The Backend Model Layer

Models for the backend define the essential properties and traversal operations for different data types, and use the Tinkerpop 2.0 Frames ORM. A crucial distinction with model classes in other frameworks is that models do not define every property that a data type can hold, only those which are required for behaviour in the core of the stack. For example, a DocumentaryUnitDescription item must have a languageCode property. Other optional properties, such as scopeAndContent are managed by high layers, such as the portal front-end or the data import classes.

Abstract Model Types

Because Tinkerpop Frames models are Java interfaces (with instances created by proxies at runtime) they can themselves extend multiple other interfaces. For better or worse, we use this extensively as a form of multiple inheritance so that "concrete" model instances can inherit behaviours (graph traversals and property accessors) from abstract base models. For example, there is a base model called Identifiable that looks like this:

package eu.ehri.project.models.base;

import com.tinkerpop.frames.Property;
import eu.ehri.project.definitions.Ontology;
import eu.ehri.project.models.annotations.Mandatory;

 * Base interface for entities that have an identifier property (other than the
 * internally assigned node ID).
public interface Identifiable extends Entity {

    String getIdentifier();

Models that inherit from Identifiable then get a getIdentifier() method which looks up the property via the identifier key. Because this property accessor is marked @Mandatory a model instance cannot be stored if the property is missing or empty.

Other abstract base model interfaces exhibit more complex behaviours. For example, the Accessible interface defines behaviour for any entity that:

  • requires access control and permissions
  • has a lifecycle stored in the audit log

These behaviours mean that any Accessible item can have certain relationships to other nodes, such as an access relationship to any node that inherits the Accessor behaviour (both UserProfile and Group models.)

Similarly, any model that inherits the PermissionScope interface is one that can contain other items to which permissions can be granted. For example, the Repository model is a PermissionScope because it contains DocumentaryUnit items, and users or groups can be granted permission to create DocumentaryUnit items within the scope of that repository.

Here are some of the most important abstract models:

All models inherit from Entity. It defines accessors called getId() and getType() which read, respectively, the __id and __type properties which are present on all model nodes in the graph.
Models that can have their access restricted by the presence of a access relationship to one or more Accessor nodes. Also defines behaviour for reading history by retrieving a set of Version nodes.
Models that can be related to Description nodes via an incoming describes relationship.
Models that can be the containing scope for permissions pertaining to other items.
Models that can have an incoming annotates relationship from an Annotation node.

Concrete model types

Concrete model types can be actually saved in the graph. Their name corresponds to their __type property (and also their Neo4j node label - this redundancy is due to node labels being added to Neo4j after we were using the __type property.)

A node representing an access point, which is EHRI lingo for a reference to another item, e.g. a <subject>> tag in an EAD file. The access point just contains text, not an actual link, but it can also be the body of a Link node.
A node representing an address.
A node representing an annotation, which can point to any Annotatable item.
A node which represents data types to which permissions can attached. The __id property corresponds to the name of the data type, e.g. Country or DocumentaryUnit.
A node representing a county. The __id is the lowercase ISO 3166-2 code.
A node representing a span of time, with startDate and endDate properties.
A node representing an archival unit, which is either heldBy a Repository or the childOf another DocumentaryUnit item.
A description of an archival unit, with properties corresponding to ISAD(G).
A group, to which UserProfile or other Group nodes can belong, via the belongsTo property.
A node representing a person, corporate body, or family.
A description of a historical agent, with properties corresponding to ISAAR.
A node representing a link between two items, via the hasLinkTarget property. A link can also have an AccessPoint body, which provides its name.
A node representing a change to an item which occured prior to its ingest.
A node representing a type of permission, with its ID as the permission identifier, e.g. create or update.
A node representing a grant of some given Permission to a particular target, where the target could be either a specific Accessible item, or a ContentType node. A PermissionGrant can also have a particular PermissionScope.
A node representing an archival institution.
A description of an archival institution, with properties corresponding to ISDIAH.
A node repesenting some ingest data we do not recognise, but someone might want to look at.
A node representing a user of the system, potentially belonging to one or more Group nodes.
A node representing a "virtual" archival entity, which does not exist in any actual archive, but serves the purpose of allowing non-virtual material to be collected together in virtual space.