Laravel, autoload classes based on an input array of class names - laravel

I have various classes, e.g.
<?php
namespace MyApp\Notifications;
class FirstNotification implements NotificationInterface {
public function getMessage() {
return 'first message';
}
}
and
<?php
namespace MyApp\Notifications;
class SecondNotification implements NotificationInterface {
public function getMessage() {
return 'second message';
}
}
I then have an array like: ['First','Second'].
I'm using:
foreach (['First','Second'] as $class_prefix) {
$class = "MyApp\Notifications\\{$class_prefix}Notification";
$object = new $class();
echo $object->getMessage();
}
but it feels a bit hacky - is there a better/more standard way to do this? The array is supplied elsewhere and will be different depending on the user - my aim is to be able to easily create classes that implement my interface and know this loop will be able to show their messages if they exist.
I ideally don't want to have to add a use statement for all the classes upfront, or pass them into a constructor, I just want magic to happen!

I'm assuming you don't actually mean autoload but rather instantiate the class. Autoloading is the process of including all the files in your application so the contents (usually classes) can be used. If you are using Laravel 5 and you follow PSR-4 (namespace matches directory structure) you don't have to do anything to make those classes available.
The code you already have looks fine and probably works. However you could make use of Laravels Service Container to resolve the class. This has quite a few advantages, one being the availability of automatic dependency injection...
foreach (['First','Second'] as $class_prefix) {
$object = app()->make("MyApp\Notifications\\{$class_prefix}Notification");
echo $object->getMessage();
}
Or even this:
foreach (['First','Second'] as $class_prefix) {
echo app()->callClass("MyApp\Notifications\\{$class_prefix}Notification#getMessage");
}
However, both will cause an exception if the class doesn't exist. You can check for that beforehand or just catch the exception:
foreach (['First','Second'] as $class_prefix) {
try{
$object = app()->make("MyApp\Notifications\\{$class_prefix}Notification");
echo $object->getMessage();
}
catch(ReflectionException $e){
// whooops
}
}

Related

How can I add extra view component class paths in Laravel?

The problem
The default namespace for view components is App\View\Components with the folder being app/View/Components. I am setting up a DDD file structure and wish to do two things:
Move "shared" view components to a namespace and folder of App\ViewComponents and src/app/ViewComponents respectively
Have view components specific to individual "apps" with their own namespace and folder of App\MyApplication\ViewComponents and /src/app/MyApplication/ViewComponets respectively.
The new App namespace/folder setup is done via composer psr-4 autoload keys and works fine. But Laravel always used the App\View\Components namespace when trying to load components.
My attempt
I have solved the first part of my problem, but I am hoping that there is a better way. For instance when I want to move views, I can just set the view.paths config directive in my AppServiceProvider but I don't see a similar way of, essentially, adding namespaces to where Laravel looks for view components. So what I ended up doing was:
Create a ViewServiceProvider class, extending Illuminate\View\ViewServiceProvider::class and point to it in bootstrap/app.php instead
In there, override the registerBladeEngine method, in there pointing towards my own BladeCompiler class instead of the built-in one
public function registerBladeEngine($resolver)
{
// The Compiler engine requires an instance of the CompilerInterface, which in
// this case will be the Blade compiler, so we'll first create the compiler
// instance to pass into the engine so it can compile the views properly.
$this->app->singleton('blade.compiler', function () {
return new BladeCompiler(
$this->app['files'],
$this->app['config']['view.compiled'],
);
});
$resolver->register('blade', function () {
return new CompilerEngine(
$this->app['blade.compiler']
);
});
}
In my own BladeCompiler class, which extends Illuminate\View\Compilers\BladeCompiler, override the component() and compileComponentTags() methods - basically anywhere that referenced View\\Components - with pretty much a carbon copy but instead using ViewComponents and also made sure that where they return a Illuminate\View\Compilers\ComponentTagCompiler I instead referenced my own ComponentTagCompiler
In my own TagCompiler I override the guessClassName() method, again with essentially a carbon copy, just renaming View\\Components to ViewComponents
As you can see, that's quite a lot of work just to change the path. And I also want to add another path. Multiple "apps" run under the same Laravel codebase, so for instance we might have App\Website\, App\Admin and App\Blog and, depending on which app is currently running, load a different namespace for the running app, i.e. the blog would be App\Blog\ViewComponents pointing to src/app/Blog/ViewComponents.
Is there a way to achieve this without as much overriding as above? If not, can you suggest a way to achieve the second part of the requirement?
Note: I haven't ruled out using sub folders and continuing with everything under the main App\View\Components namespace just yet - I don't want to fight Laravel more than I have to and am willing to concede if there's no better way, but if I can achieve the folder structure I want it would feel a lot tidier.
Update; got a working implementation by using a configuration and php 8 annotations
Follow the steps below to make it possible to add more lookup folders for the blade view components feature, based on your question and details you have provided. It would have helped to have posted that code you already had. But I have added a possible solution to get it to work, using Annotations and using a config with a namespace/path map.
Depending on how you switch between one application and the other, from which the details are not provided in your question, you have to modify the way the configuration is retreived in the MyComponentTagCompiler class.
Blade compiler
In order to change the ComponentTagCompiler we need to change the BladeCompiler class:
namespace App;
class YourBladeCompiler extends \Illuminate\View\Compilers\BladeCompiler
{
protected function compileComponentTags($value)
{
if (! $this->compilesComponentTags) {
return $value;
}
return (new \App\MyComponentTagCompiler( //it is about this line
$this->classComponentAliases, $this->classComponentNamespaces, $this
))->compile($value);
}
}
Service provider
Now register the YourBladeCompiler in YourViewServiceProvider :
class YourViewServiceProvider extends \Illuminate\View\ViewServiceProvider
{
public function registerBladeEngine($resolver)
{
$this->app->singleton('blade.compiler', function () {
return new \App\YourBladeCompiler( //it is about this line
$this->app['files'],
$this->app['config']['view.compiled'],
);
});
$resolver->register('blade', function () {
return new CompilerEngine(
$this->app['blade.compiler']
);
});
}
}
MyComponentTagCompiler
This is an implemention I created that works with PHP 8 Attributes, given below:
namespace App;
#[\Attribute]
class ViewComponentName
{
public string $name;
public string $package;
public function __construct(string $name, string $package)
{
$this->name = $name;
$this->package = $package;
}
}
With this attribute, you can declare the package name and component name on the view component class (see example at the bottom). So during lookup the component can be matched on these parameters.
But you can change it to your own requirements if needed.
What it does:
It first let's Laravel lookup the View Component through it's own mechanisms, in the parent::componentClass method.
If no component is found and an exception (InvalidArgumentException) is thrown, after which my implementation will walk through the given paths and namespaces (from the getLookupPaths method) and see if an attribute matches the component name and package name. If so it returns this class and the view component is loaded accordingly.
namespace App;
use App\View\ViewComponentName;
use Illuminate\View\Compilers\ComponentTagCompiler;
class MyComponentTagCompiler extends ComponentTagCompiler
{
protected function getLookupPaths() : array
{
/*
* add some logic here to get an application specific configuration
* since you have multiple application in one, I cannot know it works in your
* application, since the details are not provided in the question
*/
return config('view_component_paths');
}
private function getFiles(string $dir) : array
{
return scandir($dir);
}
private function isPhpFile(string $file) : bool
{
return strpos($file, ".php");
}
private function getClassNamespace(string $file, string $folderNamespace) : string
{
$class = str_replace(".php", "", $file);
$classNamespace = $folderNamespace . "\\" . $class;
return $classNamespace;
}
private function getComponentName(string $file, string $namespace) : ?ViewComponentName
{
$classNamespace = $this->getClassNamespace($file, $namespace);
$reflection = new \ReflectionClass($classNamespace);
if(method_exists($reflection, 'getAttributes')) {
$attribute = $reflection->getAttributes()[0];
if ($attribute->getName() == ViewComponentName::class) {
return $attribute->newInstance();
}
}
return null;
}
public function componentClass(string $component)
{
try {
parent::componentClass($component);
} catch(\InvalidArgumentException $e) {
list($lookupComponentPackage, $lookupComponentName) = explode("-", $component);
foreach($this->getLookupPaths() as $namespace=>$dir) {
foreach ($this->getFiles($dir) as $file) {
if ($this->isPhpFile($file)) {
if($componentName = $this->getComponentName($file, $namespace)) {
if($componentName->name == $lookupComponentName && $componentName->package == $lookupComponentPackage) {
return $this->getClassNamespace($file, $namespace);
}
}
}
}
}
throw $e;
}
}
}
Where the config contains (config/view_component_paths.php):
return [
"App\\Test"=>__DIR__ . "/Test/"
];
If you wish to replace the default laravel behavior completely or do not like my implementation based on annotations, consider implementing your own version of the method:
public function componentClass(string $component)
{
//return the class name here based the component name
//without calling parent
dd($component);
}
Example view component
namespace App\Test;
use App\View\ViewComponentName;
use Illuminate\View\Component;
#[ViewComponentName('test', 'namespace')]
class MyViewComponent extends Component
{
public function render()
{
return view('components.test');
}
}
In blade:
<x-namespace-test />
It should now be working. I think this is enough information to give you an idea of how to implement this in your own application. There seems to be no other way than to extend some base classes. But looking at this answer, it is possible to create a high level implementation based on a global lookup configuration and php annotations (or some other mechanism you wish, for example converting the class name with namespace to a view component name).
Old answer
Problem 2 as defined in your question
Have view components specific to individual "apps" with their own namespace and folder of App\MyApplication\ViewComponents and /src/app/MyApplication/ViewComponets respectively.
Sadly there seems to be no way of defining multiple class paths for view components in Laravel. But you can however change the application path and namespace prefix. As far as I found out you only have to overwrite the following properties in the Application class.
bootstrap/app.php
Replace the following lines:
$app = new Illuminate\Foundation\Application(
$_ENV['APP_BASE_PATH'] ?? dirname(__DIR__)
);
With:
class YourApplication extends \Illuminate\Foundation\Application
{
protected $namespace = "App\\MyApplication";
protected $appPath = __DIR__ . "/../app/MyApplication";
}
$app = new YourApplication(
$_ENV['APP_BASE_PATH'] ?? dirname(__DIR__)
);
This is enough to change the app folder to another and gives you an idea of how to change it dynamically for having multiple apps in different namespaces. If you now run laravel commands like php artisan make:component Test1234 it is created in your new app folder: app/MyApplication/View/Components/Test1234.php.
Hardcoded paths
Some paths like View/Components are hardcoded in Laravel, and therefor not that easy to change. If you change as defined above, in this case the view components namespace becomes: App\MyApplication\View\Components and the path: app/MyApplication/View/Components.
Problem 1 as defined in your question
Move "shared" view components to a namespace and folder of App\ViewComponents and src/app/ViewComponents respectively
When you change application paths as explained above it is not possible to have a "shared" View Component folder. Laravel, as it seems, has only one default View Components path, which is based on hard coded paths and a dynamic namespace prefix as explained above. But you can of course, create a shared namespace and register the view components manually:
View component (app/ViewComponents/ folder)
namespace App\ViewComponents;
use Illuminate\View\Component;
class Test extends Component
{
public function render()
{
return view('components.test');
}
}
Don't forget the components.test blade view.
ServiceProvider
\Blade::component("shared-test",\App\ViewComponents\Test::class);
Blade
<x-shared-test />

Mocking a service class inside controller

I am trying to write a Feature test for my controller. To simplify my current situation, imagine my controller looks like this:
public function store(Business $business)
{
try {
(new CreateApplicationAction())->execute($business);
} catch (Exception $e) {
return response()->json(['message' => 'error'], 500);
}
return response()->json(['message' => 'success']);
}
What I am trying to achieve is, instead of testing CreateApplication class logic inside my integration test, I want to write another unit test for it specifically.
Is there a way I can simply say CreateApplicationAction expects execute() and bypass testing inside it? (without executing the code inside execute())
/** #test */
public function can_create_application()
{
$business = Business:factory()->create();
$mock = $this->mock(CreateApplicationAction::class, function (MockInterface $mock) use ($business) {
$mock->shouldReceive('execute')
->once()
->with($business)
->andReturn(true);
});
$response = $this->post('/businesses/3/application', $data);
$response->assertOk();
}
I saw online that people create "MockCreateApplicationAction" class but if possible I don't want to create another class as I don't want any logic to be inside it at all.
Is it possible?
class CreateApplicationAction
{
public function execute($business) {
dd("A");
// Business Logic...
}
}
So when I do the Mock, dd() should never be called. Or I am going in the wrong direction?
You will need to use Laravels container to resolve your class. The basic approach is to use the resolve() method helper. PHP does not have dependency injection, so you need to use one to make it possible, in Laravel the container solves that.
resolve(CreateApplicationAction::class)->execute($business);
On constructors, controller methods, jobs, events, listeners and commands (rule of thumb if the method is named handle), you can inject classes into the parameters and they will resolve through the container.
public function store(Business $business, CreateApplicationAction $applicationAction)
{
try {
$applicationAction->execute($business);

Where and how to extract repeated business logic?

I have certain business logic that repeats throughout my code. Is there a standardized way of separating this out for version 4.2? I know in the newer versions, Commands are used for this type of thing. Example of repeating logic:
// detach tags
$card->tags()->detach();
foreach (Input::get('tags', []) as $tag)
{
// existing tag
if (is_numeric($tag) && Tag::whereId($tag)->exists())
{
$card->tags()->attach($tag);
}
// new tag
else
{
$new_tag = new Tag();
$new_tag->market()->associate($market);
$new_tag->name = $tag;
$new_tag->save();
$card->tags()->attach($new_tag->id);
}
}
The simplest way I could do this is create a new class, like TagManager and add methods for this logic, like createTag, attachTagToCard, etc. However, I'm not sure if there is already a "predefined" way of doing this.
The repeated code problem can be solved by using traits.
For more information on trait go here
Create a file called TagManagerTrait.php inside the folder app/traits
contents of file
<?php
namespace App\Traits;
trait TagManagerTrait {
public function createTag()
{
// your create tag logic
}
public function updateTag()
{
//your update tag logic
}
}
Update your composer.json to autoload this
{
"autoload" : {
"classmap" : [
"app/traits"
]
}
}
In the controllers where you want to use the repeated code
<?php
use App\Traits\TagManagerTrait;
class XyzController extends Controller{
use TagManagerTrait;
}
Now the controller will inherit all the functions of the trait and be used freely.

Checking passed object existence in Laravel delayed queue job

I believe if I pass an Eloquent object to a delayed Laravel job a "findOrFail" method is called to "restore" the object and pass it to the controller of my Job class.
The problem is that the DB record representing the object might be gone by the time the job is actually processed.
So "findOrFail" aborts before even calling the "handle" methods.
Everything seems fine. The problem is that the job now "gets transferred" to the failed jobs list. I know I can remove it from there manually, but that doesn't sound right.
Is there a way to "know" in my job class directly that the passed object "failed to load" or "does not exist" or anything similar?
Basically I would like to be able to do something if "ModelNotFoundException" is thrown while "rebuilding" my passed objects.
Thank you
SOLUTION:
Based on Yauheni Prakopchyk's answer I wrote my own trait and used it instead of SerializesModels where I need my altered behaviour.
Here's my new trait:
<?php
namespace App\Jobs;
use Illuminate\Queue\SerializesModels;
use Illuminate\Contracts\Database\ModelIdentifier;
use Illuminate\Database\Eloquent\ModelNotFoundException;
trait SerializesNullableModels
{
use SerializesModels {
SerializesModels::getRestoredPropertyValue as parentGetRestoredPropertyValue;
}
protected function getRestoredPropertyValue($value)
{
try
{
return $this->parentGetRestoredPropertyValue($value);
}
catch (ModelNotFoundException $e)
{
return null;
}
}
}
And that's it - now if the model loading fails I still get a null and can decide what to do with it in my handle method.
And if this is only needed in a job class or two I can keep using original trait everywhere else.
If i'm not mistaken, you have to override getRestoredPropertyValue in your job class.
protected function getRestoredPropertyValue($value)
{
try{
return $value instanceof ModelIdentifier
? (new $value->class)->findOrFail($value->id) : $value;
}catch(ModelNotFoundException $e) {
// Your handling code;
}
exit;
}

Something that escapes me about Symfony's controllers

Take a look to that code
<?php
namespace Sestante\SestanteBundle\Controller;
use Symfony\Bundle\FrameworkBundle\Controller\Controller;
use Symfony\Component\HttpFoundation\Request;
use Sestante\SestanteBundle\Model\StrutturaManager;
class MainController extends Controller
{
public function indexAction(Request $request)
{
return $this->render('SestanteSestanteBundle:Main:index.html.twig');
}
public function showLodgingsAction(Request $request)
{
$repo = $this->getDoctrine()->getRepository('SestanteSestanteBundle:Struttura');
$usr = $this->get('security.context')->getToken()->getUser();
$usrId = $usr->getId();
$sm = new StrutturaManager($repo);
$lodgingList = $sm->retrieveLodgingsFromUser($usrId);
return $this->render('SestanteSestanteBundle:Main:showLodgings.html.twig',array('lodgingList' => $lodgingList));
}
}
This is a controller for an application that I've been writing.
Take a look to showLodgingsAction. I've tryied to place all business logic into a model (StrutturaManager) that with a repository (that I've passed directly from controller because, as far I know, they're available only here or via D.I.) query my db, do some elaboration and return a list that I'll render onto a template.
First question: Is this "code separation" good, or exists a better way for do what I'm trying to do?
Second question: suppose that, now, I want to use an object of StrutturaManager type into indexAction. Remember that mine object want a repository. So, have I to declare again, and again, and again, all my object for every controller's action where I want to use them? I suppose that must exist a smarter method but, at the moment, I don't understand which.
Define StrutturaManager as a service and inject the EntityManager into it. This way the manager will have access to repositories you need and controllers won't know about Doctrine nor repositories — which is a good practice.

Resources