Merge branch '1505-threads-federation' into 'develop'

[#1505] Threads / replies federation

Closes #1505

See merge request pleroma/pleroma!2129
This commit is contained in:
Haelwenn 2020-02-24 07:28:35 +00:00
commit 1e1156b645
15 changed files with 366 additions and 25 deletions

View File

@ -75,6 +75,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
- A new users admin digest email
- OAuth: admin scopes support (relevant setting: `[:auth, :enforce_oauth_admin_scope_usage]`).
- Add an option `authorized_fetch_mode` to require HTTP signatures for AP fetches.
- ActivityPub: support for `replies` collection (output for outgoing federation & fetching on incoming federation).
<details>
<summary>API Changes</summary>
@ -117,6 +118,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
- Tag feed: `/tags/:tag.rss` - list public statuses by hashtag.
- Mastodon API: Add `reacted` property to `emoji_reactions`
- Pleroma API: Add reactions for a single emoji.
- ActivityPub: `[:activitypub, :note_replies_output_limit]` setting sets the number of note self-replies to output on outgoing federation.
</details>
### Fixed

View File

@ -328,6 +328,7 @@
unfollow_blocked: true,
outgoing_blocks: true,
follow_handshake_timeout: 500,
note_replies_output_limit: 5,
sign_object_fetches: true,
authorized_fetch_mode: false
@ -483,6 +484,7 @@
transmogrifier: 20,
scheduled_activities: 10,
background: 5,
remote_fetcher: 2,
attachments_cleanup: 5,
new_users_digest: 1
],

View File

@ -662,7 +662,7 @@
label: "Fed. incoming replies max depth",
type: :integer,
description:
"Max. depth of reply-to activities fetching on incoming federation, to prevent out-of-memory situations while" <>
"Max. depth of reply-to and reply activities fetching on incoming federation, to prevent out-of-memory situations while" <>
" fetching very long threads. If set to `nil`, threads of any depth will be fetched. Lower this value if you experience out-of-memory crashes.",
suggestions: [
100
@ -1790,6 +1790,12 @@
type: :boolean,
description: "Sign object fetches with HTTP signatures"
},
%{
key: :note_replies_output_limit,
type: :integer,
description:
"The number of Note replies' URIs to be included with outgoing federation (`5` to match Mastodon hardcoded value, `0` to disable the output)."
},
%{
key: :follow_handshake_timeout,
type: :integer,

View File

@ -41,7 +41,7 @@ On the top right you will also see a wrench icon. This opens your personal setti
This is where the interesting stuff happens!
Depending on the timeline you will see different statuses, but each status has a standard structure:
- Profile pic, name and link to profile. An optional left-arrow if it's a reply to another status (hovering will reveal the replied-to status). Clicking on the profile pic will uncollapse the user's profile.
- Profile pic, name and link to profile. An optional left-arrow if it's a reply to another status (hovering will reveal the reply-to status). Clicking on the profile pic will uncollapse the user's profile.
- A `+` button on the right allows you to Expand/Collapse an entire discussion thread. It also updates in realtime!
- An arrow icon allows you to open the status on the instance where it's originating from.
- The text of the status, including mentions and attachements. If you click on a mention, it will automatically open the profile page of that person.

View File

@ -7,7 +7,7 @@ defmodule Pleroma.Activity.Queries do
Contains queries for Activity.
"""
import Ecto.Query, only: [from: 2]
import Ecto.Query, only: [from: 2, where: 3]
@type query :: Ecto.Queryable.t() | Activity.t()
@ -63,6 +63,22 @@ def by_object_id(query, object_id) when is_binary(object_id) do
)
end
@spec by_object_in_reply_to_id(query, String.t(), keyword()) :: query
def by_object_in_reply_to_id(query, in_reply_to_id, opts \\ []) do
query =
if opts[:skip_preloading] do
Activity.with_joined_object(query)
else
Activity.with_preloaded_object(query)
end
where(
query,
[activity, object: o],
fragment("(?)->>'inReplyTo' = ?", o.data, ^to_string(in_reply_to_id))
)
end
@spec by_type(query, String.t()) :: query
def by_type(query \\ Activity, activity_type) do
from(

View File

@ -301,4 +301,26 @@ def update_data(%Object{data: data} = object, attrs \\ %{}) do
def local?(%Object{data: %{"id" => id}}) do
String.starts_with?(id, Pleroma.Web.base_url() <> "/")
end
def replies(object, opts \\ []) do
object = Object.normalize(object)
query =
Object
|> where(
[o],
fragment("(?)->>'inReplyTo' = ?", o.data, ^object.data["id"])
)
|> order_by([o], asc: o.id)
if opts[:self_only] do
actor = object.data["actor"]
where(query, [o], fragment("(?)->>'actor' = ?", o.data, ^actor))
else
query
end
end
def self_replies(object, opts \\ []),
do: replies(object, Keyword.put(opts, :self_only, true))
end

View File

@ -10,6 +10,7 @@ defmodule Pleroma.Object.Fetcher do
alias Pleroma.Signature
alias Pleroma.Web.ActivityPub.InternalFetchActor
alias Pleroma.Web.ActivityPub.Transmogrifier
alias Pleroma.Web.Federator
require Logger
require Pleroma.Constants
@ -59,20 +60,23 @@ def refetch_object(%Object{data: %{"id" => id}} = object) do
end
end
# TODO:
# This will create a Create activity, which we need internally at the moment.
# Note: will create a Create activity, which we need internally at the moment.
def fetch_object_from_id(id, options \\ []) do
with {:fetch_object, nil} <- {:fetch_object, Object.get_cached_by_ap_id(id)},
{:fetch, {:ok, data}} <- {:fetch, fetch_and_contain_remote_object_from_id(id)},
{:normalize, nil} <- {:normalize, Object.normalize(data, false)},
with {_, nil} <- {:fetch_object, Object.get_cached_by_ap_id(id)},
{_, true} <- {:allowed_depth, Federator.allowed_thread_distance?(options[:depth])},
{_, {:ok, data}} <- {:fetch, fetch_and_contain_remote_object_from_id(id)},
{_, nil} <- {:normalize, Object.normalize(data, false)},
params <- prepare_activity_params(data),
{:containment, :ok} <- {:containment, Containment.contain_origin(id, params)},
{:transmogrifier, {:ok, activity}} <-
{_, :ok} <- {:containment, Containment.contain_origin(id, params)},
{_, {:ok, activity}} <-
{:transmogrifier, Transmogrifier.handle_incoming(params, options)},
{:object, _data, %Object{} = object} <-
{_, _data, %Object{} = object} <-
{:object, data, Object.normalize(activity, false)} do
{:ok, object}
else
{:allowed_depth, false} ->
{:error, "Max thread distance exceeded."}
{:containment, _} ->
{:error, "Object containment failed."}

View File

@ -156,8 +156,9 @@ def fix_in_reply_to(%{"inReplyTo" => in_reply_to} = object, options)
when not is_nil(in_reply_to) do
in_reply_to_id = prepare_in_reply_to(in_reply_to)
object = Map.put(object, "inReplyToAtomUri", in_reply_to_id)
depth = (options[:depth] || 0) + 1
if Federator.allowed_incoming_reply_depth?(options[:depth]) do
if Federator.allowed_thread_distance?(depth) do
with {:ok, replied_object} <- get_obj_helper(in_reply_to_id, options),
%Activity{} = _ <- Activity.get_create_by_object_ap_id(replied_object.data["id"]) do
object
@ -312,7 +313,7 @@ def fix_type(object, options \\ [])
def fix_type(%{"inReplyTo" => reply_id, "name" => _} = object, options)
when is_binary(reply_id) do
with true <- Federator.allowed_incoming_reply_depth?(options[:depth]),
with true <- Federator.allowed_thread_distance?(options[:depth]),
{:ok, %{data: %{"type" => "Question"} = _} = _} <- get_obj_helper(reply_id, options) do
Map.put(object, "type", "Answer")
else
@ -406,8 +407,7 @@ def handle_incoming(
with nil <- Activity.get_create_by_object_ap_id(object["id"]),
{:ok, %User{} = user} <- User.get_or_fetch_by_ap_id(data["actor"]) do
options = Keyword.put(options, :depth, (options[:depth] || 0) + 1)
object = fix_object(data["object"], options)
object = fix_object(object, options)
params = %{
to: data["to"],
@ -424,7 +424,20 @@ def handle_incoming(
])
}
ActivityPub.create(params)
with {:ok, created_activity} <- ActivityPub.create(params) do
reply_depth = (options[:depth] || 0) + 1
if Federator.allowed_thread_distance?(reply_depth) do
for reply_id <- replies(object) do
Pleroma.Workers.RemoteFetcherWorker.enqueue("fetch_remote", %{
"id" => reply_id,
"depth" => reply_depth
})
end
end
{:ok, created_activity}
end
else
%Activity{} = activity -> {:ok, activity}
_e -> :error
@ -442,7 +455,8 @@ def handle_incoming(
|> fix_addressing
with {:ok, %User{} = user} <- User.get_or_fetch_by_ap_id(data["actor"]) do
options = Keyword.put(options, :depth, (options[:depth] || 0) + 1)
reply_depth = (options[:depth] || 0) + 1
options = Keyword.put(options, :depth, reply_depth)
object = fix_object(object, options)
params = %{
@ -903,6 +917,50 @@ def set_reply_to_uri(%{"inReplyTo" => in_reply_to} = object) when is_binary(in_r
def set_reply_to_uri(obj), do: obj
@doc """
Serialized Mastodon-compatible `replies` collection containing _self-replies_.
Based on Mastodon's ActivityPub::NoteSerializer#replies.
"""
def set_replies(obj_data) do
replies_uris =
with limit when limit > 0 <-
Pleroma.Config.get([:activitypub, :note_replies_output_limit], 0),
%Object{} = object <- Object.get_cached_by_ap_id(obj_data["id"]) do
object
|> Object.self_replies()
|> select([o], fragment("?->>'id'", o.data))
|> limit(^limit)
|> Repo.all()
else
_ -> []
end
set_replies(obj_data, replies_uris)
end
defp set_replies(obj, []) do
obj
end
defp set_replies(obj, replies_uris) do
replies_collection = %{
"type" => "Collection",
"items" => replies_uris
}
Map.merge(obj, %{"replies" => replies_collection})
end
def replies(%{"replies" => %{"first" => %{"items" => items}}}) when not is_nil(items) do
items
end
def replies(%{"replies" => %{"items" => items}}) when not is_nil(items) do
items
end
def replies(_), do: []
# Prepares the object of an outgoing create activity.
def prepare_object(object) do
object
@ -914,6 +972,7 @@ def prepare_object(object) do
|> prepare_attachments
|> set_conversation
|> set_reply_to_uri
|> set_replies
|> strip_internal_fields
|> strip_internal_tags
|> set_type

View File

@ -15,13 +15,19 @@ defmodule Pleroma.Web.Federator do
require Logger
@doc "Addresses [memory leaks on recursive replies fetching](https://git.pleroma.social/pleroma/pleroma/issues/161)"
@doc """
Returns `true` if the distance to target object does not exceed max configured value.
Serves to prevent fetching of very long threads, especially useful on smaller instances.
Addresses [memory leaks on recursive replies fetching](https://git.pleroma.social/pleroma/pleroma/issues/161).
Applies to fetching of both ancestor (reply-to) and child (reply) objects.
"""
# credo:disable-for-previous-line Credo.Check.Readability.MaxLineLength
def allowed_incoming_reply_depth?(depth) do
max_replies_depth = Pleroma.Config.get([:instance, :federation_incoming_replies_max_depth])
def allowed_thread_distance?(distance) do
max_distance = Pleroma.Config.get([:instance, :federation_incoming_replies_max_depth])
if max_replies_depth do
(depth || 1) <= max_replies_depth
if max_distance && max_distance >= 0 do
# Default depth is 0 (an object has zero distance from itself in its thread)
(distance || 0) <= max_distance
else
true
end

View File

@ -0,0 +1,20 @@
# Pleroma: A lightweight social networking server
# Copyright © 2017-2019 Pleroma Authors <https://pleroma.social/>
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Workers.RemoteFetcherWorker do
alias Pleroma.Object.Fetcher
use Pleroma.Workers.WorkerHelper, queue: "remote_fetcher"
@impl Oban.Worker
def perform(
%{
"op" => "fetch_remote",
"id" => id
} = args,
_job
) do
{:ok, _object} = Fetcher.fetch_object_from_id(id, depth: args["depth"])
end
end

View File

@ -35,6 +35,19 @@
"inReplyTo": null,
"inReplyToAtomUri": null,
"published": "2018-02-12T14:08:20Z",
"replies": {
"id": "http://mastodon.example.org/users/admin/statuses/99512778738411822/replies",
"type": "Collection",
"first": {
"type": "CollectionPage",
"next": "http://mastodon.example.org/users/admin/statuses/99512778738411822/replies?min_id=99512778738411824&page=true",
"partOf": "http://mastodon.example.org/users/admin/statuses/99512778738411822/replies",
"items": [
"http://mastodon.example.org/users/admin/statuses/99512778738411823",
"http://mastodon.example.org/users/admin/statuses/99512778738411824"
]
}
},
"sensitive": true,
"summary": "cw",
"tag": [

View File

@ -26,6 +26,31 @@ defmodule Pleroma.Object.FetcherTest do
:ok
end
describe "max thread distance restriction" do
@ap_id "http://mastodon.example.org/@admin/99541947525187367"
clear_config([:instance, :federation_incoming_replies_max_depth])
test "it returns thread depth exceeded error if thread depth is exceeded" do
Pleroma.Config.put([:instance, :federation_incoming_replies_max_depth], 0)
assert {:error, "Max thread distance exceeded."} =
Fetcher.fetch_object_from_id(@ap_id, depth: 1)
end
test "it fetches object if max thread depth is restricted to 0 and depth is not specified" do
Pleroma.Config.put([:instance, :federation_incoming_replies_max_depth], 0)
assert {:ok, _} = Fetcher.fetch_object_from_id(@ap_id)
end
test "it fetches object if requested depth does not exceed max thread depth" do
Pleroma.Config.put([:instance, :federation_incoming_replies_max_depth], 10)
assert {:ok, _} = Fetcher.fetch_object_from_id(@ap_id, depth: 10)
end
end
describe "actor origin containment" do
test "it rejects objects with a bogus origin" do
{:error, _} = Fetcher.fetch_object_from_id("https://info.pleroma.site/activity.json")

View File

@ -9,6 +9,10 @@ defmodule Pleroma.Tests.ObanHelpers do
alias Pleroma.Repo
def wipe_all do
Repo.delete_all(Oban.Job)
end
def perform_all do
Oban.Job
|> Repo.all()

View File

@ -3,7 +3,9 @@
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Web.ActivityPub.TransmogrifierTest do
use Oban.Testing, repo: Pleroma.Repo
use Pleroma.DataCase
alias Pleroma.Activity
alias Pleroma.Object
alias Pleroma.Object.Fetcher
@ -40,7 +42,7 @@ test "it ignores an incoming notice if we already have it" do
end
@tag capture_log: true
test "it fetches replied-to activities if we don't have them" do
test "it fetches reply-to activities if we don't have them" do
data =
File.read!("test/fixtures/mastodon-post-activity.json")
|> Poison.decode!()
@ -61,7 +63,7 @@ test "it fetches replied-to activities if we don't have them" do
assert returned_object.data["inReplyToAtomUri"] == "https://shitposter.club/notice/2827873"
end
test "it does not fetch replied-to activities beyond max_replies_depth" do
test "it does not fetch reply-to activities beyond max replies depth limit" do
data =
File.read!("test/fixtures/mastodon-post-activity.json")
|> Poison.decode!()
@ -73,7 +75,7 @@ test "it does not fetch replied-to activities beyond max_replies_depth" do
data = Map.put(data, "object", object)
with_mock Pleroma.Web.Federator,
allowed_incoming_reply_depth?: fn _ -> false end do
allowed_thread_distance?: fn _ -> false end do
{:ok, returned_activity} = Transmogrifier.handle_incoming(data)
returned_object = Object.normalize(returned_activity, false)
@ -1348,6 +1350,101 @@ test "it accepts Move activities" do
end
end
describe "`handle_incoming/2`, Mastodon format `replies` handling" do
clear_config([:activitypub, :note_replies_output_limit]) do
Pleroma.Config.put([:activitypub, :note_replies_output_limit], 5)
end
clear_config([:instance, :federation_incoming_replies_max_depth])
setup do
data =
"test/fixtures/mastodon-post-activity.json"
|> File.read!()
|> Poison.decode!()
items = get_in(data, ["object", "replies", "first", "items"])
assert length(items) > 0
%{data: data, items: items}
end
test "schedules background fetching of `replies` items if max thread depth limit allows", %{
data: data,
items: items
} do
Pleroma.Config.put([:instance, :federation_incoming_replies_max_depth], 10)
{:ok, _activity} = Transmogrifier.handle_incoming(data)
for id <- items do
job_args = %{"op" => "fetch_remote", "id" => id, "depth" => 1}
assert_enqueued(worker: Pleroma.Workers.RemoteFetcherWorker, args: job_args)
end
end
test "does NOT schedule background fetching of `replies` beyond max thread depth limit allows",
%{data: data} do
Pleroma.Config.put([:instance, :federation_incoming_replies_max_depth], 0)
{:ok, _activity} = Transmogrifier.handle_incoming(data)
assert all_enqueued(worker: Pleroma.Workers.RemoteFetcherWorker) == []
end
end
describe "`handle_incoming/2`, Pleroma format `replies` handling" do
clear_config([:activitypub, :note_replies_output_limit]) do
Pleroma.Config.put([:activitypub, :note_replies_output_limit], 5)
end
clear_config([:instance, :federation_incoming_replies_max_depth])
setup do
user = insert(:user)
{:ok, activity} = CommonAPI.post(user, %{"status" => "post1"})
{:ok, reply1} =
CommonAPI.post(user, %{"status" => "reply1", "in_reply_to_status_id" => activity.id})
{:ok, reply2} =
CommonAPI.post(user, %{"status" => "reply2", "in_reply_to_status_id" => activity.id})
replies_uris = Enum.map([reply1, reply2], fn a -> a.object.data["id"] end)
{:ok, federation_output} = Transmogrifier.prepare_outgoing(activity.data)
Repo.delete(activity.object)
Repo.delete(activity)
%{federation_output: federation_output, replies_uris: replies_uris}
end
test "schedules background fetching of `replies` items if max thread depth limit allows", %{
federation_output: federation_output,
replies_uris: replies_uris
} do
Pleroma.Config.put([:instance, :federation_incoming_replies_max_depth], 1)
{:ok, _activity} = Transmogrifier.handle_incoming(federation_output)
for id <- replies_uris do
job_args = %{"op" => "fetch_remote", "id" => id, "depth" => 1}
assert_enqueued(worker: Pleroma.Workers.RemoteFetcherWorker, args: job_args)
end
end
test "does NOT schedule background fetching of `replies` beyond max thread depth limit allows",
%{federation_output: federation_output} do
Pleroma.Config.put([:instance, :federation_incoming_replies_max_depth], 0)
{:ok, _activity} = Transmogrifier.handle_incoming(federation_output)
assert all_enqueued(worker: Pleroma.Workers.RemoteFetcherWorker) == []
end
end
describe "prepare outgoing" do
test "it inlines private announced objects" do
user = insert(:user)
@ -2046,4 +2143,49 @@ test "returns object with emoji when object contains map tag" do
}
end
end
describe "set_replies/1" do
clear_config([:activitypub, :note_replies_output_limit]) do
Pleroma.Config.put([:activitypub, :note_replies_output_limit], 2)
end
test "returns unmodified object if activity doesn't have self-replies" do
data = Poison.decode!(File.read!("test/fixtures/mastodon-post-activity.json"))
assert Transmogrifier.set_replies(data) == data
end
test "sets `replies` collection with a limited number of self-replies" do
[user, another_user] = insert_list(2, :user)
{:ok, %{id: id1} = activity} = CommonAPI.post(user, %{"status" => "1"})
{:ok, %{id: id2} = self_reply1} =
CommonAPI.post(user, %{"status" => "self-reply 1", "in_reply_to_status_id" => id1})
{:ok, self_reply2} =
CommonAPI.post(user, %{"status" => "self-reply 2", "in_reply_to_status_id" => id1})
# Assuming to _not_ be present in `replies` due to :note_replies_output_limit is set to 2
{:ok, _} =
CommonAPI.post(user, %{"status" => "self-reply 3", "in_reply_to_status_id" => id1})
{:ok, _} =
CommonAPI.post(user, %{
"status" => "self-reply to self-reply",
"in_reply_to_status_id" => id2
})
{:ok, _} =
CommonAPI.post(another_user, %{
"status" => "another user's reply",
"in_reply_to_status_id" => id1
})
object = Object.normalize(activity)
replies_uris = Enum.map([self_reply1, self_reply2], fn a -> a.object.data["id"] end)
assert %{"type" => "Collection", "items" => ^replies_uris} =
Transmogrifier.set_replies(object.data)["replies"]
end
end
end

View File

@ -36,6 +36,26 @@ test "renders a note activity" do
assert result["@context"]
end
describe "note activity's `replies` collection rendering" do
clear_config([:activitypub, :note_replies_output_limit]) do
Pleroma.Config.put([:activitypub, :note_replies_output_limit], 5)
end
test "renders `replies` collection for a note activity" do
user = insert(:user)
activity = insert(:note_activity, user: user)
{:ok, self_reply1} =
CommonAPI.post(user, %{"status" => "self-reply 1", "in_reply_to_status_id" => activity.id})
replies_uris = [self_reply1.object.data["id"]]
result = ObjectView.render("object.json", %{object: refresh_record(activity)})
assert %{"type" => "Collection", "items" => ^replies_uris} =
get_in(result, ["object", "replies"])
end
end
test "renders a like activity" do
note = insert(:note_activity)
object = Object.normalize(note)