Python Programming Guide

The Spark Python API (PySpark) exposes most of the Spark features available in the Scala version to Python. To learn the basics of Spark, we recommend reading through the Scala programming guide first; it should be easy to follow even if you don’t know Scala. This guide will show how to use the Spark features described there in Python.

Key Differences in the Python API

There are a few key differences between the Python and Scala APIs:

In PySpark, RDDs support the same methods as their Scala counterparts but take Python functions and return Python collection types. Short functions can be passed to RDD methods using Python’s lambda syntax:

logData = sc.textFile(logFile).cache()
errors = logData.filter(lambda line: "ERROR" in line)

You can also pass functions that are defined using the def keyword; this is useful for more complicated functions that cannot be expressed using lambda:

def is_error(line):
    return "ERROR" in line
errors = logData.filter(is_error)

Functions can access objects in enclosing scopes, although modifications to those objects within RDD methods will not be propagated to other tasks:

error_keywords = ["Exception", "Error"]
def is_error(line):
    return any(keyword in line for keyword in error_keywords)
errors = logData.filter(is_error)

PySpark will automatically ship these functions to workers, along with any objects that they reference. Instances of classes will be serialized and shipped to workers by PySpark, but classes themselves cannot be automatically distributed to workers. The Standalone Use section describes how to ship code dependencies to workers.

Installing and Configuring PySpark

PySpark requires Python 2.6 or higher. PySpark jobs are executed using a standard cPython interpreter in order to support Python modules that use C extensions. We have not tested PySpark with Python 3 or with alternative Python interpreters, such as PyPy or Jython. By default, PySpark’s scripts will run programs using python; an alternate Python executable may be specified by setting the PYSPARK_PYTHON environment variable in conf/

All of PySpark’s library dependencies, including Py4J, are bundled with PySpark and automatically imported.

Standalone PySpark jobs should be run using the pyspark script, which automatically configures the Java and Python environment using the settings in conf/ The script automatically adds the pyspark package to the PYTHONPATH.

Interactive Use

The pyspark script launches a Python interpreter that is configured to run PySpark jobs. To use pyspark interactively, first build Spark, then launch it directly from the command line without any options:

$ sbt/sbt package
$ ./pyspark

The Python shell can be used explore data interactively and is a simple way to learn the API:

>>> words = sc.textFile("/usr/share/dict/words")
>>> words.filter(lambda w: w.startswith("spar")).take(5)
[u'spar', u'sparable', u'sparada', u'sparadrap', u'sparagrass']
>>> help(pyspark) # Show all pyspark functions

By default, the pyspark shell creates SparkContext that runs jobs locally. To connect to a non-local cluster, set the MASTER environment variable. For example, to use the pyspark shell with a standalone Spark cluster:

$ MASTER=spark://IP:PORT ./pyspark

Standalone Use

PySpark can also be used from standalone Python scripts by creating a SparkContext in your script and running the script using pyspark. The Quick Start guide includes a complete example of a standalone Python job.

Code dependencies can be deployed by listing them in the pyFiles option in the SparkContext constructor:

from pyspark import SparkContext
sc = SparkContext("local", "Job Name", pyFiles=['', '', 'app.egg'])

Files listed here will be added to the PYTHONPATH and shipped to remote worker machines. Code dependencies can be added to an existing SparkContext using its addPyFile() method.

Where to Go from Here

PySpark includes several sample programs in the python/examples folder. You can run them by passing the files to the pyspark script – for example ./pyspark python/examples/ Each program prints usage help when run without arguments.

We currently provide API documentation for the Python API as Epydoc. Many of the RDD method descriptions contain doctests that provide additional usage examples.