Philosophy

There's been a great deal of momentum to embrace async functions for IO, but they split the world and are confusing, especially to new developers. Threads or green threads are often a better choice for concurrency because it doesn't require learning about coroutines before you even need concurrency.

Still, async functions are a powerful way to implement state machines. Complex workflows can be represented as async functions to allow thinking about things in a pull-based fashion, while also scaling well for processes that must take a long time. Ideally, running coroutines could be serialized and resumed, perhaps even on a different machine.

This vision motivates queueio. Routines need to coordinate with the worker because queue-level concurrency must be carefully managed. Instead of complex construction of pipelines to make this possible, queueio allows you to write complex workflows in a traditional imperative style.

queueio encourages you to use synchronous IO in routines. The yield points in async functions are only needed when a routine needs to coordinate with the worker, such as to indicate that the worker can use the capacity reserved for it to do other work.