API: add "unique=" argument to MultiIndex.get_level_values()

closes #17896

Author: toobaz

doc/source/whatsnew/v0.21.0.txt

@@ -792,6 +792,7 @@ Other API Changes
 - Pandas no longer registers matplotlib converters on import. The converters
   will be registered and used when the first plot is draw (:issue:`17710`)
 - Setting on a column with a scalar value and 0-len index now raises a ``ValueError`` (:issue:`16823`)
+- :func:`MultiIndex.get_level_values` now supports the `unique` argument (:issue:`17896`)
 
 
 .. _whatsnew_0210.deprecations:

pandas/core/indexes/multi.py

@@ -886,7 +886,7 @@ def _try_mi(k):
 
             raise InvalidIndexError(key)
 
-    def _get_level_values(self, level):
+    def _get_level_values(self, level, unique=False):
         """
         Return vector of label values for requested level,
         equal to the length of the index
@@ -896,20 +896,24 @@ def _get_level_values(self, level):
         Parameters
         ----------
         level : int level
+        unique : bool
+            if True, drop duplicated values
 
         Returns
         -------
         values : ndarray
         """
 
-        unique = self.levels[level]
+        values = self.levels[level]
         labels = self.labels[level]
-        filled = algos.take_1d(unique._values, labels,
-                               fill_value=unique._na_value)
-        values = unique._shallow_copy(filled)
+        if unique:
+            labels = np.unique(labels[labels != -1])
+        filled = algos.take_1d(values._values, labels,
+                               fill_value=values._na_value)
+        values = values._shallow_copy(filled)
         return values
 
-    def get_level_values(self, level):
+    def get_level_values(self, level, unique=False):
         """
         Return vector of label values for requested level,
         equal to the length of the index.
@@ -919,6 +923,8 @@ def get_level_values(self, level):
         level : int or str
             ``level`` is either the integer position of the level in the
             MultiIndex, or the name of the level.
+        unique : bool
+            if True, drop duplicated values
 
         Returns
         -------
@@ -942,7 +948,7 @@ def get_level_values(self, level):
         Index(['d', 'e', 'f'], dtype='object', name='level_2')
         """
         level = self._get_level_number(level)
-        values = self._get_level_values(level)
+        values = self._get_level_values(level, unique=unique)
         return values
 
     def format(self, space=2, sparsify=None, adjoin=True, names=False,

pandas/tests/indexes/test_multi.py

@@ -942,6 +942,13 @@ def test_get_level_values(self):
         tm.assert_index_equal(result, expected)
         assert result.name == 'first'
 
+        # with unique=True
+        result = self.index.get_level_values(0, unique=True)
+        expected = Index(['foo', 'bar', 'baz', 'qux'],
+                         name='first')
+        tm.assert_index_equal(result, expected)
+        assert result.name == 'first'
+
         result = self.index.get_level_values('first')
         expected = self.index.get_level_values(0)
         tm.assert_index_equal(result, expected)
@@ -988,6 +995,13 @@ def test_get_level_values_na(self):
         values = index.get_level_values(0)
         assert values.shape == (0, )
 
+        # with unique=True
+        arrays = [['a', 'b', 'b'], [2, np.nan, 2]]
+        index = pd.MultiIndex.from_arrays(arrays)
+        values = index.get_level_values(1, unique=True)
+        expected = np.array([2])
+        tm.assert_numpy_array_equal(values.values, expected)
+
     def test_reorder_levels(self):
         # this blows up
         tm.assert_raises_regex(IndexError, '^Too many levels',