2016-01-06 21:26:29 -07:00
|
|
|
# Copyright 2014-2016 OpenMarket Ltd
|
2014-10-15 03:04:55 -06:00
|
|
|
#
|
|
|
|
# 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.
|
2017-11-13 03:30:38 -07:00
|
|
|
import logging
|
2020-12-30 06:09:53 -07:00
|
|
|
from typing import (
|
|
|
|
TYPE_CHECKING,
|
|
|
|
Awaitable,
|
2021-10-12 03:44:59 -06:00
|
|
|
Collection,
|
2020-12-30 06:09:53 -07:00
|
|
|
Dict,
|
|
|
|
Iterable,
|
|
|
|
List,
|
2021-10-12 03:44:59 -06:00
|
|
|
Mapping,
|
2020-12-30 06:09:53 -07:00
|
|
|
Optional,
|
|
|
|
Set,
|
|
|
|
Tuple,
|
|
|
|
TypeVar,
|
|
|
|
)
|
2014-10-15 03:04:55 -06:00
|
|
|
|
2018-10-25 10:49:55 -06:00
|
|
|
import attr
|
2021-09-14 09:35:53 -06:00
|
|
|
from frozendict import frozendict
|
2018-10-25 10:49:55 -06:00
|
|
|
|
2018-07-25 15:10:39 -06:00
|
|
|
from synapse.api.constants import EventTypes
|
2020-07-28 14:09:53 -06:00
|
|
|
from synapse.events import EventBase
|
2021-10-12 03:44:59 -06:00
|
|
|
from synapse.types import MutableStateMap, StateKey, StateMap
|
2018-07-09 00:09:20 -06:00
|
|
|
|
2020-12-30 06:09:53 -07:00
|
|
|
if TYPE_CHECKING:
|
2021-09-14 09:35:53 -06:00
|
|
|
from typing import FrozenSet # noqa: used within quoted type hint; flake8 sad
|
|
|
|
|
2021-03-23 05:12:48 -06:00
|
|
|
from synapse.server import HomeServer
|
2020-12-30 06:09:53 -07:00
|
|
|
from synapse.storage.databases import Databases
|
|
|
|
|
2015-01-06 04:18:12 -07:00
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
2020-01-16 06:31:22 -07:00
|
|
|
# Used for generic functions below
|
|
|
|
T = TypeVar("T")
|
|
|
|
|
2014-10-15 03:04:55 -06:00
|
|
|
|
2022-01-13 06:49:28 -07:00
|
|
|
@attr.s(slots=True, frozen=True, auto_attribs=True)
|
2020-09-04 04:54:56 -06:00
|
|
|
class StateFilter:
|
2018-10-25 10:49:55 -06:00
|
|
|
"""A filter used when querying for state.
|
|
|
|
|
|
|
|
Attributes:
|
2020-07-28 14:09:53 -06:00
|
|
|
types: Map from type to set of state keys (or None). This specifies
|
|
|
|
which state_keys for the given type to fetch from the DB. If None
|
|
|
|
then all events with that type are fetched. If the set is empty
|
|
|
|
then no events with that type are fetched.
|
|
|
|
include_others: Whether to fetch events with types that do not
|
2018-10-25 10:49:55 -06:00
|
|
|
appear in `types`.
|
|
|
|
"""
|
|
|
|
|
2022-01-13 06:49:28 -07:00
|
|
|
types: "frozendict[str, Optional[FrozenSet[str]]]"
|
|
|
|
include_others: bool = False
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
def __attrs_post_init__(self):
|
|
|
|
# If `include_others` is set we canonicalise the filter by removing
|
|
|
|
# wildcards from the types dictionary
|
|
|
|
if self.include_others:
|
2021-09-14 09:35:53 -06:00
|
|
|
# this is needed to work around the fact that StateFilter is frozen
|
|
|
|
object.__setattr__(
|
|
|
|
self,
|
|
|
|
"types",
|
|
|
|
frozendict({k: v for k, v in self.types.items() if v is not None}),
|
|
|
|
)
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
@staticmethod
|
2020-07-28 14:09:53 -06:00
|
|
|
def all() -> "StateFilter":
|
2022-01-27 03:54:27 -07:00
|
|
|
"""Returns a filter that fetches everything.
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
Returns:
|
2022-01-27 03:54:27 -07:00
|
|
|
The state filter.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
2022-01-27 03:54:27 -07:00
|
|
|
return _ALL_STATE_FILTER
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
@staticmethod
|
2020-07-28 14:09:53 -06:00
|
|
|
def none() -> "StateFilter":
|
2022-01-27 03:54:27 -07:00
|
|
|
"""Returns a filter that fetches nothing.
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
The new state filter.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
2022-01-27 03:54:27 -07:00
|
|
|
return _NONE_STATE_FILTER
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
@staticmethod
|
2020-07-28 14:09:53 -06:00
|
|
|
def from_types(types: Iterable[Tuple[str, Optional[str]]]) -> "StateFilter":
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Creates a filter that only fetches the given types
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
types: A list of type and state keys to fetch. A state_key of None
|
|
|
|
fetches everything for that type
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
The new state filter.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
2021-07-15 10:46:54 -06:00
|
|
|
type_dict: Dict[str, Optional[Set[str]]] = {}
|
2018-10-25 10:49:55 -06:00
|
|
|
for typ, s in types:
|
|
|
|
if typ in type_dict:
|
|
|
|
if type_dict[typ] is None:
|
|
|
|
continue
|
|
|
|
|
|
|
|
if s is None:
|
|
|
|
type_dict[typ] = None
|
|
|
|
continue
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
type_dict.setdefault(typ, set()).add(s) # type: ignore
|
2018-10-25 10:49:55 -06:00
|
|
|
|
2021-09-14 09:35:53 -06:00
|
|
|
return StateFilter(
|
|
|
|
types=frozendict(
|
|
|
|
(k, frozenset(v) if v is not None else None)
|
|
|
|
for k, v in type_dict.items()
|
|
|
|
)
|
|
|
|
)
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
@staticmethod
|
2020-07-28 14:09:53 -06:00
|
|
|
def from_lazy_load_member_list(members: Iterable[str]) -> "StateFilter":
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Creates a filter that returns all non-member events, plus the member
|
|
|
|
events for the given users
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
members: Set of user IDs
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
The new state filter
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
2021-09-14 09:35:53 -06:00
|
|
|
return StateFilter(
|
|
|
|
types=frozendict({EventTypes.Member: frozenset(members)}),
|
|
|
|
include_others=True,
|
|
|
|
)
|
2018-10-25 10:49:55 -06:00
|
|
|
|
2021-10-12 03:44:59 -06:00
|
|
|
@staticmethod
|
|
|
|
def freeze(types: Mapping[str, Optional[Collection[str]]], include_others: bool):
|
|
|
|
"""
|
|
|
|
Returns a (frozen) StateFilter with the same contents as the parameters
|
|
|
|
specified here, which can be made of mutable types.
|
|
|
|
"""
|
|
|
|
types_with_frozen_values: Dict[str, Optional[FrozenSet[str]]] = {}
|
|
|
|
for state_types, state_keys in types.items():
|
|
|
|
if state_keys is not None:
|
|
|
|
types_with_frozen_values[state_types] = frozenset(state_keys)
|
|
|
|
else:
|
|
|
|
types_with_frozen_values[state_types] = None
|
|
|
|
|
|
|
|
return StateFilter(
|
|
|
|
frozendict(types_with_frozen_values), include_others=include_others
|
|
|
|
)
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
def return_expanded(self) -> "StateFilter":
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Creates a new StateFilter where type wild cards have been removed
|
|
|
|
(except for memberships). The returned filter is a superset of the
|
|
|
|
current one, i.e. anything that passes the current filter will pass
|
|
|
|
the returned filter.
|
|
|
|
|
|
|
|
This helps the caching as the DictionaryCache knows if it has *all* the
|
|
|
|
state, but does not know if it has all of the keys of a particular type,
|
|
|
|
which makes wildcard lookups expensive unless we have a complete cache.
|
|
|
|
Hence, if we are doing a wildcard lookup, populate the cache fully so
|
|
|
|
that we can do an efficient lookup next time.
|
|
|
|
|
|
|
|
Note that since we have two caches, one for membership events and one for
|
|
|
|
other events, we can be a bit more clever than simply returning
|
|
|
|
`StateFilter.all()` if `has_wildcards()` is True.
|
|
|
|
|
|
|
|
We return a StateFilter where:
|
|
|
|
1. the list of membership events to return is the same
|
|
|
|
2. if there is a wildcard that matches non-member events we
|
|
|
|
return all non-member events
|
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
The new state filter.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
|
|
|
|
|
|
|
if self.is_full():
|
|
|
|
# If we're going to return everything then there's nothing to do
|
|
|
|
return self
|
|
|
|
|
|
|
|
if not self.has_wildcards():
|
|
|
|
# If there are no wild cards, there's nothing to do
|
|
|
|
return self
|
|
|
|
|
|
|
|
if EventTypes.Member in self.types:
|
|
|
|
get_all_members = self.types[EventTypes.Member] is None
|
|
|
|
else:
|
|
|
|
get_all_members = self.include_others
|
|
|
|
|
|
|
|
has_non_member_wildcard = self.include_others or any(
|
|
|
|
state_keys is None
|
2020-06-15 05:03:36 -06:00
|
|
|
for t, state_keys in self.types.items()
|
2018-10-25 10:49:55 -06:00
|
|
|
if t != EventTypes.Member
|
|
|
|
)
|
|
|
|
|
|
|
|
if not has_non_member_wildcard:
|
|
|
|
# If there are no non-member wild cards we can just return ourselves
|
|
|
|
return self
|
|
|
|
|
|
|
|
if get_all_members:
|
|
|
|
# We want to return everything.
|
|
|
|
return StateFilter.all()
|
2022-02-18 07:54:31 -07:00
|
|
|
elif EventTypes.Member in self.types:
|
2018-10-25 10:49:55 -06:00
|
|
|
# We want to return all non-members, but only particular
|
|
|
|
# memberships
|
|
|
|
return StateFilter(
|
2021-09-14 09:35:53 -06:00
|
|
|
types=frozendict({EventTypes.Member: self.types[EventTypes.Member]}),
|
2018-10-25 10:49:55 -06:00
|
|
|
include_others=True,
|
|
|
|
)
|
2022-02-18 07:54:31 -07:00
|
|
|
else:
|
|
|
|
# We want to return all non-members
|
|
|
|
return _ALL_NON_MEMBER_STATE_FILTER
|
2018-10-25 10:49:55 -06:00
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
def make_sql_filter_clause(self) -> Tuple[str, List[str]]:
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Converts the filter to an SQL clause.
|
|
|
|
|
|
|
|
For example:
|
|
|
|
|
|
|
|
f = StateFilter.from_types([("m.room.create", "")])
|
|
|
|
clause, args = f.make_sql_filter_clause()
|
|
|
|
clause == "(type = ? AND state_key = ?)"
|
|
|
|
args == ['m.room.create', '']
|
|
|
|
|
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
The SQL string (may be empty) and arguments. An empty SQL string is
|
|
|
|
returned when the filter matches everything (i.e. is "full").
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
|
|
|
|
|
|
|
where_clause = ""
|
2021-07-15 10:46:54 -06:00
|
|
|
where_args: List[str] = []
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
if self.is_full():
|
|
|
|
return where_clause, where_args
|
|
|
|
|
|
|
|
if not self.include_others and not self.types:
|
|
|
|
# i.e. this is an empty filter, so we need to return a clause that
|
|
|
|
# will match nothing
|
|
|
|
return "1 = 2", []
|
|
|
|
|
|
|
|
# First we build up a lost of clauses for each type/state_key combo
|
|
|
|
clauses = []
|
2020-06-15 05:03:36 -06:00
|
|
|
for etype, state_keys in self.types.items():
|
2018-10-25 10:49:55 -06:00
|
|
|
if state_keys is None:
|
|
|
|
clauses.append("(type = ?)")
|
|
|
|
where_args.append(etype)
|
|
|
|
continue
|
|
|
|
|
|
|
|
for state_key in state_keys:
|
|
|
|
clauses.append("(type = ? AND state_key = ?)")
|
|
|
|
where_args.extend((etype, state_key))
|
|
|
|
|
|
|
|
# This will match anything that appears in `self.types`
|
|
|
|
where_clause = " OR ".join(clauses)
|
|
|
|
|
|
|
|
# If we want to include stuff that's not in the types dict then we add
|
|
|
|
# a `OR type NOT IN (...)` clause to the end.
|
|
|
|
if self.include_others:
|
|
|
|
if where_clause:
|
|
|
|
where_clause += " OR "
|
|
|
|
|
2019-04-03 03:07:29 -06:00
|
|
|
where_clause += "type NOT IN (%s)" % (",".join(["?"] * len(self.types)),)
|
2018-10-25 10:49:55 -06:00
|
|
|
where_args.extend(self.types)
|
|
|
|
|
|
|
|
return where_clause, where_args
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
def max_entries_returned(self) -> Optional[int]:
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Returns the maximum number of entries this filter will return if
|
|
|
|
known, otherwise returns None.
|
|
|
|
|
|
|
|
For example a simple state filter asking for `("m.room.create", "")`
|
|
|
|
will return 1, whereas the default state filter will return None.
|
|
|
|
|
|
|
|
This is used to bail out early if the right number of entries have been
|
|
|
|
fetched.
|
|
|
|
"""
|
|
|
|
if self.has_wildcards():
|
|
|
|
return None
|
|
|
|
|
|
|
|
return len(self.concrete_types())
|
|
|
|
|
2021-09-14 09:35:53 -06:00
|
|
|
def filter_state(self, state_dict: StateMap[T]) -> MutableStateMap[T]:
|
|
|
|
"""Returns the state filtered with by this StateFilter.
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
Args:
|
2020-01-16 06:31:22 -07:00
|
|
|
state: The state map to filter
|
2018-10-25 10:49:55 -06:00
|
|
|
|
|
|
|
Returns:
|
2021-09-14 09:35:53 -06:00
|
|
|
The filtered state map.
|
|
|
|
This is a copy, so it's safe to mutate.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
|
|
|
if self.is_full():
|
|
|
|
return dict(state_dict)
|
|
|
|
|
|
|
|
filtered_state = {}
|
2020-06-15 05:03:36 -06:00
|
|
|
for k, v in state_dict.items():
|
2018-10-25 10:49:55 -06:00
|
|
|
typ, state_key = k
|
|
|
|
if typ in self.types:
|
|
|
|
state_keys = self.types[typ]
|
|
|
|
if state_keys is None or state_key in state_keys:
|
|
|
|
filtered_state[k] = v
|
|
|
|
elif self.include_others:
|
|
|
|
filtered_state[k] = v
|
|
|
|
|
|
|
|
return filtered_state
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
def is_full(self) -> bool:
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Whether this filter fetches everything or not
|
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
True if the filter fetches everything.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
|
|
|
return self.include_others and not self.types
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
def has_wildcards(self) -> bool:
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Whether the filter includes wildcards or is attempting to fetch
|
|
|
|
specific state.
|
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
True if the filter includes wildcards.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
|
|
|
|
2019-04-03 03:07:29 -06:00
|
|
|
return self.include_others or any(
|
2020-06-15 05:03:36 -06:00
|
|
|
state_keys is None for state_keys in self.types.values()
|
2018-10-25 10:49:55 -06:00
|
|
|
)
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
def concrete_types(self) -> List[Tuple[str, str]]:
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Returns a list of concrete type/state_keys (i.e. not None) that
|
|
|
|
will be fetched. This will be a complete list if `has_wildcards`
|
|
|
|
returns False, but otherwise will be a subset (or even empty).
|
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
A list of type/state_keys tuples.
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
|
|
|
return [
|
|
|
|
(t, s)
|
2020-06-15 05:03:36 -06:00
|
|
|
for t, state_keys in self.types.items()
|
2018-10-25 10:49:55 -06:00
|
|
|
if state_keys is not None
|
|
|
|
for s in state_keys
|
|
|
|
]
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
def get_member_split(self) -> Tuple["StateFilter", "StateFilter"]:
|
2018-10-25 10:49:55 -06:00
|
|
|
"""Return the filter split into two: one which assumes it's exclusively
|
|
|
|
matching against member state, and one which assumes it's matching
|
|
|
|
against non member state.
|
|
|
|
|
|
|
|
This is useful due to the returned filters giving correct results for
|
|
|
|
`is_full()`, `has_wildcards()`, etc, when operating against maps that
|
|
|
|
either exclusively contain member events or only contain non-member
|
|
|
|
events. (Which is the case when dealing with the member vs non-member
|
|
|
|
state caches).
|
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
The member and non member filters
|
2018-10-25 10:49:55 -06:00
|
|
|
"""
|
|
|
|
|
|
|
|
if EventTypes.Member in self.types:
|
|
|
|
state_keys = self.types[EventTypes.Member]
|
|
|
|
if state_keys is None:
|
|
|
|
member_filter = StateFilter.all()
|
|
|
|
else:
|
2021-09-14 09:35:53 -06:00
|
|
|
member_filter = StateFilter(frozendict({EventTypes.Member: state_keys}))
|
2018-10-25 10:49:55 -06:00
|
|
|
elif self.include_others:
|
|
|
|
member_filter = StateFilter.all()
|
|
|
|
else:
|
|
|
|
member_filter = StateFilter.none()
|
|
|
|
|
|
|
|
non_member_filter = StateFilter(
|
2021-09-14 09:35:53 -06:00
|
|
|
types=frozendict(
|
|
|
|
{k: v for k, v in self.types.items() if k != EventTypes.Member}
|
|
|
|
),
|
2018-10-25 10:49:55 -06:00
|
|
|
include_others=self.include_others,
|
|
|
|
)
|
|
|
|
|
|
|
|
return member_filter, non_member_filter
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2021-10-12 03:44:59 -06:00
|
|
|
def _decompose_into_four_parts(
|
|
|
|
self,
|
|
|
|
) -> Tuple[Tuple[bool, Set[str]], Tuple[Set[str], Set[StateKey]]]:
|
|
|
|
"""
|
|
|
|
Decomposes this state filter into 4 constituent parts, which can be
|
|
|
|
thought of as this:
|
|
|
|
all? - minus_wildcards + plus_wildcards + plus_state_keys
|
|
|
|
|
|
|
|
where
|
|
|
|
* all represents ALL state
|
|
|
|
* minus_wildcards represents entire state types to remove
|
|
|
|
* plus_wildcards represents entire state types to add
|
|
|
|
* plus_state_keys represents individual state keys to add
|
|
|
|
|
|
|
|
See `recompose_from_four_parts` for the other direction of this
|
|
|
|
correspondence.
|
|
|
|
"""
|
|
|
|
is_all = self.include_others
|
|
|
|
excluded_types: Set[str] = {t for t in self.types if is_all}
|
|
|
|
wildcard_types: Set[str] = {t for t, s in self.types.items() if s is None}
|
|
|
|
concrete_keys: Set[StateKey] = set(self.concrete_types())
|
|
|
|
|
|
|
|
return (is_all, excluded_types), (wildcard_types, concrete_keys)
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def _recompose_from_four_parts(
|
|
|
|
all_part: bool,
|
|
|
|
minus_wildcards: Set[str],
|
|
|
|
plus_wildcards: Set[str],
|
|
|
|
plus_state_keys: Set[StateKey],
|
|
|
|
) -> "StateFilter":
|
|
|
|
"""
|
|
|
|
Recomposes a state filter from 4 parts.
|
|
|
|
|
|
|
|
See `decompose_into_four_parts` (the other direction of this
|
|
|
|
correspondence) for descriptions on each of the parts.
|
|
|
|
"""
|
|
|
|
|
|
|
|
# {state type -> set of state keys OR None for wildcard}
|
|
|
|
# (The same structure as that of a StateFilter.)
|
|
|
|
new_types: Dict[str, Optional[Set[str]]] = {}
|
|
|
|
|
|
|
|
# if we start with all, insert the excluded statetypes as empty sets
|
|
|
|
# to prevent them from being included
|
|
|
|
if all_part:
|
|
|
|
new_types.update({state_type: set() for state_type in minus_wildcards})
|
|
|
|
|
|
|
|
# insert the plus wildcards
|
|
|
|
new_types.update({state_type: None for state_type in plus_wildcards})
|
|
|
|
|
|
|
|
# insert the specific state keys
|
|
|
|
for state_type, state_key in plus_state_keys:
|
|
|
|
if state_type in new_types:
|
|
|
|
entry = new_types[state_type]
|
|
|
|
if entry is not None:
|
|
|
|
entry.add(state_key)
|
|
|
|
elif not all_part:
|
|
|
|
# don't insert if the entire type is already included by
|
|
|
|
# include_others as this would actually shrink the state allowed
|
|
|
|
# by this filter.
|
|
|
|
new_types[state_type] = {state_key}
|
|
|
|
|
|
|
|
return StateFilter.freeze(new_types, include_others=all_part)
|
|
|
|
|
|
|
|
def approx_difference(self, other: "StateFilter") -> "StateFilter":
|
|
|
|
"""
|
|
|
|
Returns a state filter which represents `self - other`.
|
|
|
|
|
|
|
|
This is useful for determining what state remains to be pulled out of the
|
|
|
|
database if we want the state included by `self` but already have the state
|
|
|
|
included by `other`.
|
|
|
|
|
|
|
|
The returned state filter
|
|
|
|
- MUST include all state events that are included by this filter (`self`)
|
|
|
|
unless they are included by `other`;
|
|
|
|
- MUST NOT include state events not included by this filter (`self`); and
|
|
|
|
- MAY be an over-approximation: the returned state filter
|
|
|
|
MAY additionally include some state events from `other`.
|
|
|
|
|
|
|
|
This implementation attempts to return the narrowest such state filter.
|
|
|
|
In the case that `self` contains wildcards for state types where
|
|
|
|
`other` contains specific state keys, an approximation must be made:
|
|
|
|
the returned state filter keeps the wildcard, as state filters are not
|
|
|
|
able to express 'all state keys except some given examples'.
|
|
|
|
e.g.
|
|
|
|
StateFilter(m.room.member -> None (wildcard))
|
|
|
|
minus
|
|
|
|
StateFilter(m.room.member -> {'@wombat:example.org'})
|
|
|
|
is approximated as
|
|
|
|
StateFilter(m.room.member -> None (wildcard))
|
|
|
|
"""
|
|
|
|
|
|
|
|
# We first transform self and other into an alternative representation:
|
|
|
|
# - whether or not they include all events to begin with ('all')
|
|
|
|
# - if so, which event types are excluded? ('excludes')
|
|
|
|
# - which entire event types to include ('wildcards')
|
|
|
|
# - which concrete state keys to include ('concrete state keys')
|
|
|
|
(self_all, self_excludes), (
|
|
|
|
self_wildcards,
|
|
|
|
self_concrete_keys,
|
|
|
|
) = self._decompose_into_four_parts()
|
|
|
|
(other_all, other_excludes), (
|
|
|
|
other_wildcards,
|
|
|
|
other_concrete_keys,
|
|
|
|
) = other._decompose_into_four_parts()
|
|
|
|
|
|
|
|
# Start with an estimate of the difference based on self
|
|
|
|
new_all = self_all
|
|
|
|
# Wildcards from the other can be added to the exclusion filter
|
|
|
|
new_excludes = self_excludes | other_wildcards
|
|
|
|
# We remove wildcards that appeared as wildcards in the other
|
|
|
|
new_wildcards = self_wildcards - other_wildcards
|
|
|
|
# We filter out the concrete state keys that appear in the other
|
|
|
|
# as wildcards or concrete state keys.
|
|
|
|
new_concrete_keys = {
|
|
|
|
(state_type, state_key)
|
|
|
|
for (state_type, state_key) in self_concrete_keys
|
|
|
|
if state_type not in other_wildcards
|
|
|
|
} - other_concrete_keys
|
|
|
|
|
|
|
|
if other_all:
|
|
|
|
if self_all:
|
|
|
|
# If self starts with all, then we add as wildcards any
|
|
|
|
# types which appear in the other's exclusion filter (but
|
|
|
|
# aren't in the self exclusion filter). This is as the other
|
|
|
|
# filter will return everything BUT the types in its exclusion, so
|
|
|
|
# we need to add those excluded types that also match the self
|
|
|
|
# filter as wildcard types in the new filter.
|
|
|
|
new_wildcards |= other_excludes.difference(self_excludes)
|
|
|
|
|
|
|
|
# If other is an `include_others` then the difference isn't.
|
|
|
|
new_all = False
|
|
|
|
# (We have no need for excludes when we don't start with all, as there
|
|
|
|
# is nothing to exclude.)
|
|
|
|
new_excludes = set()
|
|
|
|
|
|
|
|
# We also filter out all state types that aren't in the exclusion
|
|
|
|
# list of the other.
|
|
|
|
new_wildcards &= other_excludes
|
|
|
|
new_concrete_keys = {
|
|
|
|
(state_type, state_key)
|
|
|
|
for (state_type, state_key) in new_concrete_keys
|
|
|
|
if state_type in other_excludes
|
|
|
|
}
|
|
|
|
|
|
|
|
# Transform our newly-constructed state filter from the alternative
|
|
|
|
# representation back into the normal StateFilter representation.
|
|
|
|
return StateFilter._recompose_from_four_parts(
|
|
|
|
new_all, new_excludes, new_wildcards, new_concrete_keys
|
|
|
|
)
|
|
|
|
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2022-01-27 03:54:27 -07:00
|
|
|
_ALL_STATE_FILTER = StateFilter(types=frozendict(), include_others=True)
|
2022-02-18 07:54:31 -07:00
|
|
|
_ALL_NON_MEMBER_STATE_FILTER = StateFilter(
|
|
|
|
types=frozendict({EventTypes.Member: frozenset()}), include_others=True
|
|
|
|
)
|
2022-01-27 03:54:27 -07:00
|
|
|
_NONE_STATE_FILTER = StateFilter(types=frozendict(), include_others=False)
|
|
|
|
|
|
|
|
|
2020-09-04 04:54:56 -06:00
|
|
|
class StateGroupStorage:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""High level interface to fetching state for event."""
|
|
|
|
|
2020-12-30 06:09:53 -07:00
|
|
|
def __init__(self, hs: "HomeServer", stores: "Databases"):
|
2019-10-30 08:07:48 -06:00
|
|
|
self.stores = stores
|
|
|
|
|
2020-12-30 06:09:53 -07:00
|
|
|
async def get_state_group_delta(
|
|
|
|
self, state_group: int
|
|
|
|
) -> Tuple[Optional[int], Optional[StateMap[str]]]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Given a state group try to return a previous group and a delta between
|
|
|
|
the old and the new.
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
Args:
|
|
|
|
state_group: The state group used to retrieve state deltas.
|
|
|
|
|
2019-10-30 08:07:48 -06:00
|
|
|
Returns:
|
2020-12-30 06:09:53 -07:00
|
|
|
A tuple of the previous group and a state map of the event IDs which
|
|
|
|
make up the delta between the old and new state groups.
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
|
|
|
|
2021-09-15 07:54:13 -06:00
|
|
|
state_group_delta = await self.stores.state.get_state_group_delta(state_group)
|
|
|
|
return state_group_delta.prev_group, state_group_delta.delta_ids
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
async def get_state_groups_ids(
|
2022-03-04 03:25:18 -07:00
|
|
|
self, _room_id: str, event_ids: Collection[str]
|
2020-09-29 08:57:36 -06:00
|
|
|
) -> Dict[int, MutableStateMap[str]]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Get the event IDs of all the state for the state groups for the given events
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
_room_id: id of the room for these events
|
|
|
|
event_ids: ids of the events
|
2019-10-30 08:07:48 -06:00
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
dict of state_group_id -> (dict of (type, state_key) -> event id)
|
2022-04-01 06:01:49 -06:00
|
|
|
|
|
|
|
Raises:
|
|
|
|
RuntimeError if we don't have a state group for one or more of the events
|
|
|
|
(ie they are outliers or unknown)
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
|
|
|
if not event_ids:
|
|
|
|
return {}
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
event_to_groups = await self.stores.main._get_state_group_for_events(event_ids)
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2020-06-15 05:03:36 -06:00
|
|
|
groups = set(event_to_groups.values())
|
2020-07-28 14:09:53 -06:00
|
|
|
group_to_state = await self.stores.state._get_state_for_groups(groups)
|
2019-10-30 08:07:48 -06:00
|
|
|
|
|
|
|
return group_to_state
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
async def get_state_ids_for_group(self, state_group: int) -> StateMap[str]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Get the event IDs of all the state in the given state group
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
state_group: A state group for which we want to get the state IDs.
|
2019-10-30 08:07:48 -06:00
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
Resolves to a map of (type, state_key) -> event_id
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
2020-07-28 14:09:53 -06:00
|
|
|
group_to_state = await self._get_state_for_groups((state_group,))
|
2019-10-30 08:07:48 -06:00
|
|
|
|
|
|
|
return group_to_state[state_group]
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
async def get_state_groups(
|
2022-03-04 03:25:18 -07:00
|
|
|
self, room_id: str, event_ids: Collection[str]
|
2020-07-28 14:09:53 -06:00
|
|
|
) -> Dict[int, List[EventBase]]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Get the state groups for the given list of event_ids
|
2020-07-28 14:09:53 -06:00
|
|
|
|
|
|
|
Args:
|
|
|
|
room_id: ID of the room for these events.
|
|
|
|
event_ids: The event IDs to retrieve state for.
|
|
|
|
|
2019-10-30 08:07:48 -06:00
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
dict of state_group_id -> list of state events.
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
|
|
|
if not event_ids:
|
|
|
|
return {}
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
group_to_ids = await self.get_state_groups_ids(room_id, event_ids)
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
state_event_map = await self.stores.main.get_events(
|
2019-10-30 08:07:48 -06:00
|
|
|
[
|
|
|
|
ev_id
|
2020-06-15 05:03:36 -06:00
|
|
|
for group_ids in group_to_ids.values()
|
|
|
|
for ev_id in group_ids.values()
|
2019-10-30 08:07:48 -06:00
|
|
|
],
|
|
|
|
get_prev_content=False,
|
|
|
|
)
|
|
|
|
|
|
|
|
return {
|
|
|
|
group: [
|
|
|
|
state_event_map[v]
|
2020-06-15 05:03:36 -06:00
|
|
|
for v in event_id_map.values()
|
2019-10-30 08:07:48 -06:00
|
|
|
if v in state_event_map
|
|
|
|
]
|
2020-06-15 05:03:36 -06:00
|
|
|
for group, event_id_map in group_to_ids.items()
|
2019-10-30 08:07:48 -06:00
|
|
|
}
|
|
|
|
|
2020-01-16 06:31:22 -07:00
|
|
|
def _get_state_groups_from_groups(
|
|
|
|
self, groups: List[int], state_filter: StateFilter
|
2020-07-30 05:20:41 -06:00
|
|
|
) -> Awaitable[Dict[int, StateMap[str]]]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Returns the state groups for a given set of groups, filtering on
|
|
|
|
types of state events.
|
|
|
|
|
|
|
|
Args:
|
2020-01-16 06:31:22 -07:00
|
|
|
groups: list of state group IDs to query
|
|
|
|
state_filter: The state filter used to fetch state
|
2019-10-30 08:07:48 -06:00
|
|
|
from the database.
|
2020-07-28 14:09:53 -06:00
|
|
|
|
2019-10-30 08:07:48 -06:00
|
|
|
Returns:
|
2020-07-30 05:20:41 -06:00
|
|
|
Dict of state group to state map.
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
|
|
|
|
2019-12-20 03:48:24 -07:00
|
|
|
return self.stores.state._get_state_groups_from_groups(groups, state_filter)
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
async def get_state_for_events(
|
2022-03-04 03:25:18 -07:00
|
|
|
self, event_ids: Collection[str], state_filter: Optional[StateFilter] = None
|
2020-12-30 06:09:53 -07:00
|
|
|
) -> Dict[str, StateMap[EventBase]]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Given a list of event_ids and type tuples, return a list of state
|
|
|
|
dicts for each event.
|
2020-07-28 14:09:53 -06:00
|
|
|
|
2019-10-30 08:07:48 -06:00
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
event_ids: The events to fetch the state of.
|
|
|
|
state_filter: The state filter used to fetch state.
|
|
|
|
|
2019-10-30 08:07:48 -06:00
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
A dict of (event_id) -> (type, state_key) -> [state_events]
|
2022-04-01 06:01:49 -06:00
|
|
|
|
|
|
|
Raises:
|
|
|
|
RuntimeError if we don't have a state group for one or more of the events
|
|
|
|
(ie they are outliers or unknown)
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
2020-07-28 14:09:53 -06:00
|
|
|
event_to_groups = await self.stores.main._get_state_group_for_events(event_ids)
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2020-06-15 05:03:36 -06:00
|
|
|
groups = set(event_to_groups.values())
|
2020-07-28 14:09:53 -06:00
|
|
|
group_to_state = await self.stores.state._get_state_for_groups(
|
2021-04-08 15:38:54 -06:00
|
|
|
groups, state_filter or StateFilter.all()
|
2019-10-30 08:07:48 -06:00
|
|
|
)
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
state_event_map = await self.stores.main.get_events(
|
2020-06-15 05:03:36 -06:00
|
|
|
[ev_id for sd in group_to_state.values() for ev_id in sd.values()],
|
2019-10-30 08:07:48 -06:00
|
|
|
get_prev_content=False,
|
|
|
|
)
|
|
|
|
|
|
|
|
event_to_state = {
|
|
|
|
event_id: {
|
|
|
|
k: state_event_map[v]
|
2020-06-15 05:03:36 -06:00
|
|
|
for k, v in group_to_state[group].items()
|
2019-10-30 08:07:48 -06:00
|
|
|
if v in state_event_map
|
|
|
|
}
|
2020-06-15 05:03:36 -06:00
|
|
|
for event_id, group in event_to_groups.items()
|
2019-10-30 08:07:48 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
return {event: event_to_state[event] for event in event_ids}
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
async def get_state_ids_for_events(
|
2022-03-04 03:25:18 -07:00
|
|
|
self, event_ids: Collection[str], state_filter: Optional[StateFilter] = None
|
2020-12-30 06:09:53 -07:00
|
|
|
) -> Dict[str, StateMap[str]]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
|
|
|
Get the state dicts corresponding to a list of events, containing the event_ids
|
|
|
|
of the state events (as opposed to the events themselves)
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
event_ids: events whose state should be returned
|
|
|
|
state_filter: The state filter used to fetch state from the database.
|
2019-10-30 08:07:48 -06:00
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
A dict from event_id -> (type, state_key) -> event_id
|
2022-04-01 06:01:49 -06:00
|
|
|
|
|
|
|
Raises:
|
|
|
|
RuntimeError if we don't have a state group for one or more of the events
|
|
|
|
(ie they are outliers or unknown)
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
2020-07-28 14:09:53 -06:00
|
|
|
event_to_groups = await self.stores.main._get_state_group_for_events(event_ids)
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2020-06-15 05:03:36 -06:00
|
|
|
groups = set(event_to_groups.values())
|
2020-07-28 14:09:53 -06:00
|
|
|
group_to_state = await self.stores.state._get_state_for_groups(
|
2021-04-08 15:38:54 -06:00
|
|
|
groups, state_filter or StateFilter.all()
|
2019-10-30 08:07:48 -06:00
|
|
|
)
|
|
|
|
|
|
|
|
event_to_state = {
|
|
|
|
event_id: group_to_state[group]
|
2020-06-15 05:03:36 -06:00
|
|
|
for event_id, group in event_to_groups.items()
|
2019-10-30 08:07:48 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
return {event: event_to_state[event] for event in event_ids}
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
async def get_state_for_event(
|
2021-04-08 15:38:54 -06:00
|
|
|
self, event_id: str, state_filter: Optional[StateFilter] = None
|
2020-12-30 06:09:53 -07:00
|
|
|
) -> StateMap[EventBase]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
|
|
|
Get the state dict corresponding to a particular event
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
event_id: event whose state should be returned
|
|
|
|
state_filter: The state filter used to fetch state from the database.
|
2019-10-30 08:07:48 -06:00
|
|
|
|
|
|
|
Returns:
|
2020-07-28 14:09:53 -06:00
|
|
|
A dict from (type, state_key) -> state_event
|
2022-04-01 06:01:49 -06:00
|
|
|
|
|
|
|
Raises:
|
|
|
|
RuntimeError if we don't have a state group for the event (ie it is an
|
|
|
|
outlier or is unknown)
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
2021-04-08 15:38:54 -06:00
|
|
|
state_map = await self.get_state_for_events(
|
|
|
|
[event_id], state_filter or StateFilter.all()
|
|
|
|
)
|
2019-10-30 08:07:48 -06:00
|
|
|
return state_map[event_id]
|
|
|
|
|
2020-07-28 14:09:53 -06:00
|
|
|
async def get_state_ids_for_event(
|
2021-04-08 15:38:54 -06:00
|
|
|
self, event_id: str, state_filter: Optional[StateFilter] = None
|
2020-12-30 06:09:53 -07:00
|
|
|
) -> StateMap[str]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
|
|
|
Get the state dict corresponding to a particular event
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
event_id: event whose state should be returned
|
|
|
|
state_filter: The state filter used to fetch state from the database.
|
2019-10-30 08:07:48 -06:00
|
|
|
|
|
|
|
Returns:
|
2021-05-24 03:43:33 -06:00
|
|
|
A dict from (type, state_key) -> state_event_id
|
2022-04-01 06:01:49 -06:00
|
|
|
|
|
|
|
Raises:
|
|
|
|
RuntimeError if we don't have a state group for the event (ie it is an
|
|
|
|
outlier or is unknown)
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
2021-04-08 15:38:54 -06:00
|
|
|
state_map = await self.get_state_ids_for_events(
|
|
|
|
[event_id], state_filter or StateFilter.all()
|
|
|
|
)
|
2019-10-30 08:07:48 -06:00
|
|
|
return state_map[event_id]
|
|
|
|
|
2020-01-16 06:31:22 -07:00
|
|
|
def _get_state_for_groups(
|
2021-04-08 15:38:54 -06:00
|
|
|
self, groups: Iterable[int], state_filter: Optional[StateFilter] = None
|
2020-09-29 08:57:36 -06:00
|
|
|
) -> Awaitable[Dict[int, MutableStateMap[str]]]:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Gets the state at each of a list of state groups, optionally
|
|
|
|
filtering by type/state_key
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
groups: list of state groups for which we want to get the state.
|
|
|
|
state_filter: The state filter used to fetch state.
|
2019-10-30 08:07:48 -06:00
|
|
|
from the database.
|
2020-07-30 05:20:41 -06:00
|
|
|
|
2019-10-30 08:07:48 -06:00
|
|
|
Returns:
|
2020-07-30 05:20:41 -06:00
|
|
|
Dict of state group to state map.
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
2021-04-08 15:38:54 -06:00
|
|
|
return self.stores.state._get_state_for_groups(
|
|
|
|
groups, state_filter or StateFilter.all()
|
|
|
|
)
|
2019-10-30 08:07:48 -06:00
|
|
|
|
2020-08-28 07:37:55 -06:00
|
|
|
async def store_state_group(
|
2020-07-28 14:09:53 -06:00
|
|
|
self,
|
|
|
|
event_id: str,
|
|
|
|
room_id: str,
|
|
|
|
prev_group: Optional[int],
|
2021-07-26 10:49:53 -06:00
|
|
|
delta_ids: Optional[StateMap[str]],
|
|
|
|
current_state_ids: StateMap[str],
|
2020-08-28 07:37:55 -06:00
|
|
|
) -> int:
|
2019-10-30 08:07:48 -06:00
|
|
|
"""Store a new set of state, returning a newly assigned state group.
|
|
|
|
|
|
|
|
Args:
|
2020-07-28 14:09:53 -06:00
|
|
|
event_id: The event ID for which the state was calculated.
|
|
|
|
room_id: ID of the room for which the state was calculated.
|
|
|
|
prev_group: A previous state group for the room, optional.
|
|
|
|
delta_ids: The delta between state at `prev_group` and
|
2019-10-30 08:07:48 -06:00
|
|
|
`current_state_ids`, if `prev_group` was given. Same format as
|
|
|
|
`current_state_ids`.
|
2020-07-28 14:09:53 -06:00
|
|
|
current_state_ids: The state to store. Map of (type, state_key)
|
2019-10-30 08:07:48 -06:00
|
|
|
to event_id.
|
|
|
|
|
|
|
|
Returns:
|
2020-08-28 07:37:55 -06:00
|
|
|
The state group ID
|
2019-10-30 08:07:48 -06:00
|
|
|
"""
|
2020-08-28 07:37:55 -06:00
|
|
|
return await self.stores.state.store_state_group(
|
2019-10-30 08:07:48 -06:00
|
|
|
event_id, room_id, prev_group, delta_ids, current_state_ids
|
|
|
|
)
|