This library is an implementation of both the JSON-RPC and the XML-RPC specification (server-side) for the Tornado web framework. It supports the basic features of both, as well as the MultiCall / Batch support for both specifications. The JSON-RPC handler supports both the original 1.0 specification, as well as the new (proposed) 2.0 spec, which includes batch submission, keyword arguments, etc.

Asynchronous request support has been added for methods which require the use of asynchronous libraries (like Tornado's AsyncHTTPClient library.)

TornadoRPC is licensed under the Apache License, Version 2.0 (http://www.apache.org/licenses/LICENSE-2.0.html).

If you have any questions, issues, or just use the library please feel free to send a message to the mailing list at:

http://groups.google.com/group/tornadorpc

To install:

python setup.py build
sudo python setup.py install

To use this module, you'll need Tornado installed, which you can get at this address:

http://www.tornadoweb.org/

If you want to use the JSON-RPC handler, you'll also need jsonrpclib, which you can grab at:

http://github.com/joshmarshall/jsonrpclib/

The jsonrpclib library requires one of the JSON libraries. It looks first for cjson, then for the built-in JSON library (with default Python 2.6+ distributions), and finally the simplejson library.

This library is an implementation of both the JSON-RPC and the XML-RPC specification (server-side) for the Tornado web framework. It supports the basic features of both, as well as the MultiCall / Batch support for both specifications. The JSON-RPC handler supports both the original 1.0 specification, as well as the new (proposed) 2.0 spec, which includes batch submission, keyword arguments, etc.

There is also a base library that other RPC protocols could use to quickly get tied into Tornado.

The library obviously requires Tornado, which you can get at Tornado's website (http://www.tornadoweb.org). After installing Tornado (instructions included with the Tornado distribution) you should be able to use the XML-RPC handler without any other libraries.

The JSON-RPC handler requires my jsonrpclib library, which you can get at http://github.com/joshmarshall/jsonrpclib . It also requires a JSON library, although any distribution of Python past 2.5 should have it by default. (Note: Some Linuxes only include a base Python install. On Ubuntu, for instance, you may need to run sudo apt-get install python-json or sudo apt-get python-cjson to get one of the libraries.)

The library is designed to be mostly transparent in usage. You simply extend the XML/JSON RPCHandler class from either the tornadorpc.xml or the tornado.json library, resepectively, and pass that handler in to the Tornado framework just like any other handler.

For any synchronous (normal) operation, you can just return the value you want sent to the client. However, if you use any asynchronous library (like Tornado's AsyncHTTPClient) you will want to call self.result(RESULT) in your callback. See the Asynchronous section below for examples.

To set up a simple XML RPC server, this is all you need:

from tornadorpc.xml import XMLRPCHandler
from tornadorpc import private, start_server

class Handler(XMLRPCHandler):

    def add(self, x, y):
        return x+y

    def ping(self, obj):
        return obj

    @private
    def private(self):
        #should not get called
        return False

start_server(Handler, port=8080)

The @private decorator is a way to ensure that it cannot be called externally. You can also create methods that start with an underscore _ character, and they will be private by default. The start_server function is just an easy wrap around the default Tornado setup -- you can use these handlers just like you would any other Tornado RequestHandler.

A JSON-RPC server would be started with the exact same syntax, replacing XMLRPCHandler with JSONRPCHandler. Here is an example of the JSON-RPC client with "dot-attribute" support:

from tornadorpc.json import JSONRPCHandler
from tornadorpc import private, start_server

class Tree(object):

    def power(self, base, power, modulo=None):
        result = pow(base, power, modulo)
        return result

    def _private(self):
        # Won't be callable
        return False

class Handler(JSONRPCHandler):

    tree = Tree()

    def add(self, x, y):
        return x+y

    def ping(self, obj):
        return obj

start_server(Handler, port=8080)

To use this, you should be able to use either the JSON-RPC official implementation, or the jsonrpclib library (which you'd need for this to work anyway.) One of the benefits of the jsonrpclib is designed to be a parallel implementation to the xmlrpclib, so syntax should be very similar and it should be easy to experiment with existing apps.

An example of client usage would be:

from jsonrpclib import Server
server = Server('http://localhost:8080')
result = server.tree.power(2, 6)
# result should equal 64

To indicate that a request is asynchronous, simply use the "async" decorator, and call "self.result(RESULT)" in your callback. Please note that this will only work in the RPCHandler methods, not in any sub-tree methods since they do not have access to the handler's result() method.

Here is an example that uses Tornado's AsyncHTTPClient with a callback:

from tornadorpc import async
from tornadorpc.xml import XMLRPCHandler
from tornado.httpclient import AsyncHTTPClient

class Handler(XMLRPCHandler):
    
    @async
    def external(self, url):
        client = AsyncHTTPClient()
        client.fetch(url, self._handle_response)
        
    def _handle_response(self, response):
        # The underscore will make it private automatically
        # You could also use @private if you wished
        # This returns the status code of the request
        self.result(response.code)

There is a config object that is available -- it will be expanded as time goes by. Currently, it supports two options: verbose and short_errors, both of which default to True. The verbose setting just specifies whether you want to print out results to the terminal (automatically on, you'll probably want to turn that off for production, WSGI deployment, etc.) and the short_errors option determines whether to print just the last few lines of the traceback (if set to True, default) or print the full traceback. Once the logging mechanism is in place, the short_errors configuration element will apply to that as well.

The default error look something similar to this:

JSON-RPC SERVER AT http://localhost:8484
---------------
ERROR IN messup
---------------
Traceback (most recent call last):
  File "test.py", line 20, in messup
    return doesntexist['bad_key']
NameError: global name 'doesntexist' is not defined

To change the configuration, look over the following:

import tornadorpc
tornadorpc.config.verbose = False
tornadorpc.config.short_errors = False
# or...
from tornadorpc import config
config.verbose = False
config.short_errors = False

To run some basic tests, enter the following in the same directory that this README is in:

python run_tests.py

This will test a few basic utilites and the XMLRPC system. If you wish to test the JSONRPC system, run the following:

python run_tests.py --json

• Add unit tests
• Add logging mechanism
• Add proper HTTP codes for failures
• Optimize