CREATE VIEW

The CREATE VIEW statement creates a new , which is a stored query represented as a virtual table.

By default, views created in a database cannot reference objects in a different database. To enable cross-database references for views, set the sql.cross_db_views.enabled to true.

The CREATE VIEW statement performs a schema change. For more information about how online schema changes work in CockroachDB, see .

Required privileges

The user must have the CREATE on the parent database and the SELECT privilege on any table(s) referenced by the view.

Synopsis

Parameters

Parameter	Description
MATERIALIZED	Create a .
IF NOT EXISTS	Create a new view only if a view of the same name does not already exist. If one does exist, do not return an error. Note that IF NOT EXISTS checks the view name only. It does not check if an existing view has the same columns as the new view.
OR REPLACE	Create a new view if a view of the same name does not already exist. If a view of the same name already exists, replace that view. In order to replace an existing view, the new view must have the same columns as the existing view, or more. If the new view has additional columns, the old columns must be a prefix of the new columns. For example, if the existing view has columns a, b, the new view can have an additional column c, but must have columns a, b as a prefix. In this case, CREATE OR REPLACE VIEW myview (a, b, c) would be allowed, but CREATE OR REPLACE VIEW myview (b, a, c) would not.
view_name	The name of the view to create, which must be unique within its database and follow these . When the parent database is not set as the default, the name must be formatted as database.name.
name_list	An optional, comma-separated list of column names for the view. If specified, these names will be used in the response instead of the columns specified in AS select_stmt.
WITH (security_invoker = true)	Sets whether to check privileges on the underlying tables as the querying user (true) or the view owner (false). WITH (security_invoker) is equivalent to WITH (security_invoker = true). This option applies only to non-materialized views.
AS select_stmt	The to execute when the view is requested. Note that it is not currently possible to use * to select all columns from a referenced table or view; instead, you must specify specific columns.
AS OF SYSTEM TIME	When used with CREATE MATERIALIZED VIEW, populates the materialized view using historical data. This can reduce by leveraging . The timestamp must be within the . For more information, see .
opt_temp	Defines the view as a session-scoped temporary view. For more information, see . Support for temporary views is .

Example

This example highlights one key benefit to using views: simplifying complex queries. For additional benefits and examples, see .

Setup

The following examples use the .

To follow along, run to start a temporary, in-memory cluster with the startrek schema and dataset preloaded:

$ cockroach demo startrek

Create a view

The sample startrek database contains two tables, episodes and quotes. The table also contains a foreign key constraint, between the episodes.id column and the quotes.episode column. To count the number of famous quotes per season, you could run the following join:

> SELECT startrek.episodes.season, count(*)
  FROM startrek.quotes
  JOIN startrek.episodes
  ON startrek.quotes.episode = startrek.episodes.id
  GROUP BY startrek.episodes.season;

  season | count
---------+--------
       1 |    78
       2 |    76
       3 |    46
(3 rows)

Alternatively, to make it much easier to run this complex query, you could create a view:

> CREATE VIEW startrek.quotes_per_season (season, quotes)
  AS SELECT startrek.episodes.season, count(*)
  FROM startrek.quotes
  JOIN startrek.episodes
  ON startrek.quotes.episode = startrek.episodes.id
  GROUP BY startrek.episodes.season;

The view is then represented as a virtual table alongside other tables in the database:

> SHOW TABLES FROM startrek;

  schema_name |    table_name     | type  | estimated_row_count
--------------+-------------------+-------+----------------------
  public      | episodes          | table |                  79
  public      | quotes            | table |                 200
  public      | quotes_per_season | view  |                   3
(3 rows)

Executing the query is as easy as SELECTing from the view, as you would from a standard table:

> SELECT * FROM startrek.quotes_per_season;

  season | quotes
---------+---------
       3 |     46
       1 |     78
       2 |     76
(3 rows)

Replace an existing view

You can create a new view, or replace an existing view, with CREATE OR REPLACE VIEW:

> CREATE OR REPLACE VIEW startrek.quotes_per_season (season, quotes)
  AS SELECT startrek.episodes.season, count(*)
  FROM startrek.quotes
  JOIN startrek.episodes
  ON startrek.quotes.episode = startrek.episodes.id
  GROUP BY startrek.episodes.season
  ORDER BY startrek.episodes.season;

> SELECT * FROM startrek.quotes_per_season;

  season | quotes
---------+---------
       3 |     46
       1 |     78
       2 |     76
(3 rows)

Create a view with invoker privileges

To make a view check privileges on the underlying tables as the querying user, include the :

CREATE VIEW user_accounts WITH (security_invoker) AS
  SELECT type, email
  FROM accounts;

To restore the default behavior later, use .

Create a view that references routines

Views can call both scalar and set-returning in their SELECT statements.

The following example builds a view over a table and two UDFs.

Create and populate a table:

CREATE TABLE xy (x INT, y INT);
INSERT INTO xy VALUES (1, 2), (3, 4), (5, 6);

Define a scalar and a set-returning UDF:

CREATE FUNCTION f_scalar() RETURNS INT LANGUAGE SQL AS $$
    SELECT count(*) FROM xy;
  $$;

CREATE FUNCTION f_setof() RETURNS SETOF xy LANGUAGE SQL AS $$
    SELECT * FROM xy;
  $$;

Create a view that references both functions:

CREATE VIEW v_xy AS
SELECT x, y, f_scalar() AS total_rows
FROM f_setof();

Query the view:

SELECT * FROM v_xy ORDER BY x;

  x | y | total_rows
----+---+-------------
  1 | 2 |          3
  3 | 4 |          3
  5 | 6 |          3
(3 rows)

Because the view depends on f_scalar and f_setof, attempting to rename either function returns an error:

ALTER FUNCTION f_scalar RENAME TO f_scalar_renamed;

ERROR: cannot rename function "f_scalar" because other functions or views ([movr.public.v_xy]) still depend on it
SQLSTATE: 0A000

Create a materialized view with historical data using AS OF SYSTEM TIME

You can create a materialized view using historical data with the clause. This is useful for reducing by performing a when populating the view.

Historical data is available only within the .

The following example creates a materialized view using the most recent data that is available for :

CREATE MATERIALIZED VIEW overdrawn_accounts
  AS SELECT id, balance
  FROM bank
  WHERE balance < 0
  AS OF SYSTEM TIME follower_read_timestamp();

You can also specify an explicit timestamp:

CREATE MATERIALIZED VIEW overdrawn_accounts
  AS SELECT id, balance
  FROM bank
  WHERE balance < 0
  AS OF SYSTEM TIME '-10s';

See also