CREATE VIEW creates a materialized view, which lets you retrieve incrementally
updated results of a
SELECT query very quickly. Despite the simplicity of
creating a view, it’s Materialize’s most powerful feature.
CREATE VIEW computes and maintains the results of a
SELECT query in memory,
which is similar to a materialized view in the language of RDBMSes. (For more
information about materialized views, see What is
This means that as data streams in from Kafka, Materialize incrementally updates
the results of the view’s
SELECT statement. Because these results are
available in memory, you will have incredibly fast access to the query’s result
|view_name||A name for the view.|
When creating a view, Materialize’s internal Differential dataflow engine
creates a persistent dataflow for the corresponding
SELECT statement. All of
the data that is available from the view’s source’s arrangements (which is
similar to an index in the language of RDBMSes) is then used to create the first
result set of the view.
As data continues to stream in from Kafka, Differential passes the data to the appropriate dataflows, which are then responsible for making any necessary updates to maintain the views you’ve defined.
When reading from a view (e.g.
SELECT * FROM some_view), Materialize simply
returns the current result set for the persisted dataflow.
Views are maintained in memory. Because of this, one must be sure that all intermediate stages of the query, as well as its result set can fit in the memory of a single machine, while also understanding the rate at which the query’s result set will grow.
For more detail about how different clauses impact memory usage, check out our
CREATE VIEW purchase_sum_by_region AS SELECT sum(purchase.amount) AS region_sum, region.id AS region_id FROM mysql_simple_region AS region INNER JOIN mysql_simple_user AS user ON region.id = user.region_id INNER JOIN mysql_simple_purchase AS purchase ON purchase.user_id = user.id GROUP BY region.id;
In this example, as new users or purchases come in, the results of the view are
incrementally updated. For example, if a new purchase comes in for a specific
user, the underlying dataflow will determine which region that user belongs to,
and then increment the
region_sum field with those results.