From a1c4a5d7cf99247ef9964fb530a6d2d46468dc82 Mon Sep 17 00:00:00 2001 From: rinpatch Date: Fri, 18 Oct 2019 14:11:30 +0300 Subject: Fix a migration wiping user info of users that don't have any mutes And introduce safe_jsonb_set --- lib/mix/tasks/pleroma/database.ex | 4 ++-- lib/pleroma/object.ex | 4 ++-- lib/pleroma/user.ex | 6 +++--- lib/pleroma/web/activity_pub/utils.ex | 2 +- .../20190711042021_create_safe_jsonb_set.exs | 22 ++++++++++++++++++++++ ...711042024_copy_muted_to_muted_notifications.exs | 2 +- 6 files changed, 31 insertions(+), 9 deletions(-) create mode 100644 priv/repo/migrations/20190711042021_create_safe_jsonb_set.exs diff --git a/lib/mix/tasks/pleroma/database.ex b/lib/mix/tasks/pleroma/database.ex index bcc2052d6..ab7529e08 100644 --- a/lib/mix/tasks/pleroma/database.ex +++ b/lib/mix/tasks/pleroma/database.ex @@ -54,7 +54,7 @@ def run(["remove_embedded_objects" | args]) do Logger.info("Removing embedded objects") Repo.query!( - "update activities set data = jsonb_set(data, '{object}'::text[], data->'object'->'id') where data->'object'->>'id' is not null;", + "update activities set data = safe_jsonb_set(data, '{object}'::text[], data->'object'->'id') where data->'object'->>'id' is not null;", [], timeout: :infinity ) @@ -152,7 +152,7 @@ def run(["fix_likes_collections"]) do set: [ data: fragment( - "jsonb_set(?, '{likes}', '[]'::jsonb, true)", + "safe_jsonb_set(?, '{likes}', '[]'::jsonb, true)", object.data ) ] diff --git a/lib/pleroma/object.ex b/lib/pleroma/object.ex index 3fa407931..bf37b28a7 100644 --- a/lib/pleroma/object.ex +++ b/lib/pleroma/object.ex @@ -181,7 +181,7 @@ def increase_replies_count(ap_id) do data: fragment( """ - jsonb_set(?, '{repliesCount}', + safe_jsonb_set(?, '{repliesCount}', (coalesce((?->>'repliesCount')::int, 0) + 1)::varchar::jsonb, true) """, o.data, @@ -204,7 +204,7 @@ def decrease_replies_count(ap_id) do data: fragment( """ - jsonb_set(?, '{repliesCount}', + safe_jsonb_set(?, '{repliesCount}', (greatest(0, (?->>'repliesCount')::int - 1))::varchar::jsonb, true) """, o.data, diff --git a/lib/pleroma/user.ex b/lib/pleroma/user.ex index 617f160e0..f0912fb10 100644 --- a/lib/pleroma/user.ex +++ b/lib/pleroma/user.ex @@ -718,7 +718,7 @@ def increase_note_count(%User{} = user) do set: [ info: fragment( - "jsonb_set(?, '{note_count}', ((?->>'note_count')::int + 1)::varchar::jsonb, true)", + "safe_jsonb_set(?, '{note_count}', ((?->>'note_count')::int + 1)::varchar::jsonb, true)", u.info, u.info ) @@ -739,7 +739,7 @@ def decrease_note_count(%User{} = user) do set: [ info: fragment( - "jsonb_set(?, '{note_count}', (greatest(0, (?->>'note_count')::int - 1))::varchar::jsonb, true)", + "safe_jsonb_set(?, '{note_count}', (greatest(0, (?->>'note_count')::int - 1))::varchar::jsonb, true)", u.info, u.info ) @@ -812,7 +812,7 @@ def update_follower_count(%User{} = user) do set: [ info: fragment( - "jsonb_set(?, '{follower_count}', ?::varchar::jsonb, true)", + "safe_jsonb_set(?, '{follower_count}', ?::varchar::jsonb, true)", u.info, s.count ) diff --git a/lib/pleroma/web/activity_pub/utils.ex b/lib/pleroma/web/activity_pub/utils.ex index 39a532db3..f22cc2367 100644 --- a/lib/pleroma/web/activity_pub/utils.ex +++ b/lib/pleroma/web/activity_pub/utils.ex @@ -349,7 +349,7 @@ def update_follow_state_for_all( try do Ecto.Adapters.SQL.query!( Repo, - "UPDATE activities SET data = jsonb_set(data, '{state}', $1) WHERE data->>'type' = 'Follow' AND data->>'actor' = $2 AND data->>'object' = $3 AND data->>'state' = 'pending'", + "UPDATE activities SET data = safe_jsonb_set(data, '{state}', $1) WHERE data->>'type' = 'Follow' AND data->>'actor' = $2 AND data->>'object' = $3 AND data->>'state' = 'pending'", [state, actor, object] ) diff --git a/priv/repo/migrations/20190711042021_create_safe_jsonb_set.exs b/priv/repo/migrations/20190711042021_create_safe_jsonb_set.exs new file mode 100644 index 000000000..d70961431 --- /dev/null +++ b/priv/repo/migrations/20190711042021_create_safe_jsonb_set.exs @@ -0,0 +1,22 @@ +defmodule Pleroma.Repo.Migrations.CreateSafeJsonbSet do + use Ecto.Migration + alias Pleroma.User + + def change do + execute(""" + create or replace function safe_jsonb_set(target jsonb, path text[], new_value jsonb, create_missing boolean default true) returns jsonb as $$ + declare + result jsonb; + begin + result := jsonb_set(target, path, coalesce(new_value, 'null'::jsonb), create_missing); + if result is NULL then + raise 'jsonb_set tried to wipe the object, please report this incindent to Pleroma bug tracker. https://git.pleroma.social/pleroma/pleroma/issues/new'; + return target; + else + return result; + end if; + end; + $$ language plpgsql; + """) + end +end diff --git a/priv/repo/migrations/20190711042024_copy_muted_to_muted_notifications.exs b/priv/repo/migrations/20190711042024_copy_muted_to_muted_notifications.exs index b717cab2e..a3c5b52de 100644 --- a/priv/repo/migrations/20190711042024_copy_muted_to_muted_notifications.exs +++ b/priv/repo/migrations/20190711042024_copy_muted_to_muted_notifications.exs @@ -3,6 +3,6 @@ defmodule Pleroma.Repo.Migrations.CopyMutedToMutedNotifications do alias Pleroma.User def change do - execute("update users set info = jsonb_set(info, '{muted_notifications}', info->'mutes', true) where local = true") + execute("update users set info = safe_jsonb_set(info, '{muted_notifications}', info->'mutes', true) where local = true") end end -- cgit v1.2.3 From 98c9affc485a3d23d53cead9e6e40e837e467eae Mon Sep 17 00:00:00 2001 From: rinpatch Date: Fri, 18 Oct 2019 14:33:51 +0300 Subject: Add a test for safe_jsonb_set --- test/safe_jsonb_set_test.exs | 12 ++++++++++++ 1 file changed, 12 insertions(+) create mode 100644 test/safe_jsonb_set_test.exs diff --git a/test/safe_jsonb_set_test.exs b/test/safe_jsonb_set_test.exs new file mode 100644 index 000000000..748540570 --- /dev/null +++ b/test/safe_jsonb_set_test.exs @@ -0,0 +1,12 @@ +defmodule Pleroma.SafeJsonbSetTest do + use Pleroma.DataCase + + test "it doesn't wipe the object when asked to set the value to NULL" do + assert %{rows: [[%{"key" => "value", "test" => nil}]]} = + Ecto.Adapters.SQL.query!( + Pleroma.Repo, + "select safe_jsonb_set('{\"key\": \"value\"}'::jsonb, '{test}', NULL);", + [] + ) + end +end -- cgit v1.2.3 From 41c47639e3eae5d5e2840e9a9bb51cd99660af91 Mon Sep 17 00:00:00 2001 From: rinpatch Date: Fri, 18 Oct 2019 14:40:47 +0300 Subject: Add a changelog entry for yet another jsonb_set data loss --- CHANGELOG.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/CHANGELOG.md b/CHANGELOG.md index cb3785a02..3c282514a 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -3,6 +3,13 @@ All notable changes to this project will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/). +## [1.1.1] - 2019-10-18 +### Fixed +- One of the migrations between 1.0.0 and 1.1.0 wiping user info of the relay user because of unexpected behavior of postgresql's `jsonb_set`, resulting in inability to post in the default configuration. If you were affected, please run the following query in postgres console, the relay user will be recreated automatically: +``` +delete from users where ap_id = 'https://your.instance.hostname/relay'; +``` + ## [1.1.0] - 2019-10-14 **Breaking:** The stable branch has been changed from `master` to `stable`. If you want to keep using 1.0, the `release/1.0` branch will receive security updates for 6 months after 1.1 release. -- cgit v1.2.3 From 4123c9db0736f53d40792ca0a94b7c7724e544a2 Mon Sep 17 00:00:00 2001 From: rinpatch Date: Fri, 18 Oct 2019 14:53:13 +0300 Subject: Bump the version in mix.exs --- mix.exs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mix.exs b/mix.exs index 56a11cac3..56a3bfb3c 100644 --- a/mix.exs +++ b/mix.exs @@ -4,7 +4,7 @@ defmodule Pleroma.Mixfile do def project do [ app: :pleroma, - version: version("1.1.0"), + version: version("1.1.1"), elixir: "~> 1.7", elixirc_paths: elixirc_paths(Mix.env()), compilers: [:phoenix, :gettext] ++ Mix.compilers(), -- cgit v1.2.3 From 713b2187b93ebda5533e67173d82745eb15ba650 Mon Sep 17 00:00:00 2001 From: rinpatch Date: Wed, 16 Oct 2019 12:52:47 +0300 Subject: User search: Remove trigram and refactor the module - Remove trigram as it tends to rank garbage results highly, resulting in it prioritized above fts, which gives actually decent results. ACKed by kaniini and lain on irc. - Remove a test for handling misspelled requests, since we no longer have trigram - Remove a test for searching users with `nil` display names, because it is unrealistic, we don't accept usernames that are not >1 char strings - Make rank boosting for followers/followees sane again, previous values resulted in garbage matches getting on top just because the users are followers/followees --- lib/pleroma/user/search.ex | 152 +++++++++++++++------------------------------ test/user_search_test.exs | 17 ----- 2 files changed, 50 insertions(+), 119 deletions(-) diff --git a/lib/pleroma/user/search.ex b/lib/pleroma/user/search.ex index 6fb2c2352..fb2f3fedb 100644 --- a/lib/pleroma/user/search.ex +++ b/lib/pleroma/user/search.ex @@ -4,7 +4,6 @@ defmodule Pleroma.User.Search do alias Pleroma.Pagination - alias Pleroma.Repo alias Pleroma.User import Ecto.Query @@ -23,18 +22,10 @@ def search(query_string, opts \\ []) do maybe_resolve(resolve, for_user, query_string) - {:ok, results} = - Repo.transaction(fn -> - Ecto.Adapters.SQL.query( - Repo, - "select set_limit(#{@similarity_threshold})", - [] - ) - - query_string - |> search_query(for_user, following) - |> Pagination.fetch_paginated(%{"offset" => offset, "limit" => result_limit}, :offset) - end) + results = + query_string + |> search_query(for_user, following) + |> Pagination.fetch_paginated(%{"offset" => offset, "limit" => result_limit}, :offset) results end @@ -56,15 +47,52 @@ defp search_query(query_string, for_user, following) do |> base_query(following) |> filter_blocked_user(for_user) |> filter_blocked_domains(for_user) - |> search_subqueries(query_string) - |> union_subqueries - |> distinct_query() - |> boost_search_rank_query(for_user) + |> fts_subquery(query_string) |> subquery() + |> where([u], u.search_rank > @similarity_threshold) + |> boost_search_rank(for_user) |> order_by(desc: :search_rank) |> maybe_restrict_local(for_user) end + @nickname_regex ~r/^[a-zA-Z0-9.!#$%&'*+\/=?^_`{|}~\-@]+$/ + defp fts_subquery(query, query_string) do + {nickname_weight, name_weight} = + if String.match?(query_string, @nickname_regex) do + {"A", "B"} + else + {"B", "A"} + end + + query_string = to_tsquery(query_string) + + from( + u in query, + select_merge: %{ + search_rank: + fragment( + """ + ts_rank_cd((setweight(to_tsvector('simple', ?), ?) || setweight(to_tsvector('simple', ?), ?)), to_tsquery('simple', ?)) + """, + u.name, + ^name_weight, + u.nickname, + ^nickname_weight, + ^query_string + ) + } + ) + end + + defp to_tsquery(query_string) do + String.trim_trailing(query_string, "@" <> local_domain()) + |> String.replace(~r/[!-\/|@|[-`|{-~|:-?]+/, " ") + |> String.trim() + |> String.split() + |> Enum.map(&(&1 <> ":*")) + |> Enum.join(" | ") + end + defp base_query(_user, false), do: User defp base_query(user, true), do: User.get_followers_query(user) @@ -87,21 +115,6 @@ defp filter_blocked_domains(query, %User{info: %{domain_blocks: domain_blocks}}) defp filter_blocked_domains(query, _), do: query - defp union_subqueries({fts_subquery, trigram_subquery}) do - from(s in trigram_subquery, union_all: ^fts_subquery) - end - - defp search_subqueries(base_query, query_string) do - { - fts_search_subquery(base_query, query_string), - trigram_search_subquery(base_query, query_string) - } - end - - defp distinct_query(q) do - from(s in subquery(q), order_by: s.search_type, distinct: s.id) - end - defp maybe_resolve(true, user, query) do case {limit(), user} do {:all, _} -> :noop @@ -126,9 +139,9 @@ defp limit, do: Pleroma.Config.get([:instance, :limit_to_local_content], :unauth defp restrict_local(q), do: where(q, [u], u.local == true) - defp boost_search_rank_query(query, nil), do: query + defp local_domain, do: Pleroma.Config.get([Pleroma.Web.Endpoint, :url, :host]) - defp boost_search_rank_query(query, for_user) do + defp boost_search_rank(query, %User{} = for_user) do friends_ids = User.get_friends_ids(for_user) followers_ids = User.get_followers_ids(for_user) @@ -137,8 +150,8 @@ defp boost_search_rank_query(query, for_user) do search_rank: fragment( """ - CASE WHEN (?) THEN 0.5 + (?) * 1.3 - WHEN (?) THEN 0.5 + (?) * 1.2 + CASE WHEN (?) THEN (?) * 1.5 + WHEN (?) THEN (?) * 1.3 WHEN (?) THEN (?) * 1.1 ELSE (?) END """, @@ -154,70 +167,5 @@ defp boost_search_rank_query(query, for_user) do ) end - @spec fts_search_subquery(User.t() | Ecto.Query.t(), String.t()) :: Ecto.Query.t() - defp fts_search_subquery(query, term) do - processed_query = - String.trim_trailing(term, "@" <> local_domain()) - |> String.replace(~r/[!-\/|@|[-`|{-~|:-?]+/, " ") - |> String.trim() - |> String.split() - |> Enum.map(&(&1 <> ":*")) - |> Enum.join(" | ") - - from( - u in query, - select_merge: %{ - search_type: ^0, - search_rank: - fragment( - """ - ts_rank_cd( - setweight(to_tsvector('simple', regexp_replace(?, '\\W', ' ', 'g')), 'A') || - setweight(to_tsvector('simple', regexp_replace(coalesce(?, ''), '\\W', ' ', 'g')), 'B'), - to_tsquery('simple', ?), - 32 - ) - """, - u.nickname, - u.name, - ^processed_query - ) - }, - where: - fragment( - """ - (setweight(to_tsvector('simple', regexp_replace(?, '\\W', ' ', 'g')), 'A') || - setweight(to_tsvector('simple', regexp_replace(coalesce(?, ''), '\\W', ' ', 'g')), 'B')) @@ to_tsquery('simple', ?) - """, - u.nickname, - u.name, - ^processed_query - ) - ) - |> User.restrict_deactivated() - end - - @spec trigram_search_subquery(User.t() | Ecto.Query.t(), String.t()) :: Ecto.Query.t() - defp trigram_search_subquery(query, term) do - term = String.trim_trailing(term, "@" <> local_domain()) - - from( - u in query, - select_merge: %{ - # ^1 gives 'Postgrex expected a binary, got 1' for some weird reason - search_type: fragment("?", 1), - search_rank: - fragment( - "similarity(?, trim(? || ' ' || coalesce(?, '')))", - ^term, - u.nickname, - u.name - ) - }, - where: fragment("trim(? || ' ' || coalesce(?, '')) % ?", u.nickname, u.name, ^term) - ) - |> User.restrict_deactivated() - end - - defp local_domain, do: Pleroma.Config.get([Pleroma.Web.Endpoint, :url, :host]) + defp boost_search_rank(query, _for_user), do: query end diff --git a/test/user_search_test.exs b/test/user_search_test.exs index 48ce973ad..341074031 100644 --- a/test/user_search_test.exs +++ b/test/user_search_test.exs @@ -74,12 +74,6 @@ test "finds users, ranking by similarity" do assert [u4.id, u3.id, u1.id] == Enum.map(User.search("lain@ple", for_user: u1), & &1.id) end - test "finds users, handling misspelled requests" do - u1 = insert(:user, %{name: "lain"}) - - assert [u1.id] == Enum.map(User.search("laiin"), & &1.id) - end - test "finds users, boosting ranks of friends and followers" do u1 = insert(:user) u2 = insert(:user, %{name: "Doe"}) @@ -163,17 +157,6 @@ test "find all users for unauthenticated users when `limit_to_local_content` is Pleroma.Config.put([:instance, :limit_to_local_content], :unauthenticated) end - test "finds a user whose name is nil" do - _user = insert(:user, %{name: "notamatch", nickname: "testuser@pleroma.amplifie.red"}) - user_two = insert(:user, %{name: nil, nickname: "lain@pleroma.soykaf.com"}) - - assert user_two == - User.search("lain@pleroma.soykaf.com") - |> List.first() - |> Map.put(:search_rank, nil) - |> Map.put(:search_type, nil) - end - test "does not yield false-positive matches" do insert(:user, %{name: "John Doe"}) -- cgit v1.2.3 From 7a00acb3e49f441d975ff509efabb0117875c61e Mon Sep 17 00:00:00 2001 From: rinpatch Date: Wed, 16 Oct 2019 13:49:33 +0300 Subject: Order fts results by trigram --- lib/pleroma/user/search.ex | 48 +++++++++++++++++++++++++++++----------------- 1 file changed, 30 insertions(+), 18 deletions(-) diff --git a/lib/pleroma/user/search.ex b/lib/pleroma/user/search.ex index fb2f3fedb..0d697fe3d 100644 --- a/lib/pleroma/user/search.ex +++ b/lib/pleroma/user/search.ex @@ -7,7 +7,6 @@ defmodule Pleroma.User.Search do alias Pleroma.User import Ecto.Query - @similarity_threshold 0.25 @limit 20 def search(query_string, opts \\ []) do @@ -47,16 +46,16 @@ defp search_query(query_string, for_user, following) do |> base_query(following) |> filter_blocked_user(for_user) |> filter_blocked_domains(for_user) - |> fts_subquery(query_string) - |> subquery() - |> where([u], u.search_rank > @similarity_threshold) + |> fts_search(query_string) + |> trigram_rank(query_string) |> boost_search_rank(for_user) + |> subquery() |> order_by(desc: :search_rank) |> maybe_restrict_local(for_user) end @nickname_regex ~r/^[a-zA-Z0-9.!#$%&'*+\/=?^_`{|}~\-@]+$/ - defp fts_subquery(query, query_string) do + defp fts_search(query, query_string) do {nickname_weight, name_weight} = if String.match?(query_string, @nickname_regex) do {"A", "B"} @@ -68,19 +67,17 @@ defp fts_subquery(query, query_string) do from( u in query, - select_merge: %{ - search_rank: - fragment( - """ - ts_rank_cd((setweight(to_tsvector('simple', ?), ?) || setweight(to_tsvector('simple', ?), ?)), to_tsquery('simple', ?)) - """, - u.name, - ^name_weight, - u.nickname, - ^nickname_weight, - ^query_string - ) - } + where: + fragment( + """ + (setweight(to_tsvector('simple', ?), ?) || setweight(to_tsvector('simple', ?), ?)) @@ to_tsquery('simple', ?) + """, + u.name, + ^name_weight, + u.nickname, + ^nickname_weight, + ^query_string + ) ) end @@ -93,6 +90,21 @@ defp to_tsquery(query_string) do |> Enum.join(" | ") end + defp trigram_rank(query, query_string) do + from( + u in query, + select_merge: %{ + search_rank: + fragment( + "similarity(?, trim(? || ' ' || coalesce(?, '')))", + ^query_string, + u.nickname, + u.name + ) + } + ) + end + defp base_query(_user, false), do: User defp base_query(user, true), do: User.get_followers_query(user) -- cgit v1.2.3 From eca86330b4b1e834c14c4765469bf7238ffa0018 Mon Sep 17 00:00:00 2001 From: rinpatch Date: Wed, 16 Oct 2019 15:09:39 +0300 Subject: Remove a failing search test due to it being unrealistic While the case tested (`lain@ple`) gives wrong ordering due to using only trigram to order, it almost never happens in reality. In reality it would be either `lain` (in which case it's fine to list the user with `lain` in display name first), or `@lain@pleroma.soykaf.com`/`lain@pleroma.soykaf.com` (which is handled fine as well) --- test/user_search_test.exs | 9 --------- 1 file changed, 9 deletions(-) diff --git a/test/user_search_test.exs b/test/user_search_test.exs index 341074031..6ee624929 100644 --- a/test/user_search_test.exs +++ b/test/user_search_test.exs @@ -65,15 +65,6 @@ test "finds users, considering density of matched tokens" do assert [u2.id, u1.id] == Enum.map(User.search("bar word"), & &1.id) end - test "finds users, ranking by similarity" do - u1 = insert(:user, %{name: "lain"}) - _u2 = insert(:user, %{name: "ean"}) - u3 = insert(:user, %{name: "ebn", nickname: "lain@mastodon.social"}) - u4 = insert(:user, %{nickname: "lain@pleroma.soykaf.com"}) - - assert [u4.id, u3.id, u1.id] == Enum.map(User.search("lain@ple", for_user: u1), & &1.id) - end - test "finds users, boosting ranks of friends and followers" do u1 = insert(:user) u2 = insert(:user, %{name: "Doe"}) -- cgit v1.2.3 From c00b55bc80b8e5cd010d34ceb841d3a12d607056 Mon Sep 17 00:00:00 2001 From: rinpatch Date: Fri, 18 Oct 2019 14:57:13 +0300 Subject: Add a changelog entry for user search refactoring --- CHANGELOG.md | 1 + 1 file changed, 1 insertion(+) diff --git a/CHANGELOG.md b/CHANGELOG.md index 3c282514a..758d53ec4 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -9,6 +9,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/). ``` delete from users where ap_id = 'https://your.instance.hostname/relay'; ``` +- Bad user search matches ## [1.1.0] - 2019-10-14 **Breaking:** The stable branch has been changed from `master` to `stable`. If you want to keep using 1.0, the `release/1.0` branch will receive security updates for 6 months after 1.1 release. -- cgit v1.2.3