Enable pytest failures on warnings from cudf (Ensure values being set…

… are casted to the object's type) (#1358)

In 24.04, cudf issues a `FutureWarning` when setting a value to a column would change the data type of the original column. The fix involves casting the original column to the value's type first or vice versa. Could use a second eye on the correct direction to cast if a data type was used intentionally.

Additionally the test suite will now fail when a warning comes from cudf as apart of rapidsai/build-planning#26

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)

Approvers:
  - Mark Harris (https://github.com/harrism)
  - H. Thomson Comer (https://github.com/thomcom)

URL: #1358

python/cuspatial/cuspatial/core/binpreds/basic_predicates.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 import cudf
 
@@ -67,7 +67,7 @@ def _basic_intersects_count(lhs, rhs):
     is_degenerate = _multipoints_is_degenerate(intersections)
     # If all the points in the intersection are in the rhs
     if len(is_degenerate) > 0:
-        sizes[is_degenerate] = 1
+        sizes[is_degenerate] = sizes.dtype.type(1)
     return sizes
 
 

python/cuspatial/cuspatial/core/binpreds/contains_geometry_processor.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 import cupy as cp
 
@@ -231,6 +231,7 @@ def _postprocess_multipoint_rhs(
                 return count_result
             hits = result_df["point_index_x"]
             hits.index = count_result.iloc[result_df["rhs_index"]].index
+            count_result = count_result.astype(hits.dtype)
             count_result.iloc[result_df["rhs_index"]] = hits
             return count_result
 

python/cuspatial/cuspatial/core/geoseries.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2023, NVIDIA CORPORATION
+# Copyright (c) 2020-2024, NVIDIA CORPORATION
 
 from functools import cached_property
 from numbers import Integral
@@ -660,15 +660,15 @@ def _align_to_index(
                 index, how, sort, allow_non_unique
             )
         ).astype("int32")
-        aligned_union_offsets[
-            aligned_union_offsets.isna()
-        ] = Feature_Enum.NONE.value
+        aligned_union_offsets[aligned_union_offsets.isna()] = np.int32(
+            Feature_Enum.NONE.value
+        )
         aligned_input_types = self._column._meta.input_types._align_to_index(
             index, how, sort, allow_non_unique
         ).astype("int8")
-        aligned_input_types[
-            aligned_input_types.isna()
-        ] = Feature_Enum.NONE.value
+        aligned_input_types[aligned_input_types.isna()] = np.int8(
+            Feature_Enum.NONE.value
+        )
         column = GeoColumn(
             (
                 self._column.points,

python/cuspatial/cuspatial/tests/test_geodataframe.py

@@ -1,4 +1,6 @@
-# Copyright (c) 2020-2021, NVIDIA CORPORATION.
+# Copyright (c) 2020-2024, NVIDIA CORPORATION.
+import sys
+
 import geopandas as gpd
 import numpy as np
 import pandas as pd
@@ -321,8 +323,9 @@ def test_boolmask(gpdf, df_boolmask):
 
 
 @pytest.mark.xfail(
+    sys.version_info.major >= 3 and sys.version_info.minor >= 11,
     reason="Size discrepancies between Python versions. See "
-    "https://github.com/rapidsai/cuspatial/issues/1352"
+    "https://github.com/rapidsai/cuspatial/issues/1352",
 )
 def test_memory_usage(gs):
     assert gs.memory_usage() == 224

python/cuspatial/cuspatial/tests/test_geoseries.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2021, NVIDIA CORPORATION.
+# Copyright (c) 2020-2024, NVIDIA CORPORATION.
 
 from enum import Enum
 from numbers import Integral
@@ -27,7 +27,7 @@
 np.random.seed(0)
 
 
-class Test_Feature_Enum(Enum):
+class Example_Feature_Enum(Enum):
     POINT = 0
     MULTIPOINT = 1
     LINESTRING = 2
@@ -54,7 +54,7 @@ def random_multipolygon(size):
 
 
 def generate_random_shapely_feature(
-    size: Integral, has_z: bool = False, obj_type: Test_Feature_Enum = None
+    size: Integral, has_z: bool = False, obj_type: Example_Feature_Enum = None
 ):
     obj_type = obj_type.value if obj_type else np.random.randint(1, 7)
     if obj_type == 1:
@@ -78,7 +78,7 @@ def generate_random_shapely_feature(
             return random_multipolygon(size)
 
 
-def generator(size: Integral, obj_type: Test_Feature_Enum = None):
+def generator(size: Integral, obj_type: Example_Feature_Enum = None):
     geos_list = []
     for i in range(size):
         geo = generate_random_shapely_feature(3, obj_type)
@@ -257,7 +257,7 @@ def test_getitem_lines():
 
 
 def test_getitem_slice_same_index():
-    gps = gpd.GeoSeries(generator(3, Test_Feature_Enum.POINT))
+    gps = gpd.GeoSeries(generator(3, Example_Feature_Enum.POINT))
     cus = cuspatial.from_geopandas(gps)
     assert_eq_geo(cus[0:1].to_geopandas(), gps[0:1])
     assert_eq_geo(cus[0:1].to_geopandas(), gps[0:1])

python/cuspatial/cuspatial/utils/binpred_utils.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 import cupy as cp
 import numpy as np
@@ -363,7 +363,7 @@ def _points_and_lines_to_multipoints(geoseries, offsets):
     points = geoseries[points_mask]
     lines = geoseries[lines_mask]
     points_offsets = _zero_series(len(geoseries))
-    points_offsets[points_mask] = 1
+    points_offsets[points_mask] = points_offsets.dtype.type(1)
     lines_series = geoseries[lines_mask]
     lines_sizes = lines_series.sizes
     xy = _zero_series(len(points.points.xy) + len(lines.lines.xy))
@@ -372,9 +372,10 @@ def _points_and_lines_to_multipoints(geoseries, offsets):
         lines_sizes.index = points_offsets[lines_mask].index
         points_offsets[lines_mask] = lines_series.sizes.values
         sizes[lines_mask] = lines.sizes.values * 2
-    sizes[points_mask] = 2
+    sizes[points_mask] = sizes.dtype.type(2)
     # TODO Inevitable host device copy
     points_xy_mask = cp.array(np.repeat(points_mask, sizes.values_host))
+    xy = xy.astype(points.points.xy.dtype)
     xy.iloc[points_xy_mask] = points.points.xy.reset_index(drop=True)
     xy.iloc[~points_xy_mask] = lines.lines.xy.reset_index(drop=True)
     collected_offsets = cudf.concat(
@@ -446,6 +447,7 @@ def _pli_features_rebuild_offsets(pli, features):
     # Recompute the offsets for the new series
     grouped_sizes = in_sizes.groupby(level=0).sum().sort_index()
     out_sizes = _zero_series(len(pli[0]) - 1)
+    out_sizes = out_sizes.astype(grouped_sizes.dtype)
     out_sizes.iloc[grouped_sizes.index] = grouped_sizes
     offsets = cudf.concat([cudf.Series([0]), out_sizes.cumsum()])
     return offsets

python/cuspatial/pyproject.toml

@@ -1,4 +1,4 @@
-# Copyright (c) 2022-2023, NVIDIA CORPORATION.
+# Copyright (c) 2022-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -120,3 +120,9 @@ wheel.packages = ["cuspatial"]
 provider = "scikit_build_core.metadata.regex"
 input = "cuspatial/VERSION"
 regex = "(?P<value>.*)"
+
+[tool.pytest.ini_options]
+xfail_strict = true
+filterwarnings = [
+    "error:::cudf"
+]