Connection Sharing and Structuring of Model Objects - php

Question

I'm writing my own OOP framework, partially as a learning exercise, but knowing this codebase isn't going to disappear - only evolve as I learn. Below is some psuedocode of what my setup looks like. In this setup, how would I share DB connections? Do a ctrl+f for "connection" to find the places in the code I'm specifically unsure of how to structure.
A routing file would look like this:
switch($urlParameters['action']){
case 'lading':
//etc etc
break;
case 'userCP':
//etc etc
break;
case 'dashboard':
default:
$page = new dashboard();
$page->route($urlParameters);
break;
}
A page would look like
<?php
abstract class page{ // Page is a mixture of controller and view. Most view logic is frontend, so I do all my controller-view logic here
public var $models;
public var $session;
public var $user;
function __construct(){
$this->models = new models();
$this->session = $this->models->session(new MongoID($_COOKIE['session'])); // Plus some code to prevent session hijacking
$this->user = $this->models->users($this->session->userID);
}
function outputMongoData(){} // Wrapper function used by pages to output Mongo datatypes as regular json (Dates to RFC, IDs to string, rounding floats, etc)
function template($file, $data){} // This uses output buffering and extract() to use php its self as the templating language.
}
class dashboard extends page{
public var $dashboardView;
function __construct(){
parent::__construct();
$this->dashboardView = $this->models->dashboardView($this->user->dashboardViewID);
}
function route($urlParams){
$this->session->updatePageViewCountOrSomething();
echo $this->template('dashboard.php', [
'importantData' => $this->dashboardView->someTypeOfDataFromDashboardView,
'viewCount' => $this->session->pageViewCount
]);
}
}
And the models look like
<?php
class models{
private var $cache = [
'users' => [],
'dashboardViews' => [],
'sessions' => [],
];
// Cache to prevent collisions since we're doing save() vs individual updates, plus saves overhead of querying multiple times (potentially hundreds of times for certain models)
function __construct(){
// Potential connection sharing stuff here
}
private function modelFactory($collection, $modelName, $identifier){
$stringID = (string) $identifier;
if(!isset($this->cache[$collection][$stringID])){
$this->cache[$collection][$stringID] = new $modelName($identifier);
}
return $this->cache[$collection][$stringID];
}
public function session(MongoID $sessionID){
return $this->modelFactory('sessions', 'session', $sessionID);
}
public function user(MongoID $userID){
return $this->modelFactory('users', 'user', $userID);
}
public function dashboardView(MongoID $dashboardViewID){
return $this->modelFactory('dashboardViews', 'dashboardView', $dashboardViewID);
}
}
abstract class model{
public var $db;
public var $collection;
private var $data = [];
function __construct($collectionName){
$this->db = // ??? Not sure how to connect so that I don't have a new connection for every single model..
$this->collection = $this->db->{$collectionName};
}
function __destruct(){
$this->collection->save($this->data);
}
function __get($fieldName){
return $this->data[$fieldName]; // if isset, etc etc
}
}
class session extends model{
// Explanation of schema here
function __construct(MongoID $sessionID){
parent::__construct('sessions');
$this->data = $this->collection->findOne(['_id' => $sessionID]); // or if not found, create..
}
function updatePageViewCountOrSomething(){
$this->data['pageViewCount'] += 1;
$this->data['orSomething'] = ['something' => 'or another'];
}
}
class dashboardView extends model{
// Explanation of schema here
function __construct(MongoID $dashboardViewID){
parent::__construct('dashboardViews');
$this->data = $this->collection->findOne(['_id' => $sessionID]); // or if not found, create..
}
function addColumns(){}
function reorderColumns($newOrder){}
}

Answer 1

From your example, it looks like the models class is functioning as a factory for your model instances. In that case, you could use it as the connection container and then pass the appropriate MongoCollection instance to each model class instead of the collection name.
I'll also note that in the PHP driver, there is not really a concern with constructing multiple MongoClient objects. The driver uses persistent connections internally, so constructing five MongoClient instances all with the same host/port/user/password combination will still only create a single network socket behind the scenes. You can test this out for yourself using the example in this slide and the one following.
Some general concerns with your model implementation:
$data is private to the base model class, but you access it in the child classes. It should probably be protected.
Waiting until __destruct() to save an object back to the database seems like a bad idea, as it could happen at the end of the script or sometime earlier if the object is garbage-collected. Having a predictable point in the application to write data seems preferable.
Instead of using this model for persisting session data, you can read up on the SessionHandlerInterface in PHP 5.4+. That allows you to implement a custom storage backend (MongoDB in this case) to be used by the standard PHP session API (e.g. $SESSION).
A great starting point before creating a framework (even as a thought exercise), would be to read up on design patterns. Anthony Ferrara's blog is a great resource and starting point for the subject. I would also suggest learning about dependency injection and service containers, as that's very relevant to your original question about sharing a common resource (e.g. database connection). Fabien Potencier did a great blog series on the subject. The article is a few years old and references the Symfony 1.x implementation (Symfony2 is the current model he uses), but the early posts will likely serve as a useful introduction.