Perspectives

Within a graph there are often multiple smaller graphs which are connected to each other. Sometimes you need to see everything. Other times it’s helpful to have a more focused view. Defining a Perspective allows you to select what parts of the graph to show and how to show them.

Let’s use the Northwind example data set. Northwind is a fictional company with a database that tracks their product catalog, sales orders, and sales staff.

When importing Northwind directly into a Neo4j graph there is a 1:1 correspondence of tables to labels, which means that for each record a node has been created with a label that matches the original name of the table. Each foreign key reference gets converted into a graph relationship.

After starting this Neo4j database and launching Bloom for the first time, Bloom automatically generates a Perspective based on the data it finds. If you have any previously defined Perspectives, they are available for selection. You can always ask Bloom to auto-generate a Perspective. The auto-generated Perspective is a good place to start. Bloom categorizes the nodes into entities, select useful captions and assign color-coding.

Keep in mind though, that when Bloom auto-generates a Perspective, a complete scan of the database is performed. If your database is large, i.e. > 10,000,000 nodes and relationships combined, this will take a long time and you can opt for a quick scan instead. See Database scans below for more information. The Northwind example data set is not large and Bloom can quickly auto-generate a Perspective.

By contrast, when Bloom creates an Auto-perspective the first time you open it with a new database, it samples a node from each category it finds and use the properties in the sampled node. As with any other perspective, as you query more data, Bloom adds any new properties as they are found.

You are able to search and explore the entire Northwind graph.

While everyone in the organization could benefit from a graph view, not everyone needs to see everything. For instance, the shipping department of Northwind may only need to see orders, products, and customers. You can create another Perspective that highlights only those categories.

Similarly, you can create Perspectives that are specific to the sales department, purchasing department, or customer service department.

When changes are made to the database, for example by adding, removing or changing property data types, these are not automatically reflected in any existing Perspectives. In order to make the changes visible, click the "Refresh Perspective" button in the top right corner of the Perspective drawer.

Keep in mind that this operation may take anywhere from a few seconds to several minutes, depending on the size of your database.

Once you have refreshed your Perspective, new relationship types and property keys are added and data types for property keys are updated. Any relationship types or property keys that were hidden before the refresh will remain hidden after. If any labels, relationship types or property keys are deleted from the database, they are deleted from the Perspective as well.

However, new categories are not automatically created if new labels are added to the database, though the additional labels are available in the Perspective. Similarly, if a label is removed from the database, the associated category is not automatically removed from the Perspective when you refresh.

Also, note that if labels or properties have changed in the database, the change may affect styling rules and Search Phrases. Those have to be updated manually after a Perspective refresh.

A database scan is performed each time a perspective is created or refreshed, in order to determine what property keys are present for each label and relationship type. This information is used in various places in Bloom, such as the categories, search phrase editor and styling rules.

There are two types of scan, complete and quick. The complete scan will scan all the nodes and relationships in the database, whereas the quick one will scan only a random sample. The quick scan is faster than the complete, but it may not find all property keys if they are only present on a few nodes or relationships. A complete scan can take a long time if the database is large, i.e. if it contains > 10,000,000 nodes and relationships combined, and may result in a Bloom timeout.

If you know that you have a consistent schema for your database and/or if your database is large, a quick scan is recommended. On the other hand, if you have an inconsistent schema and/or if you can’t find some of your properties after a quick scan, a full scan is advisable.

When the Bloom client is used without the Bloom server plugin, Perspectives are stored locally on the user’s machine. This location depends on whether Bloom is running in Neo4j Desktop or in a web browser. In this setup, Perspectives cannot be shared between different users.

When the Bloom server plugin is used, Perspectives are stored in persistent storage (i.e. Neo4j database) provided by the server plugin. Therefore, they can be easily shared between different users who are accessing the same Neo4j server.

Perspective sharing can be explicitly controlled from the Sharing tab in the Perspective definition drawer. The best practice for assigning Perspectives is to create a custom role for each Perspective, and add that role to each user who should have access to the Perspective in question.