skip to navigation
skip to content

Not Logged In

jsonrpclib-pelix 0.2.0

This project is an implementation of the JSON-RPC v2.0 specification (backwards-compatible) as a client library. This version is a fork of jsonrpclib by Josh Marshall, usable with Pelix remote services.

Latest Version: 0.2.6

This library is an implementation of the JSON-RPC specification. It supports both the original 1.0 specification, as well as the new (proposed) 2.0 specification, which includes batch submission, keyword arguments, etc.

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

About this version

This is a patched version of the original jsonrpclib project by Josh Marshall, available at https://github.com/joshmarshall/jsonrpclib.

The suffix -pelix only indicates that this version works with Pelix Remote Services, but it is not a Pelix specific implementation.

  • This version adds support for Python 3, staying compatible with Python 2.
  • It is now possible to use the dispatch_method argument while extending the SimpleJSONRPCDispatcher, to use a custom dispatcher. This allows to use this package by Pelix Remote Services.
  • The modifications added in other forks of this project have been added:
  • The support for Unix sockets has been removed, as it is not trivial to convert to Python 3 (and I don’t use them)
  • This version cannot be installed with the original jsonrpclib, as it uses the same package name.

Summary

This library implements the JSON-RPC 2.0 proposed specification in pure Python. It is designed to be as compatible with the syntax of xmlrpclib as possible (it extends where possible), so that projects using xmlrpclib could easily be modified to use JSON and experiment with the differences.

It is backwards-compatible with the 1.0 specification, and supports all of the new proposed features of 2.0, including:

  • Batch submission (via MultiCall)
  • Keyword arguments
  • Notifications (both in a batch and ‘normal’)
  • Class translation using the __jsonclass__ key.

I’ve added a “SimpleJSONRPCServer”, which is intended to emulate the “SimpleXMLRPCServer” from the default Python distribution.

Requirements

It supports cjson and simplejson, and looks for the parsers in that order (searching first for cjson, then for the built-in json in 2.6+, and then the simplejson external library). One of these must be installed to use this library, although if you have a standard distribution of 2.6+, you should already have one. Keep in mind that cjson is supposed to be the quickest, I believe, so if you are going for full-on optimization you may want to pick it up.

Since library uses contextlib module, you should have at least Python 2.5 installed.

Installation

You can install this from PyPI with one of the following commands (sudo may be required):

easy_install jsonrpclib-pelix
pip install jsonrpclib-pelix

Alternatively, you can download the source from the GitHub repository at http://github.com/tcalmant/jsonrpclib and manually install it with the following commands:

git clone git://github.com/tcalmant/jsonrpclib.git
cd jsonrpclib
python setup.py install

SimpleJSONRPCServer

This is identical in usage (or should be) to the SimpleXMLRPCServer in the Python standard library. Some of the differences in features are that it obviously supports notification, batch calls, class translation (if left on), etc. Note: The import line is slightly different from the regular SimpleXMLRPCServer, since the SimpleJSONRPCServer is distributed within the jsonrpclib library.

from jsonrpclib.SimpleJSONRPCServer import SimpleJSONRPCServer

server = SimpleJSONRPCServer(('localhost', 8080))
server.register_function(pow)
server.register_function(lambda x,y: x+y, 'add')
server.register_function(lambda x: x, 'ping')
server.serve_forever()

Client Usage

This is (obviously) taken from a console session.

>>> import jsonrpclib
>>> server = jsonrpclib.ServerProxy('http://localhost:8080')
>>> server.add(5,6)
11
>>> server.add(x=5, y=10)
15
>>> server._notify.add(5,6)
# No result returned...
>>> batch = jsonrpclib.MultiCall(server)
>>> batch.add(5, 6)
>>> batch.ping({'key':'value'})
>>> batch._notify.add(4, 30)
>>> results = batch()
>>> for result in results:
>>> ... print(result)
11
{'key': 'value'}
# Note that there are only two responses -- this is according to spec.

# Clean up
>>> server('close')()

# Using client history
>>> history = jsonrpclib.history.History()
>>> server = jsonrpclib.ServerProxy('http://localhost:8080', history=history)
>>> server.add(5,6)
11
>>> print(history.request)
{"id": "f682b956-c8e1-4506-9db4-29fe8bc9fcaa", "jsonrpc": "2.0",
 "method": "add", "params": [5, 6]}
>>> print(history.response)
{"id": "f682b956-c8e1-4506-9db4-29fe8bc9fcaa", "jsonrpc": "2.0",
 "result": 11}

# Clean up
>>> server('close')()

If you need 1.0 functionality, there are a bunch of places you can pass that in, although the best is just to give a specific configuration to jsonrpclib.ServerProxy:

>>> import jsonrpclib
>>> jsonrpclib.config.DEFAULT.version
2.0
>>> config = jsonrpclib.config.Config(version=1.0)
>>> history = jsonrpclib.history.History()
>>> server = jsonrpclib.ServerProxy('http://localhost:8080', config=config,
                                    history=history)
>>> server.add(7, 10)
17
>>> print(history.request)
{"id": "827b2923-5b37-49a5-8b36-e73920a16d32",
 "method": "add", "params": [7, 10]}
>>> print(history.response)
{"id": "827b2923-5b37-49a5-8b36-e73920a16d32", "error": null, "result": 17}
>>> server('close')()

The equivalent loads and dumps functions also exist, although with minor modifications. The dumps arguments are almost identical, but it adds three arguments: rpcid for the ‘id’ key, version to specify the JSON-RPC compatibility, and notify if it’s a request that you want to be a notification.

Additionally, the loads method does not return the params and method like xmlrpclib, but instead a.) parses for errors, raising ProtocolErrors, and b.) returns the entire structure of the request / response for manual parsing.

Additional headers

If your remote service requires custom headers in request, you can pass them as as a headers keyword argument, when creating the ServerProxy:

>>> import jsonrpclib
>>> server = jsonrpclib.ServerProxy("http://localhost:8080",
                                    headers={'X-Test' : 'Test'})

You can also put additional request headers only for certain method invocation:

>>> import jsonrpclib
>>> server = jsonrpclib.Server("http://localhost:8080")
>>> with server._additional_headers({'X-Test' : 'Test'}) as test_server:
...     test_server.ping(42)
...
>>> # X-Test header will be no longer sent in requests

Of course _additional_headers contexts can be nested as well.

Class Translation

I’ve recently added “automatic” class translation support, although it is turned off by default. This can be devastatingly slow if improperly used, so the following is just a short list of things to keep in mind when using it.

  • Keep It (the object) Simple Stupid. (for exceptions, keep reading.)
  • Do not require init params (for exceptions, keep reading)
  • Getter properties without setters could be dangerous (read: not tested)

If any of the above are issues, use the _serialize method. (see usage below) The server and client must BOTH have use_jsonclass configuration item on and they must both have access to the same libraries used by the objects for this to work.

If you have excessively nested arguments, it would be better to turn off the translation and manually invoke it on specific objects using jsonrpclib.jsonclass.dump / jsonrpclib.jsonclass.load (since the default behavior recursively goes through attributes and lists / dicts / tuples).

Sample file: test_obj.py
# This object is /very/ simple, and the system will look through the
# attributes and serialize what it can.
class TestObj(object):
    foo = 'bar'

# This object requires __init__ params, so it uses the _serialize method
# and returns a tuple of init params and attribute values (the init params
# can be a dict or a list, but the attribute values must be a dict.)
class TestSerial(object):
    foo = 'bar'
    def __init__(self, *args):
        self.args = args
    def _serialize(self):
        return (self.args, {'foo':self.foo,})
  • Sample usage
>>> import jsonrpclib
>>> import test_obj

# History is used only to print the serialized form of beans
>>> history = jsonrpclib.history.History()
>>> testobj1 = test_obj.TestObj()
>>> testobj2 = test_obj.TestSerial()
>>> server = jsonrpclib.Server('http://localhost:8080', history=history)

# The 'ping' just returns whatever is sent
>>> ping1 = server.ping(testobj1)
>>> ping2 = server.ping(testobj2)

>>> print(history.request)
{"id": "7805f1f9-9abd-49c6-81dc-dbd47229fe13", "jsonrpc": "2.0",
 "method": "ping", "params": [{"__jsonclass__":
                               ["test_obj.TestSerial", []], "foo": "bar"}
                             ]}
>>> print(history.response)
{"id": "7805f1f9-9abd-49c6-81dc-dbd47229fe13", "jsonrpc": "2.0",
 "result": {"__jsonclass__": ["test_obj.TestSerial", []], "foo": "bar"}}

This behavior is turned by default. To deactivate it, just set the use_jsonclass member of a server Config to False. If you want to use a per-class serialization method, set its name in the serialize_method member of a server Config. Finally, if you are using classes that you have defined in the implementation (as in, not a separate library), you’ll need to add those (on BOTH the server and the client) using the config.classes.add() method.

Feedback on this “feature” is very, VERY much appreciated.

Why JSON-RPC?

In my opinion, there are several reasons to choose JSON over XML for RPC:

  • Much simpler to read (I suppose this is opinion, but I know I’m right. :)
  • Size / Bandwidth - Main reason, a JSON object representation is just much smaller.
  • Parsing - JSON should be much quicker to parse than XML.
  • Easy class passing with jsonclass (when enabled)

In the interest of being fair, there are also a few reasons to choose XML over JSON:

  • Your server doesn’t do JSON (rather obvious)
  • Wider XML-RPC support across APIs (can we change this? :))
  • Libraries are more established, i.e. more stable (Let’s change this too.)

Tests

Tests are an almost-verbatim drop from the JSON-RPC specification 2.0 page. They can be run using unittest or nosetest:

python -m unittest discover tests
python3 -m unittest discover tests
nosetests tests
 
File Type Py Version Uploaded on Size
jsonrpclib-pelix-0.2.0.tar.gz (md5) Source 2014-08-28 21KB
jsonrpclib_pelix-0.2.0-py2.py3-none-any.whl (md5) Python Wheel 2.7 2014-08-28 29KB
  • Downloads (All Versions):
  • 0 downloads in the last day
  • 414 downloads in the last week
  • 2240 downloads in the last month