summary refs log tree commit diff
path: root/synapse/storage/search.py
blob: 00f89ff02f1cf6d33f209110644765255c75b0a0 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
# -*- coding: utf-8 -*-
# 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.

from twisted.internet import defer

from .background_updates import BackgroundUpdateStore
from synapse.api.errors import SynapseError
from synapse.storage.engines import PostgresEngine, Sqlite3Engine

import logging
import re


logger = logging.getLogger(__name__)


class SearchStore(BackgroundUpdateStore):

    EVENT_SEARCH_UPDATE_NAME = "event_search"

    def __init__(self, hs):
        super(SearchStore, self).__init__(hs)
        self.register_background_update_handler(
            self.EVENT_SEARCH_UPDATE_NAME, self._background_reindex_search
        )

    @defer.inlineCallbacks
    def _background_reindex_search(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
        TYPES = ["m.room.name", "m.room.message", "m.room.topic"]

        def reindex_search_txn(txn):
            sql = (
                "SELECT stream_ordering, event_id FROM events"
                " WHERE ? <= stream_ordering AND stream_ordering < ?"
                " AND (%s)"
                " ORDER BY stream_ordering DESC"
                " LIMIT ?"
            ) % (" OR ".join("type = '%s'" % (t,) for t in TYPES),)

            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)

            event_search_rows = []
            for event in events:
                try:
                    event_id = event.event_id
                    room_id = event.room_id
                    content = event.content
                    if event.type == "m.room.message":
                        key = "content.body"
                        value = content["body"]
                    elif event.type == "m.room.topic":
                        key = "content.topic"
                        value = content["topic"]
                    elif event.type == "m.room.name":
                        key = "content.name"
                        value = content["name"]
                except (KeyError, AttributeError):
                    # If the event is missing a necessary field then
                    # skip over it.
                    continue

                event_search_rows.append((event_id, room_id, key, value))

            if isinstance(self.database_engine, PostgresEngine):
                sql = (
                    "INSERT INTO event_search (event_id, room_id, key, vector)"
                    " VALUES (?,?,?,to_tsvector('english', ?))"
                )
            elif isinstance(self.database_engine, Sqlite3Engine):
                sql = (
                    "INSERT INTO event_search (event_id, room_id, key, value)"
                    " VALUES (?,?,?,?)"
                )
            else:
                # This should be unreachable.
                raise Exception("Unrecognized database engine")

            for index in range(0, len(event_search_rows), INSERT_CLUMP_SIZE):
                clump = event_search_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(event_search_rows)
            }

            self._background_update_progress_txn(
                txn, self.EVENT_SEARCH_UPDATE_NAME, progress
            )

            return len(event_search_rows)

        result = yield self.runInteraction(
            self.EVENT_SEARCH_UPDATE_NAME, reindex_search_txn
        )

        if not result:
            yield self._end_background_update(self.EVENT_SEARCH_UPDATE_NAME)

        defer.returnValue(result)

    @defer.inlineCallbacks
    def search_msgs(self, room_ids, search_term, keys):
        """Performs a full text search over events with given keys.

        Args:
            room_ids (list): List of 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"

        Returns:
            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.
        # 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:
            local_clauses.append("key = ?")
            args.append(key)

        clauses.append(
            "(%s)" % (" OR ".join(local_clauses),)
        )

        count_args = args
        count_clauses = clauses

        if isinstance(self.database_engine, PostgresEngine):
            sql = (
                "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")

        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, *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 = {
            ev.event_id: ev
            for ev in events
        }

        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_rooms(self, room_ids, search_term, keys, limit, pagination_token=None):
        """Performs a full text search over events with given keys.

        Args:
            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"
            pagination_token (str): A pagination token previously returned

        Returns:
            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.
        # 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:
            local_clauses.append("key = ?")
            args.append(key)

        clauses.append(
            "(%s)" % (" OR ".join(local_clauses),)
        )

        count_args = args
        count_clauses = clauses

        if pagination_token:
            try:
                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(
                "(origin_server_ts < ?"
                " OR (origin_server_ts = ? AND stream_ordering < ?))"
            )
            args.extend([origin_server_ts, origin_server_ts, stream])

        if isinstance(self.database_engine, PostgresEngine):
            sql = (
                "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 @@ 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
            #
            # 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
            # to use the indexes in this order because sqlite refuses to
            # MATCH unless it uses the full text search index
            sql = (
                "SELECT rank(matchinfo) as rank, room_id, event_id,"
                " 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 "
            )
            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")

        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 origin_server_ts DESC, stream_ordering DESC LIMIT ?"

        args.append(limit)

        results = yield self._execute(
            "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 = {
            ev.event_id: ev
            for ev in events
        }

        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")