DataSourceRegistration

A wrapper for data source registration.

This instance can be accessed via spark.dataSource. Use it to register a custom DataSource subclass so it can be referenced by name in spark.read.format() and df.write.format().

Syntax

Python

spark.dataSource.register(MyDataSource)

Methods

Method	Description
register(dataSource)	Registers a Python user-defined data source. dataSource must be a subclass of DataSource.

Examples

Register a custom data source and read from it:

Python

from pyspark.sql.datasource import DataSource, DataSourceReader

class MyDataSource(DataSource):
    @classmethod
    def name(cls):
        return "my_data_source"

    def schema(self):
        return "id INT, value STRING"

    def reader(self, schema):
        return MyDataSourceReader(schema)

class MyDataSourceReader(DataSourceReader):
    def read(self, partition):
        yield (1, "hello")
        yield (2, "world")

spark.dataSource.register(MyDataSource)
df = spark.read.format("my_data_source").load()
df.show()