Materialized View

Materialized Views are an experimental feature introduced in Apache Cassandra 3.0 that provide a mechanism for server-side denormalization from a base table into a view that is updated when the base table is updated. SchemaBuilder offers API methods for creating, altering and dropping materialized views.

Creating a Materialized View (CREATE MATERIALIZED VIEW)

To start a CREATE MATERIALIZED VIEW query, use createMaterializedView in SchemaBuilder:

import static com.datastax.oss.driver.api.querybuilder.SchemaBuilder.*;

CreateMaterializedViewStart create = createMaterializedView("cycling", "cyclist_by_age");

Like all other CREATE queries, one may supply ifNotExists() to require that the view should only be created if it doesn’t already exist, i.e.:

CreateMaterializedViewStart create = createMaterializedView("cycling", "cyclist_by_age").ifNotExists();

There are a number of steps that must be executed to complete a materialized view:

  • Specify the base table using asSelectFrom
  • Specify the columns to include in the view via column or columns
  • Specify the where clause using relations
  • Specify the partition key columns using withPartitionKey and withClusteringColumn

For example, the following defines a complete CREATE MATERIALIZED VIEW statement:

createMaterializedView("cycling", "cyclist_by_age")
    .asSelectFrom("cycling", "cyclist")
    .columns("age", "name", "country")
    .whereColumn("age")
    .isNotNull()
    .whereColumn("cid")
    .isNotNull()
    .withPartitionKey("age")
    .withClusteringColumn("cid");
// CREATE MATERIALIZED VIEW cycling.cyclist_by_age AS
// SELECT age,name,country FROM cycling.cyclist WHERE age IS NOT NULL AND cid IS NOT NULL PRIMARY KEY(age,cid)

Please note that not all WHERE clause relations may be compatible with materialized views.

Like a table, one may additionally provide configuration such as clustering order, compaction options and so on. Refer to RelationStructure for documentation on additional configuration that may be provided for a view.

Altering a Materialized View (ALTER MATERIALIZED VIEW)

To start a ALTER MATERIALIZED VIEW query, use alterMaterializedView:

alterMaterializedView("cycling", "cyclist_by_age");

Unlike a table, you may not alter, drop or rename columns on a materialized view. Instead, one may only alter the options defined in RelationStructure. For example:

alterMaterializedView("cycling", "cyclist_by_age")
    .withGcGraceSeconds(0)
    .withCaching(true, RowsPerPartition.NONE);
// ALTER MATERIALIZED VIEW cycling.cyclist_by_age WITH gc_grace_seconds=0 AND caching={'keys':'ALL','rows_per_partition':'NONE'}

Dropping a View (DROP MATERIALIZED VIEW)

To create a DROP MATERIALIZED VIEW query, use dropMaterializedView:

dropMaterializedView("cycling", "cyclist_by_age");
// DROP MATERIALIZED VIEW cycling.cyclist_by_age

You may also specify ifExists:

dropTable("cyclist_by_age").ifExists();
// DROP MATERIALIZED VIEW IF EXISTS cyclist_by_age