Skip to content

Views

This documentation provides a detailed explanation of views in Databricks, covering their purpose, types, lifecycle behavior, usage patterns, and syntax. Views are fundamental for abstracting complex SQL queries and enabling reusable, logical representations of underlying data without duplicating storage.

1. What Is a View?

A view in Databricks is a virtual table:

  • It does not store physical data.
  • It represents a saved SQL query against one or more real tables.
  • Each time the view is queried, the underlying SQL logic is re-executed.

Views simplify data access and encapsulate business logic in a reusable form.

2. Types of Views in Databricks

Databricks supports three types of views:

View Type Scope Persistence Dropped When Use Case
Stored View Across sessions & users Persisted in DB Manual (via DROP VIEW) Reusable logic across notebooks
Temporary View Current Spark session In-memory On session end Short-term or one-time transformations
Global Temp View All notebooks in cluster In-memory (global_temp DB) On cluster restart Shared logic across notebooks in cluster

3. Stored Views (Persistent Views)

3.1. Description

  • Stored in the metastore of the target database.
  • Survive across sessions and cluster restarts.
  • Queryable like regular tables.

3.2. Syntax

CREATE VIEW sales_summary AS
SELECT region, SUM(total_sales) AS revenue
FROM sales_data
GROUP BY region;

3.3. Deletion

DROP VIEW sales_summary;

4. Temporary Views

4.1. Description

  • Lives only in the current Spark session.
  • Automatically deleted when the session ends.
  • Commonly used for intermediate transformations or quick exploration.

4.2. Lifecycle Triggers

Spark session is re-initialized in the following scenarios:

  • Opening a new notebook.
  • Detaching and reattaching a notebook to a cluster.
  • Installing a Python package (restarts interpreter).
  • Restarting the cluster.

4.3. Syntax

CREATE TEMP VIEW temp_sales AS
SELECT * FROM sales_data WHERE region = 'EMEA';
  • Use TEMP or TEMPORARY in the CREATE VIEW statement.

5. Global Temporary Views

5.1. Description

  • Lives across multiple sessions, but within the same cluster.
  • Stored in a reserved database: global_temp.
  • Visible to all notebooks attached to the same cluster.

5.2. Syntax

CREATE GLOBAL TEMP VIEW global_sales AS
SELECT * FROM sales_data WHERE total_sales > 1000;
  • To query:
SELECT * FROM global_temp.global_sales;

5.3. Deletion

  • Automatically dropped when the cluster restarts.

6. Summary Table

Feature Stored View Temporary View Global Temporary View
Data Stored No No No
Metadata Location Hive Metastore Session memory global_temp (cluster memory)
Lifetime Until manually dropped Until session ends Until cluster restart
Visibility Scope All sessions & clusters Current session only All sessions in same cluster
Creation Syntax CREATE VIEW CREATE TEMP VIEW CREATE GLOBAL TEMP VIEW
Required Qualifier to Query None None global_temp. prefix required
Use Case Reusable definitions Quick transformation Shared logic in dev/test cluster

7. Best Practices

  • Use stored views for production use cases where logic needs to persist across clusters and sessions.
  • Use temporary views for ad hoc data exploration or transformations inside a single notebook.
  • Use global temporary views for sharing view definitions across multiple notebooks during collaborative development in a single cluster session.