Ever used asyncio and wished you hadn't?

tinyio is a dead-simple event loop for Python, born out of my frustration with trying to get robust error handling with asyncio. (I'm not the only one running into its sharp corners: link1, link2.)

This is an alternative for the simple use-cases, where you just need an event loop, and want to crash the whole thing if anything goes wrong. (Raising an exception in every coroutine so it can clean up its resources.)

import tinyio

def slow_add_one(x: int):
    yield tinyio.sleep(1)
    return x + 1

def foo():
    four, five = yield [slow_add_one(3), slow_add_one(4)]
    return four, five

loop = tinyio.Loop()
out = loop.run(foo())
assert out == (4, 5)

• Somewhat unusually, our syntax uses yield rather than await, but the behaviour is the same. Await another coroutine with yield coro. Await on multiple with yield [coro1, coro2, ...] (a 'gather' in asyncio terminology; a 'nursery' in trio terminology).
• An error in one coroutine will cancel all coroutines across the entire event loop.
  • If the erroring coroutine is sequentially depended on by a chain of other coroutines, then we chain their tracebacks for easier debugging.
  • Errors even propagate to and from synchronous operations ran in threads.
• Can nest tinyio loops inside each other, none of this one-per-thread business.
• Ludicrously simple. No need for futures, tasks, etc. Here's the full API:

  tinyio.Loop
  tinyio.run_in_thread
  tinyio.sleep
  tinyio.CancelledError

pip install tinyio

Create a loop with tinyio.Loop(). It has a single method, .run(coro), which consumes a coroutine, and which returns the output of that coroutine.

Coroutines can yield three possible things:

• yield: yield nothing, this just pauses and gives other coroutines a chance to run.
• yield coro: wait on a single coroutine, in which case we'll resume with the output of that coroutine once it is available.
• yield [coro1, coro2, coro3]: wait on multiple coroutines by putting them in a list, and resume with a list of outputs once all have completed. This is what asyncio calls a 'gather' or 'TaskGroup', and what trio calls a 'nursery'.

You can safely yield the same coroutine multiple times, e.g. perhaps four coroutines have a diamond dependency pattern, with two coroutines each depending on a single shared one.

Synchronous functions can be ran in threads using tinyio.run_in_thread(fn, *args, **kwargs), which returns a coroutine you can yield on:

import time, tinyio

def slow_blocking_add_one(x: int) -> int:
    time.sleep(1)
    return x + 1

def foo(x: int):
    out = yield [tinyio.run_in_thread(slow_blocking_add_one, x) for _ in range(3)]
    return out

loop = tinyio.Loop()
out = loop.run(foo(x=1))  # runs in one second, not three
assert out == [2, 2, 2]

The thread will call fn(*args, **kwargs).

This is tinyio.sleep(delay_in_seconds), which is a coroutine you can yield on.

If any coroutine raises an error, then:

1. All coroutines across the entire loop will have tinyio.CancelledError raised in them (from whatever yield point they are currently waiting at).
2. Any functions ran in threads via tinyio.run_in_thread will also have tinyio.CancelledError raised in the thread.
3. The original error is raised out of loop.run(...). This behaviour can be configured (e.g. to collect errors into a BaseExceptionGroup) by setting loop.run(..., exception_group=None/False/True).

This gives every coroutine a chance to shut down gracefully. Debuggers like patdb offer the ability to navigate across exceptions in an exception group, allowing you to inspect the state of all coroutines that were related to the error.