[SPARK-11199] Improve R context management story and add getOrCreate

• Type: Sub-task
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.5.1
• Fix Version/s: 2.0.0
• Component/s: SparkR
• Assignee: Hossein Falaki
• Reporter: Felix Cheung
• Created: 20/Oct/15 00:51
• Updated: 29/Dec/15 19:44
• Resolved: 29/Dec/15 19:44

Description

Similar to SPARK-11114

Also from discussion in SPARK-10903:
"
Hossein Falaki added a comment - 08/Oct/15 13:06
+1 We have seen a lot of questions from new SparkR users about the life cycle of the context.
My question is: are we going to remove or deprecate sparkRSQL.init()? I suggest we should, because right now calling that method creates a new Java SQLContext object, and having two of them prevents users form viewing temp tables.

Felix Cheung added a comment - 08/Oct/15 17:13
+1 perhaps sparkR.init() should create sqlContext and/or hiveCtx together.
But Hossein Falaki, as of now calling sparkRSQL.init() should return the same one as you can see https://github.com/apache/spark/blob/master/R/pkg/R/sparkR.R#L224

Hossein Falaki added a comment - 08/Oct/15 17:16
I meant the SQL Context: https://github.com/apache/spark/blob/master/R/pkg/R/sparkR.R#L236
This call should have been "getOrCreate."
"

Issue Links

• links to: [Github] Pull Request #9185 (falaki)