zeitgeist team mailing list archive

[Seif Lotfy <seif@xxxxxxxxx>]

Seif Lotfy has proposed merging lp:~seif/zeitgeist/add_cache_for_get_events into lp:zeitgeist.

Requested reviews:
  Zeitgeist Framework Team (zeitgeist)


I added a cache to reduce the calls on get_events
basically i look at separate the cached ids from the ids handed to get_events
then i call the sql db to give me all uncached events then i sort them all and voila
the speed improvement is very nice

old DEBUG:zeitgeist.engine:Got 34 events in 0.001895s
new DEBUG:zeitgeist.engine:Got 34 events in 0.000248s

-- 
https://code.launchpad.net/~seif/zeitgeist/add_cache_for_get_events/+merge/42327
Your team Zeitgeist Framework Team is requested to review the proposed merge of lp:~seif/zeitgeist/add_cache_for_get_events into lp:zeitgeist.

=== modified file '_zeitgeist/engine/main.py'
--- _zeitgeist/engine/main.py	2010-11-29 09:09:13 +0000
+++ _zeitgeist/engine/main.py	2010-12-01 00:31:58 +0000
@@ -118,6 +118,7 @@
 		self._manifestation = TableLookup(cursor, "manifestation")
 		self._mimetype = TableLookup(cursor, "mimetype")
 		self._actor = TableLookup(cursor, "actor")
+		self._event_cache = {}
 	
 	@property
 	def extensions(self):
@@ -166,10 +167,14 @@
 		if not ids:
 			return []
 		
-		rows = self._cursor.execute("""
-			SELECT * FROM event_view
-			WHERE id IN (%s)
-			""" % ",".join("%d" % id for id in ids)).fetchall()
+		# Split ids into cached and uncached
+		uncached_ids = []
+		cached_ids = []
+		for id in ids:
+			if self._event_cache.has_key(id):
+				cached_ids.append(id)
+			else:
+				uncached_ids.append(id)
 		
 		id_hash = defaultdict(list)
 		for n, id in enumerate(ids):
@@ -183,6 +188,22 @@
 		# deleted
 		events = {}
 		sorted_events = [None]*len(ids)
+		
+		for id in cached_ids:
+			event = self._event_cache[id]
+			if event:
+				event = self.extensions.apply_get_hooks(event, sender)
+				if event is not None:
+					for n in id_hash[event.id]:
+						# insert the event into all necessary spots (LP: #673916)
+						sorted_events[n] = event
+		
+		# Get uncached events
+		rows = self._cursor.execute("""
+			SELECT * FROM event_view
+			WHERE id IN (%s)
+			""" % ",".join("%d" % id for id in uncached_ids)).fetchall()
+		
 		for row in rows:
 			# Assumption: all rows of a same event for its different
 			# subjects are in consecutive order.
@@ -194,12 +215,14 @@
 					events[event.id] = event
 				else:
 					event = events[event.id]
+				self._event_cache[event.id] = event
 				event.append_subject(self._get_subject_from_row(row))
 				event = self.extensions.apply_get_hooks(event, sender)
 				if event is not None:
 					for n in id_hash[event.id]:
 						# insert the event into all necessary spots (LP: #673916)
 						sorted_events[n] = event
+			
 				
 		log.debug("Got %d events in %fs" % (len(sorted_events), time.time()-t))
 		return sorted_events
@@ -603,6 +626,11 @@
 		""" % ",".join(str(int(_id)) for _id in ids))
 		timestamps = self._cursor.fetchone()
 
+		# Remove events from cache
+		for id in ids:
+			if self._event_cache.has_key(id):
+				del self._event_cache[id]
+		
 		# Make sure that we actually found some events with these ids...
 		# We can't do all(timestamps) here because the timestamps may be 0
 		if timestamps and timestamps[0] is not None and timestamps[1] is not None:
@@ -612,7 +640,6 @@
 			log.debug("Deleted %s" % map(int, ids))
 			
 			self.extensions.apply_post_delete(ids, sender)
-			
 			return timestamps
 		else:
 			log.debug("Tried to delete non-existing event(s): %s" % map(int, ids))