The CData ODBC Driver for Parquet can be configured to create a relational model of the data, treating nested object arrays 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 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 file or source will be queried once for each table (nested array) included in the query.

Example

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

Connecting String

URI=C:\people.parquet;DataModel=Relational;'

Query

The following query explicitly JOINs the people, vehicles, and maintenance tables.

SELECT 
  [people].[personal.age] AS age, 
  [people].[personal.gender] AS gender, 
  [people].[personal.name.first] AS first_name, 
  [people].[personal.name.last] AS last_name, 
  [people].[source], 
  [vehicles].[type], 
  [vehicles].[model], 
  [vehicles].[insurance.company] AS ins_company, 
  [vehicles].[insurance.policy_num] AS ins_policy_num, 
  [maintenance].[date] AS maint_date, 
  [maintenance].[desc] AS maint_desc
FROM 
  [people]
JOIN 
  [vehicles] 
ON 
  [people].[_id] = [vehicles].[people_id]
JOIN 
  [maintenance] 
ON 
  [vehicles].[_id] = [maintenance].[vehicles_id]

Results

In the example query, each maintenance object is JOINed to its parent vehicle object, which is JOINed to its parent people object to produce a table with 8 rows (2 maintenance entries for each of 2 vehicles each for 2 people).