Codename Tulip: The Making Of Async

Software has been riding on immense hardware improvement over decades. This has delayed ways to build products that are more effective, because you could cheaply rent yet another machine on the cloud and call it a day.

But hardware isn't getting better at the pace it used to, and costs are spiraling out of control. Software must become better at managing resources. Do more with less.

How? We can look at ways in which real world businesses maximize their effectiveness. Look at Starbucks! They have cashiers accepting orders, and baristas delivering them. That isn't rocket science; it's concurrency.

Would you be surprised to learn that modern programming languages aren't built for concurrency? It's as if English didn't have a syntax for asking questions! Concurrency is very much around us in real life, but Python pretends that things always happen One Damned Thing After Another.

Not all hope is lost. A few years ago, asyncio was introduced into the standard library to reboot concurrent programming. However, we're still getting used to it. It's weird! There are many things that we need to learn, and double check, and test and ultimately fix.

But what if none of that was the case? What if what we were missing was not Yet Another Standard, This Time By Guido, but a syntax?

I like my programming languages to have a semantics for concurrency, the way I like my English with a semantics for questions. And you?

Transcript

2. AsyncIO
3. AsynCore
4. @oh_duran
5. Codename Tulip Alvaro Duran 
 Senior Software Engineer 
 Kiwi.com The Making of Async
7. Device Latency in CPU cycles L1 Cache 3 L2 Cache 14 RAM 250 DISK 41,000,000 NETWORK 240,000,000 Dahl, Ryan. Introduction to Node.js @oh_duran
8. Device Latency in CPU cycles L1 Cache 3 3 seconds L2 Cache 14 14 seconds RAM 250 250 seconds DISK 41,000,000 1.3 years NETWORK 240,000,000 7.6 years Dahl, Ryan. Introduction to Node.js @oh_duran
9. Non-blocking Blocking @oh_duran
10. Non-blocking Blocking @oh_duran
11. open("f i le", "r") ; line = f.readline() f = @oh_duran
12. open("f i le", "r") line = f.readline() def foo() : f.close() return line f = @oh_duran
13. try: line = f.readline() def foo() : f.close() return line open("f i le", "r") f i nally: f = @oh_duran
14. with line = f.readline() def foo() : return line open("f i le", "r") as f: @oh_duran
15. goto
18. @oh_duran
19. @oh_duran
22. Run Run Run Run Trap Trap Trap OS Executes @oh_duran
24. Nystrom, Bob. What Color is Your Function? Independent SwitchableCallstacks @oh_duran Coroutines
25. def f i bonacci() : a, b = 0, 1 while True: yield b a, b = b, a + b Schemenauer, Neil; Peters, Tim; Lie Hetland, Magnus. PEP 255 - Simple Generators. @oh_duran
26. def averager() : total = 0.0 count = 0 
 average = None while True: term = yield average total += term count += 1 average = total / count Ramalho, Luciano. Fluent Python. O'Reilly 2015 @oh_duran
27. term = yield average sent into the generator total += term count += 1 average = total / count def averager() : total = 0.0 count = 0 
 average = None while True: yielded to the caller @oh_duran
28. def averager() : while True: results[key] = yield averager() main() @oh_duran
29. def averager() : while True:
30. def averager() : total = 0.0 count = 0 
 average = None while True: term = yield average total += term count += 1 average = total / count @coroutine primes the generator flags the generator @oh_duran
32. Confusing Subtle Bugs Limiting @oh_duran
33. Adding a new keyword 
 is a lot harder than 
 adding new syntax 
 composed of existing keywords. —Guido The Di ff erence between Yield and Yield From. Python Tulip Google Groups @oh_duran
34. Meh. Just deal with it. —Guido The Di ff erence between Yield and Yield From. Python Tulip Google Groups @oh_duran
35. AsyncIO (Interoperable Layer) @oh_duran
36. Async/Await (API) @oh_duran
37. async def handler(client) : . . . data = await client.recv(10000) @coroutine def read(f i leno, maxbytes) : yield from ('read', f i leno, maxbytes) yield def run(coro, value) : try: trap = coro.send(value) except StopIteration: return e.value Sync Space Async Space @oh_duran
38. TRAP Kernel Space User Space @oh_duran
39. @oh_duran vole.wtf/coder - serial - killer - quiz/
40. @oh_duran
41. @oh_duran
43. alvaro.duran@kiwi.com @oh_duran linkedin.com/in/alvaroduranbarata psst, my team at Kiwi.com is hiring! 
 Come see me after the Q&A