synapse-product/synapse/storage/databases/main/search.py

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

783 lines
28 KiB
Python
Raw Normal View History

2016-01-06 23:26:29 -05:00
# Copyright 2015, 2016 OpenMarket Ltd
2015-10-09 10:48:31 -04:00
#
# 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.
import logging
import re
from typing import TYPE_CHECKING, Any, Collection, Iterable, List, Optional, Set
2015-10-09 10:48:31 -04:00
import attr
2015-11-05 11:10:54 -05:00
from synapse.api.errors import SynapseError
from synapse.events import EventBase
from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause
from synapse.storage.database import (
DatabasePool,
LoggingDatabaseConnection,
LoggingTransaction,
)
from synapse.storage.databases.main.events_worker import EventRedactBehaviour
2015-10-16 09:37:14 -04:00
from synapse.storage.engines import PostgresEngine, Sqlite3Engine
from synapse.types import JsonDict
2015-10-09 10:48:31 -04:00
if TYPE_CHECKING:
from synapse.server import HomeServer
2015-11-04 12:57:44 -05:00
logger = logging.getLogger(__name__)
@attr.s(slots=True, frozen=True, auto_attribs=True)
class SearchEntry:
key: str
value: str
event_id: str
room_id: str
stream_ordering: Optional[int]
origin_server_ts: int
2015-11-04 12:57:44 -05:00
def _clean_value_for_search(value: str) -> str:
"""
Replaces any null code points in the string with spaces as
Postgres and SQLite do not like the insertion of strings with
null code points into the full-text search tables.
"""
return value.replace("\u0000", " ")
class SearchWorkerStore(SQLBaseStore):
def store_search_entries_txn(
self, txn: LoggingTransaction, entries: Iterable[SearchEntry]
) -> None:
"""Add entries to the search table
Args:
txn:
entries: entries to be added to the table
"""
if not self.hs.config.server.enable_search:
return
if isinstance(self.database_engine, PostgresEngine):
sql = (
"INSERT INTO event_search"
" (event_id, room_id, key, vector, stream_ordering, origin_server_ts)"
" VALUES (?,?,?,to_tsvector('english', ?),?,?)"
)
args1 = (
(
entry.event_id,
entry.room_id,
entry.key,
_clean_value_for_search(entry.value),
entry.stream_ordering,
entry.origin_server_ts,
)
for entry in entries
)
txn.execute_batch(sql, args1)
elif isinstance(self.database_engine, Sqlite3Engine):
sql = (
"INSERT INTO event_search (event_id, room_id, key, value)"
" VALUES (?,?,?,?)"
)
args2 = (
(
entry.event_id,
entry.room_id,
entry.key,
_clean_value_for_search(entry.value),
)
for entry in entries
)
txn.execute_batch(sql, args2)
else:
# This should be unreachable.
raise Exception("Unrecognized database engine")
class SearchBackgroundUpdateStore(SearchWorkerStore):
EVENT_SEARCH_UPDATE_NAME = "event_search"
2016-04-21 11:41:39 -04:00
EVENT_SEARCH_ORDER_UPDATE_NAME = "event_search_order"
EVENT_SEARCH_USE_GIST_POSTGRES_NAME = "event_search_postgres_gist"
2018-01-09 11:37:48 -05:00
EVENT_SEARCH_USE_GIN_POSTGRES_NAME = "event_search_postgres_gin"
EVENT_SEARCH_DELETE_NON_STRINGS = "event_search_sqlite_delete_non_strings"
def __init__(
self,
database: DatabasePool,
db_conn: LoggingDatabaseConnection,
hs: "HomeServer",
):
super().__init__(database, db_conn, hs)
self.db_pool.updates.register_background_update_handler(
self.EVENT_SEARCH_UPDATE_NAME, self._background_reindex_search
)
self.db_pool.updates.register_background_update_handler(
self.EVENT_SEARCH_ORDER_UPDATE_NAME, self._background_reindex_search_order
2016-04-21 11:41:39 -04:00
)
# we used to have a background update to turn the GIN index into a
# GIST one; we no longer do that (obviously) because we actually want
# a GIN index. However, it's possible that some people might still have
# the background update queued, so we register a handler to clear the
# background update.
self.db_pool.updates.register_noop_background_update(
self.EVENT_SEARCH_USE_GIST_POSTGRES_NAME
)
self.db_pool.updates.register_background_update_handler(
self.EVENT_SEARCH_USE_GIN_POSTGRES_NAME, self._background_reindex_gin_search
)
self.db_pool.updates.register_background_update_handler(
self.EVENT_SEARCH_DELETE_NON_STRINGS, self._background_delete_non_strings
)
async def _background_reindex_search(self, progress, batch_size):
# we work through the events table from highest stream id to lowest
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)
TYPES = ["m.room.name", "m.room.message", "m.room.topic"]
def reindex_search_txn(txn):
sql = (
2018-03-29 18:05:33 -04:00
"SELECT stream_ordering, event_id, room_id, type, json, "
" origin_server_ts FROM events"
2018-04-25 10:32:04 -04:00
" JOIN event_json USING (room_id, event_id)"
" WHERE ? <= stream_ordering AND stream_ordering < ?"
" AND (%s)"
" ORDER BY stream_ordering DESC"
" LIMIT ?"
2015-11-11 08:59:40 -05:00
) % (" OR ".join("type = '%s'" % (t,) for t in TYPES),)
2015-11-11 08:59:40 -05:00
txn.execute(sql, (target_min_stream_id, max_stream_id, batch_size))
# we could stream straight from the results into
# store_search_entries_txn with a generator function, but that
# would mean having two cursors open on the database at once.
# Instead we just build a list of results.
rows = self.db_pool.cursor_to_dict(txn)
if not rows:
2015-11-11 08:59:40 -05:00
return 0
2016-06-03 12:12:48 -04:00
min_stream_id = rows[-1]["stream_ordering"]
event_search_rows = []
2016-06-03 12:12:48 -04:00
for row in rows:
try:
2016-06-03 12:12:48 -04:00
event_id = row["event_id"]
room_id = row["room_id"]
etype = row["type"]
stream_ordering = row["stream_ordering"]
origin_server_ts = row["origin_server_ts"]
2016-06-03 12:12:48 -04:00
try:
event_json = db_to_json(row["json"])
2018-03-29 18:05:33 -04:00
content = event_json["content"]
except Exception:
2016-06-03 12:12:48 -04:00
continue
if etype == "m.room.message":
key = "content.body"
value = content["body"]
2016-06-03 12:12:48 -04:00
elif etype == "m.room.topic":
key = "content.topic"
value = content["topic"]
2016-06-03 12:12:48 -04:00
elif etype == "m.room.name":
key = "content.name"
value = content["name"]
else:
raise Exception("unexpected event type %s" % etype)
except (KeyError, AttributeError):
# If the event is missing a necessary field then
# skip over it.
continue
if not isinstance(value, str):
# If the event body, name or topic isn't a string
# then skip over it
continue
event_search_rows.append(
SearchEntry(
key=key,
value=value,
event_id=event_id,
room_id=room_id,
stream_ordering=stream_ordering,
origin_server_ts=origin_server_ts,
)
)
self.store_search_entries_txn(txn, event_search_rows)
progress = {
"target_min_stream_id_inclusive": target_min_stream_id,
"max_stream_id_exclusive": min_stream_id,
"rows_inserted": rows_inserted + len(event_search_rows),
}
self.db_pool.updates._background_update_progress_txn(
txn, self.EVENT_SEARCH_UPDATE_NAME, progress
)
return len(event_search_rows)
if self.hs.config.server.enable_search:
result = await self.db_pool.runInteraction(
self.EVENT_SEARCH_UPDATE_NAME, reindex_search_txn
)
else:
# Don't index anything if search is not enabled.
result = 0
2015-11-11 08:59:40 -05:00
if not result:
await self.db_pool.updates._end_background_update(
self.EVENT_SEARCH_UPDATE_NAME
)
return result
async def _background_reindex_gin_search(self, progress, batch_size):
"""This handles old synapses which used GIST indexes, if any;
2018-01-09 11:55:51 -05:00
converting them back to be GIN as per the actual schema.
"""
2018-01-09 11:37:48 -05:00
def create_index(conn):
conn.rollback()
# we have to set autocommit, because postgres refuses to
# CREATE INDEX CONCURRENTLY without it.
conn.set_session(autocommit=True)
2018-01-09 11:55:51 -05:00
try:
c = conn.cursor()
# if we skipped the conversion to GIST, we may already/still
# have an event_search_fts_idx; unfortunately postgres 9.4
# doesn't support CREATE INDEX IF EXISTS so we just catch the
# exception and ignore it.
import psycopg2
try:
c.execute(
"CREATE INDEX CONCURRENTLY event_search_fts_idx"
" ON event_search USING GIN (vector)"
)
except psycopg2.ProgrammingError as e:
logger.warning(
"Ignoring error %r when trying to switch from GIST to GIN", e
)
# we should now be able to delete the GIST index.
c.execute("DROP INDEX IF EXISTS event_search_fts_idx_gist")
finally:
2018-01-09 11:55:51 -05:00
conn.set_session(autocommit=False)
if isinstance(self.database_engine, PostgresEngine):
await self.db_pool.runWithConnection(create_index)
await self.db_pool.updates._end_background_update(
self.EVENT_SEARCH_USE_GIN_POSTGRES_NAME
)
return 1
async def _background_reindex_search_order(self, progress, batch_size):
2016-04-21 11:41:39 -04:00
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)
2016-04-21 12:16:11 -04:00
have_added_index = progress["have_added_indexes"]
2016-04-21 11:41:39 -04:00
2016-04-21 12:19:25 -04:00
if not have_added_index:
2016-04-21 12:19:25 -04:00
def create_index(conn):
conn.rollback()
conn.set_session(autocommit=True)
c = conn.cursor()
2016-04-21 13:09:48 -04:00
# We create with NULLS FIRST so that when we search *backwards*
# we get the ones with non null origin_server_ts *first*
2016-04-21 12:19:25 -04:00
c.execute(
2016-04-21 12:16:11 -04:00
"CREATE INDEX CONCURRENTLY event_search_room_order ON event_search("
2016-04-21 13:09:48 -04:00
"room_id, origin_server_ts NULLS FIRST, stream_ordering NULLS FIRST)"
2016-04-21 12:16:11 -04:00
)
2016-04-21 12:19:25 -04:00
c.execute(
2016-04-21 12:16:11 -04:00
"CREATE INDEX CONCURRENTLY event_search_order ON event_search("
2016-04-21 13:09:48 -04:00
"origin_server_ts NULLS FIRST, stream_ordering NULLS FIRST)"
2016-04-21 12:16:11 -04:00
)
2016-04-21 12:19:25 -04:00
conn.set_session(autocommit=False)
await self.db_pool.runWithConnection(create_index)
2016-04-21 12:39:24 -04:00
pg = dict(progress)
pg["have_added_indexes"] = True
2016-04-21 12:39:24 -04:00
await self.db_pool.runInteraction(
self.EVENT_SEARCH_ORDER_UPDATE_NAME,
self.db_pool.updates._background_update_progress_txn,
self.EVENT_SEARCH_ORDER_UPDATE_NAME,
pg,
)
2016-04-21 12:16:11 -04:00
2016-04-21 12:19:25 -04:00
def reindex_search_txn(txn):
2016-04-21 13:01:49 -04:00
sql = (
2016-04-21 13:02:36 -04:00
"UPDATE event_search AS es SET stream_ordering = e.stream_ordering,"
" origin_server_ts = e.origin_server_ts"
" FROM events AS e"
2016-04-21 13:01:49 -04:00
" WHERE e.event_id = es.event_id"
" AND ? <= e.stream_ordering AND e.stream_ordering < ?"
2016-04-21 13:01:49 -04:00
" RETURNING es.stream_ordering"
)
2016-04-21 11:41:39 -04:00
min_stream_id = max_stream_id - batch_size
txn.execute(sql, (min_stream_id, max_stream_id))
2016-04-21 11:41:39 -04:00
rows = txn.fetchall()
if min_stream_id < target_min_stream_id:
# We've recached the end.
return len(rows), False
2016-04-21 11:41:39 -04:00
progress = {
"target_min_stream_id_inclusive": target_min_stream_id,
"max_stream_id_exclusive": min_stream_id,
2016-04-21 12:16:11 -04:00
"rows_inserted": rows_inserted + len(rows),
2016-04-21 12:49:00 -04:00
"have_added_indexes": True,
2016-04-21 11:41:39 -04:00
}
self.db_pool.updates._background_update_progress_txn(
2016-04-21 11:41:39 -04:00
txn, self.EVENT_SEARCH_ORDER_UPDATE_NAME, progress
)
return len(rows), True
2016-04-21 11:41:39 -04:00
num_rows, finished = await self.db_pool.runInteraction(
2016-04-21 11:41:39 -04:00
self.EVENT_SEARCH_ORDER_UPDATE_NAME, reindex_search_txn
)
if not finished:
await self.db_pool.updates._end_background_update(
self.EVENT_SEARCH_ORDER_UPDATE_NAME
)
2016-04-21 11:41:39 -04:00
return num_rows
2016-04-21 11:41:39 -04:00
async def _background_delete_non_strings(
self, progress: JsonDict, batch_size: int
) -> int:
"""Deletes rows with non-string `value`s from `event_search` if using sqlite.
Prior to Synapse 1.44.0, malformed events received over federation could cause integers
to be inserted into the `event_search` table when using sqlite.
"""
def delete_non_strings_txn(txn: LoggingTransaction) -> None:
txn.execute("DELETE FROM event_search WHERE typeof(value) != 'text'")
await self.db_pool.runInteraction(
self.EVENT_SEARCH_DELETE_NON_STRINGS, delete_non_strings_txn
)
await self.db_pool.updates._end_background_update(
self.EVENT_SEARCH_DELETE_NON_STRINGS
)
return 1
class SearchStore(SearchBackgroundUpdateStore):
def __init__(
self,
database: DatabasePool,
db_conn: LoggingDatabaseConnection,
hs: "HomeServer",
):
super().__init__(database, db_conn, hs)
async def search_msgs(
self, room_ids: Collection[str], search_term: str, keys: Iterable[str]
) -> JsonDict:
2015-10-16 11:46:48 -04:00
"""Performs a full text search over events with given keys.
2015-10-16 06:28:12 -04:00
Args:
room_ids: List of room ids to search in
search_term: Search term to search for
keys: List of keys to search in, currently supports
2015-10-16 11:46:48 -04:00
"content.body", "content.name", "content.topic"
2015-10-16 06:28:12 -04:00
Returns:
Dictionary of results
2015-10-16 06:28:12 -04:00
"""
2015-10-09 10:48:31 -04:00
clauses = []
2015-12-02 08:28:13 -05:00
search_query = _parse_query(self.database_engine, search_term)
2015-12-02 08:28:13 -05:00
args: List[Any] = []
2015-10-09 10:48:31 -04:00
2015-10-22 08:25:22 -04:00
# Make sure we don't explode because the person is in too many rooms.
2015-10-22 11:54:56 -04:00
# We filter the results below regardless.
2015-10-22 11:18:35 -04:00
if len(room_ids) < 500:
2019-10-10 10:35:46 -04:00
clause, args = make_in_list_sql_clause(
self.database_engine, "room_id", room_ids
2019-10-02 14:07:07 -04:00
)
2019-10-10 10:35:46 -04:00
clauses = [clause]
local_clauses = []
for key in keys:
local_clauses.append("key = ?")
args.append(key)
clauses.append("(%s)" % (" OR ".join(local_clauses),))
2015-10-09 10:48:31 -04:00
count_args = args
count_clauses = clauses
2015-10-13 08:47:50 -04:00
if isinstance(self.database_engine, PostgresEngine):
sql = (
2015-12-11 06:12:57 -05:00
"SELECT ts_rank_cd(vector, to_tsquery('english', ?)) AS rank,"
" room_id, event_id"
" FROM event_search"
" WHERE vector @@ to_tsquery('english', ?)"
2015-10-13 08:47:50 -04:00
)
2015-12-11 06:12:57 -05:00
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
2015-10-16 09:37:14 -04:00
elif isinstance(self.database_engine, Sqlite3Engine):
2015-10-13 08:47:50 -04:00
sql = (
2015-10-23 08:23:48 -04:00
"SELECT rank(matchinfo(event_search)) as rank, room_id, event_id"
" FROM event_search"
2015-10-13 08:47:50 -04:00
" WHERE value MATCH ?"
)
2015-12-11 06:12:57 -05:00
args = [search_query] + args
count_sql = (
"SELECT room_id, count(*) as count FROM event_search"
" WHERE value MATCH ?"
)
count_args = [search_term] + count_args
2015-10-16 09:37:14 -04:00
else:
# This should be unreachable.
raise Exception("Unrecognized database engine")
2015-10-09 10:48:31 -04:00
for clause in clauses:
sql += " AND " + clause
for clause in count_clauses:
count_sql += " AND " + clause
2015-10-16 06:24:02 -04:00
# We add an arbitrary limit here to ensure we don't try to pull the
# entire table from the database.
2015-10-13 10:50:56 -04:00
sql += " ORDER BY rank DESC LIMIT 500"
2015-10-09 10:48:31 -04:00
results = await self.db_pool.execute(
"search_msgs", self.db_pool.cursor_to_dict, sql, *args
)
2015-10-09 10:48:31 -04:00
results = list(filter(lambda row: row["room_id"] in room_ids, results))
2015-10-22 11:54:56 -04:00
# We set redact_behaviour to block here to prevent redacted events being returned in
# search results (which is a data leak)
events = await self.get_events_as_list( # type: ignore[attr-defined]
[r["event_id"] for r in results],
redact_behaviour=EventRedactBehaviour.block,
)
2015-10-09 10:48:31 -04:00
event_map = {ev.event_id: ev for ev in events}
2015-10-09 10:48:31 -04:00
highlights = None
if isinstance(self.database_engine, PostgresEngine):
highlights = await self._find_highlights_in_postgres(search_query, events)
count_sql += " GROUP BY room_id"
count_results = await self.db_pool.execute(
"search_rooms_count", self.db_pool.cursor_to_dict, count_sql, *count_args
)
count = sum(row["count"] for row in count_results if row["room_id"] in room_ids)
return {
"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,
}
2015-11-04 12:57:44 -05:00
async def search_rooms(
self,
room_ids: Collection[str],
search_term: str,
keys: Iterable[str],
limit: int,
pagination_token: Optional[str] = None,
) -> JsonDict:
2015-11-04 12:57:44 -05:00
"""Performs a full text search over events with given keys.
Args:
room_ids: The room_ids to search in
search_term: Search term to search for
keys: List of keys to search in, currently supports "content.body",
"content.name", "content.topic"
pagination_token: A pagination token previously returned
2015-11-04 12:57:44 -05:00
Returns:
Each match as a dictionary.
2015-11-04 12:57:44 -05:00
"""
clauses = []
2015-12-02 06:38:51 -05:00
search_query = _parse_query(self.database_engine, search_term)
2015-12-02 08:28:13 -05:00
args: List[Any] = []
# 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:
2019-10-10 10:35:46 -04:00
clause, args = make_in_list_sql_clause(
self.database_engine, "room_id", room_ids
2019-10-02 14:07:07 -04:00
)
2019-10-10 10:35:46 -04:00
clauses = [clause]
2015-11-04 12:57:44 -05:00
local_clauses = []
for key in keys:
local_clauses.append("key = ?")
args.append(key)
clauses.append("(%s)" % (" OR ".join(local_clauses),))
2015-11-04 12:57:44 -05:00
# take copies of the current args and clauses lists, before adding
# pagination clauses to main query.
count_args = list(args)
count_clauses = list(clauses)
2015-11-04 12:57:44 -05:00
if pagination_token:
2015-11-05 11:10:54 -05:00
try:
origin_server_ts_str, stream_str = pagination_token.split(",")
origin_server_ts = int(origin_server_ts_str)
stream = int(stream_str)
except Exception:
2015-11-05 11:10:54 -05:00
raise SynapseError(400, "Invalid pagination token")
2015-11-04 12:57:44 -05:00
clauses.append(
"(origin_server_ts < ?"
" OR (origin_server_ts = ? AND stream_ordering < ?))"
2015-11-04 12:57:44 -05:00
)
args.extend([origin_server_ts, origin_server_ts, stream])
2015-11-04 12:57:44 -05:00
if isinstance(self.database_engine, PostgresEngine):
sql = (
2015-12-11 06:12:57 -05:00
"SELECT ts_rank_cd(vector, to_tsquery('english', ?)) as rank,"
" origin_server_ts, stream_ordering, room_id, event_id"
2015-12-11 06:12:57 -05:00
" FROM event_search"
" WHERE vector @@ to_tsquery('english', ?) AND "
2015-11-04 12:57:44 -05:00
)
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
2015-11-04 12:57:44 -05:00
elif isinstance(self.database_engine, Sqlite3Engine):
2015-11-12 10:19:56 -05:00
# We use CROSS JOIN here to ensure we use the right indexes.
# https://sqlite.org/optoverview.html#crossjoin
2015-11-12 10:33:47 -05:00
#
# We want to use the full text search index on event_search to
# extract all possible matches first, then lookup those matches
# in the events table to get the topological ordering. We need
2015-11-12 10:36:43 -05:00
# to use the indexes in this order because sqlite refuses to
2015-11-12 10:33:47 -05:00
# MATCH unless it uses the full text search index
2015-11-04 12:57:44 -05:00
sql = (
"SELECT rank(matchinfo) as rank, room_id, event_id,"
" origin_server_ts, stream_ordering"
2015-11-12 10:09:45 -05:00
" FROM (SELECT key, event_id, matchinfo(event_search) as matchinfo"
" FROM event_search"
" WHERE value MATCH ?"
" )"
" CROSS JOIN events USING (event_id)"
" WHERE "
2015-11-04 12:57:44 -05:00
)
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
2015-11-04 12:57:44 -05:00
else:
# This should be unreachable.
raise Exception("Unrecognized database engine")
sql += " AND ".join(clauses)
count_sql += " AND ".join(count_clauses)
2015-11-04 12:57:44 -05:00
# We add an arbitrary limit here to ensure we don't try to pull the
# entire table from the database.
2016-04-21 13:09:48 -04:00
if isinstance(self.database_engine, PostgresEngine):
sql += (
" ORDER BY origin_server_ts DESC NULLS LAST,"
" stream_ordering DESC NULLS LAST LIMIT ?"
)
elif isinstance(self.database_engine, Sqlite3Engine):
sql += " ORDER BY origin_server_ts DESC, stream_ordering DESC LIMIT ?"
else:
raise Exception("Unrecognized database engine")
2015-11-04 12:57:44 -05:00
args.append(limit)
results = await self.db_pool.execute(
"search_rooms", self.db_pool.cursor_to_dict, sql, *args
)
2015-11-04 12:57:44 -05:00
results = list(filter(lambda row: row["room_id"] in room_ids, results))
# We set redact_behaviour to block here to prevent redacted events being returned in
# search results (which is a data leak)
events = await self.get_events_as_list( # type: ignore[attr-defined]
[r["event_id"] for r in results],
redact_behaviour=EventRedactBehaviour.block,
)
2015-11-04 12:57:44 -05:00
event_map = {ev.event_id: ev for ev in events}
2015-11-04 12:57:44 -05:00
highlights = None
if isinstance(self.database_engine, PostgresEngine):
highlights = await self._find_highlights_in_postgres(search_query, events)
count_sql += " GROUP BY room_id"
count_results = await self.db_pool.execute(
"search_rooms_count", self.db_pool.cursor_to_dict, count_sql, *count_args
)
count = sum(row["count"] for row in count_results if row["room_id"] in room_ids)
return {
"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,
}
async def _find_highlights_in_postgres(
self, search_query: str, events: List[EventBase]
) -> Set[str]:
"""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
events: A list of events
Returns:
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:
v = _clean_value_for_search(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 += ">"
2015-12-02 08:28:13 -05:00
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 await self.db_pool.runInteraction("_find_highlights", f)
def _to_postgres_options(options_dict):
return "'%s'" % (",".join("%s=%s" % (k, v) for k, v in options_dict.items()),)
2015-12-02 06:38:51 -05:00
def _parse_query(database_engine, search_term):
2015-12-02 06:38:51 -05:00
"""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.
2015-12-02 06:38:51 -05:00
"""
# Pull out the individual words, discarding any non-word characters.
2015-12-02 06:38:51 -05:00
results = re.findall(r"([\w\-]+)", search_term, re.UNICODE)
if isinstance(database_engine, PostgresEngine):
return " & ".join(result + ":*" for result in results)
2015-12-02 08:50:43 -05:00
elif isinstance(database_engine, Sqlite3Engine):
return " & ".join(result + "*" for result in results)
2015-12-02 08:50:43 -05:00
else:
# This should be unreachable.
raise Exception("Unrecognized database engine")