Defining Types

The first step is to create the types we need to model departments and roles. We will define one object type called department from which we can later create instances of department objects.

We will also define three relations on the department object type—owner, member, and viewer—one for each of the three roles that users can have.

Object Type

We use the setObjectType mutation to create a new object type called "department".

Query

mutation CreateDepartmentObjectType {
  setObjectType(type: {
    name: "department"
    displayName: "Department"
    isSubject: false
    ordinal: 0
    status: []
  }) {
    objectType {
      name
      displayName
      isSubject
      ordinal
      status
    }
  }
}

Results

{
  "data": {
    "setObjectType": {
      "objectType": {
        "name": "department",
        "displayName": "Department",
        "isSubject": false,
        "ordinal": 0,
        "status": []
      }
    }
  }
}

The value of the name field is used to identify the object type and must be unique among all object types in the directory.

The input fields are described in more detail here.

Querying Object Types

To see that the department object types was indeed created, we can enumerate object types using the objectTypes query.

The following query retrieves the name and displayName fields of the first 10 object types that are not subject and have an no modifier flags in their status field:

Query

query ListObjectTypes {
  objectTypes(first: 10, where: {isSubject: false, status: {equals: []}}) {
    nodes {
      name
      displayName
    }
  }
}

Results

{
  "data": {
    "objectTypes": {
      "nodes": [
        {
          "name": "department",
          "displayName": "Department"
        },
        {
          "name": "application",
          "displayName": "Application"
        },
        {
          "name": "resource",
          "displayName": "Resource"
        }
      ]
    }
  }
}

Relation Types

Now that we have created an object type, we can define relations that it can have with other objects.

We define three relations, viewer, member, and owner, to represent the three roles that users can have in a department.

Viewer

We'll start witht the viewer relation.

Query

mutation CreateDepartmentViewerRelationType {
  setRelationType(type: {
    objectType: "department"
    name: "viewer",
    displayName: "department:viewer"
    ordinal: 1,
    status: []
    unions: []
  }) {
    relationType {
      name
      displayName
    }
  }
}

Results

{
  "data": {
    "setRelationType": {
      "relationType": {
        "name": "viewer",
        "displayName": "department:viewer"
      }
    }
  }
}

We specify the object type on which the relation can occur using the objectType field.

The name field of relation types needs to be unique among all relation types defined on the same object type.

By convention, the displayName of relation types is set to <object type name>:<relation type name>.

The input fields are described in more detail here.

Member and Owner

A single GraphQL request can include multiple queries and mutations. We'll create the member and owner relation types in a single request:

Query

mutation CreateDepartmentMemberAndOwnerRelationTypes {
  member: setRelationType(type: {
    name: "member",
    displayName: "department:member"
    objectType: "department"
    ordinal: 1,
    status: []
    unions: []
  }) {
    relationType {
      name
    }
  }

  owner: setRelationType(type: {
    name: "owner",
    displayName: "department:owner"
    objectType: "department"
    ordinal: 1,
    status: []
    unions: []
  }) {
    relationType {
      name
    }
  }
}

Results

{
  "data": {
    "member": {
      "relationType": {
        "name": "member"
      }
    },
    "owner": {
      "relationType": {
        "name": "owner"
      }
    }
  }
}

info

When sending multiple instances of the same query or mutation in a single request, we need to provide them with aliases to distinguish their respective results in the response's data object.

Querying Relation Types

Let's run a query to see that all three relation types were created as intended.

Query

query RelationTypes {
  relationTypes(first: 3) {
    nodes {
      name
      displayName
      objectType {
        name
      }
      ordinal
    }
  }
}

Results

{
  "data": {
    "relationTypes": {
      "nodes": [
        {
          "name": "viewer",
          "displayName": "department:viewer",
          "objectType": {
            "name": "department"
          },
          "ordinal": 1,
        },
        {
          "name": "member",
          "displayName": "department:member",
          "objectType": {
            "name": "department"
          },
          "ordinal": 1,
        },
        {
          "name": "owner",
          "displayName": "department:owner",
          "objectType": {
            "name": "department"
          },
          "ordinal": 1,
        }
      ]
    }
  }
}

info

The three relation types we created appear first because we gave them all a low ordinal value (i.e. 1). Types with lower ordinal values are retruned first. When several types have the same ordinal value, as in our case, their sort order is arbitrary but stable.