Workflow

Summary

A Workflow is a configurable stored procedure that will run one or more jobs. Jobs are independent from each other, but interconnected as you can pass response references between jobs. Jobs supports conditional running based on variables and previous job responses.

Installing

Workflow is available through Packagist (opens new window) and the repository source is at chevere/workflow (opens new window).

composer require chevere/workflow

What it does?

The Workflow package provides a robust system for defining and executing structured procedures based on the workflow pattern (opens new window). It enables to organize complex logic into a series of interconnected, independent jobs that can be executed in a controlled manner.

By breaking down monolithic procedures into modular workflow jobs, developers gain several advantages:

• Improved testability of individual components
• Better code organization and maintainability
• Reusable job definitions across different workflows
• Clear visualization of process flows
• Flexible execution patterns (sync/async)

💡 Workflow introduction

Read Workflow for PHP (opens new window) at Rodolfo's blog for a compressive introduction to this package.

How to use

The Workflow package provides a set of core functions in the Chevere\Workflow namespace that allow you to build and manage workflow processes. These functions work together to create flexible, maintainable workflow definitions.

Functions

Function	Purpose
workflow	Creates a new workflow container for organizing named jobs
sync	Defines a synchronous job that blocks until completion
async	Defines an asynchronous job that runs non-blocking
variable	Declares a workflow-level variable for job inputs
response	Creates a reference to access previous job outputs

Key concepts

• Job: Self-contained unit of work defined by Action (opens new window)
• Variable: Shared workflow-level inputs accessed by multiple jobs
• Response: Links between job outputs (response()) and inputs

Workflow example

php demo/chevere.php

Create MyAction action by extending Chevere\Action\Action. You can also use ActionTrait.

use Chevere\Action\Action;

class MyAction extends Action
{
    protected function main(string $foo): string
    {
        return "Hello, {$foo}";
    }
}

Create Workflow with your MyAction Job:

use function Chevere\Workflow\{workflow,sync,variable,response};

$workflow = workflow(
    greet: sync(
        new MyAction(),
        foo: variable('super'),
    ),
    capo: sync(
        new MyAction(),
        foo: response('greet'),
    ),
);

Run the Workflow:

use function Chevere\Workflow\run;

$hello = run(
    $workflow,
    super: 'Chevere',
);
echo $hello->response('greet')->string() . PHP_EOL;
// Hello, Chevere
echo $hello->response('capo')->string() . PHP_EOL;
// Hello, Hello, Chevere

Variable

Use function variable to declare a Workflow variable that will be injected when running the workflow. Variables allow you to pass external values into your workflow jobs during execution.

use function Chevere\Workflow\variable;

// Basic variable declaration
variable('myVar');

// Usage in a job
sync(
    new MyAction(),
    parameter: variable('myVar')
);

When running the workflow, you must provide values for all declared variables:

use function Chevere\Workflow\run;

run($workflow, myVar: 'some value');

Response

Use function response to declare a reference to a response returned by a previous Job. This allows you to chain job outputs as inputs to subsequent jobs.

🪄 When using a response it will auto declare the referenced Job as a dependency, ensuring proper execution order.

use function Chevere\Workflow\response;

// Basic response declaration
response('job1');

// Usage in a Workflow
workflow(
    job1: sync(
        new SomeAction(),
    ),
    job2: sync(
        new MyAction(),
        parameter: response('job1')
    );
);

References can be also made on a response member identified by key.

use function Chevere\Workflow\response;

response('job1', 'id');

Job

The Job class defines an Action (opens new window) that can be executed as part of a workflow.

Arguments

Job arguments can be passed in three ways:

• As-is values: Direct values passed to the Action
• Variables: Workflow-level inputs declared using the variable function
• Responses: References to previous job outputs declared using the response function

class SomeAction extends Action
{
    protected function main(
        string $context,
        int $userId,
        mixed ...$bag,
    ): void
    {
        // On runtime:
        // $context = 'public'
        // $userId = (( user.id response ))
        // $bag = ['group' => 'admin', 'mask' => 1024]
    }
}

$workflow = workflow(
    user: sync(
        new GetUser(),
        request: variable('userId')
    ),
    job1: sync(
        new SomeAction(),
        context: 'public',               // As-is value
        userId: variable('userId'),      // Variable
        group: response('user', 'group'),// Response
        mask: 1024,                      // As-is value
    );
);

run($workflow, userId: 123);

In the example above:

• The context argument is passed as-is with the value public.
• The userId argument is dynamically provided as a variable.
• The group argument is dynamically provided as a response from a previous job.
• The mask argument is passed as-is with the value 1024.

When running the Workflow, these arguments will be matched against the parameters defined in the main method (opens new window) of SomeAction.

Asynchronous

Use function async to create an asynchronous job, which runs non-blocking.

Important: When using async jobs, your Actions must support serialization (opens new window). For Actions that work with non-serializable resources like:

• Database connections
• File handles
• Stream resources
• Network sockets

You must use sync jobs instead.

In the example below a Workflow describes an image creation procedure for multiple image sizes.

use function Chevere\Workflow\{sync,async,response,variable,workflow};

workflow(
    thumb: async(
        new ImageResize(),
        image: variable('image'),
        width: 100,
        height: 100,
        fit: 'thumb'
    ),
    medium: async(
        new ImageResize(),
        image: variable('image'),
        width: 500,
        fit: 'resizeByW'
    ),
    store: sync(
        new StoreFiles(),
        response('thumb', 'filename'),
        response('medium', 'filename'),
    ),
);

• variable('image') declares a Variable.
• response('thumb', 'filename') and response('medium', 'filename') declares a Response reference.

The graph for this Workflow says that thumb, medium and poster run non-blocking in parallel. Job store runs blocking (another node).

graph TD;
    thumb-->store;
    medium-->store;
    poster-->store;

$workflow->jobs()->graph()->toArray();
// contains
[
    ['thumb', 'medium', 'poster'],
    ['store']
];

To complete the example, here's how to Run the Workflow previously defined:

use function Chevere\Workflow\run;

run(
    workflow: $workflow,
    arguments: [
        'image' => '/path/to/file',
    ]
);

Synchronous

Use function sync to create a synchronous job, which block execution until it gets resolved.

In the example below a Workflow describes an image uploading procedure.

use function Chevere\Workflow\{sync,response,variable,workflow};

workflow(
    user: sync(
        new GetUser(),
        request: variable('payload')
    ),
    validate: sync(
        new ValidateImage(),
        mime: 'image/png',
        file: variable('file')
    ),
    meta: sync(
        new GetMeta(),
        file: variable('file'),
    ),
    store: sync(
        new StoreFile(),
        file: variable('file'),
        name: response('meta', 'name'),
        user: response('user')
    ),
);

• variable('payload') and variable('file') declares a Variable.
• response('meta', 'name') and response('user') declares a Response reference.

The graph for this Workflow says that all jobs run one after each other as all jobs are defined using sync.

graph TD;
    user-->validate-->meta-->store;

$workflow->jobs()->graph()->toArray();
// contains
[
    ['user'],
    ['validate'],
    ['meta'],
    ['store']
];

To complete the example, here's how to Run the Workflow previously defined:

use function Chevere\Workflow\run;

run(
    $workflow,
    payload: $_REQUEST,
    file: '/path/to/file',
);

Conditional running

Method withRunIf enables to pass arguments of type Variable or Response for conditionally running a Job.

sync(
    new CompressImage(),
    file: variable('file')
)
    ->withRunIf(
        variable('compressImage'),
        response('SomeAction', 'doImageCompress')
    )

For the code above, all conditions must meet to run the Job and both variable compressImage and the reference SomeAction:doImageCompress must be true to run the job.

Dependencies

Use withDepends method to explicit declare previous jobs as dependencies. The dependent Job won't run until the dependencies are resolved.

job(new SomeAction())
    ->withDepends('myJob');

Running

To run a Workflow use the run function by passing a Workflow and its variables (if any).

use function Chevere\Workflow\run;

$run = run($workflow, ...$variables);

Access Job response

Use response to retrieve a job response as a CastArgument object which can be used to get a typed response.

$thumbFile = $run->response('thumb')->string();

🪄 If the response is of type array|ArrayAccess you can shortcut key access casting.

use function Chevere\Parameter\cast;

$id = $run->response('user', 'id')->int();

WorkflowTrait

The WorkflowTrait provides methods execute and run for easing handling a Workflow within a class.

use Chevere\Workflow\WorkflowTrait;

class Something
{
    use WorkflowTrait;

    public function __construct()
    {
        $workflow = workflow(
            job1: sync(
                new MyAction(),
                foo: variable('bar')
            )
        );
        // Use execute to run the Workflow
        $this->execute($workflow, bar: 'baz');
    }
}

$some = new Something();
$bar = $some->run()->response('job1')->string();

Exception handling

When running a Workflow, if a Job fails a WorkflowException will be thrown. This is an exception wrapper for the job that thrown the exception.

try {
    $run = run($workflow, ...$variables);
} catch (WorkflowException $e) {
    // Job that thrown the exception
    $e->name;
    // Job instance that thrown the exception
    $e->job;
    // The exception thrown by the Job
    $e->throwable;
}

// If using WorkflowTrait
try {
    $this->execute($workflow, ...$variables);
    $run = $this->run();
} catch (WorkflowException $e) {
    // ...
}

Demo

See the demo (opens new window) directory for a set of examples.

Testing

Workflow provides several approaches for testing your implementations. While the Workflow itself doesn't need testing (it's a configuration), you should test:

1. Job actions (unit tests)
2. Workflow execution order (graph)
3. Job responses
4. Exception handling

Testing Job Actions

The primary testing focus should be on your Action implementations:

use PHPUnit\Framework\TestCase;

class MyActionTest extends TestCase
{
    public function testAction(): void
    {
        $action = new MyAction();
        // 🪄 Chevere automatically validates Action I/O
        $response = $action(foo: 'bar');
        $this->assertSame('expected', $response);
    }
}

Testing Workflow Graph

Verify the execution order by testing the Workflow graph:

public function testWorkflowOrder(): void
{
    $expectedGraph = [
        ['job1', 'job2'], // parallel jobs
        ['job3'],         // depends on job1, job2
    ];
    $this->assertSame(
        $expectedGraph,
        $workflow->jobs()->graph()->toArray()
    );
}

Testing Job Responses

Test how jobs interact by checking their responses:

public function testJobResponses(): void
{
    $run = run($workflow, input: 'test');
    // Access typed responses
    $this->assertSame(
        123,
        $run->response('job1')->int()
    );
    $this->assertSame(
        'test',
        $run->response('job2')->string()
    );
    // Access array responses
    $this->assertSame(
        10.2,
        $run->response('job3', 'rate')->float()
    );
}

Testing Exception Handling

Use ExpectWorkflowExceptionTrait to test error scenarios:

use Chevere\Workflow\Traits\ExpectWorkflowExceptionTrait;
use function Chevere\Workflow\run;

class WorkflowTest extends TestCase
{
    use ExpectWorkflowExceptionTrait;

    public function testFailingJob(): void
    {
        $this->expectWorkflowException(
            closure: fn () => run($workflow, input: 'invalid'),
            exception: LogicException::class,
            job: 'validation',
            message: 'Invalid input'
        );
    }
}

Architecture

The architecture of the Workflow package is designed to provide a clear separation of concerns, making it easier to define, manage, and execute workflows. The following diagram illustrates the core components and their interactions:

graph TD
    subgraph Client Application
        WF[Workflow Definition]
        Run[run Function]
    end

    subgraph Core Components
        Jobs[Jobs]
        Graph[Graph]
        Job[Job]
        Action[Action]
    end

    subgraph References
        Var[Variables]
        Resp[Responses]
    end

    subgraph Execution
        Runner[Workflow Runner]
        Sync[Sync Executor]
        Async[Async Executor]
    end

    WF --> Jobs
    Jobs --> |define| Graph
    Jobs --> |manages| Job
    Job --> |executes| Action
    Job --> |depends on| Var
    Job --> |depends on| Resp
    Run --> Runner
    Runner --> |uses| Jobs
    Runner --> |resolves| Graph
    Runner --> |executes via| Sync
    Runner --> |executes via| Async