The Elasticsearch Adapter can be configured to create a relational model of the data, treating nested documents as individual tables containing a primary key and a foreign key that links to the parent document. This is particularly useful if you need to work with your Elasticsearch data in existing BI, reporting, and ETL tools that expect a relational data model.

Joining Nested Arrays as Tables

With DataModel set to "Relational", any JOINs are controlled by the query. Any time you perform a JOIN query, the Elasticsearch index will be queried once for each table (nested document) included in the query.

Example

Below is a sample query against the sample document in Raw Data, using a relational model.

Query

The following query explicitly JOINs the insured and vehiclestables.

SELECT 
  "insured"."_id", 
  "insured"."name", 
  "insured"."address.street" AS address_street, 
  "insured"."address.city.first" AS address_city, 
  "insured"."address.state.last" AS address_state, 
  "insured"."insured_ages", 
  "vehicles"."year", 
  "vehicles"."make", 
  "vehicles"."model", 
  "vehicles"."body_style",
  "vehicles"."_insured_id",
  "vehicles"."_c_id"
FROM 
  "insured"
JOIN 
  "vehicles" 
ON 
  "insured"."_id" = "vehicles"."_insured_id"

Results

In the example query, each vehicle document is JOINed to its parent insured object to produce a table with 5 rows.

See Also

• Automatic Schema Discovery: Configure the columns reported in the table schemas.
• FreeForm;: Use dot notation to select nested data.
• VerticalFlattening;: Access nested object arrays as separate tables.
• JSON Functions: Manipulate the data returned to perform client-side aggregation and transformations.