Skip to end of metadata
Go to start of metadata

MongoDB Translator

The MongoDB translator provides a relational view of data that resides in a MongoDB database. This translator is capable of converting Teiid SQL queries into MongoDB based queries. It supports a full range of SELECT, INSERT, UPDATE and DELETE calls.

MongoDB is a document based "schema-less" database with it own query language - it does not map perfectly with relational concepts or the SQL query language. More and more systems are using a MongoDB kind of NOSQL store for scalability and performance. For example, applications like storing audit logs or managing web site data fits well with MongoDB, and does not require using a structural database like Oracle, Postgres etc. MongoDB uses JSON documents as its primary storage unit, and those documents can have additional embedded documents inside the parent document. By using embedded documents it co-locates the related information to achieve de-normalization that typically requires either duplication of data or joins to achieve querying in a relational database.

To make MongoDB work with Teiid the challenge for the MongoDB translator is "How to design a MongoDB store that can achieve the balance between relational and document based storage?" In our opinion the advantages of "schema-less" design are great at development time, not much at runtime except in few special situations. "Schema-less" can also be a problem with migration of application versions and the ability to query and make use of returned information effectively.

Since it is hard and may be impossible in certain situations to derive a schema based on existing the MongoDB collection(s), Teiid approaches the problem in reverse compared to other translators. When working with MongoDB, Teiid requires the user to define the MongoDB schema upfront using Teiid metadata. Since Teiid only allows relational schema as its metadata, the user needs to define their MongoDB schema in relational terms using tables, procedures, and functions. For the purposes of MongoDB, the Teiid metadata has been extended to support extension properties that can be defined on the table to convert it into a MongoDB based document. These extension properties let users define, how a MongoDB document is structured and stored. Based on the relationships (primary-key, foreign-key) defined on a table, and the cardinality (ONE-to-ONE, ONE-to-MANY, MANY-to-ONE) relations between tables are mapped such that related information can be embedded along with the parent document for co-location (see the de-normalization comment above). Thus a relational schema based design, but document based storage in MongoDB. Currently direct mapping of MANY-to-MANY is not supported.

Who is the primary audience for this translator?

The above may not satisfy every user's needs. The document structure in MongoDB can be more complex than what Teiid can currently define. We hope this will eventually catch up in future versions of Teiid. This is currently designed for:

1. Users that are using relational databases and would like to move/migrate their data to MongoDB to take advantages scaling and performance with out modifying the end user applications currently running.

2. Users that are starting out with MongoDB and do not have experience with MongoDB, but are seasoned SQL developers. This provides a low barrier of entry compared to using MongoDB directly as an application developer.

3. Integrate other enterprise data sources with MongoDB based data.


The name of the translator to use in vdb.xml is "mongodb".
For example:

The translator does not provide a connection to the MongoDB. For that purpose, Teiid has a JCA adapter that provides a connection to MongoDB using the MongoDB Java Driver. To define such connector, use the following XML fragment in standalone-teiid.xml. See a example in "<jboss-as>/docs/teiid/datasources/mongodb"

The above defines the translator and connector. For more ways to create the connector see MongoDB Data Sources. MongoDB translator can derive the metadata based on existing document collections in some scenarios, however when working with complex documents the interpretation of metadata may be accurate, in those situations the user MUST define the metadata. For example, you can define a schema using DDL:

when INSERT operation below executed against table using Teiid,

MongoDB translator will create a below document in the MongoDB

If a PRIMARY KEY is defined on the table as

then that column name is automatically used as "_id" field in the MongoDB collection, then document structure is stored in the MongoDB as

If you defined the composite PRIMARY KEY on Customer table as

the document structure will be

Data Types

MongoDB translator supports automatic mapping of Teiid data types into MongoDB data types, including the support for Blobs, Clobs and XML. The LOB support is based on GridFS in MongoDB. Arrays are in the form of

are supported. User can get individual items in the array using function array_get, or can transform the array into tabular structure using ARRATTABLE.

Note that even though embedded documents can also be in arrays, the handling of embedded documents is different from array with scalar values.

Regular Expressions, MongoDB::Code, MongoDB::MinKey, MongoDB::MaxKey, MongoDB::OID is not currently supported.

Extension Metadata Properties To Build Complex Documents

Using the above DDL or any other metadata facility, a user can map a table in a relational store into a document in MongoDB, however to make effective use of MongoDB, you need to be able to build complex documents, that can co-locate related information, so that data can queried in a single MongoDB query. Otherwise, since MongoDB does not support join relationships like relational database, you need to issue multiple queries to retrieve and join data manually. The power of MongoDB comes from its "embedded" documents and its support of complex data types like arrays and use of the aggregation framework to be able to query them. This translator provides way to achieve that goals.

When you do not define the complex embedded documents in MongoDB, Teiid can step in for join processing and provide that functionality, however if you want to make use of the power of MongoDB itself in querying the data and avoid bringing the unnecessary data and improve performance, you need to look into building these complex documents.

MongoDB translator defines two additional metadata properties along with other Teiid metadata properties to aid in building the complex "embedded" documents. You can use the following metadata properties in your DDL.

  • teiid_mongo:EMBEDDABLE - Means that data defined in this table is allowed to be included as an "embeddable" document in any parent document. The parent document is referenced by the foreign key relationships. In this scenario, Teiid maintains more than one copy of the data in MongoDB store, one in its own collection and also a copy in each of the parent tables that have relationship to this table. You can even nest embeddable table inside another embeddable table with some limitations. Use this property on table, where table can exist, encompass all its relations on its own. For example, a "Category" table that defines a "Product"'s category is independent of Product, which can be embeddable in "Products" table.
  • teiid_mongo:MERGE - Means that data of this table is merged with the defined parent table. There is only a single copy of the data that is embedded in the parent document. Parent document is defined using the foreign key relationships.

Using the above properties and FOREIGN KEY relationships, we will illustrate how to build complex documents in MongoDB.

Please note a given table can contain either the "teiid_mongo:EMBEDDABLE" property or the "teiid_mongo:MERGE" property defining the type of nesting in MongoDB. A table is not allowed to have both properties.

ONE-2-ONE Mapping

If your current DDL structure representing ONE-2-ONE relationship is like

by default, this will produce two different collections in MongoDB, like with sample data it will look like

You can enhance the storage in MongoDB to a single collection by using "teiid_mongo:MERGE' extension property on the table's OPTIONS clause

this will produce single collection in MongoDB, like

With the above both tables are merged into a single collection that can be queried together using the JOIN clause in the SQL command. Since the existence of child/additional record has no meaning with out parent table using the "teiid_mongo:MERGE" extension property is right choice in this situation.

Note that the Foreign Key defined on child table, must refer to Primary Keys on both parent and child tables to form a One-2-One relationship.

ONE-2-MANY Mapping.

Typically there can be more than two (2) tables involved in this relationship. If MANY side is only associated single table, then use "teiid_mongo:MERGE" property on MANY side of table and define ONE as the parent. If associated with more than single table then use "teiid_mongo:EMBEDDABLE".

For example if you have DDL like

in the above a Single Customer can have MANY Orders. There are two options to define the how we store the MongoDB document. If in your schema, the Customer table's CustomerId is only referenced in Order table (i.e. Customer information used for only Order purposes), you can use

that will produce a single document for Customer table like

If Customer table is referenced in more tables other than Order table, then use "teiid_mongo:EMBEDDABLE" property

This creates three different collections in MongoDB.

Here as you can see the Customer table contents are embedded along with other table's data where they were referenced. This creates duplicated data where multiple of these embedded documents are managed automatically in the MongoDB translator.

All the SELECT, INSERT, DELETE operations that are generated against the tables with "teiid_mongo:EMBEDDABLE" property are atomic, except for UPDATES, as there can be multiple operations involved to update all the copies. Since there are no transactions in MongoDB, Teiid plans to provide automatic compensating transaction framework around this in future releases TEIID-2957.

MANY-2-ONE Mapping.

This is same as ONE-2-MANY, see above to define relationships.

A parent table can have multiple "embedded" and as well as "merge" documents inside it, it not limited so either one or other. However, please note that MongoDB imposes document size is limited can not exceed 16MB.

MANY-2-MANY Mapping.

This can also mapped with combination of "teiid_mongo:MERGE" and "teiid_mongo:EMBEDDABLE" properties (partially). For example if DDL looks like

you modify the DDL like below, to have

That will produce a document like

  • Currently nested embedding of documents has limited support due to capabilities of handling nested arrays is limited in the MongoDB. Nesting of "EMBEDDALBLE" property with multiple levels is OK, however more than two levels with MERGE is not recommended. Also, you need to be caution about not exceeding the document size of 16 MB for single row, so deep nesting is not recommended.
  • JOINS between related tables, MUST have used either of "EMBEDDABLE" or "MERGE" property, otherwise the query will result in error. In order for Teiid to correctly plan and support the JOINS, in the case that any two tables are NOT embedded in each other, use allow-joins=false property on the Foreign Key that represents the relation. For example:

with the example above, Teiid will create two collections, however when user issues query such as

instead of resulting in error, the JOIN processing will happen in the Teiid engine, without the above property it will result in an error.

When you use above properties and carefully design the MongoDB document structure, Teiid translator can intelligently collate data based on their co-location and take advantage of it while querying.

Geo Spatial function support

MongoDB translator supports geo spatial query operators in the "WHERE" clause, when the data is stored in the GeoJSon format in the MongoDB Document. The supported functions are

a sample query looks like


MongoDB translator designed on top of the MongoDB aggregation framework, use of MongoDB version that supports this framework is mandatory. Apart from SELECT queries, this translator also supports INSERT, UPDATE and DELETE queries.

This translator supports

  • grouping
  • matching
  • sorting
  • filtering
  • limit
  • support for LOBs using GridFS
  • Composite primary and foreign keys.

Native Queries

MongoDB source procedures may be created using the teiid_rel:native-query extension - see Parameterizable Native Queries. The procedure will invoke the native-query similar to a direct procedure call with the benefits that the query is predetermined and that result column types are known, rather than requiring the use of ARRAYTABLE or similar functionality.

Direct Query Procedure

This feature is turned off by default because of the security risk this exposes to execute any command against the source. To enable this feature, override the execution property called SupportsDirectQueryProcedure to true.
By default the name of the procedure that executes the queries directly is called native. Override the execution property DirectQueryProcedureName to change it to another name.

The MongoDB translator provides a procedure to execute any ad-hoc aggregate query directly against the source without Teiid parsing or resolving. Since the metadata of this procedure's results are not known to Teiid, they are returned as an object array containing single blob at array location one(1). This blob contains the JSON document. XMLTABLE can be used construct tabular output for consumption by client applications.

Example MongoDB Direct Query

In the above example, a collection called "city" is looked up with filter that matches the "city" name with "FREEDOM", using "native" procedure and then using the nested tables feature the output is passed to a XMLTABLE construct, where the output from the procedure is sent to a JSONTOXML function to construct a XML then the results of that are exposed in tabular form.

The direct query MUST be in the format

From Teiid 8.10, MongoDB translator also allows to execute Shell type java script commands like remove, drop, createIndex. For this the command needs to be in format

and example looks like

Enter labels to add to this page:
Please wait 
Looking for a label? Just start typing.
  1. Jun 18, 2014

    MongoDB is misspelled "MonogoDB" in one place. I submitted an edit but the Rich Text editor made a mess of it...

    1. Jun 19, 2014

      Yes, avoid the Rich Text editor it seems to do more harm than good.