Merge branch 'feat/rich-media-improvements' into 'develop'

Rich media improvements

See merge request pleroma/pleroma!2944

CHANGELOG.md

@@ -5,8 +5,13 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
 
 ## unreleased-patch - ???
 
+### Added
+- Rich media failure tracking (along with `:failure_backoff` option)
+
 ### Fixed
 - Mastodon API: Search parameter `following` now correctly returns the followings rather than the followers
+- Mastodon API: Timelines hanging for (`number of posts with links * rich media timeout`) in the worst case.
+  Reduced to just rich media timeout.
 
 ## [2.1.0] - 2020-08-28
 

config/config.exs

@@ -412,6 +412,7 @@
     Pleroma.Web.RichMedia.Parsers.TwitterCard,
     Pleroma.Web.RichMedia.Parsers.OEmbed
   ],
+  failure_backoff: 60_000,
   ttl_setters: [Pleroma.Web.RichMedia.Parser.TTL.AwsSignedUrl]
 
 config :pleroma, :media_proxy,

config/description.exs

@@ -2385,6 +2385,13 @@
         suggestions: [
           Pleroma.Web.RichMedia.Parser.TTL.AwsSignedUrl
         ]
+      },
+      %{
+        key: :failure_backoff,
+        type: :integer,
+        description:
+          "Amount of milliseconds after request failure, during which the request will not be retried.",
+        suggestions: [60_000]
       }
     ]
   },

docs/configuration/cheatsheet.md

@@ -361,6 +361,7 @@ config :pleroma, Pleroma.Web.MediaProxy.Invalidation.Http,
 * `ignore_hosts`: list of hosts which will be ignored by the metadata parser. For example `["accounts.google.com", "xss.website"]`, defaults to `[]`.
 * `ignore_tld`: list TLDs (top-level domains) which will ignore for parse metadata. default is ["local", "localdomain", "lan"].
 * `parsers`: list of Rich Media parsers.
+* `failure_backoff`: Amount of milliseconds after request failure, during which the request will not be retried.
 
 ## HTTP server
 

lib/pleroma/html.ex

@@ -109,8 +109,9 @@ def extract_first_external_url(object, content) do
       result =
         content
         |> Floki.parse_fragment!()
-        |> Floki.filter_out("a.mention,a.hashtag,a.attachment,a[rel~=\"tag\"]")
+        |> Floki.find("a:not(.mention,.hashtag,.attachment,[rel~=\"tag\"])")
-        |> Floki.attribute("a", "href")
+        |> Enum.take(1)
+        |> Floki.attribute("href")
         |> Enum.at(0)
 
       {:commit, {:ok, result}}

lib/pleroma/web/mastodon_api/views/status_view.ex

@@ -23,6 +23,17 @@ defmodule Pleroma.Web.MastodonAPI.StatusView do
 
   import Pleroma.Web.ActivityPub.Visibility, only: [get_visibility: 1, visible_for_user?: 2]
 
+  # This is a naive way to do this, just spawning a process per activity
+  # to fetch the preview. However it should be fine considering
+  # pagination is restricted to 40 activities at a time
+  defp fetch_rich_media_for_activities(activities) do
+    Enum.each(activities, fn activity ->
+      spawn(fn ->
+        Pleroma.Web.RichMedia.Helpers.fetch_data_for_activity(activity)
+      end)
+    end)
+  end
+
   # TODO: Add cached version.
   defp get_replied_to_activities([]), do: %{}
 
@@ -80,6 +91,11 @@ def render("index.json", opts) do
 
     # To do: check AdminAPIControllerTest on the reasons behind nil activities in the list
     activities = Enum.filter(opts.activities, & &1)
+
+    # Start fetching rich media before doing anything else, so that later calls to get the cards
+    # only block for timeout in the worst case, as opposed to
+    # length(activities_with_links) * timeout
+    fetch_rich_media_for_activities(activities)
     replied_to_activities = get_replied_to_activities(activities)
 
     parent_activities =

lib/pleroma/web/rich_media/parser.ex

@@ -17,14 +17,25 @@ def parse(url), do: parse_url(url)
   else
     @spec parse(String.t()) :: {:ok, map()} | {:error, any()}
     def parse(url) do
-      Cachex.fetch!(:rich_media_cache, url, fn _ ->
+      with {:ok, data} <- get_cached_or_parse(url),
-        with {:ok, data} <- parse_url(url) do
+           {:ok, _} <- set_ttl_based_on_image(data, url) do
-          {:commit, {:ok, data}}
+        {:ok, data}
-        else
+      else
-          error -> {:ignore, error}
+        error ->
-        end
+          Logger.error(fn -> "Rich media error: #{inspect(error)}" end)
-      end)
+      end
-      |> set_ttl_based_on_image(url)
+    end
+
+    defp get_cached_or_parse(url) do
+      case Cachex.fetch!(:rich_media_cache, url, fn _ -> {:commit, parse_url(url)} end) do
+        {:ok, _data} = res ->
+          res
+
+        {:error, _} = e ->
+          ttl = Pleroma.Config.get([:rich_media, :failure_backoff], 60_000)
+          Cachex.expire(:rich_media_cache, url, ttl)
+          e
+      end
     end
   end
 
@@ -50,22 +61,21 @@ def ttl(data, url) do
       config :pleroma, :rich_media,
         ttl_setters: [MyModule]
   """
-  @spec set_ttl_based_on_image({:ok, map()} | {:error, any()}, String.t()) ::
+  @spec set_ttl_based_on_image(map(), String.t()) ::
-          {:ok, map()} | {:error, any()}
+          {:ok, Integer.t() | :noop} | {:error, :no_key}
-  def set_ttl_based_on_image({:ok, data}, url) do
+  def set_ttl_based_on_image(data, url) do
-    with {:ok, nil} <- Cachex.ttl(:rich_media_cache, url),
+    case get_ttl_from_image(data, url) do
-         {:ok, ttl} when is_number(ttl) <- get_ttl_from_image(data, url) do
+      {:ok, ttl} when is_number(ttl) ->
-      Cachex.expire_at(:rich_media_cache, url, ttl * 1000)
+        ttl = ttl * 1000
-      {:ok, data}
-    else
-      _ ->
-        {:ok, data}
-    end
-  end
 
-  def set_ttl_based_on_image({:error, _} = error, _) do
+        case Cachex.expire_at(:rich_media_cache, url, ttl) do
-    Logger.error("parsing error: #{inspect(error)}")
+          {:ok, true} -> {:ok, ttl}
-    error
+          {:ok, false} -> {:error, :no_key}
+        end
+
+      _ ->
+        {:ok, :noop}
+    end
   end
 
   defp get_ttl_from_image(data, url) do

test/web/rich_media/aws_signed_url_test.exs

@@ -55,7 +55,7 @@ test "s3 signed url is parsed and correct ttl is set for rich media" do
 
     Cachex.put(:rich_media_cache, url, metadata)
 
-    Pleroma.Web.RichMedia.Parser.set_ttl_based_on_image({:ok, metadata}, url)
+    Pleroma.Web.RichMedia.Parser.set_ttl_based_on_image(metadata, url)
 
     {:ok, cache_ttl} = Cachex.ttl(:rich_media_cache, url)