Views

A view is a virtual table defined by a SQL SELECT statement. Views do not store data themselves; instead, their defining query is executed as a sub-query whenever the view is referenced.

What are views for?​

Views provide several benefits:

  • Abstraction: Hide complex queries behind simple table-like interfaces
  • Reusability: Define queries once, use them everywhere
  • Security: Control data access without exposing underlying tables
  • Maintainability: Single source of truth for business logic
Quick example
-- Create a view
CREATE VIEW hourly_summary AS (
  SELECT ts, symbol, sum(quantity) as volume
  FROM trades
  SAMPLE BY 1h
);

-- Query the view like a table
SELECT * FROM hourly_summary WHERE symbol = 'AAPL';

Creating views​

Use CREATE VIEW to define a new view:

Basic view
CREATE VIEW daily_prices AS (
  SELECT ts, symbol, last(price) as closing_price
  FROM trades
  SAMPLE BY 1d
)

CREATE IF NOT EXISTS​

To avoid errors when the view already exists:

CREATE VIEW IF NOT EXISTS price_view AS (
  SELECT symbol, last(price) as price FROM trades SAMPLE BY 1h
)

CREATE OR REPLACE​

To update an existing view or create it if it doesn't exist:

CREATE OR REPLACE VIEW price_view AS (
  SELECT symbol, last(price) as price, ts FROM trades SAMPLE BY 1h
)

For full syntax details, see CREATE VIEW.

Querying views​

Views are queried exactly like tables:

SELECT * FROM my_view

SELECT ts, price FROM my_view WHERE symbol = 'AAPL'

SELECT v1.ts, v2.value
FROM view1 v1
JOIN view2 v2 ON v1.id = v2.id

Optimizer transparency​

Views in QuestDB are fully transparent to the query optimizer. When you query a view, the optimizer treats it exactly as if you had written the view's query inline as a sub-query. This means views benefit from the complete suite of query optimizations:

  • Filter push-down: WHERE conditions are pushed to base tables
  • Projection push-down: Only required columns are read from storage
  • Join optimization: Join order and strategies are optimized across view boundaries
  • ORDER BY optimization: Sorting can leverage table indexes
  • Timestamp optimizations: Time-based operations use partition pruning
-- View definition
CREATE VIEW trades_view AS (
  SELECT ts, symbol, price, quantity FROM trades WHERE price > 0
)

-- This query is optimized as if written inline
SELECT ts, price FROM trades_view WHERE symbol = 'AAPL' ORDER BY ts
-- Optimizer sees: SELECT ts, price FROM trades WHERE price > 0 AND symbol = 'AAPL' ORDER BY ts
-- Only ts and price columns are read, filters applied at scan, ordering uses index

Use EXPLAIN to see how the optimizer processes view queries:

EXPLAIN SELECT * FROM trades_view WHERE symbol = 'AAPL'

There is no performance penalty for using views compared to writing equivalent sub-queries directly.

Parameterized views​

Views support the DECLARE statement to define parameters with default values. Use OVERRIDABLE to allow users to change parameter values at query time.

Creating a parameterized view​

CREATE VIEW filtered_trades AS (
  DECLARE OVERRIDABLE @min_price := 100
  SELECT ts, symbol, price FROM trades WHERE price >= @min_price
)

Querying with default parameters​

SELECT * FROM filtered_trades
-- Uses default @min_price = 100

Overriding parameters​

DECLARE @min_price := 500 SELECT * FROM filtered_trades
-- Overrides @min_price to 500

Multiple parameters​

By default, parameters are non-overridable. Use OVERRIDABLE to allow override at query time:

CREATE VIEW price_range AS (
  DECLARE OVERRIDABLE @lo := 100, OVERRIDABLE @hi := 1000
  SELECT ts, symbol, price FROM trades WHERE price >= @lo AND price <= @hi
)

-- Query with custom range
DECLARE @lo := 50, @hi := 200 SELECT * FROM price_range

Non-overridable parameters​

Parameters without OVERRIDABLE cannot be changed at query time, providing security for sensitive filters:

CREATE VIEW secure_view AS (
  DECLARE @min_value := 0
  SELECT * FROM trades WHERE value >= @min_value
)

-- This will fail with "variable is not overridable: @min_value"
DECLARE @min_value := -100 SELECT * FROM secure_view

Mixed parameters​

Combine overridable and non-overridable parameters:

CREATE VIEW mixed_params AS (
  DECLARE @fixed_filter := 'active', OVERRIDABLE @limit := 100
  SELECT * FROM data WHERE status = @fixed_filter LIMIT @limit
)

-- @limit can be overridden, @fixed_filter cannot
DECLARE @limit := 50 SELECT * FROM mixed_params

View hierarchies​

Views can reference other views, tables, and materialized views:

-- Level 1: Raw data filtering
CREATE VIEW valid_trades AS (
  SELECT * FROM trades WHERE price > 0 AND quantity > 0
)

-- Level 2: Aggregation
CREATE VIEW hourly_stats AS (
  SELECT ts, symbol, sum(quantity) as volume
  FROM valid_trades
  SAMPLE BY 1h
)

-- Level 3: Derived metrics
CREATE VIEW hourly_vwap AS (
  SELECT ts, symbol, volume, turnover / volume as vwap
  FROM hourly_stats
  WHERE volume > 0
)
tip

Keep view hierarchies shallow (3-4 levels maximum) for better query planning and maintainability.

View management​

Listing views​

SELECT * FROM views()

Returns:

ColumnDescription
view_nameName of the view
view_sqlThe SQL definition
view_table_dir_nameInternal directory name
invalidation_reasonError message if view is invalid
view_statusvalid or invalid
view_status_update_timeTimestamp of last status change

Show view definition​

SHOW CREATE VIEW my_view

Returns the CREATE VIEW statement that would recreate the view.

Show view columns​

SHOW COLUMNS FROM my_view

Altering views​

To modify an existing view's definition:

ALTER VIEW my_view AS (SELECT col1, col2 FROM my_table WHERE col1 > 0)

For full syntax, see ALTER VIEW.

Dropping views​

DROP VIEW my_view

-- Or safely:
DROP VIEW IF EXISTS my_view

For full syntax, see DROP VIEW.

View invalidation​

Views are automatically invalidated when their dependencies change:

OperationEffect
DROP TABLEView becomes invalid
RENAME TABLEView becomes invalid
DROP COLUMNView becomes invalid if column is referenced
RENAME COLUMNView becomes invalid if column is referenced
Column type changeView metadata is updated

Automatic recovery​

Views are automatically revalidated when the invalidating condition is reversed:

  • If a table is dropped and later recreated, dependent views become valid again
  • If a column is renamed back to its original name, dependent views become valid again

Checking view status​

SELECT view_name, view_status, invalidation_reason
FROM views()
WHERE view_status = 'invalid'

Views in tables() output​

Views appear in the tables() function with table_type = 'V':

SELECT table_name, table_type FROM tables()
table_typeDescription
TRegular table
VView
MMaterialized view

Views vs materialized views​

Understanding when to use each type is important for performance:

FeatureViewMaterialized View
Data storageNone (virtual)Physical storage
Query executionOn every accessPre-computed
Data freshnessAlways currentDepends on refresh
PerformanceQuery-time costRead-time benefit
Storage costZeroProportional to result size

When to use views​

  • Simple transformations that execute quickly
  • Data that must always be current
  • Ad-hoc analysis where requirements change frequently
  • Parameterized queries with DECLARE
  • Low-frequency queries

When to use materialized views​

  • Heavy aggregations over large datasets
  • Frequently accessed summary data
  • Dashboard queries that run repeatedly
  • Historical summaries that don't need real-time accuracy

For detailed comparisons and examples, see Materialized Views.

Security with views​

Views provide a security boundary between users and underlying data.

Definer security model (Enterprise)​

Views use a definer security model. When a view is created, the creator's permissions are captured. Users querying the view only need SELECT permission on the view itself - they do not need permissions on the underlying tables.

-- Admin creates a view on sensitive table
CREATE VIEW public_summary AS (
  SELECT date, region, sum(sales) as total FROM sensitive_sales GROUP BY date, region
);

-- Grant SELECT on the view to analysts
GRANT SELECT ON public_summary TO analyst_role;

-- Analysts can query the view without access to sensitive_sales
SELECT * FROM public_summary;  -- Works!
SELECT * FROM sensitive_sales; -- Access denied!

The view's definer permissions are preserved even if the creator's account is later disabled or deleted.

No column-level permissions on views​

Unlike tables, views do not support column-level permissions. You can only grant or revoke permissions on the entire view:

-- This works: grant SELECT on entire view
GRANT SELECT ON my_view TO user1;

-- Column-level permissions are NOT supported for views
-- Use separate views to expose different column subsets

To provide different column access to different users, create multiple views with different column selections.

Security patterns​

Views enable several security patterns:

  • Data subsetting: Expose only specific rows or columns
  • Column masking: Hide sensitive columns from certain users
  • Row-level security: Filter rows based on business rules
  • Aggregation-only access: Provide summaries without raw data access

Column-level security example​

-- Base table with sensitive data
CREATE TABLE employees (
  id LONG,
  name VARCHAR,
  salary DOUBLE,        -- Sensitive
  department VARCHAR,
  hire_date TIMESTAMP
);

-- View exposing only non-sensitive columns
CREATE VIEW employees_public AS (
  SELECT id, name, department, hire_date
  FROM employees
);

-- Grant access to public view only
GRANT SELECT ON employees_public TO analyst_role;

Row-level security example​

-- View for specific trading desk
CREATE VIEW desk_a_trades AS (
  SELECT * FROM trades WHERE trader_id IN (101, 102, 103)
);

GRANT SELECT ON desk_a_trades TO desk_a_users;

For more details on permissions, see Role-Based Access Control (RBAC).

Performance considerations​

Views don't cache results​

Every query against a view executes the underlying query. For expensive aggregations accessed frequently, consider materialized views.

Optimize with indexes​

Create indexes on base table columns used in view filters:

ALTER TABLE trades ALTER COLUMN symbol ADD INDEX

Check query plans​

Always examine query plans when optimizing:

EXPLAIN SELECT * FROM my_view WHERE symbol = 'AAPL'

Best practices​

  • Use indexed columns in filters for best performance
  • Use parameterized views for common filter patterns
  • Avoid deeply nested view hierarchies (>3-4 levels) for maintainability
  • Consider materialized views for expensive aggregations that run frequently

Limitations​

  1. No data storage: Views don't store data - the query runs each time
  2. No indexes: Views cannot have indexes; filtering relies on underlying table indexes
  3. Circular references: Views cannot reference themselves or create circular dependencies
  4. Read-only: You cannot INSERT, UPDATE, or DELETE on views
  5. No DDL operations: You cannot run DDL operations (like RENAME TABLE) on views
  6. No column-level permissions: Unlike tables, you cannot grant permissions on individual view columns (Enterprise)