Migrate filling operations to pylibcudf (#15225)

This PR migrates the filling operations in cuDF Python to pylibcudf.

Authors:
  - https://github.com/brandon-b-miller
  - Vyas Ramasubramani (https://github.com/vyasr)

Approvers:
  - Vyas Ramasubramani (https://github.com/vyasr)

URL: #15225

docs/cudf/source/conf.py

@@ -379,6 +379,7 @@ def _generate_namespaces(namespaces):
     "type_id",
     # Unknown base types
     "int32_t",
+    "void"
 }
 
 

docs/cudf/source/user_guide/api_docs/pylibcudf/filling.rst

@@ -0,0 +1,6 @@
+========
+filling
+========
+
+.. automodule:: cudf._lib.pylibcudf.filling
+   :members:

docs/cudf/source/user_guide/api_docs/pylibcudf/index.rst

@@ -13,6 +13,7 @@ This page provides API documentation for pylibcudf.
     column
     concatenate
     copying
+    filling
     gpumemoryview
     groupby
     join

python/cudf/cudf/_lib/filling.pyx

@@ -1,103 +1,57 @@
-# Copyright (c) 2020-2022, NVIDIA CORPORATION.
+# Copyright (c) 2020-2024, NVIDIA CORPORATION.
 
 from cudf.core.buffer import acquire_spill_lock
 
-from libcpp.memory cimport unique_ptr
-from libcpp.utility cimport move
 
-cimport cudf._lib.cpp.filling as cpp_filling
 from cudf._lib.column cimport Column
-from cudf._lib.cpp.column.column cimport column
-from cudf._lib.cpp.column.column_view cimport column_view, mutable_column_view
-from cudf._lib.cpp.scalar.scalar cimport scalar
-from cudf._lib.cpp.table.table cimport table
-from cudf._lib.cpp.table.table_view cimport table_view
-from cudf._lib.cpp.types cimport size_type
 from cudf._lib.scalar cimport DeviceScalar
-from cudf._lib.utils cimport columns_from_unique_ptr, table_view_from_columns
+from cudf._lib.utils cimport columns_from_pylibcudf_table
+
+from cudf._lib import pylibcudf
+from cudf._lib.scalar import as_device_scalar
 
 
 @acquire_spill_lock()
 def fill_in_place(Column destination, int begin, int end, DeviceScalar value):
-    cdef mutable_column_view c_destination = destination.mutable_view()
-    cdef size_type c_begin = <size_type> begin
-    cdef size_type c_end = <size_type> end
-    cdef const scalar* c_value = value.get_raw_ptr()
-
-    cpp_filling.fill_in_place(
-        c_destination,
-        c_begin,
-        c_end,
-        c_value[0]
+    pylibcudf.filling.fill_in_place(
+        destination.to_pylibcudf(mode='write'),
+        begin,
+        end,
+        (<DeviceScalar> as_device_scalar(value, dtype=destination.dtype)).c_value
     )
 
 
 @acquire_spill_lock()
 def fill(Column destination, int begin, int end, DeviceScalar value):
-    cdef column_view c_destination = destination.view()
-    cdef size_type c_begin = <size_type> begin
-    cdef size_type c_end = <size_type> end
-    cdef const scalar* c_value = value.get_raw_ptr()
-    cdef unique_ptr[column] c_result
-
-    with nogil:
-        c_result = move(cpp_filling.fill(
-            c_destination,
-            c_begin,
-            c_end,
-            c_value[0]
-        ))
-
-    return Column.from_unique_ptr(move(c_result))
+    return Column.from_pylibcudf(
+        pylibcudf.filling.fill(
+            destination.to_pylibcudf(mode='read'),
+            begin,
+            end,
+            (<DeviceScalar> as_device_scalar(value)).c_value
+        )
+    )
 
 
 @acquire_spill_lock()
 def repeat(list inp, object count):
+    ctbl = pylibcudf.Table([col.to_pylibcudf(mode="read") for col in inp])
     if isinstance(count, Column):
-        return _repeat_via_column(inp, count)
-    else:
-        return _repeat_via_size_type(inp, count)
-
-
-def _repeat_via_column(list inp, Column count):
-    cdef table_view c_inp = table_view_from_columns(inp)
-    cdef column_view c_count = count.view()
-    cdef unique_ptr[table] c_result
-
-    with nogil:
-        c_result = move(cpp_filling.repeat(
-            c_inp,
-            c_count,
-        ))
-
-    return columns_from_unique_ptr(move(c_result))
-
-
-def _repeat_via_size_type(list inp, size_type count):
-    cdef table_view c_inp = table_view_from_columns(inp)
-    cdef unique_ptr[table] c_result
-
-    with nogil:
-        c_result = move(cpp_filling.repeat(
-            c_inp,
+        count = count.to_pylibcudf(mode="read")
+    return columns_from_pylibcudf_table(
+        pylibcudf.filling.repeat(
+            ctbl,
             count
-        ))
-
-    return columns_from_unique_ptr(move(c_result))
+        )
+    )
 
 
 @acquire_spill_lock()
 def sequence(int size, DeviceScalar init, DeviceScalar step):
-    cdef size_type c_size = size
-    cdef const scalar* c_init = init.get_raw_ptr()
-    cdef const scalar* c_step = step.get_raw_ptr()
-    cdef unique_ptr[column] c_result
-
-    with nogil:
-        c_result = move(cpp_filling.sequence(
-            c_size,
-            c_init[0],
-            c_step[0]
-        ))
-
-    return Column.from_unique_ptr(move(c_result))
+    return Column.from_pylibcudf(
+        pylibcudf.filling.sequence(
+            size,
+            (<DeviceScalar> as_device_scalar(init)).c_value,
+            (<DeviceScalar> as_device_scalar(step)).c_value
+        )
+    )

python/cudf/cudf/_lib/pylibcudf/CMakeLists.txt

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

python/cudf/cudf/_lib/pylibcudf/__init__.pxd

@@ -6,6 +6,7 @@ from . cimport (
     binaryop,
     concatenate,
     copying,
+    filling,
     groupby,
     interop,
     join,
@@ -37,6 +38,7 @@ __all__ = [
     "binaryop",
     "concatenate",
     "copying",
+    "filling",
     "gpumemoryview",
     "groupby",
     "interop",

python/cudf/cudf/_lib/pylibcudf/__init__.py

@@ -5,6 +5,7 @@
     binaryop,
     concatenate,
     copying,
+    filling,
     groupby,
     interop,
     join,
@@ -35,6 +36,7 @@
     "binaryop",
     "concatenate",
     "copying",
+    "filling",
     "gpumemoryview",
     "groupby",
     "interop",

python/cudf/cudf/_lib/pylibcudf/filling.pxd

@@ -0,0 +1,35 @@
+# Copyright (c) 2024, NVIDIA CORPORATION.
+from cudf._lib.cpp.types cimport size_type
+
+from .column cimport Column
+from .scalar cimport Scalar
+from .table cimport Table
+
+ctypedef fused ColumnOrSize:
+    Column
+    size_type
+
+cpdef Column fill(
+    Column destination,
+    size_type begin,
+    size_type end,
+    Scalar value,
+)
+
+cpdef void fill_in_place(
+    Column destination,
+    size_type c_begin,
+    size_type c_end,
+    Scalar value,
+)
+
+cpdef Column sequence(
+    size_type size,
+    Scalar init,
+    Scalar step,
+)
+
+cpdef Table repeat(
+    Table input_table,
+    ColumnOrSize count
+)