jonasi / httprouter

A high performance HTTP request router that scales well

HttpRouter

HttpRouter is a lightweight high performance HTTP request router (also called multiplexer or just mux for short) for Go.

In contrast to the default mux of Go's net/http package, this router supports variables in the routing pattern and matches against the request method. It also scales better.

The router is optimized for best performance and a small memory footprint. It scales well even with very long paths and a large number of routes. A compressing dynamic trie (radix tree) structure is used for efficient matching.

Features

Zero Garbage: The matching and dispatching process generates zero bytes of garbage. In fact, the only heap allocations that are made, is by building the slice of the key-value pairs for path parameters. If the request path contains no parameters, not a single heap allocation is necessary.

Best Performance: Benchmarks speak for themselves. See below for technical details of the implementation.

Parameters in your routing pattern: Stop parsing the requested URL path, just give the path segment a name and the router delivers the dynamic value to you. Because of the design of the router, path parameters are very cheap.

Only explicit matches: With other routers, like http.ServeMux, a requested URL path could match multiple patterns. Therefore they have some awkward pattern priority rules, like longest match or first registered, first matched. By design of this router, a request can only match exactly one or no route. As a result, there are also no unintended matches, which makes it great for SEO and improves the user experience.

Stop caring about trailing slashes: Choose the URL style you like, the router automatically redirects the client if a trailing slash is missing or if there is one extra. Of course it only does so, if the new path has a handler. If you don't like it, you can turn off this behavior.

Path auto-correction: Besides detecting the missing or additional trailing slash at no extra cost, the router can also fix wrong cases and remove superfluous path elements (like ../ or //). Is CAPTAIN CAPS LOCK one of your users? HttpRouter can help him by making a case-insensitive look-up and redirecting him to the correct URL.

No more server crashes: You can set a PanicHandler to deal with panics occurring during handling a HTTP request. The router then recovers and lets the PanicHandler log what happened and deliver a nice error page.

Of course you can also set a custom NotFound handler and serve static files.

Usage

This is just a quick introduction, view the GoDoc for details.

Let's start with a trivial example:

package main

import (
    "fmt"
    "github.com/julienschmidt/httprouter"
    "net/http"
    "log"
)

func Index(w http.ResponseWriter, r *http.Request, _ httprouter.Params) {
    fmt.Fprint(w, "Welcome!\n")
}

func Hello(w http.ResponseWriter, r *http.Request, ps httprouter.Params) {
    fmt.Fprintf(w, "hello, %s!\n", ps.ByName("name"))
}

func main() {
    router := httprouter.New()
    router.GET("/", Index)
    router.GET("/hello/:name", Hello)

    log.Fatal(http.ListenAndServe(":8080", router))
}

Named parameters

As you can see, :name is a named parameter. The values are accessible via httprouter.Params, which is just a slice of httprouter.Params. You can get the value of a parameter either by its index in the slice, or by using the ByName(name) method: :name can be retrived by ByName("name").

Named parameters only match a single path segment:

Pattern: /user/:user

 /user/gordon              match
 /user/you                 match
 /user/gordon/profile      no match
 /user/                    no match

Note: Since this router has only explicit matches, you can not register static routes and parameters for the same path segment. For example you can not register the patterns /user/new and /user/:user for the same request method at the same time. The routing of different request methods is independent from each other.

Catch-All parameters

The second type are catch-all parameters and have the form *name. Like the name suggests, they match everything. Therefore they must always be at the end of the pattern:

Pattern: /src/*filepath

 /src/                     match
 /src/somefile.go          match
 /src/subdir/somefile.go   match

How does it work?

The router relies on a tree structure which makes heavy use of common prefixes, it is basically a compact prefix tree (or just Radix tree). Nodes with a common prefix also share a common parent. Here is a short example what the routing tree for the GET request method could look like:

Priority   Path             Handle
9          \                *<1>
3          ├s               nil
2          |├earch\         *<2>
1          |└upport\        *<3>
2          ├blog\           *<4>
1          |    └:post      nil
1          |         └\     *<5>
2          ├about-us\       *<6>
1          |        └team\  *<7>
1          └contact\        *<8>

Every *<num> represents the memory address of a handler function (a pointer). If you follow a path trough the tree from the root to the leaf, you get the complete route path, e.g \blog\:post\, where :post is just a placeholder (parameter) for an actual post name. Unlike hash-maps, a tree structure also allows us to use dynamic parts like the :post parameter, since we actually match against the routing patterns instead of just comparing hashes. As benchmarks show, this works very well and efficient.

Since URL paths have a hierarchical structure and make use only of a limited set of characters (byte values), it is very likely that there are a lot of common prefixes. This allows us to easily reduce the routing into ever smaller problems. Moreover the router manages a separate tree for every request method. For one thing it is more space efficient than holding a method->handle map in every single node, for another thing is also allows us to greatly reduce the routing problem before even starting the look-up in the prefix-tree.

For even better scalability, the child nodes on each tree level are ordered by priority, where the priority is just the number of handles registered in sub nodes (children, grandchildren, and so on..). This helps in two ways:

1. Nodes which are part of the most routing paths are evaluated first. This helps to make as much routes as possible to be reachable as fast as possible.
2. It is some sort of cost compensation. The longest reachable path (highest cost) can always be evaluated first. The following scheme visualizes the tree structure. Nodes are evaluated from top to bottom and from left to right.

├------------
├---------
├-----
├----
├--
├--
└-

Why doesn't this work with http.Handler?

It does! The router itself implements the http.Handler interface. Moreover the router provides convenient adapters for http.Handlers and http.HandlerFuncs which allows them to be used as a httprouter.Handle when registering a route. The only disadvantage is, that no parameter values can be retrieved when a http.Handler or http.HandlerFunc is used, since there is no efficient way to pass the values with the existing function parameters. Therefore httprouter.Handle has a third function parameter.

Just try it out for yourself, the usage of HttpRouter is very straightforward. The package is compact and minimalistic, but also probably one of the easiest routers to set up.

Where can I find Middleware X?

This package just provides a very efficient request router with a few extra features. The router is just a http.Handler, you can chain any http.Handler compatible middleware before the router, for example the Gorilla handlers. Or you could just write your own, it's very easy!

Alternatively, you could try a framework building upon HttpRouter.

Here is a quick example: Does your server serve multiple domains / hosts? You want to use sub-domains? Define a router per host!

// We need an object that implements the http.Handler interface.
// Therefore we need a type for which we implement the ServeHTTP method.
// We just use a map here, in which we map host names (with port) to http.Handlers
type HostSwitch map[string]http.Handler

// Implement the ServerHTTP method on our new type
func (hs HostSwitch) ServeHTTP(w http.ResponseWriter, r *http.Request) {
	// Check if a http.Handler is registered for the given host.
	// If yes, use it to handle the request.
	if handler := hs[r.Host]; handler != nil {
		handler.ServeHTTP(w, r)
	} else {
		// Handle host names for wich no handler is registered
		http.Error(w, "Forbidden", 403) // Or Redirect?
	}
}

func main() {
	// Initialize a router as usual
	router := httprouter.New()
	router.GET("/", Index)
	router.GET("/hello/:name", Hello)

	// Make a new HostSwitch and insert the router (our http handler)
	// for example.com and port 12345
	hs := make(HostSwitch)
	hs["example.com:12345"] = router

	// Use the HostSwitch to listen and serve on port 12345
	log.Fatal(http.ListenAndServe(":12345", hs))
}

Web Frameworks building upon HttpRouter

If the HttpRouter is a bit too minimalistic for you, you might try one of the following more high-level 3rd-party web frameworks building upon the HttpRouter package:

• Gin: Features a martini-like API with much better performance
• Hikaru: Supports standalone and Google AppEngine