routing

ozzo-routing

You may consider using golang-restful-starter-kit to jumpstart your new RESTful applications with ozzo-routing.

Description

ozzo-routing is a Go package that provides high performance and powerful HTTP routing capabilities for Web applications. It has the following features:

• middleware pipeline architecture, similar to that of the Express framework.
• extremely fast request routing with zero dynamic memory allocation (the performance is comparable to that of httprouter and gin, see the performance comparison below)
• modular code organization through route grouping
• flexible URL path matching, supporting URL parameters and regular expressions
• URL creation according to the predefined routes
• compatible with http.Handler and http.HandlerFunc
• ready-to-use handlers sufficient for building RESTful APIs

If you are using fasthttp, you may use a similar routing package fasthttp-routing which is adapted from ozzo-routing.

Requirements

Go 1.7 or above.

Installation

Run the following command to install the package:

go get github.com/go-ozzo/ozzo-routing

You may also get specified release of the package by:

go get gopkg.in/go-ozzo/ozzo-routing.v1

Getting Started

For a complete RESTful application boilerplate based on ozzo-routing, please refer to the golang-restful-starter-kit. Below we describe how to create a simple REST API using ozzo-routing.

Create a server.go file with the following content:

package main

import (
	"log"
	"net/http"
	"github.com/go-ozzo/ozzo-routing"
	"github.com/go-ozzo/ozzo-routing/access"
	"github.com/go-ozzo/ozzo-routing/slash"
	"github.com/go-ozzo/ozzo-routing/content"
	"github.com/go-ozzo/ozzo-routing/fault"
	"github.com/go-ozzo/ozzo-routing/file"
)

func main() {
	router := routing.New()

	router.Use(
		// all these handlers are shared by every route
		access.Logger(log.Printf),
		slash.Remover(http.StatusMovedPermanently),
		fault.Recovery(log.Printf),
	)

	// serve RESTful APIs
	api := router.Group("/api")
	api.Use(
		// these handlers are shared by the routes in the api group only
		content.TypeNegotiator(content.JSON, content.XML),
	)
	api.Get("/users", func(c *routing.Context) error {
		return c.Write("user list")
	})
	api.Post("/users", func(c *routing.Context) error {
		return c.Write("create a new user")
	})
	api.Put(`/users/<id:\d+>`, func(c *routing.Context) error {
		return c.Write("update user " + c.Param("id"))
	})

	// serve index file
	router.Get("/", file.Content("ui/index.html"))
	// serve files under the "ui" subdirectory
	router.Get("/*", file.Server(file.PathMap{
		"/": "/ui/",
	}))

	http.Handle("/", router)
	http.ListenAndServe(":8080", nil)
}

Create an HTML file ui/index.html with any content.

Now run the following command to start the Web server:

go run server.go

You should be able to access URLs such as http://localhost:8080, http://localhost:8080/api/users.

Routes

ozzo-routing works by building a routing table in a router and then dispatching HTTP requests to the matching handlers found in the routing table. An intuitive illustration of a routing table is as follows:

Routes	Handlers
GET /users	m1, m2, h1, ...
POST /users	m1, m2, h2, ...
PUT /users/<id>	m1, m2, h3, ...
DELETE /users/<id>	m1, m2, h4, ...

For an incoming request GET /users, the first route would match and the handlers m1, m2, and h1 would be executed. If the request is PUT /users/123, the third route would match and the corresponding handlers would be executed. Note that the token <id> can match any number of non-slash characters and the matching part can be accessed as a path parameter value in the handlers.

If an incoming request matches multiple routes in the table, the route added first to the table will take precedence. All other matching routes will be ignored.

The actual implementation of the routing table uses a variant of the radix tree data structure, which makes the routing process as fast as working with a hash table, thanks to the inspiration from httprouter.

To add a new route and its handlers to the routing table, call the To method like the following:

router := routing.New()
router.To("GET", "/users", m1, m2, h1)
router.To("POST", "/users", m1, m2, h2)

You can also use shortcut methods, such as Get, Post, Put, etc., which are named after the HTTP method names:

router.Get("/users", m1, m2, h1)
router.Post("/users", m1, m2, h2)

If you have multiple routes with the same URL path but different HTTP methods, like the above example, you can chain them together as follows,

router.Get("/users", m1, m2, h1).Post(m1, m2, h2)

If you want to use the same set of handlers to handle the same URL path but different HTTP methods, you can take the following shortcut:

router.To("GET,POST", "/users", m1, m2, h)

A route may contain parameter tokens which are in the format of <name:pattern>, where name stands for the parameter name, and pattern is a regular expression which the parameter value should match. A token <name> is equivalent to <name:[^/]*>, i.e., it matches any number of non-slash characters. At the end of a route, an asterisk character can be used to match any number of arbitrary characters. Below are some examples:

• /users/<username>: matches /users/admin
• /users/accnt-<id:\d+>: matches /users/accnt-123, but not /users/accnt-admin
• /users/<username>/*: matches /users/admin/profile/address

When a URL path matches a route, the matching parameters on the URL path can be accessed via Context.Param():

router := routing.New()

router.Get("/users/<username>", func (c *routing.Context) error {
	fmt.Fprintf(c.Response, "Name: %v", c.Param("username"))
	return nil
})

Route Groups

Route group is a way of grouping together the routes which have the same route prefix. The routes in a group also share the same handlers that are registered with the group via its Use method. For example,

router := routing.New()
api := router.Group("/api")
api.Use(m1, m2)
api.Get("/users", h1).Post(h2)
api.Put("/users/<id>", h3).Delete(h4)

The above /api route group establishes the following routing table:

Routes	Handlers
GET /api/users	m1, m2, h1, ...
POST /api/users	m1, m2, h2, ...
PUT /api/users/<id>	m1, m2, h3, ...
DELETE /api/users/<id>	m1, m2, h4, ...

As you can see, all these routes have the same route prefix /api and the handlers m1 and m2. In other similar routing frameworks, the handlers registered with a route group are also called middlewares.

Route groups can be nested. That is, a route group can create a child group by calling the Group() method. The router serves as the top level route group. A child group inherits the handlers registered with its parent group. For example,

router := routing.New()
router.Use(m1)

api := router.Group("/api")
api.Use(m2)

users := api.Group("/users")
users.Use(m3)
users.Put("/<id>", h1)

Because the router serves as the parent of the api group which is the parent of the users group, the PUT /api/users/<id> route is associated with the handlers m1, m2, m3, and h1.

Router

Router manages the routing table and dispatches incoming requests to appropriate handlers. A router instance is created by calling the routing.New() method.

Because Router implements the http.Handler interface, it can be readily used to serve subtrees on existing Go servers. For example,

router := routing.New()
http.Handle("/", router)
http.ListenAndServe(":8080", nil)

Handlers

A handler is a function with the signature func(*routing.Context) error. A handler is executed by the router if the incoming request URL path matches the route that the handler is associated with. Through the routing.Context parameter, you can access the request information in handlers.

A route may be associated with multiple handlers. These handlers will be executed in the order that they are registered to the route. The execution sequence can be terminated in the middle using one of the following two methods:

• A handler returns an error: the router will skip the rest of the handlers and handle the returned error.
• A handler calls Context.Abort(): the router will simply skip the rest of the handlers. There is no error to be handled.

A handler can call Context.Next() to explicitly execute the rest of the unexecuted handlers and take actions after they finish execution. For example, a response compression handler may start the output buffer, call Context.Next(), and then compress and send the output to response.

Context

For each incoming request, a routing.Context object is populated with the request information and passed through the handlers that need to handle the request. Handlers can get the request information via Context.Request and send a response back via Context.Response. The Context.Param() method allows handlers to access the URL path parameters that match the current route.

Using Context.Get() and Context.Set(), handlers can share data between each other. For example, an authentication handler can store the authenticated user identity by calling Context.Set(), and other handlers can retrieve back the identity information by calling Context.Get().

Reading Request Data

Context provides a few shortcut methods to read query parameters. The Context.Query() method returns the named URL query parameter value; the Context.PostForm() method returns the named parameter value in the POST or PUT body parameters; and the Context.Form() method returns the value from either POST/PUT or URL query parameters.

The Context.Read() method supports reading data from the request body and populating it into an object. The method will check the Content-Type HTTP header and parse the body data as the corresponding format. For example, if Content-Type is application/json, the request body will be parsed as JSON data. The public fields in the object being populated will receive the parsed data if the data contains the same named fields. For example,

func foo(c *routing.Context) error {
    data := &struct{
        A string
        B bool
    }{}

    // assume the body data is: {"A":"abc", "B":true}
    // data will be populated as: {A: "abc", B: true}
    if err := c.Read(&data); err != nil {
        return err
    }
}

By default, Context supports reading data that are in JSON, XML, form, and multipart-form data. You may modify routing.DataReaders to add support for other data formats.

Note that when the data is read as form data, you may use struct tag named form to customize the name of the corresponding field in the form data. The form data reader also supports populating data into embedded objects which are either named or anonymous.

Writing Response Data

The Context.Write() method can be used to write data of arbitrary type to the response. By default, if the data being written is neither a string nor a byte array, the method will will call fmt.Fprint() to write the data into the response.

You can call Context.SetWriter() to replace the default data writer with a customized one. For example, the content.TypeNegotiator will negotiate the content response type and set the data writer with an appropriate one.

Error Handling

A handler may return an error indicating some erroneous condition. Sometimes, a handler or the code it calls may cause a panic. Both should be handled properly to ensure best user experience. It is recommended that you use the fault.Recover handler or a similar error handler to handle these errors.

If an error is not handled by any handler, the router will handle it by calling its handleError() method which simply sets an appropriate HTTP status code and writes the error message to the response.

When an incoming request has no matching route, the router will call the handlers registered via the Router.NotFound() method. All the handlers registered via Router.Use() will also be called in advance. By default, the following two handlers are registered with Router.NotFound():

• routing.MethodNotAllowedHandler: a handler that sends an Allow HTTP header indicating the allowed HTTP methods for a requested URL
• routing.NotFoundHandler: a handler triggering 404 HTTP error

Serving Static Files

Static files can be served with the help of file.Server and file.Content handlers. The former serves files under the specified directories, while the latter serves the content of a single file. For example,

import (
	"github.com/go-ozzo/ozzo-routing"
	"github.com/go-ozzo/ozzo-routing/file"
)

router := routing.NewRouter()

// serve index file
router.Get("/", file.Content("ui/index.html"))
// serve files under the "ui" subdirectory
router.Get("/*", file.Server(file.PathMap{
	"/": "/ui/",
}))

Handlers

ozzo-routing comes with a few commonly used handlers in its subpackages:

Handler name	Description
access.Logger	records an entry for every incoming request
auth.Basic	provides authentication via HTTP Basic
auth.Bearer	provides authentication via HTTP Bearer
auth.Query	provides authentication via token-based query parameter
auth.JWT	provides JWT-based authentication
content.TypeNegotiator	supports content negotiation by response types
content.LanguageNegotiator	supports content negotiation by accepted languages
cors.Handler	implements the CORS (Cross Origin Resource Sharing) specification from the W3C
fault.Recovery	recovers from panics and handles errors returned by handlers
fault.PanicHandler	recovers from panics happened in the handlers
fault.ErrorHandler	handles errors returned by handlers by writing them in an appropriate format to the response
file.Server	serves the files under the specified folder as response content
file.Content	serves the content of the specified file as the response
slash.Remover	removes the trailing slashes from the request URL and redirects to the proper URL

The following code shows how these handlers may be used:

import (
	"log"
	"net/http"
	"github.com/go-ozzo/ozzo-routing"
	"github.com/go-ozzo/ozzo-routing/access"
	"github.com/go-ozzo/ozzo-routing/slash"
	"github.com/go-ozzo/ozzo-routing/fault"
)

router := routing.New()

router.Use(
	access.Logger(log.Printf),
	slash.Remover(http.StatusMovedPermanently),
	fault.Recovery(log.Printf),
)

...

Third-party Handlers

The following third-party handlers are specifically designed for ozzo-routing:

Handler name	Description
jwt.JWT	supports JWT Authorization

ozzo-routing also provides adapters to support using third-party http.HandlerFunc or http.Handler handlers. For example,

router := routing.New()

// using http.HandlerFunc
router.Use(routing.HTTPHandlerFunc(http.NotFound))

// using http.Handler
router.Use(routing.HTTPHandler(http.NotFoundHandler))

3rd-Party Extensions and Code Examples

• Simple Standard Service Endpoints (SE4)
• ozzo examples

Benchmarks

Last updated on Jan 6, 2017

Ozzo-routing is very fast, thanks to the radix tree data structure and the usage of sync.Pool (the idea was originally from HttpRouter and Gin). The following table (by running go-http-routing-benchmark) shows how ozzo-routing compares with Gin, HttpRouter, and Martini in performance.

BenchmarkOzzo_GithubAll                    50000             37989 ns/op               0 B/op          0 allocs/op
BenchmarkEcho_GithubAll                    20000             91003 ns/op            6496 B/op        203 allocs/op
BenchmarkGin_GithubAll                     50000             26717 ns/op               0 B/op          0 allocs/op
BenchmarkHttpRouter_GithubAll              50000             36052 ns/op           13792 B/op        167 allocs/op
BenchmarkMartini_GithubAll                   300           4162283 ns/op          228216 B/op       2483 allocs/op

BenchmarkOzzo_GPlusAll                   1000000              1732 ns/op               0 B/op          0 allocs/op
BenchmarkEcho_GPlusAll                    300000              4523 ns/op             416 B/op         13 allocs/op
BenchmarkGin_GPlusAll                    1000000              1171 ns/op               0 B/op          0 allocs/op
BenchmarkHttpRouter_GPlusAll             1000000              1533 ns/op             640 B/op         11 allocs/op
BenchmarkMartini_GPlusAll                  20000             75634 ns/op           14448 B/op        165 allocs/op

BenchmarkOzzo_ParseAll                    500000              3318 ns/op               0 B/op          0 allocs/op
BenchmarkEcho_ParseAll                    200000              7336 ns/op             832 B/op         26 allocs/op
BenchmarkGin_ParseAll                    1000000              2075 ns/op               0 B/op          0 allocs/op
BenchmarkHttpRouter_ParseAll             1000000              2034 ns/op             640 B/op         16 allocs/op
BenchmarkMartini_ParseAll                  10000            122002 ns/op           25600 B/op        276 allocs/op

Credits

ozzo-routing has referenced many popular routing frameworks, including Express, Martini, httprouter, and gin.