Skip to content

Matt-Esch/http-hash

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

http-hash

build status Coverage Status Davis Dependency status

NPM

HTTP router based on a strict path tree structure

Example 1 : Basic routes

var HttpHash = require('http-hash');

// Create a new http hash
var hash = HttpHash();

// Create a route mapping to /test/<anything but "/">
hash.set('/test/:foo/', function (req, res) {
    res.end();
});

// Get a valid route
var route = hash.get('/test/var');
console.log(route);
/*
-> {
    handler: function (req, res) {},
    params: {
        foo: 'var'
    },
    splat: null
}
*/

// Get an invalid route (returns null)
var missing = hash.get('/missing');
console.log(missing);
/*
-> {
    handler: null,
    params: {},
    splat: null
}
*/

Example 2 : Trailing splats

var HttpHash = require('http-hash');

// Create a new http hash
var hash = HttpHash();

// Create a route mapping to /foo/<anything but "/">/<anything>
hash.set('/foo/:test/*', function (req, res) {
    res.end();
});

var route = hash.get('/foo/val/one/two/three');
console.log(route);
/*
-> {
    handler: function (req, res) { ... },
    params: {
        test: 'val'
    },
    splat: 'one/two/three'
}
*/

Overview

The most popular node routers are based on regular expression matching. This means that the order in which the routes are defined affects the resolution of a route to handler. Sometimes this is desirable, but it would often be better to have a resolution scheme that is easier to reason about.

http-hash solves the routing problem by making route resolution independent of the order in which routes are defined. It does so by breaking a path into a tree of nodes, based on a simple split on /. For example, the route /foo/bar/baz is treated as tree nodes foo > bar > baz. We call foo, bar and baz path segments.

Theses path segments are arranged into a tree of nodes, where each segment defines a node in the tree. Each node can point to:

  • a fixed handler node.handler, otherwise known as the node value

  • a set of static paths indexed by path name node.staticPaths

  • a variable subtree node.variablePaths, that can match a single named parameter OR the remainder of a route (splat).

If the last character of a defined route is *, a variable path (or splat) will be inserted, consuming the rest of the path. This allows for subrouting, i.e. if you want to mount a static filesystem on /fs you would set the path as /fs/* where the nodes are broken down into the tree fs > *. The remainder of the route will be returned as a "splat" value, allowing for further routing.

In the simple case, the route tree is based on exact matches on the name of the segment. That is to say, for the case where we want to match /foo/bar/baz, the tree looks like

{
  staticPaths: {
    foo: {
      staticPaths: {
        bar: {
          staticPaths: {
            baz: {
              handler: function (req, res) {}
            }
          }
        }
      }
    }
  }
}

When defining routes, variable paths may be specified. This is where path segments are prefixed with : i.e. /foo/:bar/baz.

For the :bar segment, the route consumes the single variable route slot for that node in the tree. So for example, the route `/foo/:bar/baz looks like

{
  staticPaths: {
    foo: {
      variablePaths: {
        staticPaths: {
          'baz': function (req, res) {}
        }
      }
    }
  }
}

Since a node can have both static and dynamic paths associated with it, the static path will win over the variable path when we resolve the path.

Trailing slashes

In most cases the trailing / does not matter. Variables cannot be the empty string, and neither can splats. A splat value will not contain the leading slash as it is consumed by the parent node.

Path conflicts

If a path conflict occurs, an exception will be thrown. Conflicts occur when:

  • A route is defined twice, resolving to two handlers

This is the simplest case where /foo has been defined twice.

  • Variable names in the path are different

Note that /foo/:vara/ and /foo/:varb/ conflict, since they both resolve to foo.variablePaths, but have different param names.

  • A variable route is defined for a splat node

In the case of splats being defined at a level, no other other variables may be specified, as we cannot distinguish between /foo/:var and /foo/*. It is however ok to put static paths on the same level, i.e. /foo/bar and /foo/*. In this case, the static paths will be tried first before yielding the splat.

Docs

var hash = HttpHash()

http-hash := () => HttpHash

type HttpHash := {
    get: (String: url) => RouteResult,
    set: (String: path, Any: handler) => void,
    _hash: RouteNode
}

type RouteNode := {
    handler: Any,
    fixedPaths: Object<String, RouteNode>,
    variablePaths: RouteNode | null
}

type RouteResult := {
    handler: Any | null,
    params: Object<String, String>,
    splat: String | null
}

http-hash exports a safe constructor function that when called returns a new HttpHash. get and set methods are exposed for public consumption and the underlying data structure _hash is exposed for private inspection/internal use.

hash.set(path, handler)

hash.set := (String: path, Any: handler) => void

Puts a path in the route table. If the path conflicts with an existing path, an exception will be thrown.

Routes containing a * that are not part of a /* prefix will also throw an exception.

A path should look like / or /foo or /:foo or a union of theses things, or optionally end with /*

  • param names should not be repeated as they will conflict but there is no strong assertion for this. The last param name wins.

  • specifying a variable twice for a node will cause an exception

  • repeated and trailing '/' are ignored

  • paths are case sensitive

  • variables and splats are not matched by the empty string.

hash.get(path)

hash.get := (String: path) => RouteResult

Gets a route from the route table. If there is no viable route, the handler will be returned as null in the RouteResult object.

The route result contains a params hash, containing a key for each named variable in the path. Additionally, if a splat route was defined, the splat property will contain the tail portion of the route matched.

Installation

npm install http-hash

Tests

npm test

Contributors

  • Matt Esch

MIT Licensed

About

HTTP router based on a strict path tree structure

Resources

License

Stars

Watchers

Forks

Packages

No packages published