[REVIEW] Fix bug in dask-cudf caused by inconsistency with Pandas empty groupby result

CHANGELOG.md

@@ -134,6 +134,7 @@
 - PR #1607 Revert change of `column.to_dense_buffer` always return by copy for performance concerns
 - PR #1618 ORC reader: fix assert & data output when nrows/skiprows isn't aligned to stripe boundaries
 - PR #1631 Fix failure of TYPES_TEST on some gcc-7 based systems.
+- PR #1639 Fix bug caused by empty groupbys throwing an exception
 
 
 # cuDF 0.6.1 (25 Mar 2019)

python/cudf/dataframe/dataframe.py

@@ -722,7 +722,7 @@ def index(self):
     @index.setter
     def index(self, _index):
         if isinstance(_index, cudf.dataframe.multiindex.MultiIndex):
-            if len(_index) != len(self[self.columns[0]]):
+            if len(_index) != len(self):
                 msg = f"Length mismatch: Expected axis has "\
                        "%d elements, new values "\
                        "have %d elements"\

python/cudf/groupby/groupby.py

@@ -162,7 +162,30 @@ def _apply_basic_agg(self, agg_type, sort_results=False):
 
     def apply_multiindex_or_single_index(self, result):
         if len(result) == 0:
-            raise ValueError('Groupby result is empty!')
+            final_result = DataFrame()
+            for col in result.columns:
+                if col not in self._by:
+                    final_result[col] = result[col]
+            if len(self._by) == 1 or len(final_result) == 0:
+                dtype = 'float64' if len(self._by) == 1 else 'object'
+                name = self._by[0] if len(self._by) == 1 else None
+                from cudf.dataframe.index import GenericIndex
+                index = GenericIndex(Series([], dtype=dtype))
+                index.name = name
+                final_result.index = index
+            else:
+                levels = []
+                codes = []
+                names = []
+                for by in self._by:
+                    levels.append([])
+                    codes.append([])
+                    names.append(by)
+                from cudf import MultiIndex
+                mi = MultiIndex(levels, codes)
+                mi.names = names
+                final_result.index = mi
+            return final_result
         if len(self._by) == 1:
             from cudf.dataframe import index
             idx = index.as_index(result[self._by[0]])

python/cudf/tests/test_groupby.py

@@ -383,9 +383,7 @@ def test_advanced_groupby_levels():
     gdh = gdg.groupby(level=1).sum()
     assert_eq(pdh, gdh)
     pdg = pdf.groupby(['x', 'y', 'z']).sum()
-    with pytest.raises(ValueError) as raises:
-        gdg = gdf.groupby(['x', 'y', 'z']).sum()
-    raises.match("Groupby result is empty!")
+    gdg = gdf.groupby(['x', 'y', 'z']).sum()
     pdg = pdf.groupby(['z']).sum()
     gdg = gdf.groupby(['z']).sum()
     assert_eq(pdg, gdg)
@@ -432,3 +430,26 @@ def test_list_of_series():
     pdg = pdf.groupby([pdf.x, pdf.y]).y.sum()
     gdg = gdf.groupby([gdf.x, gdf.y]).y.sum()
     assert_eq(pdg, gdg)
+
+
+def test_empty_groupby():
+    pdf = pd.DataFrame({'x': [], 'y': []})
+    gdf = cudf.from_pandas(pdf)
+    pdg = pdf.groupby(['x', 'y']).sum()
+    gdg = gdf.groupby(['x', 'y']).sum()
+    assert_eq(pdg, gdg)
+    pdg = pdf.groupby(['x', 'y']).agg('sum')
+    gdg = gdf.groupby(['x', 'y']).agg('sum')
+    assert_eq(pdg, gdg)
+    pdg = pdf.groupby(['y']).sum()
+    gdg = gdf.groupby(['y']).sum()
+    assert_eq(pdg, gdg)
+    pdg = pdf.groupby(['y']).agg('sum')
+    gdg = gdf.groupby(['y']).agg('sum')
+    assert_eq(pdg, gdg)
+    pdg = pdf.groupby(['x']).sum()
+    gdg = gdf.groupby(['x']).sum()
+    assert_eq(pdg, gdg)
+    pdg = pdf.groupby(['x']).agg('sum')
+    gdg = gdf.groupby(['x']).agg('sum')
+    assert_eq(pdg, gdg)