rethinkdb-ws-server

This is a fork from mikemintz/rethinkdb-websocket-server

In the original repo there is no option to manipulate data/query in the backend. So I added an option to use session data in the query. This data DOES NOT need to be sent from the frontend client.

For the code to work you must create a session.

Install:

npm install --save rethinkdb-ws-client

Example:

Create a session that will populate the session with our special id(s)

Server:

// create a session
options.sessionCreator = function (urlQueryParams) {
  // this you get from frontend as query param
  let token = urlQueryParams.token;
  if (!token || token == "") {
    return Promise.reject('Invalid auth token');
  } else {
    // your custom session logic here
    // I use a [teade](https://www.npmjs.com/package/teade) 
    // RPC middleware to validate the token and get sessionId from MySQL
    // ... RPC Code Here
    return Promise.resolve(
        {
            sessionId: "string session id", //optional
            sessionIdInt: parseInt("integer session id") // optional
        }
    );
  }
};

add a query to whitelist so that is passes validation

//
r.table('trades')
    .orderBy({index: allowSpecificFieldOrderBy(RP, "created_at")})
    .filter(r.row('type').eq(allowAnyString(RP)))
    .filter({"userId": allowAnyString(RP)})
    .changes({includeInitial: true}).opt("db", r.db("test"))
    .validate(function (refs, session) {
        return true;
    });

Client:

Compatible Client: [mikemintz/rethinkdb-websocket-client]https://github.com/mikemintz/rethinkdb-websocket-client now you can execute the following query from your client.

let query = r.table('trades').orderBy({index: r.desc("created_at")})
    .filter(r.row('currency').eq(currencyName))
    .filter({"userId": "$sessionId$"}) // can use "$sessionIdInt$" here
    .changes({includeInitial: true});

Whenever your query data will match $sessionId$ or $sessionIdInt$ it will be replaced with a respective key from the session.

Some helper functions I use.

// helper function used
function allowAnyString(RP){
    return RP.check(function(actual, refs, session) {
        return typeof actual === 'string' && actual.trim();
    })
}

function allowBool(RP){
    return RP.check(function(actual, refs, session) {
        return typeof actual === 'boolean';
    })
}

function allowSpecificFieldOrderBy(RP, allowedField){
    return RP.check(function(actual, refs, session) {
        return ((actual[0] == 73 || actual[0] == 74) && actual[1][0] == allowedField);
    }) 
}

Documentation from original library

mikemintz/rethinkdb-websocket-server

rethinkdb-websocket-server

Node.js WebSocket server that proxies to a RethinkDB instance. Supports query validation and minor manipulations.

What is this?

This library attaches to a node.js http.Server and listens for incoming WebSocket connections at a specified path. For each incoming WebSocket connection, it opens a new TCP socket to a specified RethinkDB instance, and forwards traffic in both directions between the WebSocket and the RethinkDB socket until either side disconnects.

Each query sent from the WebSocket is parsed and validated before being forwarded to RethinkDB. This is done using a whitelist of pattern queries, described in the "Involved example" section below.

The provided WebSocket server can be used in conjunction with any of the following clients:

• rethinkdb-websocket-client is a WebSocket wrapper around the RethinkDB JavaScript driver that works in the browser.
• react-rethinkdb is a React mixin that connects using rethinkdb-websocket-client.

How do I use this?

This package should be installed with npm and should run in a node.js http.Server.

Simple example

Below is a simple example of a script that listens at ws://localhost:8000/ for incoming WebSocket connections. It blindly forwards all queries to a RethinkDB instance running locally at the default 28015 port.

var http = require('http');
var wsListen = require('rethinkdb-websocket-server').listen;

var httpServer = http.createServer();
wsListen({httpServer: httpServer, unsafelyAllowAnyQuery: true});
httpServer.listen(8000);

Involved example

In this example, we listen for WebSocket connections at ws://localhost:8000/rethinkApi and forward traffic to a RethinkDB instance running at rethink01.example.com:28015. We also serve static files in the assets directory over HTTP using express.

Rather than enabling unsafelyAllowAnyQuery, we explicitly set up a query whitelist. This one only allows two query patterns:

1. Queries that list turtles with the same herdId as the authenticated user
2. Queries that insert turtles with a non-empty name and a herdId referring to the primary key of an object in the herds table

In order to validate queries against the authenticated user, we create a "session" object from the query params of the WebSocket URL. In this case, the browser connects to ws://localhost:8000/rethinkApi?userId=foo&authToken=bar, the sessionCreator function looks up that user in the database, and user.curHerdId is stored in the custom session object that we have access to when validating queries from this client.

Note: in production, you should enable secure websockets so sensitive data is not vulnerable.

As you are developing, incoming queries will be logged to console in a format that you can copy and paste directly into your JavaScript source file. For dynamic queries, you'll likely want to generalize the pattern using RP.check() terms, RP.ref() terms, and the .validate() method. Using ES6 arrow functions can make this a bit less verbose.

var express = require('express');
var http = require('http');
var Promise = require('bluebird');
var RethinkdbWebsocketServer = require('rethinkdb-websocket-server');
var r = RethinkdbWebsocketServer.r;
var RP = RethinkdbWebsocketServer.RP;

var options = {};
options.dbHost = 'rethink01.example.com';
options.dbPort = 28015;

var rethinkConn = Promise.promisify(r.connect)({
  host: options.dbHost,
  port: options.dbPort,
  db: 'test',
});
function runQuery(query) {
  return rethinkConn.then(function(conn) {
    return query.run(conn);
  });
}

options.sessionCreator = function(urlQueryParams) {
  var userQuery = r.table('users').get(urlQueryParams.userId);
  return runQuery(userQuery).then(function(user) {
    if (user && user.authToken === urlQueryParams.authToken) {
      return {curHerdId: user.herdId};
    } else {
      return Promise.reject('Invalid auth token');
    }
  });
};

options.queryWhitelist = [
  // r.table('turtles').filter({herdId: curHerdId})
  r.table('turtles')
   .filter({"herdId": RP.ref('herdId')})
   .opt("db", r.db("test"))
   .validate(function(refs, session) {
     return session.curHerdId === refs.herdId;
   }),

  // r.table('turtles').insert({herdId: 'alpha-squadron', name: 'Speedy'})
  r.table('turtles')
   .insert({
     "herdId": RP.ref('herdId'),
     "name": RP.check(function(actual, refs, session) {
       return typeof actual === 'string' && actual.trim();
     }),
   })
   .opt("db", r.db("test"))
   .validate(function(refs) {
     var herdId = refs.herdId;
     if (typeof herdId !== 'string') return false;
     var validHerdQuery = r.table('herds').get(herdId).ne(null);
     return runQuery(validHerdQuery);
   }),
];

var app = express();
app.use('/', express.static('assets'));
var httpServer = http.createServer(app);
options.httpServer = httpServer;
options.httpPath = '/rethinkApi';

RethinkdbWebsocketServer.listen(options);
httpServer.listen(8000);

Written a bit more concisely and with some ES6 syntax, the whitelist becomes:

options.queryWhitelist = [
  r.table('turtles')
   .filter({herdId: RP.ref('herdId')})
   .opt("db", r.db("test"))
   .validate(({herdId}, session) => session.curHerdId === herdId),

  r.table('turtles')
   .insert({
     herdId: RP.ref('herdId'),
     name: RP.check(x => typeof x === 'string' && x.trim())
   })
   .opt("db", r.db("test"))
   .validate(({herdId}) => (
     typeof herdId !== 'string' &&
     runQuery(r.table('herds').get(herdId).ne(null))
   )),
];

Upgrade guide

Most new versions of rethinkdb-websocket-server are backwards compatible with previous versions. Below are exceptions with breaking changes:

Upgrading to 0.3 or 0.4 (query whitelist syntax)

Before rethinkdb-websocket-server 0.3.0, the syntax for expressing whitelist queries closely reflected the RethinkDB JSON protocol sent over the wire, using the provided RQ object to construct query patterns. This older RQ syntax was deprecated in 0.4.0 and will ultimately be removed.

In the 0.3 and 0.4 releases, the whitelist can contain pattern queries from either syntax. This makes it easy to migrate queries to the new syntax one at a time.

RQ/vanilla syntax comparison

• Both RQ and vanilla syntax chain .validate(fn) after queries to add validation functions.
• Both RQ and vanilla syntax chain .opt(key, value) after queries to set query options like db and durability.
  • However, the value argument can differ in syntax: .opt("db", RQ.DB("test")) vs .opt("db", r.db("test").
• RQ.ref(refName) in the RQ syntax has been changed to RP.ref(refName)
  • RP stands for ReQL Pattern, and separating the RQ and RP object helps ensure you are using the right syntax version.
• Pattern functions like function(actual, refs, session) {...} in the RQ syntax must now be wrapped in RP.check(function(actual, refs, session) {...}.
  • This is necessary because JavaScript functions would otherwise be ambiguous in the ReQL AST. I.e. r.filter(function(x) {...}) should only be able to specify a filter function, not a whitelist pattern function.
• The biggest difference: the underlying syntax is completely different.
  • In the RQ syntax, the expressions represent the underlying JSON protocol, whereas the vanilla syntax is the same as writing ReQL with the JavaScript driver. The following queries are equivalent:
  • RQ(RQ.FILTER(RQ.TABLE("turtles"), {"herdId": RQ.ref('herdId')}))
  • r.table("turtles").filter({"herdId": RP.ref('herdId')})

Old RQ syntax example

Below is the query whitelist excerpt of the "Involved example" section above, as it used to be written using the older RQ syntax:

var RQ = RethinkdbWebsocketServer.RQ;

options.queryWhitelist = [
  RQ(
    RQ.FILTER(
      RQ.TABLE("turtles"),
      {"herdId": RQ.ref('herdId')}
    )
  ).opt("db", RQ.DB("test"))
  .validate(function(refs, session) {
    return session.curHerdId === refs.herdId;
  }),

  RQ(
    RQ.INSERT(
      RQ.TABLE("turtles"),
      {
        "herdId": RQ.ref('herdId'),
        "name": function(actual, refs, session) {
          return typeof actual === 'string' && actual.trim();
        },
      }
    )
  ).opt("db", RQ.DB("test"))
  .validate(function(refs) {
    var herdId = refs.herdId;
    if (typeof herdId !== 'string') return false;
    var validHerdQuery = r.table('herds').get(herdId).ne(null);
    return runQuery(validHerdQuery);
  }),
];