Module Picos_std_structured.Flock

An implicit dynamic flock of fibers guaranteed to be joined at the end.

Flocks allow you to conveniently structure or delimit concurrency into nested scopes. After a flock returns or raises an exception, no fibers forked to the flock remain.

An unhandled exception, or error, within any fiber of the flock causes all of the fibers forked to the flock to be canceled and the flock to raise the error exception or error exceptions raised by all of the fibers forked into the flock.

ℹ️ This is essentially a very thin convenience wrapper for an implicitly propagated Bundle.

⚠️ All of the operations in this module, except join_after, raise the Invalid_argument exception in case they are called from outside of the dynamic multifiber scope of a flock established by calling join_after.

val join_after : ?callstack:int -> ?on_return:[ `Terminate | `Wait ] -> (unit -> 'a) -> 'a

join_after scope creates a new flock for fibers, calls scope after setting current flock to the new flock, and restores the previous flock, if any after scope exits. The flock will be implicitly propagated to all fibers forked into the flock. A call of join_after returns or raises only after scope has returned or raised and all forked fibers have terminated. If scope raises an exception, error will be called.

The optional on_return argument specifies what to do when the scope returns normally. It defaults to `Wait, which means to just wait for all the fibers to terminate on their own. When explicitly specified as ~on_return:`Terminate, then terminate ?callstack will be called on return. This can be convenient, for example, when dealing with daemon fibers.

val terminate : ?callstack:int -> unit -> unit

terminate () cancels all of the forked fibers using the Terminate exception. After terminate has been called, no new fibers can be forked to the current flock.

The optional callstack argument specifies the number of callstack entries to capture with the Terminate exception. The default is 0.

ℹ️ Calling terminate at the end of a flock can be a convenient way to cancel any background fibers started by the flock.

ℹ️ Calling terminate does not raise the Terminate exception, but blocking operations after terminate will raise the exception to propagate cancelation unless propagation of cancelation is forbidden.

val terminate_after : ?callstack:int -> seconds:float -> unit -> unit

terminate_after ~seconds () arranges to terminate the current flock after the specified timeout in seconds.

val error : ?callstack:int -> exn -> Stdlib.Printexc.raw_backtrace -> unit

error exn bt first calls terminate and then adds the exception with backtrace to the list of exceptions to be raised, unless the exception is the Terminate exception, which is not considered to signal an error by itself.

The optional callstack argument is passed to terminate.

val fork_as_promise : (unit -> 'a) -> 'a Promise.t

fork_as_promise thunk spawns a new fiber to the current flock that will run the given thunk. The result of the thunk will be written to the promise. If the thunk raises an exception, error will be called with that exception.

val fork : (unit -> unit) -> unit

fork action is equivalent to fork_as_promise action |> ignore.