from multiprocessing.managers import Valuefrom starlette.middleware import Middlewarefrom urllib.request import Request

Starlette Dispatch

Routing extensions and dependency injection library for Starlette.

Installation

Install starlette_dispatch using PIP:

pip install starlette_dispatch

Features

• Route groups. Group routes by common path prefix and common middleware.
• Route method decorators. Convenient decorators for common HTTP methods.
• Dependency injection. Route handlers can request dependencies by adding a parameter with the dependency type hint.
• Backward compatible with Starlette. You can use it with your existing Starlette application.
• No performance overhead. Dependency injection takes exact the same time as if you would write the handler manually.
• Fully typed. Starlette Dispatch is fully typed and supports type hints.
• Async support. Starlette Dispatch supports async handlers and async dependencies.

And the most important -- it does not erase route handler signature. You can compose it with any other decorators.

Quick start

Starlette Dispatch does not require any changes to your existing Starlette application. You can use it with your existing Starlette application.

Here is a simple snippet that demonstrates dependency injection and route group usage:

import typing

from starlette.applications import Starlette
from starlette.authentication import SimpleUser
from starlette.middleware import Middleware
from starlette.middleware.authentication import AuthenticationMiddleware
from starlette.requests import Request
from starlette.responses import JSONResponse
from starlette.routing import Route

from starlette_dispatch import RouteGroup

admin_middleware = [
    Middleware(AuthenticationMiddleware, backend=...)
]
admin_routes = RouteGroup('/admin', middleware=admin_middleware)

CurrentUser = typing.Annotated[SimpleUser, lambda r: r.user]


async def index_view(request: Request, user: CurrentUser) -> JSONResponse:
    """This is your landing page."""
    return JSONResponse({'message': f'Hello, {user}!'})


@admin_routes.get('/')
async def admin_index_view(request: Request) -> JSONResponse:
    """This is your admin landing page."""
    return JSONResponse({'message': 'Hello, admin!'})


app = Starlette(
    routes=[
        Route('/', index_view),  # regular Starlette route
        *admin_routes,
    ]
)

Route groups

A route group is a way to group routes by common path prefix and common middleware. Instead of writing the same prefix for each route, you can define a group and add routes to it using convenient decorators.

Route groups support all common HTTP methods and add some extra helpers like get_or_post.

from starlette.requests import Request
from starlette.responses import JSONResponse

from starlette_dispatch import RouteGroup

group = RouteGroup('/group')


@group.get('/')
def my_view(request: Request) -> JSONResponse:
    return JSONResponse({'message': 'Hello, world!'})


@group.get_or_post('/')
def form_view(request: Request) -> JSONResponse:
    return JSONResponse({'message': 'Hello, world!'})

Multiple routes on a single handler

You can call route decorators multiple times on a single handler. This way you can share the same handler for multiple routes without creating a new handler.

from starlette.requests import Request
from starlette.responses import JSONResponse

from starlette_dispatch import RouteGroup, FromPath

group = RouteGroup('/group')


@group.get('/new')
@group.get('/edit/{id}')
def create_view(request: Request, id: FromPath[int | None]) -> JSONResponse:
    ...

Route injections

Each route handler can request a dependency by adding a parameter with the dependency type hint. The dependency will be properly resolved and injected into the handler on handler call. See more about dependency injection below.

For each injection Starlette Dispatch creates a resolver function. This means, it does not add a noticeable overhead to your application and takes exact the same time as if you would write the handler manually.

import typing
from starlette.requests import Request
from starlette.responses import JSONResponse

from starlette_dispatch import RouteGroup


class User: ...


user = User()
CurrentUser = typing.Annotated[str, user]

group = RouteGroup('/')


@group.get('/')
def index_view(request: Request, user: CurrentUser) -> JSONResponse:
    return JSONResponse({'message': f'Hello, {user}!'})

Route middleware

Each route can have its own middleware.

If route group has middleware, it will be merged with route middleware. Route middleware has a higher priority.

import typing

from starlette.requests import Request
from starlette.responses import JSONResponse

from starlette_dispatch import RouteGroup


class User: ...


user = User()
CurrentUser = typing.Annotated[str, user]

group = RouteGroup('/')


@group.get('/')
def my_view(request: Request, user: CurrentUser) -> JSONResponse:
    return JSONResponse({'message': f'Hello, {user}!'})

Dependency injection

In a nutshell, the dependency is a type, annotated with a value or a factory function that resolves to the value. The factory function is called dependency resolver.

Variable dependency

Variable dependency is a resolver that returns a simple value.

import typing

from starlette_dispatch import RouteGroup, VariableDependency

Value = typing.Annotated[str, VariableDependency('hello')]

group = RouteGroup('/')


@group.get('/')
def my_view(value: Value) -> None:
    assert value == 'hello'

Factory dependency

Factory dependency is a resolver that creates a value on each call. The result can be cached. The factory can have dependencies and can be async.

import typing

from starlette_dispatch import FactoryDependency, RouteGroup


def make_dependency():
    return 'hello'


async def async_dependency():
    return 'hello'


Value = typing.Annotated[str, FactoryDependency(make_dependency)]
AsyncValue = typing.Annotated[str, FactoryDependency(async_dependency)]
CachedValue = typing.Annotated[str, FactoryDependency(make_dependency, cached=True)]

group = RouteGroup('/')


@group.get('/')
def my_view(value: Value, async_value: AsyncValue, cached_value: CachedValue) -> None:
    assert value == 'hello'
    assert async_value == 'hello'
    assert cached_value == 'hello'

Factory function dependencies

The factory function itself can have dependencies. They are defined in the same way as regular dependencies.

import typing

from starlette_dispatch import FactoryDependency, RouteGroup


def parent_dependency():
    return 'hello'


ParentValue = typing.Annotated[str, FactoryDependency(parent_dependency)]


def make_dependency(parent: ParentValue):
    return parent + ' world'


Value = typing.Annotated[str, FactoryDependency(make_dependency)]

group = RouteGroup('/')


@group.get('/')
def my_view(value: Value) -> None:
    assert value == 'hello world'

Predefined dependencies

There are several predefined dependencies: starlette.requests.Request, starlette_dispatch.injections.DependencySpec.

Request is a Starlette request object and DependencySpec is a special object that contains meta information about the dependency. DependencySpec object is very useful in complex cases.

from starlette.requests import Request

from starlette_dispatch import DependencySpec


def make_dependency(request: Request, spec: DependencySpec):
    assert request  # Starlette request object
    assert spec.param_name  # name of the parameter
    assert spec.param_type  # type of the parameter
    assert spec.optional  # is the parameter optional
    assert spec.default  # default value of the parameter
    assert spec.annotation  # type annotation of the parameter

Request resolver

If your dependency available in the request object, instead of creating a factory function, you can use a RequestDependency resolver. It takes a function that accepts Request and DependencySpec (optionally) objects.

import typing
from starlette_dispatch import RequestDependency

# example dependency that resolves to a value from query parameter
Value = typing.Annotated[str, RequestDependency(lambda request, spec,: request.query_params['value'])]
NoSpecValue = typing.Annotated[str, RequestDependency(lambda request: request.query_params['value'])]

Custom resolver

You are not limited to predefined resolvers. You can create your own resolver by subclassing DependencyResolver and implementing the resolve method.

from starlette.requests import Request

from starlette_dispatch import DependencyResolver, DependencySpec


class MyResolver(DependencyResolver):
    async def resolve(self, request: Request, spec: DependencySpec):
        """Use request and spec objects to create a value."""
        return None

Contrib and support

FromPath - inject path parameter as a dependency

from starlette_dispatch import FromPath, RouteGroup

group = RouteGroup('/')


@group.get('/{value}')
def my_view(value: FromPath[str]) -> None:
    assert value is not None

If path value does not exist in Request.path_parameters then it will fail with error. However, you can mark dependency as optional and then it will be None if path value does not exist.

from starlette_dispatch import FromPath, RouteGroup

group = RouteGroup('/')


def my_view(value: FromPath[str] | None) -> None:
    assert value is None