Taking Django's ORM Async

A talk I gave at DjangoCon AU 2020 about the plans and sequencing for making Django's ORM support asynchronous call styles.

Transcript

1. TAKING DJANGO'S ORM
   ANDREW GODWIN // @andrewgodwin
   ASYNC
   

   View Slide

2. Andrew Godwin / @andrewgodwin
   Hi, I’m
   Andrew Godwin
   • Django core developer
   • Worked on Migrations, Channels & Async
   • Not currently in Australia
   

   View Slide

3. Andrew Godwin / @andrewgodwin
   

   View Slide

4. Andrew Godwin / @andrewgodwin
   Async views are released!
   Find them in a Django 3.1 near you.
   

   View Slide

5. Andrew Godwin / @andrewgodwin
   WSGIHandler
   __call__
   WSGI Server
   WSGIRequest URLs Middleware
   View
   __call__
   ASGIHandler
   __call__
   ASGI Server
   ASGIRequest
   Asynchronous
   request path
   BaseHandler
   get_response_async
   BaseHandler
   get_response
   URLs Middleware
   Async View
   __call__
   Implemented async request flow
   

   View Slide

6. Andrew Godwin / @andrewgodwin
   Phase One: ASGI Support
   Allowing Django to be async at all
   Phase Two: Async Views
   Unlocking async use in normal apps
   Phase Three: The ORM
   High-level async use for the most common case
   

   View Slide

7. Andrew Godwin / @andrewgodwin
   The ORM is the majority of Django
   And, of course, the most complex part.
   

   View Slide

8. Andrew Godwin / @andrewgodwin
   API Design is crucial
   It must be familiar, yet safe.
   

   View Slide

9. Andrew Godwin / @andrewgodwin
   Async is… a bit different
   You can't quite do everything you're used to
   

   View Slide

10. Andrew Godwin / @andrewgodwin
    Can't tell if a function returns a coroutine!
    There are standard hints, but no actual guaranteed way
    

    View Slide

11. Andrew Godwin / @andrewgodwin
    async def calculate(x):
    result = await coroutine(x)
    return result
    # These both return a coroutine
    def calculate(x):
    result = coroutine(x)
    return result
    

    View Slide

12. Andrew Godwin / @andrewgodwin
    You have to namespace async functions
    I really, really wish we didn't have to
    

    View Slide

13. Andrew Godwin / @andrewgodwin
    instance = MyModel.objects.get(id=3)
    instance = await MyModel.objects.async_get(id=3)
    

    View Slide

14. Andrew Godwin / @andrewgodwin
    Can't do asynchronous attribute access!
    Foreign keys will be a bit less magic.
    

    View Slide

15. Andrew Godwin / @andrewgodwin
    book = Book.objects.async_get(id=3)
    # Will error - needs to run a query
    print(book.author.name)
    

    View Slide

16. Andrew Godwin / @andrewgodwin
    book = Book.objects.get(id=3)
    .select_related("author")
    # Works fine - prefetched
    print(book.author.name)
    

    View Slide

17. Andrew Godwin / @andrewgodwin
    Some things do have nice analogues!
    They have async versions of the operations that call a different special method.
    

    View Slide

18. Andrew Godwin / @andrewgodwin
    async for book in Book.objects.filter(name="Django"):
    print(book)
    

    View Slide

19. Andrew Godwin / @andrewgodwin
    The Plan
    

    View Slide

20. Andrew Godwin / @andrewgodwin
    1. Async Model API
    Querysets & model instances mostly
    2. Async Query Internals
    Django is internally async through its stack
    3. Async Database Adapters
    Async all the way down, no threads at all
    

    View Slide

21. Andrew Godwin / @andrewgodwin
    Query
    Threaded if in
    async mode
    After Phase One
    QuerySet Managers
    Model
    Connection Compiler
    Database
    Library
    

    View Slide

22. Andrew Godwin / @andrewgodwin
    Query
    Threaded if in
    async mode
    After Phase Two
    QuerySet Managers
    Model
    Connection Compiler
    Database
    Library
    

    View Slide

23. Andrew Godwin / @andrewgodwin
    Query
    After Phase Three
    QuerySet Managers
    Model
    Connection Compiler
    Database
    Library
    

    View Slide

24. Andrew Godwin / @andrewgodwin
    We may never get to Phase Three
    ...and that's alright.
    

    View Slide

25. Andrew Godwin / @andrewgodwin
    No sign of an async DBAPI... yet
    It might emerge once there's a need for it
    

    View Slide

26. Andrew Godwin / @andrewgodwin
    Databases via threads? Not the worst!
    Native async only matters for pure performance
    

    View Slide

27. Andrew Godwin / @andrewgodwin
    Transactions are tricky.
    They are very threadlocal, and that's not great with async
    

    View Slide

28. Andrew Godwin / @andrewgodwin
    Some things don't need async
    Migrations, introspection, most fields
    

    View Slide

29. Andrew Godwin / @andrewgodwin
    So what's first?
    (Apart from having to curse various non-PostgreSQL databases)
    

    View Slide

30. Andrew Godwin / @andrewgodwin
    Deciding on async API design
    Namespacing? async_ prefixes? Something else?
    

    View Slide

31. Andrew Godwin / @andrewgodwin
    Grappling with transactions
    Do we make them work across async/sync boundaries?
    

    View Slide

32. Andrew Godwin / @andrewgodwin
    Async connection management
    Namespacing? async_ prefixes? Something else?
    

    View Slide

33. Andrew Godwin / @andrewgodwin
    Asyncio only benefits IO-bound code
    Code that thrashes the CPU doesn't benefit at all
    

    View Slide

34. Andrew Godwin / @andrewgodwin
    There's lots to come!
    Hopefully starting in 3.2, but no guarantees...
    

    View Slide

35. Thanks.
    Andrew Godwin
    @andrewgodwin // aeracode.org
    

    View Slide