Download the PHP package mindy/router without Composer

PHRoute - Fast request router for PHP

This library provides a fast implementation of a regular expression based router.

• Super fast
• Route parameters and optional route parameters
• Named routes and reverse routing
• Restful controller routing
• Route filters and route groups

Credit to nikic/FastRoute.

While the bulk of the library and extensive unit tests are my own, credit for the regex matching core implementation and benchmarking goes to nikic. Please go and read nikic's blog post explaining how the implementation works and why it's fast.

Many modifications to the core have been made to suit the new library wrapper, and additional features added such as optional route parameters and reverse routing etc, but please head over and checkout nikic's library to see the origins of the core and how it works.

Installation

Install via composer

Usage

Defining routes

The routes are added by calling addRoute($method, $route, $handler) on the Phroute\RouteCollector collector instance.

  #NB. You can also call the HTTP method short cuts: 
 $routeCollector->get($route, $handler);    # match only get request methods
 $routeCollector->post($route, $handler);   # match only post request methods
 $routeCollector->delete($route, $handler); # match only delete request methods
 $routeCollector->any($route, $handler);    # match any request method
 etc...

This method accepts the HTTP method the route must match, the route pattern, an associated handler and an optional array of 'before' and 'after' filters. The handler does not necessarily have to be a callback (it could also be a controller class name and method or any other kind of data you wish to associate with the route).

By default a route pattern syntax is used where {foo} specified a placeholder with name foo and matching the string [^/]+. To adjust the pattern the placeholder matches, you can specify a custom pattern by writing {bar:[0-9]+}. However, it is also possible to adjust the pattern syntax by passing using a different route parser.

Regex Shortcuts

Named Routes for Reverse Routing

Pass in an array as the first argument, where the first item is your route and the second item is a name with which to reference it later.

Controllers

Dispatching a URI

A URI is dispatched by calling the dispatch() method of the created dispatcher. This method accepts the HTTP method and a URI. Getting those two bits of information (and normalizing them appropriately) is your job - this library is not bound to the PHP web SAPIs.

$response = (new Phroute\Dispatcher($router)) ->dispatch($_SERVER['REQUEST_METHOD'], $_SERVER['REQUEST_URI']);

The dispatch() method will call the matched route, or if no matches, throw one of the exceptions below:

# Route not found
Phroute\Exception\HttpRouteNotFoundException;

# Route found, but method not allowed
Phroute\Exception\HttpMethodNotAllowedException;

NOTE: The HTTP specification requires that a 405 Method Not Allowed response include the Allow: header to detail available methods for the requested resource. This information can be obtained from the thrown exception's message content: which will look like: "Allow: HEAD, GET, POST" etc... depending on the methods you have set You should catch the exception and use this to send a header to the client: header($e->getMessage());

A Note on HEAD Requests

The HTTP spec requires servers to [support both GET and HEAD methods][2616-511]:

The methods GET and HEAD MUST be supported by all general-purpose servers

To avoid forcing users to manually register HEAD routes for each resource we fallback to matching an available GET route for a given resource. The PHP web SAPI transparently removes the entity body from HEAD responses so this behavior has no effect on the vast majority of users.

However, implementors using Phroute outside the web SAPI environment (e.g. a custom server) MUST NOT send entity bodies generated in response to HEAD requests. If you are a non-SAPI user this is your responsibility; Phroute has no purview to prevent you from breaking HTTP in such cases.

Finally, note that applications MAY always specify their own HEAD method route for a given resource to bypass this behavior entirely.

Performance

Performed on a machine with :

• Processor 2.3 GHz Intel Core i7
• Memory 8 GB 1600 MHz DDR3

Phroute

This test is to illustrate, in part, the efficiency of the lightweight routing-core, but mostly the lack of degradation of matching speed as the number of routes grows, as compared to conventional libraries.

With 10 routes, matching 1st route (best case)

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:9943/

Finished 1000 requests

Time taken for tests:   3.062 seconds
Requests per second:    326.60 [#/sec] (mean)
Time per request:       306.181 [ms] (mean)
Time per request:       3.062 [ms] (mean, across all concurrent requests)
Transfer rate:          37.32 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%    306
  66%    307
  75%    307
  80%    308
  90%    309
  95%    309
  98%    310
  99%    310
 100%    310 (longest request)

With 10 routes, matching last route (worst case)

Note that the match is just as quick as against the first route

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:9943/thelastroute

Finished 1000 requests

Time taken for tests:   3.079 seconds
Requests per second:    324.80 [#/sec] (mean)
Time per request:       307.880 [ms] (mean)
Time per request:       3.079 [ms] (mean, across all concurrent requests)
Transfer rate:          37.11 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%    307
  66%    308
  75%    309
  80%    309
  90%    310
  95%    311
  98%    312
  99%    312
 100%    313 (longest request)

With 100 routes, matching last route (worst case)

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:9943/thelastroute

Finished 1000 requests

Time taken for tests:   3.195 seconds
Requests per second:    312.97 [#/sec] (mean)
Time per request:       319.515 [ms] (mean)
Time per request:       3.195 [ms] (mean, across all concurrent requests)
Transfer rate:          35.76 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%    318
  66%    319
  75%    320
  80%    320
  90%    322
  95%    323
  98%    323
  99%    324
 100%    324 (longest request)

With 1000 routes, matching the last route (worst case)

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:9943/thelastroute

Finished 1000 requests

Time taken for tests:   4.497 seconds
Complete requests:      1000
Requests per second:    222.39 [#/sec] (mean)
Time per request:       449.668 [ms] (mean)
Time per request:       4.497 [ms] (mean, across all concurrent requests)
Transfer rate:          25.41 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%    445
  66%    447
  75%    448
  80%    449
  90%    454
  95%    456
  98%    457
  99%    458
 100%    478 (longest request)

For comparison, Laravel 4.0 routing core

Please note, this is no slight against laravel - it is based on a routing loop, which is why the performance worsens as the number of routes grows

With 10 routes, matching first route (best case)

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:4968/

Finished 1000 requests

Time taken for tests:   13.366 seconds
Requests per second:    74.82 [#/sec] (mean)
Time per request:       1336.628 [ms] (mean)
Time per request:       13.366 [ms] (mean, across all concurrent requests)
Transfer rate:          8.55 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%   1336
  66%   1339
  75%   1340
  80%   1341
  90%   1346
  95%   1348
  98%   1349
  99%   1351
 100%   1353 (longest request)

With 10 routes, matching last route (worst case)

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:4968/thelastroute

Finished 1000 requests

Time taken for tests:   14.621 seconds
Requests per second:    68.39 [#/sec] (mean)
Time per request:       1462.117 [ms] (mean)
Time per request:       14.621 [ms] (mean, across all concurrent requests)
Transfer rate:          7.81 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%   1461
  66%   1465
  75%   1469
  80%   1472
  90%   1476
  95%   1479
  98%   1480
  99%   1482
 100%   1484 (longest request)

With 100 routes, matching last route (worst case)

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:4968/thelastroute

Finished 1000 requests

Time taken for tests:   31.254 seconds
Requests per second:    32.00 [#/sec] (mean)
Time per request:       3125.402 [ms] (mean)
Time per request:       31.254 [ms] (mean, across all concurrent requests)
Transfer rate:          3.66 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%   3124
  66%   3145
  75%   3154
  80%   3163
  90%   3188
  95%   3219
  98%   3232
  99%   3236
 100%   3241 (longest request)

With 1000 routes, matching last route (worst case)

$ /usr/local/bin/ab -n 1000 -c 100 http://127.0.0.1:5740/thelastroute

Finished 1000 requests

Time taken for tests:   197.366 seconds
Requests per second:    5.07 [#/sec] (mean)
Time per request:       19736.598 [ms] (mean)
Time per request:       197.366 [ms] (mean, across all concurrent requests)
Transfer rate:          0.58 [Kbytes/sec] received

Percentage of the requests served within a certain time (ms)
  50%  19736
  66%  19802
  75%  19827
  80%  19855
  90%  19898
  95%  19918
  98%  19945
  99%  19960
 100%  19975 (longest request)