UDF of SQL transform plugin
Use UDF SPI to extend the SQL transform functions lib.
package org.apache.seatunnel.transform.sql.zeta; public interface ZetaUDF { /** * Function name * * @return function name */ String functionName(); /** * The type of function result * * @param argsType input arguments type * @return result type */ SeaTunnelDataType<?> resultType(List<SeaTunnelDataType<?>> argsType); /** * Evaluate * * @param args input arguments * @return result value */ Object evaluate(List<Object> args); /** * Whether current udf requires row level context. */ default boolean requiresContext() { return false; } /** * Evaluate with row level context. */ default Object evaluateWithContext(List<Object> args, ZetaUDFContext context) { return evaluate(args); } /** * Initialize udf resources. */ default void open() throws Exception {} /** * Release udf resources. */ default void close() {} }
ZetaUDFContext provides runtime row-level metadata and fields:
getRawTableId()getDatabase()getSchema()getTable()getRowKind()getAllFields()Notes:
database/schema/table parsing follows TablePath.of(tableId) semantics.tableId is in an unsupported format, accessing database/schema/table throws IllegalArgumentException.evaluate(List<Object> args).Add these dependencies and provided scope to your maven project. Dependency versions should match the runtime environment.
<dependencies> <dependency> <groupId>org.apache.seatunnel</groupId> <artifactId>seatunnel-transforms-v2</artifactId> <version>${seatunnel.version}</version> <scope>provided</scope> </dependency> <dependency> <groupId>org.apache.seatunnel</groupId> <artifactId>seatunnel-api</artifactId> <version>${seatunnel.version}</version> <scope>provided</scope> </dependency> <dependency> <groupId>com.google.auto.service</groupId> <artifactId>auto-service</artifactId> <version>1.0.1</version> <scope>provided</scope> </dependency> </dependencies>
Add a Java Class implements of ZetaUDF like this:
@AutoService(ZetaUDF.class) public class ExampleUDF implements ZetaUDF { @Override public String functionName() { return "EXAMPLE"; } @Override public SeaTunnelDataType<?> resultType(List<SeaTunnelDataType<?>> argsType) { return BasicType.STRING_TYPE; } @Override public Object evaluate(List<Object> args) { String arg = (String) args.get(0); if (arg == null) return null; return "UDF: " + arg; } }
Package the UDF project and copy the jar to the path: ${SEATUNNEL_HOME}/lib. And if your UDF use third party library, you also need put it to ${SEATUNNEL_HOME}/lib.
If you use cluster mode, you need put the lib to all your node's ${SEATUNNEL_HOME}/lib folder and re-start the cluster.
@AutoService(ZetaUDF.class) public class ContextLifecycleUdf implements ZetaUDF { private transient String prefix; @Override public String functionName() { return "CTX_LIFE"; } @Override public SeaTunnelDataType<?> resultType(List<SeaTunnelDataType<?>> argsType) { return BasicType.STRING_TYPE; } @Override public boolean requiresContext() { return true; } @Override public void open() { this.prefix = "OPENED"; } @Override public Object evaluateWithContext(List<Object> args, ZetaUDFContext context) { String arg = args.get(0) == null ? null : String.valueOf(args.get(0)); if (arg == null) { return null; } return prefix + ":" + context.getRowKind().shortString() + ":" + arg; } @Override public void close() { this.prefix = null; } }
The data read from source is a table like this:
| id | name | age |
|---|---|---|
| 1 | Joy Ding | 20 |
| 2 | May Ding | 21 |
| 3 | Kin Dom | 24 |
| 4 | Joy Dom | 22 |
We use UDF of SQL query to transform the source data like this:
transform {
Sql {
plugin_input = "fake"
plugin_output = "fake1"
query = "select id, example(name) as name, age from dual"
}
}
Then the data in result table fake1 will update to
| id | name | age |
|---|---|---|
| 1 | UDF: Joy Ding | 20 |
| 2 | UDF: May Ding | 21 |
| 3 | UDF: Kin Dom | 24 |
| 4 | UDF: Joy Dom | 22 |