Http
Handlers¶
In the following text, the callable object used to handle HTTP requests is referred to as an HTTP processor.
Function Handlers¶
Using a function to handle requests is straightforward.
The @app.router.http
decorator returns the original function, so the same function can be registered under multiple routes.
from kui.asgi import Kui, request
app = Kui()
@app.router.http("/hello", name="hello")
@app.router.http("/hello/{name}", name="hello-with-name")
async def hello():
if request.path_params:
return f"hello {request.path_params['name']}"
return "hello"
You can also use required_method
to restrict the function processor to accept only specific request methods.
from kui.asgi import Kui, request, required_method
app = Kui()
@app.router.http("/hello", middlewares=[required_method("POST")])
async def need_post():
return request.method
Notice
When you use required_method
to constrain the request method, the OPTIONS
method will be automatically handled.
Notice
When you use required_method
to allow the GET
method, the HEAD
method will also be allowed.
Class Handlers¶
Using a class to handle multiple types of requests is simple. Just inherit from HttpView
and write the corresponding methods. The supported methods are "get"
, "post"
, "put"
, "patch"
, "delete"
, "head"
, "options"
, and "trace"
.
Allow more request methods
Override the class attribute HTTP_METHOD_NAMES
when inheriting the class.
from kui.asgi import Kui, request, HttpView
app = Kui()
@app.router.http("/cat")
class Cat(HttpView):
@classmethod
async def get(cls):
return request.method
@classmethod
async def post(cls):
return request.method
@classmethod
async def put(cls):
return request.method
@classmethod
async def patch(cls):
return request.method
@classmethod
async def delete(cls):
return request.method
Accessing Request Values¶
Use the following statement to access the global variable request
, which is a proxy object that allows reading, writing, and deleting various attributes of the HttpRequest
object corresponding to the current request.
In general, this is sufficient for most use cases. However, if you really need to access the original HttpRequest
object, you can use
Here are the commonly used attributes and methods of the kui.wsgi.HttpRequest
object.
Method¶
You can obtain the request method using request.method
, for example, GET
or POST
.
URL¶
You can obtain the request path using request.url
. This attribute is an object similar to a string and exposes all the components that can be parsed from the URL.
For example: request.url.path
, request.url.port
, request.url.scheme
Path Parameters¶
request.path_params
is a dictionary that contains all parsed path parameters.
Headers¶
request.headers
is a case-insensitive multi-value dictionary.
The keys obtained by calling request.headers.keys()
/request.headers.items()
are all lowercase.
Accept¶
You can retrieve all accepted response types from the client by reading the request.accepted_types
attribute.
You can determine what response types the client accepts by calling the request.accepts
function. For example: request.accepts("text/html")
.
Content Type¶
Use request.content_type
to retrieve the Content-Type
header.
Content Length¶
Use request.content_length
to retrieve the Content-Length
header.
Date¶
Use request.date
to retrieve the Date
header.
Referrer¶
Use request.referrer
to retrieve the Referer
header.
Query Parameters¶
request.query_params
is a multi-value dictionary.
For example: request.query_params['search']
Client Address¶
request.client
is a namedtuple
defined as namedtuple("Address", ["host", "port"])
.
To retrieve the client's hostname or IP address: request.client.host
.
To retrieve the port the client is using in the current connection: request.client.port
.
Notice
Any element in the tuple may be None. This depends on the values passed by the server.
Cookies¶
request.cookies
is a standard dictionary defined as Dict[str, str]
.
For example: request.cookies.get('mycookie')
Body¶
There are several ways to read the request body:
-
await request.body
: Returns abytes
object. -
await request.form
: Parses the body as a form and returns the result as a multi-value dictionary. -
await request.json
: Parses the body as a JSON string and returns the result. -
await request.data()
: Parses the body based on the information provided bycontent_type
and returns the result.
You can also use the for
syntax to read the body as a bytes
stream:
If you use request.stream()
directly to read the data, the request body will not be cached in memory.
Request Files¶
You can parse forms received in the multipart/form-data
format, including files, using await request.form
.
The file will be wrapped in an baize.datastructures.UploadFile
object, which has the following attributes:
filename: str
: The original filename of the submitted file (e.g.,myimage.jpg
).content_type: str
: The file type (MIME type / media type) (e.g.,image/jpeg
).headers: Headers
: Headers information carried by the file field in themultipart/form-data
format.file: tempfile.SpooledTemporaryFile
: A temporary file that stores the content of the file (you can read and write to this object directly, but it's better not to).
UploadFile
also has five methods:
awrite(data: bytes) -> None
: Write data to the file.aread(size: int) -> bytes
: Read data from the file.aseek(offset: int) -> None
: Move the file pointer to the specified position.asave(filepath: str) -> None
: Save the file to the specified path on disk.aclose() -> None
: Close the file.
State¶
In some cases, you may need to store additional custom information in the request
. You can use request.state
for storage.
request.state.user = User(name="Alice") # Write
user_name = request.state.user.name # Read
del request.state.user # Delete
Return Response¶
For any properly handled HTTP request, you must return an HttpResponse
object or an object of its subclass.
HttpResponse¶
Signature: HttpResponse(status_code: int = 200, headers: Mapping[str, str] = None)
status_code
- HTTP status code.headers
- String dictionary.
Set Cookie¶
HttpResponse
provides the set_cookie
method to allow you to set cookies.
Signature: HttpResponse.set_cookie(key, value="", max_age=None, expires=None, path="/", domain=None, secure=False, httponly=False, samesite="lax")
key: str
- The key that will become the cookie.value: str = ""
- The value of the cookie.max_age: int
- The lifetime of the cookie in seconds. Non-positive integers will discard the cookie immediately.expires: Optional[int]
- The number of seconds before the cookie expires.path: str = "/"
- Specifies the subset of routes to which the cookie will apply.domain: Optional[str]
- Used to specify the domain for which the cookie is valid.secure: bool = False
- Indicates that the cookie will only be sent to the server when using the HTTPS protocol.httponly: bool = False
- Indicates that the cookie cannot be accessed via JavaScript through properties likeDocument.cookie
,XMLHttpRequest
, orRequest
API.samesite: str = "lax"
- Specifies the same-site policy for the cookie. Valid values are"lax"
,"strict"
, and"none"
.
Delete Cookie¶
HttpResponse
also provides the delete_cookie
method to specify the expiration of a previously set cookie.
Signature: HttpResponse.delete_cookie(key, path='/', domain=None, secure=False, httponly=False, samesite="lax")
PlainTextResponse¶
Accepts a str
or bytes
and returns a plain text response.
from kui.asgi import PlainTextResponse
async def return_plaintext():
return PlainTextResponse('Hello, world!')
HTMLResponse¶
Accepts a str
or bytes
and returns an HTML response.
from kui.asgi import HTMLResponse
async def return_html():
return HTMLResponse('<html><body><h1>Hello, world!</h1></body></html>')
JSONResponse¶
Accepts a Python object and returns a response encoded as application/json
.
JSONResponse
exposes all options of json.dumps
as keyword arguments for customization.
RedirectResponse¶
Returns an HTTP redirect. By default, uses the 307 status code.
StreamResponse¶
Accepts a generator and streams the response body.
import asyncio
from kui.asgi import StreamResponse
async def slow_numbers(minimum, maximum):
yield b'<html><body><ul>'
for number in range(minimum, maximum + 1):
yield f'<li>{number}</li>'.encode()
await asyncio.sleep(0.5)
yield b'</ul></body></html>'
async def return_stream(scope, receive, send):
generator = slow_numbers(1, 10)
return StreamResponse(generator, content_type='text/html')
FileResponse¶
Transfers a file as a response.
Compared to other response types, it is instantiated with different parameters:
filepath
- The file path of the file to be streamed.headers
- Same as theheaders
parameter inResponse
.content_type
- The MIME media type of the file. If not set, the file name or path will be used to infer the media type.download_name
- If set, it will be included in theContent-Disposition
of the response.stat_result
- Accepts anos.stat_result
object. If not passed, it will automatically use the result ofos.stat(filepath)
.
FileResponse
automatically sets the appropriate Content-Length
, Last-Modified
, and ETag
headers. It also supports file range requests without any additional handling.
TemplateResponse¶
TemplateResponse
is a shortcut for app.templates.TemplateResponse
.
Jinja2 Template Engine¶
Kui has built-in support for Jinja2 templates. As long as you have the jinja2
module installed, you can export Jinja2Templates
from kui.wsgi.templates
. Here's a simple example: when accessing "/", it will render the homepage.html
file located in the templates
directory under the project root.
from kui.asgi import Kui, TemplateResponse, Jinja2Templates
app = Kui(templates=Jinja2Templates("templates"))
@app.router.http("/")
async def homepage():
return TemplateResponse("homepage.html")
If you want to use templates from a specific folder under a module, you can use Jinja2Templates("module_name:dirname")
.
You can also pass multiple directories to Jinja2 to search in order until it finds the first available template, for example: Jinja2Templates("templates", "module_name:dirname")
.
Other Template Engines¶
Implement the kui.wsgi.templates.BaseTemplates
interface to create your own template engine class.
SendEventResponse¶
With SendEventResponse
, you can return a Server-sent Events response, which is a type of HTTP long-polling response used for scenarios such as server-side real-time data pushing to clients.
In addition to accepting common parameters such as status_code
and headers
, SendEventResponse
requires a generator to be passed in for generating messages. Each message yielded by the generator should be a valid Server-Sent Events message.
Here's an example that sends a "hello" message every second, a total of 101 messages:
import asyncio
from typing import AsyncGenerator
from kui.asgi import Kui, SendEventResponse, ServerSentEvent
app = Kui()
@app.router.http("/message")
async def message():
async def message_gen() -> AsyncGenerator[ServerSentEvent, None]:
for i in range(101):
await asyncio.sleep(1)
yield {"id": i, "data": "hello"}
return SendEventResponse(message_gen())
Front-end Development with Browsers
In most cases, using the browser's built-in EventSource is sufficient for your needs. However, in more complex scenarios where you may need to use Server-sent events (such as the ChatGPT interface provided by OpenAI), you can use @microsoft/fetch-event-source to achieve more advanced functionality.
Simplified Writing of Responses¶
For ease of use, Kuí allows you to customize functions to handle non-HttpResponse
objects returned by HTTP handlers. The principle is to intercept the response and automatically select a processing function based on the type of the response value, converting non-HttpResponse
objects into HttpResponse
objects.
Active Conversion
To convert the return value of a function into an HttpResponse
object, you can use kui.wsgi.convert_response
.
In the following example, the view function returns a dict
object, but the client receives a JSON response. This is because Kuí provides built-in processing functions for common types:
dict | tuple | list
: Automatically converted toJSONResponse
str | bytes
: Automatically converted toPlainTextResponse
types.GeneratorType
: Automatically converted toSendEventResponse
pathlib.PurePath
: Automatically converted toFileResponse
baize.datastructures.URL
: Automatically converted toRedirectResponse
You can also return multiple values to customize the HTTP status and headers:
async def not_found():
return {"message": "Not found"}, 404
async def no_content():
return "", 301, {"location": "https://kui.aber.sh"}
Similarly, you can customize the simplified representation of response values to standardize the response format of your project (even though Python's Dict
constraint is weak, TypedDict
is more effective, but dataclass
is even better). In the following example, when you return an Error
object in a view function, it will be automatically converted to a JSONResponse
, with the default status code set to 400
:
from dataclasses import dataclass, asdict
from typing import Mapping
from kui.asgi import Kui, HttpResponse, JSONResponse
app = Kui()
@dataclass
class Error:
code: int = 0
title: str = ""
message: str = ""
@app.response_convertor.register(Error)
def _error_json(error: Error, status: int = 400, headers: Mapping[str, str] = None) -> HttpResponse:
return JSONResponse(asdict(error), status, headers)
This is equivalent to:
from dataclasses import dataclass, asdict
from typing import Mapping
from kui.asgi import Kui, HttpResponse, JSONResponse
@dataclass
class Error:
code: int = 0
title: str = ""
message: str = ""
def _error_json(error: Error, status: int = 400, headers: Mapping[str, str] = None) -> HttpResponse:
return JSONResponse(asdict(error), status, headers)
app = Kui(
response_converters={
Error: _error_json
}
)
You can also override the default conversion method, as shown in the following example.
from typing import Mapping
from kui.asgi import Kui, HttpResponse
app = Kui()
@app.response_convertor.register(tuple)
@app.response_convertor.register(list)
@app.response_convertor.register(dict)
def _more_json(body, status: int = 200, headers: Mapping[str, str] = None) -> HttpResponse:
return CustomizeJSONResponse(body, status, headers)
Exception Handling¶
HTTPException¶
The signature of this class is: HTTPException(status_code: int, headers: dict = None, content: typing.Any = None)
.
You can raise an HTTPException
to return an HTTP response (don't worry, it won't be treated as a real exception; Kuí will convert it into a regular response object). If you don't provide a content
value, it will use http.HTTPStatus(status_code).description
from the Python standard library as the final result.
Sometimes you may want to return more information. You can pass content
and headers
parameters to the HTTPException
just like you would with HttpResponse
to control the actual response object. Here's a simple example:
from kui import HTTPException
async def endpoint():
...
raise HTTPException(405, headers={"Allow": "HEAD, GET, POST"})
...
Tip
If you want to raise an HTTPException
in a lambda
function, you can use baize.exceptions.abort
.
Custom Exception Handling¶
For intentionally raised exceptions, Kuí provides a method for unified handling.
You can catch specific HTTP status codes, so when an HTTPException
with the corresponding status code occurs, Kuí will use your defined function instead of the default behavior. You can also catch other exceptions that inherit from Exception
and return specific content to the client through a custom function.
from kui.asgi import Kui, HTTPException, HttpResponse, PlainTextResponse
app = Kui()
@app.exception_handler(404)
async def not_found(exc: HTTPException) -> HttpResponse:
return PlainTextResponse("what do you want to do?", status_code=404)
@app.exception_handler(ValueError)
async def value_error(exc: ValueError) -> HttpResponse:
return PlainTextResponse("Something went wrong with the server.", status_code=500)
In addition to decorator registration, you can also use a list-based registration approach. The following example is equivalent to the previous one:
from kui.asgi import Kui, HTTPException, HttpResponse, PlainTextResponse
async def not_found(exc: HTTPException) -> HttpResponse:
return PlainTextResponse("what do you want to do?", status_code=404)
async def value_error(exc: ValueError) -> HttpResponse:
return PlainTextResponse("Something went wrong with the server.", status_code=500)
app = Kui(exception_handlers={
404: not_found,
ValueError: value_error,
})
Allowing Cross-Origin Requests¶
To solve the cross-origin issue in modern browsers, Cross-Origin Resource Sharing (CORS) is generally used. In Kuí, you can quickly configure API to allow cross-origin requests using the following code:
The allow_cors
function accepts the following parameters:
allow_origins: Iterable[Pattern]
: Allowed origins. It requires pre-compiledPattern
objects usingre.compile
. The default value is(re.compile(".*"), )
.allow_methods: Iterable[str]
: Allowed request methods. The default value is("GET", "POST", "PUT", "PATCH", "DELETE", "HEAD", "OPTIONS", "TRACE")
.allow_headers: Iterable[str]
: Allowed request headers. Corresponds toAccess-Control-Allow-Headers
.expose_headers: Iterable[str]
: Request headers that can be listed in the response. Corresponds toAccess-Control-Expose-Headers
.allow_credentials: bool
: IfTrue
, allows cross-origin requests to carry cookies; otherwise, it is not allowed. The default value isFalse
.max_age: int
: Cache time for preflight requests. The default value is600
seconds.
If you need to enable CORS globally, you can pass the cors_config
parameter to Kui
. It is a dictionary with the same key-value pairs as the allow_cors
parameters.