StateMachineBehavior for Propel2

This behavior adds a finite state machine to your model. This behavior is written by William Durand, I change it to work with Propel2.

Configuration

<behavior name="smolowik\Propel\Behavior\StateMachine\StateMachineBehavior">
    <parameter name="states" value="draft, rejected, unpublished, published" />

    <parameter name="initial_state" value="draft" />

    <parameter name="transition" value="draft to published with publish" />
    <parameter name="transition" value="draft to rejected with reject" />
    <parameter name="transition" value="published to unpublished with unpublish" />
    <parameter name="transition" value="unpublished to published with publish" />

    <!-- Optional parameters -->
    <parameter name="state_column" value="state" />
    <parameter name="timestampable" value="true" />
</behavior>

The state_machine behavior requires three parameters to work:

• states: a finite set of states as comma separated values;
• initial_state: the initial state, part of set of states;
• transition: a set of transitions. As you can see, you can add as many transition parameters as you want.

Each transition has to follow this pattern:

STATE_1 to STATE_2 with SYMBOL

A symbol, which is part of the Finite State Machine's terminology, can be considered as an event triggered on your model object.

The timestampable option can be added to log in the_given_state_at column the date when the state is setted

### ActiveRecord API ###

The behavior will generate the following constants which represent the available states of your object:

• ObjectModel::STATE_DRAFT
• ObjectModel::STATE_REJECTED
• ObjectModel::STATE_UNPUBLISHED
• ObjectModel::STATE_PUBLISHED

You can get the current state of your object:

getState()

Or get an array with all available states:

getAvailableStates()

Most of the time, you would like to display these states. Thanks to two convenient methods, it's really easy:

getHumanizedState() // 'Draft', or 'Rejected', or 'Unpublished', or 'Published'

ObjectModel::getHumanizedStates()
// array(
//  0 => 'Draft',
//  1 => 'Rejected',
//  2 => 'Unpublished',
//  3 => 'Published',
// )

The behavior will also generate a set of issers:

isDraft()

isRejected()

isPublished()

isUnpublished()

But the most interesting part is the implemenation of the FSM itself. First you have methods to determine whether you can perform, or not a transition based on the current model's state:

canPublish()

canReject()

canUnpublish()

It will also generate a set of methods for each symbol:

publish(ConnectionInterface $con = null)

unpublish(ConnectionInterface $con = null)

reject(ConnectionInterface $con = null)

To handle custom logic, new hooks are created. The methods below should return a boolean value, and can act as guards (which is not part of the FSM's terminology).

prePublish(ConnectionInterface $con = null)

preUnpublish(ConnectionInterface $con = null)

preReject(ConnectionInterface $con = null)

The methods below should contain your own logic depending on each state, and your business.

onPublish(ConnectionInterface $con = null)

onUnpublish(ConnectionInterface $con = null)

onReject(ConnectionInterface $con = null)

The methods below allow to execute code once the transition is executed.

postPublish(ConnectionInterface $con = null)

postUnpublish(ConnectionInterface $con = null)

postReject(ConnectionInterface $con = null)

ActiveQuery API

To be defined.

Usage

Let's say we have a Post model class which represents an entry in a blog engine. When we create a new post, its initial state is draft because we don't want to publish it immediately. As a draft, you can decide to publish your new post. Its state is now published. Once published, you may want to unpublish it for some reasons. Then, its state is unpublished. The last possibily is to republish an unpublished post. The new state is published.

We have three different states (draft, published, unpublished), and three transitions:

• draft to published
• published to unpublished
• unpublished to published

We can define the following configuration:

<table name="post">
    <!-- some columns -->

    <behavior name="smolowik\Propel\Behavior\StateMachine\StateMachineBehavior">
        <parameter name="states" value="draft, unpublished, published" />

        <parameter name="initial_state" value="draft" />

        <parameter name="transition" value="draft to published with publish" />
        <parameter name="transition" value="published to unpublished with unpublish" />
        <parameter name="transition" value="unpublished to published with publish" />
    </behavior>
</table>

Here is a workflow:

<?php

$post = new Post();

$post->getState();              // Post::STATE_DRAFT
$post->getAvailableStates();    // Post::STATE_DRAFT, Post::STATE_UNPUBLISHED, Post::STATE_PUBLISHED

$post->isDraft();               // true
$post->isPublished();           // false
$post->isUnpublished();         // false

$post->canPublish();            // true
$post->canUnpublish();          // false

$post->unpublish();             // throw a LogicException, no transition found from draft to unpublished

// Let's publish this post
// This is the first transition in the scenario above
$post->publish()->save();

$post->isDraft();               // false
$post->isPublished();           // true
$post->isUnpublished();         // false

$post->canPublish();            // false
$post->canUnpublish();          // true

$post->publish();               // throw a LogicException, the post is already published

// Let's unpublish this post
// This is the second transition in the scenario above
$post->unpublish()->save();

$post->isDraft();               // false
$post->isPublished();           // false
$post->isUnpublished();         // true

$post->canPublish();            // true
$post->canUnpublish();          // false

$post->unpublish();             // throw a LogicException, the post is already unpublished

// Let's (re)publish this post
// This is the last transition in the scenario above
$post->publish()->save();

$post->isDraft();               // false
$post->isPublished();           // true
$post->isUnpublished();         // false

$post->canPublish();            // false
$post->canUnpublish();          // true

Now imagine we have authors linked to each post, and once a post is published, we notify the post's author by email. Thanks to new hooks, it's really easy to extend things:

<?php

class Post extends BasePost
{
    // Assuming we have a mail manager which is able to send emails,
    // and that we injected it before.
    private $mailManager;

    public function onPublish(ConnectionInterface $con = null)
    {
        $this->mailManager->postPublished(
            $this->getAuthor(),
            $this->getTitle()
        );
    }
}

Use case in a controller:

<?php

class PostController extends Controller
{
    public function newAction()
    {
        // handle a form, etc to create a new Post
    }

    public function publishAction(Post $post)
    {
        try {
            $post->publish()->save();
        } catch (Propel\Runtime\Exception\LogicException $e) {
            // handle the exception as you wish
        }
    }

    public function unpublishAction(Post $post)
    {
        try {
            $post->unpublish()->save();
        } catch (Propel\Runtime\Exception\LogicException $e) {
            // handle the exception as you wish
        }
    }
}

Known Limitations

• You cannot use the deleted state;
• You cannot use the save, or delete symbols.

At the moment, there is no built-in solution to handle these cases.

Combining Archivable Behavior

The Archivable behavior is useful to copy a model object to an archival table. In other words, it acts as a soft delete behavior but with better performance.

In your workflow, you may want to destroy your object for some reason. I say "destroy" because you can't use the deleted status, nor the delete symbol, but it doesn't matter. Destroying an object is fine, but instead of hard deleting it, you may want to soft delete it. That means you will rely on the archivable behavior.

Just add it to your XML schema, rebuild both SQL, and model classes, and you're done. At first glance, when you destroy your object, you will expect it to be hidden, but it's not the case. It just has the destroyed state.

Just call the delete() termination method as usual, and your object will be automatically archived. No much to do.