call stats collector engine_stopped() after the engine is closed (to make sure all data from extensions has been collected), and added that method to documented api

docs/topics/stats.rst

@@ -185,6 +185,10 @@ class (which they all inherit from).
         for this spider can be accessed. This method is called automatically on
         the :signal:`spider_closed` signal.
 
+    .. method:: engine_stopped()
+
+        Called after the engine is stopped, to dump or persist global stats.
+
 Available Stats Collectors
 ==========================
 

scrapy/core/engine.py

@@ -260,3 +260,4 @@ class ExecutionEngine(object):
     @defer.inlineCallbacks
     def _finish_stopping_engine(self):
         yield send_catch_log_deferred(signal=signals.engine_stopped)
+        yield stats.engine_stopped()

scrapy/statscol.py

@@ -17,7 +17,6 @@ class StatsCollector(object):
     def __init__(self):
         self._dump = settings.getbool('STATS_DUMP')
         self._stats = {None: {}} # None is for global stats
-        dispatcher.connect(self._engine_stopped, signal=signals.engine_stopped)
 
     def get_value(self, key, default=None, spider=None):
         return self._stats[spider].get(key, default)
@@ -63,7 +62,7 @@ class StatsCollector(object):
                 spider=spider)
         self._persist_stats(stats, spider)
 
-    def _engine_stopped(self):
+    def engine_stopped(self):
         stats = self.get_stats()
         if self._dump:
             log.msg("Dumping global stats:\n" + pprint.pformat(stats))