CREATE TABLE AS SELECT

Syntax

CREATE TABLE [IF NOT EXISTS] table_name
[WITH (table_parameter = value [, ... ])] 
AS select_statement;

Description

CREATE TABLE AS is a statement that:

• Generates a DDL statement to create a new Table.

• Launches a new query to write the results of the SELECT statement into the newly created table.

Note If you specify IF NOT EXISTS and the target table relation already exists, this statement is a no-op and no query is launched.

Arguments

table_name

This specifies the name of the new table. Optionally, use <database_name>.<schema_name> as the prefix to the name to create the relation in that namespace. If the name is case-sensitive, you must wrap it in double quotes; otherwise, the system uses the lower case name.

WITH (<table_parameter> = <value> [, …​ ])

Optionally, this clause specifies Table Parameters.

select_statement

This statement specifies the SELECT statement to run.

Table Parameters

Parameter Name	Description
store	The name of the store that hosts the entity for this stream. Required: No Default value: User’s default Data Store. Type: String Valid values: See LIST STORES

Snowflake-Specific Parameters

Parameter Name	Description
snowflake.db.name	The name of the Snowflake database that would host the Snowflake Table. Required: Yes Default values: None Type: String Valid values: Database names available from LIST ENTITIES.
snowflake.schema.name	The name of the Snowflake schema that would host the Snowflake Table. Required: Yes Default value: None Type: String Valid values: Schema names available from LIST ENTITIES under the snowflake.db.name database name.
snowflake.table.name	The name of the Snowflake table to use when creating the Table. If the table doesn't exist in the store, a table with the table_name is created in the corresponding store. Required: No Default value: table_name Type: String
snowflake.buffer.millis	The amount of time to buffer events with the sink Table before writing the data into the corresponding Snowflake table. It is important to note that the longer events are buffered, the larger the internal state of the query gets. Depending on the volume and size of the events for the corresponding query, memory limitations may be reached. Required: No Default value: 1000 Type: Long Valid values: (0, ...]

Snowflake stores provide a delivery guarantee of at_least_once when producing events into a sink Table. Snowflake stores provide an insert-only mode when writing to Snowflake tables.

Databricks-Specific Parameters

Parameter Name	Description
databricks.catalog.name	The name of the Databricks catalog that would host the Databricks Table. Required: Yes Default values: None Type: String Valid values: Database names available from LIST ENTITIES.
databricks.schema.name	The name of the Databricks schema that would host the Databricks Table. Required: Yes Default value: None Type: String Valid values: Schema names available from LIST ENTITIES under the databricks.catalog.name catalog name.
databricks.table.name	The name of the Databricks table to use when creating the Table. A table with the table_name is created in the corresponding store. If that table already exists, then an error is reported back to the user. Required: No Default value: table_name Type: String
table.data.file.location	The S3 directory location for the Delta-formatted data to be written. This location is either a full S3 path, or a relative path to the data store's Cloud Provider Bucket (see CREATE STORE). The credentials for writing to S3 are given during data store creation (see CREATE STORE). Note that the S3 bucket from the location specified by this parameter must match the databricks.cloud.s3.bucket property defined in the data store. Required: Yes Default value: None Type: String

Databricks stores provide a delivery guarantee of exactly_once when producing events into a sink Table. Databricks stores provide an insert-only mode when writing to Databricks tables.

Postgres-Specific Parameters

Parameter Name	Description
postgresql.db.name	Required: Yes See CDC source parameters for the same parameter name
postgresql.schema.name	Required: Yes See CDC source parameters for the same parameter name
postgresql.table.name	Required: No See CDC source parameters for the same parameter name

Iceberg-Specific Parameters

Parameter Name	Description
iceberg.aws.glue.db.name	The name of the database (namespace) in the AWS Glue instance (used as the Iceberg Catalog implementation in the Iceberg Store) containing the existing Iceberg table. Required: Yes Type: String
iceberg.aws.glue.table.name	The name of the existing Iceberg table in the AWS Glue instance (used as the Iceberg Catalog implementation in the Iceberg Store). Required: Yes Type: String

S3-Specific Parameters

The S3 sink writes the results of a query to an S3 store as newline-delimited JSON (the jsonl format which is one JSON object per record). Records are written as part files under the configured s3.uri path and rolled into new files based on a configurable rolling policy (elapsed time, part-file size, and how often the sink checks whether a part file should be rolled). When neither a rollover interval nor a file size is set, the connector's built-in default rolling policy applies. The S3 sink is append-only: records are only added and never updated or deleted. Optionally, a PARTITION BY clause lays out the written objects into Hive-style partition directories (column=value/); partition columns are used only for the directory layout and are not included in the JSON payload. All access to S3 uses the credentials and region configured on the associated store.

Parameter Name	Description
s3.uri	The S3 URI, using the canonical s3:// scheme, that the query writes its results to. It must reference a sub-folder of the store's URI. The s3a:// and s3n:// schemes are not accepted. Required: Yes Default value: None Type: String
value.format	Format of the records written to S3. Records are written as newline-delimited JSON. When omitted, the format is inferred from the source relation; an error is reported if the source has no value format. Required: No Default value: Inferred from the source relation Type: String Valid values: json, jsonl (jsonl is treated as json)
s3.sink.rolling.policy.rollover.interval	The maximum amount of time a part file can stay open before it is rolled into a new file. Required: No Default value: The connector's default rolling policy applies when neither this property nor s3.sink.rolling.policy.file.size is set. Type: String (duration, e.g. '15 min')
s3.sink.rolling.policy.file.size	The maximum size a part file can reach before it is rolled into a new file. Required: No Default value: The connector's default rolling policy applies when neither this property nor s3.sink.rolling.policy.rollover.interval is set. Type: String (memory size, e.g. '128 mb')
s3.sink.rolling.policy.check.interval	How often the sink checks whether an open part file should be rolled based on the rollover interval. Required: No Default value: The connector's default rolling policy applies. Type: String (duration, e.g. '5 min')

Examples — Snowflake

Create a copy of a stream in a Snowflake table

The following creates a replica of the source Stream, pageviews in the Snowflake Table, PV_TABLE:

CREATE TABLE "PV_TABLE" WITH (
  'store' = 'sfstore',
  'snowflake.db.name' = 'DELTA_STREAMING',
  'snowflake.schema.name' = 'PUBLIC'
) AS SELECT * FROM pageviews;

Create a stream of changes for a changelog in a Snowflake table

The following CTAS query creates a new Snowflake Table to store incremental changes resulting from a grouping aggregation on the transactions Stream:

CREATE TABLE "CC_TYPE_USAGE" WITH (
  'store' = 'sfstore',
  'snowflake.db.name' = 'DELTA_STREAMING',
  'snowflake.schema.name' = 'PUBLIC'
) AS SELECT
  cc_type AS "CREDIT_TYPE",
  tx_time AS "TRANSACTION_TS",
  tx_id AS "TRANSACTION_ID"
FROM transactions
GROUP BY cc_type;

This query stores insert and update changes to the grouping column cc_type to the sink table CC_TYPE_USAGE.

Examples — PostgreSQL

Assumptions for the following 2 examples:

• pageviews is a stream you have already defined.

• ps_store is a PostgreSQL data store you have already created.

  • For details and an example, see CREATE STORE.

Create a new table in a PostgresSQL data store — example 1

In this example, you create a new table, named pageviews, in the given PostgreSQL data store under the public schema. The table has 3 columns:

1. viewtime

2. uid

3. pageid.

The query writes its results into this table as its sink.

CREATE TABLE pg_table WITH (
   'store'='ps_store', 
   'postgresql.db.name'='demo', 
   'postgresql.schema.name'='public', 
   'postgresql.table.name'='pageviews') AS 
SELECT viewtime, userid AS uid, pageid 
FROM pageviews;

Create a new table in a PostgresSQL data store — example 2

This query creates a new table, named pagevisits, in the given PostgreSQL data store under the public schema. The table has 2 columns:

1. userid

2. cnt

The query writes its results into this table.

CREATE TABLE pagevisits WITH (
       'store'='ps_store', 
       'postgresql.db.name'='demo', 
       'postgresql.schema.name'='public') AS 
SELECT userid, count(*) AS cnt 
FROM pageviews 
GROUP BY userid;

Note Since the query includes a GROUP BY clause, the userid column — which is the grouping column — is considered the primary key for the results. The pagevisits table in PostgreSQL is created accordingly, with userid as its primary column.

Examples — Databricks

Create a copy of a stream in a Databricks table

The following creates a replica of the source Stream, pageviews in the Databricks Table, pageviews_db:

CREATE TABLE pageviews_db WITH (
  'store' = 'databricks_store', 
  'databricks.catalog.name' = 'catalog1', 
  'databricks.schema.name' = 'schema1', 
  'databricks.table.name' = 'pageviews', 
  'table.data.file.location' = 's3://mybucket/test/0/pageviews'
) AS SELECT * FROM pageviews;

Upon issuing this query, a Databricks table is created in catalog1.schema1.pageviews that uses s3://mybucket/test/0/pageviews as its external location. This query writes the Delta-formatted parquet files and updates the Delta log in that S3 location.

Examples — S3

Create a S3 Table

The following writes the results of a query over the pageviews Database to the S3 store s3_store as newline-delimited JSON under the s3://bucket_name/ctas_output/ path:

CREATE TABLE pageviews_s3 WITH (
  'store' = 's3_store',
  's3.uri' = 's3://bucket_name/ctas_output/',
  'value.format' = 'json'
) AS SELECT viewtime, userid AS "UID", pageid 
     FROM pageviews;

Create a partitioned S3 Table

In this example, a partitioned Table is created in the pgv_parts bucket in s3_store . The written objects are partitioned by userid into Hive-style directories:

CREATE TABLE pageviews_s3_part WITH (
  'store' = 's3_store',
  's3.uri' = 's3://bucket_name/pgv_parts/',
  'value.format' = 'json'
) AS SELECT viewtime, userid, pageid 
     FROM pageviews 
     PARTITION BY userid;