Merge pull request #3 from VibhuJawa/branch-0.7

update to latest base

CHANGELOG.md

@@ -35,6 +35,7 @@
 - PR #1599 Level keyword supported in groupby
 - PR #929 Add support operations to dataframe
 - PR #1609 Groupby accept list of Series
+- PR #1658 Support `group_keys=True` keyword in groupby method
 
 ## Improvements
 
@@ -85,6 +86,7 @@
 - PR #1617 `has_nulls` and `column_dtypes` for `cudf::table`
 - PR #1590 Remove CFFI from the build / install process entirely
 - PR #1536 Convert gpuarrow CFFI to Cython
+- PR #1655 Add `Column._pointer` as a way to access underlying `gdf_column*` of a `Column`
 
 ## Bug Fixes
 
@@ -134,7 +136,7 @@
 - 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 #1641 CSV Reader: Fix skip_blank_lines behavior with Windows line terminators (\r\n) 
+- PR #1641 CSV Reader: Fix skip_blank_lines behavior with Windows line terminators (\r\n)
 - PR #1648 ORC reader: fix non-deterministic output when skiprows is non-zero
 
 

python/cudf/bindings/cudf_cpp.pxd

@@ -29,6 +29,8 @@ cdef gdf_context* create_context_view(flag_sorted, method, flag_distinct,
                                       flag_sort_result, flag_sort_inplace,
                                       null_sort_behavior)
 
+cpdef uintptr_t column_view_pointer(col)
+
 cpdef check_gdf_error(errcode)
 
 # Import cudf.h header to import all functions

python/cudf/bindings/cudf_cpp.pyx

@@ -248,6 +248,13 @@ cdef gdf_column* column_view_from_NDArrays(size, data, mask, dtype,
     return c_col
 
 
+cpdef uintptr_t column_view_pointer(col):
+    """
+    Return pointer to a view of the underlying <gdf_column*>
+    """
+    return <uintptr_t> column_view_from_column(col)
+
+
 cdef gdf_column_to_column_mem(gdf_column* input_col):
     gdf_dtype = input_col.dtype
     data_ptr = int(<uintptr_t>input_col.data)

python/cudf/dataframe/column.py

@@ -16,7 +16,7 @@
 
 from cudf.utils import cudautils, utils, ioutils
 from cudf.dataframe.buffer import Buffer
-from cudf.bindings.cudf_cpp import count_nonzero_mask
+from cudf.bindings.cudf_cpp import count_nonzero_mask, column_view_pointer
 from cudf.bindings.concat import _column_concat
 
 
@@ -588,3 +588,10 @@ def to_dlpack(self):
         """{docstring}"""
         import cudf.io.dlpack as dlpack
         return dlpack.to_dlpack(self)
+
+    @property
+    def _pointer(self):
+        """
+        Return pointer to a view of the underlying data structure
+        """
+        return column_view_pointer(self)

python/cudf/dataframe/dataframe.py

@@ -1771,7 +1771,7 @@ def join(self, other, on=None, how='left', lsuffix='', rsuffix='',
         return df
 
     def groupby(self, by=None, sort=False, as_index=True, method="hash",
-                level=None):
+                level=None, group_keys=True):
         """Groupby
 
         Parameters
@@ -1806,6 +1806,10 @@ def groupby(self, by=None, sort=False, as_index=True, method="hash",
         - Since we don't support multiindex, the *by* columns are stored
           as regular columns.
         """
+        if group_keys is not True:
+            raise NotImplementedError(
+                "The group_keys keyword is not yet implemented"
+            )
 
         if by is None and level is None:
             raise TypeError('groupby() requires either by or level to be'

python/cudf/dataframe/series.py

@@ -1568,7 +1568,13 @@ def diff(self, periods=1):
                                                     periods)
         return Series(output_dary, name=self.name, index=self.index)
 
-    def groupby(self, group_series=None, level=None, sort=False):
+    def groupby(self, group_series=None, level=None, sort=False,
+                group_keys=True):
+        if group_keys is not True:
+            raise NotImplementedError(
+                "The group_keys keyword is not yet implemented"
+            )
+
         from cudf.groupby.groupby import SeriesGroupBy
         return SeriesGroupBy(self, group_series, level, sort)
 

python/cudf/tests/test_groupby.py

@@ -61,6 +61,12 @@ def test_groupby_default(pdf, gdf):
     assert_eq(pdf, gdf)
 
 
+def test_group_keys_true(pdf, gdf):
+    gdf = gdf.groupby('y', group_keys=True).sum()
+    pdf = pdf.groupby('y', group_keys=True).sum()
+    assert_eq(pdf, gdf)
+
+
 def test_groupby_getitem_styles():
     pdf = pd.DataFrame({'x': [1, 3, 1], 'y': [1, 2, 3]})
     gdf = cudf.from_pandas(pdf)