[CRUNCH-507] Potential NPE in SparkPipeline constructor and additional constructor

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.11.0
• Fix Version/s: 0.12.0
• Component/s: Core
• Assignee: Micah Whitacre
• Reporter: Micah Whitacre
• Created: 07/Apr/15 02:14
• Updated: 18/May/15 19:03
• Resolved: 07/May/15 01:16

Description

Was looking at the SparkPipeline constructor API and was trying to maximize the number of settings I'd inherit when a Spark job was submitted with "spark-submit". This should populate the SparkContext (and JavaSparkContext) with values like the Spark Master. If you want to:

• Specify a driver class
• Hadoop Configuration (vs picking up the defaults)
• Inherit pre-populated SparkContext you'd have to use a constructor like:

JavaSparkContext sc = new JavaSparkContext(new SparkConf);
new SparkPipeline(sc.master(), sc.appName(), Driver.class, conf)

Just for convenience we could add a constructor like the following:

public SparkPipeline(JavaSparkContext sc, String appName, Class driver, Configuration conf)

Could remove the appName but since the spark context is not guaranteed to be non-null we might get a NPE. This also means that on this line[1] we could throw an NPE when trying to pull the hadoopConfiguration() off that object.

[1] - https://github.com/apache/crunch/blob/3ab0b078c47f23b3ba893fdfb05fd723f663d02b/crunch-spark/src/main/java/org/apache/crunch/impl/spark/SparkPipeline.java#L73