<meta name="twitter:description" content="Search the world's information, including webpages, images, videos and more. Google has many special features to help you find exactly what you're looking for.">
+<meta name="twitter:image" content="">

test/fixtures/rich_media/yahoo.html

@@ -0,0 +1,12 @@
+<meta property="og:url" content="https://yahoo.com">
+<meta property="og:type" content="website">
+<meta property="og:title" content="Yahoo | Mail, Weather, Search, Politics, News, Finance, Sports & Videos">
+<meta property="og:description" content="Latest news coverage, email, free stock quotes, live scores and video are just the beginning. Discover more every day at Yahoo!">
+<meta property="og:image" content="https://s.yimg.com/cv/apiv2/social/images/yahoo_default_logo.png">
+
+<meta name="twitter:card" content="summary_large_image">
+<meta property="twitter:domain" content="yahoo.com">
+<meta property="twitter:url" content="https://yahoo.com">
+<meta name="twitter:title" content="Yahoo | Mail, Weather, Search, Politics, News, Finance, Sports & Videos">
+<meta name="twitter:description" content="Latest news coverage, email, free stock quotes, live scores and video are just the beginning. Discover more every day at Yahoo!">
+<meta name="twitter:image" content="https://s.yimg.com/cv/apiv2/social/images/yahoo_default_logo.png">

test/pleroma/web/rich_media/helpers_test.exs

@@ -83,6 +83,34 @@ test "crawls valid, complete URLs" do
              Pleroma.Web.RichMedia.Helpers.fetch_data_for_activity(activity)
   end
 
+  test "recrawls URLs on updates" do
+    original_url = "https://google.com/"
+    updated_url = "https://yahoo.com/"
+
+    Pleroma.StaticStubbedConfigMock
+    |> stub(:get, fn
+      [:rich_media, :enabled] -> true
+      path -> Pleroma.Test.StaticConfig.get(path)
+    end)
+
+    user = insert(:user)
+    {:ok, activity} = CommonAPI.post(user, %{status: "I like this site #{original_url}"})
+
+    assert match?(
+             %{page_url: ^original_url, rich_media: _},
+             Pleroma.Web.RichMedia.Helpers.fetch_data_for_activity(activity)
+           )
+
+    {:ok, _} = CommonAPI.update(user, activity, %{status: "I like this site #{updated_url}"})
+
+    activity = Pleroma.Activity.get_by_id(activity.id)
+
+    assert match?(
+             %{page_url: ^updated_url, rich_media: _},
+             Pleroma.Web.RichMedia.Helpers.fetch_data_for_activity(activity)
+           )
+  end
+
   # This does not seem to work. The urls are being fetched.
   @tag skip: true
   test "refuses to crawl URLs of private network from posts" do

test/support/http_request_mock.ex

@@ -1464,6 +1464,14 @@ def get("https://misskey.io/notes/8vs6wxufd0", _, _, _) do
      }}
   end
 
+  def get("https://google.com/", _, _, _) do
+    {:ok, %Tesla.Env{status: 200, body: File.read!("test/fixtures/rich_media/google.html")}}
+  end
+
+  def get("https://yahoo.com/", _, _, _) do
+    {:ok, %Tesla.Env{status: 200, body: File.read!("test/fixtures/rich_media/yahoo.html")}}
+  end
+
   def get(url, query, body, headers) do
     {:error,
      "Mock response not implemented for GET #{inspect(url)}, #{query}, #{inspect(body)}, #{inspect(headers)}"}
@@ -1539,7 +1547,9 @@ def post(url, query, body, headers) do
   @rich_media_mocks [
     "https://example.com/ogp",
     "https://example.com/ogp-missing-data",
-    "https://example.com/twitter-card"
+    "https://example.com/twitter-card",
+    "https://google.com/",
+    "https://yahoo.com/"
   ]
   def head(url, _query, _body, _headers) when url in @rich_media_mocks do
     {:ok, %Tesla.Env{status: 404, body: ""}}