The Thread_safe module has functions that are safe to call from threads outside
Async.
All the Thread_safe.block* and Thread_safe.run* functions wake up the Async
scheduler to ensure that it continues in a timely manner with whatever jobs got
started. Some functions take an optional ?wakeup_scheduler:bool argument, which
defaults to true. One can cause the scheduler to not be woken up by supplying
~wakeup_scheduler:false, which can reduce CPU use, but increase latency, because the
scheduler may not wake up for a while to process jobs.
am_holding_async_lock () returns true if the currently running thread is holding the
Async lock.
deferred () returns (d, fill) where d is a deferred that will become determined
with value v once fill v is called.
It is ok to call deferred from inside or outside Async. fill must be called from
outside Async.
run_in_async_with_optional_cycle f acquires the Async lock and runs f () while
holding the lock. Depending on the result of f, it may also run a cycle.
run_in_async f acquires the Async lock and runs f () while holding the lock. It
returns the result of f () to the outside world. The scheduler is woken up to
ensures the code that depends on f () is run soon enough.
run_in_async doesn't run a cycle.
run_in_async does not automatically start the Async scheduler. You still need to
call Scheduler.go elsewhere in your program.
block_on_async f runs f () in the Async world and blocks until the result becomes
determined. This function can be called from the main thread (before Async is
started) or from a thread outside Async.
block_on_async will run a cycle if the deferred isn't determined, in the hope that
running the cycle will cause the deferred to become determined.
block_on_async will automatically start the scheduler if it isn't already
running.
run_in_async_wait f is like block_on_async f, except that it must be called from a
thread outside Async. Upon returning from run_in_async_wait, it is guaranteed that
the caller does not have the Async lock.