HTTPPP

HyperText Transfer Protocol Partial Parser

Overview

HTTPPP lets you parse only the bits of HTTP that you really need. It is designed specifically to aid in the creation of proxies that need to be fast and support arbitrary HTTP-like protocols. It doesn't mess with the actual data being pushed through the connection, so it trivially supports websockets and other extensions that rely on upgrade functionality or similar.

Installation

Available via npm:

$ npm install httppp

Or via git:

$ git clone git://github.com/deoxxa/httppp.git node_modules/httppp

API

httppp

The actual module.exports value (the thing you get when you require("httppp") is a factory function for creating parsers.

httppp(onHeaders);

var parser = httppp(functon onHeaders(info) { console.log(info); });

Arguments

• onHeaders - this is a function that, if supplied, is attached to #headers on the new Parser object. See below for information on the headers event.

Parser constructor

Constructs a new httppp Parser object, optionally supplying some configuration information as an object.

new httppp.Parser(options);

// basic instantiation
var parser = new httppp.Parser();
 
// instantiation with options
var parser = new httppp.Parser({maximumHeaderBytes: 4096});

Arguments

• options - an object specifying configuration parameters. The only available parameter right now is maximumHeaderBytes, which controls how many bytes the parser will try to read before it gives up and emits an error saying that the headers were too long.

#headers

The "headers" event is emitted when httppp has decided that it's parsed all the headers that are going to arrive. Note that this is only emitted once per connection, with the implication of that being that you won't know about pipelined requests.

The payload for the event is an object with properties of method, path, and headers. method and path are both strings, and headers is an object, where the keys are the header names, and the values are arrays containing the values collected for that header. The values are arrays because multiple headers with the same name may be sent (for example cookies).

NOTE: the payload for this event used to be an array, and to maintain compatibility with code that's still expecting that, the properties of the payload are aliased as method -> 0, path -> 1 and headers -> 2.

parser.on("headers", onHeaders);

parser.on("headers", function onHeaders(info) {
  // "GET" or similar
  console.log(info.method);
 
  // "/" or something
  console.log(info.path);
 
  // {host: ["127.0.0.1"], cookie: ["a=b", "c=d"]}
  console.log(info.headers);
});

Example

You might want to look at example.js as well.

var net = require("net"),
    http = require("http"),
    httppp = require("httppp");
 
var server1_port = null,
    server2_port = null;
 
var proxy = net.createServer(function(socket) {
  var parser = httppp(function(info) {
    console.log(new Date(), "proxy headers", info.method, info.path);
 
    var host = (info.headers.host && info.headers.host.length) ? info.headers.host[0] : null;
 
    // remove port from host header
    if (host) {
      host = host.split(":").shift();
    }
 
    switch (host) {
      case "localhost": parser.pipe(net.connect({port: server1_port})).pipe(socket); break;
      case "127.0.0.1": parser.pipe(net.connect({port: server2_port})).pipe(socket); break;
      default: socket.end(); break;
    }
  });
 
  parser.on("error", function() {
    socket.end();
  });
 
  socket.pipe(parser);
});
 
proxy.listen(3000, function() {
  console.log("listening on port", this.address().port);
});
 
var server1 = http.createServer(function(req, res) {
  console.log(new Date(), "http[1] request", req.url);
 
  res.end("hello there from server 1!");
});
 
server1.listen(function() {
  server1_port = this.address().port;
});
 
var server2 = http.createServer(function(req, res) {
  console.log(new Date(), "http[2] request", req.url);
 
  res.end("hello there from server 2!");
});
 
server2.listen(function() {
  server2_port = this.address().port;
});

License

3-clause BSD. A copy is included with the source.

Contact

• GitHub (deoxxa)
• Twitter (@deoxxa)
• ADN (@deoxxa)
• Email (deoxxa@fknsrs.biz)