JarCastingJWorkflow
JWorkflow is a light weight workflow library for Java. It supports pluggable persistence and concurrency providers to allow for multi-node clusters.
Installing
Using Maven
Add jworkflow to your POM file as a dependency.
<dependencies>
<dependency>
<groupId>net.jworkflow</groupId>
<artifactId>jworkflow</artifactId>
<version>0.5-SNAPSHOT</version>
</dependency>
</dependencies>
Using Gradle
dependencies {
compile 'net.jworkflow:jworkflow:0.5-SNAPSHOT'
}
Documentation
See Tutorial here.
Fluent API
Define your workflows with the fluent API.
public class HelloWorkflow implements Workflow {
@Override
public void build(WorkflowBuilder builder) {
builder
.startsWith(Task1.class)
.then(Task2.class)
.then(Task3.class);
}
}
...
...
public class Task1 implements StepBody {
@Override
public ExecutionResult run(StepExecutionContext context) {
System.out.println("Hello world");
return ExecutionResult.next();
}
}
JSON Workflow Definitions
Define your workflows in JSON
{
"id": "HelloWorld",
"version": 1,
"steps": [
{
"id": "hello",
"stepType": "com.myapp.Step1",
"nextStepId": "bye"
},
{
"id": "bye",
"stepType": "com.myapp.Step2"
}
]
}
Sample use cases
- New user workflow
public class MyData {
public String email;
public String password;
public String userId;
}
public class MyWorkflow implements Workflow<MyData> {
...
@Override
public void build(WorkflowBuilder<MyData> builder) {
builder
.startsWith(CreateUser.class)
.input((step, data) -> step.email = data.email)
.input((step, data) -> step.password = data.password)
.output((step, data) -> data.userId = step.userId)
.then(SendConfirmationEmail.class)
.waitFor("confirmation", data -> data.userId)
.then(UpdateUser.class)
.input((step, data) -> step.userId = data.userId);
}
}
- Saga Transactions
public class MyWorkflow implements Workflow<MyData> {
...
@Override
public void build(WorkflowBuilder<MyData> builder) {
builder
.startsWith(CreateCustomer.class)
.then(PushToSalesforce.class)
.onError(ErrorBehavior.RETRY)
.then(PushToERP.class)
.onError(ErrorBehavior.RETRY, Duration.ofMinutes(30));
}
}
builder
.startsWith(LogStart.class)
.saga(saga -> saga
.startsWith(Task1.class)
.compensateWith(UndoTask1.class)
.then(Task2.Class)
.compensateWith(UndoTask2.class)
.then(Task3.Class)
.compensateWith(UndoTask3.class)
)
.onError(ErrorBehavior.RETRY, Duration.ofMinutes(30));
.then(LogEnd.class);
Persistence
Since workflows are typically long running processes, they will need to be persisted to storage between steps. There are several persistence providers available as seperate packages.
- Memory Persistence Provider (Default provider, for demo and testing purposes)
- MongoDB
- Amazon DynamoDB
- (more to come soon...)
Multi-node clusters
By default, the WorkflowHost service will run as a single node using the built-in queue and locking providers for a single node configuration. Should you wish to run a multi-node cluster, you will need to configure an external queueing mechanism and a distributed lock manager to co-ordinate the cluster. These are the providers that are currently available.
Queue Providers
- SingleNodeQueueProvider (Default built-in provider)
- RabbitMQ
- Redis
- AWS Simple Queue Service
Distributed lock managers
- SingleNodeLockProvider (Default built-in provider)
- Redis
- Amazon DynamoDB
- (more to come soon...)
Samples
Contributors
- Daniel Gerlag - Initial work
License
This project is licensed under the MIT License - see the LICENSE.md file for details