Skip to content

Latest commit

 

History

History
44 lines (32 loc) · 1.59 KB

source.md

File metadata and controls

44 lines (32 loc) · 1.59 KB

Document Traits

Sometimes it is useful to seperate the knowledge of the type from the indexing logic. For this we can use the DocumentSource or DocumentMap abstraction to encapsulate the conversion logic.

These are simply traits that you can mix into your classes which provide a method that elastic4s can use to populate an index request.

For example. using DocumentMap is as simple as defining a class like this.

case class Band(name:String, albums:Seq[String], label:String) extends DocumentMap {
  def map = Map("name" -> name, "albums"->albums.mkString(" "), "label" -> label)
}

And then using that class with doc in an index request.

case class Band(name: String, albums: Seq[String], label: String)
val band = Band("coldplay", Seq("X&Y", "Parachutes"), "Parlophone")

client.execute {
  // the band object will be implicitly converted into a DocumentSource
  index into "music/bands" doc band
}

DocumentSource is similar except you return a JSON string rather than a Scala Map. Use whatever you prefer.

If you want to index directly from a Jackson JSON object then you can use the built in JacksonSource wrapper.

val myJsonDoc = ... // some jackson object
client.execute { index into "electronics/phones" doc JacksonSource(myJsonDoc) }

Or you can even index plain objects and elastic4s will use Jackson to marshall into JSON. This uses the Scala extension in Jackson and so supports scala collections, options, etc.

val anyOldObject = ... // anything that extends from AnyRef
client.execute { index into "electronics/phones" doc anyOldObject }