X-Git-Url: http://git.squeep.com/?a=blobdiff_plain;ds=inline;f=lib%2Fpleroma%2Fweb%2Frich_media%2Fparser.ex;h=1d4cad0100e2ebcb8092b96555f5b5dfa938fa70;hb=a079ec3a3cdfd42d2cbd51c7698c2c87828e5778;hp=0779065ee1bc9c00e2ddb54f7b3e8c28d9920551;hpb=764a50f8a671cca69ca1f616754660506f8c18d8;p=akkoma
diff --git a/lib/pleroma/web/rich_media/parser.ex b/lib/pleroma/web/rich_media/parser.ex
index 0779065ee..1d4cad010 100644
--- a/lib/pleroma/web/rich_media/parser.ex
+++ b/lib/pleroma/web/rich_media/parser.ex
@@ -1,14 +1,11 @@
# Pleroma: A lightweight social networking server
-# Copyright © 2017-2020 Pleroma Authors
+# Copyright © 2017-2021 Pleroma Authors
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Web.RichMedia.Parser do
- @hackney_options [
- pool: :media,
- recv_timeout: 2_000,
- max_body: 2_000_000,
- with_body: true
- ]
+ require Logger
+
+ @cachex Pleroma.Config.get([:cachex, :provider], Cachex)
defp parsers do
Pleroma.Config.get([:rich_media, :parsers])
@@ -17,19 +14,69 @@ defmodule Pleroma.Web.RichMedia.Parser do
def parse(nil), do: {:error, "No URL provided"}
if Pleroma.Config.get(:env) == :test do
- def parse(url), do: parse_url(url)
+ @spec parse(String.t()) :: {:ok, map()} | {:error, any()}
+ def parse(url), do: parse_with_timeout(url)
else
+ @spec parse(String.t()) :: {:ok, map()} | {:error, any()}
def parse(url) do
- try do
- Cachex.fetch!(:rich_media_cache, url, fn _ ->
- {:commit, parse_url(url)}
- end)
- |> set_ttl_based_on_image(url)
- rescue
- e ->
- {:error, "Cachex error: #{inspect(e)}"}
+ with {:ok, data} <- get_cached_or_parse(url),
+ {:ok, _} <- set_ttl_based_on_image(data, url) do
+ {:ok, data}
end
end
+
+ defp get_cached_or_parse(url) do
+ case @cachex.fetch(:rich_media_cache, url, fn ->
+ case parse_with_timeout(url) do
+ {:ok, _} = res ->
+ {:commit, res}
+
+ {:error, reason} = e ->
+ # Unfortunately we have to log errors here, instead of doing that
+ # along with ttl setting at the bottom. Otherwise we can get log spam
+ # if more than one process was waiting for the rich media card
+ # while it was generated. Ideally we would set ttl here as well,
+ # so we don't override it number_of_waiters_on_generation
+ # times, but one, obviously, can't set ttl for not-yet-created entry
+ # and Cachex doesn't support returning ttl from the fetch callback.
+ log_error(url, reason)
+ {:commit, e}
+ end
+ end) do
+ {action, res} when action in [:commit, :ok] ->
+ case res do
+ {:ok, _data} = res ->
+ res
+
+ {:error, reason} = e ->
+ if action == :commit, do: set_error_ttl(url, reason)
+ e
+ end
+
+ {:error, e} ->
+ {:error, {:cachex_error, e}}
+ end
+ end
+
+ defp set_error_ttl(_url, :body_too_large), do: :ok
+ defp set_error_ttl(_url, {:content_type, _}), do: :ok
+
+ # The TTL is not set for the errors above, since they are unlikely to change
+ # with time
+
+ defp set_error_ttl(url, _reason) do
+ ttl = Pleroma.Config.get([:rich_media, :failure_backoff], 60_000)
+ @cachex.expire(:rich_media_cache, url, ttl)
+ :ok
+ end
+
+ defp log_error(url, {:invalid_metadata, data}) do
+ Logger.debug(fn -> "Incomplete or invalid metadata for #{url}: #{inspect(data)}" end)
+ end
+
+ defp log_error(url, reason) do
+ Logger.warn(fn -> "Rich media error for #{url}: #{inspect(reason)}" end)
+ end
end
@doc """
@@ -54,19 +101,26 @@ defmodule Pleroma.Web.RichMedia.Parser do
config :pleroma, :rich_media,
ttl_setters: [MyModule]
"""
- def set_ttl_based_on_image({:ok, data}, url) do
- with {:ok, nil} <- Cachex.ttl(:rich_media_cache, url),
- ttl when is_number(ttl) <- get_ttl_from_image(data, url) do
- Cachex.expire_at(:rich_media_cache, url, ttl * 1000)
- {:ok, data}
- else
+ @spec set_ttl_based_on_image(map(), String.t()) ::
+ {:ok, Integer.t() | :noop} | {:error, :no_key}
+ def set_ttl_based_on_image(data, url) do
+ case get_ttl_from_image(data, url) do
+ {:ok, ttl} when is_number(ttl) ->
+ ttl = ttl * 1000
+
+ case @cachex.expire_at(:rich_media_cache, url, ttl) do
+ {:ok, true} -> {:ok, ttl}
+ {:ok, false} -> {:error, :no_key}
+ end
+
_ ->
- {:ok, data}
+ {:ok, :noop}
end
end
defp get_ttl_from_image(data, url) do
- Pleroma.Config.get([:rich_media, :ttl_setters])
+ [:rich_media, :ttl_setters]
+ |> Pleroma.Config.get()
|> Enum.reduce({:ok, nil}, fn
module, {:ok, _ttl} ->
module.ttl(data, url)
@@ -76,50 +130,54 @@ defmodule Pleroma.Web.RichMedia.Parser do
end)
end
- defp parse_url(url) do
- try do
- {:ok, %Tesla.Env{body: html}} = Pleroma.HTTP.get(url, [], adapter: @hackney_options)
-
+ def parse_url(url) do
+ with {:ok, %Tesla.Env{body: html}} <- Pleroma.Web.RichMedia.Helpers.rich_media_get(url),
+ {:ok, html} <- Floki.parse_document(html) do
html
- |> parse_html()
|> maybe_parse()
- |> Map.put(:url, url)
+ |> Map.put("url", url)
|> clean_parsed_data()
|> check_parsed_data()
- rescue
- e ->
- {:error, "Parsing error: #{inspect(e)} #{inspect(__STACKTRACE__)}"}
end
end
- defp parse_html(html), do: Floki.parse_document!(html)
+ def parse_with_timeout(url) do
+ try do
+ task =
+ Task.Supervisor.async_nolink(Pleroma.TaskSupervisor, fn ->
+ parse_url(url)
+ end)
+
+ Task.await(task, 5000)
+ catch
+ :exit, {:timeout, _} ->
+ Logger.warn("Timeout while fetching rich media for #{url}")
+ {:error, :timeout}
+ end
+ end
defp maybe_parse(html) do
Enum.reduce_while(parsers(), %{}, fn parser, acc ->
case parser.parse(html, acc) do
- {:ok, data} -> {:halt, data}
- {:error, _msg} -> {:cont, acc}
+ data when data != %{} -> {:halt, data}
+ _ -> {:cont, acc}
end
end)
end
- defp check_parsed_data(%{title: title} = data)
- when is_binary(title) and byte_size(title) > 0 do
+ defp check_parsed_data(%{"title" => title} = data)
+ when is_binary(title) and title != "" do
{:ok, data}
end
defp check_parsed_data(data) do
- {:error, "Found metadata was invalid or incomplete: #{inspect(data)}"}
+ {:error, {:invalid_metadata, data}}
end
defp clean_parsed_data(data) do
data
|> Enum.reject(fn {key, val} ->
- with {:ok, _} <- Jason.encode(%{key => val}) do
- false
- else
- _ -> true
- end
+ not match?({:ok, _}, Jason.encode(%{key => val}))
end)
|> Map.new()
end