Transform

Transform module

Transform aims to provide an easy interface for calling transformations on a Spark DataFrame, where the transformation is a function that accepts a DataFrame (df) and any number of keyword args.

koheesio.spark.transformations.transform.Transform

Transform(func: Callable, params: Dict = None, df: DataFrame = None, **kwargs)

Transform aims to provide an easy interface for calling transformations on a Spark DataFrame, where the transformation is a function that accepts a DataFrame (df) and any number of keyword args.

The implementation is inspired by and based upon: https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.DataFrame.transform.html

Parameters:

Name	Type	Description	Default
func	Callable	The function to be called on the DataFrame.	required
params	Dict	The keyword arguments to be passed to the function. Defaults to None. Alternatively, keyword arguments can be passed directly as keyword arguments - they will be merged with the params dictionary.	None

Example

a function compatible with Transform:

def some_func(df, a: str, b: str):
    return df.withColumn(a, f.lit(b))

verbose style input in Transform

Transform(func=some_func, params={"a": "foo", "b": "bar"})

shortened style notation (easier to read)

Transform(some_func, a="foo", b="bar")

when too much input is given, Transform will ignore extra input

Transform(
    some_func,
    a="foo",
    # ignored input
    c="baz",
    title=42,
    author="Adams",
    # order of params input should not matter
    b="bar",
)

using the from_func classmethod

SomeFunc = Transform.from_func(some_func, a="foo")
some_func = SomeFunc(b="bar")

Source code in src/koheesio/spark/transformations/transform.py

def __init__(self, func: Callable, params: Dict = None, df: DataFrame = None, **kwargs):
    params = {**(params or {}), **kwargs}
    super().__init__(func=func, params=params, df=df)

func class-attribute instance-attribute

func: Callable = Field(default=None, description='The function to be called on the DataFrame.')

execute

execute()

Call the function on the DataFrame with the given keyword arguments.

Source code in src/koheesio/spark/transformations/transform.py

def execute(self):
    """Call the function on the DataFrame with the given keyword arguments."""
    func, kwargs = get_args_for_func(self.func, self.params)
    self.output.df = self.df.transform(func=func, **kwargs)

from_func classmethod

from_func(func: Callable, **kwargs) -> Callable[..., Transform]

Create a Transform class from a function. Useful for creating a new class with a different name.

This method uses the functools.partial function to create a new class with the given function and keyword arguments. This way you can pre-define some of the keyword arguments for the function that might be needed for the specific use case.

Example

CustomTransform = Transform.from_func(some_func, a="foo")
some_func = CustomTransform(b="bar")

In this example, CustomTransform is a Transform class with the function some_func and the keyword argument a set to "foo". When calling some_func(b="bar"), the function some_func will be called with the keyword arguments a="foo" and b="bar".

Source code in src/koheesio/spark/transformations/transform.py

@classmethod
def from_func(cls, func: Callable, **kwargs) -> Callable[..., Transform]:
    """Create a Transform class from a function. Useful for creating a new class with a different name.

    This method uses the `functools.partial` function to create a new class with the given function and keyword
    arguments. This way you can pre-define some of the keyword arguments for the function that might be needed for
    the specific use case.

    Example
    -------
    ```python
    CustomTransform = Transform.from_func(some_func, a="foo")
    some_func = CustomTransform(b="bar")
    ```

    In this example, `CustomTransform` is a Transform class with the function `some_func` and the keyword argument
    `a` set to "foo". When calling `some_func(b="bar")`, the function `some_func` will be called with the keyword
    arguments `a="foo"` and `b="bar"`.
    """
    return partial(cls, func=func, **kwargs)