FastAPI 请求上下文及应用上下文

在fastapi的实际应用中发现，fastapi没有提供像flask中的g这样的上下文相关的东西。所以在查找了多方的资料发现，可以使用中间件，再加上Python3.7中的新特性contextvar可以实现应用上下文的管理。
contextvar可自行到官网进行学习。
这里着重讲解一下fastapi中怎么应用。
fastapi中的中间件，可实现对每次请求的request对象进行操作，然后执行请求内容，最后也可以对返回response进行处理。middleware用法参考我之前的文章。

fastapi基于starlette，所以starlette中的中间键可直接用于fastapi。
我在starlette的中间件中找到，starlette_context这个开源项目，可以直接拿来使用，但是它内置的请求只处理了几个头部信息，并不能完全符合项目要求。于是对其进行改造。这里具体讲解一下这个开源项目的原理过程，方便进行与自身需求符合的改进。
starlette_context
starlette_context目录结构：

starlette_context
	plugins
		__init__.py
		plugin.py
		.
		.
		.
	__init__.py
	ctx.py
	header_keys.py
	middleware.py

其中__init__.py
直接引入了contextvar,并对其进行初始化

from contextvars import ContextVar

_request_scope_context_storage: ContextVar[str] = ContextVar(
    "starlette_context"
)

from starlette_context.ctx import context  # noqa: E402, F401

其中ctx.py,主要对contextvar对象进行管理，并实现了一个data函数属性，以获得全部值。

from collections import UserDict
from typing import Any

from contextvars import copy_context
from libs.fd_context import _request_scope_context_storage


class _Context(UserDict):
    """
    A mapping with dict-like interface.
    It is using request context as a data store.
    Can be used only if context has been created in the middleware.

    If you know Flask, it can be compared to g object.
    """

    def __init__(self, *args: Any, **kwargs: Any):
        # not calling super on purpose
        if args or kwargs:
            raise AttributeError("Can't instantiate with attributes")

    @property
    def data(self) -> dict:
        """
        Dump this to json. Object itself it not serializable.
        """
        try:
            return _request_scope_context_storage.get()
        except LookupError as e:
            raise RuntimeError(
                "You didn't use ContextMiddleware or "
                "you're trying to access `context` object "
                "outside of the request-response cycle."
            ) from e

    def exists(self) -> bool:
        return _request_scope_context_storage in copy_context()

    def copy(self) -> dict:
        """
        Read only context data.
        """
        import copy

        return copy.copy(self.data)


context = _Context()

data属性主要通过contextvar实例的get方法获取数据。
header_keys.py，定义需要获取的信息字段。

class HeaderKeys:
    correlation_id = "X-Correlation-ID"
    request_id = "X-Request-ID"
    date = "Date"
    forwarded_for = "X-Forwarded-For"
    user_agent = "User-Agent"

middleware.py主要是继承自BaseHTTPMiddleware。初始化时传入plugins列表。实现set_context方法，功能是遍历plugins，根据request请求获取值并返回header_keys中定义的键对应的值。dispatch方法是中间件必须实现的方法，并必须调用call_next

from contextvars import Token
from typing import Optional, Sequence

from starlette.middleware.base import (
    BaseHTTPMiddleware,
    RequestResponseEndpoint,
)
from starlette.middleware import Middleware
from starlette.requests import Request
from starlette.responses import Response

from libs.fd_context import _request_scope_context_storage
from libs.fd_context.plugins import Plugin
from libs.fd_context import plugins


class ContextMiddleware(BaseHTTPMiddleware):
    """
    Middleware that creates empty context for request it's used on.
    If not used, you won't be able to use context object.
    """

    def __init__(
        self, plugins: Optional[Sequence[Plugin]] = None, *args, **kwargs
    ) -> None:
        super().__init__(*args, **kwargs)
        self.plugins = plugins or ()
        if not all([isinstance(plugin, Plugin) for plugin in self.plugins]):
            raise TypeError("This is not a valid instance of a plugin")

    async def set_context(self, request: Request) -> dict:
        """
        You might want to override this method.
        The dict it returns will be saved in the scope of a context.
        You can always do that later.
        """
        return {
            plugin.key: await plugin.process_request(request)
            for plugin in self.plugins
        }

    async def dispatch(
        self, request: Request, call_next: RequestResponseEndpoint
    ) -> Response:
        _starlette_context_token: Token = _request_scope_context_storage.set(
            await self.set_context(request)
        )
        try:
            response = await call_next(request)
            for plugin in self.plugins:
                await plugin.enrich_response(response)

        finally:
            _request_scope_context_storage.reset(_starlette_context_token)

        return response

plugins中主要的文件就是plugin.py。
Plugin类是一个元类，其中
定义了key变量用于存放header_key中的值，value变量用于存放获取到的值。
可实现方法从request中获取到头部信息，以及starlette中对requests中的所有解析信息都可以自己实现相应的方法。