Cannot inject into teardown_request handler while using request scope

[ollien]

If I have a request scoped variable, I cannot inject it into a teardown_request handler. If I do, I get the following exception.

Error on request:
Traceback (most recent call last):
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/werkzeug/local.py", line 72, in __getattr__
    return self.__storage__[self.__ident_func__()][name]
KeyError: 139781434668800

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/werkzeug/serving.py", line 304, in run_wsgi
    execute(self.server.app)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/werkzeug/serving.py", line 292, in execute
    application_iter = app(environ, start_response)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/flask/app.py", line 2463, in __call__
    return self.wsgi_app(environ, start_response)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/flask/app.py", line 2457, in wsgi_app
    ctx.auto_pop(error)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/flask/ctx.py", line 452, in auto_pop
    self.pop(exc)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/flask/ctx.py", line 415, in pop
    self.app.do_teardown_request(exc)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/flask/app.py", line 2299, in do_teardown_request
    func(exc)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/flask_injector.py", line 89, in wrapper
    return injector.call_with_injection(callable=fun, args=args, kwargs=kwargs)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/injector/__init__.py", line 972, in call_with_injection
    owner_key=self_.__class__ if self_ is not None else callable.__module__,
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/injector/__init__.py", line 96, in wrapper
    return function(*args, **kwargs)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/injector/__init__.py", line 1017, in args_to_inject
    instance = self.get(interface)  # type: Any
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/injector/__init__.py", line 915, in get
    result = scope_instance.get(interface, binding.provider).get(self)
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/flask_injector.py", line 265, in get
    return self._locals.scope[key]
  File "/home/nick/.local/share/virtualenvs/flask_injector_poc-sJYA9nHW/lib/python3.7/site-packages/werkzeug/local.py", line 74, in __getattr__
    raise AttributeError(name)
AttributeError: scope

After some digging, it seems that this is because Flask executes teardown handlers in the reverse order that they are inserted (see relevant source). Because flask_injector inserts the request scope clean up as a teardown handler, after all others are inserted, flask_injector's handler is run before mine.

I was able to hack around this by, after setting up my FlaskInjector running app.teardown_request_funcs[None].reverse(). This is not a great solution, IMO. I'm hesitant to submit a PR that just inserts this teardown handler as the first one, as the Flask docs don't seem to guarantee that the handlers are called in reverse order; it seems like an implementation detail.

What follows is an (admittedly silly) proof of concept that produces the above exception on request.

import datetime
import flask
import flask_injector
import injector

app = flask.Flask(__name__)


@app.route('/')
def index(time: datetime.datetime):
    return f'The time is {time.isoformat()}'


@app.teardown_request
def teardown_handler(err: Exception, time: datetime.datetime):
    print(f'Tearing down the request from {time.isoformat()}')


flask_injector.FlaskInjector(app=app, modules=[
    lambda binder: binder.bind(datetime.datetime, to=datetime.datetime.now(), scope=flask_injector.request)
])

if __name__ == '__main__':
    # HACK: works around the bug.
    # app.teardown_request_funcs[None].reverse()
    app.run()

[jstasiak]

Thanks for the report, this is something I think should be supported. I understand your hesitation, but I don't see any way to work around this other than to insert the teardown handler as the last one – with a proper test to make sure things keep working. I'll do that and make a release soon. The worst case scenario is this is changed in the future but I guess we'll have to cross that bridge once we reach it.

Permalink to the referenced Flask source code piece for historical visibility.

[ollien]

Well, inserted as the first one, but yes, same concept. :)

If I find the time in the next day or so, I can open a PR with these changes. Like you said, we can cross the bridge when we come to it.

[jstasiak]

Sorry, yes, the first one. I'll patch this today, but thanks anyway for wanting to help with it.

[jstasiak]

The fix has just been released in version 0.12.1.

[Chratho]

Using version 0.12.2 of the plugin I still see flask-injector's teardown handler executing before my apps teardown handler, making injections into these handlers fail. Using blueprints, fwiw.

[jstasiak]

What versions of Python and Flask are you using? Can you provide a piece of code to reproduce the incorrect behavior? I'm afraid the issue will be difficult to narrow down otherwise.

[Chratho]

Python is 3.7.4, Flask is 1.1.0. The exception prints exactly as the one posted by @ollien, but I can try to provide some simplified sample in the evening. Could also upgrade Python and Flask to the latest versions if you wish.

[jstasiak]

I think this reproduces the issue so no need for further investigation on your part:

import random
from flask import blueprint, flask
from flask_injector import flaskinjector, request

blueprint = blueprint('blueprint', __name__)


class x:
    def __init__(self, value):
        self.value = value


@blueprint.route('/')
def index(x: x):
    print(f'index, x.value: {x.value}')
    return 'ok'


@blueprint.teardown_request
def teardown_handler(err: exception, x: x):
    print(f'teardown_handler, x.value: {x.value}')


def configure(binder):
    binder.bind(x, to=lambda: x(random.randint(0, 1_000_000)), scope=request)


def create_app():
    app = flask(__name__)
    app.register_blueprint(blueprint)
    flaskinjector(app, modules=[configure])
    return app


if __name__ == '__main__':
    app = create_app()
    app.run()

[jstasiak]

Flask-Injector 0.12.3 I just released should fix this for you.

[Chratho]

Can confirm its working now. Many thanks for this amazingly fast fix!