Await for it: mixing async and blocking code - PyCon MEA - Dubai

Tips and tricks to properly mix async code with blocking code, using modern libraries, and taking advantage of the new concurrency features of the language while keeping compatibility with already existing code and libraries.

It will include the basics of what concurrency is, how to take advantage of it in the simplest use cases, and how to safely mix it with regular code.

You've heard async is great, it can improve concurrency performance, but you might have heard that you need to use only async code and use only async-compatible libraries... but the good news is, that's not entirely true.

So, how can you take advantage of these new async features in your current projects that already use blocking code and require some blocking libraries?

In this talk, I'll show you the basics of writing concurrent async code, and how to properly mix it with regular code without affecting performance (without blocking the event loop) by using thread workers underneath. All done in a simple and convenient way using modern libraries.

Transcript

1. ⏱ Await for it mixing async and blocking code
2. Who am I? Sebastián Ramírez github.com/tiangolo linkedin.com/in/tiangolo twitter.com/tiangolo tiangolo.com Berlin, Germany I created:
3. ⏱ async / await - concurrency @tiangolo
4. ⏱ async / await - concurrency @tiangolo
5. ⏱ async / await - concurrency @tiangolo
6. ⏱ async / await - concurrency Photo by John Cameron: https://unsplash.com/@john_cameron @tiangolo
7. ⏱ async / await - concurrency @tiangolo
8. ⏱ async / await - concurrency @tiangolo
9. ⏱ async / await - concurrency Photo by Kate Townsend: https://unsplash.com/@k8townsend @tiangolo
10. ⏱ async / await - concurrency @tiangolo
11. @tiangolo 📕 So many terms!
12. 📕 Concurrency and Parallelism @tiangolo • Parallel: multiple at the same time • Concurrent: multiple during the same period of time ◦ Normally in turns
13. • A lot of waiting? (e.g. multiple users) ◦ Concurrency • A lot of the same CPU work, in multiple cores or GPUs? (e.g. image processing) ◦ Parallelism 📕 When concurrency, when parallelism @tiangolo
14. • Throughput: amount of items passing through. E.g. requests per second. ◦ Concurrency helps. • Latency: how long each individual item takes. ◦ Code optimization ◦ Numpy ◦ GPUs ◦ mypyc ◦ Cython ◦ Rust 📕 Performance and Speed: throughput or latency @tiangolo
15. 📕 So many terms! @tiangolo
16. @tiangolo 🌪 Using Async
17. 󰥟 Async rules @tiangolo • You can only use await inside async functions. • To call an async function, you need to use await in front of it.
18. 󰥟 Async rules @tiangolo
19. 󰥟 Async rules @tiangolo
20. 🌪 Run an async program with AnyIO @tiangolo
21. 🌪 Run an async program with AnyIO @tiangolo
22. @tiangolo 🌪 Async with FastAPI and HTTPX
23. 🌪 Async with FastAPI and HTTPX @tiangolo
24. 🌪 Async with FastAPI and HTTPX @tiangolo
25. 🌪 Async with FastAPI and HTTPX @tiangolo
26. 🌪 Async with FastAPI and HTTPX, or not @tiangolo
27. 🌪 Async with FastAPI and HTTPX optional, or not @tiangolo
28. @tiangolo 🌪 Concurrency in your code
29. Multiple async function calls - not concurrent ⛔ @tiangolo
30. Multiple async function calls - not concurrent ⛔ @tiangolo
31. 🌪 Multiple async function calls - concurrent @tiangolo
32. 🔍 Asyncer soonify - inline errors @tiangolo
33. 🔍 Asyncer soonify - autocompletion @tiangolo
34. 🔍 Asyncer soonify - concurrent work @tiangolo
35. 🎁 Return values @tiangolo
36. 🎁 Asyncer - Soon Values @tiangolo
37. 🎁 Asyncer - Soon Values - autocompletion @tiangolo
38. 🎁 Asyncer - Soon Values - inline errors @tiangolo
39. 🤓 Autocompletion and inline errors everywhere @tiangolo
40. 😎 Async with Asyncer @tiangolo
41. @tiangolo 󰠼 Blocking sync code in async
42. 󰠼 Blocking sync code in async - don't do this! ⛔ @tiangolo
43. 󰠼 Blocking sync code in async - don't do this! ⛔ @tiangolo
44. 󰥟 Blocking sync code in async - performance rules @tiangolo • Do not call blocking sync code in async code directly. • Convert it to awaitable. • Run on another thread.
45. 󰥟 What is blocking sync code @tiangolo • Doesn't use async • Requires waiting for something • Interacts with something outside your code and variables ◦ Network ◦ External APIs ◦ External databases ◦ Files on disk • Or a lot of CPU for a long time
46. 🌪 Blocking sync code in async - asyncify @tiangolo
47. 🔍 asyncify - inline errors @tiangolo
48. 󰩂 asyncify - autocompletion @tiangolo
49. 󰩂 Blocking sync code in async @tiangolo
50. @tiangolo 🌪 Async in blocking sync in async
51. 🌪 Async in blocking sync in async - syncify @tiangolo
52. 󰩂 syncify - autocompletion - inline errors @tiangolo
53. 🌪 Async in sync in async… @tiangolo
54. 🌪 Async in blocking sync, dual - syncify @tiangolo
55. 🌪 Async in blocking sync, dual - syncify @tiangolo ⛔ Warning: performance in mainly sync code won't be the same as async.
56. 🌪 Async in blocking sync, dual - syncify @tiangolo
57. @tiangolo Libraries, what to use ❓
58. • Use AnyIO ◦ Compatible with both Asyncio and Trio ◦ Structured concurrency (task groups) • Maybe try Asyncer ◦ AnyIO + types: autocompletion, inline errors • Trio ◦ Structured concurrency (task groups) • Asyncio ◦ Comes with Python ◦ Next versions with task groups ❓ Libraries, what to use @tiangolo
60. Thank you! Sebastián Ramírez tiangolo.com fastapi.tiangolo.com github.com/tiangolo linkedin.com/in/tiangolo twitter.com/tiangolo