The Request Context

The request context keeps track of the request-level data during a request. Rather than passing the request object to each function that runs during a request, the request and session proxies are accessed instead.

This is similar to the The Application Context, which keeps track of the application-level data independent of a request. A corresponding application context is pushed when a request context is pushed.

Purpose of the Context

When the Flask application handles a request, it creates a Request object based on the environment it received from the WSGI server. Because a worker (thread, process, or coroutine depending on the server) handles only one request at a time, the request data can be considered global to that worker during that request. Flask uses the term context local for this.

Flask automatically pushes a request context when handling a request. View functions, error handlers, and other functions that run during a request will have access to the request proxy, which points to the request object for the current request.

Lifetime of the Context

When a Flask application begins handling a request, it pushes a request context, which also pushes an The Application Context. When the request ends it pops the request context then the application context.

The context is unique to each thread (or other worker type). request cannot be passed to another thread, the other thread will have a different context stack and will not know about the request the parent thread was pointing to.

Context locals are implemented in Werkzeug. See Context Locals for more information on how this works internally.

Manually Push a Context

If you try to access request, or anything that uses it, outside a request context, you’ll get this error message:

RuntimeError: Working outside of request context.

This typically means that you attempted to use functionality that
needed an active HTTP request. Consult the documentation on testing
for information about how to avoid this problem.

This should typically only happen when testing code that expects an active request. One option is to use the test client to simulate a full request. Or you can use test_request_context() in a with block, and everything that runs in the block will have access to request, populated with your test data.

def generate_report(year):
    format = request.args.get('format')

with app.test_request_context(
        '/make_report/2017', data={'format': 'short'}):

If you see that error somewhere else in your code not related to testing, it most likely indicates that you should move that code into a view function.

For information on how to use the request context from the interactive Python shell, see Working with the Shell.

How the Context Works

The Flask.wsgi_app() method is called to handle each request. It manages the contexts during the request. Internally, the request and application contexts work as stacks, _request_ctx_stack and _app_ctx_stack. When contexts are pushed onto the stack, the proxies that depend on them are available and point at information from the top context on the stack.

When the request starts, a RequestContext is created and pushed, which creates and pushes an AppContext first if a context for that application is not already the top context. While these contexts are pushed, the current_app, g, request, and session proxies are available to the original thread handling the request.

Because the contexts are stacks, other contexts may be pushed to change the proxies during a request. While this is not a common pattern, it can be used in advanced applications to, for example, do internal redirects or chain different applications together.

After the request is dispatched and a response is generated and sent, the request context is popped, which then pops the application context. Immediately before they are popped, the teardown_request() and teardown_appcontext() functions are are executed. These execute even if an unhandled exception occurred during dispatch.

Callbacks and Errors

Flask dispatches a request in multiple stages which can affect the request, response, and how errors are handled. The contexts are active during all of these stages.

A Blueprint can add handlers for these events that are specific to the blueprint. The handlers for a blueprint will run if the blueprint owns the route that matches the request.

  1. Before each request, before_request() functions are called. If one of these functions return a value, the other functions are skipped. The return value is treated as the response and the view function is not called.
  2. If the before_request() functions did not return a response, the view function for the matched route is called and returns a response.
  3. The return value of the view is converted into an actual response object and passed to the after_request() functions. Each function returns a modified or new response object.
  4. After the response is returned, the contexts are popped, which calls the teardown_request() and teardown_appcontext() functions. These functions are called even if an unhandled exception was raised at any point above.

If an exception is raised before the teardown functions, Flask tries to match it with an errorhandler() function to handle the exception and return a response. If no error handler is found, or the handler itself raises an exception, Flask returns a generic 500 Internal Server Error response. The teardown functions are still called, and are passed the exception object.

If debug mode is enabled, unhandled exceptions are not converted to a 500 response and instead are propagated to the WSGI server. This allows the development server to present the interactive debugger with the traceback.

Teardown Callbacks

The teardown callbacks are independent of the request dispatch, and are instead called by the contexts when they are popped. The functions are called even if there is an unhandled exception during dispatch, and for manually pushed contexts. This means there is no guarantee that any other parts of the request dispatch have run first. Be sure to write these functions in a way that does not depend on other callbacks and will not fail.

During testing, it can be useful to defer popping the contexts after the request ends, so that their data can be accessed in the test function. Using the test_client() as a with block to preserve the contexts until the with block exits.

from flask import Flask, request

app = Flask(__name__)

def hello():
    print('during view')
    return 'Hello, World!'

def show_teardown(exception):
    print('after with block')

with app.test_request_context():
    print('during with block')

# teardown functions are called after the context with block exits

with app.test_client():
    # the contexts are not popped even though the request ended

# the contexts are popped and teardown functions are called after
# the client with block exists


If signals_available is true, the following signals are sent:

  1. request_started is sent before the before_request() functions are called.
  2. request_finished is sent after the after_request() functions are called.
  3. got_request_exception is sent when an exception begins to be handled, but before an errorhandler() is looked up or called.
  4. request_tearing_down is sent after the teardown_request() functions are called.

Context Preservation on Error

At the end of a request, the request context is popped and all data associated with it is destroyed. If an error occurs during development, it is useful to delay destroying the data for debugging purposes.

When the development server is running in development mode (the FLASK_ENV environment variable is set to 'development'), the error and data will be preserved and shown in the interactive debugger.

This behavior can be controlled with the PRESERVE_CONTEXT_ON_EXCEPTION config. As described above, it defaults to True in the development environment.

Do not enable PRESERVE_CONTEXT_ON_EXCEPTION in production, as it will cause your application to leak memory on exceptions.

Notes On Proxies

Some of the objects provided by Flask are proxies to other objects. The proxies are accessed in the same way for each worker thread, but point to the unique object bound to each worker behind the scenes as described on this page.

Most of the time you don’t have to care about that, but there are some exceptions where it is good to know that this object is an actual proxy:

  • The proxy objects cannot fake their type as the actual object types. If you want to perform instance checks, you have to do that on the object being proxied.
  • If the specific object reference is important, for example for sending Signals or passing data to a background thread.

If you need to access the underlying object that is proxied, use the _get_current_object() method:

app = current_app._get_current_object()