Remove cudf._lib.filling in favor of inlining pylibcudf (#17459)

Contributes to #17317

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

Approvers:
  - Lawrence Mitchell (https://github.com/wence-)

URL: #17459

python/cudf/cudf/_lib/CMakeLists.txt

@@ -18,7 +18,6 @@ set(cython_sources
     column.pyx
     copying.pyx
     csv.pyx
-    filling.pyx
     groupby.pyx
     interop.pyx
     merge.pyx

python/cudf/cudf/_lib/__init__.py

@@ -5,7 +5,6 @@
     binaryop,
     copying,
     csv,
-    filling,
     groupby,
     interop,
     merge,

python/cudf/cudf/core/column/categorical.py

@@ -668,13 +668,8 @@ def _fill(
             return self if inplace else self.copy()
 
         fill_code = self._encode(fill_value)
-        fill_scalar = cudf._lib.scalar.as_device_scalar(
-            fill_code, self.codes.dtype
-        )
-
         result = self if inplace else self.copy()
-
-        libcudf.filling.fill_in_place(result.codes, begin, end, fill_scalar)
+        result.codes._fill(fill_code, begin, end, inplace=True)
         return result
 
     def slice(self, start: int, stop: int, stride: int | None = None) -> Self:

python/cudf/cudf/core/column/column.py

@@ -401,14 +401,19 @@ def _fill(
         # the scalar is None when calling `is_valid`.
         slr = cudf.Scalar(fill_value, dtype=self.dtype)
 
-        if not inplace:
-            return libcudf.filling.fill(self, begin, end, slr.device_value)
-
-        if is_string_dtype(self.dtype):
-            return self._mimic_inplace(
-                libcudf.filling.fill(self, begin, end, slr.device_value),
-                inplace=True,
-            )
+        if not inplace or is_string_dtype(self.dtype):
+            with acquire_spill_lock():
+                result = type(self).from_pylibcudf(
+                    plc.filling.fill(
+                        self.to_pylibcudf(mode="read"),
+                        begin,
+                        end,
+                        slr.device_value.c_value,
+                    )
+                )
+            if is_string_dtype(self.dtype):
+                return self._mimic_inplace(result, inplace=True)
+            return result  # type: ignore[return-value]
 
         if not slr.is_valid() and not self.nullable:
             mask = as_buffer(
@@ -418,8 +423,13 @@ def _fill(
             )
             self.set_base_mask(mask)
 
-        libcudf.filling.fill_in_place(self, begin, end, slr.device_value)
-
+        with acquire_spill_lock():
+            plc.filling.fill_in_place(
+                self.to_pylibcudf(mode="write"),
+                begin,
+                end,
+                slr.device_value.c_value,
+            )
         return self
 
     def shift(self, offset: int, fill_value: ScalarLike) -> ColumnBase:
@@ -1813,11 +1823,18 @@ def as_column(
     * range objects
     """
     if isinstance(arbitrary, (range, pd.RangeIndex, cudf.RangeIndex)):
-        column = libcudf.filling.sequence(
-            len(arbitrary),
-            as_device_scalar(arbitrary.start, dtype=cudf.dtype("int64")),
-            as_device_scalar(arbitrary.step, dtype=cudf.dtype("int64")),
-        )
+        with acquire_spill_lock():
+            column = Column.from_pylibcudf(
+                plc.filling.sequence(
+                    len(arbitrary),
+                    as_device_scalar(
+                        arbitrary.start, dtype=np.dtype(np.int64)
+                    ).c_value,
+                    as_device_scalar(
+                        arbitrary.step, dtype=np.dtype(np.int64)
+                    ).c_value,
+                )
+            )
         if cudf.get_option("default_integer_bitwidth") and dtype is None:
             dtype = cudf.dtype(
                 f'i{cudf.get_option("default_integer_bitwidth")//8}'

python/cudf/cudf/core/frame.py

@@ -1901,7 +1901,16 @@ def _repeat(
         if not is_scalar(repeats):
             repeats = as_column(repeats)
 
-        return libcudf.filling.repeat(columns, repeats)
+        with acquire_spill_lock():
+            plc_table = plc.Table(
+                [col.to_pylibcudf(mode="read") for col in columns]
+            )
+            if isinstance(repeats, ColumnBase):
+                repeats = repeats.to_pylibcudf(mode="read")
+            return [
+                libcudf.column.Column.from_pylibcudf(col)
+                for col in plc.filling.repeat(plc_table, repeats).columns()
+            ]
 
     @_performance_tracking
     @_warn_no_dask_cudf

python/cudf/cudf/core/index.py

@@ -20,7 +20,6 @@
 
 import cudf
 from cudf import _lib as libcudf
-from cudf._lib.filling import sequence
 from cudf._lib.types import size_type_dtype
 from cudf.api.extensions import no_default
 from cudf.api.types import (
@@ -3402,11 +3401,14 @@ def interval_range(
     start = start.astype(common_dtype)
     freq = freq.astype(common_dtype)
 
-    bin_edges = sequence(
-        size=periods + 1,
-        init=start.device_value,
-        step=freq.device_value,
-    )
+    with acquire_spill_lock():
+        bin_edges = libcudf.column.Column.from_pylibcudf(
+            plc.filling.sequence(
+                size=periods + 1,
+                init=start.device_value.c_value,
+                step=freq.device_value.c_value,
+            )
+        )
     return IntervalIndex.from_breaks(bin_edges, closed=closed, name=name)