Skip to content

This project contains the code to translate between Apache Spark and SFrame.

License

Notifications You must be signed in to change notification settings

kuruv/spark-sframe

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Spark Unity Codebase

This project contains the code to interact with the SFrame open-source project from within Apache Spark. Currently, the jar created by this project is included in the GraphLab Create python egg to enable translation between Apache Spark Dataframes and GraphLab Create SFrames. Users can also use this project in the scala spark shell to export Dataframes as SFrames.

Soft Dependency

The spark-sframe package leverages multiple shared libraries: cy_callback.so, cy_cpp_utils.so, cy_flexible_type.so, cy_spark_unity.so which are directly built from the open-source SFrame package.

Building spark-sframe

To build the spark-unity.jar all you need is to have the java jdk installed on your platform and run our pre-bundled mvn:

cd spark-sframe
build/mvn package

This will both test and build the spark-unity.jar on your platform.

PySpark Integration

Setup the environment

To use GraphLab Create within PySpark, you need to set the $SPARK_HOME and $PYTHONPATH environment variables on the driver. A common usage:

export PYTHONPATH=$SPARK_HOME/python/:$SPARK_HOME/python/lib/py4j-0.8.2.1-src.zip:$PYTHONPATH
export SPARK_HOME =<your-spark-home-dir>

Run from the PySpark Python Shell

cd $SPARK_HOME
bin/pyspark

Run from a standard Python Shell

Make sure you have exported the PYTHONPATH and SPARK_HOME environment variables. Then run (for example):

ipython

Then you need to start spark:

from pyspark import SparkContext
from pyspark.sql import SQLContext
# Launch spark by creating a spark context
sc = SparkContext()
# Create a SparkSQL context to manage dataframe schema information.
sql = SQLContext(sc)

Make an SFrame from an RDD

from sframe import SFrame
rdd = sc.parallelize([(x,str(x), "hello") for x in range(0,5)])
sframe = SFrame.from_rdd(rdd, sc)
print sframe

Make an SFrame from a Dataframe (preferred)

from sframe import SFrame
rdd = sc.parallelize([(x,str(x), "hello") for x in range(0,5)])
df = sql.createDataFrame(rdd)
sframe = SFrame.from_rdd(df, sc)
print sframe

Standalone Integration

Run Spark Shell

cd $SPARK_HOME
bin/spark-shell --jars spark-sframe/target/spark_unity-0.1.jar

Make an SFrame from an RDD

import org.graphlab.create.GraphLabUtil
var df = sc.parallelize(Array(1,2,3)).toDF // Must be a dataframe
val outputDir = "/tmp/graphlab_testing" // Must be an HDFS path unless running in local mode
val prefix = "test"
val sframeFileName = GraphLabUtil.toSFrame(df, outputDir, prefix)
println(sframeFileName)

Make an RDD from an SFrame

import org.graphlab.create.GraphLabUtil
val newRDD = GraphLabUtil.toRDD(sc, "/tmp/graphlab_testing/test.frame_idx")

Requirements and Caveats

The currently release requires Python 2.7, Spark 1.3 or later, and the hadoop binary must be within the PATH of the driver when running on a cluster or interacting with Hadoop (e.g., you should be able to run hadoop classpath).

We also currently only support Mac and Linux platforms but will have Windows support soon.

Recommended Settings for Spark Installation on a Cluster

We recommend downloading Pre-built for Hadoop 2.4 and later version of Apache Spark.

About

This project contains the code to translate between Apache Spark and SFrame.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Scala 87.9%
  • Shell 11.5%
  • Python 0.6%