Skip to main content

JSON-RPC v2.0 for Trio

Project description

JSON-RPC v2.0 for Trio

PyPI Python Versions MIT License Build Status codecov

This project provides an implementation of JSON-RPC v 2.0 based on sansio-jsonrpc with all of the I/O implemented using the Trio asynchronous framework.

Client Example

The following example shows a basic JSON-RPC client.

from trio_jsonrpc import open_jsonrpc_ws, JsonRpcException

async def main():
    async with open_jsonrpc_ws('ws://') as client:
            result = await client.request(
                {'employee': 'Mark', 'pin': 1234}
            print('vault open:', result['vault_open'])

            await client.notify(method='hello_world')
        except JsonRpcException as jre:
            print('RPC failed:', jre)

The example begins by opening a JSON-RPC connection using a WebSocket transport. The implementation is designed to support multiple types of transport, but currently WebSocket transport is the only one that has been implemented.

Note that JSON-RPC does not contain any framing logic, i.e. a specification for how to identify message boundaries within a stream. Therefore, if you want to use JSON-RPC with raw TCP sockets, you either need to add your own framing logic or else use a streaming JSON parser. For this reason, we have chosen to focus on WebSocket transport initially, because WebSocket does include framing.

The connection is opened inside a context manager that guarantees the connection is ready when entering the block and automatically closes the connection when leaving the block.

Within the block, we call the client's request(...) method to send a JSON-RPC request. This method sends the request to the server, waits for a response, and returns a result. If the server indicates that an error occurred, a JsonRpcException will be raised instead. The client multiplexes requests so that it can be use concurrently from multiple tasks; responses are routed back to the appropriate task that called request(...).

The client also has a notify(...) method which sends a request to the server but does not expect or wait for a response.

Server Example

The following example shows a basic JSON-RPC server. The server is more DIY (do it yourself) than the client because a server has to incorporate several disparate functionalities:

  1. Setting up the transport, especially if the transport requires a handshake as WebSocket does.
  2. Handling new connections to the server.
  3. Multiplexing requests on a single connection.
  4. Dispatching a request to an appropriate handler.
  5. Managing connection state over the course of multiple requests. (I.e. allowing one handler to indicate that the connection is authorized, so other handlers can use that authorization information to make access control decisions.)
  6. Applying pre-handler or post-handler logic to each request, for example logging each request before it is dispatched.

This library cannot feasibly implement a default solution that handles the aforementioned items in a way that satsifies every downstream project. Instead, the library gives you the pieces you need to build a server. We will go through each piece one at a time.

from dataclasses import dataclass
import trio
from trio_jsonrpc import Dispatch, JsonRpcApplicationError
import trio_websocket

class ConnectionContext:
    """ A sample implementation for request context. """
    db: typing.Any = None
    authorized_employee: str = None

dispatch = Dispatch()

In this first piece, we import a few things we need. We also define a ConnectionContext class. The purpose of this class is to share mutable connection state between different handlers on the same connection. For example, we can have one handler that authenticates a user and then sets authorization data in the connection context. Later, another handler can check that authorization data to make access control decisions.

You are free to pass any object as a connection context, as long as it can be copied with copy.copy(). A dataclass is often convenient for this purpose.

async def open_vault_door(employee, pin):
    access = await dispatch.ctx.db.check_pin(employee, pin)
    if access:
        dispatch.ctx.authorized_employee = employee
        return {"vault_open": True}
        dispatch.ctx.authorized_employee = None
        raise JsonRpcApplicationError(code=-1, message="Not authorized.")

async def close_vault_door():
    dispatch.ctx.authorized_employee = None
    return {"vault_open": False}

In this section, we define two JSON-RPC methods. Each one is annotated with @dispatch.handler, which means when we dispatch an incoming request, it will look up the Python function that matches the JSON-RPC method name. The JSON-RPC parameters are passed as arguments to the handler function.

Each handler can access the connection context as dispatch.ctx.

Also note that if a handler needs to signal an error, it can raise JsonRpcApplicationError (or any subclass of it). The dispatcher will automatically convert the exception into a JSON-RPC error to send back to the client. If a handler raises any exception that is not a subclass of JsonRpcException—i.e. if your handler is buggy and raises something like KeyError—then a generic JsonRpcInternalError is sent back to the client, and the entire exception is logged.

async def main():
    db = ...
    base_context = ConnectionContext(db=db)

    async def responder(conn, recv_channel):
        async for result in recv_channel:
            if isinstance(result, JsonRpcException):
                await conn.respond_with_error(result.get_error())
                await conn.respond_with_result(result)

    async def connection_handler(ws_request):
        ws = await ws_request.accept()
        transport = WebSocketTransport(ws)
        rpc_conn = JsonRpcConnection(transport, JsonRpcConnectionType.SERVER)
        conn_context = copy(base_context)
        result_send, result_recv = trio.open_memory_channel(10)
        async with trio.open_nursery() as nursery:
            nursery.start_soon(responder, result_recv)
            async with dispatch.connection_context(conn_context):
                async for request in rpc_conn.iter_requests():
                    nursery.start_soon(dispatch.handle_request, request, result_send)

    await trio_websocket.serve_websocket(connection_handler, 'localhost', 8000, None)

The final section has a lot going on. First of all, we set up a base connection context. This base object is used as a blueprint: for each new connection, the context is copied and then set as the context for that connection. As long as that connection stays alive, all handlers will share that same context object.

At the end of main(), the server is started by calling trio_websocket.serve_websocket(). For each new connection, the connection_handler(...) is called. This function finishes the WebSocket handshake and then wraps the WebSocket connection into a JSON-RPC connection. Then it iterates over the incoming requests and uses the dispatcher to handle each one.

Since each JSON-RPC request is dispatched in a new task, it isn't possible to directly await the result of each task. Instead, we create a Trio channel and pass it into the dispatcher. When the handler finishes, its result will be written to this channel. We use a background task called responder(...) to read from this channel and actually send the response to the client.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for trio-jsonrpc, version 0.2.0
Filename, size File type Python version Upload date Hashes
Filename, size trio_jsonrpc-0.2.0-py3-none-any.whl (10.1 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size trio-jsonrpc-0.2.0.tar.gz (12.1 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page