summary refs log tree commit diff
path: root/synapse/api/streams/event.py
diff options
context:
space:
mode:
authorErik Johnston <erik@matrix.org>2014-08-26 18:57:46 +0100
committerErik Johnston <erik@matrix.org>2014-08-26 18:57:46 +0100
commit3a2a5b959cb1f56b26af32e1ad4c1db424279eb7 (patch)
tree5f5b515cdaa586f1f7f92337471dfdc2e0d4683d /synapse/api/streams/event.py
parentMerge branch 'develop' of github.com:matrix-org/synapse into stream_refactor (diff)
downloadsynapse-3a2a5b959cb1f56b26af32e1ad4c1db424279eb7.tar.xz
WIP: Completely change how event streaming and pagination work. This reflects the change in the underlying storage model.
Diffstat (limited to 'synapse/api/streams/event.py')
-rw-r--r--synapse/api/streams/event.py194
1 files changed, 0 insertions, 194 deletions
diff --git a/synapse/api/streams/event.py b/synapse/api/streams/event.py
deleted file mode 100644
index fe44a488bc..0000000000
--- a/synapse/api/streams/event.py
+++ /dev/null
@@ -1,194 +0,0 @@
-# -*- coding: utf-8 -*-
-# Copyright 2014 matrix.org
-#
-# 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.
-
-"""This module contains classes for streaming from the event stream: /events.
-"""
-from twisted.internet import defer
-
-from synapse.api.errors import EventStreamError
-from synapse.api.events import SynapseEvent
-from synapse.api.streams import PaginationStream, StreamData
-
-import logging
-
-logger = logging.getLogger(__name__)
-
-
-class EventsStreamData(StreamData):
-    EVENT_TYPE = "EventsStream"
-
-    def __init__(self, hs, room_id=None, feedback=False):
-        super(EventsStreamData, self).__init__(hs)
-        self.room_id = room_id
-        self.with_feedback = feedback
-
-    @defer.inlineCallbacks
-    def get_rows(self, user_id, from_key, to_key, limit, direction):
-        data, latest_ver = yield self.store.get_room_events(
-            user_id=user_id,
-            from_key=from_key,
-            to_key=to_key,
-            limit=limit,
-            room_id=self.room_id,
-            with_feedback=self.with_feedback
-        )
-        defer.returnValue((data, latest_ver))
-
-    @defer.inlineCallbacks
-    def max_token(self):
-        val = yield self.store.get_room_events_max_id()
-        defer.returnValue(val)
-
-
-class EventStream(PaginationStream):
-
-    SEPARATOR = '_'
-
-    def __init__(self, user_id, stream_data_list):
-        super(EventStream, self).__init__()
-        self.user_id = user_id
-        self.stream_data = stream_data_list
-
-    @defer.inlineCallbacks
-    def fix_tokens(self, pagination_config):
-        pagination_config.from_tok = yield self.fix_token(
-            pagination_config.from_tok)
-        pagination_config.to_tok = yield self.fix_token(
-            pagination_config.to_tok)
-
-        if (
-            not pagination_config.to_tok
-            and pagination_config.direction == 'f'
-        ):
-            pagination_config.to_tok = yield self.get_current_max_token()
-
-        logger.debug("pagination_config: %s", pagination_config)
-
-        defer.returnValue(pagination_config)
-
-    @defer.inlineCallbacks
-    def fix_token(self, token):
-        """Fixes unknown values in a token to known values.
-
-        Args:
-            token (str): The token to fix up.
-        Returns:
-            The fixed-up token, which may == token.
-        """
-        if token == PaginationStream.TOK_END:
-            new_token = yield self.get_current_max_token()
-
-            logger.debug("fix_token: From %s to %s", token, new_token)
-
-            token = new_token
-
-        defer.returnValue(token)
-
-    @defer.inlineCallbacks
-    def get_current_max_token(self):
-        new_token_parts = []
-        for s in self.stream_data:
-            mx = yield s.max_token()
-            new_token_parts.append(str(mx))
-
-        new_token = EventStream.SEPARATOR.join(new_token_parts)
-
-        logger.debug("get_current_max_token: %s", new_token)
-
-        defer.returnValue(new_token)
-
-    @defer.inlineCallbacks
-    def get_chunk(self, config):
-        # no support for limit on >1 streams, makes no sense.
-        if config.limit and len(self.stream_data) > 1:
-            raise EventStreamError(
-                400, "Limit not supported on multiplexed streams."
-            )
-
-        chunk_data, next_tok = yield self._get_chunk_data(
-            config.from_tok,
-            config.to_tok,
-            config.limit,
-            config.direction,
-        )
-
-        defer.returnValue({
-            "chunk": chunk_data,
-            "start": config.from_tok,
-            "end": next_tok
-        })
-
-    @defer.inlineCallbacks
-    def _get_chunk_data(self, from_tok, to_tok, limit, direction):
-        """ Get event data between the two tokens.
-
-        Tokens are SEPARATOR separated values representing pkey values of
-        certain tables, and the position determines the StreamData invoked
-        according to the STREAM_DATA list.
-
-        The magic value '-1' can be used to get the latest value.
-
-        Args:
-            from_tok - The token to start from.
-            to_tok - The token to end at. Must have values > from_tok or be -1.
-        Returns:
-            A list of event data.
-        Raises:
-            EventStreamError if something went wrong.
-        """
-        # sanity check
-        if to_tok is not None:
-            if (from_tok.count(EventStream.SEPARATOR) !=
-                    to_tok.count(EventStream.SEPARATOR) or
-                    (from_tok.count(EventStream.SEPARATOR) + 1) !=
-                    len(self.stream_data)):
-                raise EventStreamError(400, "Token lengths don't match.")
-
-        chunk = []
-        next_ver = []
-        for i, (from_pkey, to_pkey) in enumerate(zip(
-            self._split_token(from_tok),
-            self._split_token(to_tok)
-        )):
-            if from_pkey == to_pkey:
-                # tokens are the same, we have nothing to do.
-                next_ver.append(str(to_pkey))
-                continue
-
-            (event_chunk, max_pkey) = yield self.stream_data[i].get_rows(
-                self.user_id, from_pkey, to_pkey, limit, direction,
-            )
-
-            chunk.extend([
-                e.get_dict() if isinstance(e, SynapseEvent) else e
-                for e in event_chunk
-            ])
-            next_ver.append(str(max_pkey))
-
-        defer.returnValue((chunk, EventStream.SEPARATOR.join(next_ver)))
-
-    def _split_token(self, token):
-        """Splits the given token into a list of pkeys.
-
-        Args:
-            token (str): The token with SEPARATOR values.
-        Returns:
-            A list of ints.
-        """
-        if token:
-            segments = token.split(EventStream.SEPARATOR)
-        else:
-            segments = [None] * len(self.stream_data)
-        return segments