Developing an Aggregate Function

The API for developing aggregate custom functions is at the alpha stage and intended for experimental use only. To experiment with this API, create a class within a Java package that implements Drill’s aggregate interface into the program. Include the required information for the function. Your function must include data types that Drill supports, such as INTEGER or BIGINT. For a list of supported data types, refer to the SQL Reference. Keep the following guidelines in mind:

  • Do not use complex @Workspace variables.
  • You cannot allocate a Repeated* value or have a ComplexWriter in the @Workspace.

Complete the following steps to create an aggregate function:

  1. Create a Maven project and add the following dependency:

    org.apache.drill.exec drill-java-exec 1.1.0
  2. Create a class that implements the DrillAggFunc interface and identify the scope as FunctionTemplate.FunctionScope.POINT_AGGREGATE.

**Example**

	@FunctionTemplate(name = "count", scope = FunctionTemplate.FunctionScope.POINT_AGGREGATE)
	public static class BitCount implements DrillAggFunc{   3. Provide the variables used in the code in the `Param, Workspace, `and `Output` bit holders.

**Example**

	@Param BitHolder in;
	@Workspace BitHolder value;
	@Output BitHolder out;   4. Include the setup(), add(), output(),` and `reset() methods.

**Example**
    public void setup() {
      value = new BitHolder();
      value.value = 0;
    }

    @Override
    public void add() {
      value.value++;
    }
    @Override
    public void output() {
      out.value = value.value;
    }
    @Override
    public void reset() {
        value.value = 0;   5. Use the maven-source-plugin to compile the sources and classes JAR files. Verify that an empty `drill-module.conf` is included in the resources folder of the JARs. Drill searches this module during classpath scanning. If the file is not included in the resources folder, you can add it to the JAR file or add it to `etc/drill/conf`.