Ensure that when fetching the length of a result set that the whole index isn't consumed (especially on Whoosh & Xapian).

Author: toastdriven

haystack/backends/__init__.py

@@ -396,6 +396,11 @@ def get_count(self):
         the results.
         """
         if self._hit_count is None:
+            # Limit the slice to 10 so we get a count without consuming
+            # everything.
+            if not self.end_offset:
+                self.end_offset = 10
+            
             if self._more_like_this:
                 # Special case for MLT.
                 self.run_mlt()

haystack/backends/whoosh_backend.py

@@ -344,9 +344,6 @@ def search(self, query_string, sort_by=None, start_offset=0, end_offset=None,
             if narrowed_results:
                 raw_results.filter(narrowed_results)
 
-            # Make sure we don't process bits we already have.
-            # import pdb; pdb.set_trace()
-            
             # Determine the page.
             page_num = 0
 

tests/core/tests/mocks.py

@@ -43,7 +43,7 @@ def search(self, query_string, sort_by=None, start_offset=0, end_offset=None,
         hits = len(self.mock_search_results)
         indexed_models = site.get_indexed_models()
 
-        sliced = self.mock_search_results[start_offset:end_offset]
+        sliced = self.mock_search_results
 
         for result in sliced:
             model = get_model('core', self.model_name)
@@ -57,7 +57,7 @@ def search(self, query_string, sort_by=None, start_offset=0, end_offset=None,
                 hits -= 1
 
         return {
-            'results': results,
+            'results': results[start_offset:end_offset],
             'hits': hits,
         }
 

tests/core/tests/query.py

@@ -478,6 +478,7 @@ def test_load_all(self):
 
         # If nothing is registered, you get nothing.
         haystack.site.unregister(MockModel)
+        haystack.site.unregister(CharPKMockModel)
         sqs = self.msqs.load_all()
         self.assert_(isinstance(sqs, SearchQuerySet))
         self.assertEqual(len(sqs), 0)

tests/whoosh_tests/tests/whoosh_backend.py

@@ -580,6 +580,25 @@ def test_cache_is_full(self):
         fire_the_iterator_and_fill_cache = [result for result in results]
         self.assertEqual(results._cache_is_full(), True)
         self.assertEqual(len(backends.queries), 1)
+    
+    def test_count(self):
+        more_samples = []
+        
+        for i in xrange(1, 50):
+            mock = MockModel()
+            mock.id = i
+            mock.author = 'daniel%s' % i
+            mock.pub_date = date(2009, 2, 25) - timedelta(days=i)
+            more_samples.append(mock)
+        
+        self.sb.update(self.smmi, more_samples)
+        
+        backends.reset_search_queries()
+        self.assertEqual(len(backends.queries), 0)
+        results = self.sqs.all()
+        self.assertEqual(len(results), 49)
+        self.assertEqual(results._cache_is_full(), False)
+        self.assertEqual(len(backends.queries), 1)
 
 
 class WhooshRoundTripSearchIndex(SearchIndex):