Source   Edit

Nim coroutines implementation, supports several context switching methods:

ucontextavailable on unix and alike (default)
setjmpavailable on unix and alike (x86/64 only)
fibersavailable and required on windows.
Required to build this module.
Use ucontext backend.
Use setjmp backend.
Use bundled setjmp implementation.

Unstable API.

Timer support for the realtime GC. Based on


CoroutineRef = ref object
  coro: CoroutinePtr
CoroutineRef holds a pointer to actual coroutine object. Public API always returns CoroutineRef instead of CoroutinePtr in order to allow holding a reference to coroutine object while it can be safely deallocated by coroutine scheduler loop. In this case Coroutine.reference.coro is set to nil. Public API checks for it being nil and gracefully fails if it is nil.   Source   Edit


proc alive(c: CoroutineRef): bool {....raises: [], tags: [].}
Returns true if coroutine has not returned, false otherwise.   Source   Edit
proc run() {....raises: [], tags: [TimeEffect].}
Starts main coroutine scheduler loop which exits when all coroutines exit. Calling this proc starts execution of first coroutine.   Source   Edit
proc start(c: proc (); stacksize: int = defaultStackSize): CoroutineRef {.
    discardable, ...raises: [], tags: [RootEffect].}
Schedule coroutine for execution. It does not run immediately.   Source   Edit
proc suspend(sleepTime: float = 0) {....raises: [], tags: [].}
Stops coroutine execution and resumes no sooner than after sleeptime seconds. Until then other coroutines are executed.   Source   Edit
proc wait(c: CoroutineRef; interval = 0.01) {....raises: [], tags: [].}
Returns only after coroutine c has returned. interval is time in seconds how often.   Source   Edit