1Coro(3) User Contributed Perl Documentation Coro(3)
2
6 Coro - the only real threads in perl
7
9 use Coro;
10 async {
12 # some asynchronous thread of execution
13 print "2\n";
14 cede; # yield back to main
15 print "4\n";
16 };
17 print "1\n";
18 cede; # yield to coro
19 print "3\n";
20 cede; # and again
21 # use locking
23 my $lock = new Coro::Semaphore;
24 my $locked;
25 $lock->down;
27 $locked = 1;
28 $lock->up;
29
31 For a tutorial-style introduction, please read the Coro::Intro manpage.
32 This manpage mainly contains reference information.
33 This module collection manages continuations in general, most often in
35 the form of cooperative threads (also called coros, or simply "coro" in
36 the documentation). They are similar to kernel threads but don't (in
37 general) run in parallel at the same time even on SMP machines. The
38 specific flavor of thread offered by this module also guarantees you
39 that it will not switch between threads unless necessary, at easily-
40 identified points in your program, so locking and parallel access are
41 rarely an issue, making thread programming much safer and easier than
42 using other thread models.
43 Unlike the so-called "Perl threads" (which are not actually real
45 threads but only the windows process emulation (see section of same
46 name for more details) ported to UNIX, and as such act as processes),
47 Coro provides a full shared address space, which makes communication
48 between threads very easy. And coro threads are fast, too: disabling
49 the Windows process emulation code in your perl and using Coro can
50 easily result in a two to four times speed increase for your programs.
51 A parallel matrix multiplication benchmark (very communication-
52 intensive) runs over 300 times faster on a single core than perls
53 pseudo-threads on a quad core using all four cores.
54 Coro achieves that by supporting multiple running interpreters that
56 share data, which is especially useful to code pseudo-parallel
57 processes and for event-based programming, such as multiple HTTP-GET
58 requests running concurrently. See Coro::AnyEvent to learn more on how
59 to integrate Coro into an event-based environment.
60 In this module, a thread is defined as "callchain + lexical variables +
62 some package variables + C stack), that is, a thread has its own
63 callchain, its own set of lexicals and its own set of perls most
64 important global variables (see Coro::State for more configuration and
65 background info).
66 See also the "SEE ALSO" section at the end of this document - the Coro
68 module family is quite large.
69
71 During the long and exciting (or not) life of a coro thread, it goes
72 through a number of states:
73 1. Creation
75 The first thing in the life of a coro thread is its creation -
76 obviously. The typical way to create a thread is to call the "async
77 BLOCK" function:
78 async {
80 # thread code goes here
81 };
82 You can also pass arguments, which are put in @_:
84 async {
86 print $_[1]; # prints 2
87 } 1, 2, 3;
88 This creates a new coro thread and puts it into the ready queue,
90 meaning it will run as soon as the CPU is free for it.
91 "async" will return a Coro object - you can store this for future
93 reference or ignore it - a thread that is running, ready to run or
94 waiting for some event is alive on its own.
95 Another way to create a thread is to call the "new" constructor
97 with a code-reference:
98 new Coro sub {
100 # thread code goes here
101 }, @optional_arguments;
102 This is quite similar to calling "async", but the important
104 difference is that the new thread is not put into the ready queue,
105 so the thread will not run until somebody puts it there. "async"
106 is, therefore, identical to this sequence:
107 my $coro = new Coro sub {
109 # thread code goes here
110 };
111 $coro->ready;
112 return $coro;
113 2. Startup
115 When a new coro thread is created, only a copy of the code
116 reference and the arguments are stored, no extra memory for stacks
117 and so on is allocated, keeping the coro thread in a low-memory
118 state.
119 Only when it actually starts executing will all the resources be
121 finally allocated.
122 The optional arguments specified at coro creation are available in
124 @_, similar to function calls.
125 3. Running / Blocking
127 A lot can happen after the coro thread has started running. Quite
128 usually, it will not run to the end in one go (because you could
129 use a function instead), but it will give up the CPU regularly
130 because it waits for external events.
131 As long as a coro thread runs, its Coro object is available in the
133 global variable $Coro::current.
134 The low-level way to give up the CPU is to call the scheduler,
136 which selects a new coro thread to run:
137 Coro::schedule;
139 Since running threads are not in the ready queue, calling the
141 scheduler without doing anything else will block the coro thread
142 forever - you need to arrange either for the coro to put woken up
143 (readied) by some other event or some other thread, or you can put
144 it into the ready queue before scheduling:
145 # this is exactly what Coro::cede does
147 $Coro::current->ready;
148 Coro::schedule;
149 All the higher-level synchronisation methods (Coro::Semaphore,
151 Coro::rouse_*...) are actually implemented via "->ready" and
152 "Coro::schedule".
153 While the coro thread is running it also might get assigned a
155 C-level thread, or the C-level thread might be unassigned from it,
156 as the Coro runtime wishes. A C-level thread needs to be assigned
157 when your perl thread calls into some C-level function and that
158 function in turn calls perl and perl then wants to switch
159 coroutines. This happens most often when you run an event loop and
160 block in the callback, or when perl itself calls some function such
161 as "AUTOLOAD" or methods via the "tie" mechanism.
162 4. Termination
164 Many threads actually terminate after some time. There are a number
165 of ways to terminate a coro thread, the simplest is returning from
166 the top-level code reference:
167 async {
169 # after returning from here, the coro thread is terminated
170 };
171 async {
173 return if 0.5 < rand; # terminate a little earlier, maybe
174 print "got a chance to print this\n";
175 # or here
176 };
177 Any values returned from the coroutine can be recovered using
179 "->join":
180 my $coro = async {
182 "hello, world\n" # return a string
183 };
184 my $hello_world = $coro->join;
186 print $hello_world;
188 Another way to terminate is to call "Coro::terminate", which at any
190 subroutine call nesting level:
191 async {
193 Coro::terminate "return value 1", "return value 2";
194 };
195 Yet another way is to "->cancel" (or "->safe_cancel") the coro
197 thread from another thread:
198 my $coro = async {
200 exit 1;
201 };
202 $coro->cancel; # also accepts values for ->join to retrieve
204 Cancellation can be dangerous - it's a bit like calling "exit"
206 without actually exiting, and might leave C libraries and XS
207 modules in a weird state. Unlike other thread implementations,
208 however, Coro is exceptionally safe with regards to cancellation,
209 as perl will always be in a consistent state, and for those cases
210 where you want to do truly marvellous things with your coro while
211 it is being cancelled - that is, make sure all cleanup code is
212 executed from the thread being cancelled - there is even a
213 "->safe_cancel" method.
214 So, cancelling a thread that runs in an XS event loop might not be
216 the best idea, but any other combination that deals with perl only
217 (cancelling when a thread is in a "tie" method or an "AUTOLOAD" for
218 example) is safe.
219 Last not least, a coro thread object that isn't referenced is
221 "->cancel"'ed automatically - just like other objects in Perl. This
222 is not such a common case, however - a running thread is
223 referencedy by $Coro::current, a thread ready to run is referenced
224 by the ready queue, a thread waiting on a lock or semaphore is
225 referenced by being in some wait list and so on. But a thread that
226 isn't in any of those queues gets cancelled:
227 async {
229 schedule; # cede to other coros, don't go into the ready queue
230 };
231 cede;
233 # now the async above is destroyed, as it is not referenced by anything.
234 A slightly embellished example might make it clearer:
236 async {
238 my $guard = Guard::guard { print "destroyed\n" };
239 schedule while 1;
240 };
241 cede;
243 Superficially one might not expect any output - since the "async"
245 implements an endless loop, the $guard will not be cleaned up.
246 However, since the thread object returned by "async" is not stored
247 anywhere, the thread is initially referenced because it is in the
248 ready queue, when it runs it is referenced by $Coro::current, but
249 when it calls "schedule", it gets "cancel"ed causing the guard
250 object to be destroyed (see the next section), and printing its
251 message.
252 If this seems a bit drastic, remember that this only happens when
254 nothing references the thread anymore, which means there is no way
255 to further execute it, ever. The only options at this point are
256 leaking the thread, or cleaning it up, which brings us to...
257 5. Cleanup
259 Threads will allocate various resources. Most but not all will be
260 returned when a thread terminates, during clean-up.
261 Cleanup is quite similar to throwing an uncaught exception: perl
263 will work its way up through all subroutine calls and blocks. On
264 its way, it will release all "my" variables, undo all "local"'s and
265 free any other resources truly local to the thread.
266 So, a common way to free resources is to keep them referenced only
268 by my variables:
269 async {
271 my $big_cache = new Cache ...;
272 };
273 If there are no other references, then the $big_cache object will
275 be freed when the thread terminates, regardless of how it does so.
276 What it does "NOT" do is unlock any Coro::Semaphores or similar
278 resources, but that's where the "guard" methods come in handy:
279 my $sem = new Coro::Semaphore;
281 async {
283 my $lock_guard = $sem->guard;
284 # if we return, or die or get cancelled, here,
285 # then the semaphore will be "up"ed.
286 };
287 The "Guard::guard" function comes in handy for any custom cleanup
289 you might want to do (but you cannot switch to other coroutines
290 from those code blocks):
291 async {
293 my $window = new Gtk2::Window "toplevel";
294 # The window will not be cleaned up automatically, even when $window
295 # gets freed, so use a guard to ensure its destruction
296 # in case of an error:
297 my $window_guard = Guard::guard { $window->destroy };
298 # we are safe here
300 };
301 Last not least, "local" can often be handy, too, e.g. when
303 temporarily replacing the coro thread description:
304 sub myfunction {
306 local $Coro::current->{desc} = "inside myfunction(@_)";
307 # if we return or die here, the description will be restored
309 }
310 6. Viva La Zombie Muerte
312 Even after a thread has terminated and cleaned up its resources,
313 the Coro object still is there and stores the return values of the
314 thread.
315 When there are no other references, it will simply be cleaned up
317 and freed.
318 If there areany references, the Coro object will stay around, and
320 you can call "->join" as many times as you wish to retrieve the
321 result values:
322 async {
324 print "hi\n";
325 1
326 };
327 # run the async above, and free everything before returning
329 # from Coro::cede:
330 Coro::cede;
331 {
333 my $coro = async {
334 print "hi\n";
335 1
336 };
337 # run the async above, and clean up, but do not free the coro
339 # object:
340 Coro::cede;
341 # optionally retrieve the result values
343 my @results = $coro->join;
344 # now $coro goes out of scope, and presumably gets freed
346 };
347
349 $Coro::main
350 This variable stores the Coro object that represents the main
351 program. While you can "ready" it and do most other things you can
352 do to coro, it is mainly useful to compare again $Coro::current, to
353 see whether you are running in the main program or not.
354 $Coro::current
356 The Coro object representing the current coro (the last coro that
357 the Coro scheduler switched to). The initial value is $Coro::main
358 (of course).
359 This variable is strictly read-only. You can take copies of the
361 value stored in it and use it as any other Coro object, but you
362 must not otherwise modify the variable itself.
363 $Coro::idle
365 This variable is mainly useful to integrate Coro into event loops.
366 It is usually better to rely on Coro::AnyEvent or Coro::EV, as this
367 is pretty low-level functionality.
368 This variable stores a Coro object that is put into the ready queue
370 when there are no other ready threads (without invoking any ready
371 hooks).
372 The default implementation dies with "FATAL: deadlock detected.",
374 followed by a thread listing, because the program has no other way
375 to continue.
376 This hook is overwritten by modules such as "Coro::EV" and
378 "Coro::AnyEvent" to wait on an external event that hopefully wakes
379 up a coro so the scheduler can run it.
380 See Coro::EV or Coro::AnyEvent for examples of using this
382 technique.
383
385 async { ... } [@args...]
386 Create a new coro and return its Coro object (usually unused). The
387 coro will be put into the ready queue, so it will start running
388 automatically on the next scheduler run.
389 The first argument is a codeblock/closure that should be executed
391 in the coro. When it returns argument returns the coro is
392 automatically terminated.
393 The remaining arguments are passed as arguments to the closure.
395 See the "Coro::State::new" constructor for info about the coro
397 environment in which coro are executed.
398 Calling "exit" in a coro will do the same as calling exit outside
400 the coro. Likewise, when the coro dies, the program will exit, just
401 as it would in the main program.
402 If you do not want that, you can provide a default "die" handler,
404 or simply avoid dieing (by use of "eval").
405 Example: Create a new coro that just prints its arguments.
407 async {
409 print "@_\n";
410 } 1,2,3,4;
411 async_pool { ... } [@args...]
413 Similar to "async", but uses a coro pool, so you should not call
414 terminate or join on it (although you are allowed to), and you get
415 a coro that might have executed other code already (which can be
416 good or bad :).
417 On the plus side, this function is about twice as fast as creating
419 (and destroying) a completely new coro, so if you need a lot of
420 generic coros in quick successsion, use "async_pool", not "async".
421 The code block is executed in an "eval" context and a warning will
423 be issued in case of an exception instead of terminating the
424 program, as "async" does. As the coro is being reused, stuff like
425 "on_destroy" will not work in the expected way, unless you call
426 terminate or cancel, which somehow defeats the purpose of pooling
427 (but is fine in the exceptional case).
428 The priority will be reset to 0 after each run, all "swap_sv" calls
430 will be undone, tracing will be disabled, the description will be
431 reset and the default output filehandle gets restored, so you can
432 change all these. Otherwise the coro will be re-used "as-is": most
433 notably if you change other per-coro global stuff such as $/ you
434 must needs revert that change, which is most simply done by using
435 local as in: "local $/".
436 The idle pool size is limited to 8 idle coros (this can be adjusted
438 by changing $Coro::POOL_SIZE), but there can be as many non-idle
439 coros as required.
440 If you are concerned about pooled coros growing a lot because a
442 single "async_pool" used a lot of stackspace you can e.g.
443 "async_pool { terminate }" once per second or so to slowly
444 replenish the pool. In addition to that, when the stacks used by a
445 handler grows larger than 32kb (adjustable via $Coro::POOL_RSS) it
446 will also be destroyed.
447
449 Static methods are actually functions that implicitly operate on the
450 current coro.
451 schedule
453 Calls the scheduler. The scheduler will find the next coro that is
454 to be run from the ready queue and switches to it. The next coro to
455 be run is simply the one with the highest priority that is longest
456 in its ready queue. If there is no coro ready, it will call the
457 $Coro::idle hook.
458 Please note that the current coro will not be put into the ready
460 queue, so calling this function usually means you will never be
461 called again unless something else (e.g. an event handler) calls
462 "->ready", thus waking you up.
463 This makes "schedule" the generic method to use to block the
465 current coro and wait for events: first you remember the current
466 coro in a variable, then arrange for some callback of yours to call
467 "->ready" on that once some event happens, and last you call
468 "schedule" to put yourself to sleep. Note that a lot of things can
469 wake your coro up, so you need to check whether the event indeed
470 happened, e.g. by storing the status in a variable.
471 See HOW TO WAIT FOR A CALLBACK, below, for some ways to wait for
473 callbacks.
474 cede
476 "Cede" to other coros. This function puts the current coro into the
477 ready queue and calls "schedule", which has the effect of giving up
478 the current "timeslice" to other coros of the same or higher
479 priority. Once your coro gets its turn again it will automatically
480 be resumed.
481 This function is often called "yield" in other languages.
483 Coro::cede_notself
485 Works like cede, but is not exported by default and will cede to
486 any coro, regardless of priority. This is useful sometimes to
487 ensure progress is made.
488 terminate [arg...]
490 Terminates the current coro with the given status values (see
491 cancel). The values will not be copied, but referenced directly.
492 Coro::on_enter BLOCK, Coro::on_leave BLOCK
494 These function install enter and leave winders in the current
495 scope. The enter block will be executed when on_enter is called and
496 whenever the current coro is re-entered by the scheduler, while the
497 leave block is executed whenever the current coro is blocked by the
498 scheduler, and also when the containing scope is exited (by
499 whatever means, be it exit, die, last etc.).
500 Neither invoking the scheduler, nor exceptions, are allowed within
502 those BLOCKs. That means: do not even think about calling "die"
503 without an eval, and do not even think of entering the scheduler in
504 any way.
505 Since both BLOCKs are tied to the current scope, they will
507 automatically be removed when the current scope exits.
508 These functions implement the same concept as "dynamic-wind" in
510 scheme does, and are useful when you want to localise some resource
511 to a specific coro.
512 They slow down thread switching considerably for coros that use
514 them (about 40% for a BLOCK with a single assignment, so thread
515 switching is still reasonably fast if the handlers are fast).
516 These functions are best understood by an example: The following
518 function will change the current timezone to
519 "Antarctica/South_Pole", which requires a call to "tzset", but by
520 using "on_enter" and "on_leave", which remember/change the current
521 timezone and restore the previous value, respectively, the timezone
522 is only changed for the coro that installed those handlers.
523 use POSIX qw(tzset);
525 async {
527 my $old_tz; # store outside TZ value here
528 Coro::on_enter {
530 $old_tz = $ENV{TZ}; # remember the old value
531 $ENV{TZ} = "Antarctica/South_Pole";
533 tzset; # enable new value
534 };
535 Coro::on_leave {
537 $ENV{TZ} = $old_tz;
538 tzset; # restore old value
539 };
540 # at this place, the timezone is Antarctica/South_Pole,
542 # without disturbing the TZ of any other coro.
543 };
544 This can be used to localise about any resource (locale, uid,
546 current working directory etc.) to a block, despite the existence
547 of other coros.
548 Another interesting example implements time-sliced multitasking
550 using interval timers (this could obviously be optimised, but does
551 the job):
552 # "timeslice" the given block
554 sub timeslice(&) {
555 use Time::HiRes ();
556 Coro::on_enter {
558 # on entering the thread, we set an VTALRM handler to cede
559 $SIG{VTALRM} = sub { cede };
560 # and then start the interval timer
561 Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0.01, 0.01;
562 };
563 Coro::on_leave {
564 # on leaving the thread, we stop the interval timer again
565 Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0, 0;
566 };
567 &{+shift};
569 }
570 # use like this:
572 timeslice {
573 # The following is an endless loop that would normally
574 # monopolise the process. Since it runs in a timesliced
575 # environment, it will regularly cede to other threads.
576 while () { }
577 };
578 killall
580 Kills/terminates/cancels all coros except the currently running
581 one.
582 Note that while this will try to free some of the main interpreter
584 resources if the calling coro isn't the main coro, but one cannot
585 free all of them, so if a coro that is not the main coro calls this
586 function, there will be some one-time resource leak.
587
589 These are the methods you can call on coro objects (or to create them).
590 new Coro \&sub [, @args...]
592 Create a new coro and return it. When the sub returns, the coro
593 automatically terminates as if "terminate" with the returned values
594 were called. To make the coro run you must first put it into the
595 ready queue by calling the ready method.
596 See "async" and "Coro::State::new" for additional info about the
598 coro environment.
599 $success = $coro->ready
601 Put the given coro into the end of its ready queue (there is one
602 queue for each priority) and return true. If the coro is already in
603 the ready queue, do nothing and return false.
604 This ensures that the scheduler will resume this coro automatically
606 once all the coro of higher priority and all coro of the same
607 priority that were put into the ready queue earlier have been
608 resumed.
609 $coro->suspend
611 Suspends the specified coro. A suspended coro works just like any
612 other coro, except that the scheduler will not select a suspended
613 coro for execution.
614 Suspending a coro can be useful when you want to keep the coro from
616 running, but you don't want to destroy it, or when you want to
617 temporarily freeze a coro (e.g. for debugging) to resume it later.
618 A scenario for the former would be to suspend all (other) coros
620 after a fork and keep them alive, so their destructors aren't
621 called, but new coros can be created.
622 $coro->resume
624 If the specified coro was suspended, it will be resumed. Note that
625 when the coro was in the ready queue when it was suspended, it
626 might have been unreadied by the scheduler, so an activation might
627 have been lost.
628 To avoid this, it is best to put a suspended coro into the ready
630 queue unconditionally, as every synchronisation mechanism must
631 protect itself against spurious wakeups, and the one in the Coro
632 family certainly do that.
633 $state->is_new
635 Returns true iff this Coro object is "new", i.e. has never been run
636 yet. Those states basically consist of only the code reference to
637 call and the arguments, but consumes very little other resources.
638 New states will automatically get assigned a perl interpreter when
639 they are transferred to.
640 $state->is_zombie
642 Returns true iff the Coro object has been cancelled, i.e. its
643 resources freed because they were "cancel"'ed, "terminate"'d,
644 "safe_cancel"'ed or simply went out of scope.
645 The name "zombie" stems from UNIX culture, where a process that has
647 exited and only stores and exit status and no other resources is
648 called a "zombie".
649 $is_ready = $coro->is_ready
651 Returns true iff the Coro object is in the ready queue. Unless the
652 Coro object gets destroyed, it will eventually be scheduled by the
653 scheduler.
654 $is_running = $coro->is_running
656 Returns true iff the Coro object is currently running. Only one
657 Coro object can ever be in the running state (but it currently is
658 possible to have multiple running Coro::States).
659 $is_suspended = $coro->is_suspended
661 Returns true iff this Coro object has been suspended. Suspended
662 Coros will not ever be scheduled.
663 $coro->cancel ($arg...)
665 Terminate the given Coro thread and make it return the given
666 arguments as status (default: an empty list). Never returns if the
667 Coro is the current Coro.
668 This is a rather brutal way to free a coro, with some limitations -
670 if the thread is inside a C callback that doesn't expect to be
671 canceled, bad things can happen, or if the cancelled thread insists
672 on running complicated cleanup handlers that rely on its thread
673 context, things will not work.
674 Any cleanup code being run (e.g. from "guard" blocks, destructors
676 and so on) will be run without a thread context, and is not allowed
677 to switch to other threads. A common mistake is to call "->cancel"
678 from a destructor called by die'ing inside the thread to be
679 cancelled for example.
680 On the plus side, "->cancel" will always clean up the thread, no
682 matter what. If your cleanup code is complex or you want to avoid
683 cancelling a C-thread that doesn't know how to clean up itself, it
684 can be better to "->throw" an exception, or use "->safe_cancel".
685 The arguments to "->cancel" are not copied, but instead will be
687 referenced directly (e.g. if you pass $var and after the call
688 change that variable, then you might change the return values
689 passed to e.g. "join", so don't do that).
690 The resources of the Coro are usually freed (or destructed) before
692 this call returns, but this can be delayed for an indefinite amount
693 of time, as in some cases the manager thread has to run first to
694 actually destruct the Coro object.
695 $coro->safe_cancel ($arg...)
697 Works mostly like "->cancel", but is inherently "safer", and
698 consequently, can fail with an exception in cases the thread is not
699 in a cancellable state. Essentially, "->safe_cancel" is a
700 "->cancel" with extra checks before canceling.
701 It works a bit like throwing an exception that cannot be caught -
703 specifically, it will clean up the thread from within itself, so
704 all cleanup handlers (e.g. "guard" blocks) are run with full thread
705 context and can block if they wish. The downside is that there is
706 no guarantee that the thread can be cancelled when you call this
707 method, and therefore, it might fail. It is also considerably
708 slower than "cancel" or "terminate".
709 A thread is in a safe-cancellable state if it either has never been
711 run yet, has already been canceled/terminated or otherwise
712 destroyed, or has no C context attached and is inside an SLF
713 function.
714 The first two states are trivial - a thread that hasnot started or
716 has already finished is safe to cancel.
717 The last state basically means that the thread isn't currently
719 inside a perl callback called from some C function (usually via
720 some XS modules) and isn't currently executing inside some C
721 function itself (via Coro's XS API).
722 This call returns true when it could cancel the thread, or croaks
724 with an error otherwise (i.e. it either returns true or doesn't
725 return at all).
726 Why the weird interface? Well, there are two common models on how
728 and when to cancel things. In the first, you have the expectation
729 that your coro thread can be cancelled when you want to cancel it -
730 if the thread isn't cancellable, this would be a bug somewhere, so
731 "->safe_cancel" croaks to notify of the bug.
732 In the second model you sometimes want to ask nicely to cancel a
734 thread, but if it's not a good time, well, then don't cancel. This
735 can be done relatively easy like this:
736 if (! eval { $coro->safe_cancel }) {
738 warn "unable to cancel thread: $@";
739 }
740 However, what you never should do is first try to cancel "safely"
742 and if that fails, cancel the "hard" way with "->cancel". That
743 makes no sense: either you rely on being able to execute cleanup
744 code in your thread context, or you don't. If you do, then
745 "->safe_cancel" is the only way, and if you don't, then "->cancel"
746 is always faster and more direct.
747 $coro->schedule_to
749 Puts the current coro to sleep (like "Coro::schedule"), but instead
750 of continuing with the next coro from the ready queue, always
751 switch to the given coro object (regardless of priority etc.). The
752 readyness state of that coro isn't changed.
753 This is an advanced method for special cases - I'd love to hear
755 about any uses for this one.
756 $coro->cede_to
758 Like "schedule_to", but puts the current coro into the ready queue.
759 This has the effect of temporarily switching to the given coro, and
760 continuing some time later.
761 This is an advanced method for special cases - I'd love to hear
763 about any uses for this one.
764 $coro->throw ([$scalar])
766 If $throw is specified and defined, it will be thrown as an
767 exception inside the coro at the next convenient point in time.
768 Otherwise clears the exception object.
769 Coro will check for the exception each time a schedule-like-
771 function returns, i.e. after each "schedule", "cede",
772 "Coro::Semaphore->down", "Coro::Handle->readable" and so on. Most
773 of those functions (all that are part of Coro itself) detect this
774 case and return early in case an exception is pending.
775 The exception object will be thrown "as is" with the specified
777 scalar in $@, i.e. if it is a string, no line number or newline
778 will be appended (unlike with "die").
779 This can be used as a softer means than either "cancel" or
781 "safe_cancel "to ask a coro to end itself, although there is no
782 guarantee that the exception will lead to termination, and if the
783 exception isn't caught it might well end the whole program.
784 You might also think of "throw" as being the moral equivalent of
786 "kill"ing a coro with a signal (in this case, a scalar).
787 $coro->join
789 Wait until the coro terminates and return any values given to the
790 "terminate" or "cancel" functions. "join" can be called
791 concurrently from multiple threads, and all will be resumed and
792 given the status return once the $coro terminates.
793 $coro->on_destroy (\&cb)
795 Registers a callback that is called when this coro thread gets
796 destroyed, that is, after its resources have been freed but before
797 it is joined. The callback gets passed the terminate/cancel
798 arguments, if any, and must not die, under any circumstances.
799 There can be any number of "on_destroy" callbacks per coro, and
801 there is currently no way to remove a callback once added.
802 $oldprio = $coro->prio ($newprio)
804 Sets (or gets, if the argument is missing) the priority of the coro
805 thread. Higher priority coro get run before lower priority coros.
806 Priorities are small signed integers (currently -4 .. +3), that you
807 can refer to using PRIO_xxx constants (use the import tag :prio to
808 get then):
809 PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN
811 3 > 1 > 0 > -1 > -3 > -4
812 # set priority to HIGH
814 current->prio (PRIO_HIGH);
815 The idle coro thread ($Coro::idle) always has a lower priority than
817 any existing coro.
818 Changing the priority of the current coro will take effect
820 immediately, but changing the priority of a coro in the ready queue
821 (but not running) will only take effect after the next schedule (of
822 that coro). This is a bug that will be fixed in some future
823 version.
824 $newprio = $coro->nice ($change)
826 Similar to "prio", but subtract the given value from the priority
827 (i.e. higher values mean lower priority, just as in UNIX's nice
828 command).
829 $olddesc = $coro->desc ($newdesc)
831 Sets (or gets in case the argument is missing) the description for
832 this coro thread. This is just a free-form string you can associate
833 with a coro.
834 This method simply sets the "$coro->{desc}" member to the given
836 string. You can modify this member directly if you wish, and in
837 fact, this is often preferred to indicate major processing states
838 that can then be seen for example in a Coro::Debug session:
839 sub my_long_function {
841 local $Coro::current->{desc} = "now in my_long_function";
842 ...
843 $Coro::current->{desc} = "my_long_function: phase 1";
844 ...
845 $Coro::current->{desc} = "my_long_function: phase 2";
846 ...
847 }
848
850 Coro::nready
851 Returns the number of coro that are currently in the ready state,
852 i.e. that can be switched to by calling "schedule" directory or
853 indirectly. The value 0 means that the only runnable coro is the
854 currently running one, so "cede" would have no effect, and
855 "schedule" would cause a deadlock unless there is an idle handler
856 that wakes up some coro.
857 my $guard = Coro::guard { ... }
859 This function still exists, but is deprecated. Please use the
860 "Guard::guard" function instead.
861 unblock_sub { ... }
863 This utility function takes a BLOCK or code reference and
864 "unblocks" it, returning a new coderef. Unblocking means that
865 calling the new coderef will return immediately without blocking,
866 returning nothing, while the original code ref will be called (with
867 parameters) from within another coro.
868 The reason this function exists is that many event libraries (such
870 as the venerable Event module) are not thread-safe (a weaker form
871 of reentrancy). This means you must not block within event
872 callbacks, otherwise you might suffer from crashes or worse. The
873 only event library currently known that is safe to use without
874 "unblock_sub" is EV (but you might still run into deadlocks if all
875 event loops are blocked).
876 Coro will try to catch you when you block in the event loop
878 ("FATAL: $Coro::idle blocked itself"), but this is just best effort
879 and only works when you do not run your own event loop.
880 This function allows your callbacks to block by executing them in
882 another coro where it is safe to block. One example where blocking
883 is handy is when you use the Coro::AIO functions to save results to
884 disk, for example.
885 In short: simply use "unblock_sub { ... }" instead of "sub { ... }"
887 when creating event callbacks that want to block.
888 If your handler does not plan to block (e.g. simply sends a message
890 to another coro, or puts some other coro into the ready queue),
891 there is no reason to use "unblock_sub".
892 Note that you also need to use "unblock_sub" for any other
894 callbacks that are indirectly executed by any C-based event loop.
895 For example, when you use a module that uses AnyEvent (and you use
896 Coro::AnyEvent) and it provides callbacks that are the result of
897 some event callback, then you must not block either, or use
898 "unblock_sub".
899 $cb = rouse_cb
901 Create and return a "rouse callback". That's a code reference that,
902 when called, will remember a copy of its arguments and notify the
903 owner coro of the callback.
904 Only the first invocation will store agruments and signal any
906 waiter - further calls will effectively be ignored, but it is ok to
907 try.
908 Also see the next function.
910 @args = rouse_wait [$cb]
912 Wait for the specified rouse callback to be invoked (or if the
913 argument is missing, use the most recently created callback in the
914 current coro).
915 As soon as the callback is invoked (or when the callback was
917 invoked before "rouse_wait"), it will return the arguments
918 originally passed to the rouse callback. In scalar context, that
919 means you get the last argument, just as if "rouse_wait" had a
920 "return ($a1, $a2, $a3...)" statement at the end.
921 You are only allowed to wait once for a given rouse callback.
923 See the section HOW TO WAIT FOR A CALLBACK for an actual usage
925 example.
926 As of Coro 6.57, you can reliably wait for a rouse callback in a
928 different thread than from where it was created.
929
931 It is very common for a coro to wait for some callback to be called.
932 This occurs naturally when you use coro in an otherwise event-based
933 program, or when you use event-based libraries.
934 These typically register a callback for some event, and call that
936 callback when the event occurred. In a coro, however, you typically
937 want to just wait for the event, simplyifying things.
938 For example "AnyEvent->child" registers a callback to be called when a
940 specific child has exited:
941 my $child_watcher = AnyEvent->child (pid => $pid, cb => sub { ... });
943 But from within a coro, you often just want to write this:
945 my $status = wait_for_child $pid;
947 Coro offers two functions specifically designed to make this easy,
949 "rouse_cb" and "rouse_wait".
950 The first function, "rouse_cb", generates and returns a callback that,
952 when invoked, will save its arguments and notify the coro that created
953 the callback.
954 The second function, "rouse_wait", waits for the callback to be called
956 (by calling "schedule" to go to sleep) and returns the arguments
957 originally passed to the callback.
958 Using these functions, it becomes easy to write the "wait_for_child"
960 function mentioned above:
961 sub wait_for_child($) {
963 my ($pid) = @_;
964 my $watcher = AnyEvent->child (pid => $pid, cb => rouse_cb);
966 my ($rpid, $rstatus) = rouse_wait;
968 $rstatus
969 }
970 In the case where "rouse_cb" and "rouse_wait" are not flexible enough,
972 you can roll your own, using "schedule" and "ready":
973 sub wait_for_child($) {
975 my ($pid) = @_;
976 # store the current coro in $current,
978 # and provide result variables for the closure passed to ->child
979 my $current = $Coro::current;
980 my ($done, $rstatus);
981 # pass a closure to ->child
983 my $watcher = AnyEvent->child (pid => $pid, cb => sub {
984 $rstatus = $_[1]; # remember rstatus
985 $done = 1; # mark $rstatus as valid
986 $current->ready; # wake up the waiting thread
987 });
988 # wait until the closure has been called
990 schedule while !$done;
991 $rstatus
993 }
994
996 fork with pthread backend
997 When Coro is compiled using the pthread backend (which isn't
998 recommended but required on many BSDs as their libcs are completely
999 broken), then coro will not survive a fork. There is no known
1000 workaround except to fix your libc and use a saner backend.
1001 perl process emulation ("threads")
1003 This module is not perl-pseudo-thread-safe. You should only ever
1004 use this module from the first thread (this requirement might be
1005 removed in the future to allow per-thread schedulers, but
1006 Coro::State does not yet allow this). I recommend disabling thread
1007 support and using processes, as having the windows process
1008 emulation enabled under unix roughly halves perl performance, even
1009 when not used.
1010 Attempts to use threads created in another emulated process will
1012 crash ("cleanly", with a null pointer exception).
1013 coro switching is not signal safe
1015 You must not switch to another coro from within a signal handler
1016 (only relevant with %SIG - most event libraries provide safe
1017 signals), unless you are sure you are not interrupting a Coro
1018 function.
1019 That means you MUST NOT call any function that might "block" the
1021 current coro - "cede", "schedule" "Coro::Semaphore->down" or
1022 anything that calls those. Everything else, including calling
1023 "ready", works.
1024
1026 A great many people seem to be confused about ithreads (for example,
1027 Chip Salzenberg called me unintelligent, incapable, stupid and
1028 gullible, while in the same mail making rather confused statements
1029 about perl ithreads (for example, that memory or files would be
1030 shared), showing his lack of understanding of this area - if it is hard
1031 to understand for Chip, it is probably not obvious to everybody).
1032 What follows is an ultra-condensed version of my talk about threads in
1034 scripting languages given on the perl workshop 2009:
1035 The so-called "ithreads" were originally implemented for two reasons:
1037 first, to (badly) emulate unix processes on native win32 perls, and
1038 secondly, to replace the older, real thread model ("5.005-threads").
1039 It does that by using threads instead of OS processes. The difference
1041 between processes and threads is that threads share memory (and other
1042 state, such as files) between threads within a single process, while
1043 processes do not share anything (at least not semantically). That means
1044 that modifications done by one thread are seen by others, while
1045 modifications by one process are not seen by other processes.
1046 The "ithreads" work exactly like that: when creating a new ithreads
1048 process, all state is copied (memory is copied physically, files and
1049 code is copied logically). Afterwards, it isolates all modifications.
1050 On UNIX, the same behaviour can be achieved by using operating system
1051 processes, except that UNIX typically uses hardware built into the
1052 system to do this efficiently, while the windows process emulation
1053 emulates this hardware in software (rather efficiently, but of course
1054 it is still much slower than dedicated hardware).
1055 As mentioned before, loading code, modifying code, modifying data
1057 structures and so on is only visible in the ithreads process doing the
1058 modification, not in other ithread processes within the same OS
1059 process.
1060 This is why "ithreads" do not implement threads for perl at all, only
1062 processes. What makes it so bad is that on non-windows platforms, you
1063 can actually take advantage of custom hardware for this purpose (as
1064 evidenced by the forks module, which gives you the (i-) threads API,
1065 just much faster).
1066 Sharing data is in the i-threads model is done by transferring data
1068 structures between threads using copying semantics, which is very slow
1069 - shared data simply does not exist. Benchmarks using i-threads which
1070 are communication-intensive show extremely bad behaviour with i-threads
1071 (in fact, so bad that Coro, which cannot take direct advantage of
1072 multiple CPUs, is often orders of magnitude faster because it shares
1073 data using real threads, refer to my talk for details).
1074 As summary, i-threads *use* threads to implement processes, while the
1076 compatible forks module *uses* processes to emulate, uhm, processes.
1077 I-threads slow down every perl program when enabled, and outside of
1078 windows, serve no (or little) practical purpose, but disadvantages
1079 every single-threaded Perl program.
1080 This is the reason that I try to avoid the name "ithreads", as it is
1082 misleading as it implies that it implements some kind of thread model
1083 for perl, and prefer the name "windows process emulation", which
1084 describes the actual use and behaviour of it much better.
1085
1087 Event-Loop integration: Coro::AnyEvent, Coro::EV, Coro::Event.
1088 Debugging: Coro::Debug.
1090 Support/Utility: Coro::Specific, Coro::Util.
1092 Locking and IPC: Coro::Signal, Coro::Channel, Coro::Semaphore,
1094 Coro::SemaphoreSet, Coro::RWLock.
1095 I/O and Timers: Coro::Timer, Coro::Handle, Coro::Socket, Coro::AIO.
1097 Compatibility with other modules: Coro::LWP (but see also
1099 AnyEvent::HTTP for a better-working alternative), Coro::BDB,
1100 Coro::Storable, Coro::Select.
1101 XS API: Coro::MakeMaker.
1103 Low level Configuration, Thread Environment, Continuations:
1105 Coro::State.
1106
1108 Marc A. Lehmann <schmorp@schmorp.de>
1109 http://software.schmorp.de/pkg/Coro.html
1110perl v5.38.0 2023-07-20 Coro(3)