itholic commented on a change in pull request #33858:
URL: https://github.com/apache/spark/pull/33858#discussion_r704045118
##########
File path: python/pyspark/pandas/series.py
##########
@@ -4475,6 +4477,146 @@ def replace(
return self._with_new_scol(current) # TODO: dtype?
+ def combine(
+ self,
+ other: "Series",
+ func: Callable,
+ fill_value: Optional[Any] = None,
+ return_type: Union[Union[AtomicType, str], ArrayType] = "string",
+ ) -> "Series":
+ """
+ Combine the Series with a Series or scalar according to `func`.
+ Combine the Series and `other` using `func` to perform elementwise
+ selection for combined Series.
+ `fill_value` is assumed when value is missing at some index
+ from one of the two objects being combined.
+
+ .. versionadded:: 3.3.0
+
+ Parameters
+ ----------
+ other : Series or scalar
+ The value(s) to be combined with the `Series`.
+ func : function
+ Function that takes two scalars as inputs and returns an element.
+ fill_value : scalar, optional
+ The value to assume when an index is missing from
+ one Series or the other. The default specifies to use the
+ appropriate NaN value for the underlying dtype of the Series.
+ return_type : :class:`pyspark.sql.types.DataType` or str
+ the return type of the output Series. The value can be either a
+ :class:`pyspark.sql.types.DataType` object or a DDL-formatted type
string.
+
+ Returns
+ -------
+ Series
+ The result of combining the Series with the other object.
+
+ See Also
+ --------
+ Series.combine_first : Combine Series values, choosing the calling
+ Series' values first.
+
+ Examples
+ --------
+ Consider 2 Datasets ``s1`` and ``s2`` containing
+ highest clocked speeds of different birds.
+
+ >>> from pyspark.pandas.config import set_option, reset_option
+ >>> set_option("compute.ops_on_diff_frames", True)
+ >>> s1 = ps.Series({'falcon': 330.0, 'eagle': 160.0})
+ >>> s1
+ falcon 330.0
+ eagle 160.0
+ dtype: float64
+ >>> s2 = ps.Series({'falcon': 345.0, 'eagle': 200.0, 'duck': 30.0})
+ >>> s2
+ falcon 345.0
+ eagle 200.0
+ duck 30.0
+ dtype: float64
+
+ Now, to combine the two datasets and view the highest speeds
+ of the birds across the two datasets
+
+ >>> s1.combine(s2, max, return_type="double")
+ duck NaN
+ eagle 200.0
+ falcon 345.0
+ dtype: float64
+
+ In the previous example, the resulting value for duck is missing,
+ because the maximum of a NaN and a float is a NaN.
+ So, in the example, we set ``fill_value=0``,
+ so the maximum value returned will be the value from some dataset.
+
+ >>> s1.combine(s2, max, fill_value=0, return_type="double")
+ duck 30.0
+ eagle 200.0
+ falcon 345.0
+ dtype: float64
+ >>> reset_option("compute.ops_on_diff_frames")
+ """
+ if not isinstance(other, Series) and not np.isscalar(other):
+ raise TypeError("unsupported type: %s" % type(other))
+
+ tmp_other_col = verify_temp_column_name(self._internal.spark_frame,
"__tmp_other_col__")
+ if np.isscalar(other):
+ combined = DataFrame(
+ InternalFrame(
+ spark_frame=self._internal.spark_frame.select(
+ *self._internal.spark_columns
+ ).withColumn(tmp_other_col, SF.lit(other)),
+ index_spark_columns=self._internal.index_spark_columns,
+ )
+ )
+ elif same_anchor(self, other):
+ combined = self._psdf[self._column_label, other._column_label]
+ elif fill_value is None:
+ combined = combine_frames(self.to_frame(), other.to_frame())
+ else:
+ combined = self._combine_frame_with_fill_value(other,
fill_value=fill_value)
+
+ @pandas_udf(returnType=return_type) # type: ignore
+ def wrapped_func(x: pd.Series, y: pd.Series) -> pd.Series:
+ return x.combine(y, func)
+
+ internal = InternalFrame(
+ spark_frame=combined._internal.spark_frame.select(
+ *combined._internal.index_spark_columns,
+ wrapped_func(*combined._internal.data_spark_columns),
+ NATURAL_ORDER_COLUMN_NAME
+ ),
+ index_spark_columns=combined._internal.index_spark_columns,
+ column_labels=self._internal.column_labels if np.isscalar(other)
else [None],
+ column_label_names=[None],
+ )
Review comment:
Oh, yeah it's my mistake.
I think we should get a return type from the function, not `self`.
Let's try to use the `infer_return_type`.
For example,
```python
return_type = infer_return_type(func).spark_type
@pandas_udf(returnType=return_type) # type: ignore
def wrapped_func(x: pd.Series, y: pd.Series) -> pd.Series:
return x.combine(y, func)
```
Also, can you refer the implementation for `Series.apply` ??
I think that might be a hint for this implementation.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
