Flask测试中preprocess_request等方法调用逻辑及相关疑问

Hey there! Let's break down your Flask testing questions and the core request handling flow clearly, since you're working with Python3.7, Flask==1.0.2, and pytest==4.3.1.

1. Why only call dispatch_request() if preprocess_request() returns None? #

This is all about Flask's request interception design. The preprocess_request() method runs all your registered before_request hooks in order. If any of those hooks returns a non-None value (like a string, tuple, or Response object), Flask treats that as a "short-circuit" signal—it means you want to return a response immediately without processing the actual request (think: authentication failures returning a 401, or redirects for unlogged users).

Only when every before_request hook returns None does Flask know it's safe to proceed to the actual request routing and view function execution, which is exactly what dispatch_request() handles.

2. Why call process_response() no matter if rv is None or not? #

process_response() is Flask's way of ensuring every response gets consistent post-processing, regardless of where it came from. Whether the response was generated by a before_request hook (the rv != None case) or your view function (the dispatch_request() case), this method takes care of:

• Running all after_request hooks (e.g., adding security headers, logging)
• Handling session cookies if the session was modified
• Normalizing the response into a proper WSGI-compatible format
• Setting default headers like Content-Type if they're missing

Skipping this step would leave you with an incomplete response that might not work correctly with your WSGI server.

3. Is my understanding wrong that rv has nothing to do with request dispatch? #

Yep, that's a common misconception! The rv returned by preprocess_request() is the return value of the last executed before_request hook, and it's a critical branching point in the request flow:

• If rv is non-None: Flask aborts the normal dispatch flow entirely and uses this value to build the response.
• If rv is None: Flask proceeds to route the request and run the matching view function via dispatch_request().

So rv directly controls whether your request ever reaches the view function—it's deeply tied to request dispatch.

Let's unpack what each of these methods actually does under the hood:

preprocess_request() #

This method executes all registered before_request hooks in a specific order: first global app-level hooks, then blueprint-level hooks in the order the blueprints were registered. It stops executing hooks as soon as one returns a non-None value, and returns that value. If all hooks return None, it returns None to signal the request should continue.

make_response(rv) #

Flask lets you return all kinds of values from views/hooks (strings, dicts, tuples, Response objects)—this method normalizes all those into a standard flask.Response object:

• Strings become responses with a 200 status code and text/html content type
• Dicts are automatically converted to JSON responses using jsonify()
• Tuples like (body, 404, {"X-Custom-Header": "value"}) are parsed into a Response with the given body, status code, and headers
• Existing Response objects are returned as-is

dispatch_request() #

This is Flask's routing workhorse. It looks up the current request's URL and HTTP method in the app's route map, finds the matching view function, and runs it. If no route matches, it returns a placeholder that gets converted to a 404 Response later. The return value is whatever the view function returns (which then gets passed to make_response()).

process_response(response) #

This is the final step in the request flow. It takes the raw Response object and polishes it for the WSGI server:

1. Runs all after_request hooks (each can modify the Response)
2. Updates session cookies if the session was changed during the request
3. Ensures required headers are set (e.g., Content-Length)
4. Converts the Response into a WSGI-compatible tuple: (status_code, headers_list, response_body_generator)

内容的提问来源于stack exchange，提问作者Jerry