We’re overhauling Dgraph’s docs to make them clearer and more approachable. If you notice any issues during this transition or have suggestions, please let us know.
Dgraph provides two types of built-in identifiers: the ID scalar type and the @id directive.
  • The ID scalar type is used when you don’t need to set an identifier outside of Dgraph.
  • The @id directive is used for external identifiers, such as email addresses.

​
The @id directive

For some types, you’ll need a unique identifier set from outside Dgraph. A common example is a username. The @id directive tells Dgraph to keep that field’s values unique and use them as identifiers. For example, you might set the following type in a schema: 
type User {
    username: String! @id
    ...
}
Dgraph requires a unique username when creating a new user. It generates the input type for addUser with username: String!, so you can’t make an add mutation without setting a username; and when processing the mutation, Dgraph will ensure that the username isn’t already set for another node of the User type. In a single-page app, you could render the page for http://.../user/Erik when a user clicks to view the author bio page for that user. Your app can then use a getUser(username: "Erik") { ... } GraphQL query to fetch the data and generate the page. Identities created with @id are reusable. If you delete an existing user, you can reuse the username. Fields with the @id directive must have the type String!. As with ID types, Dgraph generates queries and mutations so you can query, update, and delete data in nodes, using the fields with the @id directive as references. It is possible to use the @id directive on more than one field in a type. For example, you can define a type like the following: 
type Book {
    name: String! @id
    isbn: String! @id
    genre: String!
    ...
}
You can then use multiple @id fields in arguments to get queries, and while searching, these fields will be combined with the AND operator, resulting in a Boolean AND operation. For example, for the above schema, you can send a getBook query like the following: 
query {
  getBook(name: "The Metamorphosis", isbn: "9871165072") {
    name
    genre
    ...
  }
}
This will yield a positive response if both the name and isbn match any data in the database.

​
@id and interfaces

By default, if used in an interface, the @id directive will ensure field uniqueness for each implementing type separately. In this case, the @id field in the interface won’t be unique for the interface but for each of its implementing types. This allows two different types implementing the same interface to have the same value for the inherited @id field. There are scenarios where this behavior might not be desired, and you may want to constrain the @id field to be unique across all the implementing types. In that case, you can set the interface argument of the @id directive to true, and Dgraph will ensure that the field has unique values across all the implementing types of an interface. For example: 
interface Item {
  refID: Int! @id(interface: true) # if there is a Book with refID = 1, then there can't be a chair with that refID.
  itemID: Int! @id # If there is a Book with itemID = 1, there can still be a Chair with the same itemID.
}

type Book implements Item { ... }
type Chair implements Item { ... }
In the above example, itemID won’t be present as an argument to the getItem query as it might return more than one Item.
get queries generated for an interface will have only the @id(interface: true) fields as arguments.

​
Combining ID and @id

You can use both the ID type and the @id directive on another field definition to have both a unique identifier and a generated identifier. For example, you might define the following type in a schema: 
type User {
    id: ID!
    username: String! @id
    ...
}
With this schema, Dgraph requires a unique username when creating a new user. This schema provides the benefits of both of the previous examples above. Your app can then use the getUser(...) { ... } query to provide either the Dgraph-generated id or the externally-generated username.
If in a type there are multiple @id fields, then in a get query these arguments will be optional. If in a type there’s only one field defined with either @id or ID, then that will be a required field in the get query’s arguments.