Extending the Aerospike JDBC driver with UDFs

Preface

This post continues series of articles that present the Aerosplike SQL driver.

Introduction

Previous post explained the built-in functions provided by the Aerospike JDBC Driver. Here I am going to explain how to extend the functionality of the driver by implementing User Defined Functions (UDF).

Implementation of UDF

Right now aggregate functions cannot be customized, however you can implement scalar and collection functions using Java programming language.

Function is a named block of reusable code that used to perform single action. Typically functions return result. Functions may accept arguments.

Java 8 and higher provides interfaces Supplier, Function, BiFunction. Implementations of these interfaces can be defined as UDF for the Aerospike JDBC driver. For example here is definition of function now() that returns epoch in milliseconds:

public class Now implements new Supplier<Long>() {
    @Override
    public Long get() {
        return System.currentTimeMillis();
    }
};

This is example of function that receives one argument:

public class Lower implements new Function<String, String>() {
    @Override
    public String apply(String s) {
        return s == null ? null : s.toLowerCase();
    }
};

Implementation of function that accepts 2 arguments is similar. Create class that implements BiFunction. Unfortunately JDK does not define interface for function that accepts 3, 4 or more arguments as well as function that accepts any number of arguments. Such interface is defined by the driver:

package com.nosqldriver.util;
@FunctionalInterface
public interface VarargsFunction<T, R> {
    R apply(T ... t);
}

If you want to implement UDF that accepts more than 2 arguments you have to implement this interface. In this case you have to add the driver to the compilation class path. Implementation of other types of functions does not require this dependency. Implementation of VarargsFunction has yet another complexity: the programmer is responsible on casting and verification of accepted arguments.

Deploying of UDF

Once UDF is implemented and verified it should be packed into jar file that should be added to the classpath of the driver. Various tools allow this. For example Dbeaver and SquirreL allow defining driver packaged in several jar files. Once this is done the the function should be registered using connection parameter:

custom.function.NAME=FULLY_QUALIFIED_CLASS_NAME

e.g.

custom.function.now=com.mycompany.Now

The connection parameter can be supplied as a parameter of JDBC URL or as a connection property.

jdbc:aeropspike:myhost?custom.function.now=com.mycompany.Now

Once function is registered it can be used in SQL query:

select first_name, now() from people

Conclusions

Aerospike JDBC Driver provides a comprehensive set of built-in functions. Moreover it is extendable using User-Defined Functions that can be easily implemented using Java programming language.

Project home

The project is available in GitHub.

What's next

Next article of this series will explain how to work with complex types.