Tasks

• Using Tasks
• Full Syntax vs Expressions
• Development
  • Complete Example
  • Creating Tasks
  • Task Output
  • Injectable Services
  • Call Context
  • Using External Artifacts
  • Environment Defaults
  • Task Output and Error Handling
  • Unit Tests
  • Integration Tests

Using Tasks

In order to be able to use a task a URL to the JAR containing the implementation has to be added as a dependency.

Typically, the JAR is published to a Maven repository or a remote host and a URL pointing to the JAR in the repository is used.

You can invoke tasks in multiple ways. Following are a number of examples, check the Task Calls section for more details:

configuration:
  dependencies:
    - "http://repo.example.com/my-concord-task.jar"

flows:
  default:
    # call methods directly using expressions
    - ${myTask.call("hello")}

    # call the task using "task" syntax
    # use "out" to save the task's output and "error" to handle errors
    - task: myTask
      in:
        taskVar: ${processVar}
        anotherTaskVar: "a literal value"
      out: myResult
      error:
        - log: myTask failed with ${lastError}

Full Syntax vs Expressions

There are two ways how the task can be invoked: the task syntax and using expressions. Consider the task syntax for tasks with multiple parameters. Use expressions for simple tasks that return data:

# use the `task` syntax when you need to pass multiple parameters and/or complex data structures
- task: myTask
  in:
    param1: 123
    param2: "abc"
    nestedParams:
      x: true
      y: false
      
# use expressions for tasks returning data
- log: "${myTask.getAListOfThings()}"

Development

We recommend running Concord using Java 8 or Java 11. If you’re building a task using Java 11 Concord must use Java 11 as well.

Complete Example

Check out the hello-world-task project for a complete example of a Concord task including end to end testing using testcontainers-concord.

Creating Tasks

Tasks must implement com.walmartlabs.concord.runtime.v2.sdk.Task Java interface and must be annotated with javax.inject.Named.

The following section describes the necessary Maven project setup steps.

Add concord-targetplatform to your dependencyManagement section:

<dependencyManagement>
  <dependencies>
    <dependency>
      <groupId>com.walmartlabs.concord</groupId>
      <artifactId>concord-targetplatform</artifactId>
      <version>1.103.0</version>
      <type>pom</type>
      <scope>import</scope>
    </dependency>
  </dependencies>
</dependencyManagement>

Add the following dependencies to your pom.xml:

<dependencies>
  <dependency>
    <groupId>com.walmartlabs.concord.runtime.v2</groupId>
    <artifactId>concord-runtime-sdk-v2</artifactId>
    <version>1.103.0</version>
    <scope>provided</scope>
  </dependency>

  <dependency>
    <groupId>javax.inject</groupId>
    <artifactId>javax.inject</artifactId>
    <version>1</version>
    <scope>provided</scope>
  </dependency>
</dependencies>

Add sisu-maven-plugin to the build section:

<build>
  <plugins>
    <plugin>
      <groupId>org.eclipse.sisu</groupId>
      <artifactId>sisu-maven-plugin</artifactId>
    </plugin>
  </plugins>
</build>

Some dependencies are provided by the runtime. It is recommended to mark such dependencies as provided in the POM file to avoid classpath conflicts:

• com.fasterxml.jackson.core/*
• javax.inject/javax.inject
• org.slf4j/slf4j-api

Implement the com.walmartlabs.concord.runtime.v2.sdk.Task interface and add javax.inject.Named annotation with the name of the task.

Here’s an example of a simple task:

import com.walmartlabs.concord.runtime.v2.sdk.*;
import javax.inject.Named;

@Named("myTask")
public class MyTask implements Task {

    public void sayHello(String name) {
        System.out.println("Hello, " + name + "!");
    }

    public int sum(int a, int b) {
        return a + b;
    }
}

This task can be called using an expression:

flows:
  default:
  - ${myTask.sayHello("world")}         # short form

  - expr: ${myTask.sum(1, 2)}           # full form
    out: mySum

If a task implements Task#execute method, it can be started using task step type:

@Named("myTask")
public class MyTask implements Task {

    @Override
    public TaskResult execute(Variables input) throws Exception {
        String name = input.assertString("name");
        return TaskResult.success()
                    .value("msg", "Hello, " + name + "!");
    }
}

The task receives a Variables object as input. It contains all in parameters of the call and provides some utility methods to validate the presence of required parameters, convert between types, etc.

Tasks can use the TaskResult object to return data to the flow. See the Task Output section for more details.

To call a task with an execute method, use the task syntax:

flows:
  default:
    - task: myTask
      in:
        name: "world"
      out: myResult

    - log: "${myResult.msg}" # prints out "Hello, world!"

This form allows use of in and out variables and error-handling blocks. See the Task Call section for more details.

In the example above, the task’s result is saved as myResult variable. The runtime converts the TaskResult object into a regular Java Map object:

{
  "ok": true,
  "msg": "Hello, world!"
}

The ok value depends on whether the result was constructed as TaskResult#success() or TaskResult#error(String). In the latter case, the resulting object also contains an error key with the specified error message.

The task syntax is recommended for most use cases, especially when dealing with multiple input parameters.

Task Output

The task must return a TaskResult instance. The TaskResult class provides methods to return additional values as the task call’s result. A task can return multiple values:

return TaskResult.success()
    .value("foo", "bar")
    .value("baz", 123);

Values of any type can be returned, but we recommend returning standard JDK types. Preferably Serializable to avoid serialization issues (e.g. when using forms).

If you need to return some complex data structure, consider converting it to regular Java collections. The runtime provides Jackson as the default JSON/YAML library which can also be used to convert arbitrary data classes into regular Map’s and List’s:

import com.fasterxml.jackson.databind.ObjectMapper;

@Named("myTask")
public class MyTask implements Task {

    @Override
    public TaskResult execute(Variables input) throws Exception {
        MyResult result = new MyResult();
        ObjectMapper om = new ObjectMapper();
        return TaskResult.success()
                .values(om.convertValue(result, Map.class));
    }

    public static class MyResult implements Serializable {
        String data;
        List<String> stuff;
    }
}

In the example above, the properties of MyResult instance became values in the result Map:

- task: myTask
  out: result

- log: |
    data = ${result.data}
    stuff = ${result.stuff}

Injectable Services

The SDK provides a number of services that can be injected into task classes using the javax.inject.Inject annotation:

• Context - provides access to the current call’s environment, low-level access to the runtime, etc. See the Call Context section for more details;
• DependencyManager - a common way for tasks to work with external dependencies. See the Using External Artifacts section for details.

Call Context

To access the current task call’s environment, com.walmartlabs.concord.runtime.v2.sdk.Context can be injected into the task class:

@Named("myTask")
public class MyTask implements Task {

    private final Context ctx;

    @Inject
    public MyTask(Context ctx) {
        this.ctx = ctx;
    }
}

The Context object provides access to multiple features, such as:

• workingDirectory() - returns Path, the working directory of the current process;
• processInstanceId() - returns UUID, the current process’ unique indentifier;
• variables() - provides access to the current flow’s Variables, i.e. all variables defined before the current task call;
• defaultVariables() - default input parameters for the current task. See the Environment Defaults section for more details.

For the complete list of provided features please refer to Javadoc of the Context interface.

Using External Artifacts

The runtime provides a way for tasks to download and cache external artifacts:

import com.walmartlabs.concord.runtime.v2.sdk.*;

@Named("myTask")
public class MyTask implements Task {
    
    private final DependencyManager dependencyManager;
    
    @Inject
    public MyTask(DependencyManager dependencyManager) {
        this.dependencyManager = dependencyManager;
    }
    
    @Override
    public TaskResult execute(Variables input) throws Exception {
        URI uri = ...
        Path p = dependencyManager.resolve(uri);
        // ...do something with the returned path
    }
}

The DependencyManager is an @Inject-able service that takes care of resolving, downloading and caching URLs. It supports all URL types as the regular dependencies section in Concord YAML files - http(s), mvn, etc.

Typically, cached copies are persistent between process executions (depends on the Concord’s environment configuration).

The tasks shouldn’t expect the returning path to be writable (i.e. assume only read-only access).

DependencyManager shouldn’t be used as a way to download deployment artifacts. It’s not a replacement for Ansible or any other deployment tool.

Environment Defaults

Instead of hard coding parameters like endpoint URLs, credentials and other environment-specific values, use Context#defaultVariables:

import com.walmartlabs.concord.runtime.v2.sdk.*;

@Named("myTask")
public class MyTask implements Task {

    private final Context ctx;

    @Inject
    public MyTask(Context ctx) {
        this.ctx = ctx;
    }
    
    @Override
    public TaskResult execute(Variables input) throws Exception {
        Map<String, Object> defaults = ctx.defaultVariables().toMap();
        ...
    }
}

The environment-specific defaults are provided using a Default Process Configuration Rule policy. A defaultTaskVariables entry matching the plugin’s @Named value is provided to the plugin at runtime via the ctx.defaultVariables() method.

{
  "defaultProcessCfg": {
    "defaultTaskVariables": {
      "github": {
        "apiUrl": "https://github.example.com/api/v3"
      }
    }
  }
}

Check out the GitHub task as the example.

Error Handling

By default, the task should throw an exception in case of any execution errors or invalid input parameters. Consider adding the ignoreErrors parameter to catch all execution errors except for input validation errors.

Throw an exception:

- task: myTask
  in:
    url: "https://httpstat.us/404"

Save the error in the result variable:

- task: myTask
  in:
    url: "https://httpstat.us/404"
    ignoreErrors: true
  out: result

- log: "${result.errorCode}"

Unit Tests

Consider using unit tests to quickly test the task without publishing SNAPSHOT versions. Use a library like Mockito to replace the dependencies in your task with “mocks”:

@Test
public void test() throws Exception {
    Map<String, Object> input = new HashMap();
    input.put("name", "Concord");
    
    MyTask t = new MyTask(someService);
    TaskResult.SimpleResult result = (TaskResult.SimpleResult) t.execute(new MapBackedVariables(input));

    assertEquals("Hello, Concord", result.toMap().get("msg"));
}

Integration Tests

The testcontainers-concord project provides a JUnit4 test rule to run Concord in Docker. See the complete example for more details.

Alternatively, it is possible to test a task using a running Concord instance without publishing the task’s JAR. Concord automatically adds lib/*.jar files from the payload archive to the process’ classpath. This mechanism can be used to upload local JAR files and, consequently, to test locally-built JARs. Check out the custom_task example. It uses Maven to collect all compile dependencies of the task and creates a payload archive with the dependencies and the task’s JAR.

Note: It is important to use provided scope for the dependencies that are already included in the runtime. See Creating Tasks section for the list of provided dependencies.