I'm programming in PHP procedurally (is this even a word?) for about five years now and decided to try an OOP approach but ran into some concept/design problems. Let's say you have some modules in the program, every module has the possibility to list, add, edit and delete an entity. An entity can be..dunno, user, client, product etc.
How would you design the classes to manipulate these entityes?
Two possibilities came in my mind:
create classes for every entity with methods like getUsersList, addUser, editUser, delUser
This approach seems resource consumingbecause in the script for the listing you only need the getUsersList and maybe delUser methods, while in the add user popup script, you only need the addUser method and in the edit user popup script only the editUser method. So, you have to instanciate an object and only use two or one of it's methods...
create general classes: listing, add, edit and delete and extend them for every entity this way you only have to instanciate one class at a time (the one you really need)
Thanks in advance,
I would create an interface defining your list, add, edit, and delete methods. This gives you a class "template". If your classes (User, Client, Product, etc.) implement this interface, then the methods in the interface must be defined in those classes.
This will give you a similar "API" to access all the functionality of every class that implements your interface. Since each of your listed objects contains different data, the details of the methods will be different, and thus separate, but the interface will be the same.
Aside:
Your inclusion of "list" in your list of methods concerns me a little. It seems to imply that you are seeing your objects as collections of Users, Clients, Products, etc, where there should most likely be a User class that represents a single user, a Client class that represents a single client, etc.
On the other hand, "list" may be handled as a static method - a method that can be called without an instance of the class.
$bob = new User('bob');
$bob->add(); // Adds bob to the database
$fred = new User('fred');
$fred->add(); // Adds fred to the database
$users = User::list(); // Gives an array of all the users in the database
That's how I would handle things, anyway.
You will need to create a solid architecture and framework for managing your data model. This is not easy and will only get more complex as the data model grows. I would highly recommend using a PHP framework (Symfony, CakePHP, etc), or at least, an ORM (Doctrine, Propel, etc).
If you still want to roll your own, I would start with an architecture similar to below.
You will want a DbRecord class that is used for individual record operations (like saving, deleting, etc). This DbRecord class will be extended by specific entities and will provide the foundation for basic entity operations.
class DbRecord {
public function save() {
// save logic (create or update)
}
public function delete() {
// delete logic
}
// other record methods
}
class User extends DbRecord {
private $name;
private $email;
public function setName($name_) {
$this->name = $name_;
}
public function setEmail($email_) {
$this->email = $email_;
}
}
From which, you can perform individual record operations:
$user = new User();
$user->setName('jim');
$user->setEmail('jim#domain.com');
$user->save();
You will now want a DbTable class that is used for bulk operations on the table (like reading all entities, etc).
class DbTable {
public function readAll() {
// read all
}
public function deleteAll() {
// delete all logic
}
public function fetch($sql) {
// fetch logic
}
// other table methods
}
class UserTable extends DbTable {
public function validateAllUsers() {
// validation logic
}
// ...
}
From which, you can perform bulk/table operations:
$userTable = new UserTable();
$users = $userTable->readAll();
foreach ($users as $user) {
// etc
}
Code architecture is the key to a website scaling properly. It is important to divide the data model into the appropriate classes and hierarchy.
Again, as your website grows, it can get very complicated to manage the data model manually. It is then when you will really see the benefit of a PHP framework or ORM.
NOTE: DbRecord and DbTable are arbitrary names - use w/e name you like.
Use your first method, where you create a reusable object with methods. It is not a waste of time as you only code it once.
class User {
function __construct() { /* Constructor code */ }
function load($id) { ... }
function save() { ... }
function delete() { ... }
}
You're on the right track with 'general classes' (also called base classes, or abstract classes in case their behaviour NEEDS to be complemented by child classes before they can be put to use).
The OOP approach would be to put all behavior that is common to all entities in the base classes.
If you use something akin to ActiveRecord, you already have a general (abstract) interface for create-update-delete operations. Use that to your advantage, and let your base classes operate ONLY on those interface methods. They don't need to know they are updating a Product, or a a User, they just need to know they can call the update() method on an entity.
But even without using something quite feature-heavy like an AR framework (check out Doctrine if you're interested in a very flexible ORM..) you can use interfaces to abstract behavior.
Let me give you a more elaborate example...
/**
* Interface for all entities to use
*/
interface Entity {
static function newEntity();
static function fetch($id);
function save();
function setProperties(array $properties);
function delete();
}
/**
* A concrete product entity which implements the interface
*/
class Product implements Entity {
public $productId;
public $name;
public $price;
public $description;
/**
* Factory method to create a new Product
*
* #param integer $id Optional, if you have auto-increment keys you don't need to set it
* #return Product
*/
public static function newEntity($id=NULL) {
$product = new Product();
$product->productId = $id;
return $product;
}
/**
* Factory method to fetch an existing entity from the database
*
* #param integer $id
* #return Product
*/
public static function fetch($id) {
// make select with supplied id
// let $row be resultset
if (!$row) {
return NULL; // you might devise different strategies for handling not-found cases; in this case you need to check if fetch returned NULL
}
$product = new Product();
$product->productId = $id;
$product->name = $row['name'];
$product->price = $row['price'];
$product->description = $row['description'];
return $product;
}
/**
* Update properties from a propreties array
* #param array $properties
* #return void
*/
public function setProperties(array $properties) {
$this->name = $properties['name'];
$this->price = $properties['price'];
$this->description = $properties['description'];
}
public function save() {
// save current product properties to database
}
public function delete() {
// delete product with $this->productId from database
}
}
/**
* An abstract CRUD controller for entities
*/
abstract class EntityCrudController {
protected $entityClass = 'UNDEFINED'; // Override this property in child controllers to define the entity class name
protected $editTemplate = NULL; // Override this to set an edit template for the specific entity
protected $templateEngine; // Pseudo-Templating engine for this example
/**
* Display the edit form for this entity
* #param integer $entityId
* #return string
*/
public function editAction($entityId) {
// Fetch entity - this is not the most clean way to fetch, you should probably consider building a factory that encapsulates this.
$entity = call_user_func($this->entityClass, 'fetch', $entityId);
// Assign entity to your edit template, in this example I'm assuming we're using a template engine similar to Smarty
// You can generate the HTML output in any other way you might like to use.
$this->templateEngine->setTemplate($this->editTemplate);
$this->templateEngine->assign('entity', $entity);
return $this->template->render();
}
/**
* Update an existing entity
*
* #param integer $entityId
* #param array $postArray
* #return string
*/
public function updateAction($entityId, array $formArray) {
// Be sure to validate form data first here, if there are errors call $this->editAction() instead and be sure to set some error information
$entity = call_user_func($this->entityClass, 'fetch', $entityId);
$entity->setProperties($formArray);
$entity->save();
// Again, using our imaginary templating engine to display...
$this->templateEngine->setTemplate($this->editTemplate);
$this->templateEngine->assign('entity', $entity);
$this->templateEngine->assign('message', 'Saved successfully!');
return $this->template->render();
}
// Devise similar generic methods for newAction/insertAction here
}
/**
* Concrete controller class for products
* This controller doesn't do much more than extend the abstract controller and override the 2 relevant properties.
*/
class ProductCrudController extends EntityCrudController {
protected $entityClass = 'Product';
protected $editTemplate = 'editProduct.tpl';
}
// Usage example:
// Display edit form:
$controller = new ProductCrudController();
$htmlOutput = $controller->editAction(1);
// Save product:
$htmlOutput = $controller->updateAction(1, array('name' => 'Test Product', 'price' => '9.99', 'description' => 'This is a test product'));
Of course, there is much to improve.. e.g. you generally don't want to make a query everytime you call fetch() on an entity, but instead only query once and store the resulting object in an IdentityMap, which also ensures data integrity.
Hope this helps, got a bit more than I intended, but I think it's commendable you try to tackle this without throwing a framework on the problem :)
Related
For historical reasons, my pattern of running databases using Symfony is mixed. That is, the query uses DBAL and the insert uses ORM. Now you need to write a lot of data to the database. The flush in ORM can help me achieve business at the lowest cost.
All flush operations have been removed from the project. Put it in the __destruct of the controller.
However, doing so will cause DBAL to not find the latest changed data. Of course, these data ORMs can be obtained normally.
This is a very difficult problem. I hope to get guidance.
class BaseController extends Controller
{
public function __destruct()
{
$this->getDoctrine()->getManager()->flush();
}
public function indexAction()
{
$model = new CompanyModel();
$model->install(['company_name' => '1234']);
$model->update(['company_name' => 'abcd'], $model->lastInsertId);
}
}
class CompanyModel extends BaseController
{
public function validate($data, $id = false)
{
$this->entityManager = $this->getDoctrine()->getManager();
if(empty($id)){
$this->company_class = new Company();
}else{
if(!$this->is_exist($id)){
return false;
}
$this->company_class = $this->entityManager->getRepository(Company::class)->find($id);
}
if(array_key_exists('company_name', $data)){
$this->company_class->setCompanyName($data['company_name']);
}
if(self::$error->validate($this->company_class)){
return false;
}
return true;
}
public function insert($data)
{
if(!$this->validate($data)){
return false;
}
$this->company_class->setCreateAt(new \DateTime());
$this->entityManager->persist($this->company_class);
//$this->entityManager->flush();
$this->lastInsertId = $this->company_class->getId();
return true;
}
public function update($data, $id)
{
if(empty($id)){
self::$error->setError('param id is not null');
return false;
}
if(!$this->validate($data, $id)){
return false;
}
$this->company_class->setUpdateAt(new \DateTime());
//$this->entityManager->flush();
return true;
}
public function is_exist($id)
{
return $this->get('database_connection')->fetchColumn('...');
}
}
The final result of executing indexAction company_name is 1234; $ model-> update() was not executed successfully. The reason is that the $this-> is_exist() method that took the DBAL query did not find the ORM insert but did not flush the message.
Unchanging conditions,run
$this->entityManager->getRepository(Company::class)->find($id);
Is successful。
The problem is not the entity manager or dbal, as far as I can tell, but the usage of an anti-pattern, which I would call ... entanglement. What you should strive for is separation of concerns. Essentially: Your "CompanyModel" is an insufficient and bad wrapper for the EntityManager and/or EntityRepository.
No object should know about the entity manager. It should only be concerned with holding the data.
The entity manager should be concerned with persistence and ensuring integrity.
The controller is meant to orchestrate one "action", that can be adding one company, editing one company, batch-importing/updatig many companies.
Services can be implemented, when actions become to business-logic-heavy or when functionality is repeated.
(Note: the following code samples could be made way more elegant with using all the features that symfony provide, like ParamConverters, the Form component, the Validation component, I usually wouldn't write code this way, but I assume everything else would go way over your head - no offence.)
handling actions in the controller
controller actions (or service actions, really) are when you look at your problem from the task perspective. Like "I want to update that object with this data"). That's when you fetch/create that object, then give it the data.
use Doctrine\ORM\EntityManagerInterface;
class BaseController extends Controller {
public function __construct(EntityManagerInterface $em) {
$this->em = $em;
}
public function addAction() {
$company = new Company(['name' => '1234']); // initial setting in constructor
$this->em->persist($company);
// since you have the object, you can do any changes to it.
// just change the object
$company->update(['name' => 'abcd']); // <-- don't need id
// updates will be flushed as well!
$this->em->flush();
}
public function editAction($id, $newData) {
$company = $this->em->find(Company::class, $id);
if(!$company) {
throw $this->createNotFoundException();
}
$company->update($newData);
$this->em->flush();
}
// $companiesData should be an array of arrays, each containing
// a company with an id for update, or without an id for creation
public function batchAction(array $companiesData) {
foreach($companies as $companyData) {
if($companyData['id']) {
// has id -> update existing company
$company = $this->em->find(Company::class, $companyData['id']);
//// optional:
// if(!$company) { // id was given, but company does not exist
// continue; // skip
// // OR
// $company = new Company($companyData); // create
// // OR
// throw new \Exception('company not found: '.$companyData['id']);
// }
$company->update($companyData);
} else {
// no id -> create new company
$company = new Company($companyData);
$this->em->persist($company);
}
}
$this->em->flush(); // one flush.
}
}
the base controller should handle creating objects, and persisting it, so very basic business logic. some would argue, that some of those operations should be done in an adapted Repository for that class, or should be encapsulated in a Service. And they would be right, generally.
the entity handles it's internal state
Now, the Company class handles its own properties and tries to stay consistent. You just have to make some assumptions here. First of all: the object itself shouldn't care if it exists in the database or not. it's not its purpose! it should handle itself. Separation of concerns! The functions inside the Company entity should concern simple business logic, that concerns its INNER state. It doesn't need the database, and it should not have any reference to the database, it only cares about it's fields.
class Company {
/**
* all the database fields as public $fieldname;
*/
// ...
/**
* constructor for the inital state. You should never want
* an inconsistent state!
*/
public function __construct(array $data=[]) {
$this->validate($data); // set values
if(empty($this->createAt)) {
$this->createAt = new \DateTime();
}
}
/**
* update the data
*/
public function update(array $data) {
$this->validate($data); // set new values
$this->updateAt = new \DateTime();
}
public function validate(array $data) {
// this is simplified, but you can also validate
// here and throw exceptions and stuff
foreach($array as $key => $value) {
$this->$key = $value;
}
}
}
some notes
Now, there should be NO use case, where you get an object to persist and at the same time an update - with an id - that refers to the new object ... unless that object was given the id beforehand! HOWEVER. If you persist an object, that has an ID and you call $this->em->find(Company::class, $id) you would get that object back.
if you have many relations, there are always good ways to solve this problem without destroying separation of concerns! you should never inject an entity manager into an entity. the entity should not manage its own persistence! nor should it manage the persistence of linked objects. handling persistence is the purpose of the entity manager or entity repository. you should never need a wrapper around an object just to handle that object. be careful not to mix responsibilities of services, entities (objects) and controllers. In my example code, I have merged services and controllers, because in simple cases, it's good enough.
In my application I use Laravel's authentication system and I use dependency injection (or the Facade) to access the logged in user. I tend to make the logged in user accessible through my base controller so I can access it easily in my child classes:
class Controller extends BaseController
{
protected $user;
public function __construct()
{
$this->user = \Auth::user();
}
}
My user has a number of different relationships, that I tend to eager load like this:
$this->user->load(['relationshipOne', 'relationshipTwo']);
As in this project I'm expecting to receive consistently high volumes of traffic, I want to make the application run as smoothly and efficiently as possible so I am looking to implement some caching.
I ideally, need to be able to avoid repeatedly querying the database, particularly for the user's related records. As such I need to look into caching the user object, after loading relationships.
I had the idea to do something like this:
public function __construct()
{
$userId = \Auth::id();
if (!is_null($userId)) {
$this->user = \Cache::remember("user-{$userId}", 60, function() use($userId) {
return User::with(['relationshipOne', 'relationshipTwo'])->find($userId);
});
}
}
However, I'm unsure whether or not it's safe to rely on whether or not \Auth::id() returning a non-null value to pass authentication. Has anyone faced any similar issues?
I would suggest you used a package like the following one. https://github.com/spatie/laravel-responsecache
It caches the response and you can use it for more than just the user object.
Well, after some messing about I've come up with kind of a solution for myself which I thought I would share.
I thought I would give up on caching the actual User object, and just let the authentication happen as normal and just focus on trying to cache the user's relations. This feels like quite a dirty way to do it, since my logic is in the model:
class User extends Model
{
// ..
/**
* This is the relationship I want to cache
*/
public function related()
{
return $this->hasMany(Related::class);
}
/**
* This method can be used when we want to utilise a cache
*/
public function getRelated()
{
return \Cache::remember("relatedByUser({$this->id})", 60, function() {
return $this->related;
});
}
/**
* Do something with the cached relationship
*/
public function totalRelated()
{
return $this->getRelated()->count();
}
}
In my case, I needed to be able to cache the related items inside the User model because I had some methods inside the user that would use that relationship. Like in the pretty trivial example of the totalRelated method above (My project is a bit more complex).
Of course, if I didn't have internal methods like that on my User model it would have been just as easy to call the relationship from outside my model and cache that (In a controller for example)
class MyController extends Controller
{
public function index()
{
$related = \Cache::remember("relatedByUser({$this->user->id})", 60, function() {
return $this->user->related;
});
// Do something with the $related items...
}
}
Again, this doesn't feel like the best solution to me and I am open to try other suggestions.
Cheers
Edit: I've went a step further and implemented a couple of methods on my parent Model class to help with caching relationships and implemented getter methods for all my relatonships that accept a $useCache parameter, to make things a bit more flexible:
Parent Model class:
class Model extends BaseModel
{
/**
* Helper method to get a value from the cache if it exists, or using the provided closure, caching the result for
* the default cache time.
*
* #param $key
* #param Closure|null $callback
* #return mixed
*/
protected function cacheRemember($key, Closure $callback = null)
{
return Cache::remember($key, Cache::getDefaultCacheTime(), $callback);
}
/**
* Another helper method to either run a closure to get a value, or if useCache is true, attempt to get the value
* from the cache, using the provided key and the closure as a means of getting the value if it doesn't exist.
*
* #param $useCache
* #param $key
* #param Closure $callback
* #return mixed
*/
protected function getOrCacheRemember($useCache, $key, Closure $callback)
{
return !$useCache ? $callback() : $this->cacheRemember($key, $callback);
}
}
My User class:
class User extends Model
{
public function related()
{
return $this->hasMany(Related::class);
}
public function getRelated($useCache = false)
{
return $this->getOrCacheRemember($useCache, "relatedByUser({$this->id})", function() {
return $this->related;
});
}
}
Usage:
$related = $user->getRelated(); // Gets related from the database
$relatedTwo = $user->getRelated(true); // Gets related from the cache if present (Or from database and caches result)
We have a fairly large symfony2 code base. Generally our Controller actions would look something like
public function landingPageAction(Request $request) {
//do stuff
return $this->render("view_to_render", $template_data);
}
We have two functionalities that are very generic between all of our controllers:
We tend to pass Controller level template parameters to all of the actions in a specific controller - let's call these "Default Parameters"
We set HTTP cache headers at the end of each Action
Understandably we want to abstract this logic away. In doing so we came up with two approaches. We are not certain which approach is better, both in terms of general OO and SOLID principles, but also in terms of performance and how SF2 recommends things be done.
Both approaches rely on having the controller extend an interface that indicates if the controller has "Default Parameters" (later we are considering also adding Cacheable interface)
use Symfony\Component\HttpFoundation\Request;
interface InjectDefaultTemplateVariablesController {
public function getDefaultTemplateVariables(Request $request);
}
Approach 1
This approach is based on events. We define an object that will store our template variables, as well as (in the future) cache indicators
class TemplateVariables {
protected $template_name;
protected $template_data;
public function __construct($template_name, $template_data) {
$this->template_name = $template_name;
$this->template_data = $template_data;
}
/**
* #param mixed $template_data
* #return $this
*/
public function setTemplateData($template_data) {
$this->template_data = $template_data;
return $this;
}
/**
* #return mixed
*/
public function getTemplateData() {
return $this->template_data;
}
/**
* #param mixed $template_name
* #return $this
*/
public function setTemplateName($template_name) {
$this->template_name = $template_name;
return $this;
}
/**
* #return mixed
*/
public function getTemplateName() {
return $this->template_name;
}
}
We also define events that will be triggered on render and which call the views
class InjectDefaultTemplateVariablesControllerEventListener {
/** #var DelegatingEngine */
private $templating;
private $default_template_variables;
public function __construct($templating) {
$this->templating = $templating;
}
public function onKernelController(FilterControllerEvent $event) {
$controller = $event->getController();
if (!is_array($controller)) {
return;
}
if ($controller[0] instanceof InjectDefaultTemplateVariablesController) {
$this->default_template_variables = $controller[0]->getDefaultTemplateVariables($event->getRequest());
}
}
public function onKernelView(GetResponseForControllerResultEvent $event) {
$controller_data = $event->getControllerResult();
if ($controller_data instanceof TemplateVariables) {
$template_data = (array)$controller_data->getTemplateData();
$template_data = array_merge($this->default_template_variables, $template_data);
$event->setResponse($this->templating->renderResponse($controller_data->getTemplateName(), $template_data));
}
}
}
Finally our Action now becomes
public function landingPageAction(Request $request) {
//do stuff
return new TemplateVariables("view_to_render", $template_data);
}
Approach 2
This approach is based on putting the common logic into a BaseController from which every other controller inherits. We are still keeping the approach of having Child controllers also extend an interface in case they want to use "Default Parameters".
The following is the new method in the base controller that determines if Default Parameters need to be merged with the specific template parameters. Later this method will also handle cache headers using ttl parameter.
public function renderWithDefaultsAndCache($view, array $parameters = array(), Response $response = null, $ttl = null)
{
$default_template_variables = array();
if ($this instanceof InjectDefaultTemplateVariablesController ) {
$default_template_variables = $this->getDefaultTemplateVariables();
}
$template_data = array_merge($default_template_variables, $parameters);
return $this->render($view, $template_data, $response);
}
Action now becomes
public function landingPageAction(Request $request) {
//do stuff
return $this->renderWithDefaultsAndCache("view_to_render", $template_data);
}
Discussion
So far the main arguments for the first approach were that it follows SOLID principles and is easier to extend - iin case more common logic were to be added, it can be put directly into Event Listeners without affecting the controllers.
The main arguments for the second approach were that the logic we are trying to abstract away actually does belong to the controller and not an external event. In addition there was a concern that using events in this manner will result in a poor performance.
We would be really grateful to hear from the experts on which approach is better or possibly suggest a third one that we have missed.
Thank you!
First off I am in no way claiming to be a Symfony 2 architecture expert.
I have a game schedule program which outputs a number of different types of schedules (public, team, referee etc). The various schedules are all similar in that they deal with a set of games but vary in details. The schedules need to be displayed in various formats (html,pdf,xls etc). I also wanted to be able to further tweak things for individual tournaments.
I originally used your second approach by creating a ScheduleBaseController and then deriving various individual schedule controllers from it. It did not work well. I tried to abstract common functionality but the schedules were just different enough that common functionality became complicated and difficult to update.
So I went with an event driven approach very similar to yours. And to answer one of your questions, adding some event listeners will not have any noticeable impact on performance.
Instead of focusing on template data I created what I call an Action Model. Action models are responsible for loading the games based on request parameters and (in some cases) updating the games themselves based on posted data.
Action models are created in the Controller event listener, stored in the request object and then passed to the controller's action method as an argument.
// KernelEvents::CONTROLLER listener
$modelFactoryServiceId = $request->attributes->get('_model');
$modelFactory = $this->container->get($modelFactoryServiceId);
$model = $modelFactory->create($request);
$request->attributes->set('model',$model);
// Controller action
public function action($request,$model)
{
// do stuff
// No template processing at all, just return null
return null;
}
// KernelEvents::VIEW listener
$model = $request->attributes->get('model')
$response = $view->renderResponse($model);
So the controller is mostly responsible for form stuff. It can get data from the model if need be but let's the model handle most of the data related stuff. The controller does no template processing stuff at all. It just returns null which in turn kicks off a VIEW event for rendering.
Lot's of objects? You bet. The key is wiring this up in the route definition:
// Referee Schedule Route
cerad_game__project__schedule_referee__show:
path: /project/{_project}/schedule-referee.{_format}
defaults:
_controller: cerad_game__project__schedule_referee__show_controller:action
_model: cerad_game__project__schedule_referee__show_model_factory
_form: cerad_game__project__schedule_referee__show_form_factory
_template: '#CeradGame\Project\Schedule\Referee\Show\ScheduleRefereeShowTwigPage.html.twig'
_format: html
_views:
csv: cerad_game__project__schedule_referee__show_view_csv
xls: cerad_game__project__schedule_referee__show_view_xls
html: cerad_game__project__schedule_referee__show_view_html
requirements:
_format: html|csv|xls|pdf
Each part is broken up into individual services which, for me at least, makes it easier to customize individual sections and to see what is going on. Is it a good approach? I don't really know but it works well for me.
I preferred double layer models (mapper and model) over doctrine in my zend framework 2 project and trying to make them work little bit like doctrine so I can access relational data from the models (entities). Following example demonstrates what I am trying to achieve.
class User
{
protected $userTable;
public $id;
public $name;
public function __construct($userTable)
{
$this->userTable = $userTable
}
public function getUserArticles()
{
return $this->userTable->getUserArticles($this->id);
}
}
Problem is I cannot inject my user table in my user model, because table gateway uses model class as array object prototype which gets later injected to create user table gateway (mapper).
I don't want to inject service manager in my models as it is considered as a bad practice. How can I inject my user table in my user model? is it possible? what is the best way to achieve what I am trying to do
What you are trying to do is mix two design patterns: Active Record and Data Mapper.
If you take a look at the Data Mapper pattern, you have the Mapper that accesses both the Model and the database. The Model is passive - usually does not call external resources (it's a POPO - Plain Old PHP Object).
A solution for your issue is to inject the related information into the Model, thus keeping the Model only as a data structure.
Here is a working scenario for an MVC application:
Controller - used for input validation & retrieving data from services
<?php
...
public function viewAction()
{
$id = (int) $this->params()->fromQuery('id');
$service = $this->getServiceLocator()->get('your-user-service-name');
$user = $service->getUser($id);
...
}
Service - used for executing the business logic; calls multiple Data Mappers
<?php
...
public function getUser($id)
{
// get user
$mapper = $this->getServiceLocator()->get('your-user-mapper');
$user = $mapper->getUserById($id);
// get articles
$article_mapper = $this->getServiceLocator()->get('your-article-mapper');
$user->articles = $article_mapper->getArticlesByUser($id);
return $user;
}
Data Mapper - used to manipulate one type of Domain entity - it should be composed with a tableGateway if you are accessing the database
<?php
...
public function getUserById($id)
{
$select = $this->tableGateway->getSql()->select();
$select = $select->where(array('id' => $value));
$row = $this->tableGateway->selectWith($select)->current();
return $row;
}
Domain Model - used for data representation
<?php
...
class User
{
public $name; // user name
...
public $articles; // holds the user articles
}
Advantages
Passive Models are easy to read - understand the data structure and it's relations.
Passive Models are easy to test - you don't need external dependencies.
You separate the persistence layer from the Domain layer.
Hope this helps!
You should not inject your mapper into your model, that's exactly the other way around. Important for you to understand is the way the relations work and models shouldn't have any knowledge how their data is mapped to a persistency framework.
You refer to Doctrine, so I'd suggest you also look at how Doctrine solves this problem. The way they do it is via a Proxy. A proxy is a generated class (you need to write your own generator or write all proxies yourself) which extends the model and have the mapper injected:
class Foo
{
protected $id;
protected $name;
public function getId()
{
return $this->id;
}
public function getName()
{
return $this->name;
}
public function setName($name)
{
$this->name = $name;
}
}
class ProxyFoo extends Foo
{
protected $mapper;
public function __construct(FooMapper $mapper)
{
$this->mapper = $mapper;
}
public function getName()
{
if (null === $this->name) {
$this->load();
}
return parent::getName();
}
protected function load()
{
$data = $this->mapper->findById($this->id);
// Populate this model with $data
}
}
My suggestion: look either at the default mapper pattern Zend Framework 2 applies and forget lazy loading, or just use Doctrine. This is too much work to get this done properly.
I have an entity that needs to return an instance of another entity - the stored record or a new one if one has not been stored.
<?php
/**
* #Entity
*/
class User {
/**
* #OneToOne(targetEntity="Basket")
* #JoinColumn(nullable=true)
*/
protected $activeBasket;
public function getActiveBasket() {
if (!isset($this->activeBasket)) {
$this->activeBasket = new Basket($this);
// $em->persist($this->activeBasket);
// $em->flush();
}
return $this->activeBasket;
}
}
My problem is that I don't have an EntityManager to use to persist this new Basket (and obviously don't want one in the Model). I'm not sure as to the best way to do this. I do want to be able to call $user->getActiveBasket() and retrieve a basket, whether it's previously created or a new one.
It feels like this should be a common problem, or that there's a better way to structure this (I'm hoping there's a way to hook one in in an EntityRepository or something). How can I do this?
I wouldn't return a new Basket() when it is null or not set. The model (entity) should only serve for setting and getting its properties.
The logic when $user->getActiveBasket() returned null should be elsewhere - in controller or in entity's reporitory object...
So I would move public function getActiveBasket() { ... } into just public function getActiveBasket() { return $this->activeBasket; } and somewhere in the controller I would do:
$if(!$user->getActiveBasket()) {
$user->setActiveBasket(new Basket($user));
$this->_em->persist($user);
$this->_em->flush();
}
// now do something with that activeBasket...