serverwamp 0.2.1

Components that add Web Application Messaging Protocol features to WebSocket servers.

serverwamp

Components that add Web Application Messaging Protocol features to WebSocket servers. With serverwamp, a server can act as both a WAMP broker and dealer. Currently supports ASGI servers using asyncio or Trio and aiohttp servers.

Usage

Example

This service enables retrieval and removal of documents from a database via two RPCs exposed to WAMP clients.

from aiohttp import web
from serverwamp.adapters.aiohttp import WAMPApplication
from serverwamp.rpc import RPCRouteSet, RPCError

docs_api = RPCRouteSet()

@docs_api.route('docs.getDocument')
async def get_doc(document_id):
    record = await mydb.retrieve(document_id)
    return {'status': 'SUCCESS', 'document': record}

@docs_api.route('docs.deleteDocument')
async def delete_doc(document_id):
    succeeded = await mydb.delete(document_id)
    if succeeded:
        return {'status': 'SUCCESS'}
    else:
        raise RPCError('wamp.error.delete_failed', {'status': 'FAILURE'})


if __name__ == '__main__':
    wamp = WAMPApplication()
    wamp.add_rpc_routes(docs_api)

    app = web.Application()
    app.add_routes((web.get('/', wamp.handle),))
    web.run_app(app)

Type Marshalling

RPC route handlers can supply type hints that are used to transform call arguments passed in from the client.

@rpc_api.route('slowlyCountToNumber')
async def slowly_count(count_to: int, interval: float, taunt: str = 'ah ah ah!'):
    for i in range(count_to):
        print(f'{i}, {taunt}')
        await asyncio.sleep(interval)

from decimal import Decimal

@rpc_api.route('storeHighPrecisionTimestamp')
async def store_timestamp(timestamp: Decimal):
    pass

Without type hints, arguments are injected as they were parsed by the deserializer (e.g. float for JSON Number).

Built-in and Custom RPC Handler Arguments

If an RPC route contains arguments of certain names, these arguments are automatically populated.

In this example for trio, a custom nursery argument is supplied for running background tasks. The built-in argument session is requested so a background task can later send an event to the session if it subscribed to the relevant topic.

import trio
from serverwamp.rpc import RPCRouteSet
from serverwamp.adapters.asgi_trio import WAMPApplication

async def long_running_job(session):
    await session.send_event('job_events', job_status='STARTED')
    await trio.sleep(3600)
    await session.send_event('job_events', job_status='COMPLETED')

rpc_api = RPCRouteSet()

@rpc_api.route('doJob')
async def do_job(self, nursery, session):
    nursery.start_soon(long_running_job(session))
    return 'Job scheduled.'

async def application(*args, **kwargs):
    async with trio.open_nursery() as rpc_nursery:
        wamp = WAMPApplication()
        wamp.set_default_rpc_arg('nursery', rpc_nursery)
        wamp.add_rpc_routes(rpc_api)

        return await wamp.asgi_application(*args, **kwargs)

Compared to Traditional WAMP

Traditionally, WAMP clients connect to a WAMP router either to provide functionality or to use functionality provided by other clients. This library enables a server to respond to calls and subscriptions like a WAMP router would, but serve them in a custom way instead of routing them to other clients.

You might do this to provide Python-based web services over a WebSocket using a typical client and server model, while taking advantage of the structure the WAMP protocol provides.

In summary, this is not a WAMP router, but you could use it to build one if you wanted.

This library is good if…

• you like developing in micro-frameworks like Flask and aiohttp, but want to branch into serving over WebSockets.
• you want a structure for communicating over WebSockets, but want more request/response features than the socket.io protocol provides.
• you want to build a WAMP router.

It's not useful if…

• you want to build a WAMP client
  • Consider Autobahn|Python instead.
• you want a WAMP router out of the box.
  • Consider Crossbar.io instead.

Known Deficiencies

• Project is in the prototyping stages, so documentation isn't available on low-level features yet.

Development

Unit tests can be run with:

pip install -e .
pytest