Making Decisions with a Workflow

Having a state for a model is quite common. The comment state is only determined by the spam checker. What if we add more decision factors?

We might want to let the website admin moderate all comments after the spam checker. The process would be something along the lines of:

Start with a submitted state when a comment is submitted by a user;
Let the spam checker analyze the comment and switch the state to either potential_spam, ham, or rejected;
If not rejected, wait for the website admin to decide if the comment is good enough by switching the state to published or rejected.

Implementing this logic is not too complex, but you can imagine that adding more rules would greatly increase the complexity. Instead of coding the logic ourselves, we can use the Symfony Workflow Component:

        1
        $ symfony composer req workflow
    

Describing Workflows

The comment workflow can be described in the config/packages/workflow.yaml file:

config/packages/workflow.yaml

        1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
        framework:
    workflows:
        comment:
            type: state_machine
            audit_trail:
                enabled: "%kernel.debug%"
            marking_store:
                type: 'method'
                property: 'state'
            supports:
                - App\Entity\Comment
            initial_marking: submitted
            places:
                - submitted
                - ham
                - potential_spam
                - spam
                - rejected
                - published
            transitions:
                accept:
                    from: submitted
                    to:   ham
                might_be_spam:
                    from: submitted
                    to:   potential_spam
                reject_spam:
                    from: submitted
                    to:   spam
                publish:
                    from: potential_spam
                    to:   published
                reject:
                    from: potential_spam
                    to:   rejected
                publish_ham:
                    from: ham
                    to:   published
                reject_ham:
                    from: ham
                    to:   rejected
    

To validate the workflow, generate a visual representation:

        1
        $ symfony console workflow:dump comment | dot -Tpng -o workflow.png
    

Note

The dot command is a part of the Graphviz utility.

Using a Workflow

Replace the current logic in the message handler with the workflow:

        1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
        --- i/src/MessageHandler/CommentMessageHandler.php
+++ w/src/MessageHandler/CommentMessageHandler.php
@@ -6,7 +6,10 @@ use App\Message\CommentMessage;
 use App\Repository\CommentRepository;
 use App\SpamChecker;
 use Doctrine\ORM\EntityManagerInterface;
+use Psr\Log\LoggerInterface;
 use Symfony\Component\Messenger\Attribute\AsMessageHandler;
+use Symfony\Component\Messenger\MessageBusInterface;
+use Symfony\Component\Workflow\WorkflowInterface;

 #[AsMessageHandler]
 class CommentMessageHandler
@@ -15,6 +18,9 @@ class CommentMessageHandler
         private EntityManagerInterface $entityManager,
         private SpamChecker $spamChecker,
         private CommentRepository $commentRepository,
+        private MessageBusInterface $bus,
+        private WorkflowInterface $commentStateMachine,
+        private ?LoggerInterface $logger = null,
     ) {
     }

@@ -25,12 +31,18 @@ class CommentMessageHandler
             return;
         }

-        if (2 === $this->spamChecker->getSpamScore($comment, $message->getContext())) {
-            $comment->setState('spam');
-        } else {
-            $comment->setState('published');
+        if ($this->commentStateMachine->can($comment, 'accept')) {
+            $score = $this->spamChecker->getSpamScore($comment, $message->getContext());
+            $transition = match ($score) {
+                2 => 'reject_spam',
+                1 => 'might_be_spam',
+                default => 'accept',
+            };
+            $this->commentStateMachine->apply($comment, $transition);
+            $this->entityManager->flush();
+            $this->bus->dispatch($message);
+        } elseif ($this->logger) {
+            $this->logger->debug('Dropping comment message', ['comment' => $comment->getId(), 'state' => $comment->getState()]);
         }
-
-        $this->entityManager->flush();
     }
 }
    

The new logic reads as follows:

If the accept transition is available for the comment in the message, check for spam;
Depending on the outcome, choose the right transition to apply;
Call apply() to update the Comment via a call to the setState() method;
Call flush() to commit the changes to the database;
Re-dispatch the message to allow the workflow to transition again.

As we haven't implemented the admin validation, the next time the message is consumed, the "Dropping comment message" will be logged.

Let's implement an auto-validation until the next chapter:

        1
2
3
4
5
6
7
8
9
10
11
12
        --- i/src/MessageHandler/CommentMessageHandler.php
+++ w/src/MessageHandler/CommentMessageHandler.php
@@ -41,6 +41,9 @@ class CommentMessageHandler
             $this->commentStateMachine->apply($comment, $transition);
             $this->entityManager->flush();
             $this->bus->dispatch($message);
+        } elseif ($this->commentStateMachine->can($comment, 'publish') || $this->commentStateMachine->can($comment, 'publish_ham')) {
+            $this->commentStateMachine->apply($comment, $this->commentStateMachine->can($comment, 'publish') ? 'publish' : 'publish_ham');
+            $this->entityManager->flush();
         } elseif ($this->logger) {
             $this->logger->debug('Dropping comment message', ['comment' => $comment->getId(), 'state' => $comment->getState()]);
         }
    

Run symfony server:log and add a comment in the frontend to see all transitions happening one after the other.

Finding Services from the Dependency Injection Container

When using dependency injection, we get services from the dependency injection container by type hinting an interface or sometimes a concrete implementation class name. But when an interface has several implementations, Symfony cannot guess which one you need. We need a way to be explicit.

We have just come across such an example with the injection of a WorkflowInterface in the previous section.

As we inject any instance of the generic WorkflowInterface interface in the contructor, how can Symfony guess which workflow implementation to use? Symfony uses a convention based on the argument name: $commentStateMachine refers to the comment workflow in the configuration (which type is state_machine). Try any other argument name and it will fail.

If you don't remember the convention, use the debug:container command. Search for all services containing "workflow":

        1
2
3
4
5
6
7
8
9
10
11
12
13
14
        $ symfony console debug:container workflow

 Select one of the following services to display its information:
  [0] console.command.workflow_dump
  [1] workflow.abstract
  [2] workflow.marking_store.method
  [3] workflow.registry
  [4] workflow.security.expression_language
  [5] workflow.twig_extension
  [6] monolog.logger.workflow
  [7] Symfony\Component\Workflow\Registry
  [8] Symfony\Component\Workflow\WorkflowInterface $commentStateMachine
  [9] Psr\Log\LoggerInterface $workflowLogger
 >
    

Notice choice 8, Symfony\Component\Workflow\WorkflowInterface $commentStateMachine which tells you that using $commentStateMachine as an argument name has a special meaning.

Note

We could have used the debug:autowiring command as seen in a previous chapter:

        1
        $ symfony console debug:autowiring workflow
    

Workflows and State Machines and when to choose each one;
The Symfony Workflow docs.

This work, including the code samples, is licensed under a Creative Commons BY-NC-SA 4.0 license.

TOC

Version