pjrpc is an extensible JSON-RPC client/server library with an intuitive interface
that can be easily extended and integrated in your project without writing a lot of boilerplate code.
Features:
- framework agnostic
- intuitive api
- extendability
- synchronous and asynchronous client backed
- synchronous and asynchronous server support
- popular frameworks integration
- builtin parameter validation
- pytest integration
- openapi schema generation support
- web ui support (SwaggerUI, RapiDoc, ReDoc)
You can install pjrpc with pip:
$ pip install pjrpcDocumentation is available at Read the Docs.
pjrpc client interface is very simple and intuitive. Methods may be called by name, using proxy object
or by sending handmade pjrpc.common.Request class object. Notification requests can be made using
pjrpc.client.AbstractClient.notify method or by sending a pjrpc.common.Request object without id.
import pjrpc
from pjrpc.client.backend import requests as pjrpc_client
client = pjrpc_client.Client('http://localhost/api/v1')
response: pjrpc.Response = client.send(pjrpc.Request('sum', params=[1, 2], id=1))
print(f"1 + 2 = {response.result}")
result = client('sum', a=1, b=2)
print(f"1 + 2 = {result}")
result = client.proxy.sum(1, 2)
print(f"1 + 2 = {result}")
client.notify('tick')Asynchronous client api looks pretty much the same:
import pjrpc
from pjrpc.client.backend import aiohttp as pjrpc_client
client = pjrpc_client.Client('http://localhost/api/v1')
response = await client.send(pjrpc.Request('sum', params=[1, 2], id=1))
print(f"1 + 2 = {response.result}")
result = await client('sum', a=1, b=2)
print(f"1 + 2 = {result}")
result = await client.proxy.sum(1, 2)
print(f"1 + 2 = {result}")
await client.notify('tick')Batch requests also supported. You can build pjrpc.common.BatchRequest request by your hand and then send it to the
server. The result is a pjrpc.common.BatchResponse instance you can iterate over to get all the results or get
each one by index:
import pjrpc
from pjrpc.client.backend import requests as pjrpc_client
client = pjrpc_client.Client('http://localhost/api/v1')
with client.batch() as batch:
batch.send(pjrpc.Request('sum', [2, 2], id=1))
batch.send(pjrpc.Request('sub', [2, 2], id=2))
batch.send(pjrpc.Request('div', [2, 2], id=3))
batch.send(pjrpc.Request('mult', [2, 2], id=4))
batch_response = batch.get_response()
print(f"2 + 2 = {batch_response[0].result}")
print(f"2 - 2 = {batch_response[1].result}")
print(f"2 / 2 = {batch_response[2].result}")
print(f"2 * 2 = {batch_response[3].result}")There are also several alternative approaches which are a syntactic sugar for the first one (note that the result
is not a pjrpc.common.BatchResponse object anymore but a tuple of "plain" method invocation results):
- using call notation:
async with client.batch() as batch:
batch('sum', 2, 2)
batch('sub', 2, 2)
batch('div', 2, 2)
batch('mult', 2, 2)
result = batch.get_results()
print(f"2 + 2 = {result[0]}")
print(f"2 - 2 = {result[1]}")
print(f"2 / 2 = {result[2]}")
print(f"2 * 2 = {result[3]}")- using proxy call:
async with client.batch() as batch:
batch.proxy.sum(2, 2)
batch.proxy.sub(2, 2)
batch.proxy.div(2, 2)
batch.proxy.mult(2, 2)
result = batch.get_results()
print(f"2 + 2 = {result[0]}")
print(f"2 - 2 = {result[1]}")
print(f"2 / 2 = {result[2]}")
print(f"2 * 2 = {result[3]}")Which one to use is up to you but be aware that if any of the requests returns an error the result of the other ones will be lost. In such case the first approach can be used to iterate over all the responses and get the results of the succeeded ones like this:
import pjrpc
from pjrpc.client.backend import requests as pjrpc_client
client = pjrpc_client.Client('http://localhost/api/v1')
batch_response = client.send(
pjrpc.BatchRequest(
pjrpc.Request('sum', [2, 2], id=1),
pjrpc.Request('sub', [2, 2], id=2),
pjrpc.Request('div', [2, 2], id=3),
pjrpc.Request('mult', [2, 2], id=4),
)
)
for response in batch_response:
if response.is_success:
print(response.result)
else:
print(response.error)Batch notifications:
import pjrpc
from pjrpc.client.backend import requests as pjrpc_client
client = pjrpc_client.Client('http://localhost/api/v1')
with client.batch() as batch:
batch.notify('tick')
batch.notify('tack')
batch.notify('tick')
batch.notify('tack')pjrpc supports popular backend frameworks like aiohttp,
flask and message brokers like aio_pika.
Running of aiohttp based JSON-RPC server is a very simple process. Just define methods, add them to the registry and run the server:
import uuid
from aiohttp import web
import pjrpc.server
from pjrpc.server.integration import aiohttp
methods = pjrpc.server.MethodRegistry()
@methods.add(pass_context='request')
async def add_user(request: web.Request, user: dict) -> dict:
user_id = uuid.uuid4().hex
request.app['users'][user_id] = user
return {'id': user_id, **user}
jsonrpc_app = aiohttp.Application('/api/v1')
jsonrpc_app.add_methods(methods)
jsonrpc_app.app['users'] = {}
if __name__ == "__main__":
web.run_app(jsonrpc_app.http_app, host='localhost', port=8080)Very often besides dumb method parameters validation it is necessary to implement more "deep" validation and provide
comprehensive errors description to clients. Fortunately pjrpc has builtin parameter validation based on
pydantic library which uses python type annotation for validation.
Look at the following example: all you need to annotate method parameters (or describe more complex types beforehand if
necessary). pjrpc will be validating method parameters and returning informative errors to clients.
import enum
import uuid
import pydantic
from aiohttp import web
import pjrpc.server
from pjrpc.server.validators import pydantic as validators
from pjrpc.server.integration import aiohttp
methods = pjrpc.server.MethodRegistry(
validator_factory=validators.PydanticValidatorFactory(exclude=aiohttp.is_aiohttp_request),
)
class ContactType(enum.Enum):
PHONE = 'phone'
EMAIL = 'email'
class Contact(pydantic.BaseModel):
type: ContactType
value: str
class User(pydantic.BaseModel):
name: str
surname: str
age: int
contacts: list[Contact]
@methods.add(pass_context='request')
async def add_user(request: web.Request, user: User):
user_id = uuid.uuid4()
request.app['users'][user_id] = user
return {'id': user_id, **user.dict()}
class JSONEncoder(pjrpc.server.JSONEncoder):
def default(self, o):
if isinstance(o, uuid.UUID):
return o.hex
if isinstance(o, enum.Enum):
return o.value
return super().default(o)
jsonrpc_app = aiohttp.Application('/api/v1', json_encoder=JSONEncoder)
jsonrpc_app.add_methods(methods)
jsonrpc_app.http_app['users'] = {}
if __name__ == "__main__":
web.run_app(jsonrpc_app.http_app, host='localhost', port=8080)pjrpc implements all the errors listed in protocol specification
which can be found in pjrpc.common.exceptions module so that error handling is very simple and "pythonic-way":
import pjrpc
from pjrpc.client.backend import requests as pjrpc_client
client = pjrpc_client.Client('http://localhost/api/v1')
try:
result = client.proxy.sum(1, 2)
except pjrpc.MethodNotFound as e:
print(e)Default error list can be easily extended. All you need to create an error class inherited from
pjrpc.exc.TypedError and define an error code and a description message. pjrpc will be automatically
deserializing custom errors for you:
import pjrpc
from pjrpc.client.backend import requests as pjrpc_client
class UserNotFound(pjrpc.exc.TypedError):
CODE = 1
MESSAGE = 'user not found'
client = pjrpc_client.Client('http://localhost/api/v1')
try:
result = client.proxy.get_user(user_id=1)
except UserNotFound as e:
print(e)On the server side everything is also pretty straightforward:
import uuid
import flask
import pjrpc
from pjrpc.server import MethodRegistry
from pjrpc.server.integration import flask as integration
methods = pjrpc.server.MethodRegistry()
class UserNotFound(pjrpc.exc.TypedError):
CODE = 1
MESSAGE = 'user not found'
@methods.add()
def add_user(user: dict) -> dict:
user_id = uuid.uuid4().hex
flask.current_app.users[user_id] = user
return {'id': user_id, **user}
@methods.add()
def get_user(user_id: str) -> dict:
user = flask.current_app.users.get(user_id)
if not user:
raise UserNotFound(data=user_id)
return user
json_rpc = integration.JsonRPC('/api/v1')
json_rpc.add_methods(methods)
json_rpc.http_app.users = {}
if __name__ == "__main__":
json_rpc.http_app.run(port=80)pjrpc has built-in OpenAPI and OpenRPC
specification generation support and integrated web UI as an extra dependency. Three UI types are supported:
- SwaggerUI (https://swagger.io/tools/swagger-ui/)
- RapiDoc (https://mrin9.github.io/RapiDoc/)
- ReDoc (https://github.com/Redocly/redoc)
Web UI extra dependency can be installed using the following code:
$ pip install pjrpc[openapi-ui-bundles]The following example illustrates how to configure OpenAPI specification generation and Swagger UI web tool with basic auth:
import uuid
from typing import Annotated, Any
import aiohttp.typedefs
import pydantic as pd
from aiohttp import web
import pjrpc.server.specs.extractors.pydantic
import pjrpc.server.specs.openapi.ui
from pjrpc.server.integration import aiohttp as integration
from pjrpc.server.specs import extractors
from pjrpc.server.specs import openapi as specs
from pjrpc.server.validators import pydantic as validators
methods = pjrpc.server.MethodRegistry(
validator_factory=validators.PydanticValidatorFactory(exclude=integration.is_aiohttp_request),
metadata_processors=[
specs.MethodSpecificationGenerator(
extractor=extractors.pydantic.PydanticMethodInfoExtractor(
exclude=integration.is_aiohttp_request,
),
),
],
)
UserName = Annotated[
str,
pd.Field(description="User name", examples=["John"]),
]
UserSurname = Annotated[
str,
pd.Field(description="User surname", examples=['Doe']),
]
UserAge = Annotated[
int,
pd.Field(description="User age", examples=[36]),
]
UserId = Annotated[
uuid.UUID,
pd.Field(description="User identifier", examples=["226a2c23-c98b-4729-b398-0dae550e99ff"]),
]
class UserIn(pd.BaseModel):
"""
User registration data.
"""
name: UserName
surname: UserSurname
age: UserAge
class UserOut(UserIn):
"""
Registered user data.
"""
id: UserId
class AlreadyExistsError(pjrpc.exc.TypedError):
"""
User already registered error.
"""
CODE = 2001
MESSAGE = "user already exists"
class NotFoundError(pjrpc.exc.TypedError):
"""
User not found error.
"""
CODE = 2002
MESSAGE = "user not found"
@methods.add(
pass_context='request',
metadata=[
specs.metadata(
summary='Creates a user',
tags=['users'],
errors=[AlreadyExistsError],
),
],
)
def add_user(request: web.Request, user: UserIn) -> UserOut:
for existing_user in request.config_dict['users'].values():
if user.name == existing_user.name:
raise AlreadyExistsError()
user_id = uuid.uuid4()
request.config_dict['users'][user_id] = user
return UserOut(id=user_id, **user.model_dump())
@methods.add(
pass_context='request',
metadata=[
specs.metadata(
summary='Returns a user',
tags=['users'],
errors=[NotFoundError],
),
],
)
def get_user(request: web.Request, user_id: UserId) -> UserOut:
user = request.config_dict['users'].get(user_id.hex)
if not user:
raise NotFoundError()
return UserOut(id=user_id, **user.model_dump())
@methods.add(
pass_context='request',
metadata=[
specs.metadata(
summary='Deletes a user',
tags=['users'],
errors=[NotFoundError],
),
],
)
def delete_user(request: web.Request, user_id: UserId) -> None:
user = request.config_dict['users'].pop(user_id.hex, None)
if not user:
raise NotFoundError()
class JSONEncoder(pjrpc.server.JSONEncoder):
def default(self, o: Any) -> Any:
if isinstance(o, pd.BaseModel):
return o.model_dump()
if isinstance(o, uuid.UUID):
return str(o)
return super().default(o)
openapi_spec = specs.OpenAPI(
info=specs.Info(version="1.0.0", title="User storage"),
servers=[
specs.Server(
url='http://127.0.0.1:8080',
),
],
security_schemes=dict(
basicAuth=specs.SecurityScheme(
type=specs.SecuritySchemeType.HTTP,
scheme='basic',
),
),
security=[
dict(basicAuth=[]),
],
)
http_app = web.Application()
http_app['users'] = {}
jsonrpc_app = integration.Application('/api')
jsonrpc_app.add_spec(openapi_spec, path='openapi.json')
jsonrpc_app.add_spec_ui('swagger', specs.ui.SwaggerUI(), spec_url='../openapi.json')
jsonrpc_app.add_spec_ui('redoc', specs.ui.ReDoc(), spec_url='../openapi.json')
jsonrpc_v1_app = integration.Application(http_app=web.Application(), json_encoder=JSONEncoder)
jsonrpc_v1_app.add_methods(methods)
jsonrpc_app.add_subapp('/v1', jsonrpc_v1_app)
http_app.add_subapp('/rpc', jsonrpc_app.http_app)
if __name__ == "__main__":
web.run_app(http_app, host='localhost', port=8080)Specification is available on http://localhost:8080/rpc/api/openapi.json
Web UI is running on http://localhost:8080/rpc/api/swagger/ and http://localhost:8080/rpc/api/redoc/