Create a new stream with the specified columns and properties.
If the IF NOT EXISTS clause is present, the statement won't fail if a stream with the same name already exists.
A stream can be created as read-only if the SOURCE clause is provided. INSERTS statements and the DELETE TOPIC on DROP STREAM statements aren't permitted. Source streams do not support pull queries. Only source tables support running pull queries on them.
To disable the SOURCE stream feature, set
false in your ksqlDB server properties file.
Assign the PARTITIONS property in the WITH clause to specify the number of partitions in the stream's backing topic. Partitioning streams and tables is especially important for stateful or otherwise intensive queries. For more information, see Parallelization.
KEY and VALUE columns¶
A ksqlDB STREAM is a stream of facts. Each fact is immutable and unique.
A stream can store its data in either
VALUE columns can be NULL. No special processing is done if two rows have the same
key. This situation is handled differently by ksqlDB TABLEs, as shown in the following table.
|Key column type||
|NON NULL key constraint||No||Yes
Records in the Kafka topic with a NULL
|Unique key constraint||No
Messages with the same key as another have no special meaning.
Later messages with the same key replace earlier.
Messages with NULL values are ignored.
NULL message values are treated as a tombstone
Any existing row with a matching key is deleted.
Each column is defined by:
column_name: the name of the column. If unquoted, the name must be a valid SQL identifier and ksqlDB converts it to uppercase. The name can be quoted if case needs to be preserved or if the name is not a valid SQL identifier, for example
data_type: the SQL type of the column. Columns can be any of the data types supported by ksqlDB.
KEY: columns that are stored in the Kafka message's key should be marked as
KEYcolumns. If a column is not marked as a
KEYcolumn, ksqlDB loads it from the Kafka message's value. Unlike a table's
PRIMARY KEY, a stream's keys can be NULL.
Each row within the stream has a
ROWTIME pseudo column, which represents the event time
of the row. The timestamp has milliseconds accuracy. The timestamp is used by ksqlDB during any
windowing operations and during joins, where data from each side of a join is generally processed
in time order.
ROWTIME is populated from the corresponding Kafka message timestamp. Set
WITH clause to populate
ROWTIME from a column in the Kafka message key or value.
ROWPARTITION and ROWOFFSET¶
ROWOFFSET are pseudo columns. They represent the
partition and offset of the source topic. For example, if issuing a push query
ROWOFFSET in the SELECT clause on a stream or table backed
by topic x, the push query's projection will contain the partition and offset information
of the underlying records in topic x.
Specify details about your stream by using the WITH clause, which supports the following properties:
|KAFKA_TOPIC (required)||The name of the Kafka topic that backs this source. The topic must either already exist in Kafka, or PARTITIONS must be specified to create the topic. Command will fail if the topic exists with different partition/replica counts.|
|KEY_FORMAT||Specifies the serialization format of the message key in the topic. For supported formats, see Serialization Formats.
If not supplied, the system default, defined by ksql.persistence.default.format.key, is used. If the default is also not set the statement will be rejected as invalid.
|VALUE_FORMAT||Specifies the serialization format of the message value in the topic. For supported formats, see Serialization Formats.
If not supplied, the system default, defined by ksql.persistence.default.format.value, is used. If the default is also not set the statement will be rejected as invalid.
|FORMAT||Specifies the serialization format of both the message key and value in the topic. It is not valid to supply this property alongside either
|PARTITIONS||The number of partitions in the backing topic. This property must be set if creating a STREAM without an existing topic (the command will fail if the topic does not exist). You can't change the number of partitions on a stream. To change the partition count, you must drop the stream and create it again.|
|REPLICAS||The number of replicas in the backing topic. If this property is not set but PARTITIONS is set, then the default Kafka cluster configuration for replicas will be used for creating a new topic.|
|VALUE_DELIMITER||Used when VALUE_FORMAT='DELIMITED'. Supports single character to be a delimiter, defaults to ','. For space and tab delimited values you must use the special values 'SPACE' or 'TAB', not an actual space or tab character.|
|TIMESTAMP||By default, the pseudo
|TIMESTAMP_FORMAT||Used in conjunction with TIMESTAMP. If not set, ksqlDB timestamp column must be of type
|WRAP_SINGLE_VALUE||Controls how values are deserialized where the value schema contains only a single column. The setting controls how ksqlDB will deserialize the value of the records in the supplied
If set to
If set to
If not supplied, the system default, defined by ksql.persistence.wrap.single.values and defaulting to
Note: Supplying this property for formats that do not support wrapping, for example
|WINDOW_TYPE||By default, the topic is assumed to contain non-windowed data. If the data is windowed, i.e., was created using ksqlDB using a query that contains a
|WINDOW_SIZE||By default, the topic is assumed to contain non-windowed data. If the data is windowed, i.e., was created using ksqlDB using a query that contains a
For more information on timestamp formats, see DateTimeFormatter.
- To use Avro or Protobuf, you must have Schema Registry enabled and
ksql.schema.registry.urlmust be set in the ksqlDB Server configuration file. See Configure ksqlDB for Avro, Protobuf, and JSON schemas.
- Avro and Protobuf field names are not case sensitive in ksqlDB. This matches the ksqlDB column name behavior.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33