Get item

Transformation to wrap around the pyspark getItem function

koheesio.spark.transformations.get_item.GetItem #

Get item from list or map (dictionary)

Wrapper around pyspark.sql.functions.getItem

GetItem is strict about the data type of the column. If the column is not a list or a map, an error will be raised.

Note

Only MapType and ArrayType are supported.

Parameters:

Name	Type	Description	Default
`columns`	`Union[str, List[str]]`	The column (or list of columns) to get the item from. Alias: column	required
`target_column`	`Optional[str]`	The column to store the result in. If not provided, the result will be stored in the source column. Alias: target_suffix - if multiple columns are given as source, this will be used as a suffix.	`None`
`key`	`Union[int, str]`	The key (or index) to get from the list or map. If the column is a list (ArrayType), this should be an integer. If the column is a dict (MapType), this should be a string. Alias: index	required

Example

Example with list (ArrayType)#

By specifying an integer for the parameter "key", getItem knows to get the element at index n of a list (index starts at 0).

input_df:

id	content
1	[1, 2, 3]
2	[4, 5]
3	[6]
4	[]

output_df = GetItem(
    column="content",
    index=1,  # get the second element of the list
    target_column="item",
).transform(input_df)

output_df:

id	content	item
1	[1, 2, 3]	2
2	[4, 5]	5
3	[6]	null
4	[]	null

Example with a dict (MapType)#

input_df:

id	content
1	{key1 -> value1}
2	{key1 -> value2}
3	{key2 -> hello}
4	{key2 -> world}

output_df = GetItem(
    column= "content",
    key="key2,
    target_column="item",
).transform(input_df)

As we request the key to be "key2", the first 2 rows will be null, because it does not have "key2".

output_df:

id	content	item
1	{key1 -> value1}	null
2	{key1 -> value2}	null
3	{key2 -> hello}	hello
4	{key2 -> world}	world

key `class-attribute` `instance-attribute` #

key: Union[int, str] = Field(
    default=...,
    alias="index",
    description="The key (or index) to get from the list or map. If the column is a list (ArrayType), this should be an integer. If the column is a dict (MapType), this should be a string. Alias: index",
)

ColumnConfig #

Limit the data types to ArrayType and MapType.

data_type_strict_mode `class-attribute` `instance-attribute` #

data_type_strict_mode = True

limit_data_type `class-attribute` `instance-attribute` #

limit_data_type = run_for_all_data_type

run_for_all_data_type `class-attribute` `instance-attribute` #

run_for_all_data_type = [ARRAY, MAP]

func #

func(column: Column) -> Column

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

def func(self, column: Column) -> Column:
    return get_item(column, self.key)

koheesio.spark.transformations.get_item.get_item #

get_item(column: Column, key: Union[str, int]) -> Column

Wrapper around pyspark.sql.functions.getItem