Skip to content

Latest commit

 

History

History
106 lines (79 loc) · 3.52 KB

15_python.md

File metadata and controls

106 lines (79 loc) · 3.52 KB
Raw

Documentation

PySpark with Data Frames

With the inclusion of the Cassandra Data Source, PySpark can now be used with the Connector to access Cassandra data. This does not require DataStax Enterprise but you are limited to DataFrame only operations.

Setup

To enable Cassandra access the Spark Cassandra Connector assembly jar must be included on both the driver and executor classpath for the PySpark Java Gateway. This can be done by starting the PySpark shell similarly to how the spark shell is started. The preferred method is now to use the maven artifact.

./bin/pyspark \
  --packages com.datastax.spark:spark-cassandra-connector_2.12:3.5.1 \
  --conf spark.sql.extensions=com.datastax.spark.connector.CassandraSparkExtensions

Catalogs

Spark allows you to manipulate external data with and without a Catalog. For a short intro and more details about Catalogs see Quick Start and Data Frames.

Loading a DataFrame

Loading a data set with DatasourceV2 requires creating a Catalog Reference to your Cassandra Cluster.

spark.conf.set("spark.sql.catalog.myCatalog", "com.datastax.spark.connector.datasource.CassandraCatalog")
spark.read.table("myCatalog.myKs.myTab").show()

Saving a DataFrame to Cassandra

A DataFrame can be saved to an existing Cassandra table by using the the saveAsTable method with a catalog, keyspace and a table name specified.

spark.range(1, 10)\
    .selectExpr("id as k")\
    .write\
    .mode("append")\
    .partitionBy("k")\
    .saveAsTable("myCatalog.myKs.myTab")

Manipulating data without a Catalog

Loading a DataFrame

A DataFrame can be created which links to Cassandra by using the the org.apache.spark.sql.cassandra source and by specifying keyword arguments for keyspace and table.

Example Loading a Cassandra Table as a Pyspark DataFrame

 spark.read\
    .format("org.apache.spark.sql.cassandra")\
    .options(table="kv", keyspace="test")\
    .load().show()
+-+-+
|k|v|
+-+-+
|5|5|
|1|1|
|2|2|
|4|4|
|3|3|
+-+-+

Saving a DataFrame to Cassandra

A DataFrame can be saved to an existing Cassandra table by using the the org.apache.spark.sql.cassandra source and by specifying keyword arguments for keyspace and table and saving mode (append, overwrite, error or ignore, see Data Sources API doc).

Example Saving to a Cassandra Table as a Pyspark DataFrame
 df.write\
    .format("org.apache.spark.sql.cassandra")\
    .mode('append')\
    .options(table="kv", keyspace="test")\
    .save()

The options and parameters are identical to the Scala Data Frames Api so please see Data Frames for more information.

Passing options with periods to the DataFrameReader

Python does not support using periods(".") in variable names. This makes it slightly more difficult to pass SCC options to the DataFrameReader. The options function takes kwargs** which means you can't directly pass in keys. There is a workaround though. Python allows you to pass a dictionary as a representation of kwargs and dictionaries can have keys with periods.

Example of using a dictionary as kwargs

load_options = { "table": "kv", "keyspace": "test", "spark.cassandra.input.split.size_in_mb": "10"}
spark.read.format("org.apache.spark.sql.cassandra").options(**load_options).load().show()

Next - Spark Partitioners