Configure PolyBase to access external data in MongoDB
The article explains how to use PolyBase on a SQL Server instance to query external data in MongoDB.
If you haven't installed PolyBase, see PolyBase installation.
Before creating a database scoped credential a Master Key must be created.
Configure a MongoDB external data source
To query the data from a MongoDB data source, you must create external tables to reference the external data. This section provides sample code to create these external tables.
The following Transact-SQL commands are used in this section:
- CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL)
- CREATE EXTERNAL DATA SOURCE (Transact-SQL)
- CREATE STATISTICS (Transact-SQL)
Create a database scoped credential for accessing the MongoDB source.
/* specify credentials to external data source * IDENTITY: user name for external source. * SECRET: password for external source. */ CREATE DATABASE SCOPED CREDENTIAL credential_name WITH IDENTITY = 'username', Secret = 'password';
The MongoDB ODBC Connector for PolyBase supports only basic authentication, not Kerberos authentication.
Create an external data source with CREATE EXTERNAL DATA SOURCE.
/* LOCATION: Location string should be of format '<type>://<server>[:<port>]'. * PUSHDOWN: specify whether computation should be pushed down to the source. ON by default. *CONNECTION_OPTIONS: Specify driver location * CREDENTIAL: the database scoped credential, created above. */ CREATE EXTERNAL DATA SOURCE external_data_source_name WITH (LOCATION = 'mongodb://<server>[:<port>]', -- PUSHDOWN = ON | OFF, CREDENTIAL = credential_name);
Optional: Create statistics on an external table.
We recommend creating statistics on external table columns, especially the ones used for joins, filters and aggregates, for optimal query performance.
CREATE STATISTICS statistics_name ON customer (C_CUSTKEY) WITH FULLSCAN;
Once you have created an external data source, you can use the CREATE EXTERNAL TABLE command to create a queryable table over that source.
For an example, see Create an external table for MongoDB.
Flattening is enabled for nested and repeated data from MongoDB document collections. User is required to enable
create an external table and explicitly specify a relational schema over MongoDB document collections that may have nested and/or repeated data.
JSON nested/repeated data types will be flattened as follows
Object: unordered key/value collection enclosed in curly braces (nested)
SQL Server creates a table column for each object key
- Column Name: objectname_keyname
Array: ordered values, separated by commas, enclosed in square brackets (repeated)
SQL Server adds a new table row for each array item
SQL Server creates a column per array to store the array item index
Column Name: arrayname_index
Data Type: bigint
There are several potential issues with this technique, two of them being:
An empty repeated field will effectively mask the data contained in the flat fields of the same record
The presence of multiple repeated fields can result in an explosion of the number of produced rows
As an example, SQL Server evaluates the MongoDB sample dataset restaurant collection stored in non-relational JSON format. Each restaurant has a nested address field and an array of grades it was assigned on different days. The figure below illustrates a typical restaurant with nested address and nested-repeated grades.
Object address will be flattened as below:
- Nested field restaurant.address.building becomes restaurant.address_building
- Nested field restaurant.address.coord becomes restaurant.address_coord
- Nested field restaurant.address.street becomes restaurant.address_street
- Nested field restaurant.address.zipcode becomes restaurant.address_zipcode
Array grades will be flattened as below:
Cosmos DB Connection
Using the Cosmos DB mongo api and the Mongo DB PolyBase connector you can create an external table of a Cosmos DB instance. This accomplished by following the same steps listed above. Make sure the Database scoped credential, Server address, port, and location string reflect that of the Cosmos DB server.
To learn more about PolyBase, see Overview of SQL Server PolyBase.