Dispatching in tinyrpc is very similiar to url-routing in web frameworks. Functions are registered with a specific name and made public, i.e. callable, to remote clients.


Exposing a few functions:

from tinyrpc.dispatch import RPCDispatcher

dispatch = RPCDispatcher()

def foo():
    # ...

def bar(arg):
    # ...

# later on, assuming we know we want to call foo(*args, **kwargs):

f = dispatch.get_method('foo')
f(*args, **kwargs)

Using prefixes and instance registration:

from tinyrpc.dispatch import public

class SomeWebsite(object):
    def __init__(self, ...):
        # note: this method will not be exposed

    def secret(self):
        # another unexposed method

    def get_user_info(self, user):
        # ...

    # using a different name
    def get_comment(self, comment_id):
        # ...

The code above declares an RPC interface for SomeWebsite objects, consisting of two visible methods: get_user_info(user) and get_user_comment(comment_id).

These can be used with a dispatcher now:

def hello():
    # ...

website1 = SomeWebsite(...)
website2 = SomeWebsite(...)

from tinyrpc.dispatch import RPCDispatcher

dispatcher = RPCDispatcher()

# directly register version method
def version():
    # ...

# add earlier defined method

# register the two website instances
dispatcher.register_instance(website1, 'sitea.')
dispatcher.register_instance(website2, 'siteb.')

In the example above, the RPCDispatcher now knows a total of six registered methods: version, hello, sitea.get_user_info, sitea.get_user_comment, siteb.get_user_info, siteb.get_user_comment.

Automatic dispatching

When writing a server application, a higher level dispatching method is available with dispatch():

from tinyrpc.dispatch import RPCDispatcher

dispatcher = RPCDispatcher()

# register methods like in the examples above
# ...
# now assumes that a valid RPCRequest has been obtained, as `request`

response = dispatcher.dispatch(request)

# response can be directly processed back to the client, all Exceptions have
# been handled already

Class, static and unbound method dispatching

Although you will only rarely use these method types they do work and here we show you how.

Class methods do not have self as the initial parameter but rather a reference to their class. You may want to use such methods to instantiate class instances.

class ShowClassMethod:
    def func(cls, a, b):
        return a-b

Note the ordering of the decorators. Ordering them differently will not work. You call dispatch to the func method just as you would dispatch to any other method.

Static methods have neither a class nor instance reference as first parameter:

class ShowStaticMethod:
    def func(a, b):
        return a-b

Again the ordering of the decorators is critical and you dispatch them as any other method.

Finally it is possible to dispatch to unbound methods but I strongly advise against it. If you really want to do that see the tests to learn how. Everyone else should use static methods instead.

API reference

class tinyrpc.dispatch.RPCDispatcher

Bases: object

Stores name-to-method mappings.


Convenient decorator.

Allows easy registering of functions to this dispatcher. Example:

dispatch = RPCDispatcher()

def foo(bar):
    # ...

class Baz(object):
    def not_exposed(self):
        # ...

    def visible_method(arg1)
        # ...
Parameters:name (str) – Name to register callable with.
add_subdispatch(dispatcher: tinyrpc.dispatch.RPCDispatcher, prefix: str = '')

Adds a subdispatcher, possibly in its own namespace.

  • dispatcher (RPCDispatcher) – The dispatcher to add as a subdispatcher.
  • prefix (str) – A prefix. All of the new subdispatchers methods will be available as prefix + their original name.
add_method(f: Callable, name: str = None) → None

Add a method to the dispatcher.

  • f (callable) – Callable to be added.
  • name (str) – Name to register it with. If None, f.__name__ will be used.

RPCError – When the name is already registered.

get_method(name: str) → Callable

Retrieve a previously registered method.

Checks if a method matching name has been registered.

If get_method() cannot find a method, every subdispatcher with a prefix matching the method name is checked as well.

Parameters:name (str) – Function to find.
Returns:The callable implementing the function.
Return type:callable
register_instance(obj: object, prefix: str = '') → None

Create new subdispatcher and register all public object methods on it.

To be used in conjunction with the public() decorator (not RPCDispatcher.public()).

  • obj (object) – The object whose public methods should be made available.
  • prefix (str) – A prefix for the new subdispatcher.
dispatch(request: Union[tinyrpc.protocols.RPCRequest, tinyrpc.protocols.RPCBatchRequest], caller: Callable = None) → Union[tinyrpc.protocols.RPCResponse, tinyrpc.protocols.RPCBatchResponse]

Fully handle request.

The dispatch method determines which method to call, calls it and returns a response containing a result.

No exceptions will be thrown, rather, every exception will be turned into a response using error_respond().

If a method isn’t found, a MethodNotFoundError response will be returned. If any error occurs outside of the requested method, a ServerError without any error information will be returned.

If the method is found and called but throws an exception, the exception thrown is used as a response instead. This is the only case in which information from the exception is possibly propagated back to the client, as the exception is part of the requested method.

RPCBatchRequest instances are handled by handling all its children in order and collecting the results, then returning an RPCBatchResponse with the results.

  • request (RPCRequest or RPCBatchRequest) – The request containing the function to be called and its parameters.
  • caller (callable) – An optional callable used to invoke the method.

The result produced by calling the requested function.

Return type:

RPCResponse or RPCBatchResponse



The ServerError is raised for any kind of exception not raised by the called function itself or MethodNotFoundError.

static validate_parameters(method: Callable, args: List[Any], kwargs: Dict[str, Any]) → None

Verify that *args and **kwargs are appropriate parameters for method.


This function has changed to a static function. This will make it easier to replace it with a regular function instead of having to subclass only to replace it.

  • method – A callable.
  • args – List of positional arguments for method
  • kwargs – Keyword arguments for method

InvalidParamsError – Raised when the provided arguments are not acceptable for method.

static validator(method: Callable, args: List[Any], kwargs: Dict[str, Any]) → None

Dispatched function parameter validation.


By default this attribute is set to validate_parameters(). The value can be set to any callable implementing the same interface as validate_parameters() or to None to disable validation entirely.

Classes can be made to support an RPC interface without coupling it to a dispatcher using a decorator:


Decorator. Mark a method as eligible for registration by a dispatcher.

The dispatchers register_instance() function will do the actual registration of the marked method.

The difference with public() is that this decorator does not register with a dispatcher, therefore binding the marked methods with a dispatcher is delayed until runtime. It also becomes possible to bind with multiple dispatchers.

Parameters:name – The name to register the function with.


def class Baz(object):
    def not_exposed(self);
        # ...

    def visible_method(self, arg1):
        # ...

baz = Baz()
dispatch = RPCDispatcher()
dispatch.register_instance(baz, 'bazzies`)
# Baz.visible_method is now callable via RPC as bazzies.do_something('hello')

@public is a shortcut for @public().