performance improvement in DataFrame.duplicated

doc/source/whatsnew/v0.16.0.txt

@@ -267,6 +267,7 @@ Performance
 - Performance and memory usage improvements in ``merge`` when key space exceeds ``int64`` bounds (:issue:`9151`)
 - Performance improvements in multi-key ``groupby`` (:issue:`9429`)
 - Performance improvements in ``MultiIndex.sortlevel`` (:issue:`9445`)
+- Performance and memory usage improvements in ``DataFrame.duplicated`` (:issue:`9398`)
 - Cythonized ``Period`` (:issue:`9440`)
 
 Bug Fixes

pandas/core/frame.py

@@ -2831,30 +2831,32 @@ def duplicated(self, subset=None, take_last=False):
         -------
         duplicated : Series
         """
-        # kludge for #1833
-        def _m8_to_i8(x):
-            if issubclass(x.dtype.type, np.datetime64):
-                return x.view(np.int64)
-            return x
+        from pandas.core.groupby import get_group_index
+        from pandas.hashtable import duplicated_int64, _SIZE_HINT_LIMIT
+
+        size_hint = min(len(self), _SIZE_HINT_LIMIT)
+
+        def factorize(vals):
+            (hash_klass, vec_klass), vals = \
+                    algos._get_data_algo(vals, algos._hashtables)
+
+            uniques, table = vec_klass(), hash_klass(size_hint)
+            labels = table.get_labels(vals, uniques, 0, -1)
+
+            return labels.astype('i8', copy=False), len(uniques)
 
-        # if we are only duplicating on Categoricals this can be much faster
         if subset is None:
-            values = list(_m8_to_i8(self.get_values().T))
-        else:
-            if np.iterable(subset) and not isinstance(subset, compat.string_types):
-                if isinstance(subset, tuple):
-                    if subset in self.columns:
-                        values = [self[subset].get_values()]
-                    else:
-                        values = [_m8_to_i8(self[x].get_values()) for x in subset]
-                else:
-                    values = [_m8_to_i8(self[x].get_values()) for x in subset]
-            else:
-                values = [self[subset].get_values()]
+            subset = self.columns
+        elif not np.iterable(subset) or \
+                isinstance(subset, compat.string_types) or \
+                isinstance(subset, tuple) and subset in self.columns:
+            subset = subset,
+
+        vals = (self[col].values for col in subset)
+        labels, shape = map(list, zip( * map(factorize, vals)))
 
-        keys = lib.fast_zip_fillna(values)
-        duplicated = lib.duplicated(keys, take_last=take_last)
-        return Series(duplicated, index=self.index)
+        ids = get_group_index(labels, shape, sort=False, xnull=False)
+        return Series(duplicated_int64(ids, take_last), index=self.index)
 
     #----------------------------------------------------------------------
     # Sorting

pandas/lib.pyx

@@ -1278,7 +1278,7 @@ def fast_zip_fillna(list ndarrays, fill_value=pandas_null):
 def duplicated(ndarray[object] values, take_last=False):
     cdef:
         Py_ssize_t i, n
-        dict seen = {}
+        set seen = set()
         object row
 
     n = len(values)
@@ -1291,15 +1291,15 @@ def duplicated(ndarray[object] values, take_last=False):
             if row in seen:
                 result[i] = 1
             else:
-                seen[row] = None
+                seen.add(row)
                 result[i] = 0
     else:
         for i from 0 <= i < n:
             row = values[i]
             if row in seen:
                 result[i] = 1
             else:
-                seen[row] = None
+                seen.add(row)
                 result[i] = 0
 
     return result.view(np.bool_)

vb_suite/frame_methods.py

@@ -506,3 +506,17 @@ def get_data(n=100000):
                                 setup,
                                 name='frame_from_records_generator_nrows',
                                 start_date=datetime(2013,10,04))  # issue-4911
+
+setup = common_setup + '''
+n = 1 << 20
+
+t = date_range('2015-01-01', freq='S', periods=n // 64)
+xs = np.random.randn(n // 64).round(2)
+
+df = DataFrame({'a':np.random.randint(- 1 << 8, 1 << 8, n),
+                'b':np.random.choice(t, n),
+                'c':np.random.choice(xs, n)})
+'''
+
+frame_duplicated = Benchmark('df.duplicated()', setup,
+                             name='frame_duplicated')