Advanced Indexing

This page discusses a number of advanced indexing topics. Be sure to read about Titan's indexing system first.

Dual Indexing

For some use cases, it can be beneficial to register a property key with multiple index backends. This is supported by Titan.

For instance, consider the scenario where user vertices have the property key name which stores the full name of the user – first, middle, and last name. We define the name key as follows:

g.makeType().name("name").dataType(String.class).indexed(Vertex.class).indexed("search",Vertex.class).unique(Direction.OUT).makePropertyKey()

Registering the name key with the standard index and an external full-text index has the following advantages:

• Quick querying for users by their entire name using the standard index: g.query().has("name",EQUAL,"John Doe").vertices()
• Using the full-text index to search for user’s by any part of their name: g.query().has("name",CONTAINS,"John").vertices()

Titan will automatically route the index request to the correct index backend. Likewise, Titan can also handle composite queries that cover multiple indexes. Titan will attempt to determine an optimal query execution plan.