Custom Source Stream¶
Compile code to obtain data from the desired cloud ecosystem or open-source ecosystem as the input data of Flink jobs.
Syntax¶
CREATE SOURCE STREAM stream_id (attr_name attr_type (',' attr_name attr_type)* )
WITH (
type = "user_defined",
type_class_name = "",
type_class_parameter = ""
)
(TIMESTAMP BY timeindicator (',' timeindicator)?);timeindicator:PROCTIME '.' PROCTIME| ID '.' ROWTIME
Keywords¶
Parameter | Mandatory | Description |
|---|---|---|
type | Yes | Data source type. The value user_defined indicates that the data source is a user-defined data source. |
type_class_name | Yes | Name of the source class for obtaining source data. The value must contain the complete package path. |
type_class_parameter | Yes | Input parameter of the user-defined source class. Only one parameter of the string type is supported. |
Precautions¶
The user-defined source class needs to inherit the RichParallelSourceFunction class and specify the data type as Row. For example, define MySource class: public class MySource extends RichParallelSourceFunction<Row>{}. It aims to implement the open, run, and close functions.
Dependency pom:
<dependency>
<groupId>org.apache.flink</groupId>
<artifactId>flink-streaming-java_2.11</artifactId>
<version>${flink.version}</version>
<scope>provided</scope>
</dependency>
<dependency>
<groupId>org.apache.flink</groupId>
<artifactId>flink-core</artifactId>
<version>${flink.version}</version>
<scope>provided</scope>
</dependency>
Example¶
A data record is generated in each period. The data record contains only one field of the INT type. The initial value is 1 and the period is 60 seconds. The period is specified by an input parameter.
CREATE SOURCE STREAM user_in_data (
count INT
)
WITH (
type = "user_defined",
type_class_name = "mySourceSink.MySource",
type_class_parameter = "60"
)
TIMESTAMP BY car_timestamp.rowtime;
Note
To customize the implementation of the source class, you need to pack the class in a JAR package and upload the UDF function on the SQL editing page.