JSON-RPC for python

There are two implementations hosted on json-rpc.org: The which has support for JSON-RPC over TCP and HTTP. And there is the rewrite of JSON-RPC for python, it is described in this document. You can the code or check it out using subversion from the .

The current implementation provides functionality for making JSON-RPC services available through CGI and through a mod-python handler. It also provides an easy to use proxy class for calling JSON-RPC services from python.

Installation

Download the source using bazaar.

$ bzr checkout http://bzr.json-rpc.org/trunk

As root/administrator install the package running the provided setup script.

$ cd trunk/python-jsonrpc$ python setup.py install

If you do not wish to install the package you can simply copy the jsonrpc folder to where python can find it when it searches for modules to be imported. E.g. this can be the same place where you python script resides in.

Using the ServiceProxy class

If everything worked you should now be able to call JSON-RPC services. Start your favourite python shell and enter the code below:

>>> from jsonrpc import ServiceProxy>>> s = ServiceProxy("http://jsolait.net/services/test.jsonrpc")>>> print s.echo("foobar")

The example above creates a proxy object for a service hosted on jsolait.net It calls the service's echo method and prints out the result of the call.

Creating CGI based services

To provide your own service you can create a python CGI-Script and use jsonrpc's handleCGI method for handling requests.

#!/usr/bin/env pythonfrom jsonrpc import handleCGI, ServiceMethod@ServiceMethoddef echo(msg):    return msgif __name__ == "__main__":    handleCGI()

This is the simplest way to create a service using as CGI script. All methods in the script decorated using the ServiceMethod decorator are available to remote callers. all other methods are inaccessible to the "outside world".

handleCGI() gives you some flexibility to define what to use as the service. By default, as seen above it uses the __main__ module as a service. You can though, specify a particular service to be used by passing it to handleCGI(service) as first parameter:

#!/usr/bin/env pythonfrom jsonrpc import handleCGI, ServiceMethodclass MyService(object):    @ServiceMethod    def echo(self, msg):        return msgif __name__ == "__main__":    service = MyService()    handleCGI(service)

Creating mod-python based services

Similar to the way the CGI handling works, you can use jsonrpc's mod-python handler. First you need to install and setup mod-python to handle service URLs using jsonrpc's mod-python handler. E.g. in your .htaccess file in any folder that is being served by apache add:

AddHandler mod_python .pyPythonHandler jsonrpc

Make sure apache is setup to allow you to add the AddType Directive in .htaccess files. Alternatively you can create an apache config file which gets loaded by apache upon startup. In a Location or Directory section you should add the python handler:

Alias /services/ /var/www/json-rpc-services/
       AddHandler mod_python .py  PythonHandler jsonrpc

If you have not installed jsonrpc using it's setup script, you will need to add it's location to python's sys-path so that mod-python can find it. In your apache config or .htaccess file add:

...  PythonHandler jsonrpc  PythonPath "sys.path+["/path/to/where/jsorpc/package/is/located/"]

Now you need to create a python script that will be used as a service. Place it in a sub folder that is covered by the Directives above. E.g. in the folder of where .htaccess is located or a subfolder thereof or in any sub-folder of /var/www/json-rpc-services/ for the second config example.

Similar to the CGI based service you can create a script with methods decorated using the ServiceMethod decorator:

from jsonrpc import handleCGI, ServiceMethod@ServiceMethoddef echo(msg):    return msg

Again, this is probably the simplest way to create a service.

You can also create a script which exposes a service , which will then be used as the service.

from jsonrpc import ServiceMethodclass MyService(object):    @ServiceMethod    def echo(self, msg):        return msgservice = MyService()

or you create a script which exposes a Service class. A service object will be created using this class and used as a service.

from jsonrpc import ServiceMethodclass Service(object):    @ServiceMethod    def echo(self, msg):        return msg

Testing your services

Lets assume you created a mod-python based or a CGI based service using a script called test.py and that that script can be found under the URL http://localhost/services/test.py (check your web server configuration).

Now you should be able to use your services with any JSON-RPC client that supports json-rpc over HTTP. The simplest way to try it out, is to start your python shell an use jsonrpc's ServiceProxy class.

>>> from jsonrpc import ServiceProxy>>> s = ServiceProxy("http://localhost/services/test.py")>>> print s.echo("foobar")

Error handling

Any error that the ServiceProxy received through the JSON-RPC protocoll will be raised as a JSONRPCException before the called method returns. The exception raised will contain a service specific error object, which can be accessed using the exception's error property.

try:  print s.echo("foobar")except JSONRPCException, e:  print repr(e.error)

Any exception raised in a Service's method during invokation will be converted into an error object and transmitted back to the caller by jsonrpc. The error object will use the exception's class name as a name property and it's message property as the message property of the error object being returned.