PolarDB:Real-time materialized views

Terms

• Base table: a common table used in the definition of a materialized view.

• Delta: a collection of data that is added or removed when the data in the base table changes, compared with the data in the materialized view.

• Refresh: maintains a materialized view so that the data in the materialized view is consistent with the data obtained by querying the current base table based on the view definition.

• Apply Delta: inserts or deletes calculated incremental data from a real-time materialized view to maintain data consistency between the real-time materialized view and the base table.

Limits

Performance degradation

• Real-time materialized views greatly improve query performance but have a significant impact on the write performance of the base tables. If the number of read operations is greater than the number of write operations, we recommend that you use real-time materialized views.

• The impact of real-time materialized views on the write performance of the base tables depends on factors such as the view definitions and the write loads, structures, and indexes of the base tables. Before you create real-time materialized views in the production environment, we recommend that you first test the impact of real-time materialized views on the write performance of the base tables in the test environment. You can use real-time materialized views in the production environment if the write performance meets the requirements.

• The following methods can be used to reduce the costs of real-time materialized views:

  • Create a few real-time materialized views on the same base table.

  • Batch write data to the base table. For example, you can execute the COPY or INSERT INTO SELECT statement to batch import data.

  • Create primary keys for all base tables, and include the primary keys for all base tables in the projected column for the definitions of real-time materialized views.

How it works

• Create a real-time materialized view

  • Rewrite the query of the materialized view and calculate the hidden columns required to maintain the real-time materialized view.

  • Create a trigger for the base table to refresh the real-time materialized view.

  • Create unique indexes for the real-time materialized view when specified conditions are met to accelerate delta refresh.

• Refresh a real-time materialized view

  • Data changes in the base table activate the trigger.

  • Obtain incremental data from the base table by using the trigger.

  • Calculate the delta of the real-time materialized view based on the definitions and the incremental data of the current base table.

  • Apply the calculated incremental data to the real-time materialized view to implement delta refresh.

• Delete a real-time materialized view

  • Delete the delta refresh trigger from the base table of the real-time materialized view.

  • Delete the real-time materialized view.

Usage notes

• Prerequisites

  Create the polar_ivm extension in the database to use real-time materialized views.

  CREATE EXTENSION polar_ivm WITH SCHEMA pg_catalog;

• Create a real-time materialized view

  CREATE MATERIALIZED VIEW table_name[ (column_name [, ...] ) ]
  REFRESH FAST
  ON COMMIT
  AS query
  [ WITH [ NO ] DATA ]

  The following table describes the parameters.

  Parameter	Description
  table_name	The name of the real-time materialized view that you want to create, which can be schema-qualified.
  column_name	The name of a column in the real-time materialized view. If you do not specify column names, the materialized view uses the column names from the result set of the query specified in the AS clause.
  WITH DATA	Populates data immediately after the view is created. This is the default option.
  WITH NO DATA	Creates the view, but populates data later such as upon the first refresh. When you query the view, no error is reported. However, no data is returned until you execute the REFRESH MATERIALIZED VIEW statement on the view.
  query	The SQL query whose results are stored in the materialized view, which can be a SELECT, TABLE, or VALUES statement. The query runs in a secure and restricted operation.

• Incrementally refresh a real-time materialized view

  REFRESH MATERIALIZED VIEW table_name

  Note

  • table_name: the name of the real-time materialized view that you want to refresh.

  • You do not need to manually refresh a real-time materialized view that is created by using the WITH DATA parameter.

  • After you refresh a real-time materialized view that is created by using the WITH NO DATA parameter, the view is populated with data based on the view definition, and subsequent modifications made to the base table are updated to the view in real time.

• Delete a real-time materialized view

  DROP MATERIALIZED VIEW [ IF EXISTS ] table_name [, ...] [ CASCADE | RESTRICT ]

  The following table describes the parameters.

  Parameter	Description
  IF EXISTS	If the real-time materialized view does not exist, a prompt, instead of an error, is returned.
  table_name	The name of the real-time materialized view that you want to delete, which can be schema-qualified.
  CASCADE	Automatically deletes the objects that depend on the view, such as other materialized views or regular views, and all objects that depend on the automatically deleted objects.
  RESTRICT	If an object depends on the real-time materialized view, the view is not deleted. By default, this parameter is used.

Examples

1. Install the polar_ivm extension.

   CREATE EXTENSION IF NOT EXISTS polar_ivm WITH SCHEMA pg_catalog ;

2. Create a base table and insert data into the table.

   CREATE TABLE t( a INT, b VARCHAR);
   INSERT INTO t VALUES
     (1,'a'),
     (2,'b'),
     (3,'c'),
     (4,'d'),
     (5,'e');

3. Create a real-time materialized view.

   CREATE MATERIALIZED VIEW mv
   REFRESH FAST
   ON COMMIT
   AS
   SELECT max(a),min(a),b FROM t GROUP BY b;

4. Execute a DML statement on the base table.

   • Query the data of the real-time materialized view.

     SELECT * FROM mv ORDER BY b;

     Sample result:

      max | min | b
     -----+-----+---
        1 |   1 | a
        2 |   2 | b
        3 |   3 | c
        4 |   4 | d
        5 |   5 | e
     (5 rows)

     The result shows that the data of the real-time materialized view is consistent with the data of the base table.

   • Insert new data into the base table and query the data of the real-time materialized view.

     INSERT INTO t VALUES(6,'f');
     
     SELECT * FROM mv ORDER BY b;

     Sample result:

      max | min | b
     -----+-----+---
        1 |   1 | a
        2 |   2 | b
        3 |   3 | c
        4 |   4 | d
        5 |   5 | e
        6 |   6 | f
     (6 rows)

     The result shows that the data of the real-time materialized view is consistent with the data of the base table.

   • Delete specific data from the base table and query the data of the real-time materialized view.

     DELETE FROM t WHERE a = 2;
     
     SELECT * FROM mv ORDER BY b;

     Sample result:

      max | min | b
     -----+-----+---
        1 |   1 | a
        3 |   3 | c
        4 |   4 | d
        5 |   5 | e
        6 |   6 | f
     (5 rows)

     The result shows that the data of the real-time materialized view is consistent with the data of the base table.

   • Update the data of the base table and query the data of the real-time materialized view.

     UPDATE t SET a = a + 1;
     
     SELECT * FROM mv ORDER BY b;

     Sample result:

      max | min | b
     -----+-----+---
        2 |   2 | a
        4 |   4 | c
        5 |   5 | d
        6 |   6 | e
        7 |   7 | f
     (5 rows)

     The result shows that the data of the real-time materialized view is consistent with the data of the base table.

5. Delete the real-time materialized view.

   DROP MATERIALIZED VIEW mv;