[#3213] Speedup of HashtagsTableMigrator (query optimization). State handling fix.

This commit is contained in:
Ivan Tashkinov 2021-01-24 23:27:02 +03:00
parent 218c519606
commit f264d930cc
2 changed files with 19 additions and 3 deletions

View File

@ -72,6 +72,8 @@ def handle_continue(:init_state, _state) do
@impl true @impl true
def handle_info(:migrate_hashtags, state) do def handle_info(:migrate_hashtags, state) do
State.clear()
data_migration = data_migration() data_migration = data_migration()
persistent_data = Map.take(data_migration.data, ["max_processed_id"]) persistent_data = Map.take(data_migration.data, ["max_processed_id"])
@ -152,8 +154,6 @@ defp query do
# Note: most objects have Mention-type AS2 tags and no hashtags (but we can't filter them out) # Note: most objects have Mention-type AS2 tags and no hashtags (but we can't filter them out)
from( from(
object in Object, object in Object,
left_join: hashtag in assoc(object, :hashtags),
where: is_nil(hashtag.id),
where: where:
fragment("(?)->'tag' IS NOT NULL AND (?)->'tag' != '[]'::jsonb", object.data, object.data), fragment("(?)->'tag' IS NOT NULL AND (?)->'tag' != '[]'::jsonb", object.data, object.data),
select: %{ select: %{
@ -161,12 +161,24 @@ defp query do
tag: fragment("(?)->'tag'", object.data) tag: fragment("(?)->'tag'", object.data)
} }
) )
|> join(:left, [o], hashtags_objects in fragment("SELECT object_id FROM hashtags_objects"),
on: hashtags_objects.object_id == o.id
)
|> where([_o, hashtags_objects], is_nil(hashtags_objects.object_id))
end end
defp transfer_object_hashtags(object) do defp transfer_object_hashtags(object) do
embedded_tags = (Map.has_key?(object, :tag) && object.tag) || object.data["tag"] embedded_tags = if Map.has_key?(object, :tag), do: object.tag, else: object.data["tag"]
hashtags = Object.object_data_hashtags(%{"tag" => embedded_tags}) hashtags = Object.object_data_hashtags(%{"tag" => embedded_tags})
if Enum.any?(hashtags) do
transfer_object_hashtags(object, hashtags)
else
{:ok, object.id}
end
end
defp transfer_object_hashtags(object, hashtags) do
Repo.transaction(fn -> Repo.transaction(fn ->
with {:ok, hashtag_records} <- Hashtag.get_or_create_by_names(hashtags) do with {:ok, hashtag_records} <- Hashtag.get_or_create_by_names(hashtags) do
for hashtag_record <- hashtag_records do for hashtag_record <- hashtag_records do

View File

@ -12,6 +12,10 @@ def start_link(_) do
Agent.start_link(fn -> @init_state end, name: @reg_name) Agent.start_link(fn -> @init_state end, name: @reg_name)
end end
def clear do
Agent.update(@reg_name, fn _state -> @init_state end)
end
def get do def get do
Agent.get(@reg_name, & &1) Agent.get(@reg_name, & &1)
end end