Remove cudf.Scalar from interval_range (#17844)

Towards #17843

Needed to change some `interval_range` test to account for pandas-dev/pandas#57268

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

Approvers:
  - Matthew Murray (https://github.com/Matt711)

URL: #17844

Author: mroeschke

python/cudf/cudf/core/index.py

@@ -48,11 +48,14 @@
 from cudf.core.dtypes import IntervalDtype
 from cudf.core.join._join_helpers import _match_join_keys
 from cudf.core.mixins import BinaryOperand
+from cudf.core.scalar import pa_scalar_to_plc_scalar
 from cudf.core.single_column_frame import SingleColumnFrame
 from cudf.utils.docutils import copy_docstring
 from cudf.utils.dtypes import (
     SIZE_TYPE_DTYPE,
     _maybe_convert_to_default_type,
+    cudf_dtype_from_pa_type,
+    cudf_dtype_to_pa_type,
     find_common_type,
     is_mixed_with_object_dtype,
 )
@@ -3346,50 +3349,56 @@ def interval_range(
             "freq, exactly three must be specified"
         )
 
-    start = cudf.Scalar(start) if start is not None else start
-    end = cudf.Scalar(end) if end is not None else end
     if periods is not None and not cudf.api.types.is_integer(periods):
         warnings.warn(
             "Non-integer 'periods' in cudf.date_range, and cudf.interval_range"
             " are deprecated and will raise in a future version.",
             FutureWarning,
         )
-    periods = cudf.Scalar(int(periods)) if periods is not None else periods
-    freq = cudf.Scalar(freq) if freq is not None else freq
-
     if start is None:
         start = end - freq * periods
     elif freq is None:
-        quotient, remainder = divmod((end - start).value, periods.value)
+        quotient, remainder = divmod(end - start, periods)
         if remainder:
             freq = (end - start) / periods
         else:
-            freq = cudf.Scalar(int(quotient))
+            freq = int(quotient)
     elif periods is None:
-        periods = cudf.Scalar(int((end - start) / freq))
+        periods = int((end - start) / freq)
     elif end is None:
         end = start + periods * freq
 
+    pa_start = pa.scalar(start)
+    pa_end = pa.scalar(end)
+    pa_freq = pa.scalar(freq)
+
     if any(
-        not _is_non_decimal_numeric_dtype(x.dtype)
-        for x in (start, periods, freq, end)
+        not _is_non_decimal_numeric_dtype(cudf_dtype_from_pa_type(x.type))
+        for x in (pa_start, pa.scalar(periods), pa_freq, pa_end)
     ):
         raise ValueError("start, end, periods, freq must be numeric values.")
 
-    periods = periods.astype("int64")
-    common_dtype = find_common_type((start.dtype, freq.dtype, end.dtype))
-    start = start.astype(common_dtype)
-    freq = freq.astype(common_dtype)
+    common_dtype = find_common_type(
+        (
+            cudf_dtype_from_pa_type(pa_start.type),
+            cudf_dtype_from_pa_type(pa_freq.type),
+            cudf_dtype_from_pa_type(pa_end.type),
+        )
+    )
+    pa_start = pa_start.cast(cudf_dtype_to_pa_type(common_dtype))
+    pa_freq = pa_freq.cast(cudf_dtype_to_pa_type(common_dtype))
 
     with acquire_spill_lock():
         bin_edges = libcudf.column.Column.from_pylibcudf(
             plc.filling.sequence(
                 size=periods + 1,
-                init=start.device_value,
-                step=freq.device_value,
+                init=pa_scalar_to_plc_scalar(pa_start),
+                step=pa_scalar_to_plc_scalar(pa_freq),
             )
         )
-    return IntervalIndex.from_breaks(bin_edges, closed=closed, name=name)
+    return IntervalIndex.from_breaks(
+        bin_edges.astype(common_dtype), closed=closed, name=name
+    )
 
 
 class IntervalIndex(Index):

python/cudf/cudf/tests/indexes/test_interval.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2023-2024, NVIDIA CORPORATION.
+# Copyright (c) 2023-2025, NVIDIA CORPORATION.
 import numpy as np
 import pandas as pd
 import pyarrow as pa
@@ -30,7 +30,6 @@ def test_interval_to_arrow():
     np.int64,
     np.float32,
     np.float64,
-    cudf.Scalar,
 ]
 
 PERIODS_TYPES = [
@@ -39,10 +38,23 @@ def test_interval_to_arrow():
     np.int16,
     np.int32,
     np.int64,
-    cudf.Scalar,
 ]
 
 
+def assert_with_pandas_2_bug(pindex, gindex):
+    # pandas upcasts to 64 bit https://github.com/pandas-dev/pandas/issues/57268
+    # using Series to use check_dtype
+    if gindex.dtype.subtype.kind == "f":
+        gindex = gindex.astype(
+            cudf.IntervalDtype(subtype="float64", closed=gindex.dtype.closed)
+        )
+    elif gindex.dtype.subtype.kind == "i":
+        gindex = gindex.astype(
+            cudf.IntervalDtype(subtype="int64", closed=gindex.dtype.closed)
+        )
+    assert_eq(pd.Series(pindex), cudf.Series(gindex), check_dtype=False)
+
+
 @pytest.mark.parametrize("closed", ["left", "right", "both", "neither"])
 @pytest.mark.parametrize("start", [0, 1, 2, 3])
 @pytest.mark.parametrize("end", [4, 5, 6, 7])
@@ -57,9 +69,7 @@ def test_interval_range_basic(start, end, closed):
 @pytest.mark.parametrize("end_t", INTERVAL_BOUNDARY_TYPES)
 def test_interval_range_dtype_basic(start_t, end_t):
     start, end = start_t(24), end_t(42)
-    start_val = start.value if isinstance(start, cudf.Scalar) else start
-    end_val = end.value if isinstance(end, cudf.Scalar) else end
-    pindex = pd.interval_range(start=start_val, end=end_val, closed="left")
+    pindex = pd.interval_range(start=start, end=end, closed="left")
     gindex = cudf.interval_range(start=start, end=end, closed="left")
 
     assert_eq(pindex, gindex)
@@ -91,27 +101,11 @@ def test_interval_range_freq_basic(start, end, freq, closed):
 @pytest.mark.parametrize("freq_t", INTERVAL_BOUNDARY_TYPES)
 def test_interval_range_freq_basic_dtype(start_t, end_t, freq_t):
     start, end, freq = start_t(5), end_t(70), freq_t(3)
-    start_val = start.value if isinstance(start, cudf.Scalar) else start
-    end_val = end.value if isinstance(end, cudf.Scalar) else end
-    freq_val = freq.value if isinstance(freq, cudf.Scalar) else freq
-    pindex = pd.interval_range(
-        start=start_val, end=end_val, freq=freq_val, closed="left"
-    )
+    pindex = pd.interval_range(start=start, end=end, freq=freq, closed="left")
     gindex = cudf.interval_range(
         start=start, end=end, freq=freq, closed="left"
     )
-    if gindex.dtype.subtype.kind == "f":
-        gindex = gindex.astype(
-            cudf.IntervalDtype(subtype="float64", closed=gindex.dtype.closed)
-        )
-    elif gindex.dtype.subtype.kind == "i":
-        gindex = gindex.astype(
-            cudf.IntervalDtype(subtype="int64", closed=gindex.dtype.closed)
-        )
-
-    # pandas upcasts to 64 bit https://github.com/pandas-dev/pandas/issues/57268
-    # using Series to use check_dtype
-    assert_eq(pd.Series(pindex), cudf.Series(gindex), check_dtype=False)
+    assert_with_pandas_2_bug(pindex, gindex)
 
 
 @pytest.mark.parametrize("closed", ["left", "right", "both", "neither"])
@@ -134,13 +128,8 @@ def test_interval_range_periods_basic(start, end, periods, closed):
 @pytest.mark.parametrize("periods_t", PERIODS_TYPES)
 def test_interval_range_periods_basic_dtype(start_t, end_t, periods_t):
     start, end, periods = start_t(0), end_t(4), periods_t(1)
-    start_val = start.value if isinstance(start, cudf.Scalar) else start
-    end_val = end.value if isinstance(end, cudf.Scalar) else end
-    periods_val = (
-        periods.value if isinstance(periods, cudf.Scalar) else periods
-    )
     pindex = pd.interval_range(
-        start=start_val, end=end_val, periods=periods_val, closed="left"
+        start=start, end=end, periods=periods, closed="left"
     )
     gindex = cudf.interval_range(
         start=start, end=end, periods=periods, closed="left"
@@ -188,19 +177,13 @@ def test_interval_range_periods_freq_end(end, freq, periods, closed):
 @pytest.mark.parametrize("end_t", INTERVAL_BOUNDARY_TYPES)
 def test_interval_range_periods_freq_end_dtype(periods_t, freq_t, end_t):
     periods, freq, end = periods_t(2), freq_t(3), end_t(10)
-    freq_val = freq.value if isinstance(freq, cudf.Scalar) else freq
-    end_val = end.value if isinstance(end, cudf.Scalar) else end
-    periods_val = (
-        periods.value if isinstance(periods, cudf.Scalar) else periods
-    )
     pindex = pd.interval_range(
-        end=end_val, freq=freq_val, periods=periods_val, closed="left"
+        end=end, freq=freq, periods=periods, closed="left"
     )
     gindex = cudf.interval_range(
         end=end, freq=freq, periods=periods, closed="left"
     )
-
-    assert_eq(pindex, gindex)
+    assert_with_pandas_2_bug(pindex, gindex)
 
 
 @pytest.mark.parametrize("closed", ["left", "right", "both", "neither"])
@@ -223,21 +206,13 @@ def test_interval_range_periods_freq_start(start, freq, periods, closed):
 @pytest.mark.parametrize("start_t", INTERVAL_BOUNDARY_TYPES)
 def test_interval_range_periods_freq_start_dtype(periods_t, freq_t, start_t):
     periods, freq, start = periods_t(2), freq_t(3), start_t(9)
-    freq_val = freq.value if isinstance(freq, cudf.Scalar) else freq
-    start_val = start.value if isinstance(start, cudf.Scalar) else start
-    periods_val = (
-        periods.value if isinstance(periods, cudf.Scalar) else periods
-    )
     pindex = pd.interval_range(
-        start=start_val, freq=freq_val, periods=periods_val, closed="left"
+        start=start, freq=freq, periods=periods, closed="left"
     )
     gindex = cudf.interval_range(
         start=start, freq=freq, periods=periods, closed="left"
     )
-
-    # pandas upcasts to 64 bit https://github.com/pandas-dev/pandas/issues/57268
-    # using Series to use check_dtype
-    assert_eq(pd.Series(pindex), cudf.Series(gindex), check_dtype=False)
+    assert_with_pandas_2_bug(pindex, gindex)
 
 
 @pytest.mark.parametrize("closed", ["right", "left", "both", "neither"])