2016-01-07 04:26:29 +00:00
|
|
|
# Copyright 2014-2016 OpenMarket Ltd
|
2019-05-30 10:22:59 +00:00
|
|
|
# Copyright 2018-2019 New Vector Ltd
|
|
|
|
# Copyright 2019 The Matrix.org Foundation C.I.C.
|
2015-03-20 13:52:56 +00:00
|
|
|
#
|
|
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
# you may not use this file except in compliance with the License.
|
|
|
|
# You may obtain a copy of the License at
|
|
|
|
#
|
|
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
#
|
|
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
# See the License for the specific language governing permissions and
|
|
|
|
# limitations under the License.
|
2018-04-17 17:30:53 +00:00
|
|
|
import itertools
|
2018-03-29 21:57:28 +00:00
|
|
|
import logging
|
2020-05-13 12:38:22 +00:00
|
|
|
from collections import OrderedDict, namedtuple
|
2021-01-11 16:09:22 +00:00
|
|
|
from typing import (
|
|
|
|
TYPE_CHECKING,
|
|
|
|
Any,
|
|
|
|
Dict,
|
|
|
|
Generator,
|
|
|
|
Iterable,
|
|
|
|
List,
|
|
|
|
Optional,
|
|
|
|
Set,
|
|
|
|
Tuple,
|
|
|
|
)
|
2018-07-09 06:09:20 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
import attr
|
2019-10-23 11:00:21 +00:00
|
|
|
from prometheus_client import Counter
|
2018-06-28 13:49:57 +00:00
|
|
|
|
2018-07-09 06:09:20 +00:00
|
|
|
import synapse.metrics
|
2020-07-07 13:20:40 +00:00
|
|
|
from synapse.api.constants import EventContentFields, EventTypes, RelationTypes
|
2020-01-29 11:01:32 +00:00
|
|
|
from synapse.api.room_versions import RoomVersions
|
2020-05-13 12:38:22 +00:00
|
|
|
from synapse.crypto.event_signing import compute_event_reference_hash
|
2018-07-09 06:09:20 +00:00
|
|
|
from synapse.events import EventBase # noqa: F401
|
|
|
|
from synapse.events.snapshot import EventContext # noqa: F401
|
2019-07-03 14:07:04 +00:00
|
|
|
from synapse.logging.utils import log_function
|
2020-07-16 15:32:19 +00:00
|
|
|
from synapse.storage._base import db_to_json, make_in_list_sql_clause
|
2020-08-05 20:38:57 +00:00
|
|
|
from synapse.storage.database import DatabasePool, LoggingTransaction
|
|
|
|
from synapse.storage.databases.main.search import SearchEntry
|
2021-02-23 12:33:24 +00:00
|
|
|
from synapse.storage.types import Connection
|
2020-09-14 09:16:41 +00:00
|
|
|
from synapse.storage.util.id_generators import MultiWriterIdGenerator
|
2021-02-24 10:13:53 +00:00
|
|
|
from synapse.storage.util.sequence import SequenceGenerator
|
2020-05-13 12:38:22 +00:00
|
|
|
from synapse.types import StateMap, get_domain_from_id
|
2020-10-28 15:51:15 +00:00
|
|
|
from synapse.util import json_encoder
|
2021-01-11 16:09:22 +00:00
|
|
|
from synapse.util.iterutils import batch_iter, sorted_topologically
|
2015-03-20 13:52:56 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
if TYPE_CHECKING:
|
|
|
|
from synapse.server import HomeServer
|
2020-08-05 20:38:57 +00:00
|
|
|
from synapse.storage.databases.main import DataStore
|
2020-05-13 12:38:22 +00:00
|
|
|
|
|
|
|
|
2018-05-22 00:48:57 +00:00
|
|
|
logger = logging.getLogger(__name__)
|
2015-03-20 13:52:56 +00:00
|
|
|
|
2018-05-22 00:48:57 +00:00
|
|
|
persist_event_counter = Counter("synapse_storage_events_persisted_events", "")
|
2019-03-28 13:37:16 +00:00
|
|
|
event_counter = Counter(
|
|
|
|
"synapse_storage_events_persisted_events_sep",
|
|
|
|
"",
|
|
|
|
["type", "origin_type", "origin_entity"],
|
|
|
|
)
|
2018-03-27 12:13:38 +00:00
|
|
|
|
2016-06-06 10:58:09 +00:00
|
|
|
|
2016-06-03 16:55:32 +00:00
|
|
|
_EventCacheEntry = namedtuple("_EventCacheEntry", ("event", "redacted_event"))
|
|
|
|
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
@attr.s(slots=True)
|
|
|
|
class DeltaState:
|
|
|
|
"""Deltas to use to update the `current_state_events` table.
|
2019-05-28 17:52:41 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Attributes:
|
|
|
|
to_delete: List of type/state_keys to delete from current state
|
|
|
|
to_insert: Map of state to upsert into current state
|
|
|
|
no_longer_in_room: The server is not longer in the room, so the room
|
|
|
|
should e.g. be removed from `current_state_events` table.
|
|
|
|
"""
|
2019-06-13 12:40:52 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
to_delete = attr.ib(type=List[Tuple[str, str]])
|
|
|
|
to_insert = attr.ib(type=StateMap[str])
|
|
|
|
no_longer_in_room = attr.ib(type=bool, default=False)
|
2019-06-13 12:40:52 +00:00
|
|
|
|
2019-07-03 08:31:27 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
class PersistEventsStore:
|
|
|
|
"""Contains all the functions for writing events to the database.
|
2019-06-13 12:40:52 +00:00
|
|
|
|
2020-05-13 09:37:11 +00:00
|
|
|
Should only be instantiated on one process (when using a worker mode setup).
|
2020-05-13 12:38:22 +00:00
|
|
|
|
|
|
|
Note: This is not part of the `DataStore` mixin.
|
|
|
|
"""
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
def __init__(
|
2021-02-23 12:33:24 +00:00
|
|
|
self,
|
|
|
|
hs: "HomeServer",
|
|
|
|
db: DatabasePool,
|
|
|
|
main_data_store: "DataStore",
|
|
|
|
db_conn: Connection,
|
2020-08-05 20:38:57 +00:00
|
|
|
):
|
2020-05-13 12:38:22 +00:00
|
|
|
self.hs = hs
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool = db
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store = main_data_store
|
|
|
|
self.database_engine = db.engine
|
|
|
|
self._clock = hs.get_clock()
|
2020-09-14 09:16:41 +00:00
|
|
|
self._instance_name = hs.get_instance_name()
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2019-12-03 19:19:45 +00:00
|
|
|
self._ephemeral_messages_enabled = hs.config.enable_ephemeral_messages
|
2020-01-15 14:59:33 +00:00
|
|
|
self.is_mine_id = hs.is_mine_id
|
2019-12-03 19:19:45 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
# Ideally we'd move these ID gens here, unfortunately some other ID
|
|
|
|
# generators are chained off them so doing so is a bit of a PITA.
|
2020-09-14 09:16:41 +00:00
|
|
|
self._backfill_id_gen = (
|
|
|
|
self.store._backfill_id_gen
|
|
|
|
) # type: MultiWriterIdGenerator
|
|
|
|
self._stream_id_gen = self.store._stream_id_gen # type: MultiWriterIdGenerator
|
2019-06-13 12:40:52 +00:00
|
|
|
|
2020-05-22 15:11:35 +00:00
|
|
|
# This should only exist on instances that are configured to write
|
2020-05-13 12:38:22 +00:00
|
|
|
assert (
|
2020-09-14 09:16:41 +00:00
|
|
|
hs.get_instance_name() in hs.config.worker.writers.events
|
2020-05-22 15:11:35 +00:00
|
|
|
), "Can only instantiate EventsStore on master"
|
2019-06-13 12:40:52 +00:00
|
|
|
|
2020-08-14 14:05:19 +00:00
|
|
|
async def _persist_events_and_state_updates(
|
2019-10-23 11:00:21 +00:00
|
|
|
self,
|
2020-01-20 18:07:20 +00:00
|
|
|
events_and_contexts: List[Tuple[EventBase, EventContext]],
|
|
|
|
current_state_for_room: Dict[str, StateMap[str]],
|
|
|
|
state_delta_for_room: Dict[str, DeltaState],
|
|
|
|
new_forward_extremeties: Dict[str, List[str]],
|
|
|
|
backfilled: bool = False,
|
2020-08-14 14:05:19 +00:00
|
|
|
) -> None:
|
2019-10-30 13:33:38 +00:00
|
|
|
"""Persist a set of events alongside updates to the current state and
|
|
|
|
forward extremities tables.
|
2017-03-17 11:51:13 +00:00
|
|
|
|
|
|
|
Args:
|
2020-01-20 18:07:20 +00:00
|
|
|
events_and_contexts:
|
|
|
|
current_state_for_room: Map from room_id to the current state of
|
|
|
|
the room based on forward extremities
|
|
|
|
state_delta_for_room: Map from room_id to the delta to apply to
|
|
|
|
room state
|
|
|
|
new_forward_extremities: Map from room_id to list of event IDs
|
|
|
|
that are the new forward extremities of the room.
|
|
|
|
backfilled
|
2017-03-17 11:51:13 +00:00
|
|
|
|
|
|
|
Returns:
|
2020-08-14 14:05:19 +00:00
|
|
|
Resolves when the events have been persisted
|
2017-03-17 11:51:13 +00:00
|
|
|
"""
|
2015-06-25 16:18:19 +00:00
|
|
|
|
2019-10-23 11:00:21 +00:00
|
|
|
# We want to calculate the stream orderings as late as possible, as
|
|
|
|
# we only notify after all events with a lesser stream ordering have
|
|
|
|
# been persisted. I.e. if we spend 10s inside the with block then
|
|
|
|
# that will delay all subsequent events from being notified about.
|
|
|
|
# Hence why we do it down here rather than wrapping the entire
|
|
|
|
# function.
|
|
|
|
#
|
|
|
|
# Its safe to do this after calculating the state deltas etc as we
|
|
|
|
# only need to protect the *persistence* of the events. This is to
|
|
|
|
# ensure that queries of the form "fetch events since X" don't
|
|
|
|
# return events and stream positions after events that are still in
|
|
|
|
# flight, as otherwise subsequent requests "fetch event since Y"
|
|
|
|
# will not return those events.
|
|
|
|
#
|
|
|
|
# Note: Multiple instances of this function cannot be in flight at
|
|
|
|
# the same time for the same room.
|
|
|
|
if backfilled:
|
2020-09-23 15:11:18 +00:00
|
|
|
stream_ordering_manager = self._backfill_id_gen.get_next_mult(
|
2019-10-23 11:00:21 +00:00
|
|
|
len(events_and_contexts)
|
|
|
|
)
|
|
|
|
else:
|
2020-09-23 15:11:18 +00:00
|
|
|
stream_ordering_manager = self._stream_id_gen.get_next_mult(
|
2019-10-23 11:00:21 +00:00
|
|
|
len(events_and_contexts)
|
|
|
|
)
|
2019-08-06 12:27:22 +00:00
|
|
|
|
2020-09-23 15:11:18 +00:00
|
|
|
async with stream_ordering_manager as stream_orderings:
|
2019-10-23 11:00:21 +00:00
|
|
|
for (event, context), stream in zip(events_and_contexts, stream_orderings):
|
|
|
|
event.internal_metadata.stream_ordering = stream
|
2019-08-06 12:27:22 +00:00
|
|
|
|
2020-08-14 14:05:19 +00:00
|
|
|
await self.db_pool.runInteraction(
|
2019-10-23 11:00:21 +00:00
|
|
|
"persist_events",
|
|
|
|
self._persist_events_txn,
|
|
|
|
events_and_contexts=events_and_contexts,
|
|
|
|
backfilled=backfilled,
|
|
|
|
state_delta_for_room=state_delta_for_room,
|
|
|
|
new_forward_extremeties=new_forward_extremeties,
|
|
|
|
)
|
|
|
|
persist_event_counter.inc(len(events_and_contexts))
|
2019-08-06 12:27:22 +00:00
|
|
|
|
|
|
|
if not backfilled:
|
2019-10-23 11:00:21 +00:00
|
|
|
# backfilled events have negative stream orderings, so we don't
|
|
|
|
# want to set the event_persisted_position to that.
|
|
|
|
synapse.metrics.event_persisted_position.set(
|
|
|
|
events_and_contexts[-1][0].internal_metadata.stream_ordering
|
2016-08-30 15:54:40 +00:00
|
|
|
)
|
2017-01-20 14:28:53 +00:00
|
|
|
|
2019-10-23 11:00:21 +00:00
|
|
|
for event, context in events_and_contexts:
|
|
|
|
if context.app_service:
|
|
|
|
origin_type = "local"
|
|
|
|
origin_entity = context.app_service.id
|
|
|
|
elif self.hs.is_mine_id(event.sender):
|
|
|
|
origin_type = "local"
|
|
|
|
origin_entity = "*client*"
|
|
|
|
else:
|
|
|
|
origin_type = "remote"
|
|
|
|
origin_entity = get_domain_from_id(event.sender)
|
|
|
|
|
|
|
|
event_counter.labels(event.type, origin_type, origin_entity).inc()
|
|
|
|
|
2020-06-15 11:03:36 +00:00
|
|
|
for room_id, new_state in current_state_for_room.items():
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store.get_current_state_ids.prefill((room_id,), new_state)
|
2019-10-23 11:00:21 +00:00
|
|
|
|
2020-06-15 11:03:36 +00:00
|
|
|
for room_id, latest_event_ids in new_forward_extremeties.items():
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store.get_latest_event_ids_in_room.prefill(
|
2019-10-23 11:00:21 +00:00
|
|
|
(room_id,), list(latest_event_ids)
|
|
|
|
)
|
2017-01-20 14:28:53 +00:00
|
|
|
|
2020-08-14 14:05:19 +00:00
|
|
|
async def _get_events_which_are_prevs(self, event_ids: Iterable[str]) -> List[str]:
|
2018-10-02 22:33:29 +00:00
|
|
|
"""Filter the supplied list of event_ids to get those which are prev_events of
|
2018-10-03 09:19:41 +00:00
|
|
|
existing (non-outlier/rejected) events.
|
2018-10-02 22:33:29 +00:00
|
|
|
|
|
|
|
Args:
|
2020-08-14 14:05:19 +00:00
|
|
|
event_ids: event ids to filter
|
2018-10-02 22:33:29 +00:00
|
|
|
|
|
|
|
Returns:
|
2020-08-14 14:05:19 +00:00
|
|
|
Filtered event ids
|
2018-10-02 22:33:29 +00:00
|
|
|
"""
|
2020-09-11 11:22:55 +00:00
|
|
|
results = [] # type: List[str]
|
2018-10-02 22:33:29 +00:00
|
|
|
|
2019-05-29 10:56:24 +00:00
|
|
|
def _get_events_which_are_prevs_txn(txn, batch):
|
2018-10-02 22:33:29 +00:00
|
|
|
sql = """
|
2019-05-21 15:10:54 +00:00
|
|
|
SELECT prev_event_id, internal_metadata
|
2018-10-02 22:33:29 +00:00
|
|
|
FROM event_edges
|
|
|
|
INNER JOIN events USING (event_id)
|
|
|
|
LEFT JOIN rejections USING (event_id)
|
2019-05-21 15:10:54 +00:00
|
|
|
LEFT JOIN event_json USING (event_id)
|
2018-10-02 22:33:29 +00:00
|
|
|
WHERE
|
2019-10-02 18:07:07 +00:00
|
|
|
NOT events.outlier
|
2018-10-02 22:33:29 +00:00
|
|
|
AND rejections.event_id IS NULL
|
2019-10-02 18:07:07 +00:00
|
|
|
AND
|
|
|
|
"""
|
|
|
|
|
|
|
|
clause, args = make_in_list_sql_clause(
|
|
|
|
self.database_engine, "prev_event_id", batch
|
2018-10-02 22:33:29 +00:00
|
|
|
)
|
2017-01-20 14:28:53 +00:00
|
|
|
|
2019-10-02 18:07:07 +00:00
|
|
|
txn.execute(sql + clause, args)
|
2020-07-16 15:32:19 +00:00
|
|
|
results.extend(r[0] for r in txn if not db_to_json(r[1]).get("soft_failed"))
|
2018-10-02 22:33:29 +00:00
|
|
|
|
|
|
|
for chunk in batch_iter(event_ids, 100):
|
2020-08-14 14:05:19 +00:00
|
|
|
await self.db_pool.runInteraction(
|
2019-06-13 12:40:52 +00:00
|
|
|
"_get_events_which_are_prevs", _get_events_which_are_prevs_txn, chunk
|
2019-05-29 10:56:24 +00:00
|
|
|
)
|
2018-10-02 22:33:29 +00:00
|
|
|
|
2019-07-23 13:00:55 +00:00
|
|
|
return results
|
2017-01-20 14:28:53 +00:00
|
|
|
|
2020-08-14 14:05:19 +00:00
|
|
|
async def _get_prevs_before_rejected(self, event_ids: Iterable[str]) -> Set[str]:
|
2019-05-29 10:56:24 +00:00
|
|
|
"""Get soft-failed ancestors to remove from the extremities.
|
|
|
|
|
|
|
|
Given a set of events, find all those that have been soft-failed or
|
|
|
|
rejected. Returns those soft failed/rejected events and their prev
|
|
|
|
events (whether soft-failed/rejected or not), and recurses up the
|
|
|
|
prev-event graph until it finds no more soft-failed/rejected events.
|
|
|
|
|
|
|
|
This is used to find extremities that are ancestors of new events, but
|
|
|
|
are separated by soft failed events.
|
|
|
|
|
|
|
|
Args:
|
2020-08-14 14:05:19 +00:00
|
|
|
event_ids: Events to find prev events for. Note that these must have
|
|
|
|
already been persisted.
|
2019-05-29 10:56:24 +00:00
|
|
|
|
|
|
|
Returns:
|
2020-08-14 14:05:19 +00:00
|
|
|
The previous events.
|
2019-05-29 10:56:24 +00:00
|
|
|
"""
|
|
|
|
|
|
|
|
# The set of event_ids to return. This includes all soft-failed events
|
|
|
|
# and their prev events.
|
|
|
|
existing_prevs = set()
|
|
|
|
|
|
|
|
def _get_prevs_before_rejected_txn(txn, batch):
|
|
|
|
to_recursively_check = batch
|
|
|
|
|
|
|
|
while to_recursively_check:
|
|
|
|
sql = """
|
|
|
|
SELECT
|
|
|
|
event_id, prev_event_id, internal_metadata,
|
|
|
|
rejections.event_id IS NOT NULL
|
|
|
|
FROM event_edges
|
|
|
|
INNER JOIN events USING (event_id)
|
|
|
|
LEFT JOIN rejections USING (event_id)
|
|
|
|
LEFT JOIN event_json USING (event_id)
|
|
|
|
WHERE
|
2019-10-02 18:07:07 +00:00
|
|
|
NOT events.outlier
|
|
|
|
AND
|
|
|
|
"""
|
|
|
|
|
|
|
|
clause, args = make_in_list_sql_clause(
|
|
|
|
self.database_engine, "event_id", to_recursively_check
|
2019-05-29 10:56:24 +00:00
|
|
|
)
|
|
|
|
|
2019-10-02 18:07:07 +00:00
|
|
|
txn.execute(sql + clause, args)
|
2019-05-29 10:56:24 +00:00
|
|
|
to_recursively_check = []
|
|
|
|
|
|
|
|
for event_id, prev_event_id, metadata, rejected in txn:
|
|
|
|
if prev_event_id in existing_prevs:
|
|
|
|
continue
|
|
|
|
|
2020-07-16 15:32:19 +00:00
|
|
|
soft_failed = db_to_json(metadata).get("soft_failed")
|
2019-05-29 10:56:24 +00:00
|
|
|
if soft_failed or rejected:
|
|
|
|
to_recursively_check.append(prev_event_id)
|
|
|
|
existing_prevs.add(prev_event_id)
|
|
|
|
|
|
|
|
for chunk in batch_iter(event_ids, 100):
|
2020-08-14 14:05:19 +00:00
|
|
|
await self.db_pool.runInteraction(
|
2019-06-13 12:40:52 +00:00
|
|
|
"_get_prevs_before_rejected", _get_prevs_before_rejected_txn, chunk
|
2019-05-29 10:56:24 +00:00
|
|
|
)
|
|
|
|
|
2019-07-23 13:00:55 +00:00
|
|
|
return existing_prevs
|
2019-05-29 10:56:24 +00:00
|
|
|
|
2015-06-25 16:18:19 +00:00
|
|
|
@log_function
|
2019-03-28 13:37:16 +00:00
|
|
|
def _persist_events_txn(
|
|
|
|
self,
|
2020-01-20 18:07:20 +00:00
|
|
|
txn: LoggingTransaction,
|
|
|
|
events_and_contexts: List[Tuple[EventBase, EventContext]],
|
|
|
|
backfilled: bool,
|
2021-04-08 21:38:54 +00:00
|
|
|
state_delta_for_room: Optional[Dict[str, DeltaState]] = None,
|
|
|
|
new_forward_extremeties: Optional[Dict[str, List[str]]] = None,
|
2019-03-28 13:37:16 +00:00
|
|
|
):
|
2016-07-26 09:49:52 +00:00
|
|
|
"""Insert some number of room events into the necessary database tables.
|
|
|
|
|
|
|
|
Rejected events are only inserted into the events table, the events_json table,
|
|
|
|
and the rejections table. Things reading from those table will need to check
|
|
|
|
whether the event was rejected.
|
2016-08-04 14:02:15 +00:00
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
Args:
|
2020-01-20 18:07:20 +00:00
|
|
|
txn
|
|
|
|
events_and_contexts: events to persist
|
|
|
|
backfilled: True if the events were backfilled
|
|
|
|
delete_existing True to purge existing table rows for the events
|
|
|
|
from the database. This is useful when retrying due to
|
2017-03-17 11:51:13 +00:00
|
|
|
IntegrityError.
|
2020-01-20 18:07:20 +00:00
|
|
|
state_delta_for_room: The current-state delta for each room.
|
|
|
|
new_forward_extremetie: The new forward extremities for each room.
|
|
|
|
For each room, a list of the event ids which are the forward
|
|
|
|
extremities.
|
2017-03-17 11:51:13 +00:00
|
|
|
|
2016-07-26 09:49:52 +00:00
|
|
|
"""
|
2021-04-08 21:38:54 +00:00
|
|
|
state_delta_for_room = state_delta_for_room or {}
|
|
|
|
new_forward_extremeties = new_forward_extremeties or {}
|
|
|
|
|
2018-02-20 12:33:04 +00:00
|
|
|
all_events_and_contexts = events_and_contexts
|
|
|
|
|
2019-04-02 11:42:39 +00:00
|
|
|
min_stream_order = events_and_contexts[0][0].internal_metadata.stream_ordering
|
2017-01-20 14:28:53 +00:00
|
|
|
max_stream_order = events_and_contexts[-1][0].internal_metadata.stream_ordering
|
2017-05-30 13:41:42 +00:00
|
|
|
|
2020-10-05 13:43:14 +00:00
|
|
|
# stream orderings should have been assigned by now
|
|
|
|
assert min_stream_order
|
|
|
|
assert max_stream_order
|
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
self._update_forward_extremities_txn(
|
|
|
|
txn,
|
|
|
|
new_forward_extremities=new_forward_extremeties,
|
|
|
|
max_stream_order=max_stream_order,
|
|
|
|
)
|
|
|
|
|
|
|
|
# Ensure that we don't have the same event twice.
|
|
|
|
events_and_contexts = self._filter_events_and_contexts_for_duplicates(
|
2019-03-28 13:37:16 +00:00
|
|
|
events_and_contexts
|
2017-03-17 11:51:13 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
self._update_room_depths_txn(
|
2019-03-28 13:37:16 +00:00
|
|
|
txn, events_and_contexts=events_and_contexts, backfilled=backfilled
|
2017-03-17 11:51:13 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
# _update_outliers_txn filters out any events which have already been
|
|
|
|
# persisted, and returns the filtered list.
|
|
|
|
events_and_contexts = self._update_outliers_txn(
|
2019-03-28 13:37:16 +00:00
|
|
|
txn, events_and_contexts=events_and_contexts
|
2017-03-17 11:51:13 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
# From this point onwards the events are only events that we haven't
|
|
|
|
# seen before.
|
|
|
|
|
2019-03-28 13:37:16 +00:00
|
|
|
self._store_event_txn(txn, events_and_contexts=events_and_contexts)
|
2017-03-17 11:51:13 +00:00
|
|
|
|
2020-10-13 11:07:56 +00:00
|
|
|
self._persist_transaction_ids_txn(txn, events_and_contexts)
|
|
|
|
|
2018-02-06 14:31:24 +00:00
|
|
|
# Insert into event_to_state_groups.
|
|
|
|
self._store_event_state_mappings_txn(txn, events_and_contexts)
|
2017-03-17 14:30:16 +00:00
|
|
|
|
2021-01-11 16:09:22 +00:00
|
|
|
self._persist_event_auth_chain_txn(txn, [e for e, _ in events_and_contexts])
|
|
|
|
|
|
|
|
# _store_rejected_events_txn filters out any events which were
|
|
|
|
# rejected, and returns the filtered list.
|
|
|
|
events_and_contexts = self._store_rejected_events_txn(
|
|
|
|
txn, events_and_contexts=events_and_contexts
|
|
|
|
)
|
|
|
|
|
|
|
|
# From this point onwards the events are only ones that weren't
|
|
|
|
# rejected.
|
|
|
|
|
|
|
|
self._update_metadata_tables_txn(
|
|
|
|
txn,
|
|
|
|
events_and_contexts=events_and_contexts,
|
|
|
|
all_events_and_contexts=all_events_and_contexts,
|
|
|
|
backfilled=backfilled,
|
|
|
|
)
|
|
|
|
|
|
|
|
# We call this last as it assumes we've inserted the events into
|
|
|
|
# room_memberships, where applicable.
|
|
|
|
self._update_current_state_txn(txn, state_delta_for_room, min_stream_order)
|
|
|
|
|
|
|
|
def _persist_event_auth_chain_txn(
|
|
|
|
self,
|
|
|
|
txn: LoggingTransaction,
|
|
|
|
events: List[EventBase],
|
|
|
|
) -> None:
|
|
|
|
|
|
|
|
# We only care about state events, so this if there are no state events.
|
|
|
|
if not any(e.is_state() for e in events):
|
|
|
|
return
|
|
|
|
|
2018-10-16 13:01:53 +00:00
|
|
|
# We want to store event_auth mappings for rejected events, as they're
|
|
|
|
# used in state res v2.
|
|
|
|
# This is only necessary if the rejected event appears in an accepted
|
|
|
|
# event's auth chain, but its easier for now just to store them (and
|
|
|
|
# it doesn't take much storage compared to storing the entire event
|
|
|
|
# anyway).
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2018-10-16 13:01:53 +00:00
|
|
|
txn,
|
|
|
|
table="event_auth",
|
|
|
|
values=[
|
|
|
|
{
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"room_id": event.room_id,
|
|
|
|
"auth_id": auth_id,
|
|
|
|
}
|
2021-01-11 16:09:22 +00:00
|
|
|
for event in events
|
2018-11-05 13:35:15 +00:00
|
|
|
for auth_id in event.auth_event_ids()
|
2018-10-16 13:01:53 +00:00
|
|
|
if event.is_state()
|
|
|
|
],
|
|
|
|
)
|
|
|
|
|
2021-01-11 16:09:22 +00:00
|
|
|
# We now calculate chain ID/sequence numbers for any state events we're
|
|
|
|
# persisting. We ignore out of band memberships as we're not in the room
|
|
|
|
# and won't have their auth chain (we'll fix it up later if we join the
|
|
|
|
# room).
|
|
|
|
#
|
|
|
|
# See: docs/auth_chain_difference_algorithm.md
|
|
|
|
|
|
|
|
# We ignore legacy rooms that we aren't filling the chain cover index
|
|
|
|
# for.
|
|
|
|
rows = self.db_pool.simple_select_many_txn(
|
|
|
|
txn,
|
|
|
|
table="rooms",
|
|
|
|
column="room_id",
|
|
|
|
iterable={event.room_id for event in events if event.is_state()},
|
|
|
|
keyvalues={},
|
|
|
|
retcols=("room_id", "has_auth_chain_index"),
|
2017-03-17 11:51:13 +00:00
|
|
|
)
|
2021-01-11 16:09:22 +00:00
|
|
|
rooms_using_chain_index = {
|
|
|
|
row["room_id"] for row in rows if row["has_auth_chain_index"]
|
|
|
|
}
|
2017-03-17 11:51:13 +00:00
|
|
|
|
2021-01-11 16:09:22 +00:00
|
|
|
state_events = {
|
|
|
|
event.event_id: event
|
|
|
|
for event in events
|
|
|
|
if event.is_state() and event.room_id in rooms_using_chain_index
|
|
|
|
}
|
2017-03-17 11:51:13 +00:00
|
|
|
|
2021-01-11 16:09:22 +00:00
|
|
|
if not state_events:
|
|
|
|
return
|
|
|
|
|
|
|
|
# We need to know the type/state_key and auth events of the events we're
|
|
|
|
# calculating chain IDs for. We don't rely on having the full Event
|
|
|
|
# instances as we'll potentially be pulling more events from the DB and
|
|
|
|
# we don't need the overhead of fetching/parsing the full event JSON.
|
|
|
|
event_to_types = {
|
|
|
|
e.event_id: (e.type, e.state_key) for e in state_events.values()
|
|
|
|
}
|
|
|
|
event_to_auth_chain = {
|
|
|
|
e.event_id: e.auth_event_ids() for e in state_events.values()
|
|
|
|
}
|
2021-01-14 15:18:27 +00:00
|
|
|
event_to_room_id = {e.event_id: e.room_id for e in state_events.values()}
|
|
|
|
|
|
|
|
self._add_chain_cover_index(
|
2021-01-14 17:19:35 +00:00
|
|
|
txn,
|
|
|
|
self.db_pool,
|
2021-02-24 10:13:53 +00:00
|
|
|
self.store.event_chain_id_gen,
|
2021-01-14 17:19:35 +00:00
|
|
|
event_to_room_id,
|
|
|
|
event_to_types,
|
|
|
|
event_to_auth_chain,
|
2021-01-14 15:18:27 +00:00
|
|
|
)
|
|
|
|
|
2021-01-21 17:00:12 +00:00
|
|
|
@classmethod
|
2021-01-14 15:18:27 +00:00
|
|
|
def _add_chain_cover_index(
|
2021-01-21 17:00:12 +00:00
|
|
|
cls,
|
2021-01-14 15:18:27 +00:00
|
|
|
txn,
|
2021-01-14 17:19:35 +00:00
|
|
|
db_pool: DatabasePool,
|
2021-02-24 10:13:53 +00:00
|
|
|
event_chain_id_gen: SequenceGenerator,
|
2021-01-14 15:18:27 +00:00
|
|
|
event_to_room_id: Dict[str, str],
|
|
|
|
event_to_types: Dict[str, Tuple[str, str]],
|
|
|
|
event_to_auth_chain: Dict[str, List[str]],
|
|
|
|
) -> None:
|
|
|
|
"""Calculate the chain cover index for the given events.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
event_to_room_id: Event ID to the room ID of the event
|
|
|
|
event_to_types: Event ID to type and state_key of the event
|
|
|
|
event_to_auth_chain: Event ID to list of auth event IDs of the
|
|
|
|
event (events with no auth events can be excluded).
|
|
|
|
"""
|
|
|
|
|
|
|
|
# Map from event ID to chain ID/sequence number.
|
|
|
|
chain_map = {} # type: Dict[str, Tuple[int, int]]
|
2021-01-11 16:09:22 +00:00
|
|
|
|
|
|
|
# Set of event IDs to calculate chain ID/seq numbers for.
|
2021-01-14 15:18:27 +00:00
|
|
|
events_to_calc_chain_id_for = set(event_to_room_id)
|
2021-01-11 16:09:22 +00:00
|
|
|
|
|
|
|
# We check if there are any events that need to be handled in the rooms
|
|
|
|
# we're looking at. These should just be out of band memberships, where
|
|
|
|
# we didn't have the auth chain when we first persisted.
|
2021-01-14 17:19:35 +00:00
|
|
|
rows = db_pool.simple_select_many_txn(
|
2017-03-17 11:51:13 +00:00
|
|
|
txn,
|
2021-01-11 16:09:22 +00:00
|
|
|
table="event_auth_chain_to_calculate",
|
|
|
|
keyvalues={},
|
|
|
|
column="room_id",
|
2021-01-14 15:18:27 +00:00
|
|
|
iterable=set(event_to_room_id.values()),
|
2021-01-11 16:09:22 +00:00
|
|
|
retcols=("event_id", "type", "state_key"),
|
2017-03-17 11:51:13 +00:00
|
|
|
)
|
2021-01-11 16:09:22 +00:00
|
|
|
for row in rows:
|
|
|
|
event_id = row["event_id"]
|
|
|
|
event_type = row["type"]
|
|
|
|
state_key = row["state_key"]
|
|
|
|
|
|
|
|
# (We could pull out the auth events for all rows at once using
|
|
|
|
# simple_select_many, but this case happens rarely and almost always
|
|
|
|
# with a single row.)
|
2021-01-14 17:19:35 +00:00
|
|
|
auth_events = db_pool.simple_select_onecol_txn(
|
2021-01-11 16:09:22 +00:00
|
|
|
txn,
|
|
|
|
"event_auth",
|
|
|
|
keyvalues={"event_id": event_id},
|
|
|
|
retcol="auth_id",
|
|
|
|
)
|
2017-03-17 11:51:13 +00:00
|
|
|
|
2021-01-11 16:09:22 +00:00
|
|
|
events_to_calc_chain_id_for.add(event_id)
|
|
|
|
event_to_types[event_id] = (event_type, state_key)
|
|
|
|
event_to_auth_chain[event_id] = auth_events
|
|
|
|
|
|
|
|
# First we get the chain ID and sequence numbers for the events'
|
|
|
|
# auth events (that aren't also currently being persisted).
|
|
|
|
#
|
|
|
|
# Note that there there is an edge case here where we might not have
|
|
|
|
# calculated chains and sequence numbers for events that were "out
|
|
|
|
# of band". We handle this case by fetching the necessary info and
|
|
|
|
# adding it to the set of events to calculate chain IDs for.
|
|
|
|
|
|
|
|
missing_auth_chains = {
|
|
|
|
a_id
|
|
|
|
for auth_events in event_to_auth_chain.values()
|
|
|
|
for a_id in auth_events
|
|
|
|
if a_id not in events_to_calc_chain_id_for
|
|
|
|
}
|
|
|
|
|
|
|
|
# We loop here in case we find an out of band membership and need to
|
|
|
|
# fetch their auth event info.
|
|
|
|
while missing_auth_chains:
|
|
|
|
sql = """
|
|
|
|
SELECT event_id, events.type, state_key, chain_id, sequence_number
|
|
|
|
FROM events
|
|
|
|
INNER JOIN state_events USING (event_id)
|
|
|
|
LEFT JOIN event_auth_chains USING (event_id)
|
|
|
|
WHERE
|
|
|
|
"""
|
|
|
|
clause, args = make_in_list_sql_clause(
|
|
|
|
txn.database_engine,
|
|
|
|
"event_id",
|
|
|
|
missing_auth_chains,
|
|
|
|
)
|
|
|
|
txn.execute(sql + clause, args)
|
|
|
|
|
|
|
|
missing_auth_chains.clear()
|
|
|
|
|
|
|
|
for auth_id, event_type, state_key, chain_id, sequence_number in txn:
|
|
|
|
event_to_types[auth_id] = (event_type, state_key)
|
|
|
|
|
|
|
|
if chain_id is None:
|
|
|
|
# No chain ID, so the event was persisted out of band.
|
|
|
|
# We add to list of events to calculate auth chains for.
|
|
|
|
|
|
|
|
events_to_calc_chain_id_for.add(auth_id)
|
|
|
|
|
2021-01-14 17:19:35 +00:00
|
|
|
event_to_auth_chain[auth_id] = db_pool.simple_select_onecol_txn(
|
2021-01-11 16:09:22 +00:00
|
|
|
txn,
|
|
|
|
"event_auth",
|
|
|
|
keyvalues={"event_id": auth_id},
|
|
|
|
retcol="auth_id",
|
|
|
|
)
|
|
|
|
|
|
|
|
missing_auth_chains.update(
|
|
|
|
e
|
|
|
|
for e in event_to_auth_chain[auth_id]
|
|
|
|
if e not in event_to_types
|
|
|
|
)
|
|
|
|
else:
|
|
|
|
chain_map[auth_id] = (chain_id, sequence_number)
|
|
|
|
|
|
|
|
# Now we check if we have any events where we don't have auth chain,
|
|
|
|
# this should only be out of band memberships.
|
|
|
|
for event_id in sorted_topologically(event_to_auth_chain, event_to_auth_chain):
|
|
|
|
for auth_id in event_to_auth_chain[event_id]:
|
|
|
|
if (
|
|
|
|
auth_id not in chain_map
|
|
|
|
and auth_id not in events_to_calc_chain_id_for
|
|
|
|
):
|
|
|
|
events_to_calc_chain_id_for.discard(event_id)
|
|
|
|
|
|
|
|
# If this is an event we're trying to persist we add it to
|
|
|
|
# the list of events to calculate chain IDs for next time
|
|
|
|
# around. (Otherwise we will have already added it to the
|
|
|
|
# table).
|
2021-01-14 15:18:27 +00:00
|
|
|
room_id = event_to_room_id.get(event_id)
|
|
|
|
if room_id:
|
|
|
|
e_type, state_key = event_to_types[event_id]
|
2021-01-14 17:19:35 +00:00
|
|
|
db_pool.simple_insert_txn(
|
2021-01-11 16:09:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_auth_chain_to_calculate",
|
|
|
|
values={
|
2021-01-14 15:18:27 +00:00
|
|
|
"event_id": event_id,
|
|
|
|
"room_id": room_id,
|
|
|
|
"type": e_type,
|
|
|
|
"state_key": state_key,
|
2021-01-11 16:09:22 +00:00
|
|
|
},
|
|
|
|
)
|
|
|
|
|
|
|
|
# We stop checking the event's auth events since we've
|
|
|
|
# discarded it.
|
|
|
|
break
|
|
|
|
|
|
|
|
if not events_to_calc_chain_id_for:
|
|
|
|
return
|
|
|
|
|
2021-01-21 17:00:12 +00:00
|
|
|
# Allocate chain ID/sequence numbers to each new event.
|
|
|
|
new_chain_tuples = cls._allocate_chain_ids(
|
|
|
|
txn,
|
|
|
|
db_pool,
|
2021-02-24 10:13:53 +00:00
|
|
|
event_chain_id_gen,
|
2021-01-21 17:00:12 +00:00
|
|
|
event_to_room_id,
|
|
|
|
event_to_types,
|
|
|
|
event_to_auth_chain,
|
|
|
|
events_to_calc_chain_id_for,
|
|
|
|
chain_map,
|
|
|
|
)
|
|
|
|
chain_map.update(new_chain_tuples)
|
2021-01-11 16:09:22 +00:00
|
|
|
|
2021-01-14 17:19:35 +00:00
|
|
|
db_pool.simple_insert_many_txn(
|
2021-01-11 16:09:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_auth_chains",
|
|
|
|
values=[
|
|
|
|
{"event_id": event_id, "chain_id": c_id, "sequence_number": seq}
|
|
|
|
for event_id, (c_id, seq) in new_chain_tuples.items()
|
|
|
|
],
|
|
|
|
)
|
|
|
|
|
2021-01-14 17:19:35 +00:00
|
|
|
db_pool.simple_delete_many_txn(
|
2021-01-11 16:09:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_auth_chain_to_calculate",
|
|
|
|
keyvalues={},
|
|
|
|
column="event_id",
|
|
|
|
iterable=new_chain_tuples,
|
|
|
|
)
|
|
|
|
|
|
|
|
# Now we need to calculate any new links between chains caused by
|
|
|
|
# the new events.
|
|
|
|
#
|
|
|
|
# Links are pairs of chain ID/sequence numbers such that for any
|
|
|
|
# event A (CA, SA) and any event B (CB, SB), B is in A's auth chain
|
|
|
|
# if and only if there is at least one link (CA, S1) -> (CB, S2)
|
|
|
|
# where SA >= S1 and S2 >= SB.
|
|
|
|
#
|
|
|
|
# We try and avoid adding redundant links to the table, e.g. if we
|
|
|
|
# have two links between two chains which both start/end at the
|
|
|
|
# sequence number event (or cross) then one can be safely dropped.
|
|
|
|
#
|
|
|
|
# To calculate new links we look at every new event and:
|
|
|
|
# 1. Fetch the chain ID/sequence numbers of its auth events,
|
|
|
|
# discarding any that are reachable by other auth events, or
|
|
|
|
# that have the same chain ID as the event.
|
|
|
|
# 2. For each retained auth event we:
|
|
|
|
# a. Add a link from the event's to the auth event's chain
|
|
|
|
# ID/sequence number; and
|
|
|
|
# b. Add a link from the event to every chain reachable by the
|
|
|
|
# auth event.
|
|
|
|
|
|
|
|
# Step 1, fetch all existing links from all the chains we've seen
|
|
|
|
# referenced.
|
|
|
|
chain_links = _LinkMap()
|
2021-01-14 17:19:35 +00:00
|
|
|
rows = db_pool.simple_select_many_txn(
|
2021-01-11 16:09:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_auth_chain_links",
|
|
|
|
column="origin_chain_id",
|
|
|
|
iterable={chain_id for chain_id, _ in chain_map.values()},
|
|
|
|
keyvalues={},
|
|
|
|
retcols=(
|
|
|
|
"origin_chain_id",
|
|
|
|
"origin_sequence_number",
|
|
|
|
"target_chain_id",
|
|
|
|
"target_sequence_number",
|
|
|
|
),
|
|
|
|
)
|
|
|
|
for row in rows:
|
|
|
|
chain_links.add_link(
|
|
|
|
(row["origin_chain_id"], row["origin_sequence_number"]),
|
|
|
|
(row["target_chain_id"], row["target_sequence_number"]),
|
|
|
|
new=False,
|
|
|
|
)
|
|
|
|
|
|
|
|
# We do this in toplogical order to avoid adding redundant links.
|
|
|
|
for event_id in sorted_topologically(
|
|
|
|
events_to_calc_chain_id_for, event_to_auth_chain
|
|
|
|
):
|
|
|
|
chain_id, sequence_number = chain_map[event_id]
|
|
|
|
|
|
|
|
# Filter out auth events that are reachable by other auth
|
|
|
|
# events. We do this by looking at every permutation of pairs of
|
|
|
|
# auth events (A, B) to check if B is reachable from A.
|
|
|
|
reduction = {
|
|
|
|
a_id
|
2021-01-14 15:18:27 +00:00
|
|
|
for a_id in event_to_auth_chain.get(event_id, [])
|
2021-01-11 16:09:22 +00:00
|
|
|
if chain_map[a_id][0] != chain_id
|
|
|
|
}
|
|
|
|
for start_auth_id, end_auth_id in itertools.permutations(
|
2021-01-14 15:18:27 +00:00
|
|
|
event_to_auth_chain.get(event_id, []),
|
|
|
|
r=2,
|
2021-01-11 16:09:22 +00:00
|
|
|
):
|
|
|
|
if chain_links.exists_path_from(
|
|
|
|
chain_map[start_auth_id], chain_map[end_auth_id]
|
|
|
|
):
|
|
|
|
reduction.discard(end_auth_id)
|
|
|
|
|
|
|
|
# Step 2, figure out what the new links are from the reduced
|
|
|
|
# list of auth events.
|
|
|
|
for auth_id in reduction:
|
|
|
|
auth_chain_id, auth_sequence_number = chain_map[auth_id]
|
|
|
|
|
|
|
|
# Step 2a, add link between the event and auth event
|
|
|
|
chain_links.add_link(
|
|
|
|
(chain_id, sequence_number), (auth_chain_id, auth_sequence_number)
|
|
|
|
)
|
|
|
|
|
|
|
|
# Step 2b, add a link to chains reachable from the auth
|
|
|
|
# event.
|
|
|
|
for target_id, target_seq in chain_links.get_links_from(
|
|
|
|
(auth_chain_id, auth_sequence_number)
|
|
|
|
):
|
|
|
|
if target_id == chain_id:
|
|
|
|
continue
|
|
|
|
|
|
|
|
chain_links.add_link(
|
|
|
|
(chain_id, sequence_number), (target_id, target_seq)
|
|
|
|
)
|
|
|
|
|
2021-01-14 17:19:35 +00:00
|
|
|
db_pool.simple_insert_many_txn(
|
2021-01-11 16:09:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_auth_chain_links",
|
|
|
|
values=[
|
|
|
|
{
|
|
|
|
"origin_chain_id": source_id,
|
|
|
|
"origin_sequence_number": source_seq,
|
|
|
|
"target_chain_id": target_id,
|
|
|
|
"target_sequence_number": target_seq,
|
|
|
|
}
|
|
|
|
for (
|
|
|
|
source_id,
|
|
|
|
source_seq,
|
|
|
|
target_id,
|
|
|
|
target_seq,
|
|
|
|
) in chain_links.get_additions()
|
|
|
|
],
|
|
|
|
)
|
2019-07-17 14:33:37 +00:00
|
|
|
|
2021-01-21 17:00:12 +00:00
|
|
|
@staticmethod
|
|
|
|
def _allocate_chain_ids(
|
|
|
|
txn,
|
|
|
|
db_pool: DatabasePool,
|
2021-02-24 10:13:53 +00:00
|
|
|
event_chain_id_gen: SequenceGenerator,
|
2021-01-21 17:00:12 +00:00
|
|
|
event_to_room_id: Dict[str, str],
|
|
|
|
event_to_types: Dict[str, Tuple[str, str]],
|
|
|
|
event_to_auth_chain: Dict[str, List[str]],
|
|
|
|
events_to_calc_chain_id_for: Set[str],
|
|
|
|
chain_map: Dict[str, Tuple[int, int]],
|
|
|
|
) -> Dict[str, Tuple[int, int]]:
|
|
|
|
"""Allocates, but does not persist, chain ID/sequence numbers for the
|
|
|
|
events in `events_to_calc_chain_id_for`. (c.f. _add_chain_cover_index
|
|
|
|
for info on args)
|
|
|
|
"""
|
|
|
|
|
|
|
|
# We now calculate the chain IDs/sequence numbers for the events. We do
|
|
|
|
# this by looking at the chain ID and sequence number of any auth event
|
|
|
|
# with the same type/state_key and incrementing the sequence number by
|
|
|
|
# one. If there was no match or the chain ID/sequence number is already
|
|
|
|
# taken we generate a new chain.
|
|
|
|
#
|
|
|
|
# We try to reduce the number of times that we hit the database by
|
|
|
|
# batching up calls, to make this more efficient when persisting large
|
|
|
|
# numbers of state events (e.g. during joins).
|
|
|
|
#
|
|
|
|
# We do this by:
|
|
|
|
# 1. Calculating for each event which auth event will be used to
|
|
|
|
# inherit the chain ID, i.e. converting the auth chain graph to a
|
|
|
|
# tree that we can allocate chains on. We also keep track of which
|
|
|
|
# existing chain IDs have been referenced.
|
|
|
|
# 2. Fetching the max allocated sequence number for each referenced
|
|
|
|
# existing chain ID, generating a map from chain ID to the max
|
|
|
|
# allocated sequence number.
|
|
|
|
# 3. Iterating over the tree and allocating a chain ID/seq no. to the
|
|
|
|
# new event, by incrementing the sequence number from the
|
|
|
|
# referenced event's chain ID/seq no. and checking that the
|
|
|
|
# incremented sequence number hasn't already been allocated (by
|
|
|
|
# looking in the map generated in the previous step). We generate a
|
|
|
|
# new chain if the sequence number has already been allocated.
|
|
|
|
#
|
|
|
|
|
|
|
|
existing_chains = set() # type: Set[int]
|
|
|
|
tree = [] # type: List[Tuple[str, Optional[str]]]
|
|
|
|
|
|
|
|
# We need to do this in a topologically sorted order as we want to
|
|
|
|
# generate chain IDs/sequence numbers of an event's auth events before
|
|
|
|
# the event itself.
|
|
|
|
for event_id in sorted_topologically(
|
|
|
|
events_to_calc_chain_id_for, event_to_auth_chain
|
|
|
|
):
|
|
|
|
for auth_id in event_to_auth_chain.get(event_id, []):
|
|
|
|
if event_to_types.get(event_id) == event_to_types.get(auth_id):
|
|
|
|
existing_chain_id = chain_map.get(auth_id)
|
|
|
|
if existing_chain_id:
|
|
|
|
existing_chains.add(existing_chain_id[0])
|
|
|
|
|
|
|
|
tree.append((event_id, auth_id))
|
|
|
|
break
|
|
|
|
else:
|
|
|
|
tree.append((event_id, None))
|
|
|
|
|
|
|
|
# Fetch the current max sequence number for each existing referenced chain.
|
|
|
|
sql = """
|
|
|
|
SELECT chain_id, MAX(sequence_number) FROM event_auth_chains
|
|
|
|
WHERE %s
|
|
|
|
GROUP BY chain_id
|
|
|
|
"""
|
|
|
|
clause, args = make_in_list_sql_clause(
|
|
|
|
db_pool.engine, "chain_id", existing_chains
|
|
|
|
)
|
|
|
|
txn.execute(sql % (clause,), args)
|
|
|
|
|
|
|
|
chain_to_max_seq_no = {row[0]: row[1] for row in txn} # type: Dict[Any, int]
|
|
|
|
|
|
|
|
# Allocate the new events chain ID/sequence numbers.
|
|
|
|
#
|
|
|
|
# To reduce the number of calls to the database we don't allocate a
|
|
|
|
# chain ID number in the loop, instead we use a temporary `object()` for
|
|
|
|
# each new chain ID. Once we've done the loop we generate the necessary
|
|
|
|
# number of new chain IDs in one call, replacing all temporary
|
|
|
|
# objects with real allocated chain IDs.
|
|
|
|
|
|
|
|
unallocated_chain_ids = set() # type: Set[object]
|
|
|
|
new_chain_tuples = {} # type: Dict[str, Tuple[Any, int]]
|
|
|
|
for event_id, auth_event_id in tree:
|
|
|
|
# If we reference an auth_event_id we fetch the allocated chain ID,
|
|
|
|
# either from the existing `chain_map` or the newly generated
|
|
|
|
# `new_chain_tuples` map.
|
|
|
|
existing_chain_id = None
|
|
|
|
if auth_event_id:
|
|
|
|
existing_chain_id = new_chain_tuples.get(auth_event_id)
|
|
|
|
if not existing_chain_id:
|
|
|
|
existing_chain_id = chain_map[auth_event_id]
|
|
|
|
|
|
|
|
new_chain_tuple = None # type: Optional[Tuple[Any, int]]
|
|
|
|
if existing_chain_id:
|
|
|
|
# We found a chain ID/sequence number candidate, check its
|
|
|
|
# not already taken.
|
|
|
|
proposed_new_id = existing_chain_id[0]
|
|
|
|
proposed_new_seq = existing_chain_id[1] + 1
|
|
|
|
|
|
|
|
if chain_to_max_seq_no[proposed_new_id] < proposed_new_seq:
|
|
|
|
new_chain_tuple = (
|
|
|
|
proposed_new_id,
|
|
|
|
proposed_new_seq,
|
|
|
|
)
|
|
|
|
|
|
|
|
# If we need to start a new chain we allocate a temporary chain ID.
|
|
|
|
if not new_chain_tuple:
|
|
|
|
new_chain_tuple = (object(), 1)
|
|
|
|
unallocated_chain_ids.add(new_chain_tuple[0])
|
|
|
|
|
|
|
|
new_chain_tuples[event_id] = new_chain_tuple
|
|
|
|
chain_to_max_seq_no[new_chain_tuple[0]] = new_chain_tuple[1]
|
|
|
|
|
|
|
|
# Generate new chain IDs for all unallocated chain IDs.
|
2021-02-24 10:13:53 +00:00
|
|
|
newly_allocated_chain_ids = event_chain_id_gen.get_next_mult_txn(
|
2021-01-21 17:00:12 +00:00
|
|
|
txn, len(unallocated_chain_ids)
|
|
|
|
)
|
|
|
|
|
|
|
|
# Map from potentially temporary chain ID to real chain ID
|
|
|
|
chain_id_to_allocated_map = dict(
|
|
|
|
zip(unallocated_chain_ids, newly_allocated_chain_ids)
|
|
|
|
) # type: Dict[Any, int]
|
|
|
|
chain_id_to_allocated_map.update((c, c) for c in existing_chains)
|
|
|
|
|
|
|
|
return {
|
|
|
|
event_id: (chain_id_to_allocated_map[chain_id], seq)
|
|
|
|
for event_id, (chain_id, seq) in new_chain_tuples.items()
|
|
|
|
}
|
|
|
|
|
2020-10-13 11:07:56 +00:00
|
|
|
def _persist_transaction_ids_txn(
|
|
|
|
self,
|
|
|
|
txn: LoggingTransaction,
|
|
|
|
events_and_contexts: List[Tuple[EventBase, EventContext]],
|
|
|
|
):
|
|
|
|
"""Persist the mapping from transaction IDs to event IDs (if defined)."""
|
|
|
|
|
|
|
|
to_insert = []
|
|
|
|
for event, _ in events_and_contexts:
|
|
|
|
token_id = getattr(event.internal_metadata, "token_id", None)
|
|
|
|
txn_id = getattr(event.internal_metadata, "txn_id", None)
|
|
|
|
if token_id and txn_id:
|
|
|
|
to_insert.append(
|
|
|
|
{
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"room_id": event.room_id,
|
|
|
|
"user_id": event.sender,
|
|
|
|
"token_id": token_id,
|
|
|
|
"txn_id": txn_id,
|
|
|
|
"inserted_ts": self._clock.time_msec(),
|
|
|
|
}
|
|
|
|
)
|
|
|
|
|
|
|
|
if to_insert:
|
|
|
|
self.db_pool.simple_insert_many_txn(
|
|
|
|
txn,
|
|
|
|
table="event_txn_id",
|
|
|
|
values=to_insert,
|
|
|
|
)
|
|
|
|
|
2020-01-20 18:07:20 +00:00
|
|
|
def _update_current_state_txn(
|
|
|
|
self,
|
|
|
|
txn: LoggingTransaction,
|
|
|
|
state_delta_by_room: Dict[str, DeltaState],
|
|
|
|
stream_id: int,
|
|
|
|
):
|
2020-06-15 11:03:36 +00:00
|
|
|
for room_id, delta_state in state_delta_by_room.items():
|
2020-01-20 18:07:20 +00:00
|
|
|
to_delete = delta_state.to_delete
|
|
|
|
to_insert = delta_state.to_insert
|
2017-01-20 15:40:04 +00:00
|
|
|
|
2020-01-29 11:01:32 +00:00
|
|
|
if delta_state.no_longer_in_room:
|
|
|
|
# Server is no longer in the room so we delete the room from
|
|
|
|
# current_state_events, being careful we've already updated the
|
|
|
|
# rooms.room_version column (which gets populated in a
|
|
|
|
# background task).
|
|
|
|
self._upsert_room_version_txn(txn, room_id)
|
|
|
|
|
|
|
|
# Before deleting we populate the current_state_delta_stream
|
|
|
|
# so that async background tasks get told what happened.
|
|
|
|
sql = """
|
|
|
|
INSERT INTO current_state_delta_stream
|
2020-10-09 12:10:33 +00:00
|
|
|
(stream_id, instance_name, room_id, type, state_key, event_id, prev_event_id)
|
|
|
|
SELECT ?, ?, room_id, type, state_key, null, event_id
|
2020-01-29 11:01:32 +00:00
|
|
|
FROM current_state_events
|
|
|
|
WHERE room_id = ?
|
|
|
|
"""
|
2020-10-09 12:10:33 +00:00
|
|
|
txn.execute(sql, (stream_id, self._instance_name, room_id))
|
2020-01-29 11:01:32 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_delete_txn(
|
2020-01-29 11:01:32 +00:00
|
|
|
txn,
|
|
|
|
table="current_state_events",
|
|
|
|
keyvalues={"room_id": room_id},
|
2017-01-20 15:40:04 +00:00
|
|
|
)
|
2020-01-29 11:01:32 +00:00
|
|
|
else:
|
|
|
|
# We're still in the room, so we update the current state as normal.
|
|
|
|
|
|
|
|
# First we add entries to the current_state_delta_stream. We
|
|
|
|
# do this before updating the current_state_events table so
|
|
|
|
# that we can use it to calculate the `prev_event_id`. (This
|
|
|
|
# allows us to not have to pull out the existing state
|
|
|
|
# unnecessarily).
|
|
|
|
#
|
|
|
|
# The stream_id for the update is chosen to be the minimum of the stream_ids
|
|
|
|
# for the batch of the events that we are persisting; that means we do not
|
|
|
|
# end up in a situation where workers see events before the
|
|
|
|
# current_state_delta updates.
|
|
|
|
#
|
|
|
|
sql = """
|
|
|
|
INSERT INTO current_state_delta_stream
|
2020-10-09 12:10:33 +00:00
|
|
|
(stream_id, instance_name, room_id, type, state_key, event_id, prev_event_id)
|
|
|
|
SELECT ?, ?, ?, ?, ?, ?, (
|
2020-01-29 11:01:32 +00:00
|
|
|
SELECT event_id FROM current_state_events
|
|
|
|
WHERE room_id = ? AND type = ? AND state_key = ?
|
2019-03-28 13:37:16 +00:00
|
|
|
)
|
2020-01-29 11:01:32 +00:00
|
|
|
"""
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-01-29 11:01:32 +00:00
|
|
|
sql,
|
2019-03-28 13:37:16 +00:00
|
|
|
(
|
2020-01-29 11:01:32 +00:00
|
|
|
(
|
|
|
|
stream_id,
|
2020-10-09 12:10:33 +00:00
|
|
|
self._instance_name,
|
2020-01-29 11:01:32 +00:00
|
|
|
room_id,
|
|
|
|
etype,
|
|
|
|
state_key,
|
|
|
|
to_insert.get((etype, state_key)),
|
|
|
|
room_id,
|
|
|
|
etype,
|
|
|
|
state_key,
|
|
|
|
)
|
|
|
|
for etype, state_key in itertools.chain(to_delete, to_insert)
|
|
|
|
),
|
|
|
|
)
|
|
|
|
# Now we actually update the current_state_events table
|
2019-01-30 10:53:17 +00:00
|
|
|
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-01-29 11:01:32 +00:00
|
|
|
"DELETE FROM current_state_events"
|
|
|
|
" WHERE room_id = ? AND type = ? AND state_key = ?",
|
|
|
|
(
|
|
|
|
(room_id, etype, state_key)
|
|
|
|
for etype, state_key in itertools.chain(to_delete, to_insert)
|
|
|
|
),
|
|
|
|
)
|
2019-01-30 10:53:17 +00:00
|
|
|
|
2020-01-29 11:01:32 +00:00
|
|
|
# We include the membership in the current state table, hence we do
|
|
|
|
# a lookup when we insert. This assumes that all events have already
|
|
|
|
# been inserted into room_memberships.
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-01-29 11:01:32 +00:00
|
|
|
"""INSERT INTO current_state_events
|
|
|
|
(room_id, type, state_key, event_id, membership)
|
|
|
|
VALUES (?, ?, ?, ?, (SELECT membership FROM room_memberships WHERE event_id = ?))
|
|
|
|
""",
|
|
|
|
[
|
|
|
|
(room_id, key[0], key[1], ev_id, ev_id)
|
2020-06-15 11:03:36 +00:00
|
|
|
for key, ev_id in to_insert.items()
|
2020-01-29 11:01:32 +00:00
|
|
|
],
|
|
|
|
)
|
2017-01-20 15:40:04 +00:00
|
|
|
|
2020-01-29 11:01:32 +00:00
|
|
|
# We now update `local_current_membership`. We do this regardless
|
|
|
|
# of whether we're still in the room or not to handle the case where
|
|
|
|
# e.g. we just got banned (where we need to record that fact here).
|
2017-01-20 15:40:04 +00:00
|
|
|
|
2020-01-15 14:59:33 +00:00
|
|
|
# Note: Do we really want to delete rows here (that we do not
|
|
|
|
# subsequently reinsert below)? While technically correct it means
|
|
|
|
# we have no record of the fact the user *was* a member of the
|
|
|
|
# room but got, say, state reset out of it.
|
|
|
|
if to_delete or to_insert:
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-01-15 14:59:33 +00:00
|
|
|
"DELETE FROM local_current_membership"
|
|
|
|
" WHERE room_id = ? AND user_id = ?",
|
|
|
|
(
|
|
|
|
(room_id, state_key)
|
|
|
|
for etype, state_key in itertools.chain(to_delete, to_insert)
|
|
|
|
if etype == EventTypes.Member and self.is_mine_id(state_key)
|
|
|
|
),
|
|
|
|
)
|
|
|
|
|
|
|
|
if to_insert:
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-01-15 14:59:33 +00:00
|
|
|
"""INSERT INTO local_current_membership
|
|
|
|
(room_id, user_id, event_id, membership)
|
|
|
|
VALUES (?, ?, ?, (SELECT membership FROM room_memberships WHERE event_id = ?))
|
|
|
|
""",
|
|
|
|
[
|
|
|
|
(room_id, key[1], ev_id, ev_id)
|
|
|
|
for key, ev_id in to_insert.items()
|
|
|
|
if key[0] == EventTypes.Member and self.is_mine_id(key[1])
|
|
|
|
],
|
|
|
|
)
|
|
|
|
|
2019-01-30 10:53:17 +00:00
|
|
|
txn.call_after(
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store._curr_state_delta_stream_cache.entity_has_changed,
|
2019-03-28 13:37:16 +00:00
|
|
|
room_id,
|
2019-04-02 11:42:39 +00:00
|
|
|
stream_id,
|
2019-01-30 10:53:17 +00:00
|
|
|
)
|
2017-06-13 08:56:18 +00:00
|
|
|
|
2019-01-30 10:53:17 +00:00
|
|
|
# Invalidate the various caches
|
|
|
|
|
|
|
|
# Figure out the changes of membership to invalidate the
|
|
|
|
# `get_rooms_for_user` cache.
|
|
|
|
# We find out which membership events we may have deleted
|
2021-02-12 16:01:48 +00:00
|
|
|
# and which we have added, then we invalidate the caches for all
|
2019-01-30 10:53:17 +00:00
|
|
|
# those users.
|
2020-02-21 12:15:07 +00:00
|
|
|
members_changed = {
|
2019-01-30 10:53:17 +00:00
|
|
|
state_key
|
|
|
|
for ev_type, state_key in itertools.chain(to_delete, to_insert)
|
|
|
|
if ev_type == EventTypes.Member
|
2020-02-21 12:15:07 +00:00
|
|
|
}
|
2019-01-30 10:53:17 +00:00
|
|
|
|
2019-04-02 11:42:39 +00:00
|
|
|
for member in members_changed:
|
|
|
|
txn.call_after(
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store.get_rooms_for_user_with_stream_ordering.invalidate,
|
|
|
|
(member,),
|
2019-04-02 11:42:39 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store._invalidate_state_caches_and_stream(
|
|
|
|
txn, room_id, members_changed
|
|
|
|
)
|
2019-01-30 10:53:17 +00:00
|
|
|
|
2020-01-29 11:01:32 +00:00
|
|
|
def _upsert_room_version_txn(self, txn: LoggingTransaction, room_id: str):
|
|
|
|
"""Update the room version in the database based off current state
|
|
|
|
events.
|
|
|
|
|
|
|
|
This is used when we're about to delete current state and we want to
|
|
|
|
ensure that the `rooms.room_version` column is up to date.
|
|
|
|
"""
|
|
|
|
|
|
|
|
sql = """
|
|
|
|
SELECT json FROM event_json
|
|
|
|
INNER JOIN current_state_events USING (room_id, event_id)
|
|
|
|
WHERE room_id = ? AND type = ? AND state_key = ?
|
|
|
|
"""
|
|
|
|
txn.execute(sql, (room_id, EventTypes.Create, ""))
|
|
|
|
row = txn.fetchone()
|
|
|
|
if row:
|
2020-07-16 15:32:19 +00:00
|
|
|
event_json = db_to_json(row[0])
|
2020-01-29 11:01:32 +00:00
|
|
|
content = event_json.get("content", {})
|
|
|
|
creator = content.get("creator")
|
|
|
|
room_version_id = content.get("room_version", RoomVersions.V1.identifier)
|
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_upsert_txn(
|
2020-01-29 11:01:32 +00:00
|
|
|
txn,
|
|
|
|
table="rooms",
|
|
|
|
keyvalues={"room_id": room_id},
|
|
|
|
values={"room_version": room_version_id},
|
|
|
|
insertion_values={"is_public": False, "creator": creator},
|
|
|
|
)
|
|
|
|
|
2019-03-28 13:37:16 +00:00
|
|
|
def _update_forward_extremities_txn(
|
|
|
|
self, txn, new_forward_extremities, max_stream_order
|
|
|
|
):
|
2020-06-15 11:03:36 +00:00
|
|
|
for room_id, new_extrem in new_forward_extremities.items():
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_delete_txn(
|
2019-03-28 13:37:16 +00:00
|
|
|
txn, table="event_forward_extremities", keyvalues={"room_id": room_id}
|
2017-01-20 14:28:53 +00:00
|
|
|
)
|
2020-05-13 12:38:22 +00:00
|
|
|
txn.call_after(
|
|
|
|
self.store.get_latest_event_ids_in_room.invalidate, (room_id,)
|
|
|
|
)
|
2017-01-20 14:28:53 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2017-01-20 14:28:53 +00:00
|
|
|
txn,
|
|
|
|
table="event_forward_extremities",
|
|
|
|
values=[
|
2019-03-28 13:37:16 +00:00
|
|
|
{"event_id": ev_id, "room_id": room_id}
|
2020-06-15 11:03:36 +00:00
|
|
|
for room_id, new_extrem in new_forward_extremities.items()
|
2017-01-20 14:28:53 +00:00
|
|
|
for ev_id in new_extrem
|
|
|
|
],
|
|
|
|
)
|
|
|
|
# We now insert into stream_ordering_to_exterm a mapping from room_id,
|
|
|
|
# new stream_ordering to new forward extremeties in the room.
|
|
|
|
# This allows us to later efficiently look up the forward extremeties
|
|
|
|
# for a room before a given stream_ordering
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2017-01-20 14:28:53 +00:00
|
|
|
txn,
|
|
|
|
table="stream_ordering_to_exterm",
|
|
|
|
values=[
|
|
|
|
{
|
|
|
|
"room_id": room_id,
|
|
|
|
"event_id": event_id,
|
|
|
|
"stream_ordering": max_stream_order,
|
|
|
|
}
|
2020-06-15 11:03:36 +00:00
|
|
|
for room_id, new_extrem in new_forward_extremities.items()
|
2017-01-20 14:28:53 +00:00
|
|
|
for event_id in new_extrem
|
2019-03-28 13:37:16 +00:00
|
|
|
],
|
2017-01-20 14:28:53 +00:00
|
|
|
)
|
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
@classmethod
|
2020-09-11 11:22:55 +00:00
|
|
|
def _filter_events_and_contexts_for_duplicates(
|
|
|
|
cls, events_and_contexts: List[Tuple[EventBase, EventContext]]
|
|
|
|
) -> List[Tuple[EventBase, EventContext]]:
|
2017-03-17 11:51:13 +00:00
|
|
|
"""Ensure that we don't have the same event twice.
|
|
|
|
|
|
|
|
Pick the earliest non-outlier if there is one, else the earliest one.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
events_and_contexts (list[(EventBase, EventContext)]):
|
|
|
|
Returns:
|
|
|
|
list[(EventBase, EventContext)]: filtered list
|
|
|
|
"""
|
2020-09-11 11:22:55 +00:00
|
|
|
new_events_and_contexts = (
|
|
|
|
OrderedDict()
|
|
|
|
) # type: OrderedDict[str, Tuple[EventBase, EventContext]]
|
2016-08-03 10:23:39 +00:00
|
|
|
for event, context in events_and_contexts:
|
|
|
|
prev_event_context = new_events_and_contexts.get(event.event_id)
|
|
|
|
if prev_event_context:
|
|
|
|
if not event.internal_metadata.is_outlier():
|
|
|
|
if prev_event_context[0].internal_metadata.is_outlier():
|
|
|
|
# To ensure correct ordering we pop, as OrderedDict is
|
|
|
|
# ordered by first insertion.
|
|
|
|
new_events_and_contexts.pop(event.event_id, None)
|
|
|
|
new_events_and_contexts[event.event_id] = (event, context)
|
|
|
|
else:
|
|
|
|
new_events_and_contexts[event.event_id] = (event, context)
|
2018-05-31 09:03:47 +00:00
|
|
|
return list(new_events_and_contexts.values())
|
2016-08-03 10:23:39 +00:00
|
|
|
|
2020-09-11 11:22:55 +00:00
|
|
|
def _update_room_depths_txn(
|
|
|
|
self,
|
|
|
|
txn,
|
|
|
|
events_and_contexts: List[Tuple[EventBase, EventContext]],
|
|
|
|
backfilled: bool,
|
|
|
|
):
|
2017-03-17 11:51:13 +00:00
|
|
|
"""Update min_depth for each room
|
2016-08-03 10:23:39 +00:00
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
Args:
|
|
|
|
txn (twisted.enterprise.adbapi.Connection): db connection
|
|
|
|
events_and_contexts (list[(EventBase, EventContext)]): events
|
|
|
|
we are persisting
|
|
|
|
backfilled (bool): True if the events were backfilled
|
|
|
|
"""
|
2020-09-11 11:22:55 +00:00
|
|
|
depth_updates = {} # type: Dict[str, int]
|
2016-02-09 16:19:15 +00:00
|
|
|
for event, context in events_and_contexts:
|
|
|
|
# Remove the any existing cache entries for the event_ids
|
2020-05-13 12:38:22 +00:00
|
|
|
txn.call_after(self.store._invalidate_get_event_cache, event.event_id)
|
2016-01-28 15:02:37 +00:00
|
|
|
if not backfilled:
|
|
|
|
txn.call_after(
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store._events_stream_cache.entity_has_changed,
|
2019-03-28 13:37:16 +00:00
|
|
|
event.room_id,
|
|
|
|
event.internal_metadata.stream_ordering,
|
2016-01-28 15:02:37 +00:00
|
|
|
)
|
|
|
|
|
2016-07-25 17:44:30 +00:00
|
|
|
if not event.internal_metadata.is_outlier() and not context.rejected:
|
2016-02-09 16:19:15 +00:00
|
|
|
depth_updates[event.room_id] = max(
|
|
|
|
event.depth, depth_updates.get(event.room_id, event.depth)
|
|
|
|
)
|
|
|
|
|
2020-06-15 11:03:36 +00:00
|
|
|
for room_id, depth in depth_updates.items():
|
2015-06-25 16:18:19 +00:00
|
|
|
self._update_min_depth_for_room_txn(txn, room_id, depth)
|
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
def _update_outliers_txn(self, txn, events_and_contexts):
|
|
|
|
"""Update any outliers with new event info.
|
|
|
|
|
|
|
|
This turns outliers into ex-outliers (unless the new event was
|
|
|
|
rejected).
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn (twisted.enterprise.adbapi.Connection): db connection
|
|
|
|
events_and_contexts (list[(EventBase, EventContext)]): events
|
|
|
|
we are persisting
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
list[(EventBase, EventContext)] new list, without events which
|
|
|
|
are already in the events table.
|
|
|
|
"""
|
2015-06-25 16:18:19 +00:00
|
|
|
txn.execute(
|
2019-03-28 13:37:16 +00:00
|
|
|
"SELECT event_id, outlier FROM events WHERE event_id in (%s)"
|
|
|
|
% (",".join(["?"] * len(events_and_contexts)),),
|
|
|
|
[event.event_id for event, _ in events_and_contexts],
|
2015-03-20 13:52:56 +00:00
|
|
|
)
|
2016-07-25 17:44:30 +00:00
|
|
|
|
2019-03-28 13:37:16 +00:00
|
|
|
have_persisted = {event_id: outlier for event_id, outlier in txn}
|
2015-06-25 16:18:19 +00:00
|
|
|
|
|
|
|
to_remove = set()
|
|
|
|
for event, context in events_and_contexts:
|
|
|
|
if event.event_id not in have_persisted:
|
|
|
|
continue
|
|
|
|
|
|
|
|
to_remove.add(event)
|
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
if context.rejected:
|
|
|
|
# If the event is rejected then we don't care if the event
|
|
|
|
# was an outlier or not.
|
|
|
|
continue
|
|
|
|
|
2015-06-25 16:18:19 +00:00
|
|
|
outlier_persisted = have_persisted[event.event_id]
|
|
|
|
if not event.internal_metadata.is_outlier() and outlier_persisted:
|
2016-07-26 09:49:52 +00:00
|
|
|
# We received a copy of an event that we had already stored as
|
|
|
|
# an outlier in the database. We now have some state at that
|
|
|
|
# so we need to update the state_groups table with that state.
|
|
|
|
|
2018-02-06 14:31:24 +00:00
|
|
|
# insert into event_to_state_groups.
|
2016-09-02 09:41:38 +00:00
|
|
|
try:
|
2018-02-06 14:31:24 +00:00
|
|
|
self._store_event_state_mappings_txn(txn, ((event, context),))
|
2016-09-02 09:41:38 +00:00
|
|
|
except Exception:
|
|
|
|
logger.exception("")
|
|
|
|
raise
|
2015-03-20 13:52:56 +00:00
|
|
|
|
2021-03-17 12:33:18 +00:00
|
|
|
# update the stored internal_metadata to update the "outlier" flag.
|
|
|
|
# TODO: This is unused as of Synapse 1.31. Remove it once we are happy
|
|
|
|
# to drop backwards-compatibility with 1.30.
|
2020-10-28 15:51:15 +00:00
|
|
|
metadata_json = json_encoder.encode(event.internal_metadata.get_dict())
|
2019-11-21 12:00:14 +00:00
|
|
|
sql = "UPDATE event_json SET internal_metadata = ? WHERE event_id = ?"
|
2019-03-28 13:37:16 +00:00
|
|
|
txn.execute(sql, (metadata_json, event.event_id))
|
2015-03-20 13:52:56 +00:00
|
|
|
|
2016-07-26 09:49:52 +00:00
|
|
|
# Add an entry to the ex_outlier_stream table to replicate the
|
|
|
|
# change in outlier status to our workers.
|
2016-03-30 16:19:56 +00:00
|
|
|
stream_order = event.internal_metadata.stream_ordering
|
2016-08-31 09:09:46 +00:00
|
|
|
state_group_id = context.state_group
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_txn(
|
2016-03-30 16:19:56 +00:00
|
|
|
txn,
|
|
|
|
table="ex_outlier_stream",
|
|
|
|
values={
|
|
|
|
"event_stream_ordering": stream_order,
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"state_group": state_group_id,
|
2020-10-09 12:10:33 +00:00
|
|
|
"instance_name": self._instance_name,
|
2019-03-28 13:37:16 +00:00
|
|
|
},
|
2016-03-30 16:19:56 +00:00
|
|
|
)
|
|
|
|
|
2019-11-21 12:00:14 +00:00
|
|
|
sql = "UPDATE events SET outlier = ? WHERE event_id = ?"
|
2019-03-28 13:37:16 +00:00
|
|
|
txn.execute(sql, (False, event.event_id))
|
2016-07-25 17:44:30 +00:00
|
|
|
|
2016-07-26 09:49:52 +00:00
|
|
|
# Update the event_backward_extremities table now that this
|
|
|
|
# event isn't an outlier any more.
|
2017-01-20 14:40:31 +00:00
|
|
|
self._update_backward_extremeties(txn, [event])
|
2015-09-15 15:34:42 +00:00
|
|
|
|
2019-03-28 13:37:16 +00:00
|
|
|
return [ec for ec in events_and_contexts if ec[0] not in to_remove]
|
2015-03-24 16:20:26 +00:00
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
def _store_event_txn(self, txn, events_and_contexts):
|
2021-01-11 13:57:33 +00:00
|
|
|
"""Insert new events into the event, event_json, redaction and
|
|
|
|
state_events tables.
|
2017-03-17 11:51:13 +00:00
|
|
|
|
|
|
|
Args:
|
|
|
|
txn (twisted.enterprise.adbapi.Connection): db connection
|
|
|
|
events_and_contexts (list[(EventBase, EventContext)]): events
|
|
|
|
we are persisting
|
|
|
|
"""
|
|
|
|
|
|
|
|
if not events_and_contexts:
|
|
|
|
# nothing to do here
|
|
|
|
return
|
|
|
|
|
|
|
|
def event_dict(event):
|
2017-03-24 10:57:02 +00:00
|
|
|
d = event.get_dict()
|
|
|
|
d.pop("redacted", None)
|
|
|
|
d.pop("redacted_because", None)
|
|
|
|
return d
|
2016-08-04 14:02:15 +00:00
|
|
|
|
2021-03-17 12:33:18 +00:00
|
|
|
def get_internal_metadata(event):
|
|
|
|
im = event.internal_metadata.get_dict()
|
|
|
|
|
|
|
|
# temporary hack for database compatibility with Synapse 1.30 and earlier:
|
|
|
|
# store the `outlier` flag inside the internal_metadata json as well as in
|
|
|
|
# the `events` table, so that if anyone rolls back to an older Synapse,
|
|
|
|
# things keep working. This can be removed once we are happy to drop support
|
|
|
|
# for that
|
|
|
|
if event.internal_metadata.is_outlier():
|
|
|
|
im["outlier"] = True
|
|
|
|
|
|
|
|
return im
|
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2015-06-25 16:18:19 +00:00
|
|
|
txn,
|
|
|
|
table="event_json",
|
|
|
|
values=[
|
|
|
|
{
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"room_id": event.room_id,
|
2020-10-28 15:51:15 +00:00
|
|
|
"internal_metadata": json_encoder.encode(
|
2021-03-17 12:33:18 +00:00
|
|
|
get_internal_metadata(event)
|
2018-08-01 14:54:06 +00:00
|
|
|
),
|
2020-10-28 15:51:15 +00:00
|
|
|
"json": json_encoder.encode(event_dict(event)),
|
2019-01-23 11:11:52 +00:00
|
|
|
"format_version": event.format_version,
|
2015-06-25 16:18:19 +00:00
|
|
|
}
|
|
|
|
for event, _ in events_and_contexts
|
|
|
|
],
|
2015-04-15 09:24:24 +00:00
|
|
|
)
|
2015-03-20 13:52:56 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2015-06-25 16:18:19 +00:00
|
|
|
txn,
|
|
|
|
table="events",
|
|
|
|
values=[
|
|
|
|
{
|
2020-09-14 09:16:41 +00:00
|
|
|
"instance_name": self._instance_name,
|
2015-06-25 16:18:19 +00:00
|
|
|
"stream_ordering": event.internal_metadata.stream_ordering,
|
|
|
|
"topological_ordering": event.depth,
|
|
|
|
"depth": event.depth,
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"room_id": event.room_id,
|
|
|
|
"type": event.type,
|
|
|
|
"processed": True,
|
|
|
|
"outlier": event.internal_metadata.is_outlier(),
|
2015-11-30 17:45:31 +00:00
|
|
|
"origin_server_ts": int(event.origin_server_ts),
|
2016-04-19 13:24:36 +00:00
|
|
|
"received_ts": self._clock.time_msec(),
|
2016-07-14 14:15:22 +00:00
|
|
|
"sender": event.sender,
|
|
|
|
"contains_url": (
|
2020-06-16 12:51:47 +00:00
|
|
|
"url" in event.content and isinstance(event.content["url"], str)
|
2016-07-14 14:15:22 +00:00
|
|
|
),
|
2015-06-25 16:18:19 +00:00
|
|
|
}
|
2020-08-06 16:15:35 +00:00
|
|
|
for event, _ in events_and_contexts
|
2015-06-25 16:18:19 +00:00
|
|
|
],
|
2015-03-20 13:52:56 +00:00
|
|
|
)
|
|
|
|
|
2021-04-16 13:44:55 +00:00
|
|
|
# If we're persisting an unredacted event we go and ensure
|
|
|
|
# that we mark any redactions that reference this event as
|
|
|
|
# requiring censoring.
|
|
|
|
sql = "UPDATE redactions SET have_censored = ? WHERE redacts = ?"
|
|
|
|
txn.execute_batch(
|
|
|
|
sql,
|
|
|
|
(
|
|
|
|
(
|
|
|
|
False,
|
|
|
|
event.event_id,
|
2019-10-01 10:05:48 +00:00
|
|
|
)
|
2021-04-16 13:44:55 +00:00
|
|
|
for event, _ in events_and_contexts
|
|
|
|
if not event.internal_metadata.is_redacted()
|
|
|
|
),
|
|
|
|
)
|
2019-10-01 10:05:48 +00:00
|
|
|
|
2021-01-11 13:57:33 +00:00
|
|
|
state_events_and_contexts = [
|
|
|
|
ec for ec in events_and_contexts if ec[0].is_state()
|
|
|
|
]
|
|
|
|
|
|
|
|
state_values = []
|
|
|
|
for event, context in state_events_and_contexts:
|
|
|
|
vals = {
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"room_id": event.room_id,
|
|
|
|
"type": event.type,
|
|
|
|
"state_key": event.state_key,
|
|
|
|
}
|
|
|
|
|
|
|
|
# TODO: How does this work with backfilling?
|
|
|
|
if hasattr(event, "replaces_state"):
|
|
|
|
vals["prev_state"] = event.replaces_state
|
|
|
|
|
|
|
|
state_values.append(vals)
|
|
|
|
|
|
|
|
self.db_pool.simple_insert_many_txn(
|
|
|
|
txn, table="state_events", values=state_values
|
|
|
|
)
|
|
|
|
|
2017-03-17 11:51:13 +00:00
|
|
|
def _store_rejected_events_txn(self, txn, events_and_contexts):
|
|
|
|
"""Add rows to the 'rejections' table for received events which were
|
|
|
|
rejected
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn (twisted.enterprise.adbapi.Connection): db connection
|
|
|
|
events_and_contexts (list[(EventBase, EventContext)]): events
|
|
|
|
we are persisting
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
list[(EventBase, EventContext)] new list, without the rejected
|
|
|
|
events.
|
|
|
|
"""
|
2016-07-26 09:49:52 +00:00
|
|
|
# Remove the rejected events from the list now that we've added them
|
|
|
|
# to the events table and the events_json table.
|
2016-07-25 17:44:30 +00:00
|
|
|
to_remove = set()
|
2016-07-25 15:12:16 +00:00
|
|
|
for event, context in events_and_contexts:
|
|
|
|
if context.rejected:
|
2016-07-26 09:49:52 +00:00
|
|
|
# Insert the event_id into the rejections table
|
2019-03-28 13:37:16 +00:00
|
|
|
self._store_rejections_txn(txn, event.event_id, context.rejected)
|
2016-07-26 09:49:52 +00:00
|
|
|
to_remove.add(event)
|
2016-07-25 17:44:30 +00:00
|
|
|
|
2019-03-28 13:37:16 +00:00
|
|
|
return [ec for ec in events_and_contexts if ec[0] not in to_remove]
|
2016-07-25 17:44:30 +00:00
|
|
|
|
2019-03-28 13:37:16 +00:00
|
|
|
def _update_metadata_tables_txn(
|
|
|
|
self, txn, events_and_contexts, all_events_and_contexts, backfilled
|
|
|
|
):
|
2017-03-17 11:51:13 +00:00
|
|
|
"""Update all the miscellaneous tables for new events
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn (twisted.enterprise.adbapi.Connection): db connection
|
|
|
|
events_and_contexts (list[(EventBase, EventContext)]): events
|
|
|
|
we are persisting
|
2018-02-20 12:33:04 +00:00
|
|
|
all_events_and_contexts (list[(EventBase, EventContext)]): all
|
|
|
|
events that we were going to persist. This includes events
|
|
|
|
we've already persisted, etc, that wouldn't appear in
|
|
|
|
events_and_context.
|
2017-03-17 11:51:13 +00:00
|
|
|
backfilled (bool): True if the events were backfilled
|
|
|
|
"""
|
|
|
|
|
2018-02-20 12:29:50 +00:00
|
|
|
# Insert all the push actions into the event_push_actions table.
|
|
|
|
self._set_push_actions_for_event_and_users_txn(
|
|
|
|
txn,
|
|
|
|
events_and_contexts=events_and_contexts,
|
2018-02-20 12:33:04 +00:00
|
|
|
all_events_and_contexts=all_events_and_contexts,
|
2018-02-20 12:29:50 +00:00
|
|
|
)
|
|
|
|
|
2016-07-25 17:44:30 +00:00
|
|
|
if not events_and_contexts:
|
2017-03-17 11:51:13 +00:00
|
|
|
# nothing to do here
|
2016-07-25 17:44:30 +00:00
|
|
|
return
|
|
|
|
|
|
|
|
for event, context in events_and_contexts:
|
2016-07-26 10:05:39 +00:00
|
|
|
if event.type == EventTypes.Redaction and event.redacts is not None:
|
|
|
|
# Remove the entries in the event_push_actions table for the
|
|
|
|
# redacted event.
|
|
|
|
self._remove_push_actions_for_event_id_txn(
|
|
|
|
txn, event.room_id, event.redacts
|
|
|
|
)
|
|
|
|
|
2019-05-20 09:13:05 +00:00
|
|
|
# Remove from relations table.
|
|
|
|
self._handle_redaction(txn, event.redacts)
|
|
|
|
|
2016-07-26 09:49:52 +00:00
|
|
|
# Update the event_forward_extremities, event_backward_extremities and
|
|
|
|
# event_edges tables.
|
2016-07-25 17:44:30 +00:00
|
|
|
self._handle_mult_prev_events(
|
2019-03-28 13:37:16 +00:00
|
|
|
txn, events=[event for event, _ in events_and_contexts]
|
2016-07-25 17:44:30 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
for event, _ in events_and_contexts:
|
|
|
|
if event.type == EventTypes.Name:
|
2019-08-21 12:16:28 +00:00
|
|
|
# Insert into the event_search table.
|
2016-07-25 17:44:30 +00:00
|
|
|
self._store_room_name_txn(txn, event)
|
|
|
|
elif event.type == EventTypes.Topic:
|
2019-08-21 12:16:28 +00:00
|
|
|
# Insert into the event_search table.
|
2016-07-25 17:44:30 +00:00
|
|
|
self._store_room_topic_txn(txn, event)
|
|
|
|
elif event.type == EventTypes.Message:
|
2016-07-26 09:49:52 +00:00
|
|
|
# Insert into the event_search table.
|
2016-07-25 17:44:30 +00:00
|
|
|
self._store_room_message_txn(txn, event)
|
2020-01-23 15:19:03 +00:00
|
|
|
elif event.type == EventTypes.Redaction and event.redacts is not None:
|
2016-07-26 09:49:52 +00:00
|
|
|
# Insert into the redactions table.
|
2016-07-25 17:44:30 +00:00
|
|
|
self._store_redaction(txn, event)
|
2019-11-04 17:09:22 +00:00
|
|
|
elif event.type == EventTypes.Retention:
|
|
|
|
# Update the room_retention table.
|
|
|
|
self._store_retention_policy_for_room_txn(txn, event)
|
2016-07-25 17:44:30 +00:00
|
|
|
|
2019-05-14 15:59:21 +00:00
|
|
|
self._handle_event_relations(txn, event)
|
|
|
|
|
2019-10-29 18:35:49 +00:00
|
|
|
# Store the labels for this event.
|
2019-11-01 16:22:44 +00:00
|
|
|
labels = event.content.get(EventContentFields.LABELS)
|
2019-10-29 18:35:49 +00:00
|
|
|
if labels:
|
2019-11-01 11:47:28 +00:00
|
|
|
self.insert_labels_for_event_txn(
|
|
|
|
txn, event.event_id, labels, event.room_id, event.depth
|
|
|
|
)
|
2019-10-29 18:35:49 +00:00
|
|
|
|
2019-12-03 19:19:45 +00:00
|
|
|
if self._ephemeral_messages_enabled:
|
|
|
|
# If there's an expiry timestamp on the event, store it.
|
|
|
|
expiry_ts = event.content.get(EventContentFields.SELF_DESTRUCT_AFTER)
|
|
|
|
if isinstance(expiry_ts, int) and not event.is_state():
|
|
|
|
self._insert_event_expiry_txn(txn, event.event_id, expiry_ts)
|
|
|
|
|
2016-07-26 09:49:52 +00:00
|
|
|
# Insert into the room_memberships table.
|
2016-07-25 17:44:30 +00:00
|
|
|
self._store_room_members_txn(
|
|
|
|
txn,
|
|
|
|
[
|
|
|
|
event
|
|
|
|
for event, _ in events_and_contexts
|
|
|
|
if event.type == EventTypes.Member
|
|
|
|
],
|
|
|
|
backfilled=backfilled,
|
|
|
|
)
|
|
|
|
|
2016-07-26 09:49:52 +00:00
|
|
|
# Insert event_reference_hashes table.
|
2015-06-25 16:18:19 +00:00
|
|
|
self._store_event_reference_hashes_txn(
|
|
|
|
txn, [event for event, _ in events_and_contexts]
|
|
|
|
)
|
|
|
|
|
2016-07-26 10:05:39 +00:00
|
|
|
# Prefill the event cache
|
2016-06-06 10:08:12 +00:00
|
|
|
self._add_to_cache(txn, events_and_contexts)
|
|
|
|
|
|
|
|
def _add_to_cache(self, txn, events_and_contexts):
|
|
|
|
to_prefill = []
|
|
|
|
|
|
|
|
rows = []
|
|
|
|
N = 200
|
|
|
|
for i in range(0, len(events_and_contexts), N):
|
2019-03-28 13:37:16 +00:00
|
|
|
ev_map = {e[0].event_id: e[0] for e in events_and_contexts[i : i + N]}
|
2016-06-06 10:08:12 +00:00
|
|
|
if not ev_map:
|
|
|
|
break
|
|
|
|
|
|
|
|
sql = (
|
|
|
|
"SELECT "
|
|
|
|
" e.event_id as event_id, "
|
|
|
|
" r.redacts as redacts,"
|
|
|
|
" rej.event_id as rejects "
|
|
|
|
" FROM events as e"
|
|
|
|
" LEFT JOIN rejections as rej USING (event_id)"
|
|
|
|
" LEFT JOIN redactions as r ON e.event_id = r.redacts"
|
2019-10-02 18:07:07 +00:00
|
|
|
" WHERE "
|
|
|
|
)
|
|
|
|
|
|
|
|
clause, args = make_in_list_sql_clause(
|
|
|
|
self.database_engine, "e.event_id", list(ev_map)
|
|
|
|
)
|
2016-06-06 10:08:12 +00:00
|
|
|
|
2019-10-02 18:07:07 +00:00
|
|
|
txn.execute(sql + clause, args)
|
2020-08-05 20:38:57 +00:00
|
|
|
rows = self.db_pool.cursor_to_dict(txn)
|
2016-06-06 10:08:12 +00:00
|
|
|
for row in rows:
|
|
|
|
event = ev_map[row["event_id"]]
|
|
|
|
if not row["rejects"] and not row["redacts"]:
|
2019-03-28 13:37:16 +00:00
|
|
|
to_prefill.append(
|
|
|
|
_EventCacheEntry(event=event, redacted_event=None)
|
|
|
|
)
|
2016-06-06 10:08:12 +00:00
|
|
|
|
|
|
|
def prefill():
|
|
|
|
for cache_entry in to_prefill:
|
2020-10-19 11:20:29 +00:00
|
|
|
self.store._get_event_cache.set((cache_entry[0].event_id,), cache_entry)
|
2019-03-28 13:37:16 +00:00
|
|
|
|
2016-06-06 10:08:12 +00:00
|
|
|
txn.call_after(prefill)
|
|
|
|
|
2015-03-20 13:52:56 +00:00
|
|
|
def _store_redaction(self, txn, event):
|
|
|
|
# invalidate the cache for the redacted event
|
2020-05-13 12:38:22 +00:00
|
|
|
txn.call_after(self.store._invalidate_get_event_cache, event.redacts)
|
2019-10-01 12:23:34 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_txn(
|
2019-10-01 12:23:34 +00:00
|
|
|
txn,
|
|
|
|
table="redactions",
|
|
|
|
values={
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"redacts": event.redacts,
|
|
|
|
"received_ts": self._clock.time_msec(),
|
|
|
|
},
|
2015-03-20 13:52:56 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def insert_labels_for_event_txn(
|
|
|
|
self, txn, event_id, labels, room_id, topological_ordering
|
|
|
|
):
|
|
|
|
"""Store the mapping between an event's ID and its labels, with one row per
|
|
|
|
(event_id, label) tuple.
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Args:
|
|
|
|
txn (LoggingTransaction): The transaction to execute.
|
|
|
|
event_id (str): The event's ID.
|
|
|
|
labels (list[str]): A list of text labels.
|
|
|
|
room_id (str): The ID of the room the event was sent to.
|
|
|
|
topological_ordering (int): The position of the event in the room's topology.
|
2019-08-29 16:38:51 +00:00
|
|
|
"""
|
2020-08-05 20:38:57 +00:00
|
|
|
return self.db_pool.simple_insert_many_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn=txn,
|
|
|
|
table="event_labels",
|
|
|
|
values=[
|
|
|
|
{
|
|
|
|
"event_id": event_id,
|
|
|
|
"label": label,
|
|
|
|
"room_id": room_id,
|
|
|
|
"topological_ordering": topological_ordering,
|
|
|
|
}
|
|
|
|
for label in labels
|
|
|
|
],
|
|
|
|
)
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _insert_event_expiry_txn(self, txn, event_id, expiry_ts):
|
|
|
|
"""Save the expiry timestamp associated with a given event ID.
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Args:
|
|
|
|
txn (LoggingTransaction): The database transaction to use.
|
|
|
|
event_id (str): The event ID the expiry timestamp is associated with.
|
|
|
|
expiry_ts (int): The timestamp at which to expire (delete) the event.
|
2019-08-29 16:38:51 +00:00
|
|
|
"""
|
2020-08-05 20:38:57 +00:00
|
|
|
return self.db_pool.simple_insert_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn=txn,
|
|
|
|
table="event_expiry",
|
|
|
|
values={"event_id": event_id, "expiry_ts": expiry_ts},
|
2019-12-04 13:52:46 +00:00
|
|
|
)
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_event_reference_hashes_txn(self, txn, events):
|
|
|
|
"""Store a hash for a PDU
|
|
|
|
Args:
|
|
|
|
txn (cursor):
|
|
|
|
events (list): list of Events.
|
|
|
|
"""
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
vals = []
|
|
|
|
for event in events:
|
|
|
|
ref_alg, ref_hash_bytes = compute_event_reference_hash(event)
|
|
|
|
vals.append(
|
|
|
|
{
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"algorithm": ref_alg,
|
|
|
|
"hash": memoryview(ref_hash_bytes),
|
|
|
|
}
|
2019-08-29 16:38:51 +00:00
|
|
|
)
|
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
|
|
|
txn, table="event_reference_hashes", values=vals
|
|
|
|
)
|
2019-08-29 16:38:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_room_members_txn(self, txn, events, backfilled):
|
|
|
|
"""Store a room member in the database."""
|
2020-09-23 15:42:14 +00:00
|
|
|
|
|
|
|
def str_or_none(val: Any) -> Optional[str]:
|
|
|
|
return val if isinstance(val, str) else None
|
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2019-12-03 19:19:45 +00:00
|
|
|
txn,
|
2020-05-13 12:38:22 +00:00
|
|
|
table="room_memberships",
|
|
|
|
values=[
|
|
|
|
{
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"user_id": event.state_key,
|
|
|
|
"sender": event.user_id,
|
|
|
|
"room_id": event.room_id,
|
|
|
|
"membership": event.membership,
|
2020-09-23 15:42:14 +00:00
|
|
|
"display_name": str_or_none(event.content.get("displayname")),
|
|
|
|
"avatar_url": str_or_none(event.content.get("avatar_url")),
|
2020-05-13 12:38:22 +00:00
|
|
|
}
|
|
|
|
for event in events
|
|
|
|
],
|
2019-12-03 19:19:45 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
for event in events:
|
|
|
|
txn.call_after(
|
|
|
|
self.store._membership_stream_cache.entity_has_changed,
|
|
|
|
event.state_key,
|
|
|
|
event.internal_metadata.stream_ordering,
|
2016-03-01 14:49:41 +00:00
|
|
|
)
|
2020-05-13 12:38:22 +00:00
|
|
|
txn.call_after(
|
|
|
|
self.store.get_invited_rooms_for_local_user.invalidate,
|
|
|
|
(event.state_key,),
|
2016-03-01 14:49:41 +00:00
|
|
|
)
|
2016-03-30 16:19:56 +00:00
|
|
|
|
2020-07-07 13:20:40 +00:00
|
|
|
# We update the local_current_membership table only if the event is
|
|
|
|
# "current", i.e., its something that has just happened.
|
|
|
|
#
|
|
|
|
# This will usually get updated by the `current_state_events` handling,
|
|
|
|
# unless its an outlier, and an outlier is only "current" if it's an "out of
|
|
|
|
# band membership", like a remote invite or a rejection of a remote invite.
|
|
|
|
if (
|
|
|
|
self.is_mine_id(event.state_key)
|
|
|
|
and not backfilled
|
|
|
|
and event.internal_metadata.is_outlier()
|
|
|
|
and event.internal_metadata.is_out_of_band_membership()
|
|
|
|
):
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_upsert_txn(
|
2020-07-07 13:20:40 +00:00
|
|
|
txn,
|
|
|
|
table="local_current_membership",
|
|
|
|
keyvalues={"room_id": event.room_id, "user_id": event.state_key},
|
|
|
|
values={
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"membership": event.membership,
|
|
|
|
},
|
|
|
|
)
|
2016-07-05 09:28:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _handle_event_relations(self, txn, event):
|
|
|
|
"""Handles inserting relation data during peristence of events
|
2019-10-30 15:12:49 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Args:
|
|
|
|
txn
|
|
|
|
event (EventBase)
|
2016-07-04 15:02:50 +00:00
|
|
|
"""
|
2020-05-13 12:38:22 +00:00
|
|
|
relation = event.content.get("m.relates_to")
|
|
|
|
if not relation:
|
|
|
|
# No relations
|
|
|
|
return
|
2016-07-04 15:02:50 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
rel_type = relation.get("rel_type")
|
|
|
|
if rel_type not in (
|
|
|
|
RelationTypes.ANNOTATION,
|
|
|
|
RelationTypes.REFERENCE,
|
|
|
|
RelationTypes.REPLACE,
|
|
|
|
):
|
|
|
|
# Unknown relation type
|
|
|
|
return
|
2016-07-05 09:28:51 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
parent_id = relation.get("event_id")
|
|
|
|
if not parent_id:
|
|
|
|
# Invalid relation
|
|
|
|
return
|
2018-02-14 16:41:12 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
aggregation_key = relation.get("key")
|
2018-02-14 16:41:12 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_relations",
|
|
|
|
values={
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"relates_to_id": parent_id,
|
|
|
|
"relation_type": rel_type,
|
|
|
|
"aggregation_key": aggregation_key,
|
|
|
|
},
|
2018-02-14 16:41:12 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
txn.call_after(self.store.get_relations_for_event.invalidate_many, (parent_id,))
|
|
|
|
txn.call_after(
|
|
|
|
self.store.get_aggregation_groups_for_event.invalidate_many, (parent_id,)
|
2016-07-04 15:02:50 +00:00
|
|
|
)
|
2018-02-14 11:02:22 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
if rel_type == RelationTypes.REPLACE:
|
|
|
|
txn.call_after(self.store.get_applicable_edit.invalidate, (parent_id,))
|
2018-02-14 11:02:22 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _handle_redaction(self, txn, redacted_event_id):
|
|
|
|
"""Handles receiving a redaction and checking whether we need to remove
|
|
|
|
any redacted relations from the database.
|
2018-09-13 14:05:52 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Args:
|
|
|
|
txn
|
|
|
|
redacted_event_id (str): The event that was redacted.
|
|
|
|
"""
|
2018-09-13 14:05:52 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_delete_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn, table="event_relations", keyvalues={"event_id": redacted_event_id}
|
2018-09-13 14:05:52 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_room_topic_txn(self, txn, event):
|
|
|
|
if hasattr(event, "content") and "topic" in event.content:
|
|
|
|
self.store_event_search_txn(
|
|
|
|
txn, event, "content.topic", event.content["topic"]
|
|
|
|
)
|
2016-09-02 09:41:38 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_room_name_txn(self, txn, event):
|
|
|
|
if hasattr(event, "content") and "name" in event.content:
|
|
|
|
self.store_event_search_txn(
|
|
|
|
txn, event, "content.name", event.content["name"]
|
|
|
|
)
|
2017-05-10 16:46:41 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_room_message_txn(self, txn, event):
|
|
|
|
if hasattr(event, "content") and "body" in event.content:
|
|
|
|
self.store_event_search_txn(
|
|
|
|
txn, event, "content.body", event.content["body"]
|
|
|
|
)
|
2016-07-04 15:02:50 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_retention_policy_for_room_txn(self, txn, event):
|
2020-10-14 11:00:52 +00:00
|
|
|
if not event.is_state():
|
|
|
|
logger.debug("Ignoring non-state m.room.retention event")
|
|
|
|
return
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
if hasattr(event, "content") and (
|
|
|
|
"min_lifetime" in event.content or "max_lifetime" in event.content
|
|
|
|
):
|
|
|
|
if (
|
|
|
|
"min_lifetime" in event.content
|
2020-06-16 12:51:47 +00:00
|
|
|
and not isinstance(event.content.get("min_lifetime"), int)
|
2020-05-13 12:38:22 +00:00
|
|
|
) or (
|
|
|
|
"max_lifetime" in event.content
|
2020-06-16 12:51:47 +00:00
|
|
|
and not isinstance(event.content.get("max_lifetime"), int)
|
2020-05-13 12:38:22 +00:00
|
|
|
):
|
|
|
|
# Ignore the event if one of the value isn't an integer.
|
|
|
|
return
|
2017-05-10 16:46:41 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn=txn,
|
|
|
|
table="room_retention",
|
|
|
|
values={
|
|
|
|
"room_id": event.room_id,
|
|
|
|
"event_id": event.event_id,
|
|
|
|
"min_lifetime": event.content.get("min_lifetime"),
|
|
|
|
"max_lifetime": event.content.get("max_lifetime"),
|
|
|
|
},
|
|
|
|
)
|
2016-07-15 13:23:15 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store._invalidate_cache_and_stream(
|
|
|
|
txn, self.store.get_retention_policy_for_room, (event.room_id,)
|
|
|
|
)
|
2016-07-15 13:23:15 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def store_event_search_txn(self, txn, event, key, value):
|
|
|
|
"""Add event to the search table
|
2017-05-10 16:46:41 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Args:
|
|
|
|
txn (cursor):
|
|
|
|
event (EventBase):
|
|
|
|
key (str):
|
|
|
|
value (str):
|
2019-03-28 13:37:16 +00:00
|
|
|
"""
|
2020-05-13 12:38:22 +00:00
|
|
|
self.store.store_search_entries_txn(
|
|
|
|
txn,
|
|
|
|
(
|
|
|
|
SearchEntry(
|
|
|
|
key=key,
|
|
|
|
value=value,
|
|
|
|
event_id=event.event_id,
|
|
|
|
room_id=event.room_id,
|
|
|
|
stream_ordering=event.internal_metadata.stream_ordering,
|
|
|
|
origin_server_ts=event.origin_server_ts,
|
|
|
|
),
|
|
|
|
),
|
2019-03-28 13:37:16 +00:00
|
|
|
)
|
2016-09-05 13:49:08 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _set_push_actions_for_event_and_users_txn(
|
|
|
|
self, txn, events_and_contexts, all_events_and_contexts
|
|
|
|
):
|
|
|
|
"""Handles moving push actions from staging table to main
|
|
|
|
event_push_actions table for all events in `events_and_contexts`.
|
2018-10-04 14:18:52 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Also ensures that all events in `all_events_and_contexts` are removed
|
|
|
|
from the push action staging area.
|
2016-07-04 15:02:50 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Args:
|
|
|
|
events_and_contexts (list[(EventBase, EventContext)]): events
|
|
|
|
we are persisting
|
|
|
|
all_events_and_contexts (list[(EventBase, EventContext)]): all
|
|
|
|
events that we were going to persist. This includes events
|
|
|
|
we've already persisted, etc, that wouldn't appear in
|
|
|
|
events_and_context.
|
|
|
|
"""
|
2017-05-10 16:46:41 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
sql = """
|
|
|
|
INSERT INTO event_push_actions (
|
|
|
|
room_id, event_id, user_id, actions, stream_ordering,
|
2020-09-02 16:19:37 +00:00
|
|
|
topological_ordering, notif, highlight, unread
|
2016-07-04 15:02:50 +00:00
|
|
|
)
|
2020-09-02 16:19:37 +00:00
|
|
|
SELECT ?, event_id, user_id, actions, ?, ?, notif, highlight, unread
|
2020-05-13 12:38:22 +00:00
|
|
|
FROM event_push_actions_staging
|
|
|
|
WHERE event_id = ?
|
|
|
|
"""
|
2016-07-04 15:02:50 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
if events_and_contexts:
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-05-13 12:38:22 +00:00
|
|
|
sql,
|
|
|
|
(
|
|
|
|
(
|
|
|
|
event.room_id,
|
|
|
|
event.internal_metadata.stream_ordering,
|
|
|
|
event.depth,
|
|
|
|
event.event_id,
|
|
|
|
)
|
|
|
|
for event, _ in events_and_contexts
|
|
|
|
),
|
|
|
|
)
|
2018-02-14 15:44:51 +00:00
|
|
|
|
2021-04-16 13:44:55 +00:00
|
|
|
room_to_event_ids = {} # type: Dict[str, List[str]]
|
|
|
|
for e, _ in events_and_contexts:
|
|
|
|
room_to_event_ids.setdefault(e.room_id, []).append(e.event_id)
|
2018-02-14 15:44:51 +00:00
|
|
|
|
2021-04-16 13:44:55 +00:00
|
|
|
for room_id, event_ids in room_to_event_ids.items():
|
|
|
|
rows = self.db_pool.simple_select_many_txn(
|
|
|
|
txn,
|
|
|
|
table="event_push_actions_staging",
|
|
|
|
column="event_id",
|
|
|
|
iterable=event_ids,
|
|
|
|
keyvalues={},
|
|
|
|
retcols=("user_id",),
|
2020-05-13 12:38:22 +00:00
|
|
|
)
|
|
|
|
|
2021-04-16 13:44:55 +00:00
|
|
|
user_ids = {row["user_id"] for row in rows}
|
|
|
|
|
|
|
|
for user_id in user_ids:
|
|
|
|
txn.call_after(
|
|
|
|
self.store.get_unread_event_push_actions_by_room_for_user.invalidate_many,
|
|
|
|
(room_id, user_id),
|
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
# Now we delete the staging area for *all* events that were being
|
|
|
|
# persisted.
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-05-13 12:38:22 +00:00
|
|
|
"DELETE FROM event_push_actions_staging WHERE event_id = ?",
|
|
|
|
((event.event_id,) for event, _ in all_events_and_contexts),
|
2018-02-14 11:02:22 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _remove_push_actions_for_event_id_txn(self, txn, room_id, event_id):
|
|
|
|
# Sad that we have to blow away the cache for the whole room here
|
|
|
|
txn.call_after(
|
|
|
|
self.store.get_unread_event_push_actions_by_room_for_user.invalidate_many,
|
2019-03-28 13:37:16 +00:00
|
|
|
(room_id,),
|
|
|
|
)
|
2018-02-09 12:13:34 +00:00
|
|
|
txn.execute(
|
2020-05-13 12:38:22 +00:00
|
|
|
"DELETE FROM event_push_actions WHERE room_id = ? AND event_id = ?",
|
|
|
|
(room_id, event_id),
|
2018-02-09 12:13:34 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_rejections_txn(self, txn, event_id, reason):
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn,
|
|
|
|
table="rejections",
|
|
|
|
values={
|
|
|
|
"event_id": event_id,
|
|
|
|
"reason": reason,
|
|
|
|
"last_check": self._clock.time_msec(),
|
|
|
|
},
|
|
|
|
)
|
2019-08-22 09:42:59 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _store_event_state_mappings_txn(
|
|
|
|
self, txn, events_and_contexts: Iterable[Tuple[EventBase, EventContext]]
|
|
|
|
):
|
|
|
|
state_groups = {}
|
|
|
|
for event, context in events_and_contexts:
|
|
|
|
if event.internal_metadata.is_outlier():
|
|
|
|
continue
|
2019-11-04 13:36:57 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
# if the event was rejected, just give it the same state as its
|
|
|
|
# predecessor.
|
|
|
|
if context.rejected:
|
|
|
|
state_groups[event.event_id] = context.state_group_before_event
|
|
|
|
continue
|
2019-08-22 09:42:59 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
state_groups[event.event_id] = context.state_group
|
2019-08-22 09:42:59 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_to_state_groups",
|
|
|
|
values=[
|
|
|
|
{"state_group": state_group_id, "event_id": event_id}
|
2020-06-15 11:03:36 +00:00
|
|
|
for event_id, state_group_id in state_groups.items()
|
2020-05-13 12:38:22 +00:00
|
|
|
],
|
2019-08-22 09:42:59 +00:00
|
|
|
)
|
|
|
|
|
2020-06-15 11:03:36 +00:00
|
|
|
for event_id, state_group_id in state_groups.items():
|
2020-05-13 12:38:22 +00:00
|
|
|
txn.call_after(
|
|
|
|
self.store._get_state_group_for_event.prefill,
|
|
|
|
(event_id,),
|
|
|
|
state_group_id,
|
2019-08-22 09:42:59 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _update_min_depth_for_room_txn(self, txn, room_id, depth):
|
|
|
|
min_depth = self.store._get_min_depth_interaction(txn, room_id)
|
2019-08-22 09:42:59 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
if min_depth is not None and depth >= min_depth:
|
|
|
|
return
|
2017-04-11 16:34:09 +00:00
|
|
|
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_upsert_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn,
|
|
|
|
table="room_depth",
|
|
|
|
keyvalues={"room_id": room_id},
|
|
|
|
values={"min_depth": depth},
|
2017-04-11 16:34:09 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _handle_mult_prev_events(self, txn, events):
|
2019-11-01 10:30:51 +00:00
|
|
|
"""
|
2020-05-13 12:38:22 +00:00
|
|
|
For the given event, update the event edges table and forward and
|
|
|
|
backward extremities tables.
|
|
|
|
"""
|
2020-08-05 20:38:57 +00:00
|
|
|
self.db_pool.simple_insert_many_txn(
|
2020-05-13 12:38:22 +00:00
|
|
|
txn,
|
|
|
|
table="event_edges",
|
2019-11-01 11:47:28 +00:00
|
|
|
values=[
|
|
|
|
{
|
2020-05-13 12:38:22 +00:00
|
|
|
"event_id": ev.event_id,
|
|
|
|
"prev_event_id": e_id,
|
|
|
|
"room_id": ev.room_id,
|
|
|
|
"is_state": False,
|
2019-11-01 11:47:28 +00:00
|
|
|
}
|
2020-05-13 12:38:22 +00:00
|
|
|
for ev in events
|
|
|
|
for e_id in ev.prev_event_ids()
|
2019-11-01 11:47:28 +00:00
|
|
|
],
|
2019-10-29 18:35:49 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
self._update_backward_extremeties(txn, events)
|
2019-12-03 19:19:45 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
def _update_backward_extremeties(self, txn, events):
|
|
|
|
"""Updates the event_backward_extremities tables based on the new/updated
|
|
|
|
events being persisted.
|
2019-12-03 19:19:45 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
This is called for new events *and* for events that were outliers, but
|
|
|
|
are now being persisted as non-outliers.
|
2019-12-03 19:19:45 +00:00
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
Forward extremities are handled when we first start persisting the events.
|
2019-12-03 19:19:45 +00:00
|
|
|
"""
|
2020-09-11 11:22:55 +00:00
|
|
|
events_by_room = {} # type: Dict[str, List[EventBase]]
|
2020-05-13 12:38:22 +00:00
|
|
|
for ev in events:
|
|
|
|
events_by_room.setdefault(ev.room_id, []).append(ev)
|
|
|
|
|
|
|
|
query = (
|
|
|
|
"INSERT INTO event_backward_extremities (event_id, room_id)"
|
|
|
|
" SELECT ?, ? WHERE NOT EXISTS ("
|
|
|
|
" SELECT 1 FROM event_backward_extremities"
|
|
|
|
" WHERE event_id = ? AND room_id = ?"
|
|
|
|
" )"
|
|
|
|
" AND NOT EXISTS ("
|
|
|
|
" SELECT 1 FROM events WHERE event_id = ? AND room_id = ? "
|
|
|
|
" AND outlier = ?"
|
|
|
|
" )"
|
2019-12-03 19:19:45 +00:00
|
|
|
)
|
|
|
|
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-05-13 12:38:22 +00:00
|
|
|
query,
|
|
|
|
[
|
|
|
|
(e_id, ev.room_id, e_id, ev.room_id, e_id, ev.room_id, False)
|
|
|
|
for ev in events
|
|
|
|
for e_id in ev.prev_event_ids()
|
|
|
|
if not ev.internal_metadata.is_outlier()
|
|
|
|
],
|
2019-12-03 19:19:45 +00:00
|
|
|
)
|
|
|
|
|
2020-05-13 12:38:22 +00:00
|
|
|
query = (
|
|
|
|
"DELETE FROM event_backward_extremities"
|
|
|
|
" WHERE event_id = ? AND room_id = ?"
|
|
|
|
)
|
2021-01-21 10:22:53 +00:00
|
|
|
txn.execute_batch(
|
2020-05-13 12:38:22 +00:00
|
|
|
query,
|
|
|
|
[
|
|
|
|
(ev.event_id, ev.room_id)
|
|
|
|
for ev in events
|
|
|
|
if not ev.internal_metadata.is_outlier()
|
|
|
|
],
|
|
|
|
)
|
2021-01-11 16:09:22 +00:00
|
|
|
|
|
|
|
|
|
|
|
@attr.s(slots=True)
|
|
|
|
class _LinkMap:
|
|
|
|
"""A helper type for tracking links between chains."""
|
|
|
|
|
|
|
|
# Stores the set of links as nested maps: source chain ID -> target chain ID
|
|
|
|
# -> source sequence number -> target sequence number.
|
|
|
|
maps = attr.ib(type=Dict[int, Dict[int, Dict[int, int]]], factory=dict)
|
|
|
|
|
|
|
|
# Stores the links that have been added (with new set to true), as tuples of
|
|
|
|
# `(source chain ID, source sequence no, target chain ID, target sequence no.)`
|
|
|
|
additions = attr.ib(type=Set[Tuple[int, int, int, int]], factory=set)
|
|
|
|
|
|
|
|
def add_link(
|
|
|
|
self,
|
|
|
|
src_tuple: Tuple[int, int],
|
|
|
|
target_tuple: Tuple[int, int],
|
|
|
|
new: bool = True,
|
|
|
|
) -> bool:
|
|
|
|
"""Add a new link between two chains, ensuring no redundant links are added.
|
|
|
|
|
|
|
|
New links should be added in topological order.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
src_tuple: The chain ID/sequence number of the source of the link.
|
|
|
|
target_tuple: The chain ID/sequence number of the target of the link.
|
|
|
|
new: Whether this is a "new" link, i.e. should it be returned
|
|
|
|
by `get_additions`.
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
True if a link was added, false if the given link was dropped as redundant
|
|
|
|
"""
|
|
|
|
src_chain, src_seq = src_tuple
|
|
|
|
target_chain, target_seq = target_tuple
|
|
|
|
|
|
|
|
current_links = self.maps.setdefault(src_chain, {}).setdefault(target_chain, {})
|
|
|
|
|
|
|
|
assert src_chain != target_chain
|
|
|
|
|
|
|
|
if new:
|
|
|
|
# Check if the new link is redundant
|
|
|
|
for current_seq_src, current_seq_target in current_links.items():
|
|
|
|
# If a link "crosses" another link then its redundant. For example
|
|
|
|
# in the following link 1 (L1) is redundant, as any event reachable
|
|
|
|
# via L1 is *also* reachable via L2.
|
|
|
|
#
|
|
|
|
# Chain A Chain B
|
|
|
|
# | |
|
|
|
|
# L1 |------ |
|
|
|
|
# | | |
|
|
|
|
# L2 |---- | -->|
|
|
|
|
# | | |
|
|
|
|
# | |--->|
|
|
|
|
# | |
|
|
|
|
# | |
|
|
|
|
#
|
|
|
|
# So we only need to keep links which *do not* cross, i.e. links
|
|
|
|
# that both start and end above or below an existing link.
|
|
|
|
#
|
|
|
|
# Note, since we add links in topological ordering we should never
|
|
|
|
# see `src_seq` less than `current_seq_src`.
|
|
|
|
|
|
|
|
if current_seq_src <= src_seq and target_seq <= current_seq_target:
|
|
|
|
# This new link is redundant, nothing to do.
|
|
|
|
return False
|
|
|
|
|
|
|
|
self.additions.add((src_chain, src_seq, target_chain, target_seq))
|
|
|
|
|
|
|
|
current_links[src_seq] = target_seq
|
|
|
|
return True
|
|
|
|
|
|
|
|
def get_links_from(
|
|
|
|
self, src_tuple: Tuple[int, int]
|
|
|
|
) -> Generator[Tuple[int, int], None, None]:
|
|
|
|
"""Gets the chains reachable from the given chain/sequence number.
|
|
|
|
|
|
|
|
Yields:
|
|
|
|
The chain ID and sequence number the link points to.
|
|
|
|
"""
|
|
|
|
src_chain, src_seq = src_tuple
|
|
|
|
for target_id, sequence_numbers in self.maps.get(src_chain, {}).items():
|
|
|
|
for link_src_seq, target_seq in sequence_numbers.items():
|
|
|
|
if link_src_seq <= src_seq:
|
|
|
|
yield target_id, target_seq
|
|
|
|
|
|
|
|
def get_links_between(
|
|
|
|
self, source_chain: int, target_chain: int
|
|
|
|
) -> Generator[Tuple[int, int], None, None]:
|
|
|
|
"""Gets the links between two chains.
|
|
|
|
|
|
|
|
Yields:
|
|
|
|
The source and target sequence numbers.
|
|
|
|
"""
|
|
|
|
|
|
|
|
yield from self.maps.get(source_chain, {}).get(target_chain, {}).items()
|
|
|
|
|
|
|
|
def get_additions(self) -> Generator[Tuple[int, int, int, int], None, None]:
|
|
|
|
"""Gets any newly added links.
|
|
|
|
|
|
|
|
Yields:
|
|
|
|
The source chain ID/sequence number and target chain ID/sequence number
|
|
|
|
"""
|
|
|
|
|
|
|
|
for src_chain, src_seq, target_chain, _ in self.additions:
|
|
|
|
target_seq = self.maps.get(src_chain, {}).get(target_chain, {}).get(src_seq)
|
|
|
|
if target_seq is not None:
|
|
|
|
yield (src_chain, src_seq, target_chain, target_seq)
|
|
|
|
|
|
|
|
def exists_path_from(
|
|
|
|
self,
|
|
|
|
src_tuple: Tuple[int, int],
|
|
|
|
target_tuple: Tuple[int, int],
|
|
|
|
) -> bool:
|
|
|
|
"""Checks if there is a path between the source chain ID/sequence and
|
|
|
|
target chain ID/sequence.
|
|
|
|
"""
|
|
|
|
src_chain, src_seq = src_tuple
|
|
|
|
target_chain, target_seq = target_tuple
|
|
|
|
|
|
|
|
if src_chain == target_chain:
|
|
|
|
return target_seq <= src_seq
|
|
|
|
|
|
|
|
links = self.get_links_between(src_chain, target_chain)
|
|
|
|
for link_start_seq, link_end_seq in links:
|
|
|
|
if link_start_seq <= src_seq and target_seq <= link_end_seq:
|
|
|
|
return True
|
|
|
|
|
|
|
|
return False
|