CREATE SOURCE

A source describes an external system you want Materialize to read data from, and provides details about how to decode and interpret that data. To create a source, you must specify a connector, a format and an envelope. Like other relations, sources are namespaced by a database and schema.

Connectors

Materialize bundles native connectors that allow ingesting data from the following external systems:

For details on the syntax, supported formats and features of each connector, check out the dedicated CREATE SOURCE documentation pages.

Sample data

To get started with no external dependencies, you can use the load generator source to produce sample data that is suitable for demo and performance test scenarios.

Formats

To read from an external data source, Materialize must be able to determine how to decode raw bytes from different formats into data structures it can understand at runtime. This is handled by specifying a FORMAT in the CREATE SOURCE statement.

Avro

Syntax: FORMAT AVRO

Materialize can decode Avro messages by integrating with a schema registry to retrieve a schema, and automatically determine the columns and data types to use in the source.

Schema versioning

The latest schema is retrieved using the TopicNameStrategy strategy at the time the CREATE SOURCE statement is issued. In the future, we expect to support specifying a different subject name strategy (#6170).

Schema evolution

As long as the writer schema changes in a compatible way, Materialize will continue using the original reader schema definition by mapping values from the new to the old schema version. To use the new version of the writer schema in Materialize, you need to drop and recreate the source.

Name collision

To avoid case-sensitivity conflicts with Materialize identifiers, we recommend double-quoting all field names when working with Avro-formatted sources.

Supported types

Materialize supports all Avro types, except for recursive types (#5803) and union types in arrays (#8917).

JSON

Syntax: FORMAT JSON

Materialize can decode JSON messages into a single column named data with type jsonb. Refer to the jsonb type documentation for the supported operations on this type.

If your JSON messages have a consistent shape, consider creating a view that extracts the fields of the JSON into columns:

-- extract jsonb into typed columns
CREATE VIEW my_typed_source AS
  SELECT
    (data->>'field1')::boolean AS field_1,
    (data->>'field2')::int AS field_2,
    (data->>'field3')::float AS field_3
  FROM my_jsonb_source;

Schema registry integration

Retrieving schemas from a schema registry is not supported yet for JSON-formatted sources (#7186). This means that Materialize cannot decode messages serialized using the JSON Schema serialization format (JSON_SR).

Protobuf

Syntax: FORMAT PROTOBUF

Materialize can decode Protobuf messages by integrating with a schema registry or parsing an inline schema to retrieve a .proto schema definition. It can then automatically define the columns and data types to use in the source. Unlike Avro, Protobuf does not serialize a schema with the message, so Materialize expects:

• A FileDescriptorSet that encodes the Protobuf message schema. You can generate the FileDescriptorSet with protoc, for example:

  protoc --include_imports --descriptor_set_out=SCHEMA billing.proto
  

• A top-level message name and its package name, so Materialize knows which message from the FileDescriptorSet is the top-level message to decode, in the following format:

  <package name>.<top-level message>
  

  For example, if the FileDescriptorSet were from a .proto file in the billing package, and the top-level message was called Batch, the message_name value would be billing.Batch.

Schema versioning

The latest schema is retrieved using the TopicNameStrategy strategy at the time the CREATE SOURCE statement is issued. In the future, we expect to support specifying a different subject name strategy (#6170).

Schema evolution

As long as the .proto schema definition changes in a compatible way, Materialize will continue using the original schema definition by mapping values from the new to the old schema version. To use the new version of the schema in Materialize, you need to drop and recreate the source.

Supported types

Materialize supports all well-known Protobuf types from the proto2 and proto3 specs, except for recursive Struct values (#5803) and map types.

Multiple message schemas

When using a schema registry with Protobuf sources, the registered schemas must contain exactly one Message definition. In the future, we expect to support schemas with multiple messages (#9598).

Text/bytes

Text

Syntax: FORMAT TEXT

Materialize can parse new-line delimited data as plain text. Data is assumed to be valid unicode (UTF-8), and discarded if it cannot be converted to UTF-8. Text-formatted sources have a single column, by default named text.

For details on casting, check the text documentation.

Bytes

Syntax: FORMAT BYTES

Materialize can read raw bytes without applying any formatting or decoding. Raw byte-formatted sources have a single column, by default named data.

For details on encodings and casting, check the bytea documentation.

CSV

Syntax: FORMAT CSV

Materialize can parse CSV-formatted data using different methods to determine the number of columns to create and their respective names:

The data in CSV sources is read as text. You can then handle the conversion to other types using explicit casts when creating views.

Invalid rows

Any row that doesn’t match the number of columns determined by the format is ignored, and Materialize logs an error.

Envelopes

In addition to determining how to decode incoming records, Materialize also needs to understand how to interpret them. Whether a new record inserts, updates, or deletes existing data in Materialize depends on the ENVELOPE specified in the CREATE SOURCE statement.

Append-only envelope

Syntax: ENVELOPE NONE

The append-only envelope treats all records as inserts. This is the default envelope, if no envelope is specified.

Upsert envelope

Syntax: ENVELOPE UPSERT

The upsert envelope treats all records as having a key and a value, and supports inserts, updates and deletes within Materialize:

• If the key does not match a preexisting record, it inserts the record’s key and value.

• If the key matches a preexisting record and the value is non-null, Materialize updates the existing record with the new value.

• If the key matches a preexisting record and the value is null, Materialize deletes the record.

Debezium envelope

Syntax: ENVELOPE DEBEZIUM

Materialize provides a dedicated envelope to decode messages produced by Debezium. This envelope treats all records as change events with a diff structure that indicates whether each record should be interpreted as an insert, update or delete within Materialize:

• If the before field is null, the record represents an upstream create event and Materialize inserts the record’s key and value.

• If the before and after fields are non-null, the record represents an upstream update event and Materialize updates the existing record with the new value.

• If the after field is null, the record represents an upstream delete event and Materialize deletes the record.

Materialize expects a specific message structure that includes the row data before and after the change event, which is not guaranteed for every Debezium connector. For more details, check the Debezium integration guide.

Truncation

The Debezium envelope does not support upstream truncate events (#6596).

Debezium metadata

The envelope exposes the before and after value fields from change events. In the future, we expect to support additional metadata with information about the original context of the events, like source.ts_ms, source.database and source.table (#12077).

Duplicate handling

Debezium may produce duplicate records if the connector is interrupted. Materialize makes a best-effort attempt to detect and filter out duplicates.

Best practices

Sizing a source

Some sources are low traffic and require relatively few resources to handle data ingestion, while others are high traffic and require hefty resource allocations. You choose the amount of CPU, memory, and disk available to a source using the SIZE option, and adjust the provisioned size after source creation using the ALTER SOURCE command.

It’s a good idea to size up a source when:

• You want to increase throughput. Larger sources will typically ingest data faster, as there is more CPU available to read and decode data from the upstream external system.

• You are using the upsert envelope or Debezium envelope, and your source contains many unique keys. These envelopes maintain state proportional to the number of unique keys in the upstream external system. Larger sizes can store more unique keys.

  In this case, it’s also possible to enable spill to disk to accommodate larger state sizes without sizing up. See the CREATE CLUSTER: Disk documentation for more details.

Sources that specify the SIZE option are linked to a single-purpose cluster dedicated to maintaining that source.

You can also choose to place a source in an existing cluster by using the IN CLUSTER option. Sources in a cluster share the resource allocation of the cluster with all other objects in the cluster.

Colocating multiple sources onto the same cluster can be more resource efficient when you have many low-traffic sources that occasionally need some burst capacity.

Privileges

The privileges required to execute this statement are:

• CREATE privileges on the containing schema.
• CREATE privileges on the containing cluster if the source is created in an existing cluster.
• CREATECLUSTER privileges on the system if the source is not created in an existing cluster.
• USAGE privileges on all connections and secrets used in the source definition.
• USAGE privileges on the schemas that all connections and secrets in the statement are contained in.

Related pages

• Key Concepts
• SHOW SOURCES
• SHOW COLUMNS
• SHOW CREATE SOURCE