Laravel - enum like variable? - php

I would like to be able to pass enums to functions. I know PHP doesn't support them per se, but there are instances where an enum would make sense.
For instance, take:
function doPayment($payment_method = 'paypal')
{ // ... code goes here }
You can call it via doPayment('paypal'), doPayment(), doPayment('creditcard'), or doPayment('creditcadr');
This introduces the chance of typos plus it's a pain to keep strings consistent and to debug. We could pass them as ints such as doPayment(2) but that's ugly and we'd have to keep track of them or put them in defines, which is the next example: doPayment(PAYMENT_METHOD_PAYPAL); but then we'd have to put the define in some global file that is accessible by all. Also I find that you end up with a class of hundreds of defines and it's hard to keep track of what belongs to what.
Meanwhile in Java, it's easy:
doPayment(PaymentClass.PaymentMethod.PayPal); // or doPayment(PaymentMethod.PayPal);
That's really what I'm looking for if possible. I guess I could create an external class for each set of enums but that adds a lot of files as well (while the Java example you can create an enum within the parent class).
I was wondering if Laravel has added a method to do this, if it's not doable in PHP.

You could do something like:
class PaymentMethod extends SplEnum {
const __default = 1;
// Our enum values
const PAYPAL = 1;
const IDEAL = 2;
//..etc
}
Then, calling your function:
doPayment(PaymentMethod::PAYPAL);
Laravel is properly namespaced, so I would go on to create a helper class and add this to the autoloader, this way you just need to include the namespace as a "use" in top of the Controllers where you want to use it. You can also add the path to the class in your project's composer.json file
an example from composer documentation:
{
"autoload": {
"psr-4": {
"Monolog\\": "src/",
"helper\\": "path/to/helpers"
}
}
}

Related

make NAMESPACE declaration conditional?

I include myfile.php file 2 times.. I want to know if there is way to put in the start of myfile.php something like this:
if($authorized){
Namespace MyProject;
}
function ABC(){}
function EFG(){}
(I do this to avoid function-redefine errors. Please dont ask me why I do such thing.. I know there are if function_exists and etc.. but I need answers to what I ask).
update:
I do this, because I need to define function XYZ() and using that before other framework is loaded (which newly defines globally function XYZ()). So I want to use that function (with different functionality) before framework loads, and after framework loads, it should behave as framework decides.
Is there a way to make namespace's conditional?
Although this is impossible directly, I could suggest using the define() method to create this yourself. Where the file requiring the class needs to have a allowed definition to access the file.
define('IN_NAMESPACE', 0);
if(defined('IN_NAMESPACE')) {
// authorized
}
But if you're worrying about a class name being repeated, namespaces are for the declaration of environments so you do not get duplicates, for example:
namespace Environments\One;
class Example { }
namespace Environments\Two;
Class Example { }
use Environments\One\Example as ExampleOne;
use Environments\Two\Example as ExampleTwo;
$e_o = new ExampleOne();
$e_t = new ExampleTwo();
Or simply, directly say you will use this environment like so:
$e_o = new Environments\One\Example();
$e_t = new Environments\Two\Example();
But again, this issue is not canonical to PHP. Use MVC methodologies to over-come these issues and Singleton/Dependency Injection design patterns.

PSR-4: Autoloader (composer) and extending namespaces ensuring fallback php

I am having a problem with my namespace fallbacks and using PSR-4 loader in Composer.
What I am trying to do is this:
Have a core which can overwritten / extended.
The core is based off an interface.
The directory structure is like so:
site/app/View/Example.php
site/src/ACME/app/View/Example.php
site/src/ACME/app/Interface/View.php
I am not set on this configuration so if you have a better suggestion then go for it.
My composer json is like so for psr-4:
"autoload": {
"psr-4": {
"ACME\\App\\Site\\" : "app/",
"ACME\\App\\" : "src/AMCE/app/"
}
}
I thought this would make ACME\App\Site\View fallback to ACME\App\View if the site one was not found (Note I haven't done the interface part yet...).
My code for site/app/View/Example.php is like so:
namespace ACME\App\Site\View;
class ViewExample extends View {
Which works, when I have site/app/View/View.php as well. That looks like:
namespace ACME\App\Site\View;
class View extends \ACME\App\View\View {
The site/src/app/View/View.php look like this:
namespace ACME\APP\View;
class View {
This one should use the interface (I haven't tried yet).
So what I really want to do is make it so I don't have to have site/app/View/View.php, and I don't have to have site/app/View/Example.php - it can use site/src/ACME/app/View/Example.php.
Sorry I'm new to namespaces so I may not of phrased it very well.
What I am getting at is I thought ACME\App\Site would fallback to ACME\App - it doesn't? Or I am doing it wrong? At the moment it needs all the files in place.
Edit: Turns out I was originally wrong, it is possible to get your example working with PSR-4! You just need to specify an array of directories for namespaces that can be loaded from different places.
Easy solution
{
"autoload": {
"psr-4": {
"ACME\\App\\Site\\": ["app/", "src/ACME/app"],
"ACME\\App\\": "src/ACME/app/"
}
}
}
Personally, I would rather name my namespaces more explicitly, see below.
Original Answer
The composer PSR-4 loader does not fall back when trying to load files that do not exist. It just fails immediately. Its flow looks like:
\ACME\App\Site\View is not loaded
Scan PSR-4 entries for matching namespaces
Class name matches the namespace \ACME\App\Site (your first PSR-4 entry).
Load file app/View.php
File does not exist. Error.
It never goes back to step 3 and tries the next namespace.
So how do I fix it?
It looks like you want to separate your reusable library code from your site code. If that's the case, I would use separate namespaces. For example, use the ACME\Site namespace to hold your reusable code, and use ACME\MySiteName for your site-specific code. Then there will be no ambiguity, and composer will have no trouble loading your classes.
But I don't want to rearrange my namespaces!
Ok, that's fine, but you'll have to use a hack to get around your problem. Composer has a classmap loader, and you'll have to use that instead of the preferred PSR-4 loader.
{
"autoload": {
"classmap": ["app/", "src/"]
}
}
Let's separate the things a bit, because they are all mixed up for now.
What I am trying to do is this:
Have a core which can overwritten / extended.
The core is based off an interface.
This sounds like basic object oriented inheritance. An interface defines the proposed public behaviour, the core implements the needed basics, and the detail implementation changes some parts, and reuses the others.
Let's write your example code in a way PHP sees it with absolute namespace names:
class \ACME\App\Site\View\ViewExample extends \ACME\App\Site\View\View {}
class \ACME\App\Site\View\View extends \ACME\App\View\View {}
class \ACME\App\View\View {}
You have three classes explicitly named. You'd need three files that match the namespace and class name. The autoloading does not need to do any detection whether or not a class is present - because you cannot optionally inherit from a class that isn't there, or omit it otherwise.
On the other hand, implementing three levels of inheritance by default very likely is too much. It looks like bad design to me, and will make maintaining the code harder than necessary. Depending on what you want to achieve, there are plenty of alternatives to get what you want easier. For example, to change some details of behavior, there are the decorator pattern or the strategy pattern.
So what I really want to do is make it so I don't have to have site/app/View/View.php, and I don't have to have site/app/View/Example.php - it can use site/src/ACME/app/View/Example.php.
You cannot have this. Your code explicitly states that it inherits from \ACME\App\Site\View\View, so this class MUST be present somewhere.
This is independent of any autoloading. To experiment, you can add all your code into one single file and then run it. This will make all classes known to PHP immediately, and the problem will become obvious: You cannot remove a class when at the same time other classes inherit it.
Sorry I'm new to namespaces so I may not of phrased it very well.
Namespaces are nothing really fancy, the same problem would arise if you would use the PSR-0 style classnames with underscores:
class ACME_App_Site_View_ViewExample extends ACME_App_Site_View_View {}
// This class MUST be present for the above class to work
class ACME_App_Site_View_View extends ACME_App_View_View {}
class ACME_App_View_View {}
The main new feature with namespaces is that you can import one class under a second name within a file with use OtherNamespace\Classname. But this is only an alias within the scope of this file (i.e. it does not affect other files or the global scope).
Namespaces and autoloading are not the right tool for this job. A namespace is just a way of making sure two people (or parts of your code) don't use the same name to mean different things. Autoloading is just a way to avoid having to list every source file you want to load code from.
When you override the behaviour of one class in another, these are not the same class; often, you'll want to inherit the default actions and reuse parts of them.
You might want to create several sub-classes for different purposes, so you need somewhere to hold the logic of which to use. The component which deals with this is called a "service locator" or sometimes a "DI container".
Namespaces let you map short names to longer, unique class names; autoloading let's you map a specific unique class name to a source file; service location is how you choose which unique class you want to use in a specific circumstance.

How to split a PHP class to separate files?

I want to do something like class categories in Objective-C, to define a class in one file, implement some core methods there, then implement some helper methods in another without subclassing or interfaces, just "continue" the class.
Possible in PHP?
As of PHP 5.4, you can accomplish this with Traits. To quote the official documentation:
Traits [enable] a developer to reuse sets of methods freely in several independent classes living in different class hierarchies. [...] A Trait is similar to a class, but only intended to group functionality in a fine-grained and consistent way. It is not possible to instantiate a Trait on its own. It is an addition to traditional inheritance and enables horizontal composition of behavior; that is, the application of class members without requiring inheritance.
Using a trait allows you to store helper methods that address cross-cutting concerns in a central place and use these methods in whatever classes you need.
// File "HelperMethodTrait.php"
trait HelperMethodTrait
{
public function myIncredibleUsefulHelperFunction()
{
return 42;
}
}
// File "MyClass.php"
class MyClass
{
use HelperMethodTrait;
public function myDomainSpecificFunction()
{
return "foobar";
}
}
$instance = new MyClass();
$instance->myIncredibleUsefulHelperFunction(); // 42
a-priori PHP doest not give you this feature by language construction, the whole class must be in a single file.
Now there are tricks that could do the stuff, but they are quite heavy CPU demanding:
You can for example dissociate your classes between different files, then load the files and build a single string with the whole code, then do an eval(string) to interprete the classes and build them into the runnable code area.
not really a good idea btw for many reasons
If your only goal is to make the class definition file smaller, you can include files within a function definition
class MyClass {
public function doStuff($foo,$bar) {
global $quz;
include('MyClass.doStuff.php');
}
}
Within the include, your local scope applies ($foo,$bar and $quz will be defined)
I'm not sure what the overhead is - the include seems to happen at run-time, not compile time (since an include can actually return a value ..)
I think you better split your class in n parts along functional criteria, and then use the dependency injection pattern to "inject" class A into class B in the constructor for B, without a need for subclassing/interfaces.
You could use this method to add functions to some objects externally, which is different from adding methods to the class itself.
https://stackoverflow.com/a/2938020/2277620
Another not-so-good-way is it create a multiple subclasses so every level of subclasses add only a 'group' of functions.
I used this trick in a single case where a class was more than 10k rows.
Now I've 6 levels of inheritance where every level add a single 'category' of functions the the "previous level" class.
Not so cpu-consuming, ugly to see, not so difficult to maintain
It is highly dependant on what you wish to achieve. I had this bright idea doing what you wish to do. I likes the key words set, get run delete etc. Suppose there was a class and I wanted set of delete. As Ph.T pointed out class needs to be in single file
So make abc->set in one file and abc->get in another file made no use and sense.
So had to lot of rethinking and think of classes as micro services doing specific tasks broken down in functionality.
Solution went with name spacing.
namespace A;
class Get{public static function myCode(){}
public static function myAnotherCode(){}}
this would then be called A\Get::myCode(); or A\Get::myAnotherCode();
Similarly set would be
namespace A;
class Set{public static function myCode(){}
public static function myAnotherCode(){}}
This would then be called A\Set::myCode(); or A\Set::myAnotherCode();
Then user spl_autoload_register to load classes remember to replace \ with /
spl_autoload_register(function ($class) {
require_once __DIR__ . '/' . strtolower(str_replace('\\', '/', $class) . '.php')
});
and directory structure is as following
main (folder)
-- A (folder)
-- set.php (file)
-- get.php (file)
we asked for A\Get::myCode()
require would see it as A/Get.php then run its logic whether it is right name space or class etc.
good side of it if planned out right is it would force you to run things run in logical not just wishy washy there and there. Second you would only load functionality what you need not just whole bunch of this and that.
down side two parts of classes wont share between them unless one extends one from another which would defy the whole logic because other one would be loaded as a consequence any ways.
Careful negation has to be made as in what logic and planning data will be shared and has to be worked out before hand.
namespace A;
class Universal {
protected static $share_one;
protected static $share_two;
}
namespace A;
class Set extends Universal {
public static function myCode(){ parent::$share_one ='wow';}
}
namespace A;
class Get extends Universal {
public static function myCode(){ return parent::$share_one;}
}
main (folder)
-- A (folder)
-- universal.php (file)
-- set.php (file)
-- get.php (file)
run the code and see the magic
A\Set::myCode();
echo A\Get::myCode();
if you only need one set of functionality e.g. get then just use get echo A\Get::myCode(); would work without errors.
To be honest without playing around with name spaces its just expenditure of time earning headaches.
Side note: if done right it can speed up things as class universal wont load twice. even if called gazillion times whats what we want load defaults and then populate them where ever we need them. I have made huge classes where besides defaults only needed few functions rest was just memory load. breaking down just spead up things.

Is this a good way to use Namespaces in PHP

I have studied the use of Namespaces in PHP a while back but recently looking at a project that used the use keyword and then accessed the namespaced object as if they were normal without namespace.
My question is, is the code below correct, it hs a file index.php and uses the namespace MyLibrary\Base it then uses use to bring in \MyLibrary\Registry \MyLibrary\User and \MyLibrary\Request
It then can access any of these object without putting there namespace in front of them, so the actual code below the use section looks like a normal pre-namespace php file.
I am asking if this is how you use namespaces? Or am I missing something?
File: index.php
<?php
namespace MyLibrary\Base;
use \MyLibrary\Registry;
use \MyLibrary\User;
use \MyLibrary\Request;
class Base
{
public $registry;
function __construct($registry)
{
$this->registry = $registry;
$this->user = New User;
$this->request = new Request;
# code...
}
}
?>
File: registry.class.php
<?php
namespace MyLibrary\Registry;
class Registry
{
public $user;
function __construct($user)
{
$this->user = $user;
# code...
}
}
?>
Yes. The use-statement imports the class- or namespace-name into the current scope. To write everything that short is the reason, why the PHP-devs implemented namespaces ;)
namespace MyFirstNamespace {
class Foo {}
}
namespace MySecondNamespace {
use \MyFirstNamespace\Foo as Bar;
$foo = new Bar;
}
a) it make everything more readable, because its much shorter, than Vendor_Package_Foo_Bar_XyzClass and b) you can exchange the classes to use very fast.
# use \MyFirstNamespace\Foo as Bar; // I don't like Foo anymore
use \MyFirstNamespace\SimilarToFoo as Bar;
Namespacing has a lot of advantages to it.
The first on is you can reuse method names and even class names if it makes sense provided they exist within a different namespace. Example:
namespace \myNamespace\data\postgres;
class DataBase extends \PDO
{
}
namespace \myNamespace\data\mysql;
class DataBase extends \PDO
{
}
You could even reuse names that are normally reserved for PHP functions
namespace \myNamespace\dir;
function makedir ()
{
if (// some condition is true)
{
\makedir ();
}
}
All of this is intended to make it easier to use code from different sources together without having to worry about naming conflicts. Provided programmers are courteous enough to observe a few simple rules then the chances of name conflicts are hugely reduced. Actually, pretty much the only rule you need to concern yourself with to avoid naming conflicts is to make the first level of your namespace your own. For example, use your company name, or some other way of identifying you as a unique vendor as the first level of your namespace and everything should be good.
For example I use \gordian as the root namespace in all the code I write, as I can then call my classes under that namespace anything I like without worrying about them colliding with someone who chose a different root namespace.
So what's wrong with PEAR conventions, you might ask? A lot of projects follow them, including the popular Zend framework.
The answer is names become very unwieldy very quickly. For instance, Zend, as it follows the PEAR convention, uses a sort of pseudo-namespacing. All the classes in the collection start with Zend_ and with each level of class hierachy add a further part to the name.
Ze
As a result, you end up with class names like Zend_Db_Adaptor_Abstract and Zend_Dojo_Form_Decorator_TabContainer.
Should Zend update their framework to use namespaces (which I'm told is happening with Zend Framework 2.0) then they'd be replaced with \Zend\Db\Adaptor\Abstract and \Zend\Dojo\Form\Decorator\TabContainer instead. So what, you might ask? The answer is that you can alias them to much shorter names with the Use keyword, as you've already seen. This means you don't have to keep writing the full class name out, but only as far as to what you've aliased.
use \Zend\Dojo\Forn\Decorator as Dec;
$a = new Dec\TabContainer; // Not easy to do without namespaces!
Further more, if you're already in a given namespace, then you don't even have to use the use keyword to access other items within the same namespace by a short name, as it happens automatically for you in that case. For framework writers this is a huge timesaver.
For example, you might see something like the followign in Zend Framework 2 (as I'm not working on it in any way, this is purely an example and not from the actual ZF2 source).
namespace \Zend\Db\Adaptor;
class Postgres extends Abstract // We don't need to use \Zend\Db\Adaptor\Abstract here because it's in the same namespace already anyway
{
}
There are other benefits too, such as it makes autoloaders ridiculously simple to make (provided your namespace structure maps exactly onto your filesystem directory structure).
Namespaces can seem like one of those features that aren't really very important, or don't even seem to make any sense, but after using them for a little while their usefulness will suddenly become very obvious.

Dynamically Create Subclass

I am using Kohana and just found this piece of code in their autoloading method
// Class extension to be evaluated
$extension = 'class '.$class.' extends '.$class.'_Core { }';
// Start class analysis
$core = new ReflectionClass($class.'_Core');
if ($core->isAbstract())
{
// Make the extension abstract
$extension = 'abstract '.$extension;
}
// Transparent class extensions are handled using eval. This is
// a disgusting hack, but it gets the job done.
eval($extension);
basically what it does is when I am referring to a class that doesn't exist (through object instantiation, calling class_exists() etc.), Kohana will create a class (eg. Foo) that extends a library class that follows certain naming convention (eg. Foo_Core). just being curious, is there any way to do something similar but without the use of eval?
If you are looking to create a dynamic class then eval() is the goto function (pun intended.) However kind of related, I've found that you can put a class declaration within an if-then statement. So you can do the following:
if(true)
{
class foo
{
// methods
}
}
I use this to check to see if dynamically created classes (from a configuration file) are current... if so then load the class, otherwise... regenerate the class, and load the new one. So if you're looking to create dynamic classes for similar reasons this might be a solution.
I think you're stuck with eval() for that.
It's marked as a "disgusting hack" so that makes it OK :)
I'd be interested to know what you do with such an empty class...
If you were wanting to be able to cache your dynamically created classes, you could write out to a file and require it instead. This could be considered equally hack-ish, but it is an option. For classes that are created once and used often, it might be a good solution. For classes that need to be dynamic every time, sticking with eval is probably the best solution.
$proxyClassOnDisk = '/path/to/proxyCodeCache/' . $clazz .'.cachedProxyClass';
if ( ! file_exists($proxyClassOnDisk) ) {
// Generate the proxy and put it into the proxy class on disk.
file_put_contents($proxyClassOnDisk, $this->generateProxy($object));
}
require_once($proxyClassOnDisk);
In this example, the idea is that you are creating dynamic proxies for the class of $object. $this->generateProxy($object) will return the string that looks more or less like what $extension looks like in the original question.
This is by no means a complete implementation, just some pseudo code to show what I'm describing.

Categories