Property Graphs

The Structure of a Property Graph

The nodes in a property graph can represent anything from people, places, things, or concepts. Nodes can be labelled to indicate their type or class and can also have one or more properties associated with them. For example, a node labelled "Person" might have properties like "Name", "Age", "Address", "Occupation", etc. Properties can have different data types, such as strings, numbers, dates, or even complex data structures like arrays or maps.

The edges in a property graph represent relationships between nodes. Edges can be labelled to indicate the type of relationship and can also have properties associated with them. For example, an edge labelled "Friend" might have a property like "DateAdded", which indicates when two people became friends.

One of the key features of property graphs is that they allow for properties to be attached to both nodes and edges. This allows for a rich and expressive data model where relationships can have characteristics of their own, such as timestamps or ratings. This makes it possible to represent complex data structures with high accuracy and detail.

In addition to nodes, edges, and properties, property graphs also support indexes and queries. Indexes are used to speed up queries and enable efficient retrieval of data. Queries in property graphs are expressed using a query language, such as Cypher or Gremlin, which allows for powerful querying capabilities that can traverse the graph in various ways.

Labelled Property Graphs (LPGs) & Typed Property Graphs

The simplicity and flexibility of property graphs makes them ideal for use in rapid development and prototyping scenarios.

In order to provide to provide additional context and semantics to the data in a property graph, labels or types can be added as properties of nodes. In a labelled property graph (LPG) it is mandatory that a node has a label property. The label is a string or identifier that represents a category or class of nodes, such as "Person", "Place", "Organization", or "Product". The label provides additional information about the nature of the node and helps to classify it within the graph.

A Typed Property Graph is very similar to a labelled property graph but a type rather than a label is the enforced property on a node. The type property of a node determines the set of properties that it can have. For example, a node representing a person might have a type of "Person", which allows it to have properties such as name, age, and email. The addition of types is useful for data integrity and ensuring that only valid properties are associated with nodes.

Labels and types can also be added to the edges of graphs. For example, an edge (relationship) between a person and a company might have a type of "WorksAt", which allows it to have properties such as start date and end date. A label on an edge could describe different types of relationships, such as "Friend", "Follower", or "Colleague".

There are pros and cons adding types or labels to a property graph, some important considerations are explored here:

Analysis

When labels or types are enforced on nodes, users can more easily identify patterns and relationships in the data. Without these, property graphs may lack the contextual information required to understand relationships between nodes.

Data Quality

By enforcing a strict data schema, labelled or typed property graphs can help ensure data quality. Labels and types can be used to enforce consistency in data entry and ensure that data is categorised correctly. Without a strict schema, property graphs are more susceptible to data inconsistencies and errors. This can be problematic for larger and more complex graphs.

Flexibility

Property graphs are highly flexible and can be easily modified or extended to accommodate new data types or relationships. Labelled and typed property graphs are more rigid in terms of their data schema. Once a label or type is assigned to a node, it can be difficult to change or modify without affecting the entire graph.

Query Performance

The addition of labels or types to property graphs can help to improve query performance by limiting the scope of searches to a particular type of node. For example, if you are only interested in people in a social network graph, you can restrict your search to nodes with the "Person" label.

Simplicity

The addition of labels or types to a property graph can increase the complexity of the data model. This can make it more difficult to manage and maintain, especially for large and complex graphs.

Storage

The addition of labels or types often means that a property graphs require more storage space.