diff options
author | Erik Johnston <erik@matrix.org> | 2016-01-19 11:27:27 +0000 |
---|---|---|
committer | Erik Johnston <erik@matrix.org> | 2016-01-19 11:27:27 +0000 |
commit | 47e7963e50704a1a981b8c009f8403085be721a6 (patch) | |
tree | 1a8b4930d67c314c21d40e8e48436a0b246476f1 /synapse/push/push_rule_evaluator.py | |
parent | Merge pull request #503 from matrix-org/daniel/nonghosts (diff) | |
parent | Take a deepcopy of push rules before mutating them (diff) | |
download | synapse-47e7963e50704a1a981b8c009f8403085be721a6.tar.xz |
Merge pull request #502 from matrix-org/erikj/push_notif_perf
Unread notification performance.
Diffstat (limited to 'synapse/push/push_rule_evaluator.py')
-rw-r--r-- | synapse/push/push_rule_evaluator.py | 260 |
1 files changed, 166 insertions, 94 deletions
diff --git a/synapse/push/push_rule_evaluator.py b/synapse/push/push_rule_evaluator.py index b0283743a2..379652c513 100644 --- a/synapse/push/push_rule_evaluator.py +++ b/synapse/push/push_rule_evaluator.py @@ -15,17 +15,22 @@ from twisted.internet import defer -from synapse.types import UserID - import baserules import logging import simplejson as json import re +from synapse.types import UserID + logger = logging.getLogger(__name__) +GLOB_REGEX = re.compile(r'\\\[(\\\!|)(.*)\\\]') +IS_GLOB = re.compile(r'[\?\*\[\]]') +INEQUALITY_EXPR = re.compile("^([=<>]*)([0-9]*)$") + + @defer.inlineCallbacks def evaluator_for_user_id_and_profile_tag(user_id, profile_tag, room_id, store): rawrules = yield store.get_push_rules_for_user(user_id) @@ -42,9 +47,34 @@ def evaluator_for_user_id_and_profile_tag(user_id, profile_tag, room_id, store): )) +def _room_member_count(ev, condition, room_member_count): + if 'is' not in condition: + return False + m = INEQUALITY_EXPR.match(condition['is']) + if not m: + return False + ineq = m.group(1) + rhs = m.group(2) + if not rhs.isdigit(): + return False + rhs = int(rhs) + + if ineq == '' or ineq == '==': + return room_member_count == rhs + elif ineq == '<': + return room_member_count < rhs + elif ineq == '>': + return room_member_count > rhs + elif ineq == '>=': + return room_member_count >= rhs + elif ineq == '<=': + return room_member_count <= rhs + else: + return False + + class PushRuleEvaluator: DEFAULT_ACTIONS = [] - INEQUALITY_EXPR = re.compile("^([=<>]*)([0-9]*)$") def __init__(self, user_id, profile_tag, raw_rules, enabled_map, room_id, our_member_event, store): @@ -61,8 +91,7 @@ class PushRuleEvaluator: rule['actions'] = json.loads(raw_rule['actions']) rules.append(rule) - user = UserID.from_string(self.user_id) - self.rules = baserules.list_with_base_rules(rules, user) + self.rules = baserules.list_with_base_rules(rules) self.enabled_map = enabled_map @@ -98,28 +127,19 @@ class PushRuleEvaluator: room_members = yield self.store.get_users_in_room(room_id) room_member_count = len(room_members) + evaluator = PushRuleEvaluatorForEvent(ev, room_member_count) + for r in self.rules: - if r['rule_id'] in self.enabled_map: - r['enabled'] = self.enabled_map[r['rule_id']] - elif 'enabled' not in r: - r['enabled'] = True - if not r['enabled']: + enabled = self.enabled_map.get(r['rule_id'], None) + if enabled is not None and not enabled: + continue + + if not r.get("enabled", True): continue - matches = True conditions = r['conditions'] actions = r['actions'] - for c in conditions: - matches &= self._event_fulfills_condition( - ev, c, display_name=my_display_name, - room_member_count=room_member_count, - profile_tag=self.profile_tag - ) - logger.debug( - "Rule %s %s", - r['rule_id'], "matches" if matches else "doesn't match" - ) # ignore rules with no actions (we have an explict 'dont_notify') if len(actions) == 0: logger.warn( @@ -127,8 +147,22 @@ class PushRuleEvaluator: r['rule_id'], self.user_id ) continue + + matches = True + for c in conditions: + matches = evaluator.matches( + c, self.user_id, my_display_name, self.profile_tag + ) + if not matches: + break + + logger.debug( + "Rule %s %s", + r['rule_id'], "matches" if matches else "doesn't match" + ) + if matches: - logger.info( + logger.debug( "%s matches for user %s, event %s", r['rule_id'], self.user_id, ev['event_id'] ) @@ -139,94 +173,132 @@ class PushRuleEvaluator: defer.returnValue(actions) - logger.info( + logger.debug( "No rules match for user %s, event %s", self.user_id, ev['event_id'] ) defer.returnValue(PushRuleEvaluator.DEFAULT_ACTIONS) - @staticmethod - def _glob_to_regexp(glob): - r = re.escape(glob) - r = re.sub(r'\\\*', r'.*?', r) - r = re.sub(r'\\\?', r'.', r) - # handle [abc], [a-z] and [!a-z] style ranges. - r = re.sub(r'\\\[(\\\!|)(.*)\\\]', - lambda x: ('[%s%s]' % (x.group(1) and '^' or '', - re.sub(r'\\\-', '-', x.group(2)))), r) - return r +class PushRuleEvaluatorForEvent(object): + def __init__(self, event, room_member_count): + self._event = event + self._room_member_count = room_member_count - @staticmethod - def _event_fulfills_condition(ev, condition, - display_name, room_member_count, profile_tag): - if condition['kind'] == 'event_match': - if 'pattern' not in condition: - logger.warn("event_match condition with no pattern") - return False - # XXX: optimisation: cache our pattern regexps - if condition['key'] == 'content.body': - r = r'\b%s\b' % PushRuleEvaluator._glob_to_regexp(condition['pattern']) - else: - r = r'^%s$' % PushRuleEvaluator._glob_to_regexp(condition['pattern']) - val = _value_for_dotted_key(condition['key'], ev) - if val is None: - return False - return re.search(r, val, flags=re.IGNORECASE) is not None + # Maps strings of e.g. 'content.body' -> event["content"]["body"] + self._value_cache = _flatten_dict(event) + def matches(self, condition, user_id, display_name, profile_tag): + if condition['kind'] == 'event_match': + return self._event_match(condition, user_id) elif condition['kind'] == 'device': if 'profile_tag' not in condition: return True return condition['profile_tag'] == profile_tag - elif condition['kind'] == 'contains_display_name': - # This is special because display names can be different - # between rooms and so you can't really hard code it in a rule. - # Optimisation: we should cache these names and update them from - # the event stream. - if 'content' not in ev or 'body' not in ev['content']: - return False - if not display_name: - return False - return re.search( - r"\b%s\b" % re.escape(display_name), ev['content']['body'], - flags=re.IGNORECASE - ) is not None - + return self._contains_display_name(display_name) elif condition['kind'] == 'room_member_count': - if 'is' not in condition: - return False - m = PushRuleEvaluator.INEQUALITY_EXPR.match(condition['is']) - if not m: - return False - ineq = m.group(1) - rhs = m.group(2) - if not rhs.isdigit(): + return _room_member_count( + self._event, condition, self._room_member_count + ) + else: + return True + + def _event_match(self, condition, user_id): + pattern = condition.get('pattern', None) + + if not pattern: + pattern_type = condition.get('pattern_type', None) + if pattern_type == "user_id": + pattern = user_id + elif pattern_type == "user_localpart": + pattern = UserID.from_string(user_id).localpart + + if not pattern: + logger.warn("event_match condition with no pattern") + return False + + # XXX: optimisation: cache our pattern regexps + if condition['key'] == 'content.body': + body = self._event["content"].get("body", None) + if not body: return False - rhs = int(rhs) - - if ineq == '' or ineq == '==': - return room_member_count == rhs - elif ineq == '<': - return room_member_count < rhs - elif ineq == '>': - return room_member_count > rhs - elif ineq == '>=': - return room_member_count >= rhs - elif ineq == '<=': - return room_member_count <= rhs - else: + + return _glob_matches(pattern, body, word_boundary=True) + else: + haystack = self._get_value(condition['key']) + if haystack is None: return False + + return _glob_matches(pattern, haystack) + + def _contains_display_name(self, display_name): + if not display_name: + return False + + body = self._event["content"].get("body", None) + if not body: + return False + + return _glob_matches(display_name, body, word_boundary=True) + + def _get_value(self, dotted_key): + return self._value_cache.get(dotted_key, None) + + +def _glob_matches(glob, value, word_boundary=False): + """Tests if value matches glob. + + Args: + glob (string) + value (string): String to test against glob. + word_boundary (bool): Whether to match against word boundaries or entire + string. Defaults to False. + + Returns: + bool + """ + if IS_GLOB.search(glob): + r = re.escape(glob) + + r = r.replace(r'\*', '.*?') + r = r.replace(r'\?', '.') + + # handle [abc], [a-z] and [!a-z] style ranges. + r = GLOB_REGEX.sub( + lambda x: ( + '[%s%s]' % ( + x.group(1) and '^' or '', + x.group(2).replace(r'\\\-', '-') + ) + ), + r, + ) + if word_boundary: + r = r"\b%s\b" % (r,) + r = re.compile(r, flags=re.IGNORECASE) + + return r.search(value) else: - return True + r = r + "$" + r = re.compile(r, flags=re.IGNORECASE) + + return r.match(value) + elif word_boundary: + r = re.escape(glob) + r = r"\b%s\b" % (r,) + r = re.compile(r, flags=re.IGNORECASE) + + return r.search(value) + else: + return value.lower() == glob.lower() + +def _flatten_dict(d, prefix=[], result={}): + for key, value in d.items(): + if isinstance(value, basestring): + result[".".join(prefix + [key])] = value.lower() + elif hasattr(value, "items"): + _flatten_dict(value, prefix=(prefix+[key]), result=result) -def _value_for_dotted_key(dotted_key, event): - parts = dotted_key.split(".") - val = event - while len(parts) > 0: - if parts[0] not in val: - return None - val = val[parts[0]] - parts = parts[1:] - return val + return result |