Support args= in Series.apply

python/cudf/cudf/core/dataframe.py

@@ -63,6 +63,7 @@
 from cudf.core.multiindex import MultiIndex
 from cudf.core.resample import DataFrameResampler
 from cudf.core.series import Series
+from cudf.core.udf.row_function import _get_row_kernel
 from cudf.utils import applyutils, docutils, ioutils, queryutils, utils
 from cudf.utils.docutils import copy_docstring
 from cudf.utils.dtypes import (
@@ -3926,10 +3927,8 @@ def apply(
             raise ValueError("The `raw` kwarg is not yet supported.")
         if result_type is not None:
             raise ValueError("The `result_type` kwarg is not yet supported.")
-        if kwargs:
-            raise ValueError("UDFs using **kwargs are not yet supported.")
 
-        return self._apply(func, *args)
+        return self._apply(func, _get_row_kernel, *args, **kwargs)
 
     @applyutils.doc_apply()
     def apply_rows(

python/cudf/cudf/core/frame.py

@@ -45,7 +45,6 @@
 )
 from cudf.core.column_accessor import ColumnAccessor
 from cudf.core.join import Merge, MergeSemi
-from cudf.core.udf.pipeline import compile_or_get, supported_cols_from_frame
 from cudf.core.window import Rolling
 from cudf.utils import ioutils
 from cudf.utils.docutils import copy_docstring
@@ -1367,39 +1366,6 @@ def _quantiles(
         result._copy_type_metadata(self)
         return result
 
-    @annotate("APPLY", color="purple", domain="cudf_python")
-    def _apply(self, func, *args):
-        """
-        Apply `func` across the rows of the frame.
-        """
-        kernel, retty = compile_or_get(self, func, args)
-
-        # Mask and data column preallocated
-        ans_col = cupy.empty(len(self), dtype=retty)
-        ans_mask = cudf.core.column.column_empty(len(self), dtype="bool")
-        launch_args = [(ans_col, ans_mask), len(self)]
-        offsets = []
-
-        # if compile_or_get succeeds, it is safe to create a kernel that only
-        # consumes the columns that are of supported dtype
-        for col in supported_cols_from_frame(self).values():
-            data = col.data
-            mask = col.mask
-            if mask is None:
-                launch_args.append(data)
-            else:
-                launch_args.append((data, mask))
-            offsets.append(col.offset)
-        launch_args += offsets
-        launch_args += list(args)
-        kernel.forall(len(self))(*launch_args)
-
-        col = as_column(ans_col)
-        col.set_base_mask(libcudf.transform.bools_to_mask(ans_mask))
-        result = cudf.Series._from_data({None: col}, self._index)
-
-        return result
-
     def rank(
         self,
         axis=0,

python/cudf/cudf/core/indexed_frame.py

@@ -24,11 +24,12 @@
     is_integer_dtype,
     is_list_like,
 )
-from cudf.core.column import arange
+from cudf.core.column import arange, as_column
 from cudf.core.column_accessor import ColumnAccessor
 from cudf.core.frame import Frame
 from cudf.core.index import Index, RangeIndex, _index_from_columns
 from cudf.core.multiindex import MultiIndex
+from cudf.core.udf.utils import _compile_or_get, _supported_cols_from_frame
 from cudf.utils.utils import cached_property
 
 doc_reset_index_template = """
@@ -756,6 +757,51 @@ def add_suffix(self, suffix):
                 Use `Series.add_suffix` or `DataFrame.add_suffix`"
         )
 
+    @annotate("APPLY", color="purple", domain="cudf_python")
+    def _apply(self, func, kernel_getter, *args, **kwargs):
+        """Apply `func` across the rows of the frame."""
+        if kwargs:
+            raise ValueError("UDFs using **kwargs are not yet supported.")
+
+        try:
+            kernel, retty = _compile_or_get(
+                self, func, args, kernel_getter=kernel_getter
+            )
+        except Exception as e:
+            raise ValueError(
+                "user defined function compilation failed."
+            ) from e
+
+        # Mask and data column preallocated
+        ans_col = cp.empty(len(self), dtype=retty)
+        ans_mask = cudf.core.column.column_empty(len(self), dtype="bool")
+        launch_args = [(ans_col, ans_mask), len(self)]
+        offsets = []
+
+        # if _compile_or_get succeeds, it is safe to create a kernel that only
+        # consumes the columns that are of supported dtype
+        for col in _supported_cols_from_frame(self).values():
+            data = col.data
+            mask = col.mask
+            if mask is None:
+                launch_args.append(data)
+            else:
+                launch_args.append((data, mask))
+            offsets.append(col.offset)
+        launch_args += offsets
+        launch_args += list(args)
+
+        try:
+            kernel.forall(len(self))(*launch_args)
+        except Exception as e:
+            raise RuntimeError("UDF kernel execution failed.") from e
+
+        col = as_column(ans_col)
+        col.set_base_mask(libcudf.transform.bools_to_mask(ans_mask))
+        result = cudf.Series._from_data({None: col}, self._index)
+
+        return result
+
     def sort_values(
         self,
         by,

python/cudf/cudf/core/series.py

@@ -14,7 +14,6 @@
 import cupy
 import numpy as np
 import pandas as pd
-from numba import cuda
 from pandas._config import get_option
 
 import cudf
@@ -67,6 +66,7 @@
     doc_reset_index_template,
 )
 from cudf.core.single_column_frame import SingleColumnFrame
+from cudf.core.udf.scalar_function import _get_scalar_kernel
 from cudf.utils import cudautils, docutils
 from cudf.utils.docutils import copy_docstring
 from cudf.utils.dtypes import (
@@ -2374,7 +2374,7 @@ def apply(self, func, convert_dtype=True, args=(), **kwargs):
             by numba based on the function logic and argument types.
             See examples for details.
         args : tuple
-            Not supported
+            Positional arguments passed to func after the series value.
         **kwargs
             Not supported
 
@@ -2440,20 +2440,9 @@ def apply(self, func, convert_dtype=True, args=(), **kwargs):
         2     4.5
         dtype: float64
         """
-        if args or kwargs:
-            raise ValueError(
-                "UDFs using *args or **kwargs are not yet supported."
-            )
-
-        # these functions are generally written as functions of scalar
-        # values rather than rows. Rather than writing an entirely separate
-        # numba kernel that is not built around a row object, its simpler
-        # to just turn this into the equivalent single column dataframe case
-        name = self.name or "__temp_srname"
-        df = cudf.DataFrame({name: self})
-        f_ = cuda.jit(device=True)(func)
-
-        return df.apply(lambda row: f_(row[name]))
+        if convert_dtype is not True:
+            raise ValueError("Series.apply only supports convert_dtype=True")
+        return self._apply(func, _get_scalar_kernel, *args, **kwargs)
 
     def applymap(self, udf, out_dtype=None):
         """Apply an elementwise function to transform the values in the Column.