ParentWithBackslash.scala

import org.apache.spark.SparkContext
import org.apache.spark.SparkContext._
import org.apache.spark.SparkConf
import org.elasticsearch.spark._
import org.elasticsearch.spark.rdd.EsSpark 

object ParentWithBackslash { 

  	def main(args: Array[String]) {

	    val conf = new SparkConf()
	    conf.set("es.nodes", "localhost:9200")
	    conf.set("es.mapping.parent","parentId")
	    
	    val sc = new SparkContext(conf)
		    
	    val simpleDocWithParent = Map("itemId"->"1", "parentId"->"""a\s""")

	    val docs = sc.makeRDD(Seq(simpleDocWithParent))

		docs.saveToEs("sparkdemo2/revenue", Map("es.mapping.id" -> "itemId"))		
  	}

}

Just tried this in master and works correctly:

val doc = Map("itemId" -> "1", "parent" -> """a\s""")
sc.makeRDD(Seq(doc)).saveToEs("spark-test/escaped-char", Map("es.mapping.id" -> "itemId"))

Tx [PUT]@[192.168.1.50:9500][spark-test/escaped-char/_bulk] w/ payload [{"index":{"_id":"1"}}
{"itemId":"1","parent":"a\\s"}
]

It's not the writing of the field value in the source that's failing, it's the setting of the parent Id (i.e. parent child).

val doc = Map("itemId" -> "1", "parent" -> """a\s""")
sc.makeRDD(Seq(doc)).saveToEs("spark-test/escaped-char", Map("es.mapping.id" -> "itemId", "es.mapping.parent" -> "parent"))

I've managed to get the case class version using Metadata/Source PairRdd to work though. I'm setting both ID and PARENT in the metadata Map. Still can't get the json or plain old Map versions to work though.

Incidentally, the parent mapping pre-exists in my elasticsearch index.