Merge lp:~seif/zeitgeist/fix-738555 into lp:zeitgeist/0.1

[Seif Lotfy]

Add a try except block in the _write_to_disk method
Add a counter to write to disk upon every 20 events
Fixed some indents
----
OK so a primary solution for this was putting the content of _write_to_disk
in datasource_registery.py in a normal " try - catch " block
My primary fear was that the next extension or zeitgeist would crash trying
to write to the disk. This was not the case and the error was caught.
To explain why the other stuff did not crash is:
1) This was the only extension writing to disk upon unloading
2) The engine does not write to disk upon exit since all events are written
to disk upon entrance.
My primary solution would be to put the _write_to_disk in a normal "try -
catch" block for now...

[Seif Lotfy (seif)]

Add a try except block in the _write_to_disk method
Add a counter to write to disk upon every 20 events
Fixed some indents

[lp:~seif/zeitgeist/fix-738555]

1716. By Siegfried Gevatter on 2011-04-22

Enable query filtering by StorageState.

[Siegfried Gevatter (rainct)]

> Add a try except block in the _write_to_disk method
> Fixed some indents
Looks good. Feel free to commit this part already.

> Add a counter to write to disk upon every 20 events
Why every 20 events? I don't like this, too arbitrary (and it ends writing to disk a lot again).

Also, shouldn't really matter much, but I don't like the modulo approach, since it's doing a useless division and the variable can overflow. Something like this would look nicer to me:
if self._counter == 0:
    self._write_to_disk()
    self._counter = 19
else:
    self._counter -= 1

[lp:~seif/zeitgeist/fix-738555]

1717. By Siegfried Gevatter on 2011-04-22

Fix typo in singleton.py comment.

1718. By Siegfried Gevatter on 2011-04-22

Update NEWS.

1719. By Siegfried Gevatter on 2011-04-22

test/engine-test.py:
 - Fix typo in method names (MimetType -> MimeType).
 - Rename *Origin methods to *SubjectOrigin.

1720. By Seif Lotfy on 2011-04-22

replace the counter in datasource_registry with a timeout to write to disk

1721. By Seif Lotfy on 2011-04-22

added a dirty flag to mimize writing to disk

1722. By Seif Lotfy on 2011-04-22

merge with trunk

[Siegfried Gevatter (rainct)]

Haven't tested it, but looks good :).

review: Approve

=== modified file '_zeitgeist/engine/extensions/datasource_registry.py'
--- _zeitgeist/engine/extensions/datasource_registry.py	2011-01-17 15:54:47 +0000
+++ _zeitgeist/engine/extensions/datasource_registry.py	2011-04-22 21:58:00 +0000
@@ -23,6 +23,7 @@
 import cPickle as pickle
 import dbus
 import dbus.service
+import gobject
 import logging
 
 from zeitgeist.datamodel import get_timestamp_for_now
@@ -36,6 +37,7 @@
 REGISTRY_DBUS_OBJECT_PATH = "/org/gnome/zeitgeist/data_source_registry"
 REGISTRY_DBUS_INTERFACE = "org.gnome.zeitgeist.DataSourceRegistry"
 SIG_FULL_DATASOURCE = "(sssa("+constants.SIG_EVENT+")bxb)"
+DISK_WRITE_TIMEOUT = 5*60*1000 # 5 Minutes
 
 class DataSource(OrigDataSource):
 	@classmethod
@@ -88,16 +90,25 @@
 		
 		# Connect to client disconnection signals
 		dbus.SessionBus().add_signal_receiver(self._name_owner_changed,
-		    signal_name="NameOwnerChanged",
-		    dbus_interface=dbus.BUS_DAEMON_IFACE,
-		    arg2="", # only match services with no new owner
-	    )
-	
+			signal_name="NameOwnerChanged",
+			dbus_interface=dbus.BUS_DAEMON_IFACE,
+			arg2="", # only match services with no new owner
+		)
+		
+		self._dirty = True
+		gobject.timeout_add(DISK_WRITE_TIMEOUT, self._write_to_disk)
+		
 	def _write_to_disk(self):
 		data = [DataSource.get_plain(datasource) for datasource in
 			self._registry.itervalues()]
-		with open(DATA_FILE, "w") as data_file:
-			pickle.dump(data, data_file)
+		try:
+			if self._dirty:
+				self._dirty = False
+				with open(DATA_FILE, "w") as data_file:
+					pickle.dump(data, data_file)
+		except Exception, e:
+			log.warn("Failed to write to data file %s: %s" % (DATA_FILE, e))
+		return True
 		#log.debug("Data-source registry update written to disk.")
 	
 	def pre_insert_event(self, event, sender):
@@ -106,6 +117,7 @@
 				datasource = self._registry[unique_id]
 				# Update LastSeen time
 				datasource.last_seen = get_timestamp_for_now()
+				self._dirty = True
 				# Check whether the data-source is allowed to insert events
 				if not datasource.enabled:
 					return None