Ensure Cython bindings only interact with pylibcudf objects (#1523)

Follow up to #1514

Addressing the review in #1514 (comment) by ensuring the Cython bindings return pylibcudf objects where possible instead of python objects (like cudf Python columns) so we can also have better return annotations for these methods where appropriate. Subsequently, the conversion from pylibcudf to cudf Python has been moved to the cuspatial Python layer.

Additionally, this PR will be necessary for rapidsai/cudf#17760 not to break cuspatial

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)
  - Michael Wang (https://github.com/isVoid)
  - Bradley Dice (https://github.com/bdice)

Approvers:
  - Michael Wang (https://github.com/isVoid)
  - Bradley Dice (https://github.com/bdice)

URL: #1523

python/cuspatial/cuspatial/_lib/distance.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport make_shared, shared_ptr, unique_ptr
 from libcpp.utility cimport move, pair
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column, Table as plc_Table
 from pylibcudf.libcudf.column.column cimport column
 from pylibcudf.libcudf.column.column_view cimport column_view
@@ -26,7 +25,7 @@ from cuspatial._lib.cpp.types cimport collection_type_id, geometry_type_id
 from cuspatial._lib.types cimport collection_type_py_to_c
 
 
-cpdef haversine_distance(
+cpdef plc_Column haversine_distance(
     plc_Column x1,
     plc_Column y1,
     plc_Column x2,
@@ -42,7 +41,7 @@ cpdef haversine_distance(
     with nogil:
         c_result = move(cpp_haversine_distance(c_x1, c_y1, c_x2, c_y2))
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(c_result)))
+    return plc_Column.from_libcudf(move(c_result))
 
 
 def directed_hausdorff_distance(
@@ -65,17 +64,14 @@ def directed_hausdorff_distance(
             )
         )
 
-    owner_col = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(result.first)),
-        data_ptr_exposed=True
-    )
+    cdef plc_Column owner_col = plc_Column.from_libcudf(move(result.first))
     cdef plc_Table plc_owner_table = plc_Table(
-        [owner_col.to_pylibcudf(mode="read")] * result.second.num_columns()
+        [owner_col] * result.second.num_columns()
     )
     cdef plc_Table plc_result_table = plc_Table.from_table_view(
         result.second, plc_owner_table
     )
-    return [Column.from_pylibcudf(col) for col in plc_result_table.columns()]
+    return plc_result_table.columns()
 
 
 def pairwise_point_distance(
@@ -108,7 +104,7 @@ def pairwise_point_distance(
             c_multipoints_lhs.get()[0],
             c_multipoints_rhs.get()[0],
         ))
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(c_result)))
+    return plc_Column.from_libcudf(move(c_result))
 
 
 def pairwise_linestring_distance(
@@ -134,7 +130,7 @@ def pairwise_linestring_distance(
             c_multilinestring_rhs.get()[0],
         ))
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(c_result)))
+    return plc_Column.from_libcudf(move(c_result))
 
 
 def pairwise_point_linestring_distance(
@@ -164,7 +160,7 @@ def pairwise_point_linestring_distance(
             c_multilinestrings.get()[0],
         ))
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(c_result)))
+    return plc_Column.from_libcudf(move(c_result))
 
 
 def pairwise_point_polygon_distance(
@@ -195,7 +191,7 @@ def pairwise_point_polygon_distance(
             c_multipoints.get()[0], c_multipolygons.get()[0]
         ))
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(c_result)))
+    return plc_Column.from_libcudf(move(c_result))
 
 
 def pairwise_linestring_polygon_distance(
@@ -221,7 +217,7 @@ def pairwise_linestring_polygon_distance(
             c_multilinestrings.get()[0], c_multipolygons.get()[0]
         ))
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(c_result)))
+    return plc_Column.from_libcudf(move(c_result))
 
 
 def pairwise_polygon_distance(plc_Column lhs, plc_Column rhs):
@@ -244,4 +240,4 @@ def pairwise_polygon_distance(plc_Column lhs, plc_Column rhs):
             c_lhs.get()[0], c_rhs.get()[0]
         ))
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(c_result)))
+    return plc_Column.from_libcudf(move(c_result))

python/cuspatial/cuspatial/_lib/intersection.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport make_shared, shared_ptr
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column
 
 from cuspatial._lib.types import CollectionType, GeometryType
@@ -26,8 +25,6 @@ def pairwise_linestring_intersection(plc_Column lhs, plc_Column rhs):
     """
     Compute the intersection of two (multi)linestrings.
     """
-    from cuspatial.core._column.geometa import Feature_Enum
-
     cdef linestring_intersection_column_result c_result
     cdef collection_type_id multi_type = <collection_type_id>(
         <underlying_collection_type_id_t>(CollectionType.MULTI.value)
@@ -52,39 +49,21 @@ def pairwise_linestring_intersection(plc_Column lhs, plc_Column rhs):
             c_lhs.get()[0], c_rhs.get()[0]
         ))
 
-    geometry_collection_offset = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.geometry_collection_offset))
-    )
-
-    types_buffer = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.types_buffer))
-    )
-    offset_buffer = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.offset_buffer))
-    )
-    points = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.points))
+    geometry_collection_offset = plc_Column.from_libcudf(
+        move(c_result.geometry_collection_offset)
     )
-    segments = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.segments))
+    types_buffer = plc_Column.from_libcudf(move(c_result.types_buffer))
+    offset_buffer = plc_Column.from_libcudf(move(c_result.offset_buffer))
+    points = plc_Column.from_libcudf(move(c_result.points))
+    segments = plc_Column.from_libcudf(move(c_result.segments))
+    lhs_linestring_id = plc_Column.from_libcudf(
+        move(c_result.lhs_linestring_id)
     )
-    lhs_linestring_id = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.lhs_linestring_id))
-    )
-    lhs_segment_id = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.lhs_segment_id))
-    )
-    rhs_linestring_id = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.rhs_linestring_id))
-    )
-    rhs_segment_id = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.rhs_segment_id))
-    )
-
-    # Map linestring type codes from libcuspatial to cuspatial
-    types_buffer[types_buffer == GeometryType.LINESTRING.value] = (
-        Feature_Enum.LINESTRING.value
+    lhs_segment_id = plc_Column.from_libcudf(move(c_result.lhs_segment_id))
+    rhs_linestring_id = plc_Column.from_libcudf(
+        move(c_result.rhs_linestring_id)
     )
+    rhs_segment_id = plc_Column.from_libcudf(move(c_result.rhs_segment_id))
 
     return ((geometry_collection_offset,
             types_buffer,

python/cuspatial/cuspatial/_lib/linestring_bounding_boxes.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport unique_ptr
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column, Table as plc_Table
 from pylibcudf.libcudf.column.column_view cimport column_view
 from pylibcudf.libcudf.table.table cimport table
@@ -13,9 +12,12 @@ from cuspatial._lib.cpp.linestring_bounding_boxes cimport (
 )
 
 
-cpdef linestring_bounding_boxes(plc_Column poly_offsets,
-                                plc_Column x, plc_Column y,
-                                double R):
+cpdef list linestring_bounding_boxes(
+    plc_Column poly_offsets,
+    plc_Column x,
+    plc_Column y,
+    double R
+):
     cdef column_view c_poly_offsets = poly_offsets.view()
     cdef column_view c_x = x.view()
     cdef column_view c_y = y.view()
@@ -25,7 +27,4 @@ cpdef linestring_bounding_boxes(plc_Column poly_offsets,
             c_poly_offsets, c_x, c_y, R
         ))
     cdef plc_Table plc_table = plc_Table.from_libcudf(move(result))
-    return [
-        Column.from_pylibcudf(col)
-        for col in plc_table.columns()
-    ]
+    return plc_table.columns()

python/cuspatial/cuspatial/_lib/nearest_points.pyx

@@ -2,7 +2,6 @@
 
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column
 from pylibcudf.libcudf.column.column_view cimport column_view
 
@@ -42,22 +41,18 @@ def pairwise_point_linestring_nearest_points(
 
     multipoint_geometry_id = None
     if multipoint_geometry_offset is not None:
-        multipoint_geometry_id = Column.from_pylibcudf(plc_Column.from_libcudf(
-            move(c_result.nearest_point_geometry_id.value())))
+        multipoint_geometry_id = plc_Column.from_libcudf(
+            move(c_result.nearest_point_geometry_id.value()))
 
     multilinestring_geometry_id = None
     if multilinestring_geometry_offset is not None:
-        multilinestring_geometry_id = Column.from_pylibcudf(
-            plc_Column.from_libcudf(
-                move(c_result.nearest_linestring_geometry_id.value())
-            )
+        multilinestring_geometry_id = plc_Column.from_libcudf(
+            move(c_result.nearest_linestring_geometry_id.value())
         )
 
-    segment_id = Column.from_pylibcudf(
-        plc_Column.from_libcudf(move(c_result.nearest_segment_id))
-    )
-    point_on_linestring_xy = Column.from_pylibcudf(plc_Column.from_libcudf(
-        move(c_result.nearest_point_on_linestring_xy)))
+    segment_id = plc_Column.from_libcudf(move(c_result.nearest_segment_id))
+    point_on_linestring_xy = plc_Column.from_libcudf(
+        move(c_result.nearest_point_on_linestring_xy))
 
     return (
         multipoint_geometry_id,

python/cuspatial/cuspatial/_lib/pairwise_multipoint_equals_count.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport make_shared, shared_ptr, unique_ptr
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column
 from pylibcudf.libcudf.column.column cimport column
 
@@ -42,4 +41,4 @@ def pairwise_multipoint_equals_count(
             )
         )
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(result)))
+    return plc_Column.from_libcudf(move(result))

python/cuspatial/cuspatial/_lib/pairwise_point_in_polygon.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport unique_ptr
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column
 from pylibcudf.libcudf.column.column cimport column
 from pylibcudf.libcudf.column.column_view cimport column_view
@@ -42,4 +41,4 @@ def pairwise_point_in_polygon(
             )
         )
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(result)))
+    return plc_Column.from_libcudf(move(result))

python/cuspatial/cuspatial/_lib/point_in_polygon.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport unique_ptr
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column
 from pylibcudf.libcudf.column.column cimport column
 from pylibcudf.libcudf.column.column_view cimport column_view
@@ -42,4 +41,4 @@ def point_in_polygon(
             )
         )
 
-    return Column.from_pylibcudf(plc_Column.from_libcudf(move(result)))
+    return plc_Column.from_libcudf(move(result))

python/cuspatial/cuspatial/_lib/points_in_range.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport unique_ptr
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column, Table as plc_Table
 from pylibcudf.libcudf.column.column_view cimport column_view
 from pylibcudf.libcudf.table.table cimport table
@@ -13,7 +12,7 @@ from cuspatial._lib.cpp.points_in_range cimport (
 )
 
 
-cpdef points_in_range(
+cpdef plc_Table points_in_range(
     double range_min_x,
     double range_max_x,
     double range_min_y,
@@ -37,12 +36,4 @@ cpdef points_in_range(
                 y_v
             )
         )
-
-    cdef plc_Table plc_table = plc_Table.from_libcudf(move(c_result))
-    return (
-        {
-            name: Column.from_pylibcudf(col)
-            for name, col in zip(["x", "y"], plc_table.columns())
-        },
-        None
-    )
+    return plc_Table.from_libcudf(move(c_result))

python/cuspatial/cuspatial/_lib/polygon_bounding_boxes.pyx

@@ -3,7 +3,6 @@
 from libcpp.memory cimport unique_ptr
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column, Table as plc_Table
 from pylibcudf.libcudf.column.column_view cimport column_view
 from pylibcudf.libcudf.table.table cimport table
@@ -13,9 +12,12 @@ from cuspatial._lib.cpp.polygon_bounding_boxes cimport (
 )
 
 
-cpdef polygon_bounding_boxes(plc_Column poly_offsets,
-                             plc_Column ring_offsets,
-                             plc_Column x, plc_Column y):
+cpdef list polygon_bounding_boxes(
+    plc_Column poly_offsets,
+    plc_Column ring_offsets,
+    plc_Column x,
+    plc_Column y,
+):
     cdef column_view c_poly_offsets = poly_offsets.view()
     cdef column_view c_ring_offsets = ring_offsets.view()
     cdef column_view c_x = x.view()
@@ -26,7 +28,4 @@ cpdef polygon_bounding_boxes(plc_Column poly_offsets,
             c_poly_offsets, c_ring_offsets, c_x, c_y
         ))
     cdef plc_Table table_result = plc_Table.from_libcudf(move(result))
-    return [
-        Column.from_pylibcudf(col)
-        for col in table_result.columns()
-    ]
+    return table_result.columns()

python/cuspatial/cuspatial/_lib/quadtree.pyx

@@ -5,7 +5,6 @@ from libcpp.memory cimport unique_ptr
 from libcpp.pair cimport pair
 from libcpp.utility cimport move
 
-from cudf.core.column.column import Column
 from pylibcudf cimport Column as plc_Column, Table as plc_Table
 from pylibcudf.libcudf.column.column cimport column
 from pylibcudf.libcudf.column.column_view cimport column_view
@@ -17,34 +16,25 @@ from cuspatial._lib.cpp.quadtree cimport (
 )
 
 
-cpdef quadtree_on_points(plc_Column x, plc_Column y,
-                         double x_min, double x_max,
-                         double y_min, double y_max,
-                         double scale,
-                         int8_t max_depth,
-                         size_type min_size):
+cpdef tuple quadtree_on_points(
+    plc_Column x,
+    plc_Column y,
+    double x_min,
+    double x_max,
+    double y_min,
+    double y_max,
+    double scale,
+    int8_t max_depth,
+    size_type min_size,
+):
     cdef column_view c_x = x.view()
     cdef column_view c_y = y.view()
     cdef pair[unique_ptr[column], unique_ptr[table]] result
     with nogil:
         result = move(cpp_quadtree_on_points(
             c_x, c_y, x_min, x_max, y_min, y_max, scale, max_depth, min_size
         ))
-    cdef plc_Table plc_table = plc_Table.from_libcudf(move(result.second))
-    result_names = [
-        "key",
-        "level",
-        "is_internal_node",
-        "length",
-        "offset"
-    ]
     return (
-        Column.from_pylibcudf(plc_Column.from_libcudf(move(result.first))),
-        (
-            {
-                name: Column.from_pylibcudf(col)
-                for name, col in zip(result_names, plc_table.columns())
-            },
-            None
-        )
+        plc_Column.from_libcudf(move(result.first)),
+        plc_Table.from_libcudf(move(result.second))
     )