2014-08-12 10:10:52 -04:00
|
|
|
# -*- coding: utf-8 -*-
|
2016-01-06 23:26:29 -05:00
|
|
|
# Copyright 2014-2016 OpenMarket Ltd
|
2019-05-17 14:37:31 -04:00
|
|
|
# Copyright 2017-2018 New Vector Ltd
|
|
|
|
# Copyright 2019 The Matrix.org Foundation C.I.C.
|
2014-08-12 10:10:52 -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.
|
2019-02-19 06:34:40 -05:00
|
|
|
import itertools
|
2014-08-12 10:10:52 -04:00
|
|
|
import logging
|
2019-05-28 11:47:42 -04:00
|
|
|
import random
|
2018-07-09 02:09:20 -04:00
|
|
|
import sys
|
|
|
|
import threading
|
|
|
|
import time
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2018-08-30 10:19:58 -04:00
|
|
|
from six import PY2, iteritems, iterkeys, itervalues
|
2018-10-19 20:16:55 -04:00
|
|
|
from six.moves import builtins, intern, range
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2018-08-30 10:19:58 -04:00
|
|
|
from canonicaljson import json
|
2018-05-21 20:48:57 -04:00
|
|
|
from prometheus_client import Histogram
|
2015-04-09 06:41:36 -04:00
|
|
|
|
2014-11-14 06:16:50 -05:00
|
|
|
from twisted.internet import defer
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2018-07-09 02:09:20 -04:00
|
|
|
from synapse.api.errors import StoreError
|
2019-01-24 06:57:41 -05:00
|
|
|
from synapse.metrics.background_process_metrics import run_as_background_process
|
2019-01-28 10:43:32 -05:00
|
|
|
from synapse.storage.engines import PostgresEngine, Sqlite3Engine
|
2019-02-18 12:53:31 -05:00
|
|
|
from synapse.types import get_domain_from_id
|
2019-02-27 05:28:37 -05:00
|
|
|
from synapse.util import batch_iter
|
2018-07-09 02:09:20 -04:00
|
|
|
from synapse.util.caches.descriptors import Cache
|
|
|
|
from synapse.util.logcontext import LoggingContext, PreserveLoggingContext
|
2018-12-04 05:55:52 -05:00
|
|
|
from synapse.util.stringutils import exception_to_unicode
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2019-06-19 16:04:58 -04:00
|
|
|
# import a function which will return a monotonic time, in seconds
|
|
|
|
try:
|
|
|
|
# on python 3, use time.monotonic, since time.clock can go backwards
|
|
|
|
from time import monotonic as monotonic_time
|
|
|
|
except ImportError:
|
|
|
|
# ... but python 2 doesn't have it
|
|
|
|
from time import clock as monotonic_time
|
|
|
|
|
2014-08-12 10:10:52 -04:00
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
2018-04-30 10:38:23 -04:00
|
|
|
try:
|
|
|
|
MAX_TXN_ID = sys.maxint - 1
|
|
|
|
except AttributeError:
|
|
|
|
# python 3 does not have a maximum int value
|
2019-04-03 05:07:29 -04:00
|
|
|
MAX_TXN_ID = 2 ** 63 - 1
|
2018-04-30 10:38:23 -04:00
|
|
|
|
2014-09-12 09:37:55 -04:00
|
|
|
sql_logger = logging.getLogger("synapse.storage.SQL")
|
2014-10-28 07:18:04 -04:00
|
|
|
transaction_logger = logging.getLogger("synapse.storage.txn")
|
2015-04-08 08:12:38 -04:00
|
|
|
perf_logger = logging.getLogger("synapse.storage.TIME")
|
2014-09-12 09:37:55 -04:00
|
|
|
|
2018-05-28 05:10:27 -04:00
|
|
|
sql_scheduling_timer = Histogram("synapse_storage_schedule_time", "sec")
|
2014-09-12 09:37:55 -04:00
|
|
|
|
2018-05-28 05:10:27 -04:00
|
|
|
sql_query_timer = Histogram("synapse_storage_query_time", "sec", ["verb"])
|
|
|
|
sql_txn_timer = Histogram("synapse_storage_transaction_time", "sec", ["desc"])
|
2015-03-06 11:18:21 -05:00
|
|
|
|
2015-08-07 13:14:49 -04:00
|
|
|
|
2019-02-08 13:30:46 -05:00
|
|
|
# Unique indexes which have been added in background updates. Maps from table name
|
|
|
|
# to the name of the background update which added the unique index to that table.
|
|
|
|
#
|
|
|
|
# This is used by the upsert logic to figure out which tables are safe to do a proper
|
|
|
|
# UPSERT on: until the relevant background update has completed, we
|
|
|
|
# have to emulate an upsert by locking the table.
|
|
|
|
#
|
|
|
|
UNIQUE_INDEX_BACKGROUND_UPDATES = {
|
|
|
|
"user_ips": "user_ips_device_unique_index",
|
|
|
|
"device_lists_remote_extremeties": "device_lists_remote_extremeties_unique_idx",
|
|
|
|
"device_lists_remote_cache": "device_lists_remote_cache_unique_idx",
|
|
|
|
"event_search": "event_search_event_id_idx",
|
|
|
|
}
|
|
|
|
|
2019-02-18 12:53:31 -05:00
|
|
|
# This is a special cache name we use to batch multiple invalidations of caches
|
|
|
|
# based on the current state when notifying workers over replication.
|
|
|
|
_CURRENT_STATE_CACHE_NAME = "cs_cache_fake"
|
|
|
|
|
2019-02-08 13:30:46 -05:00
|
|
|
|
2014-09-12 09:37:55 -04:00
|
|
|
class LoggingTransaction(object):
|
|
|
|
"""An object that almost-transparently proxies for the 'txn' object
|
2015-03-04 14:45:16 -05:00
|
|
|
passed to the constructor. Adds logging and metrics to the .execute()
|
|
|
|
method."""
|
2019-04-03 05:07:29 -04:00
|
|
|
|
2017-06-07 12:34:20 -04:00
|
|
|
__slots__ = [
|
2019-04-03 05:07:29 -04:00
|
|
|
"txn",
|
|
|
|
"name",
|
|
|
|
"database_engine",
|
|
|
|
"after_callbacks",
|
|
|
|
"exception_callbacks",
|
2017-06-07 12:34:20 -04:00
|
|
|
]
|
2014-09-12 09:37:55 -04:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def __init__(
|
|
|
|
self, txn, name, database_engine, after_callbacks, exception_callbacks
|
|
|
|
):
|
2014-09-12 09:37:55 -04:00
|
|
|
object.__setattr__(self, "txn", txn)
|
2014-10-28 07:18:04 -04:00
|
|
|
object.__setattr__(self, "name", name)
|
2015-04-01 09:12:33 -04:00
|
|
|
object.__setattr__(self, "database_engine", database_engine)
|
2015-05-05 12:32:21 -04:00
|
|
|
object.__setattr__(self, "after_callbacks", after_callbacks)
|
2018-03-02 09:43:29 -05:00
|
|
|
object.__setattr__(self, "exception_callbacks", exception_callbacks)
|
2015-05-05 12:32:21 -04:00
|
|
|
|
2017-05-02 05:40:31 -04:00
|
|
|
def call_after(self, callback, *args, **kwargs):
|
2015-05-05 12:45:11 -04:00
|
|
|
"""Call the given callback on the main twisted thread after the
|
|
|
|
transaction has finished. Used to invalidate the caches on the
|
|
|
|
correct thread.
|
|
|
|
"""
|
2017-05-02 05:40:31 -04:00
|
|
|
self.after_callbacks.append((callback, args, kwargs))
|
2014-09-12 09:37:55 -04:00
|
|
|
|
2018-03-02 09:43:29 -05:00
|
|
|
def call_on_exception(self, callback, *args, **kwargs):
|
|
|
|
self.exception_callbacks.append((callback, args, kwargs))
|
2017-06-07 12:34:20 -04:00
|
|
|
|
2014-10-28 06:53:11 -04:00
|
|
|
def __getattr__(self, name):
|
|
|
|
return getattr(self.txn, name)
|
2014-09-12 09:37:55 -04:00
|
|
|
|
|
|
|
def __setattr__(self, name, value):
|
2014-10-28 06:53:11 -04:00
|
|
|
setattr(self.txn, name, value)
|
2014-09-12 09:37:55 -04:00
|
|
|
|
2017-03-23 13:53:49 -04:00
|
|
|
def __iter__(self):
|
|
|
|
return self.txn.__iter__()
|
|
|
|
|
2019-02-20 07:03:30 -05:00
|
|
|
def execute_batch(self, sql, args):
|
|
|
|
if isinstance(self.database_engine, PostgresEngine):
|
|
|
|
from psycopg2.extras import execute_batch
|
2019-04-03 05:07:29 -04:00
|
|
|
|
2019-02-20 07:03:30 -05:00
|
|
|
self._do_execute(lambda *x: execute_batch(self.txn, *x), sql, args)
|
|
|
|
else:
|
|
|
|
for val in args:
|
|
|
|
self.execute(sql, val)
|
|
|
|
|
2015-05-05 10:13:25 -04:00
|
|
|
def execute(self, sql, *args):
|
|
|
|
self._do_execute(self.txn.execute, sql, *args)
|
|
|
|
|
|
|
|
def executemany(self, sql, *args):
|
|
|
|
self._do_execute(self.txn.executemany, sql, *args)
|
|
|
|
|
2017-02-23 06:15:31 -05:00
|
|
|
def _make_sql_one_line(self, sql):
|
|
|
|
"Strip newlines out of SQL so that the loggers in the DB are on one line"
|
|
|
|
return " ".join(l.strip() for l in sql.splitlines() if l.strip())
|
|
|
|
|
2015-05-05 10:13:25 -04:00
|
|
|
def _do_execute(self, func, sql, *args):
|
2017-02-23 06:15:31 -05:00
|
|
|
sql = self._make_sql_one_line(sql)
|
|
|
|
|
2014-09-12 09:37:55 -04:00
|
|
|
# TODO(paul): Maybe use 'info' and 'debug' for values?
|
2014-10-28 07:18:04 -04:00
|
|
|
sql_logger.debug("[SQL] {%s} %s", self.name, sql)
|
2015-03-04 14:45:16 -05:00
|
|
|
|
2015-04-01 09:12:33 -04:00
|
|
|
sql = self.database_engine.convert_param_style(sql)
|
2015-05-05 10:13:25 -04:00
|
|
|
if args:
|
2015-04-02 05:06:22 -04:00
|
|
|
try:
|
2019-04-03 05:07:29 -04:00
|
|
|
sql_logger.debug("[SQL values] {%s} %r", self.name, args[0])
|
2017-10-23 10:52:32 -04:00
|
|
|
except Exception:
|
2015-04-02 05:06:22 -04:00
|
|
|
# Don't let logging failures stop SQL from working
|
|
|
|
pass
|
2014-09-12 09:37:55 -04:00
|
|
|
|
2018-05-28 05:10:27 -04:00
|
|
|
start = time.time()
|
2015-04-02 05:06:22 -04:00
|
|
|
|
2014-10-28 06:34:05 -04:00
|
|
|
try:
|
2019-04-03 05:07:29 -04:00
|
|
|
return func(sql, *args)
|
2015-04-07 07:05:36 -04:00
|
|
|
except Exception as e:
|
2015-04-08 08:11:28 -04:00
|
|
|
logger.debug("[SQL FAIL] {%s} %s", self.name, e)
|
|
|
|
raise
|
2014-10-28 06:34:05 -04:00
|
|
|
finally:
|
2018-05-28 05:10:27 -04:00
|
|
|
secs = time.time() - start
|
|
|
|
sql_logger.debug("[SQL time] {%s} %f sec", self.name, secs)
|
|
|
|
sql_query_timer.labels(sql.split()[0]).observe(secs)
|
2014-09-12 09:37:55 -04:00
|
|
|
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2015-02-10 09:50:53 -05:00
|
|
|
class PerformanceCounters(object):
|
|
|
|
def __init__(self):
|
|
|
|
self.current_counters = {}
|
|
|
|
self.previous_counters = {}
|
|
|
|
|
|
|
|
def update(self, key, start_time, end_time=None):
|
|
|
|
if end_time is None:
|
2018-05-28 05:10:27 -04:00
|
|
|
end_time = time.time()
|
2015-02-10 09:50:53 -05:00
|
|
|
duration = end_time - start_time
|
|
|
|
count, cum_time = self.current_counters.get(key, (0, 0))
|
|
|
|
count += 1
|
|
|
|
cum_time += duration
|
|
|
|
self.current_counters[key] = (count, cum_time)
|
|
|
|
return end_time
|
|
|
|
|
|
|
|
def interval(self, interval_duration, limit=3):
|
|
|
|
counters = []
|
2018-05-15 11:52:41 -04:00
|
|
|
for name, (count, cum_time) in iteritems(self.current_counters):
|
2015-02-10 09:50:53 -05:00
|
|
|
prev_count, prev_time = self.previous_counters.get(name, (0, 0))
|
2019-04-03 05:07:29 -04:00
|
|
|
counters.append(
|
|
|
|
((cum_time - prev_time) / interval_duration, count - prev_count, name)
|
|
|
|
)
|
2015-02-10 09:50:53 -05:00
|
|
|
|
|
|
|
self.previous_counters = dict(self.current_counters)
|
|
|
|
|
|
|
|
counters.sort(reverse=True)
|
|
|
|
|
|
|
|
top_n_counters = ", ".join(
|
|
|
|
"%s(%d): %.3f%%" % (name, count, 100 * ratio)
|
2015-02-10 09:54:07 -05:00
|
|
|
for ratio, count, name in counters[:limit]
|
2015-02-10 09:50:53 -05:00
|
|
|
)
|
|
|
|
|
|
|
|
return top_n_counters
|
|
|
|
|
|
|
|
|
2014-08-12 10:10:52 -04:00
|
|
|
class SQLBaseStore(object):
|
2014-10-28 07:18:04 -04:00
|
|
|
_TXN_ID = 0
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2017-11-09 13:51:27 -05:00
|
|
|
def __init__(self, db_conn, hs):
|
2014-08-13 13:15:23 -04:00
|
|
|
self.hs = hs
|
2016-06-02 10:23:09 -04:00
|
|
|
self._clock = hs.get_clock()
|
2014-08-12 10:10:52 -04:00
|
|
|
self._db_pool = hs.get_db_pool()
|
|
|
|
|
2015-02-09 09:22:52 -05:00
|
|
|
self._previous_txn_total_time = 0
|
|
|
|
self._current_txn_total_time = 0
|
|
|
|
self._previous_loop_ts = 0
|
2015-03-05 10:58:03 -05:00
|
|
|
|
|
|
|
# TODO(paul): These can eventually be removed once the metrics code
|
|
|
|
# is running in mainline, and we have some nice monitoring frontends
|
|
|
|
# to watch it
|
2015-02-10 09:50:53 -05:00
|
|
|
self._txn_perf_counters = PerformanceCounters()
|
|
|
|
self._get_event_counters = PerformanceCounters()
|
2015-02-09 09:22:52 -05:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
self._get_event_cache = Cache(
|
|
|
|
"*getEvent*", keylen=3, max_entries=hs.config.event_cache_size
|
|
|
|
)
|
2015-03-09 16:39:17 -04:00
|
|
|
|
2015-05-15 06:35:04 -04:00
|
|
|
self._event_fetch_lock = threading.Condition()
|
2015-05-14 10:34:02 -04:00
|
|
|
self._event_fetch_list = []
|
2015-05-14 10:40:21 -04:00
|
|
|
self._event_fetch_ongoing = 0
|
2015-05-14 10:34:02 -04:00
|
|
|
|
2015-05-15 05:54:04 -04:00
|
|
|
self._pending_ds = []
|
|
|
|
|
2015-04-01 09:12:33 -04:00
|
|
|
self.database_engine = hs.database_engine
|
|
|
|
|
2019-01-24 05:31:54 -05:00
|
|
|
# A set of tables that are not safe to use native upserts in.
|
2019-02-08 13:30:46 -05:00
|
|
|
self._unsafe_to_upsert_tables = set(UNIQUE_INDEX_BACKGROUND_UPDATES.keys())
|
2019-01-24 05:31:54 -05:00
|
|
|
|
2019-05-17 14:37:31 -04:00
|
|
|
self._account_validity = self.hs.config.account_validity
|
|
|
|
|
2019-01-28 10:43:32 -05:00
|
|
|
# We add the user_directory_search table to the blacklist on SQLite
|
|
|
|
# because the existing search table does not have an index, making it
|
|
|
|
# unsafe to use native upserts.
|
|
|
|
if isinstance(self.database_engine, Sqlite3Engine):
|
|
|
|
self._unsafe_to_upsert_tables.add("user_directory_search")
|
|
|
|
|
2019-01-24 05:31:54 -05:00
|
|
|
if self.database_engine.can_native_upsert:
|
|
|
|
# Check ASAP (and then later, every 1s) to see if we have finished
|
|
|
|
# background updates of tables that aren't safe to update.
|
2019-01-24 06:57:41 -05:00
|
|
|
self._clock.call_later(
|
|
|
|
0.0,
|
|
|
|
run_as_background_process,
|
|
|
|
"upsert_safety_check",
|
2019-04-03 05:07:29 -04:00
|
|
|
self._check_safe_to_upsert,
|
2019-01-24 06:57:41 -05:00
|
|
|
)
|
2019-01-24 05:31:54 -05:00
|
|
|
|
2019-05-28 11:47:42 -04:00
|
|
|
self.rand = random.SystemRandom()
|
|
|
|
|
2019-05-17 14:37:31 -04:00
|
|
|
if self._account_validity.enabled:
|
|
|
|
self._clock.call_later(
|
|
|
|
0.0,
|
|
|
|
run_as_background_process,
|
|
|
|
"account_validity_set_expiration_dates",
|
|
|
|
self._set_expiration_date_when_missing,
|
|
|
|
)
|
|
|
|
|
2019-01-24 05:31:54 -05:00
|
|
|
@defer.inlineCallbacks
|
|
|
|
def _check_safe_to_upsert(self):
|
|
|
|
"""
|
|
|
|
Is it safe to use native UPSERT?
|
|
|
|
|
|
|
|
If there are background updates, we will need to wait, as they may be
|
|
|
|
the addition of indexes that set the UNIQUE constraint that we require.
|
|
|
|
|
2019-01-24 06:57:41 -05:00
|
|
|
If the background updates have not completed, wait 15 sec and check again.
|
2019-01-24 05:31:54 -05:00
|
|
|
"""
|
|
|
|
updates = yield self._simple_select_list(
|
|
|
|
"background_updates",
|
|
|
|
keyvalues=None,
|
|
|
|
retcols=["update_name"],
|
|
|
|
desc="check_background_updates",
|
|
|
|
)
|
|
|
|
updates = [x["update_name"] for x in updates]
|
|
|
|
|
2019-02-08 13:30:46 -05:00
|
|
|
for table, update_name in UNIQUE_INDEX_BACKGROUND_UPDATES.items():
|
|
|
|
if update_name not in updates:
|
|
|
|
logger.debug("Now safe to upsert in %s", table)
|
|
|
|
self._unsafe_to_upsert_tables.discard(table)
|
2019-01-24 05:31:54 -05:00
|
|
|
|
2019-02-08 13:30:46 -05:00
|
|
|
# If there's any updates still running, reschedule to run.
|
2019-01-29 05:04:23 -05:00
|
|
|
if updates:
|
2019-01-24 06:57:41 -05:00
|
|
|
self._clock.call_later(
|
|
|
|
15.0,
|
|
|
|
run_as_background_process,
|
|
|
|
"upsert_safety_check",
|
2019-04-03 05:07:29 -04:00
|
|
|
self._check_safe_to_upsert,
|
2019-01-24 06:57:41 -05:00
|
|
|
)
|
2019-01-24 05:31:54 -05:00
|
|
|
|
2019-05-17 14:37:31 -04:00
|
|
|
@defer.inlineCallbacks
|
|
|
|
def _set_expiration_date_when_missing(self):
|
|
|
|
"""
|
|
|
|
Retrieves the list of registered users that don't have an expiration date, and
|
|
|
|
adds an expiration date for each of them.
|
|
|
|
"""
|
|
|
|
|
|
|
|
def select_users_with_no_expiration_date_txn(txn):
|
|
|
|
"""Retrieves the list of registered users with no expiration date from the
|
2019-06-07 10:30:54 -04:00
|
|
|
database, filtering out deactivated users.
|
2019-05-17 14:37:31 -04:00
|
|
|
"""
|
|
|
|
sql = (
|
|
|
|
"SELECT users.name FROM users"
|
|
|
|
" LEFT JOIN account_validity ON (users.name = account_validity.user_id)"
|
2019-06-07 10:30:54 -04:00
|
|
|
" WHERE account_validity.user_id is NULL AND users.deactivated = 0;"
|
2019-05-17 14:37:31 -04:00
|
|
|
)
|
|
|
|
txn.execute(sql, [])
|
|
|
|
|
2019-05-21 08:38:51 -04:00
|
|
|
res = self.cursor_to_dict(txn)
|
|
|
|
if res:
|
|
|
|
for user in res:
|
2019-05-28 11:47:42 -04:00
|
|
|
self.set_expiration_date_for_user_txn(
|
2019-06-20 05:32:02 -04:00
|
|
|
txn, user["name"], use_delta=True
|
2019-05-28 11:47:42 -04:00
|
|
|
)
|
2019-05-21 08:38:51 -04:00
|
|
|
|
|
|
|
yield self.runInteraction(
|
2019-05-17 14:37:31 -04:00
|
|
|
"get_users_with_no_expiration_date",
|
|
|
|
select_users_with_no_expiration_date_txn,
|
|
|
|
)
|
|
|
|
|
2019-05-28 11:47:42 -04:00
|
|
|
def set_expiration_date_for_user_txn(self, txn, user_id, use_delta=False):
|
2019-05-17 14:37:31 -04:00
|
|
|
"""Sets an expiration date to the account with the given user ID.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
user_id (str): User ID to set an expiration date for.
|
2019-05-28 11:47:42 -04:00
|
|
|
use_delta (bool): If set to False, the expiration date for the user will be
|
|
|
|
now + validity period. If set to True, this expiration date will be a
|
2019-05-31 06:09:34 -04:00
|
|
|
random value in the [now + period - d ; now + period] range, d being a
|
2019-05-31 04:54:46 -04:00
|
|
|
delta equal to 10% of the validity period.
|
2019-05-17 14:37:31 -04:00
|
|
|
"""
|
|
|
|
now_ms = self._clock.time_msec()
|
|
|
|
expiration_ts = now_ms + self._account_validity.period
|
2019-05-28 11:47:42 -04:00
|
|
|
|
2019-05-31 04:54:46 -04:00
|
|
|
if use_delta:
|
2019-05-28 11:47:42 -04:00
|
|
|
expiration_ts = self.rand.randrange(
|
2019-05-31 06:09:34 -04:00
|
|
|
expiration_ts - self._account_validity.startup_job_max_delta,
|
2019-05-28 11:47:42 -04:00
|
|
|
expiration_ts,
|
|
|
|
)
|
|
|
|
|
2019-05-17 14:37:31 -04:00
|
|
|
self._simple_insert_txn(
|
|
|
|
txn,
|
|
|
|
"account_validity",
|
|
|
|
values={
|
|
|
|
"user_id": user_id,
|
|
|
|
"expiration_ts_ms": expiration_ts,
|
|
|
|
"email_sent": False,
|
|
|
|
},
|
|
|
|
)
|
|
|
|
|
2015-02-09 09:22:52 -05:00
|
|
|
def start_profiling(self):
|
2019-06-19 16:04:58 -04:00
|
|
|
self._previous_loop_ts = monotonic_time()
|
2015-02-09 09:22:52 -05:00
|
|
|
|
|
|
|
def loop():
|
|
|
|
curr = self._current_txn_total_time
|
|
|
|
prev = self._previous_txn_total_time
|
|
|
|
self._previous_txn_total_time = curr
|
|
|
|
|
2019-06-19 16:04:58 -04:00
|
|
|
time_now = monotonic_time()
|
2015-02-09 09:22:52 -05:00
|
|
|
time_then = self._previous_loop_ts
|
|
|
|
self._previous_loop_ts = time_now
|
|
|
|
|
2016-02-02 12:18:50 -05:00
|
|
|
ratio = (curr - prev) / (time_now - time_then)
|
2015-02-09 09:22:52 -05:00
|
|
|
|
2015-02-10 09:50:53 -05:00
|
|
|
top_three_counters = self._txn_perf_counters.interval(
|
|
|
|
time_now - time_then, limit=3
|
|
|
|
)
|
|
|
|
|
|
|
|
top_3_event_counters = self._get_event_counters.interval(
|
|
|
|
time_now - time_then, limit=3
|
2015-02-09 12:55:56 -05:00
|
|
|
)
|
|
|
|
|
2015-04-08 08:12:38 -04:00
|
|
|
perf_logger.info(
|
2015-02-10 09:54:07 -05:00
|
|
|
"Total database time: %.3f%% {%s} {%s}",
|
2019-04-03 05:07:29 -04:00
|
|
|
ratio * 100,
|
|
|
|
top_three_counters,
|
|
|
|
top_3_event_counters,
|
2015-02-09 12:55:56 -05:00
|
|
|
)
|
2015-02-09 09:22:52 -05:00
|
|
|
|
2015-02-09 09:45:15 -05:00
|
|
|
self._clock.looping_call(loop, 10000)
|
2015-02-09 09:22:52 -05:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _new_transaction(
|
|
|
|
self, conn, desc, after_callbacks, exception_callbacks, func, *args, **kwargs
|
|
|
|
):
|
2019-06-19 16:04:58 -04:00
|
|
|
start = monotonic_time()
|
2015-05-14 11:54:35 -04:00
|
|
|
txn_id = self._TXN_ID
|
|
|
|
|
|
|
|
# We don't really need these to be unique, so lets stop it from
|
|
|
|
# growing really large.
|
2018-04-30 10:38:23 -04:00
|
|
|
self._TXN_ID = (self._TXN_ID + 1) % (MAX_TXN_ID)
|
2015-05-14 11:54:35 -04:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
name = "%s-%x" % (desc, txn_id)
|
2015-05-14 11:54:35 -04:00
|
|
|
|
|
|
|
transaction_logger.debug("[TXN START] {%s}", name)
|
|
|
|
|
|
|
|
try:
|
|
|
|
i = 0
|
|
|
|
N = 5
|
|
|
|
while True:
|
|
|
|
try:
|
|
|
|
txn = conn.cursor()
|
|
|
|
txn = LoggingTransaction(
|
2019-04-03 05:07:29 -04:00
|
|
|
txn,
|
|
|
|
name,
|
|
|
|
self.database_engine,
|
|
|
|
after_callbacks,
|
2018-03-02 09:43:29 -05:00
|
|
|
exception_callbacks,
|
2015-05-14 11:54:35 -04:00
|
|
|
)
|
2015-05-15 05:54:04 -04:00
|
|
|
r = func(txn, *args, **kwargs)
|
|
|
|
conn.commit()
|
|
|
|
return r
|
2015-05-14 11:54:35 -04:00
|
|
|
except self.database_engine.module.OperationalError as e:
|
|
|
|
# This can happen if the database disappears mid
|
|
|
|
# transaction.
|
2018-12-04 05:55:52 -05:00
|
|
|
logger.warning(
|
2015-05-14 11:54:35 -04:00
|
|
|
"[TXN OPERROR] {%s} %s %d/%d",
|
2019-04-03 05:07:29 -04:00
|
|
|
name,
|
|
|
|
exception_to_unicode(e),
|
|
|
|
i,
|
|
|
|
N,
|
2015-05-14 11:54:35 -04:00
|
|
|
)
|
|
|
|
if i < N:
|
|
|
|
i += 1
|
|
|
|
try:
|
|
|
|
conn.rollback()
|
|
|
|
except self.database_engine.module.Error as e1:
|
2018-12-04 05:55:52 -05:00
|
|
|
logger.warning(
|
2019-04-03 05:07:29 -04:00
|
|
|
"[TXN EROLL] {%s} %s", name, exception_to_unicode(e1)
|
2015-05-14 11:54:35 -04:00
|
|
|
)
|
|
|
|
continue
|
|
|
|
raise
|
|
|
|
except self.database_engine.module.DatabaseError as e:
|
|
|
|
if self.database_engine.is_deadlock(e):
|
2018-12-04 05:55:52 -05:00
|
|
|
logger.warning("[TXN DEADLOCK] {%s} %d/%d", name, i, N)
|
2015-05-14 11:54:35 -04:00
|
|
|
if i < N:
|
|
|
|
i += 1
|
|
|
|
try:
|
|
|
|
conn.rollback()
|
|
|
|
except self.database_engine.module.Error as e1:
|
2018-12-04 05:55:52 -05:00
|
|
|
logger.warning(
|
2015-05-14 11:54:35 -04:00
|
|
|
"[TXN EROLL] {%s} %s",
|
2019-04-03 05:07:29 -04:00
|
|
|
name,
|
|
|
|
exception_to_unicode(e1),
|
2015-05-14 11:54:35 -04:00
|
|
|
)
|
|
|
|
continue
|
|
|
|
raise
|
|
|
|
except Exception as e:
|
|
|
|
logger.debug("[TXN FAIL] {%s} %s", name, e)
|
|
|
|
raise
|
|
|
|
finally:
|
2019-06-19 16:04:58 -04:00
|
|
|
end = monotonic_time()
|
2015-05-14 11:54:35 -04:00
|
|
|
duration = end - start
|
|
|
|
|
2018-07-10 11:12:36 -04:00
|
|
|
LoggingContext.current_context().add_database_transaction(duration)
|
2015-12-07 12:56:11 -05:00
|
|
|
|
2018-05-28 05:10:27 -04:00
|
|
|
transaction_logger.debug("[TXN END] {%s} %f sec", name, duration)
|
2015-05-14 11:54:35 -04:00
|
|
|
|
|
|
|
self._current_txn_total_time += duration
|
|
|
|
self._txn_perf_counters.update(desc, start, end)
|
2018-05-21 20:48:57 -04:00
|
|
|
sql_txn_timer.labels(desc).observe(duration)
|
2015-05-14 11:54:35 -04:00
|
|
|
|
2014-10-29 21:21:33 -04:00
|
|
|
@defer.inlineCallbacks
|
2014-10-28 07:18:04 -04:00
|
|
|
def runInteraction(self, desc, func, *args, **kwargs):
|
2018-01-16 11:17:24 -05:00
|
|
|
"""Starts a transaction on the database and runs a given function
|
2014-11-20 12:26:36 -05:00
|
|
|
|
2018-01-16 11:17:24 -05:00
|
|
|
Arguments:
|
|
|
|
desc (str): description of the transaction, for logging and metrics
|
|
|
|
func (func): callback function, which will be called with a
|
|
|
|
database transaction (twisted.enterprise.adbapi.Transaction) as
|
|
|
|
its first argument, followed by `args` and `kwargs`.
|
|
|
|
|
|
|
|
args (list): positional args to pass to `func`
|
|
|
|
kwargs (dict): named args to pass to `func`
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
Deferred: The result of func
|
|
|
|
"""
|
2015-05-05 12:32:21 -04:00
|
|
|
after_callbacks = []
|
2018-03-02 09:43:29 -05:00
|
|
|
exception_callbacks = []
|
2015-05-05 12:32:21 -04:00
|
|
|
|
2018-07-25 05:47:57 -04:00
|
|
|
if LoggingContext.current_context() == LoggingContext.sentinel:
|
2019-04-03 05:07:29 -04:00
|
|
|
logger.warn("Starting db txn '%s' from sentinel context", desc)
|
2018-07-25 05:47:57 -04:00
|
|
|
|
2016-08-15 04:45:44 -04:00
|
|
|
try:
|
2018-07-10 11:12:36 -04:00
|
|
|
result = yield self.runWithConnection(
|
|
|
|
self._new_transaction,
|
2019-04-03 05:07:29 -04:00
|
|
|
desc,
|
|
|
|
after_callbacks,
|
|
|
|
exception_callbacks,
|
|
|
|
func,
|
|
|
|
*args,
|
|
|
|
**kwargs
|
2018-07-10 11:12:36 -04:00
|
|
|
)
|
2017-06-07 12:34:20 -04:00
|
|
|
|
2017-05-02 05:40:31 -04:00
|
|
|
for after_callback, after_args, after_kwargs in after_callbacks:
|
|
|
|
after_callback(*after_args, **after_kwargs)
|
2018-03-02 09:43:29 -05:00
|
|
|
except: # noqa: E722, as we reraise the exception this is fine.
|
|
|
|
for after_callback, after_args, after_kwargs in exception_callbacks:
|
2017-06-07 12:34:20 -04:00
|
|
|
after_callback(*after_args, **after_kwargs)
|
2018-03-02 09:43:29 -05:00
|
|
|
raise
|
2017-06-07 12:34:20 -04:00
|
|
|
|
2015-05-14 11:54:35 -04:00
|
|
|
defer.returnValue(result)
|
2014-11-14 06:16:50 -05:00
|
|
|
|
2015-05-14 11:54:35 -04:00
|
|
|
@defer.inlineCallbacks
|
|
|
|
def runWithConnection(self, func, *args, **kwargs):
|
2018-01-16 11:17:24 -05:00
|
|
|
"""Wraps the .runWithConnection() method on the underlying db_pool.
|
|
|
|
|
|
|
|
Arguments:
|
|
|
|
func (func): callback function, which will be called with a
|
|
|
|
database connection (twisted.enterprise.adbapi.Connection) as
|
|
|
|
its first argument, followed by `args` and `kwargs`.
|
|
|
|
args (list): positional args to pass to `func`
|
|
|
|
kwargs (dict): named args to pass to `func`
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
Deferred: The result of func
|
|
|
|
"""
|
2018-07-10 11:12:36 -04:00
|
|
|
parent_context = LoggingContext.current_context()
|
|
|
|
if parent_context == LoggingContext.sentinel:
|
2018-07-19 06:14:20 -04:00
|
|
|
logger.warn(
|
2019-04-03 05:07:29 -04:00
|
|
|
"Starting db connection from sentinel context: metrics will be lost"
|
2018-07-19 06:14:20 -04:00
|
|
|
)
|
2018-07-10 11:12:36 -04:00
|
|
|
parent_context = None
|
2015-05-14 11:54:35 -04:00
|
|
|
|
2019-06-19 16:04:58 -04:00
|
|
|
start_time = monotonic_time()
|
2015-05-14 11:54:35 -04:00
|
|
|
|
|
|
|
def inner_func(conn, *args, **kwargs):
|
2018-07-10 11:12:36 -04:00
|
|
|
with LoggingContext("runWithConnection", parent_context) as context:
|
2019-06-19 16:04:58 -04:00
|
|
|
sched_duration_sec = monotonic_time() - start_time
|
2018-05-28 05:10:27 -04:00
|
|
|
sql_scheduling_timer.observe(sched_duration_sec)
|
2018-07-10 11:12:36 -04:00
|
|
|
context.add_database_scheduled(sched_duration_sec)
|
2015-02-09 09:22:52 -05:00
|
|
|
|
2015-05-14 11:54:35 -04:00
|
|
|
if self.database_engine.is_connection_closed(conn):
|
|
|
|
logger.debug("Reconnecting closed database connection")
|
|
|
|
conn.reconnect()
|
|
|
|
|
|
|
|
return func(conn, *args, **kwargs)
|
2015-03-05 10:58:03 -05:00
|
|
|
|
2016-02-04 05:22:44 -05:00
|
|
|
with PreserveLoggingContext():
|
2019-04-03 05:07:29 -04:00
|
|
|
result = yield self._db_pool.runWithConnection(inner_func, *args, **kwargs)
|
2015-05-08 11:32:18 -04:00
|
|
|
|
2014-10-29 21:21:33 -04:00
|
|
|
defer.returnValue(result)
|
2014-09-12 08:57:24 -04:00
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@staticmethod
|
|
|
|
def cursor_to_dict(cursor):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Converts a SQL cursor into an list of dicts.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
cursor : The DBAPI cursor which has executed a query.
|
|
|
|
Returns:
|
|
|
|
A list of dicts where the key is the column header.
|
|
|
|
"""
|
2018-01-06 12:13:56 -05:00
|
|
|
col_headers = list(intern(str(column[0])) for column in cursor.description)
|
2019-04-03 05:07:29 -04:00
|
|
|
results = list(dict(zip(col_headers, row)) for row in cursor)
|
2014-08-12 10:10:52 -04:00
|
|
|
return results
|
|
|
|
|
2015-03-11 13:19:17 -04:00
|
|
|
def _execute(self, desc, decoder, query, *args):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Runs a single query for a result set.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
decoder - The function which can resolve the cursor results to
|
|
|
|
something meaningful.
|
|
|
|
query - The query string to execute
|
|
|
|
*args - Query args.
|
|
|
|
Returns:
|
|
|
|
The result of decoder(results)
|
|
|
|
"""
|
2019-04-03 05:07:29 -04:00
|
|
|
|
2014-08-12 10:10:52 -04:00
|
|
|
def interaction(txn):
|
2015-03-19 11:59:48 -04:00
|
|
|
txn.execute(query, args)
|
2014-08-14 11:02:10 -04:00
|
|
|
if decoder:
|
2015-03-19 11:59:48 -04:00
|
|
|
return decoder(txn)
|
2014-08-14 11:02:10 -04:00
|
|
|
else:
|
2015-03-19 11:59:48 -04:00
|
|
|
return txn.fetchall()
|
2014-08-14 11:02:10 -04:00
|
|
|
|
2015-03-11 13:19:17 -04:00
|
|
|
return self.runInteraction(desc, interaction)
|
2014-08-12 10:10:52 -04:00
|
|
|
|
|
|
|
# "Simple" SQL API methods that operate on a single table with no JOINs,
|
|
|
|
# no complex WHERE clauses, just a dict of values for columns.
|
|
|
|
|
2015-04-15 09:51:21 -04:00
|
|
|
@defer.inlineCallbacks
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_insert(self, table, values, or_ignore=False, desc="_simple_insert"):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes an INSERT query on the named table.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table : string giving the table name
|
|
|
|
values : dict of new column names and values for them
|
2019-06-06 12:34:07 -04:00
|
|
|
or_ignore : bool stating whether an exception should be raised
|
|
|
|
when a conflicting row already exists. If True, False will be
|
|
|
|
returned by the function instead
|
|
|
|
desc : string giving a description of the transaction
|
2017-01-25 09:27:27 -05:00
|
|
|
|
|
|
|
Returns:
|
|
|
|
bool: Whether the row was inserted or not. Only useful when
|
|
|
|
`or_ignore` is True
|
2014-08-12 10:10:52 -04:00
|
|
|
"""
|
2015-04-15 09:51:21 -04:00
|
|
|
try:
|
2019-04-03 05:07:29 -04:00
|
|
|
yield self.runInteraction(desc, self._simple_insert_txn, table, values)
|
2015-04-15 09:51:21 -04:00
|
|
|
except self.database_engine.module.IntegrityError:
|
|
|
|
# We have to do or_ignore flag at this layer, since we can't reuse
|
|
|
|
# a cursor after we receive an error from the db.
|
|
|
|
if not or_ignore:
|
|
|
|
raise
|
2017-01-25 09:27:27 -05:00
|
|
|
defer.returnValue(False)
|
|
|
|
defer.returnValue(True)
|
2014-08-26 09:31:48 -04:00
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@staticmethod
|
|
|
|
def _simple_insert_txn(txn, table, values):
|
2015-05-05 10:13:25 -04:00
|
|
|
keys, vals = zip(*values.items())
|
|
|
|
|
2015-03-23 11:38:56 -04:00
|
|
|
sql = "INSERT INTO %s (%s) VALUES(%s)" % (
|
2014-08-12 10:10:52 -04:00
|
|
|
table,
|
2015-05-05 10:13:25 -04:00
|
|
|
", ".join(k for k in keys),
|
2019-04-03 05:07:29 -04:00
|
|
|
", ".join("?" for _ in keys),
|
2014-08-12 10:10:52 -04:00
|
|
|
)
|
2014-09-12 12:11:00 -04:00
|
|
|
|
2015-05-05 10:13:25 -04:00
|
|
|
txn.execute(sql, vals)
|
|
|
|
|
2017-06-01 09:50:46 -04:00
|
|
|
def _simple_insert_many(self, table, values, desc):
|
2019-04-03 05:07:29 -04:00
|
|
|
return self.runInteraction(desc, self._simple_insert_many_txn, table, values)
|
2017-06-01 09:50:46 -04:00
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@staticmethod
|
|
|
|
def _simple_insert_many_txn(txn, table, values):
|
2015-05-05 10:13:25 -04:00
|
|
|
if not values:
|
|
|
|
return
|
|
|
|
|
2015-05-05 12:06:55 -04:00
|
|
|
# This is a *slight* abomination to get a list of tuples of key names
|
|
|
|
# and a list of tuples of value names.
|
|
|
|
#
|
|
|
|
# i.e. [{"a": 1, "b": 2}, {"c": 3, "d": 4}]
|
|
|
|
# => [("a", "b",), ("c", "d",)] and [(1, 2,), (3, 4,)]
|
|
|
|
#
|
|
|
|
# The sort is to ensure that we don't rely on dictionary iteration
|
|
|
|
# order.
|
2019-04-03 05:07:29 -04:00
|
|
|
keys, vals = zip(
|
|
|
|
*[zip(*(sorted(i.items(), key=lambda kv: kv[0]))) for i in values if i]
|
|
|
|
)
|
2015-05-05 10:13:25 -04:00
|
|
|
|
|
|
|
for k in keys:
|
|
|
|
if k != keys[0]:
|
2019-04-03 05:07:29 -04:00
|
|
|
raise RuntimeError("All items must have the same keys")
|
2015-05-05 10:13:25 -04:00
|
|
|
|
|
|
|
sql = "INSERT INTO %s (%s) VALUES(%s)" % (
|
|
|
|
table,
|
|
|
|
", ".join(k for k in keys[0]),
|
2019-04-03 05:07:29 -04:00
|
|
|
", ".join("?" for _ in keys[0]),
|
2014-09-12 12:11:00 -04:00
|
|
|
)
|
|
|
|
|
2015-05-05 10:13:25 -04:00
|
|
|
txn.executemany(sql, vals)
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2017-11-16 10:30:15 -05:00
|
|
|
@defer.inlineCallbacks
|
2019-01-24 05:31:54 -05:00
|
|
|
def _simple_upsert(
|
|
|
|
self,
|
|
|
|
table,
|
|
|
|
keyvalues,
|
|
|
|
values,
|
|
|
|
insertion_values={},
|
|
|
|
desc="_simple_upsert",
|
2019-04-03 05:07:29 -04:00
|
|
|
lock=True,
|
2019-01-24 05:31:54 -05:00
|
|
|
):
|
2014-12-18 09:49:22 -05:00
|
|
|
"""
|
2017-11-16 10:30:15 -05:00
|
|
|
|
|
|
|
`lock` should generally be set to True (the default), but can be set
|
|
|
|
to False if either of the following are true:
|
|
|
|
|
|
|
|
* there is a UNIQUE INDEX on the key columns. In this case a conflict
|
|
|
|
will cause an IntegrityError in which case this function will retry
|
|
|
|
the update.
|
|
|
|
|
|
|
|
* we somehow know that we are the only thread which will be updating
|
|
|
|
this table.
|
|
|
|
|
2015-01-28 09:44:41 -05:00
|
|
|
Args:
|
|
|
|
table (str): The table to upsert into
|
2019-04-04 09:21:16 -04:00
|
|
|
keyvalues (dict): The unique key columns and their new values
|
2015-01-28 09:44:41 -05:00
|
|
|
values (dict): The nonunique columns and their new values
|
2017-11-16 10:30:15 -05:00
|
|
|
insertion_values (dict): additional key/values to use only when
|
|
|
|
inserting
|
|
|
|
lock (bool): True to lock the table when doing the upsert.
|
2016-05-13 06:25:02 -04:00
|
|
|
Returns:
|
2019-01-24 05:31:54 -05:00
|
|
|
Deferred(None or bool): Native upserts always return None. Emulated
|
|
|
|
upserts return True if a new entry was created, False if an existing
|
|
|
|
one was updated.
|
2014-12-18 09:49:22 -05:00
|
|
|
"""
|
2017-11-27 06:56:57 -05:00
|
|
|
attempts = 0
|
2017-11-16 10:30:15 -05:00
|
|
|
while True:
|
|
|
|
try:
|
|
|
|
result = yield self.runInteraction(
|
|
|
|
desc,
|
2019-01-24 05:31:54 -05:00
|
|
|
self._simple_upsert_txn,
|
|
|
|
table,
|
|
|
|
keyvalues,
|
|
|
|
values,
|
|
|
|
insertion_values,
|
|
|
|
lock=lock,
|
2017-11-16 10:30:15 -05:00
|
|
|
)
|
|
|
|
defer.returnValue(result)
|
2017-11-16 11:03:38 -05:00
|
|
|
except self.database_engine.module.IntegrityError as e:
|
2017-11-27 06:56:57 -05:00
|
|
|
attempts += 1
|
|
|
|
if attempts >= 5:
|
|
|
|
# don't retry forever, because things other than races
|
|
|
|
# can cause IntegrityErrors
|
|
|
|
raise
|
|
|
|
|
2017-11-16 10:30:15 -05:00
|
|
|
# presumably we raced with another transaction: let's retry.
|
|
|
|
logger.warn(
|
2019-04-04 09:21:16 -04:00
|
|
|
"IntegrityError when upserting into %s; retrying: %s", table, e
|
2017-11-16 10:30:15 -05:00
|
|
|
)
|
2014-12-18 09:49:22 -05:00
|
|
|
|
2019-01-24 05:31:54 -05:00
|
|
|
def _simple_upsert_txn(
|
2019-04-03 05:07:29 -04:00
|
|
|
self, txn, table, keyvalues, values, insertion_values={}, lock=True
|
2019-01-24 05:31:54 -05:00
|
|
|
):
|
|
|
|
"""
|
|
|
|
Pick the UPSERT method which works best on the platform. Either the
|
|
|
|
native one (Pg9.5+, recent SQLites), or fall back to an emulated method.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn: The transaction to use.
|
|
|
|
table (str): The table to upsert into
|
|
|
|
keyvalues (dict): The unique key tables and their new values
|
|
|
|
values (dict): The nonunique columns and their new values
|
|
|
|
insertion_values (dict): additional key/values to use only when
|
|
|
|
inserting
|
|
|
|
lock (bool): True to lock the table when doing the upsert.
|
|
|
|
Returns:
|
2019-01-24 06:57:41 -05:00
|
|
|
None or bool: Native upserts always return None. Emulated
|
2019-01-24 05:31:54 -05:00
|
|
|
upserts return True if a new entry was created, False if an existing
|
|
|
|
one was updated.
|
|
|
|
"""
|
|
|
|
if (
|
|
|
|
self.database_engine.can_native_upsert
|
|
|
|
and table not in self._unsafe_to_upsert_tables
|
|
|
|
):
|
|
|
|
return self._simple_upsert_txn_native_upsert(
|
2019-04-03 05:07:29 -04:00
|
|
|
txn, table, keyvalues, values, insertion_values=insertion_values
|
2019-01-24 05:31:54 -05:00
|
|
|
)
|
|
|
|
else:
|
|
|
|
return self._simple_upsert_txn_emulated(
|
|
|
|
txn,
|
|
|
|
table,
|
|
|
|
keyvalues,
|
|
|
|
values,
|
|
|
|
insertion_values=insertion_values,
|
|
|
|
lock=lock,
|
|
|
|
)
|
|
|
|
|
|
|
|
def _simple_upsert_txn_emulated(
|
|
|
|
self, txn, table, keyvalues, values, insertion_values={}, lock=True
|
|
|
|
):
|
2019-01-24 06:57:41 -05:00
|
|
|
"""
|
|
|
|
Args:
|
|
|
|
table (str): The table to upsert into
|
|
|
|
keyvalues (dict): The unique key tables and their new values
|
|
|
|
values (dict): The nonunique columns and their new values
|
|
|
|
insertion_values (dict): additional key/values to use only when
|
|
|
|
inserting
|
|
|
|
lock (bool): True to lock the table when doing the upsert.
|
|
|
|
Returns:
|
|
|
|
bool: Return True if a new entry was created, False if an existing
|
|
|
|
one was updated.
|
|
|
|
"""
|
2015-05-01 05:46:48 -04:00
|
|
|
# We need to lock the table :(, unless we're *really* careful
|
|
|
|
if lock:
|
|
|
|
self.database_engine.lock_table(txn, table)
|
2015-04-27 08:22:30 -04:00
|
|
|
|
2019-01-09 06:26:25 -05:00
|
|
|
def _getwhere(key):
|
|
|
|
# If the value we're passing in is None (aka NULL), we need to use
|
|
|
|
# IS, not =, as NULL = NULL equals NULL (False).
|
|
|
|
if keyvalues[key] is None:
|
|
|
|
return "%s IS ?" % (key,)
|
|
|
|
else:
|
|
|
|
return "%s = ?" % (key,)
|
|
|
|
|
2019-02-20 07:03:30 -05:00
|
|
|
if not values:
|
|
|
|
# If `values` is empty, then all of the values we care about are in
|
|
|
|
# the unique key, so there is nothing to UPDATE. We can just do a
|
|
|
|
# SELECT instead to see if it exists.
|
|
|
|
sql = "SELECT 1 FROM %s WHERE %s" % (
|
|
|
|
table,
|
2019-04-03 05:07:29 -04:00
|
|
|
" AND ".join(_getwhere(k) for k in keyvalues),
|
2019-02-20 07:03:30 -05:00
|
|
|
)
|
|
|
|
sqlargs = list(keyvalues.values())
|
|
|
|
txn.execute(sql, sqlargs)
|
|
|
|
if txn.fetchall():
|
|
|
|
# We have an existing record.
|
|
|
|
return False
|
|
|
|
else:
|
|
|
|
# First try to update.
|
|
|
|
sql = "UPDATE %s SET %s WHERE %s" % (
|
|
|
|
table,
|
|
|
|
", ".join("%s = ?" % (k,) for k in values),
|
2019-04-03 05:07:29 -04:00
|
|
|
" AND ".join(_getwhere(k) for k in keyvalues),
|
2019-02-20 07:03:30 -05:00
|
|
|
)
|
|
|
|
sqlargs = list(values.values()) + list(keyvalues.values())
|
2014-12-18 09:49:22 -05:00
|
|
|
|
2019-02-20 07:03:30 -05:00
|
|
|
txn.execute(sql, sqlargs)
|
|
|
|
if txn.rowcount > 0:
|
|
|
|
# successfully updated at least one row.
|
|
|
|
return False
|
2016-05-13 06:25:02 -04:00
|
|
|
|
2019-02-20 07:03:30 -05:00
|
|
|
# We didn't find any existing rows, so insert a new one
|
2017-11-16 10:29:10 -05:00
|
|
|
allvalues = {}
|
|
|
|
allvalues.update(keyvalues)
|
|
|
|
allvalues.update(values)
|
|
|
|
allvalues.update(insertion_values)
|
|
|
|
|
|
|
|
sql = "INSERT INTO %s (%s) VALUES (%s)" % (
|
|
|
|
table,
|
|
|
|
", ".join(k for k in allvalues),
|
2019-01-24 05:31:54 -05:00
|
|
|
", ".join("?" for _ in allvalues),
|
2017-11-16 10:29:10 -05:00
|
|
|
)
|
2018-05-15 11:52:41 -04:00
|
|
|
txn.execute(sql, list(allvalues.values()))
|
2017-11-16 10:29:10 -05:00
|
|
|
# successfully inserted
|
|
|
|
return True
|
|
|
|
|
2019-01-24 05:31:54 -05:00
|
|
|
def _simple_upsert_txn_native_upsert(
|
|
|
|
self, txn, table, keyvalues, values, insertion_values={}
|
|
|
|
):
|
|
|
|
"""
|
|
|
|
Use the native UPSERT functionality in recent PostgreSQL versions.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table (str): The table to upsert into
|
|
|
|
keyvalues (dict): The unique key tables and their new values
|
|
|
|
values (dict): The nonunique columns and their new values
|
|
|
|
insertion_values (dict): additional key/values to use only when
|
|
|
|
inserting
|
|
|
|
Returns:
|
|
|
|
None
|
|
|
|
"""
|
|
|
|
allvalues = {}
|
|
|
|
allvalues.update(keyvalues)
|
|
|
|
allvalues.update(insertion_values)
|
|
|
|
|
2019-03-11 09:35:31 -04:00
|
|
|
if not values:
|
|
|
|
latter = "NOTHING"
|
|
|
|
else:
|
|
|
|
allvalues.update(values)
|
2019-04-03 05:07:29 -04:00
|
|
|
latter = "UPDATE SET " + ", ".join(k + "=EXCLUDED." + k for k in values)
|
2019-03-11 09:35:31 -04:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
sql = ("INSERT INTO %s (%s) VALUES (%s) " "ON CONFLICT (%s) DO %s") % (
|
2019-01-24 05:31:54 -05:00
|
|
|
table,
|
|
|
|
", ".join(k for k in allvalues),
|
|
|
|
", ".join("?" for _ in allvalues),
|
|
|
|
", ".join(k for k in keyvalues),
|
2019-04-03 05:07:29 -04:00
|
|
|
latter,
|
2019-01-24 05:31:54 -05:00
|
|
|
)
|
|
|
|
txn.execute(sql, list(allvalues.values()))
|
|
|
|
|
2019-02-20 07:03:30 -05:00
|
|
|
def _simple_upsert_many_txn(
|
|
|
|
self, txn, table, key_names, key_values, value_names, value_values
|
|
|
|
):
|
|
|
|
"""
|
|
|
|
Upsert, many times.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table (str): The table to upsert into
|
|
|
|
key_names (list[str]): The key column names.
|
|
|
|
key_values (list[list]): A list of each row's key column values.
|
|
|
|
value_names (list[str]): The value column names. If empty, no
|
|
|
|
values will be used, even if value_values is provided.
|
|
|
|
value_values (list[list]): A list of each row's value column values.
|
|
|
|
Returns:
|
|
|
|
None
|
|
|
|
"""
|
|
|
|
if (
|
|
|
|
self.database_engine.can_native_upsert
|
|
|
|
and table not in self._unsafe_to_upsert_tables
|
|
|
|
):
|
|
|
|
return self._simple_upsert_many_txn_native_upsert(
|
|
|
|
txn, table, key_names, key_values, value_names, value_values
|
|
|
|
)
|
|
|
|
else:
|
|
|
|
return self._simple_upsert_many_txn_emulated(
|
|
|
|
txn, table, key_names, key_values, value_names, value_values
|
|
|
|
)
|
|
|
|
|
|
|
|
def _simple_upsert_many_txn_emulated(
|
|
|
|
self, txn, table, key_names, key_values, value_names, value_values
|
|
|
|
):
|
|
|
|
"""
|
|
|
|
Upsert, many times, but without native UPSERT support or batching.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table (str): The table to upsert into
|
|
|
|
key_names (list[str]): The key column names.
|
|
|
|
key_values (list[list]): A list of each row's key column values.
|
|
|
|
value_names (list[str]): The value column names. If empty, no
|
|
|
|
values will be used, even if value_values is provided.
|
|
|
|
value_values (list[list]): A list of each row's value column values.
|
|
|
|
Returns:
|
|
|
|
None
|
|
|
|
"""
|
|
|
|
# No value columns, therefore make a blank list so that the following
|
|
|
|
# zip() works correctly.
|
|
|
|
if not value_names:
|
|
|
|
value_values = [() for x in range(len(key_values))]
|
|
|
|
|
|
|
|
for keyv, valv in zip(key_values, value_values):
|
|
|
|
_keys = {x: y for x, y in zip(key_names, keyv)}
|
|
|
|
_vals = {x: y for x, y in zip(value_names, valv)}
|
|
|
|
|
|
|
|
self._simple_upsert_txn_emulated(txn, table, _keys, _vals)
|
|
|
|
|
|
|
|
def _simple_upsert_many_txn_native_upsert(
|
|
|
|
self, txn, table, key_names, key_values, value_names, value_values
|
|
|
|
):
|
|
|
|
"""
|
|
|
|
Upsert, many times, using batching where possible.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table (str): The table to upsert into
|
|
|
|
key_names (list[str]): The key column names.
|
|
|
|
key_values (list[list]): A list of each row's key column values.
|
|
|
|
value_names (list[str]): The value column names. If empty, no
|
|
|
|
values will be used, even if value_values is provided.
|
|
|
|
value_values (list[list]): A list of each row's value column values.
|
|
|
|
Returns:
|
|
|
|
None
|
|
|
|
"""
|
|
|
|
allnames = []
|
|
|
|
allnames.extend(key_names)
|
|
|
|
allnames.extend(value_names)
|
|
|
|
|
|
|
|
if not value_names:
|
|
|
|
# No value columns, therefore make a blank list so that the
|
|
|
|
# following zip() works correctly.
|
|
|
|
latter = "NOTHING"
|
|
|
|
value_values = [() for x in range(len(key_values))]
|
|
|
|
else:
|
2019-04-03 05:07:29 -04:00
|
|
|
latter = "UPDATE SET " + ", ".join(
|
|
|
|
k + "=EXCLUDED." + k for k in value_names
|
2019-02-20 07:03:30 -05:00
|
|
|
)
|
|
|
|
|
|
|
|
sql = "INSERT INTO %s (%s) VALUES (%s) ON CONFLICT (%s) DO %s" % (
|
|
|
|
table,
|
|
|
|
", ".join(k for k in allnames),
|
|
|
|
", ".join("?" for _ in allnames),
|
|
|
|
", ".join(key_names),
|
|
|
|
latter,
|
|
|
|
)
|
|
|
|
|
|
|
|
args = []
|
|
|
|
|
|
|
|
for x, y in zip(key_values, value_values):
|
|
|
|
args.append(tuple(x) + tuple(y))
|
|
|
|
|
|
|
|
return txn.execute_batch(sql, args)
|
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_one(
|
|
|
|
self, table, keyvalues, retcols, allow_none=False, desc="_simple_select_one"
|
|
|
|
):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes a SELECT query on the named table, which is expected to
|
2017-11-29 13:27:05 -05:00
|
|
|
return a single row, returning multiple columns from it.
|
2014-08-12 10:10:52 -04:00
|
|
|
|
|
|
|
Args:
|
|
|
|
table : string giving the table name
|
|
|
|
keyvalues : dict of column names and values to select the row with
|
|
|
|
retcols : list of strings giving the names of the columns to return
|
|
|
|
|
|
|
|
allow_none : If true, return None instead of failing if the SELECT
|
|
|
|
statement returns no rows
|
|
|
|
"""
|
2015-03-20 10:59:48 -04:00
|
|
|
return self.runInteraction(
|
2019-04-03 05:07:29 -04:00
|
|
|
desc, self._simple_select_one_txn, table, keyvalues, retcols, allow_none
|
2014-08-12 10:10:52 -04:00
|
|
|
)
|
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_one_onecol(
|
|
|
|
self,
|
|
|
|
table,
|
|
|
|
keyvalues,
|
|
|
|
retcol,
|
|
|
|
allow_none=False,
|
|
|
|
desc="_simple_select_one_onecol",
|
|
|
|
):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes a SELECT query on the named table, which is expected to
|
2015-10-09 10:48:31 -04:00
|
|
|
return a single row, returning a single column from it.
|
2014-08-12 10:10:52 -04:00
|
|
|
|
|
|
|
Args:
|
|
|
|
table : string giving the table name
|
|
|
|
keyvalues : dict of column names and values to select the row with
|
|
|
|
retcol : string giving the name of the column to return
|
|
|
|
"""
|
2014-10-28 12:42:35 -04:00
|
|
|
return self.runInteraction(
|
2015-03-20 11:59:18 -04:00
|
|
|
desc,
|
2014-10-28 12:42:35 -04:00
|
|
|
self._simple_select_one_onecol_txn,
|
2019-04-03 05:07:29 -04:00
|
|
|
table,
|
|
|
|
keyvalues,
|
|
|
|
retcol,
|
|
|
|
allow_none=allow_none,
|
2014-10-28 12:42:35 -04:00
|
|
|
)
|
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@classmethod
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_one_onecol_txn(
|
|
|
|
cls, txn, table, keyvalues, retcol, allow_none=False
|
|
|
|
):
|
2016-01-26 10:51:06 -05:00
|
|
|
ret = cls._simple_select_onecol_txn(
|
2019-04-03 05:07:29 -04:00
|
|
|
txn, table=table, keyvalues=keyvalues, retcol=retcol
|
2014-08-12 10:10:52 -04:00
|
|
|
)
|
|
|
|
|
|
|
|
if ret:
|
2014-10-29 12:59:24 -04:00
|
|
|
return ret[0]
|
2014-08-12 10:10:52 -04:00
|
|
|
else:
|
2014-10-28 12:42:35 -04:00
|
|
|
if allow_none:
|
|
|
|
return None
|
|
|
|
else:
|
|
|
|
raise StoreError(404, "No row found")
|
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@staticmethod
|
|
|
|
def _simple_select_onecol_txn(txn, table, keyvalues, retcol):
|
2019-04-03 05:07:29 -04:00
|
|
|
sql = ("SELECT %(retcol)s FROM %(table)s") % {"retcol": retcol, "table": table}
|
2014-10-28 12:42:35 -04:00
|
|
|
|
2017-11-22 13:02:15 -05:00
|
|
|
if keyvalues:
|
2018-05-15 11:52:41 -04:00
|
|
|
sql += " WHERE %s" % " AND ".join("%s = ?" % k for k in iterkeys(keyvalues))
|
|
|
|
txn.execute(sql, list(keyvalues.values()))
|
2017-11-22 13:02:15 -05:00
|
|
|
else:
|
|
|
|
txn.execute(sql)
|
2014-10-28 12:42:35 -04:00
|
|
|
|
2017-03-23 13:53:49 -04:00
|
|
|
return [r[0] for r in txn]
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_onecol(
|
|
|
|
self, table, keyvalues, retcol, desc="_simple_select_onecol"
|
|
|
|
):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes a SELECT query on the named table, which returns a list
|
|
|
|
comprising of the values of the named column from the selected rows.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table (str): table name
|
2017-11-22 13:02:15 -05:00
|
|
|
keyvalues (dict|None): column names and values to select the rows with
|
2014-08-12 10:10:52 -04:00
|
|
|
retcol (str): column whos value we wish to retrieve.
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
Deferred: Results in a list
|
|
|
|
"""
|
2014-10-28 12:42:35 -04:00
|
|
|
return self.runInteraction(
|
2019-04-03 05:07:29 -04:00
|
|
|
desc, self._simple_select_onecol_txn, table, keyvalues, retcol
|
2014-10-28 12:42:35 -04:00
|
|
|
)
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_list(
|
|
|
|
self, table, keyvalues, retcols, desc="_simple_select_list"
|
|
|
|
):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes a SELECT query on the named table, which may return zero or
|
|
|
|
more rows, returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Args:
|
2016-07-20 10:12:42 -04:00
|
|
|
table (str): the table name
|
|
|
|
keyvalues (dict[str, Any] | None):
|
|
|
|
column names and values to select the rows with, or None to not
|
|
|
|
apply a WHERE clause.
|
|
|
|
retcols (iterable[str]): the names of the columns to return
|
|
|
|
Returns:
|
|
|
|
defer.Deferred: resolves to list[dict[str, Any]]
|
2014-08-12 10:10:52 -04:00
|
|
|
"""
|
2014-11-06 10:10:55 -05:00
|
|
|
return self.runInteraction(
|
2019-04-03 05:07:29 -04:00
|
|
|
desc, self._simple_select_list_txn, table, keyvalues, retcols
|
2014-11-06 10:10:55 -05:00
|
|
|
)
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@classmethod
|
|
|
|
def _simple_select_list_txn(cls, txn, table, keyvalues, retcols):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes a SELECT query on the named table, which may return zero or
|
|
|
|
more rows, returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Args:
|
2014-11-06 10:10:55 -05:00
|
|
|
txn : Transaction object
|
2016-07-20 10:12:42 -04:00
|
|
|
table (str): the table name
|
|
|
|
keyvalues (dict[str, T] | None):
|
|
|
|
column names and values to select the rows with, or None to not
|
|
|
|
apply a WHERE clause.
|
|
|
|
retcols (iterable[str]): the names of the columns to return
|
2014-08-12 10:10:52 -04:00
|
|
|
"""
|
2015-03-02 05:16:24 -05:00
|
|
|
if keyvalues:
|
2015-03-19 11:59:48 -04:00
|
|
|
sql = "SELECT %s FROM %s WHERE %s" % (
|
2015-03-02 05:16:24 -05:00
|
|
|
", ".join(retcols),
|
|
|
|
table,
|
2019-04-03 05:07:29 -04:00
|
|
|
" AND ".join("%s = ?" % (k,) for k in keyvalues),
|
2015-03-02 05:16:24 -05:00
|
|
|
)
|
2018-05-15 11:52:41 -04:00
|
|
|
txn.execute(sql, list(keyvalues.values()))
|
2015-03-02 05:16:24 -05:00
|
|
|
else:
|
2019-04-03 05:07:29 -04:00
|
|
|
sql = "SELECT %s FROM %s" % (", ".join(retcols), table)
|
2015-03-02 05:16:24 -05:00
|
|
|
txn.execute(sql)
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
return cls.cursor_to_dict(txn)
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2016-01-25 08:36:02 -05:00
|
|
|
@defer.inlineCallbacks
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_many_batch(
|
|
|
|
self,
|
|
|
|
table,
|
|
|
|
column,
|
|
|
|
iterable,
|
|
|
|
retcols,
|
|
|
|
keyvalues={},
|
|
|
|
desc="_simple_select_many_batch",
|
|
|
|
batch_size=100,
|
|
|
|
):
|
2016-01-25 08:36:02 -05:00
|
|
|
"""Executes a SELECT query on the named table, which may return zero or
|
|
|
|
more rows, returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Filters rows by if value of `column` is in `iterable`.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table : string giving the table name
|
|
|
|
column : column name to test for inclusion against `iterable`
|
|
|
|
iterable : list
|
|
|
|
keyvalues : dict of column names and values to select the rows with
|
|
|
|
retcols : list of strings giving the names of the columns to return
|
|
|
|
"""
|
|
|
|
results = []
|
|
|
|
|
2016-01-25 10:59:29 -05:00
|
|
|
if not iterable:
|
|
|
|
defer.returnValue(results)
|
|
|
|
|
2018-05-15 11:52:41 -04:00
|
|
|
# iterables can not be sliced, so convert it to a list first
|
|
|
|
it_list = list(iterable)
|
|
|
|
|
2016-02-02 12:18:50 -05:00
|
|
|
chunks = [
|
2019-04-03 05:07:29 -04:00
|
|
|
it_list[i : i + batch_size] for i in range(0, len(it_list), batch_size)
|
2016-02-02 12:18:50 -05:00
|
|
|
]
|
2016-01-25 08:36:02 -05:00
|
|
|
for chunk in chunks:
|
|
|
|
rows = yield self.runInteraction(
|
|
|
|
desc,
|
|
|
|
self._simple_select_many_txn,
|
2019-04-03 05:07:29 -04:00
|
|
|
table,
|
|
|
|
column,
|
|
|
|
chunk,
|
|
|
|
keyvalues,
|
|
|
|
retcols,
|
2016-01-25 08:36:02 -05:00
|
|
|
)
|
|
|
|
|
|
|
|
results.extend(rows)
|
|
|
|
|
|
|
|
defer.returnValue(results)
|
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@classmethod
|
|
|
|
def _simple_select_many_txn(cls, txn, table, column, iterable, keyvalues, retcols):
|
2016-01-25 08:36:02 -05:00
|
|
|
"""Executes a SELECT query on the named table, which may return zero or
|
|
|
|
more rows, returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Filters rows by if value of `column` is in `iterable`.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn : Transaction object
|
|
|
|
table : string giving the table name
|
|
|
|
column : column name to test for inclusion against `iterable`
|
|
|
|
iterable : list
|
|
|
|
keyvalues : dict of column names and values to select the rows with
|
|
|
|
retcols : list of strings giving the names of the columns to return
|
|
|
|
"""
|
2016-01-25 10:59:29 -05:00
|
|
|
if not iterable:
|
|
|
|
return []
|
|
|
|
|
2016-01-25 08:36:02 -05:00
|
|
|
sql = "SELECT %s FROM %s" % (", ".join(retcols), table)
|
|
|
|
|
|
|
|
clauses = []
|
|
|
|
values = []
|
2019-04-03 05:07:29 -04:00
|
|
|
clauses.append("%s IN (%s)" % (column, ",".join("?" for _ in iterable)))
|
2016-01-25 08:36:02 -05:00
|
|
|
values.extend(iterable)
|
|
|
|
|
2018-05-15 11:52:41 -04:00
|
|
|
for key, value in iteritems(keyvalues):
|
2016-01-25 08:36:02 -05:00
|
|
|
clauses.append("%s = ?" % (key,))
|
|
|
|
values.append(value)
|
|
|
|
|
|
|
|
if clauses:
|
2019-04-03 05:07:29 -04:00
|
|
|
sql = "%s WHERE %s" % (sql, " AND ".join(clauses))
|
2016-01-25 08:36:02 -05:00
|
|
|
|
|
|
|
txn.execute(sql, values)
|
2016-01-26 10:51:06 -05:00
|
|
|
return cls.cursor_to_dict(txn)
|
2016-01-25 08:36:02 -05:00
|
|
|
|
2017-08-25 06:11:37 -04:00
|
|
|
def _simple_update(self, table, keyvalues, updatevalues, desc):
|
|
|
|
return self.runInteraction(
|
2019-04-03 05:07:29 -04:00
|
|
|
desc, self._simple_update_txn, table, keyvalues, updatevalues
|
2017-08-25 06:11:37 -04:00
|
|
|
)
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def _simple_update_txn(txn, table, keyvalues, updatevalues):
|
|
|
|
if keyvalues:
|
2018-05-15 11:52:41 -04:00
|
|
|
where = "WHERE %s" % " AND ".join("%s = ?" % k for k in iterkeys(keyvalues))
|
2017-08-25 06:11:37 -04:00
|
|
|
else:
|
|
|
|
where = ""
|
|
|
|
|
|
|
|
update_sql = "UPDATE %s SET %s %s" % (
|
|
|
|
table,
|
|
|
|
", ".join("%s = ?" % (k,) for k in updatevalues),
|
|
|
|
where,
|
|
|
|
)
|
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
txn.execute(update_sql, list(updatevalues.values()) + list(keyvalues.values()))
|
2017-08-25 06:11:37 -04:00
|
|
|
|
|
|
|
return txn.rowcount
|
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_update_one(
|
|
|
|
self, table, keyvalues, updatevalues, desc="_simple_update_one"
|
|
|
|
):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes an UPDATE query on the named table, setting new values for
|
|
|
|
columns in a row matching the key values.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table : string giving the table name
|
|
|
|
keyvalues : dict of column names and values to select the row with
|
|
|
|
updatevalues : dict giving column names and values to update
|
|
|
|
retcols : optional list of column names to return
|
|
|
|
|
|
|
|
If present, retcols gives a list of column names on which to perform
|
|
|
|
a SELECT statement *before* performing the UPDATE statement. The values
|
|
|
|
of these will be returned in a dict.
|
|
|
|
|
|
|
|
These are performed within the same transaction, allowing an atomic
|
|
|
|
get-and-set. This can be used to implement compare-and-set by putting
|
|
|
|
the update column in the 'keyvalues' dict as well.
|
|
|
|
"""
|
2015-03-20 10:59:48 -04:00
|
|
|
return self.runInteraction(
|
2019-04-03 05:07:29 -04:00
|
|
|
desc, self._simple_update_one_txn, table, keyvalues, updatevalues
|
2015-03-20 10:59:48 -04:00
|
|
|
)
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2017-08-25 06:11:37 -04:00
|
|
|
@classmethod
|
|
|
|
def _simple_update_one_txn(cls, txn, table, keyvalues, updatevalues):
|
|
|
|
rowcount = cls._simple_update_txn(txn, table, keyvalues, updatevalues)
|
2015-03-20 10:59:48 -04:00
|
|
|
|
2017-08-25 06:11:37 -04:00
|
|
|
if rowcount == 0:
|
2018-12-04 05:57:39 -05:00
|
|
|
raise StoreError(404, "No row found (%s)" % (table,))
|
2017-08-25 06:11:37 -04:00
|
|
|
if rowcount > 1:
|
2018-12-04 05:57:39 -05:00
|
|
|
raise StoreError(500, "More than one row matched (%s)" % (table,))
|
2015-03-20 10:59:48 -04:00
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@staticmethod
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_one_txn(txn, table, keyvalues, retcols, allow_none=False):
|
2015-03-24 12:19:01 -04:00
|
|
|
select_sql = "SELECT %s FROM %s WHERE %s" % (
|
2015-03-20 10:59:48 -04:00
|
|
|
", ".join(retcols),
|
|
|
|
table,
|
2019-04-03 05:07:29 -04:00
|
|
|
" AND ".join("%s = ?" % (k,) for k in keyvalues),
|
2015-03-20 10:59:48 -04:00
|
|
|
)
|
|
|
|
|
2018-05-15 11:52:41 -04:00
|
|
|
txn.execute(select_sql, list(keyvalues.values()))
|
2015-03-20 10:59:48 -04:00
|
|
|
row = txn.fetchone()
|
2019-06-06 12:34:07 -04:00
|
|
|
|
2015-03-20 10:59:48 -04:00
|
|
|
if not row:
|
|
|
|
if allow_none:
|
|
|
|
return None
|
2018-12-04 05:57:39 -05:00
|
|
|
raise StoreError(404, "No row found (%s)" % (table,))
|
2015-03-20 10:59:48 -04:00
|
|
|
if txn.rowcount > 1:
|
2018-12-04 05:57:39 -05:00
|
|
|
raise StoreError(500, "More than one row matched (%s)" % (table,))
|
2015-03-20 10:59:48 -04:00
|
|
|
|
|
|
|
return dict(zip(retcols, row))
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2015-03-20 11:59:18 -04:00
|
|
|
def _simple_delete_one(self, table, keyvalues, desc="_simple_delete_one"):
|
2014-08-12 10:10:52 -04:00
|
|
|
"""Executes a DELETE query on the named table, expecting to delete a
|
|
|
|
single row.
|
|
|
|
|
2016-03-01 08:35:37 -05:00
|
|
|
Args:
|
|
|
|
table : string giving the table name
|
|
|
|
keyvalues : dict of column names and values to select the row with
|
|
|
|
"""
|
2019-04-03 05:07:29 -04:00
|
|
|
return self.runInteraction(desc, self._simple_delete_one_txn, table, keyvalues)
|
2016-03-01 08:35:37 -05:00
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def _simple_delete_one_txn(txn, table, keyvalues):
|
|
|
|
"""Executes a DELETE query on the named table, expecting to delete a
|
|
|
|
single row.
|
|
|
|
|
2014-08-12 10:10:52 -04:00
|
|
|
Args:
|
|
|
|
table : string giving the table name
|
|
|
|
keyvalues : dict of column names and values to select the row with
|
|
|
|
"""
|
|
|
|
sql = "DELETE FROM %s WHERE %s" % (
|
|
|
|
table,
|
2019-04-03 05:07:29 -04:00
|
|
|
" AND ".join("%s = ?" % (k,) for k in keyvalues),
|
2014-08-12 10:10:52 -04:00
|
|
|
)
|
|
|
|
|
2018-05-15 11:52:41 -04:00
|
|
|
txn.execute(sql, list(keyvalues.values()))
|
2016-03-01 08:35:37 -05:00
|
|
|
if txn.rowcount == 0:
|
2018-12-04 05:57:39 -05:00
|
|
|
raise StoreError(404, "No row found (%s)" % (table,))
|
2016-03-01 08:35:37 -05:00
|
|
|
if txn.rowcount > 1:
|
2018-12-04 05:57:39 -05:00
|
|
|
raise StoreError(500, "More than one row matched (%s)" % (table,))
|
2014-10-28 07:18:04 -04:00
|
|
|
|
2016-06-30 10:40:58 -04:00
|
|
|
def _simple_delete(self, table, keyvalues, desc):
|
2019-04-03 05:07:29 -04:00
|
|
|
return self.runInteraction(desc, self._simple_delete_txn, table, keyvalues)
|
2016-06-30 10:40:58 -04:00
|
|
|
|
2016-01-26 10:51:06 -05:00
|
|
|
@staticmethod
|
|
|
|
def _simple_delete_txn(txn, table, keyvalues):
|
2014-10-28 07:18:04 -04:00
|
|
|
sql = "DELETE FROM %s WHERE %s" % (
|
|
|
|
table,
|
2019-04-03 05:07:29 -04:00
|
|
|
" AND ".join("%s = ?" % (k,) for k in keyvalues),
|
2014-10-28 07:18:04 -04:00
|
|
|
)
|
|
|
|
|
2019-05-29 06:58:32 -04:00
|
|
|
txn.execute(sql, list(keyvalues.values()))
|
|
|
|
return txn.rowcount
|
2014-08-12 10:10:52 -04:00
|
|
|
|
2017-03-13 13:53:23 -04:00
|
|
|
def _simple_delete_many(self, table, column, iterable, keyvalues, desc):
|
|
|
|
return self.runInteraction(
|
|
|
|
desc, self._simple_delete_many_txn, table, column, iterable, keyvalues
|
|
|
|
)
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def _simple_delete_many_txn(txn, table, column, iterable, keyvalues):
|
|
|
|
"""Executes a DELETE query on the named table.
|
|
|
|
|
|
|
|
Filters rows by if value of `column` is in `iterable`.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn : Transaction object
|
|
|
|
table : string giving the table name
|
|
|
|
column : column name to test for inclusion against `iterable`
|
|
|
|
iterable : list
|
|
|
|
keyvalues : dict of column names and values to select the rows with
|
2019-05-29 06:58:32 -04:00
|
|
|
|
|
|
|
Returns:
|
|
|
|
int: Number rows deleted
|
2017-03-13 13:53:23 -04:00
|
|
|
"""
|
|
|
|
if not iterable:
|
2019-05-29 06:58:32 -04:00
|
|
|
return 0
|
2017-03-13 13:53:23 -04:00
|
|
|
|
|
|
|
sql = "DELETE FROM %s" % table
|
|
|
|
|
|
|
|
clauses = []
|
|
|
|
values = []
|
2019-04-03 05:07:29 -04:00
|
|
|
clauses.append("%s IN (%s)" % (column, ",".join("?" for _ in iterable)))
|
2017-03-13 13:53:23 -04:00
|
|
|
values.extend(iterable)
|
|
|
|
|
2018-05-15 11:52:41 -04:00
|
|
|
for key, value in iteritems(keyvalues):
|
2017-03-13 13:53:23 -04:00
|
|
|
clauses.append("%s = ?" % (key,))
|
|
|
|
values.append(value)
|
|
|
|
|
|
|
|
if clauses:
|
2019-04-03 05:07:29 -04:00
|
|
|
sql = "%s WHERE %s" % (sql, " AND ".join(clauses))
|
2019-05-29 06:58:32 -04:00
|
|
|
txn.execute(sql, values)
|
|
|
|
|
|
|
|
return txn.rowcount
|
2017-03-13 13:53:23 -04:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _get_cache_dict(
|
|
|
|
self, db_conn, table, entity_column, stream_column, max_value, limit=100000
|
|
|
|
):
|
2016-04-06 08:05:19 -04:00
|
|
|
# Fetch a mapping of room_id -> max stream position for "recent" rooms.
|
|
|
|
# It doesn't really matter how many we get, the StreamChangeCache will
|
|
|
|
# do the right thing to ensure it respects the max size of cache.
|
|
|
|
sql = (
|
|
|
|
"SELECT %(entity)s, MAX(%(stream)s) FROM %(table)s"
|
2017-01-10 09:34:50 -05:00
|
|
|
" WHERE %(stream)s > ? - %(limit)s"
|
2016-04-06 08:05:19 -04:00
|
|
|
" GROUP BY %(entity)s"
|
|
|
|
) % {
|
|
|
|
"table": table,
|
|
|
|
"entity": entity_column,
|
|
|
|
"stream": stream_column,
|
2017-01-10 09:34:50 -05:00
|
|
|
"limit": limit,
|
2016-04-06 08:05:19 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
sql = self.database_engine.convert_param_style(sql)
|
|
|
|
|
|
|
|
txn = db_conn.cursor()
|
|
|
|
txn.execute(sql, (int(max_value),))
|
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
cache = {row[0]: int(row[1]) for row in txn}
|
2016-04-06 08:05:19 -04:00
|
|
|
|
2017-03-23 13:53:49 -04:00
|
|
|
txn.close()
|
|
|
|
|
2016-04-06 08:05:19 -04:00
|
|
|
if cache:
|
2018-05-15 11:52:41 -04:00
|
|
|
min_val = min(itervalues(cache))
|
2016-04-06 08:05:19 -04:00
|
|
|
else:
|
|
|
|
min_val = max_value
|
|
|
|
|
|
|
|
return cache, min_val
|
|
|
|
|
2016-08-15 05:21:25 -04:00
|
|
|
def _invalidate_cache_and_stream(self, txn, cache_func, keys):
|
2016-08-15 06:45:57 -04:00
|
|
|
"""Invalidates the cache and adds it to the cache stream so slaves
|
|
|
|
will know to invalidate their caches.
|
|
|
|
|
|
|
|
This should only be used to invalidate caches where slaves won't
|
|
|
|
otherwise know from other replication streams that the cache should
|
|
|
|
be invalidated.
|
|
|
|
"""
|
2016-08-15 05:21:25 -04:00
|
|
|
txn.call_after(cache_func.invalidate, keys)
|
2019-02-18 12:53:31 -05:00
|
|
|
self._send_invalidation_to_replication(txn, cache_func.__name__, keys)
|
|
|
|
|
|
|
|
def _invalidate_state_caches_and_stream(self, txn, room_id, members_changed):
|
|
|
|
"""Special case invalidation of caches based on current state.
|
|
|
|
|
|
|
|
We special case this so that we can batch the cache invalidations into a
|
|
|
|
single replication poke.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn
|
2019-02-19 06:24:59 -05:00
|
|
|
room_id (str): Room where state changed
|
2019-02-19 06:38:40 -05:00
|
|
|
members_changed (iterable[str]): The user_ids of members that have changed
|
2019-02-18 12:53:31 -05:00
|
|
|
"""
|
|
|
|
txn.call_after(self._invalidate_state_caches, room_id, members_changed)
|
|
|
|
|
2019-02-27 05:28:37 -05:00
|
|
|
# We need to be careful that the size of the `members_changed` list
|
|
|
|
# isn't so large that it causes problems sending over replication, so we
|
|
|
|
# send them in chunks.
|
|
|
|
# Max line length is 16K, and max user ID length is 255, so 50 should
|
|
|
|
# be safe.
|
|
|
|
for chunk in batch_iter(members_changed, 50):
|
|
|
|
keys = itertools.chain([room_id], chunk)
|
2019-04-03 05:07:29 -04:00
|
|
|
self._send_invalidation_to_replication(txn, _CURRENT_STATE_CACHE_NAME, keys)
|
2019-02-18 12:53:31 -05:00
|
|
|
|
|
|
|
def _invalidate_state_caches(self, room_id, members_changed):
|
|
|
|
"""Invalidates caches that are based on the current state, but does
|
|
|
|
not stream invalidations down replication.
|
|
|
|
|
|
|
|
Args:
|
2019-02-19 06:24:59 -05:00
|
|
|
room_id (str): Room where state changed
|
2019-02-19 06:38:40 -05:00
|
|
|
members_changed (iterable[str]): The user_ids of members that have
|
|
|
|
changed
|
2019-02-18 12:53:31 -05:00
|
|
|
"""
|
|
|
|
for host in set(get_domain_from_id(u) for u in members_changed):
|
2019-04-03 05:07:29 -04:00
|
|
|
self._attempt_to_invalidate_cache("is_host_joined", (room_id, host))
|
|
|
|
self._attempt_to_invalidate_cache("was_host_joined", (room_id, host))
|
2019-02-22 09:38:14 -05:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
self._attempt_to_invalidate_cache("get_users_in_room", (room_id,))
|
|
|
|
self._attempt_to_invalidate_cache("get_room_summary", (room_id,))
|
|
|
|
self._attempt_to_invalidate_cache("get_current_state_ids", (room_id,))
|
2019-02-22 09:38:14 -05:00
|
|
|
|
|
|
|
def _attempt_to_invalidate_cache(self, cache_name, key):
|
|
|
|
"""Attempts to invalidate the cache of the given name, ignoring if the
|
|
|
|
cache doesn't exist. Mainly used for invalidating caches on workers,
|
|
|
|
where they may not have the cache.
|
2019-02-18 12:53:31 -05:00
|
|
|
|
2019-02-22 09:38:14 -05:00
|
|
|
Args:
|
|
|
|
cache_name (str)
|
|
|
|
key (tuple)
|
|
|
|
"""
|
|
|
|
try:
|
|
|
|
getattr(self, cache_name).invalidate(key)
|
|
|
|
except AttributeError:
|
|
|
|
# We probably haven't pulled in the cache in this worker,
|
|
|
|
# which is fine.
|
|
|
|
pass
|
2019-02-18 12:53:31 -05:00
|
|
|
|
|
|
|
def _send_invalidation_to_replication(self, txn, cache_name, keys):
|
|
|
|
"""Notifies replication that given cache has been invalidated.
|
|
|
|
|
|
|
|
Note that this does *not* invalidate the cache locally.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn
|
|
|
|
cache_name (str)
|
2019-02-19 06:24:59 -05:00
|
|
|
keys (iterable[str])
|
2019-02-18 12:53:31 -05:00
|
|
|
"""
|
2016-08-15 05:21:25 -04:00
|
|
|
|
2016-08-15 06:16:45 -04:00
|
|
|
if isinstance(self.database_engine, PostgresEngine):
|
2016-08-15 08:45:26 -04:00
|
|
|
# get_next() returns a context manager which is designed to wrap
|
|
|
|
# the transaction. However, we want to only get an ID when we want
|
|
|
|
# to use it, here, so we need to call __enter__ manually, and have
|
|
|
|
# __exit__ called after the transaction finishes.
|
2016-08-15 06:16:45 -04:00
|
|
|
ctx = self._cache_id_gen.get_next()
|
|
|
|
stream_id = ctx.__enter__()
|
2018-03-02 09:43:29 -05:00
|
|
|
txn.call_on_exception(ctx.__exit__, None, None, None)
|
|
|
|
txn.call_after(ctx.__exit__, None, None, None)
|
2016-08-15 13:10:54 -04:00
|
|
|
txn.call_after(self.hs.get_notifier().on_new_replication_data)
|
2016-08-15 06:16:45 -04:00
|
|
|
|
|
|
|
self._simple_insert_txn(
|
|
|
|
txn,
|
2016-08-15 06:45:57 -04:00
|
|
|
table="cache_invalidation_stream",
|
2016-08-15 06:16:45 -04:00
|
|
|
values={
|
|
|
|
"stream_id": stream_id,
|
2019-02-18 12:53:31 -05:00
|
|
|
"cache_func": cache_name,
|
2016-08-15 06:16:45 -04:00
|
|
|
"keys": list(keys),
|
|
|
|
"invalidation_ts": self.clock.time_msec(),
|
2019-04-03 05:07:29 -04:00
|
|
|
},
|
2016-08-15 06:16:45 -04:00
|
|
|
)
|
2016-08-15 05:21:25 -04:00
|
|
|
|
2016-08-15 06:16:45 -04:00
|
|
|
def get_all_updated_caches(self, last_id, current_id, limit):
|
2016-08-16 12:05:34 -04:00
|
|
|
if last_id == current_id:
|
|
|
|
return defer.succeed([])
|
|
|
|
|
2016-08-15 06:16:45 -04:00
|
|
|
def get_all_updated_caches_txn(txn):
|
|
|
|
# We purposefully don't bound by the current token, as we want to
|
|
|
|
# send across cache invalidations as quickly as possible. Cache
|
|
|
|
# invalidations are idempotent, so duplicates are fine.
|
|
|
|
sql = (
|
2016-08-15 06:45:57 -04:00
|
|
|
"SELECT stream_id, cache_func, keys, invalidation_ts"
|
|
|
|
" FROM cache_invalidation_stream"
|
2016-08-15 06:16:45 -04:00
|
|
|
" WHERE stream_id > ? ORDER BY stream_id ASC LIMIT ?"
|
|
|
|
)
|
2019-04-03 05:07:29 -04:00
|
|
|
txn.execute(sql, (last_id, limit))
|
2016-08-15 06:16:45 -04:00
|
|
|
return txn.fetchall()
|
2019-04-03 05:07:29 -04:00
|
|
|
|
|
|
|
return self.runInteraction("get_all_updated_caches", get_all_updated_caches_txn)
|
2016-08-15 05:21:25 -04:00
|
|
|
|
2016-08-15 06:16:45 -04:00
|
|
|
def get_cache_stream_token(self):
|
|
|
|
if self._cache_id_gen:
|
|
|
|
return self._cache_id_gen.get_current_token()
|
|
|
|
else:
|
|
|
|
return 0
|
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_list_paginate(
|
2019-04-04 09:21:16 -04:00
|
|
|
self,
|
|
|
|
table,
|
|
|
|
keyvalues,
|
|
|
|
orderby,
|
|
|
|
start,
|
|
|
|
limit,
|
|
|
|
retcols,
|
|
|
|
order_direction="ASC",
|
|
|
|
desc="_simple_select_list_paginate",
|
2019-04-03 05:07:29 -04:00
|
|
|
):
|
2019-04-04 09:21:16 -04:00
|
|
|
"""
|
|
|
|
Executes a SELECT query on the named table with start and limit,
|
2017-02-02 08:02:26 -05:00
|
|
|
of row numbers, which may return zero or number of rows from start to limit,
|
|
|
|
returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table (str): the table name
|
2019-04-04 09:21:16 -04:00
|
|
|
keyvalues (dict[str, T] | None):
|
2017-02-02 08:02:26 -05:00
|
|
|
column names and values to select the rows with, or None to not
|
|
|
|
apply a WHERE clause.
|
2019-04-04 09:21:16 -04:00
|
|
|
orderby (str): Column to order the results by.
|
|
|
|
start (int): Index to begin the query at.
|
|
|
|
limit (int): Number of results to return.
|
2017-02-02 08:02:26 -05:00
|
|
|
retcols (iterable[str]): the names of the columns to return
|
2019-04-04 09:21:16 -04:00
|
|
|
order_direction (str): Whether the results should be ordered "ASC" or "DESC".
|
2017-02-02 08:02:26 -05:00
|
|
|
Returns:
|
|
|
|
defer.Deferred: resolves to list[dict[str, Any]]
|
|
|
|
"""
|
|
|
|
return self.runInteraction(
|
|
|
|
desc,
|
|
|
|
self._simple_select_list_paginate_txn,
|
2019-04-03 05:07:29 -04:00
|
|
|
table,
|
|
|
|
keyvalues,
|
2019-04-04 09:21:16 -04:00
|
|
|
orderby,
|
|
|
|
start,
|
|
|
|
limit,
|
2019-04-03 05:07:29 -04:00
|
|
|
retcols,
|
2019-04-04 09:21:16 -04:00
|
|
|
order_direction=order_direction,
|
2017-02-02 08:02:26 -05:00
|
|
|
)
|
|
|
|
|
|
|
|
@classmethod
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_select_list_paginate_txn(
|
2019-04-04 09:21:16 -04:00
|
|
|
cls,
|
|
|
|
txn,
|
|
|
|
table,
|
|
|
|
keyvalues,
|
|
|
|
orderby,
|
|
|
|
start,
|
|
|
|
limit,
|
|
|
|
retcols,
|
|
|
|
order_direction="ASC",
|
2019-04-03 05:07:29 -04:00
|
|
|
):
|
2019-04-04 09:21:16 -04:00
|
|
|
"""
|
|
|
|
Executes a SELECT query on the named table with start and limit,
|
2017-02-02 08:02:26 -05:00
|
|
|
of row numbers, which may return zero or number of rows from start to limit,
|
|
|
|
returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn : Transaction object
|
|
|
|
table (str): the table name
|
|
|
|
keyvalues (dict[str, T] | None):
|
|
|
|
column names and values to select the rows with, or None to not
|
|
|
|
apply a WHERE clause.
|
2019-04-04 09:21:16 -04:00
|
|
|
orderby (str): Column to order the results by.
|
|
|
|
start (int): Index to begin the query at.
|
|
|
|
limit (int): Number of results to return.
|
2017-02-02 08:02:26 -05:00
|
|
|
retcols (iterable[str]): the names of the columns to return
|
2019-04-04 09:21:16 -04:00
|
|
|
order_direction (str): Whether the results should be ordered "ASC" or "DESC".
|
2017-02-02 08:02:26 -05:00
|
|
|
Returns:
|
|
|
|
defer.Deferred: resolves to list[dict[str, Any]]
|
|
|
|
"""
|
2019-04-04 09:21:16 -04:00
|
|
|
if order_direction not in ["ASC", "DESC"]:
|
|
|
|
raise ValueError("order_direction must be one of 'ASC' or 'DESC'.")
|
|
|
|
|
2017-02-02 08:02:26 -05:00
|
|
|
if keyvalues:
|
2019-04-04 09:21:16 -04:00
|
|
|
where_clause = "WHERE " + " AND ".join("%s = ?" % (k,) for k in keyvalues)
|
2017-02-02 08:02:26 -05:00
|
|
|
else:
|
2019-04-04 09:21:16 -04:00
|
|
|
where_clause = ""
|
2017-02-02 08:02:26 -05:00
|
|
|
|
2019-04-04 09:21:16 -04:00
|
|
|
sql = "SELECT %s FROM %s %s ORDER BY %s %s LIMIT ? OFFSET ?" % (
|
|
|
|
", ".join(retcols),
|
2019-04-03 05:07:29 -04:00
|
|
|
table,
|
2019-04-04 09:21:16 -04:00
|
|
|
where_clause,
|
|
|
|
orderby,
|
|
|
|
order_direction,
|
2017-02-02 08:02:26 -05:00
|
|
|
)
|
2019-04-04 09:21:16 -04:00
|
|
|
txn.execute(sql, list(keyvalues.values()) + [limit, start])
|
|
|
|
|
|
|
|
return cls.cursor_to_dict(txn)
|
2017-02-02 08:02:26 -05:00
|
|
|
|
|
|
|
def get_user_count_txn(self, txn):
|
2018-08-16 04:46:50 -04:00
|
|
|
"""Get a total number of registered users in the users list.
|
2017-02-02 08:02:26 -05:00
|
|
|
|
|
|
|
Args:
|
|
|
|
txn : Transaction object
|
|
|
|
Returns:
|
2018-08-16 04:46:50 -04:00
|
|
|
int : number of users
|
2017-02-02 08:02:26 -05:00
|
|
|
"""
|
|
|
|
sql_count = "SELECT COUNT(*) FROM users WHERE is_guest = 0;"
|
|
|
|
txn.execute(sql_count)
|
2018-08-16 04:46:50 -04:00
|
|
|
return txn.fetchone()[0]
|
2017-02-02 08:02:26 -05:00
|
|
|
|
2019-04-03 05:07:29 -04:00
|
|
|
def _simple_search_list(
|
|
|
|
self, table, term, col, retcols, desc="_simple_search_list"
|
|
|
|
):
|
2017-02-02 08:02:26 -05:00
|
|
|
"""Executes a SELECT query on the named table, which may return zero or
|
|
|
|
more rows, returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
table (str): the table name
|
|
|
|
term (str | None):
|
|
|
|
term for searching the table matched to a column.
|
|
|
|
col (str): column to query term should be matched to
|
|
|
|
retcols (iterable[str]): the names of the columns to return
|
|
|
|
Returns:
|
|
|
|
defer.Deferred: resolves to list[dict[str, Any]] or None
|
|
|
|
"""
|
|
|
|
|
|
|
|
return self.runInteraction(
|
2019-04-03 05:07:29 -04:00
|
|
|
desc, self._simple_search_list_txn, table, term, col, retcols
|
2017-02-02 08:02:26 -05:00
|
|
|
)
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def _simple_search_list_txn(cls, txn, table, term, col, retcols):
|
|
|
|
"""Executes a SELECT query on the named table, which may return zero or
|
|
|
|
more rows, returning the result as a list of dicts.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
txn : Transaction object
|
|
|
|
table (str): the table name
|
|
|
|
term (str | None):
|
|
|
|
term for searching the table matched to a column.
|
|
|
|
col (str): column to query term should be matched to
|
|
|
|
retcols (iterable[str]): the names of the columns to return
|
|
|
|
Returns:
|
|
|
|
defer.Deferred: resolves to list[dict[str, Any]] or None
|
|
|
|
"""
|
|
|
|
if term:
|
2019-04-03 05:07:29 -04:00
|
|
|
sql = "SELECT %s FROM %s WHERE %s LIKE ?" % (", ".join(retcols), table, col)
|
2017-02-02 08:02:26 -05:00
|
|
|
termvalues = ["%%" + term + "%%"]
|
|
|
|
txn.execute(sql, termvalues)
|
|
|
|
else:
|
|
|
|
return 0
|
|
|
|
|
|
|
|
return cls.cursor_to_dict(txn)
|
|
|
|
|
2019-02-27 05:21:49 -05:00
|
|
|
@property
|
|
|
|
def database_engine_name(self):
|
|
|
|
return self.database_engine.module.__name__
|
|
|
|
|
|
|
|
def get_server_version(self):
|
|
|
|
"""Returns a string describing the server version number"""
|
|
|
|
return self.database_engine.server_version
|
|
|
|
|
2014-09-23 10:28:32 -04:00
|
|
|
|
2015-03-20 09:52:56 -04:00
|
|
|
class _RollbackButIsFineException(Exception):
|
|
|
|
""" This exception is used to rollback a transaction without implying
|
|
|
|
something went wrong.
|
|
|
|
"""
|
2019-04-03 05:07:29 -04:00
|
|
|
|
2015-03-20 09:52:56 -04:00
|
|
|
pass
|
2018-08-30 10:19:58 -04:00
|
|
|
|
|
|
|
|
|
|
|
def db_to_json(db_content):
|
|
|
|
"""
|
|
|
|
Take some data from a database row and return a JSON-decoded object.
|
|
|
|
|
|
|
|
Args:
|
|
|
|
db_content (memoryview|buffer|bytes|bytearray|unicode)
|
|
|
|
"""
|
|
|
|
# psycopg2 on Python 3 returns memoryview objects, which we need to
|
|
|
|
# cast to bytes to decode
|
|
|
|
if isinstance(db_content, memoryview):
|
|
|
|
db_content = db_content.tobytes()
|
|
|
|
|
|
|
|
# psycopg2 on Python 2 returns buffer objects, which we need to cast to
|
|
|
|
# bytes to decode
|
2018-10-19 20:16:55 -04:00
|
|
|
if PY2 and isinstance(db_content, builtins.buffer):
|
2018-08-30 10:19:58 -04:00
|
|
|
db_content = bytes(db_content)
|
|
|
|
|
|
|
|
# Decode it to a Unicode string before feeding it to json.loads, so we
|
|
|
|
# consistenty get a Unicode-containing object out.
|
|
|
|
if isinstance(db_content, (bytes, bytearray)):
|
2019-06-20 05:32:02 -04:00
|
|
|
db_content = db_content.decode("utf8")
|
2018-08-30 10:19:58 -04:00
|
|
|
|
|
|
|
try:
|
|
|
|
return json.loads(db_content)
|
|
|
|
except Exception:
|
|
|
|
logging.warning("Tried to decode '%r' as JSON and failed", db_content)
|
|
|
|
raise
|