The driver automatically infers a relational schema by inspecting a series of MongoDB documents in a collection. You can use the RowScanDepth property to define the number of documents the driver will scan to do so. The columns identified during the discovery process depend on the FlattenArrays and FlattenObjects properties.

Flattening Objects

If FlattenObjects is set, all nested objects will be flattened into a series of columns. For example, consider the following document:

{
  id: 12,
  name: "Lohia Manufacturers Inc.",
  address: {street: "Main Street", city: "Chapel Hill", state: "NC"},
  offices: ["Chapel Hill", "London", "New York"],
  annual_revenue: 35,600,000
}

This document will be represented by the following columns:

Column Name	Data Type	Example Value
id	Integer	12
name	String	Lohia Manufacturers Inc.
address.street	String	Main Street
address.city	String	Chapel Hill
address.state	String	NC
offices	String	["Chapel Hill", "London", "New York"]
annual_revenue	Double	35,600,000

If FlattenObjects is not set, then the address.street, address.city, and address.state columns will not be broken apart. The address column of type string will instead represent the entire object. Its value would be {street: "Main Street", city: "Chapel Hill", state: "NC"}. See JSON Functions for more details on working with JSON aggregates.

Flattening Arrays

The FlattenArrays property can be used to flatten array values into columns of their own. This is only recommended for arrays that are expected to be short, for example the coordinates below:

"coord": [ -73.856077, 40.848447 ]

The FlattenArrays property can be set to 2 to represent the array above as follows:

Column Name	Data Type	Example Value
coord.0	Float	-73.856077
coord.1	Float	40.848447

It is best to leave other unbounded arrays as they are and piece out the data for them as needed using JSON Functions.