Thin controller for RESTful applications and APIs.
- Very thin and lightweight.
- Don't try to change PHP, small learning curve.
- Completely RESTful, the right way to build apps.
Packages available on PEAR and Composer. Autoloading is PSR-0 compatible.
Bootstrapping is easy. Just create an instance of Respect\Rest\Router.
<?php
use Respect\Rest\Router;
$r3 = new Router;
This assumes you have a .htaccess
file that redirects every request to this PHP file and
you're running this from the domain root (http://example.com/ without any subfolder).
If you want to use it from a subfolder, you can pass the virtual root to the Router:
<?php
$r3 = new Router('/myapp');
This will instruct the router to work from http://example.com/myapp/.
You can also use the Router without a .htaccess
file. This uses the CGI PATH_INFO
variable,
and can be declared as:
<?php
$r3 = new Router('/index.php/');
The same goes for folders:
<?php
$r3 = new Router('/myapp/index.php/');
This assumes that every URL in the project will begin with these namespaces.
The Router is auto-dispatched, which means that you don't have to call anything more than declaring routes to run it. If you want to omit this behavior, you can set:
<?php
$r3->isAutoDispatched = false;
You can then dispatch it yourself at the end of the proccess:
<?php
print $r3->run();
You can print the output or store in a variable if you want. This allows you to better test and integrate the Router into existing applications.
The Hello World route goes something like this:
<?php
$r3->get('/', function() {
return 'Hello World';
});
Hitting http://localhost/
(consider your local configuration for this) will print
"Hello World" in the browser. You can declare as many routes as you want:
<?php
$r3->get('/hello', function() {
return 'Hello from Path';
});
Hitting http://localhost/hello
will now print "Hello from Path".
You can declare routes that receives parameters from the URL. For this, every parameter
is a /*
on the route path. Considering the previous sample model:
<?php
$r3->get('/users/*', function($screenName) {
echo "User {$screenName}";
});
Accessing http://localhost/users/alganet
or any other username besides alganet
will
now print "User alganet" (or the username of your choosing).
Multiple parameters can be defined:
<?php
$r3->get('/users/*/lists/*', function($user, $list) {
return "List {$list} from user {$user}.";
});
Last parameters on the route path are optional by default, so declaring just
a ->get('/posts/*'
will match http://localhost/posts/
without any
parameter. You can declare a second ->get('/posts'
, now the Router will
match it properly, or treat the missing parameter yourself by making them
null
able on the passed function:
<?php
$r3->get('/posts/*/*/*', function($year,$month=null,$day=null) {
//list posts, month and day are optional
});
- This will match /posts/2010/10/10, /posts/2011/01 and /posts/2010
- Optional parameters are allowed only on the end of the route path. This
does not allow optional parameters:
/posts/*/*/*/comments/*
Sometimes you need to catch an undefined number of parameters. You can use Routes with catch-all parameters like this:
<?php
$r3->get('/users/*/documents/**', function($user, $documentPath) {
return readfile(PATH_STORAGE. implode('/', $documentPath));
});
- The above sample will match
/users/alganet/documents/foo/bar/baz/anything
. Callback $user parameter will receive alganet and $documentPath will receive an array filled with [foo,bar,baz,anything]. - Catch-all parameters are defined by a double asterisk
/**
. - Catch-all parameters must appear only on the end of the path. Double asterisks in any other position will be converted to single asterisks.
- Catch-all parameters will match after any other route that matches the same pattern.
Things can became very complex quick. We have simple routes, routes with parameters, optional parameters and catch-all parameters. A simple rule to keep in mind is that Respect\Rest matches the routes from the most specific to the most generic.
- Routes with the most slashes
/
are more specific and will be matched first. - Routes with parameters are less specific than routes without parameters.
- Routes with multiple parameters are even less specific than a route with one parameter.
- Routes with catch-all parameters are the least specific and will be matched last.
Summing up: Slashes and asterisks places your route at the top of the priority list to match first.
Respect\Rest sort routes automatically, but it is highly recommended to declare routes from the most specific to the most generic. This will improve performance and maintainability of the code.
Sometimes you need to use a router to proxy requests to some other router or map
requests to a class. By using the magic method any
, you can pass any HTTP method to a given
function.
<?php
$r3->any('/users/*', function($userName) {
//do anything
});
- Any HTTP method will match this same route.
- You can figure out the method using the standard PHP
$_SERVER['REQUEST_METHOD']
The any
method is extremely useful to bind classes to controllers, one of Respect\Rest's most
awesome features:
<?php
use Respect\Rest\Routable;
class MyArticle implements Routable {
public function get($id) { }
public function delete($id) { }
public function put($id) { }
}
$r3->any('/article/*', 'MyArticle');
- This route will bind the class methods to the HTTP methods for the given path.
- Parameters will be sent to the class methods just like the callbacks from the previous examples.
- Controllers are lazy loaded and persistent. The MyArticle class will be instantiated only when a route matches one of its methods, and this instance will be reused on subsequent callbacks (redirects, etc).
- Classes must implement the interface Respect\Rest\Routable for safety reasons. (Imagine someone mapping HTTP to a PDO class automatically, that wouldn't be right).
Passing constructor arguments to the class is also possible:
<?php
$r3->any('/images/*', 'ImageController', array($myImageHandler, $myDb));
- This will pass
$myImageHandler
and$myDb
as parameters for the ImageController class constructor.
You can also instantiate the class yourself if you want:
<?php
$r3->any('/downloads/*', $myDownloadManager);
- Sample above will assign the existent
$myDownloadManager
as a controller. - This instance is also reused by Respect\Rest
And you can even use a factory or DI container to build the controller class:
<?php
$r3->any('/downloads/*', 'MyControllerClass', array('Factory', 'getController'));
- Sample above will use the MyController class returned by Factory::getController
- This instance is also reused by Respect\Rest
- Third parameter is any callable variable, so you can put a closure there to build an instance if you want.
Sometimes you need to route users to streams. The Router doesn't have to first handle large files or wait for streams to finish before serving them.
<?php
$r3->get('/images/*/hi-res', function($imageName) {
header('Content-type: image/jpg');
return fopen("/path/to/hi/images/{$imageName}.jpg", 'r');
});
This will redirect the file directly to the browser without keeping it in memory.
CAUTION: We created a possible security vulnerability in the sample: passing a parameter
directly to a fopen
handle. Please validate user input parameters before using them.
This is for demonstrational purposes only.
No surprises here, you can make a route return a plain string:
<?php
$r3->get('/greetings', 'Hello!');
Respect\Rest has an internal forwarding mechanism. First you'll need to understand that every route declaration returns an instance:
<?php
$usersRoute = $r3->any('/users', 'UsersController');
Then you can use
and return
this route in another one:
<?php
$r3->any('/premium', function($user) use ($db, $usersRoute) {
if (!$db->userPremium($user)) {
return $usersRoute;
}
});
Illustrative sample above will redirect internally when an user is not privileged to another route that handle normal users.
Respect\Rest uses a different approach to validate route parameters:
<?php
$r3->get('/documents/*', function($documentId) {
//do something
})->when(function($documentId) {
return is_numeric($documentId) && $documentId > 0;
});
- This will match the route only if the callback on when is matched.
- The
$documentId
param must have the same name in the action and the condition (but does not need to appear in the same order). - You can specify more than one parameter per condition callback.
- You can specify more than one callback:
when($cb1)->when($cb2)->when($etc)
- Conditions will also sync with parameters on binded classes and instance methods.
This makes it possible for the user to validate parameters using any custom routine and
not just data types such as int
or string
.
We highly recommend that you use a strong validation library when using this. Consider Respect\Validation.
<?php
$r3->get('/images/*/hi-res', function($imageName) {
header('Content-type: image/jpg');
return fopen("/path/to/hi/images/{$imageName}.jpg", 'r');
})->when(function($imageName) {
// Using Respect Validation alias to `V`
return V::alphanum(".")->length(5,155)
->noWhitespace()->validate($imageName);
});
Sometimes you need to run something before a route does its job. This is useful for logging, authentication and similar purposes.
<?php
$r3->get('/artists/*/albums/*', function($artistName, $albumName) {
//do something
})->by(function($albumName) use ($myLogger) {
$myLogger->logAlbumVisit($albumName);
});
- This will execute the callback defined with by before the route action which needs to match a route.
- Parameters are also synced by name and not by order, like with
when
. - You can specify more than one parameter per proxy callback.
- You can specify more than one proxy:
by($cb1)->by($cb2)->by($etc)
- A
return false
from a proxy will stop the execution of any following proxies as well as the route action. - Proxies will also sync with parameters on binded classes and instance methods.
If your By routine returns false
, then the route method/function will not be
processed. If you return an instance of another route, an internal forward
will be performed.
Similar to ->by
, but runs after the route did its job. In the sample
below we're showing something similar to invalidating a cache after
saving some new information.
<?php
$r3->post('/artists/*/albums/*', function($artistName, $albumName) {
//save some artist info
})->through(function() use($myCache) {
$myCache->clear($artistName, $albumName);
});
by
proxies will be executed before the route action,through proxies
will be executed after.- You are free to use them separately or in tandem.
through
can also receive parameters by name.
Sample above allows you to do something based on the route parameters, but when procesing something after the route has run, its desirable to process its output as well. This can be achieved with a nested closure:
<?php
$r3->any('/settings', 'SetingsController')->through(function(){
return function($data) {
if (isset($settings['admin_user'])) {
unset($settings['admin_user']);
}
return $data;
};
});
The illustrative sample above removes sensitive keys from a settings controller before outputing the result.
When using routines you are encouraged to separate the controller logic into components. You can reuse them:
<?php
$logRoutine = function() use ($myLogger, $r3) {
$myLogger->logVisit($r3->request->path);
};
$r3->any('/users', 'UsersController')->by($logRoutine);
$r3->any('/products', 'ProductsController')->by($logRoutine);
A simple way of applying routines to every route on the router is:
<?php
$r3->always('By', $logRoutine);
You can use the param sync to take advantage of this:
<?php
$r3->always('When', function($user=null) {
if ($user) {
return strlen($user) > 3;
}
});
$r3->any('/products', function () { /***/ });
$r3->any('/users/*', function ($user) { /***/ });
$r3->any('/users/*/products', function ($user) { /***/ });
$r3->any('/listeners/*', function ($user) { /***/ });
Since there are three routes with the $user
parameter, when
will
verify them all automatically by its name.
Respect/Rest currently supports the four distinct types of Accept header content-negotiation: Mimetype, Encoding, Language and Charset. Usage sample:
<?php
$r3->get('/about', function() {
return array('v' => 2.0);
})->acceptLanguage(array(
'en' => function($data) { return array("Version" => $data['v']); },
'pt' => function($data) { return array("Versão" => $data['v']); }
))->accept(array(
'text/html' => function($data) {
list($k,$v)=each($data);
return "<strong>$k</strong>: $v";
},
'application/json' => 'json_encode'
));
As in every routine, conneg routines are executed in the same order in which
you appended them to the route. You can also use ->always
to apply this
routine to every route on the Router.
Please note that when returning streams, conneg routines are also called. You may take advantage of this when processing streams. The hardcore example below serves text, using the deflate encoding, directly to the browser:
<?php
$r3->get('/text/*', function($filename) {
return fopen('data/'.$filename, 'r+');
})->acceptEncoding(array(
'deflate' => function($stream) {
stream_filter_append($stream, 'zlib.deflate', STREAM_FILTER_READ);
return $stream; //now deflated on demand
}
));
When applying conneg routines to multiple routes that can return streams you
(really) should check for is_resource()
before doing anything.
Support for Basic HTTP Authentication is already implemented as a routine:
<?php
$r3->get('/home', 'HomeController')->authBasic('My Realm', function($user, $pass) {
return $user === 'admin' && $user === 'p4ss';
});
You'll receive an username and password provided by the user, and you just need to return true or false. True means that the user could be authenticated.
Respect\Rest will handle the authentication flow, sending the appropriate headers when unauthenticated. You can also return another route, which will act as an internal forward (see the section on forwarding above).
Below is an illustrative sample of how to block requests from mobile devices:
<?php
$r3->get('/videos/*', 'VideosController')->userAgent(array(
'iphone|android' => function(){
header('HTTP/1.1 403 Forbidden');
return false; //do not process the route.
}
));
You can pass several items in the array, like any conneg routine. The array key is a regular expression matcher without delimiters.
Note that this is not currently implemented.
By default, HTML forms send POST data as multipart/form-data
, but API clients
may send any other format. PUT requests often send other mime types. You can pre-process
this data before doing anything:
<?php
$r3->post('/timeline', function() {
return file_get_contents('php://input');
})->contentType(array(
'multipart/form-data' => function($input) {
parse_str($input, $output);
return $output;
},
'application/json' => function($input) {
return my_json_converter($input);
},
'text/xml' => function($input) {
return my_xml_converter($input);
},
));
Respect\Rest currently handle the following errors by default:
- 404, when no matching route paths are found.
- 401, when the client sends an unauthenticated request to a route using
authBasic
routine. - 405, when a matching path is found but the method isn't specified.
- 400, when a
when
validation fails. - 406, when the route path and method matches but content-negotiation doesn't.
- A HEAD request automatically works sending all GET headers without body. You can override
this behavior declaring custom
head
routes. - An OPTIONS request to
*
or any route path returns theAllow
headers properly. - When returning 405,
Allow
headers are also set properly.
- You can set
$r3->methodOverriding = true
to allow?_method=ANYMETHOD
on the URI to override default HTTP methods. This isfalse
by default.
Routines are classes in the Respect\Rest\Routines namespace, but you can add your own routines by instance using:
<?php
$r3->get('/greetings', 'Hello World')->appendRoutine(new MyRoutine);
In the sample above, MyRoutine
is a user provided routine declared as a class and
appended to the router. Custom routines have the option of several different interfaces
which can be implemented:
- IgnorableFileExtension - Instructs the router to ignore the file extension in requests
- ParamSynced - Syncs parameters with the route function/method.
- ProxyableBy - Instructs the router to run method
by()
before the route. - ProxyableThrough - Instructs the router to run method
through()
after the route. - ProxyableWhen - Instructs the router to run method
when()
to validate the route match. - Unique - Makes this routine be replaced, not appended, if more than one is declared for the same type.
You can use any combination of the above but also need to implement the Routinable interface.