michael-golfi/azure-functions-java-worker

Java runtime for Microsoft Azure Functions

Building Microsoft Azure Functions in Java

Prerequisites

• JDK 8
• TBD

Programming Model

Your Azure function should be a stateless method to process input and produce output. Although you are allowed to write instance methods, your function must not depend on any instance fields of the class. You need to make sure all the function methods are public accessible.

Typically an Azure function is invoked because of one trigger. Your function needs to process that trigger (sometimes with additional inputs) and gives one or more output values.

All the input and output bindings can be defined in function.json, or in the Java method by using annotations. All the types and annotations used in this document are included in the azure-functions-java-core package.

Here is an example for a simple Azure function written in Java:

package com.example;

public class MyClass {
    public static String echo(String in) {
        return "Hello, " + in + ".";
    }
}

and the corresponding function.json would be:

{
  "scriptFile": "azure-functions-example.jar",
  "entryPoint": "com.example.MyClass.echo",
  "bindings": [
    {
      "type": "httpTrigger",
      "name": "req",
      "direction": "in",
      "authLevel": "anonymous",
      "methods": [ "post" ]
    },
    {
      "type": "http",
      "name": "$return",
      "direction": "out"
    }
  ]
}

Data Types

You are free to use all the data types in Java for the input and output data, including native types; customized POJO types and specialized Azure types defined in azure-functions-java-core package. And Azure Functions runtime will try its best to convert the actual input value to the type you need (for example, a String input will be treated as a JSON string and be parsed to a POJO type defined in your code).

The POJO types you defined have the same accessible requirements as the function methods, it needs to be public accessible. While the POJO class fields are not; for example a JSON string { "x": 3 } is able to be converted to the following POJO type:

public class MyData {
    private int x;
}

You are also allowed to overload methods with the same name but with different types. For example, you can have both String echo(String s) and String echo(MyType s) in one class, and Azure Functions runtime will decide which one to invoke by examine the actual input type (for HTTP input, MIME type text/plain leads to String while application/json represents MyType).

Inputs

Inputs are divided into two categories in Azure Functions: one is the trigger input and the other is the additional input. Although they are different in function.json, the usages are identical in Java code. Let's take the following code snippet as an example:

package com.example;

import com.microsoft.azure.serverless.functions.annotation.BindingName;

public class MyClass {
    public static String echo(String in, @BindingName("item") MyObject obj) {
        return "Hello, " + in + " and " + obj.getKey() + ".";
    }

    public static class MyObject {
        public String getKey() { return this.RowKey; }
        private String RowKey;
    }
}

Here we use two inputs for a function. The @BindingName annotation accepts a String property which represents the name of the binding/trigger defined in function.json:

{
  "scriptFile": "azure-functions-example.jar",
  "entryPoint": "com.example.MyClass.echo",
  "bindings": [
    {
      "type": "httpTrigger",
      "name": "req",
      "direction": "in",
      "authLevel": "anonymous",
      "methods": [ "put" ],
      "route": "items/{id}"
    },
    {
      "type": "table",
      "name": "item",
      "direction": "in",
      "tableName": "items",
      "partitionKey": "Example",
      "rowKey": "{id}",
      "connection": "ExampleStorageAccount"
    },
    {
      "type": "http",
      "name": "$return",
      "direction": "out"
    }
  ]
}

So when this function is invoked, the HTTP request payload will be passed as the String for argument in; and one specific item will be retrieved from the Azure Table Storage and be parsed to MyObject type and be passed to argument obj.

Outputs

Outputs can be expressed both in return value or output parameters. If there is only one output, you are recommended to use the return value. For multiple outputs, you have to use output parameters.

Return value is the simplest form of output, you just return the value of any type, and Azure Functions runtime will try to marshal it back to the actual type (such as an HTTP response). In functions.json, you use $return as the name of the output binding.

To produce multiple output values, use OutputBinding<T> type defined in the azure-functions-java-core package. If you need to make an HTTP response and push a message to a queue as well, you can write something like:

package com.example;

import com.microsoft.azure.serverless.functions.OutputBinding;
import com.microsoft.azure.serverless.functions.annotation.BindingName;

public class MyClass {
    public static String echo(String body, @BindingName("message") OutputBinding<String> queue) {
        String result = "Hello, " + body + ".";
        queue.setValue(result);
        return result;
    }
}

and define the output binding in function.json:

{
  "scriptFile": "azure-functions-example.jar",
  "entryPoint": "com.example.MyClass.echo",
  "bindings": [
    {
      "type": "httpTrigger",
      "name": "req",
      "direction": "in",
      "authLevel": "anonymous",
      "methods": [ "post" ]
    },
    {
      "type": "queue",
      "name": "message",
      "direction": "out",
      "queueName": "myqueue",
      "connection": "ExampleStorageAccount"
    },
    {
      "type": "http",
      "name": "$return",
      "direction": "out"
    }
  ]
}

Binary Data

Binary data is represented as byte[] in your Azure functions code. Let's say you have a binary file in blob, you need to reference it as a blob input binding in function.json (the trick here is dataType is defined as binary, and dataType is available for all kinds of bindings/triggers):

{
  "scriptFile": "azure-functions-example.jar",
  "entryPoint": "com.example.MyClass.echo",
  "bindings": [
    {
      "type": "httpTrigger",
      "name": "req",
      "direction": "in",
      "authLevel": "anonymous",
      "methods": [ "get" ]
    },
    {
      "type": "blob",
      "name": "content",
      "direction": "in",
      "dataType": "binary",
      "path": "container/myfile.bin",
      "connection": "ExampleStorageAccount"
    },
    {
      "type": "http",
      "name": "$return",
      "direction": "out"
    }
  ]
}

And use it in your function code simply as (or if you have too many parameters, you can also use @BindingName("content") byte[] content to reference it):

// Class definition and imports are omitted here
public static String echoLength(byte[] content) {
}

Similarly as the outputs, you need to use OutputBinding<byte[]> type to make a binary output binding.

Context

You interact with Azure Functions execution environment via the ExecutionContext object defined in the azure-functions-java-core package. You are able to get the invocation ID and a built-in logger (which is integrated prefectly with Azure Function Portal experience as well as App Insights) from the context object.

What you need to do is just add one more ExecutionContext typed parameter to your function method (leave all other stuffs unchanged including function.json). Let's take a timer triggered function as an example:

package com.example;

import com.microsoft.azure.serverless.functions.ExecutionContext;

public class MyClass {
    public static String heartbeat(String timerInfo, ExecutionContext context) {
        context.getLogger().info("Heartbeat triggered by " + context.getInvocationId());
        return "Processed " + timerInfo;
    }
}

Specialized Types

Sometimes a function need to take a more detailed control of the input and output, and that's why we also provide some specialized types in the azure-functions-java-core package for you to manipulate:

Specialized Type	Target	Typical Usage
HttpRequestMessage	HTTP Trigger	Get method, headers or queries
HttpResponseMessage	HTTP Output Binding	Return status other than 200

[NOTE] You can also use @BindingName annotation to get HTTP headers and queries. For example, @BindingName("name") String query will try to iterate the HTTP request headers and queries and pass that value to the method; query will be "test" if the request URL is http://example.org/api/echo?name=test.