PERF: don't materialize arrays on checking in groupby (#16413)

asv_bench/benchmarks/groupby.py

@@ -368,6 +368,11 @@ def setup(self):
         self.dates = (np.datetime64('now') + self.offsets)
         self.df = DataFrame({'key1': np.random.randint(0, 500, size=self.n), 'key2': np.random.randint(0, 100, size=self.n), 'value1': np.random.randn(self.n), 'value2': np.random.randn(self.n), 'value3': np.random.randn(self.n), 'dates': self.dates, })
 
+        N = 1000000
+        self.draws = pd.Series(np.random.randn(N))
+        labels = pd.Series(['foo', 'bar', 'baz', 'qux'] * (N // 4))
+        self.cats = labels.astype('category')
+
     def time_groupby_multi_size(self):
         self.df.groupby(['key1', 'key2']).size()
 
@@ -377,6 +382,10 @@ def time_groupby_dt_size(self):
     def time_groupby_dt_timegrouper_size(self):
         self.df.groupby(TimeGrouper(key='dates', freq='M')).size()
 
+    def time_groupby_size(self):
+        self.draws.groupby(self.cats).size()
+
+
 
 #----------------------------------------------------------------------
 # groupby with a variable value for ngroups

doc/source/whatsnew/v0.20.2.txt

@@ -29,7 +29,7 @@ Performance Improvements
 - Performance regression fix when indexing with a list-like (:issue:`16285`)
 - Performance regression fix for MultiIndexes (:issue:`16319`, :issue:`16346`)
 - Improved performance of ``.clip()`` with scalar arguments (:issue:`15400`)
-
+- Improved performance of groupby with categorical groupers (:issue:`16413`)
 
 .. _whatsnew_0202.bug_fixes:
 

pandas/core/indexes/base.py

@@ -2388,7 +2388,6 @@ def get_loc(self, key, method=None, tolerance=None):
             if tolerance is not None:
                 raise ValueError('tolerance argument only valid if using pad, '
                                  'backfill or nearest lookups')
-            key = _values_from_object(key)
             try:
                 return self._engine.get_loc(key)
             except KeyError: