Fixing issue with duplicated rows when looping through groupby object (#459)

* Fixing issue with duplicated rows when looping through groupby object

* Resolves #458
* Only gets the unique values of the index
* Add test to validate changes

* Revert change to axis=1

* Lint

* Lint

* Add fix that Resolves #460

Author: devin-petersohn

modin/pandas/groupby.py

@@ -41,7 +41,7 @@ def __init__(
         self._columns = self._query_compiler.columns
         self._by = by
         # This tells us whether or not there are multiple columns/rows in the groupby
-        self._is_multi_by = all(obj in self._df for obj in self._by)
+        self._is_multi_by = all(obj in self._df for obj in self._by) and axis == 0
         self._level = level
         self._kwargs = {
             "sort": sort,
@@ -113,7 +113,7 @@ def _iter(self):
                     k,
                     DataFrame(
                         query_compiler=self._query_compiler.getitem_row_array(
-                            self._index.get_indexer_for(self._index_grouped[k])
+                            self._index.get_indexer_for(self._index_grouped[k].unique())
                         )
                     ),
                 )
@@ -125,7 +125,7 @@ def _iter(self):
                     k,
                     DataFrame(
                         query_compiler=self._query_compiler.getitem_column_array(
-                            self._index_grouped[k]
+                            self._index_grouped[k].unique()
                         )
                     ),
                 )

modin/pandas/test/test_groupby.py

@@ -764,3 +764,32 @@ def test_groups(ray_groupby, pandas_groupby):
 @pytest.fixture
 def test_shift(ray_groupby, pandas_groupby):
     assert ray_groupby.groups == pandas_groupby.groups
+
+
+def test_groupby_on_index_values_with_loop():
+    length = 2 ** 6
+    data = {
+        "a": np.random.randint(0, 100, size=length),
+        "b": np.random.randint(0, 100, size=length),
+        "c": np.random.randint(0, 100, size=length),
+    }
+    idx = ["g1" if i % 3 != 0 else "g2" for i in range(length)]
+    modin_df = pd.DataFrame(data, index=idx, columns=list("aba"))
+    pandas_df = pandas.DataFrame(data, index=idx, columns=list("aba"))
+    modin_groupby_obj = modin_df.groupby(modin_df.index)
+    pandas_groupby_obj = pandas_df.groupby(pandas_df.index)
+
+    modin_dict = {k: v for k, v in modin_groupby_obj}
+    pandas_dict = {k: v for k, v in pandas_groupby_obj}
+
+    for k in modin_dict:
+        ray_df_equals_pandas(modin_dict[k], pandas_dict[k])
+
+    modin_groupby_obj = modin_df.groupby(modin_df.columns, axis=1)
+    pandas_groupby_obj = pandas_df.groupby(pandas_df.columns, axis=1)
+
+    modin_dict = {k: v for k, v in modin_groupby_obj}
+    pandas_dict = {k: v for k, v in pandas_groupby_obj}
+
+    for k in modin_dict:
+        ray_df_equals_pandas(modin_dict[k], pandas_dict[k])