asyncio

asyncio

New in version 2.0.

Scrapy has partial support for asyncio. After you install the asyncio reactor, you may use asyncio and asyncio-powered libraries in any coroutine.

Installing the asyncio reactor

To enable asyncio support, set the TWISTED_REACTOR setting to 'twisted.internet.asyncioreactor.AsyncioSelectorReactor'.

If you are using CrawlerRunner, you also need to install the AsyncioSelectorReactor reactor manually. You can do that using install_reactor():

install_reactor('twisted.internet.asyncioreactor.AsyncioSelectorReactor')

Using custom asyncio loops

You can also use custom asyncio event loops with the asyncio reactor. Set the ASYNCIO_EVENT_LOOP setting to the import path of the desired event loop class to use it instead of the default asyncio event loop.

Windows-specific notes

The Windows implementation of asyncio can use two event loop implementations:

SelectorEventLoop, default before Python 3.8, required when using Twisted.
ProactorEventLoop, default since Python 3.8, cannot work with Twisted.

So on Python 3.8+ the event loop class needs to be changed.

Changed in version 2.6.0: The event loop class is changed automatically when you change the TWISTED_REACTOR setting or call install_reactor().

To change the event loop class manually, call the following code before installing the reactor:

import asyncio
asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())

You can put this in the same function that installs the reactor, if you do that yourself, or in some code that runs before the reactor is installed, e.g. settings.py.

Note

Other libraries you use may require ProactorEventLoop, e.g. because it supports subprocesses (this is the case with playwright), so you cannot use them together with Scrapy on Windows (but you should be able to use them on WSL or native Linux).

Awaiting on Deferreds

When the asyncio reactor isn’t installed, you can await on Deferreds in the coroutines directly. When it is installed, this is not possible anymore, due to specifics of the Scrapy coroutine integration (the coroutines are wrapped into asyncio.Future objects, not into Deferred directly), and you need to wrap them into Futures. Scrapy provides two helpers for this:

scrapy.utils.defer.deferred_to_future(d: Deferred) → Future[source]

New in version 2.6.0.

Return an asyncio.Future object that wraps d.

When using the asyncio reactor, you cannot await on Deferred objects from Scrapy callables defined as coroutines, you can only await on Future objects. Wrapping Deferred objects into Future objects allows you to wait on them:

class MySpider(Spider):
    ...
    async def parse(self, response):
        d = treq.get('https://example.com/additional')
        additional_response = await deferred_to_future(d)

scrapy.utils.defer.maybe_deferred_to_future(d: Deferred) → Union[Deferred, Future][source]

New in version 2.6.0.

Return d as an object that can be awaited from a Scrapy callable defined as a coroutine.

What you can await in Scrapy callables defined as coroutines depends on the value of TWISTED_REACTOR:

When not using the asyncio reactor, you can only await on Deferred objects.
When using the asyncio reactor, you can only await on asyncio.Future objects.

If you want to write code that uses Deferred objects but works with any reactor, use this function on all Deferred objects:

class MySpider(Spider):
    ...
    async def parse(self, response):
        d = treq.get('https://example.com/additional')
        extra_response = await maybe_deferred_to_future(d)

Tip

If you need to use these functions in code that aims to be compatible with lower versions of Scrapy that do not provide these functions, down to Scrapy 2.0 (earlier versions do not support asyncio), you can copy the implementation of these functions into your own code.