diff options
Diffstat (limited to 'synapse/storage')
| -rw-r--r-- | synapse/storage/__init__.py | 2 | ||||
| -rw-r--r-- | synapse/storage/_base.py | 9 | ||||
| -rw-r--r-- | synapse/storage/account_data.py | 211 | ||||
| -rw-r--r-- | synapse/storage/events.py | 79 | ||||
| -rw-r--r-- | synapse/storage/prepare_database.py | 2 | ||||
| -rw-r--r-- | synapse/storage/registration.py | 2 | ||||
| -rw-r--r-- | synapse/storage/roommember.py | 73 | ||||
| -rw-r--r-- | synapse/storage/schema/delta/15/v15.sql | 23 | ||||
| -rw-r--r-- | synapse/storage/schema/delta/25/fts.py | 2 | ||||
| -rw-r--r-- | synapse/storage/schema/delta/27/account_data.sql | 36 | ||||
| -rw-r--r-- | synapse/storage/schema/delta/27/forgotten_memberships.sql | 26 | ||||
| -rw-r--r-- | synapse/storage/schema/delta/27/ts.py | 57 | ||||
| -rw-r--r-- | synapse/storage/search.py | 263 | ||||
| -rw-r--r-- | synapse/storage/tags.py | 4 |
14 files changed, 726 insertions, 63 deletions
diff --git a/synapse/storage/__init__.py b/synapse/storage/__init__.py index e7443f2838..c46b653f11 100644 --- a/synapse/storage/__init__.py +++ b/synapse/storage/__init__.py @@ -42,6 +42,7 @@ from .end_to_end_keys import EndToEndKeyStore from .receipts import ReceiptsStore from .search import SearchStore from .tags import TagsStore +from .account_data import AccountDataStore import logging @@ -73,6 +74,7 @@ class DataStore(RoomMemberStore, RoomStore, EndToEndKeyStore, SearchStore, TagsStore, + AccountDataStore, ): def __init__(self, hs): diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py index 218e708054..17a14e001c 100644 --- a/synapse/storage/_base.py +++ b/synapse/storage/_base.py @@ -214,7 +214,8 @@ class SQLBaseStore(object): self._clock.looping_call(loop, 10000) - def _new_transaction(self, conn, desc, after_callbacks, func, *args, **kwargs): + def _new_transaction(self, conn, desc, after_callbacks, logging_context, + func, *args, **kwargs): start = time.time() * 1000 txn_id = self._TXN_ID @@ -277,6 +278,9 @@ class SQLBaseStore(object): end = time.time() * 1000 duration = end - start + if logging_context is not None: + logging_context.add_database_transaction(duration) + transaction_logger.debug("[TXN END] {%s} %f", name, duration) self._current_txn_total_time += duration @@ -302,7 +306,8 @@ class SQLBaseStore(object): current_context.copy_to(context) return self._new_transaction( - conn, desc, after_callbacks, func, *args, **kwargs + conn, desc, after_callbacks, current_context, + func, *args, **kwargs ) result = yield preserve_context_over_fn( diff --git a/synapse/storage/account_data.py b/synapse/storage/account_data.py new file mode 100644 index 0000000000..d1829f84e8 --- /dev/null +++ b/synapse/storage/account_data.py @@ -0,0 +1,211 @@ +# -*- coding: utf-8 -*- +# Copyright 2014, 2015 OpenMarket Ltd +# +# 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. + +from ._base import SQLBaseStore +from twisted.internet import defer + +import ujson as json +import logging + +logger = logging.getLogger(__name__) + + +class AccountDataStore(SQLBaseStore): + + def get_account_data_for_user(self, user_id): + """Get all the client account_data for a user. + + Args: + user_id(str): The user to get the account_data for. + Returns: + A deferred pair of a dict of global account_data and a dict + mapping from room_id string to per room account_data dicts. + """ + + def get_account_data_for_user_txn(txn): + rows = self._simple_select_list_txn( + txn, "account_data", {"user_id": user_id}, + ["account_data_type", "content"] + ) + + global_account_data = { + row["account_data_type"]: json.loads(row["content"]) for row in rows + } + + rows = self._simple_select_list_txn( + txn, "room_account_data", {"user_id": user_id}, + ["room_id", "account_data_type", "content"] + ) + + by_room = {} + for row in rows: + room_data = by_room.setdefault(row["room_id"], {}) + room_data[row["account_data_type"]] = json.loads(row["content"]) + + return (global_account_data, by_room) + + return self.runInteraction( + "get_account_data_for_user", get_account_data_for_user_txn + ) + + def get_account_data_for_room(self, user_id, room_id): + """Get all the client account_data for a user for a room. + + Args: + user_id(str): The user to get the account_data for. + room_id(str): The room to get the account_data for. + Returns: + A deferred dict of the room account_data + """ + def get_account_data_for_room_txn(txn): + rows = self._simple_select_list_txn( + txn, "room_account_data", {"user_id": user_id, "room_id": room_id}, + ["account_data_type", "content"] + ) + + return { + row["account_data_type"]: json.loads(row["content"]) for row in rows + } + + return self.runInteraction( + "get_account_data_for_room", get_account_data_for_room_txn + ) + + def get_updated_account_data_for_user(self, user_id, stream_id): + """Get all the client account_data for a that's changed. + + Args: + user_id(str): The user to get the account_data for. + stream_id(int): The point in the stream since which to get updates + Returns: + A deferred pair of a dict of global account_data and a dict + mapping from room_id string to per room account_data dicts. + """ + + def get_updated_account_data_for_user_txn(txn): + sql = ( + "SELECT account_data_type, content FROM account_data" + " WHERE user_id = ? AND stream_id > ?" + ) + + txn.execute(sql, (user_id, stream_id)) + + global_account_data = { + row[0]: json.loads(row[1]) for row in txn.fetchall() + } + + sql = ( + "SELECT room_id, account_data_type, content FROM room_account_data" + " WHERE user_id = ? AND stream_id > ?" + ) + + txn.execute(sql, (user_id, stream_id)) + + account_data_by_room = {} + for row in txn.fetchall(): + room_account_data = account_data_by_room.setdefault(row[0], {}) + room_account_data[row[1]] = json.loads(row[2]) + + return (global_account_data, account_data_by_room) + + return self.runInteraction( + "get_updated_account_data_for_user", get_updated_account_data_for_user_txn + ) + + @defer.inlineCallbacks + def add_account_data_to_room(self, user_id, room_id, account_data_type, content): + """Add some account_data to a room for a user. + Args: + user_id(str): The user to add a tag for. + room_id(str): The room to add a tag for. + account_data_type(str): The type of account_data to add. + content(dict): A json object to associate with the tag. + Returns: + A deferred that completes once the account_data has been added. + """ + content_json = json.dumps(content) + + def add_account_data_txn(txn, next_id): + self._simple_upsert_txn( + txn, + table="room_account_data", + keyvalues={ + "user_id": user_id, + "room_id": room_id, + "account_data_type": account_data_type, + }, + values={ + "stream_id": next_id, + "content": content_json, + } + ) + self._update_max_stream_id(txn, next_id) + + with (yield self._account_data_id_gen.get_next(self)) as next_id: + yield self.runInteraction( + "add_room_account_data", add_account_data_txn, next_id + ) + + result = yield self._account_data_id_gen.get_max_token(self) + defer.returnValue(result) + + @defer.inlineCallbacks + def add_account_data_for_user(self, user_id, account_data_type, content): + """Add some account_data to a room for a user. + Args: + user_id(str): The user to add a tag for. + account_data_type(str): The type of account_data to add. + content(dict): A json object to associate with the tag. + Returns: + A deferred that completes once the account_data has been added. + """ + content_json = json.dumps(content) + + def add_account_data_txn(txn, next_id): + self._simple_upsert_txn( + txn, + table="account_data", + keyvalues={ + "user_id": user_id, + "account_data_type": account_data_type, + }, + values={ + "stream_id": next_id, + "content": content_json, + } + ) + self._update_max_stream_id(txn, next_id) + + with (yield self._account_data_id_gen.get_next(self)) as next_id: + yield self.runInteraction( + "add_user_account_data", add_account_data_txn, next_id + ) + + result = yield self._account_data_id_gen.get_max_token(self) + defer.returnValue(result) + + def _update_max_stream_id(self, txn, next_id): + """Update the max stream_id + + Args: + txn: The database cursor + next_id(int): The the revision to advance to. + """ + update_max_id_sql = ( + "UPDATE account_data_max_stream_id" + " SET stream_id = ?" + " WHERE stream_id < ?" + ) + txn.execute(update_max_id_sql, (next_id, next_id)) diff --git a/synapse/storage/events.py b/synapse/storage/events.py index 5d35ca90b9..fc5725097c 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -51,6 +51,14 @@ EVENT_QUEUE_TIMEOUT_S = 0.1 # Timeout when waiting for requests for events class EventsStore(SQLBaseStore): + EVENT_ORIGIN_SERVER_TS_NAME = "event_origin_server_ts" + + def __init__(self, hs): + super(EventsStore, self).__init__(hs) + self.register_background_update_handler( + self.EVENT_ORIGIN_SERVER_TS_NAME, self._background_reindex_origin_server_ts + ) + @defer.inlineCallbacks def persist_events(self, events_and_contexts, backfilled=False, is_new_state=True): @@ -365,6 +373,7 @@ class EventsStore(SQLBaseStore): "processed": True, "outlier": event.internal_metadata.is_outlier(), "content": encode_json(event.content).decode("UTF-8"), + "origin_server_ts": int(event.origin_server_ts), } for event, _ in events_and_contexts ], @@ -640,7 +649,7 @@ class EventsStore(SQLBaseStore): ] rows = self._new_transaction( - conn, "do_fetch", [], self._fetch_event_rows, event_ids + conn, "do_fetch", [], None, self._fetch_event_rows, event_ids ) row_dict = { @@ -964,3 +973,71 @@ class EventsStore(SQLBaseStore): ret = yield self.runInteraction("count_messages", _count_messages) defer.returnValue(ret) + + @defer.inlineCallbacks + def _background_reindex_origin_server_ts(self, progress, batch_size): + target_min_stream_id = progress["target_min_stream_id_inclusive"] + max_stream_id = progress["max_stream_id_exclusive"] + rows_inserted = progress.get("rows_inserted", 0) + + INSERT_CLUMP_SIZE = 1000 + + def reindex_search_txn(txn): + sql = ( + "SELECT stream_ordering, event_id FROM events" + " WHERE ? <= stream_ordering AND stream_ordering < ?" + " ORDER BY stream_ordering DESC" + " LIMIT ?" + ) + + txn.execute(sql, (target_min_stream_id, max_stream_id, batch_size)) + + rows = txn.fetchall() + if not rows: + return 0 + + min_stream_id = rows[-1][0] + event_ids = [row[1] for row in rows] + + events = self._get_events_txn(txn, event_ids) + + rows = [] + for event in events: + try: + event_id = event.event_id + origin_server_ts = event.origin_server_ts + except (KeyError, AttributeError): + # If the event is missing a necessary field then + # skip over it. + continue + + rows.append((origin_server_ts, event_id)) + + sql = ( + "UPDATE events SET origin_server_ts = ? WHERE event_id = ?" + ) + + for index in range(0, len(rows), INSERT_CLUMP_SIZE): + clump = rows[index:index + INSERT_CLUMP_SIZE] + txn.executemany(sql, clump) + + progress = { + "target_min_stream_id_inclusive": target_min_stream_id, + "max_stream_id_exclusive": min_stream_id, + "rows_inserted": rows_inserted + len(rows) + } + + self._background_update_progress_txn( + txn, self.EVENT_ORIGIN_SERVER_TS_NAME, progress + ) + + return len(rows) + + result = yield self.runInteraction( + self.EVENT_ORIGIN_SERVER_TS_NAME, reindex_search_txn + ) + + if not result: + yield self._end_background_update(self.EVENT_ORIGIN_SERVER_TS_NAME) + + defer.returnValue(result) diff --git a/synapse/storage/prepare_database.py b/synapse/storage/prepare_database.py index 9800fd4203..16eff62544 100644 --- a/synapse/storage/prepare_database.py +++ b/synapse/storage/prepare_database.py @@ -25,7 +25,7 @@ logger = logging.getLogger(__name__) # Remember to update this number every time a change is made to database # schema files, so the users will be informed on server restarts. -SCHEMA_VERSION = 26 +SCHEMA_VERSION = 27 dir_path = os.path.abspath(os.path.dirname(__file__)) diff --git a/synapse/storage/registration.py b/synapse/storage/registration.py index 2e5eddd259..09a05b08ef 100644 --- a/synapse/storage/registration.py +++ b/synapse/storage/registration.py @@ -258,10 +258,10 @@ class RegistrationStore(SQLBaseStore): @defer.inlineCallbacks def user_add_threepid(self, user_id, medium, address, validated_at, added_at): yield self._simple_upsert("user_threepids", { - "user_id": user_id, "medium": medium, "address": address, }, { + "user_id": user_id, "validated_at": validated_at, "added_at": added_at, }) diff --git a/synapse/storage/roommember.py b/synapse/storage/roommember.py index ae1ad56d9a..4e0e9ab59a 100644 --- a/synapse/storage/roommember.py +++ b/synapse/storage/roommember.py @@ -18,7 +18,7 @@ from twisted.internet import defer from collections import namedtuple from ._base import SQLBaseStore -from synapse.util.caches.descriptors import cached +from synapse.util.caches.descriptors import cached, cachedInlineCallbacks from synapse.api.constants import Membership from synapse.types import UserID @@ -121,7 +121,7 @@ class RoomMemberStore(SQLBaseStore): return self.get_rooms_for_user_where_membership_is( user_id, [Membership.INVITE] ).addCallback(lambda invites: self._get_events([ - invites.event_id for invite in invites + invite.event_id for invite in invites ])) def get_leave_and_ban_events_for_user(self, user_id): @@ -160,7 +160,7 @@ class RoomMemberStore(SQLBaseStore): def _get_rooms_for_user_where_membership_is_txn(self, txn, user_id, membership_list): - where_clause = "user_id = ? AND (%s)" % ( + where_clause = "user_id = ? AND (%s) AND forgotten = 0" % ( " OR ".join(["membership = ?" for _ in membership_list]), ) @@ -269,3 +269,70 @@ class RoomMemberStore(SQLBaseStore): ret = len(room_id_lists.pop(0).intersection(*room_id_lists)) > 0 defer.returnValue(ret) + + @defer.inlineCallbacks + def forget(self, user_id, room_id): + """Indicate that user_id wishes to discard history for room_id.""" + def f(txn): + sql = ( + "UPDATE" + " room_memberships" + " SET" + " forgotten = 1" + " WHERE" + " user_id = ?" + " AND" + " room_id = ?" + ) + txn.execute(sql, (user_id, room_id)) + yield self.runInteraction("forget_membership", f) + self.was_forgotten_at.invalidate_all() + self.did_forget.invalidate((user_id, room_id)) + + @cachedInlineCallbacks(num_args=2) + def did_forget(self, user_id, room_id): + """Returns whether user_id has elected to discard history for room_id. + + Returns False if they have since re-joined.""" + def f(txn): + sql = ( + "SELECT" + " COUNT(*)" + " FROM" + " room_memberships" + " WHERE" + " user_id = ?" + " AND" + " room_id = ?" + " AND" + " forgotten = 0" + ) + txn.execute(sql, (user_id, room_id)) + rows = txn.fetchall() + return rows[0][0] + count = yield self.runInteraction("did_forget_membership", f) + defer.returnValue(count == 0) + + @cachedInlineCallbacks(num_args=3) + def was_forgotten_at(self, user_id, room_id, event_id): + """Returns whether user_id has elected to discard history for room_id at event_id. + + event_id must be a membership event.""" + def f(txn): + sql = ( + "SELECT" + " forgotten" + " FROM" + " room_memberships" + " WHERE" + " user_id = ?" + " AND" + " room_id = ?" + " AND" + " event_id = ?" + ) + txn.execute(sql, (user_id, room_id, event_id)) + rows = txn.fetchall() + return rows[0][0] + forgot = yield self.runInteraction("did_forget_membership_at", f) + defer.returnValue(forgot == 1) diff --git a/synapse/storage/schema/delta/15/v15.sql b/synapse/storage/schema/delta/15/v15.sql index f5b2a08ca4..9523d2bcc3 100644 --- a/synapse/storage/schema/delta/15/v15.sql +++ b/synapse/storage/schema/delta/15/v15.sql @@ -1,23 +1,22 @@ -- Drop, copy & recreate pushers table to change unique key -- Also add access_token column at the same time CREATE TABLE IF NOT EXISTS pushers2 ( - id INTEGER PRIMARY KEY AUTOINCREMENT, + id BIGINT PRIMARY KEY, user_name TEXT NOT NULL, - access_token INTEGER DEFAULT NULL, - profile_tag varchar(32) NOT NULL, - kind varchar(8) NOT NULL, - app_id varchar(64) NOT NULL, - app_display_name varchar(64) NOT NULL, - device_display_name varchar(128) NOT NULL, - pushkey blob NOT NULL, + access_token BIGINT DEFAULT NULL, + profile_tag VARCHAR(32) NOT NULL, + kind VARCHAR(8) NOT NULL, + app_id VARCHAR(64) NOT NULL, + app_display_name VARCHAR(64) NOT NULL, + device_display_name VARCHAR(128) NOT NULL, + pushkey bytea NOT NULL, ts BIGINT NOT NULL, - lang varchar(8), - data blob, + lang VARCHAR(8), + data bytea, last_token TEXT, last_success BIGINT, failing_since BIGINT, - FOREIGN KEY(user_name) REFERENCES users(name), - UNIQUE (app_id, pushkey, user_name) + UNIQUE (app_id, pushkey) ); INSERT INTO pushers2 (id, user_name, profile_tag, kind, app_id, app_display_name, device_display_name, pushkey, ts, lang, data, last_token, last_success, failing_since) SELECT id, user_name, profile_tag, kind, app_id, app_display_name, device_display_name, pushkey, ts, lang, data, last_token, last_success, failing_since FROM pushers; diff --git a/synapse/storage/schema/delta/25/fts.py b/synapse/storage/schema/delta/25/fts.py index 5239d69073..ba48e43792 100644 --- a/synapse/storage/schema/delta/25/fts.py +++ b/synapse/storage/schema/delta/25/fts.py @@ -38,7 +38,7 @@ CREATE INDEX event_search_ev_ridx ON event_search(room_id); SQLITE_TABLE = ( - "CREATE VIRTUAL TABLE IF NOT EXISTS event_search" + "CREATE VIRTUAL TABLE event_search" " USING fts4 ( event_id, room_id, sender, key, value )" ) diff --git a/synapse/storage/schema/delta/27/account_data.sql b/synapse/storage/schema/delta/27/account_data.sql new file mode 100644 index 0000000000..9f25416005 --- /dev/null +++ b/synapse/storage/schema/delta/27/account_data.sql @@ -0,0 +1,36 @@ +/* Copyright 2015 OpenMarket Ltd + * + * 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. + */ + +CREATE TABLE IF NOT EXISTS account_data( + user_id TEXT NOT NULL, + account_data_type TEXT NOT NULL, -- The type of the account_data. + stream_id BIGINT NOT NULL, -- The version of the account_data. + content TEXT NOT NULL, -- The JSON content of the account_data + CONSTRAINT account_data_uniqueness UNIQUE (user_id, account_data_type) +); + + +CREATE TABLE IF NOT EXISTS room_account_data( + user_id TEXT NOT NULL, + room_id TEXT NOT NULL, + account_data_type TEXT NOT NULL, -- The type of the account_data. + stream_id BIGINT NOT NULL, -- The version of the account_data. + content TEXT NOT NULL, -- The JSON content of the account_data + CONSTRAINT room_account_data_uniqueness UNIQUE (user_id, room_id, account_data_type) +); + + +CREATE INDEX account_data_stream_id on account_data(user_id, stream_id); +CREATE INDEX room_account_data_stream_id on room_account_data(user_id, stream_id); diff --git a/synapse/storage/schema/delta/27/forgotten_memberships.sql b/synapse/storage/schema/delta/27/forgotten_memberships.sql new file mode 100644 index 0000000000..beeb8a288b --- /dev/null +++ b/synapse/storage/schema/delta/27/forgotten_memberships.sql @@ -0,0 +1,26 @@ +/* Copyright 2015 OpenMarket Ltd + * + * 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. + */ + +/* + * Keeps track of what rooms users have left and don't want to be able to + * access again. + * + * If all users on this server have left a room, we can delete the room + * entirely. + * + * This column should always contain either 0 or 1. + */ + + ALTER TABLE room_memberships ADD COLUMN forgotten INTEGER DEFAULT 0; diff --git a/synapse/storage/schema/delta/27/ts.py b/synapse/storage/schema/delta/27/ts.py new file mode 100644 index 0000000000..8d4a981975 --- /dev/null +++ b/synapse/storage/schema/delta/27/ts.py @@ -0,0 +1,57 @@ +# Copyright 2015 OpenMarket Ltd +# +# 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. + +import logging + +from synapse.storage.prepare_database import get_statements + +import ujson + +logger = logging.getLogger(__name__) + + +ALTER_TABLE = ( + "ALTER TABLE events ADD COLUMN origin_server_ts BIGINT;" + "CREATE INDEX events_ts ON events(origin_server_ts, stream_ordering);" +) + + +def run_upgrade(cur, database_engine, *args, **kwargs): + for statement in get_statements(ALTER_TABLE.splitlines()): + cur.execute(statement) + + cur.execute("SELECT MIN(stream_ordering) FROM events") + rows = cur.fetchall() + min_stream_id = rows[0][0] + + cur.execute("SELECT MAX(stream_ordering) FROM events") + rows = cur.fetchall() + max_stream_id = rows[0][0] + + if min_stream_id is not None and max_stream_id is not None: + progress = { + "target_min_stream_id_inclusive": min_stream_id, + "max_stream_id_exclusive": max_stream_id + 1, + "rows_inserted": 0, + } + progress_json = ujson.dumps(progress) + + sql = ( + "INSERT into background_updates (update_name, progress_json)" + " VALUES (?, ?)" + ) + + sql = database_engine.convert_param_style(sql) + + cur.execute(sql, ("event_origin_server_ts", progress_json)) diff --git a/synapse/storage/search.py b/synapse/storage/search.py index 380270b009..6cb5e73b6e 100644 --- a/synapse/storage/search.py +++ b/synapse/storage/search.py @@ -20,6 +20,7 @@ from synapse.api.errors import SynapseError from synapse.storage.engines import PostgresEngine, Sqlite3Engine import logging +import re logger = logging.getLogger(__name__) @@ -84,6 +85,11 @@ class SearchStore(BackgroundUpdateStore): # skip over it. continue + if not isinstance(value, basestring): + # If the event body, name or topic isn't a string + # then skip over it + continue + event_search_rows.append((event_id, room_id, key, value)) if isinstance(self.database_engine, PostgresEngine): @@ -139,6 +145,9 @@ class SearchStore(BackgroundUpdateStore): list of dicts """ clauses = [] + + search_query = search_query = _parse_query(self.database_engine, search_term) + args = [] # Make sure we don't explode because the person is in too many rooms. @@ -158,18 +167,36 @@ class SearchStore(BackgroundUpdateStore): "(%s)" % (" OR ".join(local_clauses),) ) + count_args = args + count_clauses = clauses + if isinstance(self.database_engine, PostgresEngine): sql = ( - "SELECT ts_rank_cd(vector, query) AS rank, room_id, event_id" - " FROM plainto_tsquery('english', ?) as query, event_search" - " WHERE vector @@ query" + "SELECT ts_rank_cd(vector, to_tsquery('english', ?)) AS rank," + " room_id, event_id" + " FROM event_search" + " WHERE vector @@ to_tsquery('english', ?)" + ) + args = [search_query, search_query] + args + + count_sql = ( + "SELECT room_id, count(*) as count FROM event_search" + " WHERE vector @@ to_tsquery('english', ?)" ) + count_args = [search_query] + count_args elif isinstance(self.database_engine, Sqlite3Engine): sql = ( "SELECT rank(matchinfo(event_search)) as rank, room_id, event_id" " FROM event_search" " WHERE value MATCH ?" ) + args = [search_query] + args + + count_sql = ( + "SELECT room_id, count(*) as count FROM event_search" + " WHERE value MATCH ?" + ) + count_args = [search_term] + count_args else: # This should be unreachable. raise Exception("Unrecognized database engine") @@ -177,12 +204,15 @@ class SearchStore(BackgroundUpdateStore): for clause in clauses: sql += " AND " + clause + for clause in count_clauses: + count_sql += " AND " + clause + # We add an arbitrary limit here to ensure we don't try to pull the # entire table from the database. sql += " ORDER BY rank DESC LIMIT 500" results = yield self._execute( - "search_msgs", self.cursor_to_dict, sql, *([search_term] + args) + "search_msgs", self.cursor_to_dict, sql, *args ) results = filter(lambda row: row["room_id"] in room_ids, results) @@ -194,21 +224,37 @@ class SearchStore(BackgroundUpdateStore): for ev in events } - defer.returnValue([ - { - "event": event_map[r["event_id"]], - "rank": r["rank"], - } - for r in results - if r["event_id"] in event_map - ]) + highlights = None + if isinstance(self.database_engine, PostgresEngine): + highlights = yield self._find_highlights_in_postgres(search_query, events) + + count_sql += " GROUP BY room_id" + + count_results = yield self._execute( + "search_rooms_count", self.cursor_to_dict, count_sql, *count_args + ) + + count = sum(row["count"] for row in count_results if row["room_id"] in room_ids) + + defer.returnValue({ + "results": [ + { + "event": event_map[r["event_id"]], + "rank": r["rank"], + } + for r in results + if r["event_id"] in event_map + ], + "highlights": highlights, + "count": count, + }) @defer.inlineCallbacks - def search_room(self, room_id, search_term, keys, limit, pagination_token=None): + def search_rooms(self, room_ids, search_term, keys, limit, pagination_token=None): """Performs a full text search over events with given keys. Args: - room_id (str): The room_id to search in + room_id (list): The room_ids to search in search_term (str): Search term to search for keys (list): List of keys to search in, currently supports "content.body", "content.name", "content.topic" @@ -218,7 +264,18 @@ class SearchStore(BackgroundUpdateStore): list of dicts """ clauses = [] - args = [search_term, room_id] + + search_query = search_query = _parse_query(self.database_engine, search_term) + + args = [] + + # Make sure we don't explode because the person is in too many rooms. + # We filter the results below regardless. + if len(room_ids) < 500: + clauses.append( + "room_id IN (%s)" % (",".join(["?"] * len(room_ids)),) + ) + args.extend(room_ids) local_clauses = [] for key in keys: @@ -229,28 +286,40 @@ class SearchStore(BackgroundUpdateStore): "(%s)" % (" OR ".join(local_clauses),) ) + # take copies of the current args and clauses lists, before adding + # pagination clauses to main query. + count_args = list(args) + count_clauses = list(clauses) + if pagination_token: try: - topo, stream = pagination_token.split(",") - topo = int(topo) + origin_server_ts, stream = pagination_token.split(",") + origin_server_ts = int(origin_server_ts) stream = int(stream) except: raise SynapseError(400, "Invalid pagination token") clauses.append( - "(topological_ordering < ?" - " OR (topological_ordering = ? AND stream_ordering < ?))" + "(origin_server_ts < ?" + " OR (origin_server_ts = ? AND stream_ordering < ?))" ) - args.extend([topo, topo, stream]) + args.extend([origin_server_ts, origin_server_ts, stream]) if isinstance(self.database_engine, PostgresEngine): sql = ( - "SELECT ts_rank_cd(vector, query) as rank," - " topological_ordering, stream_ordering, room_id, event_id" - " FROM plainto_tsquery('english', ?) as query, event_search" + "SELECT ts_rank_cd(vector, to_tsquery('english', ?)) as rank," + " origin_server_ts, stream_ordering, room_id, event_id" + " FROM event_search" " NATURAL JOIN events" - " WHERE vector @@ query AND room_id = ?" + " WHERE vector @@ to_tsquery('english', ?) AND " + ) + args = [search_query, search_query] + args + + count_sql = ( + "SELECT room_id, count(*) as count FROM event_search" + " WHERE vector @@ to_tsquery('english', ?) AND " ) + count_args = [search_query] + count_args elif isinstance(self.database_engine, Sqlite3Engine): # We use CROSS JOIN here to ensure we use the right indexes. # https://sqlite.org/optoverview.html#crossjoin @@ -262,24 +331,31 @@ class SearchStore(BackgroundUpdateStore): # MATCH unless it uses the full text search index sql = ( "SELECT rank(matchinfo) as rank, room_id, event_id," - " topological_ordering, stream_ordering" + " origin_server_ts, stream_ordering" " FROM (SELECT key, event_id, matchinfo(event_search) as matchinfo" " FROM event_search" " WHERE value MATCH ?" " )" " CROSS JOIN events USING (event_id)" - " WHERE room_id = ?" + " WHERE " ) + args = [search_query] + args + + count_sql = ( + "SELECT room_id, count(*) as count FROM event_search" + " WHERE value MATCH ? AND " + ) + count_args = [search_term] + count_args else: # This should be unreachable. raise Exception("Unrecognized database engine") - for clause in clauses: - sql += " AND " + clause + sql += " AND ".join(clauses) + count_sql += " AND ".join(count_clauses) # We add an arbitrary limit here to ensure we don't try to pull the # entire table from the database. - sql += " ORDER BY topological_ordering DESC, stream_ordering DESC LIMIT ?" + sql += " ORDER BY origin_server_ts DESC, stream_ordering DESC LIMIT ?" args.append(limit) @@ -287,6 +363,8 @@ class SearchStore(BackgroundUpdateStore): "search_rooms", self.cursor_to_dict, sql, *args ) + results = filter(lambda row: row["room_id"] in room_ids, results) + events = yield self._get_events([r["event_id"] for r in results]) event_map = { @@ -294,14 +372,119 @@ class SearchStore(BackgroundUpdateStore): for ev in events } - defer.returnValue([ - { - "event": event_map[r["event_id"]], - "rank": r["rank"], - "pagination_token": "%s,%s" % ( - r["topological_ordering"], r["stream_ordering"] - ), - } - for r in results - if r["event_id"] in event_map - ]) + highlights = None + if isinstance(self.database_engine, PostgresEngine): + highlights = yield self._find_highlights_in_postgres(search_query, events) + + count_sql += " GROUP BY room_id" + + count_results = yield self._execute( + "search_rooms_count", self.cursor_to_dict, count_sql, *count_args + ) + + count = sum(row["count"] for row in count_results if row["room_id"] in room_ids) + + defer.returnValue({ + "results": [ + { + "event": event_map[r["event_id"]], + "rank": r["rank"], + "pagination_token": "%s,%s" % ( + r["origin_server_ts"], r["stream_ordering"] + ), + } + for r in results + if r["event_id"] in event_map + ], + "highlights": highlights, + "count": count, + }) + + def _find_highlights_in_postgres(self, search_query, events): + """Given a list of events and a search term, return a list of words + that match from the content of the event. + + This is used to give a list of words that clients can match against to + highlight the matching parts. + + Args: + search_query (str) + events (list): A list of events + + Returns: + deferred : A set of strings. + """ + def f(txn): + highlight_words = set() + for event in events: + # As a hack we simply join values of all possible keys. This is + # fine since we're only using them to find possible highlights. + values = [] + for key in ("body", "name", "topic"): + v = event.content.get(key, None) + if v: + values.append(v) + + if not values: + continue + + value = " ".join(values) + + # We need to find some values for StartSel and StopSel that + # aren't in the value so that we can pick results out. + start_sel = "<" + stop_sel = ">" + + while start_sel in value: + start_sel += "<" + while stop_sel in value: + stop_sel += ">" + + query = "SELECT ts_headline(?, to_tsquery('english', ?), %s)" % ( + _to_postgres_options({ + "StartSel": start_sel, + "StopSel": stop_sel, + "MaxFragments": "50", + }) + ) + txn.execute(query, (value, search_query,)) + headline, = txn.fetchall()[0] + + # Now we need to pick the possible highlights out of the haedline + # result. + matcher_regex = "%s(.*?)%s" % ( + re.escape(start_sel), + re.escape(stop_sel), + ) + + res = re.findall(matcher_regex, headline) + highlight_words.update([r.lower() for r in res]) + + return highlight_words + + return self.runInteraction("_find_highlights", f) + + +def _to_postgres_options(options_dict): + return "'%s'" % ( + ",".join("%s=%s" % (k, v) for k, v in options_dict.items()), + ) + + +def _parse_query(database_engine, search_term): + """Takes a plain unicode string from the user and converts it into a form + that can be passed to database. + We use this so that we can add prefix matching, which isn't something + that is supported by default. + """ + + # Pull out the individual words, discarding any non-word characters. + results = re.findall(r"([\w\-]+)", search_term, re.UNICODE) + + if isinstance(database_engine, PostgresEngine): + return " & ".join(result + ":*" for result in results) + elif isinstance(database_engine, Sqlite3Engine): + return " & ".join(result + "*" for result in results) + else: + # This should be unreachable. + raise Exception("Unrecognized database engine") diff --git a/synapse/storage/tags.py b/synapse/storage/tags.py index f6d826cc59..f520f60c6c 100644 --- a/synapse/storage/tags.py +++ b/synapse/storage/tags.py @@ -48,8 +48,8 @@ class TagsStore(SQLBaseStore): Args: user_id(str): The user to get the tags for. Returns: - A deferred dict mapping from room_id strings to lists of tag - strings. + A deferred dict mapping from room_id strings to dicts mapping from + tag strings to tag content. """ deferred = self._simple_select_list( |