node package manager


Routing web requests based on the request's URL

Woute: A Web Request Router
Build Status Dependency Status NPM version Stories in Ready

The most natural way to route web requests is to use matching rules, not unlike Sinatra in Ruby and Express in JavaScript.

However, in NoFlo, it's a network of blackboxes you connect to make a program. The intuitive way is to connect incoming requests to a series of matchers; failure to match forwards to the next matcher.

When successful matches occur, the request is sent to wherever the programmer wants, which hopefully produces some result to be sent to a responder. An abstract example would be:

                       /login                /get_images
                         +                      +
                         |                      |
                         |                      |
+-----------+        +---v-------+         +----v------+           +------------+
|           |        |           |         |           |           |            |
|           |  Req   |  First    |  Fail   |  Second   |  Fail     |  Third     |
| Webserver +-------->  Matcher  +--------->  Matcher  +----------->  Matcher   |
|           |        |           |         |           |           |            |
+-----------+        +---+-------+         +----+------+           +---+--------+
                         |                      |                      |
                         |Success               |Success               |Success
                         |                      |                      |
                     +---v-------+         +----v------+           +---v--------+
                     |           |         |           |           |            |
                     |           |         |           |           |            |
                     |           |         | Fetch     |           |            |
                     | Login     |         | Images    |           | 404        |
                     |           |         |           |           |            |
                     +-----+-----+         +----+------+           +---+--------+
                           |                    |                      |
                           |                    |Res                   |
                           |Res                 |                      |Res
                           |               +----v------+               |
                           |               |           |               |
                           +--------------->           <---------------+
                                           | Respond   |
                                           |           |


npm install --save noflo-woute

Quick & Dirty Usage

To use noflo-woute in its most basic form, you only need the Matcher component:

  • Inport MATCH: optional takes a URL segment to match. Default to always match
  • Inport METHOD: optional an HTTP method. Default to any method
  • Inport IN: takes a request/response pair produced by noflo-webserver
  • Outport OUT: the request/repsonse pair if match is successful
  • Outport FAIL: the request/response pair if match is unsuccessful, most likely attached to another matcher

Simply connect some matchers together like the abstract example shown above and you're good to go!

More Advanced Usage: Adapters

Matchers are agnostic to the actual request/response, meaning that whoever handling a successfully matched case is handed the same thing that they would get from noflo-webserver. Two pairs of adapter components are there to help you to "split" the request into different parts so manipulation is easier: woute/ToGroups, woute/ToPorts, and their other halves woute/FromGroups and woute/FromPorts.

Both adapter pairs break the request/response object into these parts:

  • url: ditto
  • headers: the HTTP headers broken down into pairs (as an object of headers)
  • query: the query string broken down into pairs (as an object of queries)
  • body: the body passed through as-is (i.e. always a string)
  • request: the request/response object

ToGroups converts the incoming request/response object into the listed parts grouped by the names. Groups are constructed and sent in the order of the list above. ToPorts converts the object into the parts via ports by those names.

For instance, out comes from port ToGroups within a single connection:

DATA: /login
DATA: { "x-http-destination": "NoFlo Awesomeness" }
DATA: { this: "is sent", as: "an object" }
DATA: {"this":[{"is":"JSON","but":"is"},{"still":"sent"}],"as":"a string"}
DATA: <The request/response object>

For ToPorts, the same data packets would be sent to ports URL, HEADERS, QUERY, BODY, and REQUEST, respectively. Both ToGroups and ToPorts retain all the groups emitted from noflo-webserver.

Preparing for response

When you are done and are ready to send back a response, remember to feed your content to these two components' significant others: woute/FromGroups and woute/FromPorts. These two components take the disassembled data packets, apply them on a response object, and splice the connection back into request/repsonse pair so it's ready to be sent to webserver/SendResponse.

The components take these types of data packets:

  • status: the status code to set
  • headers: the response headers to be sent back
  • body: the body to be sent back
  • request: the request/response object

Which way is best?

ToPorts makes the HTTP request much more manageable as it breaks down the main parts of an HTTP request into separate connections. However, there is a down side: you need to be careful when asynchronous operation is involved.

Asynchronous operation removes a connection on a port from another, rendering FromPorts unable to splice them back into a request/response object. Yes, noflo-webserver does wrap the request around with a unique UUID for identification of a single HTTP request, but noflo-woute ignores that (but still retains it as a group). It is the responsibility of the programmer or a different package built on top of noflo-woute to handle asynchronicity.

Also, ToPorts triggers the re-assembly of the request/response object as soon as it receives a request/response object because it would otherwise not know when to apply the content of the packets on the response object. This means that anything that is sent after the request/response object is received is ignored.

ToGroups on the other hand puts everything within the same connection so there isn't any synchronicity problem as you pass the entire object around all at once. However, every time you need to find what you want and manipulate it, you need to weed through all the groups. Both approaches have pros and cons, hence the options.


  • In order for noflo-woute to recognize the BODY part, you need to run the request through webserver/BodyParser
  • FromPorts and FromGroups both expect the incoming body to be a string. You must JSONify or perform any conversion before feeding the body to the two components.
  • Remember to apply any desired middleware before passing the request/response object to any matcher.


Echo Server

Run the server:

> cd examples/echo_server
> npm install
> npm run-script main

Get back exactly what you pass in:

> curl "http://localhost:1337/echo" -d '{"a":"b"}' -H "Content-Type: application/json"

Get back the string 'empty-body':

> curl "http://localhost:1337/empty-body" -d '{"a":"b"}' -H "Content-Type: application/json"

Print to console of incoming body, but get nothing in return:

> curl "http://localhost:1337/anything-here" -X POST -d '{"a":"b"}' -H "Content-Type: application/json"

File not found:

> curl "http://localhost:1337/abc" -I
HTTP/1.1 404 Not Found
Date: Sat, 10 Aug 2013 08:35:04 GMT
Connection: keep-alive

To look at how these work, read the FBP code.