1 |
=head1 NAME |
2 |
|
3 |
Coro - coroutine process abstraction |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use Coro; |
8 |
|
9 |
async { |
10 |
# some asynchronous thread of execution |
11 |
print "2\n"; |
12 |
cede; # yield back to main |
13 |
print "4\n"; |
14 |
}; |
15 |
print "1\n"; |
16 |
cede; # yield to coroutine |
17 |
print "3\n"; |
18 |
cede; # and again |
19 |
|
20 |
# use locking |
21 |
my $lock = new Coro::Semaphore; |
22 |
my $locked; |
23 |
|
24 |
$lock->down; |
25 |
$locked = 1; |
26 |
$lock->up; |
27 |
|
28 |
=head1 DESCRIPTION |
29 |
|
30 |
This module collection manages coroutines. Coroutines are similar |
31 |
to threads but don't run in parallel at the same time even on SMP |
32 |
machines. The specific flavor of coroutine used in this module also |
33 |
guarantees you that it will not switch between coroutines unless |
34 |
necessary, at easily-identified points in your program, so locking and |
35 |
parallel access are rarely an issue, making coroutine programming much |
36 |
safer than threads programming. |
37 |
|
38 |
(Perl, however, does not natively support real threads but instead does a |
39 |
very slow and memory-intensive emulation of processes using threads. This |
40 |
is a performance win on Windows machines, and a loss everywhere else). |
41 |
|
42 |
In this module, coroutines are defined as "callchain + lexical variables + |
43 |
@_ + $_ + $@ + $/ + C stack), that is, a coroutine has its own callchain, |
44 |
its own set of lexicals and its own set of perls most important global |
45 |
variables (see L<Coro::State> for more configuration). |
46 |
|
47 |
=cut |
48 |
|
49 |
package Coro; |
50 |
|
51 |
use strict; |
52 |
no warnings "uninitialized"; |
53 |
|
54 |
use Coro::State; |
55 |
|
56 |
use base qw(Coro::State Exporter); |
57 |
|
58 |
our $idle; # idle handler |
59 |
our $main; # main coroutine |
60 |
our $current; # current coroutine |
61 |
|
62 |
our $VERSION = 4.6; |
63 |
|
64 |
our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub); |
65 |
our %EXPORT_TAGS = ( |
66 |
prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)], |
67 |
); |
68 |
our @EXPORT_OK = (@{$EXPORT_TAGS{prio}}, qw(nready)); |
69 |
|
70 |
=over 4 |
71 |
|
72 |
=item $main |
73 |
|
74 |
This coroutine represents the main program. |
75 |
|
76 |
=cut |
77 |
|
78 |
$main = new Coro; |
79 |
|
80 |
=item $current (or as function: current) |
81 |
|
82 |
The current coroutine (the last coroutine switched to). The initial value |
83 |
is C<$main> (of course). |
84 |
|
85 |
This variable is B<strictly> I<read-only>. It is provided for performance |
86 |
reasons. If performance is not essential you are encouraged to use the |
87 |
C<Coro::current> function instead. |
88 |
|
89 |
=cut |
90 |
|
91 |
$main->{desc} = "[main::]"; |
92 |
|
93 |
# maybe some other module used Coro::Specific before... |
94 |
$main->{_specific} = $current->{_specific} |
95 |
if $current; |
96 |
|
97 |
_set_current $main; |
98 |
|
99 |
sub current() { $current } |
100 |
|
101 |
=item $idle |
102 |
|
103 |
A callback that is called whenever the scheduler finds no ready coroutines |
104 |
to run. The default implementation prints "FATAL: deadlock detected" and |
105 |
exits, because the program has no other way to continue. |
106 |
|
107 |
This hook is overwritten by modules such as C<Coro::Timer> and |
108 |
C<Coro::Event> to wait on an external event that hopefully wake up a |
109 |
coroutine so the scheduler can run it. |
110 |
|
111 |
Please note that if your callback recursively invokes perl (e.g. for event |
112 |
handlers), then it must be prepared to be called recursively itself. |
113 |
|
114 |
=cut |
115 |
|
116 |
$idle = sub { |
117 |
require Carp; |
118 |
Carp::croak ("FATAL: deadlock detected"); |
119 |
}; |
120 |
|
121 |
sub _cancel { |
122 |
my ($self) = @_; |
123 |
|
124 |
# free coroutine data and mark as destructed |
125 |
$self->_destroy |
126 |
or return; |
127 |
|
128 |
# call all destruction callbacks |
129 |
$_->(@{$self->{_status}}) |
130 |
for @{(delete $self->{_on_destroy}) || []}; |
131 |
} |
132 |
|
133 |
# this coroutine is necessary because a coroutine |
134 |
# cannot destroy itself. |
135 |
my @destroy; |
136 |
my $manager; |
137 |
|
138 |
$manager = new Coro sub { |
139 |
while () { |
140 |
(shift @destroy)->_cancel |
141 |
while @destroy; |
142 |
|
143 |
&schedule; |
144 |
} |
145 |
}; |
146 |
$manager->desc ("[coro manager]"); |
147 |
$manager->prio (PRIO_MAX); |
148 |
|
149 |
=back |
150 |
|
151 |
=head2 STATIC METHODS |
152 |
|
153 |
Static methods are actually functions that operate on the current coroutine only. |
154 |
|
155 |
=over 4 |
156 |
|
157 |
=item async { ... } [@args...] |
158 |
|
159 |
Create a new asynchronous coroutine and return it's coroutine object |
160 |
(usually unused). When the sub returns the new coroutine is automatically |
161 |
terminated. |
162 |
|
163 |
See the C<Coro::State::new> constructor for info about the coroutine |
164 |
environment in which coroutines run. |
165 |
|
166 |
Calling C<exit> in a coroutine will do the same as calling exit outside |
167 |
the coroutine. Likewise, when the coroutine dies, the program will exit, |
168 |
just as it would in the main program. |
169 |
|
170 |
# create a new coroutine that just prints its arguments |
171 |
async { |
172 |
print "@_\n"; |
173 |
} 1,2,3,4; |
174 |
|
175 |
=cut |
176 |
|
177 |
sub async(&@) { |
178 |
my $coro = new Coro @_; |
179 |
$coro->ready; |
180 |
$coro |
181 |
} |
182 |
|
183 |
=item async_pool { ... } [@args...] |
184 |
|
185 |
Similar to C<async>, but uses a coroutine pool, so you should not call |
186 |
terminate or join (although you are allowed to), and you get a coroutine |
187 |
that might have executed other code already (which can be good or bad :). |
188 |
|
189 |
Also, the block is executed in an C<eval> context and a warning will be |
190 |
issued in case of an exception instead of terminating the program, as |
191 |
C<async> does. As the coroutine is being reused, stuff like C<on_destroy> |
192 |
will not work in the expected way, unless you call terminate or cancel, |
193 |
which somehow defeats the purpose of pooling. |
194 |
|
195 |
The priority will be reset to C<0> after each job, tracing will be |
196 |
disabled, the description will be reset and the default output filehandle |
197 |
gets restored, so you can change alkl these. Otherwise the coroutine will |
198 |
be re-used "as-is": most notably if you change other per-coroutine global |
199 |
stuff such as C<$/> you need to revert that change, which is most simply |
200 |
done by using local as in C< local $/ >. |
201 |
|
202 |
The pool size is limited to 8 idle coroutines (this can be adjusted by |
203 |
changing $Coro::POOL_SIZE), and there can be as many non-idle coros as |
204 |
required. |
205 |
|
206 |
If you are concerned about pooled coroutines growing a lot because a |
207 |
single C<async_pool> used a lot of stackspace you can e.g. C<async_pool |
208 |
{ terminate }> once per second or so to slowly replenish the pool. In |
209 |
addition to that, when the stacks used by a handler grows larger than 16kb |
210 |
(adjustable with $Coro::POOL_RSS) it will also exit. |
211 |
|
212 |
=cut |
213 |
|
214 |
our $POOL_SIZE = 8; |
215 |
our $POOL_RSS = 16 * 1024; |
216 |
our @async_pool; |
217 |
|
218 |
sub pool_handler { |
219 |
my $cb; |
220 |
|
221 |
while () { |
222 |
eval { |
223 |
while () { |
224 |
_pool_1 $cb; |
225 |
&$cb; |
226 |
_pool_2 $cb; |
227 |
&schedule; |
228 |
} |
229 |
}; |
230 |
|
231 |
last if $@ eq "\3async_pool terminate\2\n"; |
232 |
warn $@ if $@; |
233 |
} |
234 |
} |
235 |
|
236 |
sub async_pool(&@) { |
237 |
# this is also inlined into the unlock_scheduler |
238 |
my $coro = (pop @async_pool) || new Coro \&pool_handler; |
239 |
|
240 |
$coro->{_invoke} = [@_]; |
241 |
$coro->ready; |
242 |
|
243 |
$coro |
244 |
} |
245 |
|
246 |
=item schedule |
247 |
|
248 |
Calls the scheduler. Please note that the current coroutine will not be put |
249 |
into the ready queue, so calling this function usually means you will |
250 |
never be called again unless something else (e.g. an event handler) calls |
251 |
ready. |
252 |
|
253 |
The canonical way to wait on external events is this: |
254 |
|
255 |
{ |
256 |
# remember current coroutine |
257 |
my $current = $Coro::current; |
258 |
|
259 |
# register a hypothetical event handler |
260 |
on_event_invoke sub { |
261 |
# wake up sleeping coroutine |
262 |
$current->ready; |
263 |
undef $current; |
264 |
}; |
265 |
|
266 |
# call schedule until event occurred. |
267 |
# in case we are woken up for other reasons |
268 |
# (current still defined), loop. |
269 |
Coro::schedule while $current; |
270 |
} |
271 |
|
272 |
=item cede |
273 |
|
274 |
"Cede" to other coroutines. This function puts the current coroutine into the |
275 |
ready queue and calls C<schedule>, which has the effect of giving up the |
276 |
current "timeslice" to other coroutines of the same or higher priority. |
277 |
|
278 |
=item Coro::cede_notself |
279 |
|
280 |
Works like cede, but is not exported by default and will cede to any |
281 |
coroutine, regardless of priority, once. |
282 |
|
283 |
=item terminate [arg...] |
284 |
|
285 |
Terminates the current coroutine with the given status values (see L<cancel>). |
286 |
|
287 |
=item killall |
288 |
|
289 |
Kills/terminates/cancels all coroutines except the currently running |
290 |
one. This is useful after a fork, either in the child or the parent, as |
291 |
usually only one of them should inherit the running coroutines. |
292 |
|
293 |
=cut |
294 |
|
295 |
sub terminate { |
296 |
$current->cancel (@_); |
297 |
} |
298 |
|
299 |
sub killall { |
300 |
for (Coro::State::list) { |
301 |
$_->cancel |
302 |
if $_ != $current && UNIVERSAL::isa $_, "Coro"; |
303 |
} |
304 |
} |
305 |
|
306 |
=back |
307 |
|
308 |
=head2 COROUTINE METHODS |
309 |
|
310 |
These are the methods you can call on coroutine objects. |
311 |
|
312 |
=over 4 |
313 |
|
314 |
=item new Coro \&sub [, @args...] |
315 |
|
316 |
Create a new coroutine and return it. When the sub returns the coroutine |
317 |
automatically terminates as if C<terminate> with the returned values were |
318 |
called. To make the coroutine run you must first put it into the ready queue |
319 |
by calling the ready method. |
320 |
|
321 |
See C<async> and C<Coro::State::new> for additional info about the |
322 |
coroutine environment. |
323 |
|
324 |
=cut |
325 |
|
326 |
sub _run_coro { |
327 |
terminate &{+shift}; |
328 |
} |
329 |
|
330 |
sub new { |
331 |
my $class = shift; |
332 |
|
333 |
$class->SUPER::new (\&_run_coro, @_) |
334 |
} |
335 |
|
336 |
=item $success = $coroutine->ready |
337 |
|
338 |
Put the given coroutine into the ready queue (according to it's priority) |
339 |
and return true. If the coroutine is already in the ready queue, do nothing |
340 |
and return false. |
341 |
|
342 |
=item $is_ready = $coroutine->is_ready |
343 |
|
344 |
Return wether the coroutine is currently the ready queue or not, |
345 |
|
346 |
=item $coroutine->cancel (arg...) |
347 |
|
348 |
Terminates the given coroutine and makes it return the given arguments as |
349 |
status (default: the empty list). Never returns if the coroutine is the |
350 |
current coroutine. |
351 |
|
352 |
=cut |
353 |
|
354 |
sub cancel { |
355 |
my $self = shift; |
356 |
$self->{_status} = [@_]; |
357 |
|
358 |
if ($current == $self) { |
359 |
push @destroy, $self; |
360 |
$manager->ready; |
361 |
&schedule while 1; |
362 |
} else { |
363 |
$self->_cancel; |
364 |
} |
365 |
} |
366 |
|
367 |
=item $coroutine->join |
368 |
|
369 |
Wait until the coroutine terminates and return any values given to the |
370 |
C<terminate> or C<cancel> functions. C<join> can be called concurrently |
371 |
from multiple coroutines. |
372 |
|
373 |
=cut |
374 |
|
375 |
sub join { |
376 |
my $self = shift; |
377 |
|
378 |
unless ($self->{_status}) { |
379 |
my $current = $current; |
380 |
|
381 |
push @{$self->{_on_destroy}}, sub { |
382 |
$current->ready; |
383 |
undef $current; |
384 |
}; |
385 |
|
386 |
&schedule while $current; |
387 |
} |
388 |
|
389 |
wantarray ? @{$self->{_status}} : $self->{_status}[0]; |
390 |
} |
391 |
|
392 |
=item $coroutine->on_destroy (\&cb) |
393 |
|
394 |
Registers a callback that is called when this coroutine gets destroyed, |
395 |
but before it is joined. The callback gets passed the terminate arguments, |
396 |
if any. |
397 |
|
398 |
=cut |
399 |
|
400 |
sub on_destroy { |
401 |
my ($self, $cb) = @_; |
402 |
|
403 |
push @{ $self->{_on_destroy} }, $cb; |
404 |
} |
405 |
|
406 |
=item $oldprio = $coroutine->prio ($newprio) |
407 |
|
408 |
Sets (or gets, if the argument is missing) the priority of the |
409 |
coroutine. Higher priority coroutines get run before lower priority |
410 |
coroutines. Priorities are small signed integers (currently -4 .. +3), |
411 |
that you can refer to using PRIO_xxx constants (use the import tag :prio |
412 |
to get then): |
413 |
|
414 |
PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN |
415 |
3 > 1 > 0 > -1 > -3 > -4 |
416 |
|
417 |
# set priority to HIGH |
418 |
current->prio(PRIO_HIGH); |
419 |
|
420 |
The idle coroutine ($Coro::idle) always has a lower priority than any |
421 |
existing coroutine. |
422 |
|
423 |
Changing the priority of the current coroutine will take effect immediately, |
424 |
but changing the priority of coroutines in the ready queue (but not |
425 |
running) will only take effect after the next schedule (of that |
426 |
coroutine). This is a bug that will be fixed in some future version. |
427 |
|
428 |
=item $newprio = $coroutine->nice ($change) |
429 |
|
430 |
Similar to C<prio>, but subtract the given value from the priority (i.e. |
431 |
higher values mean lower priority, just as in unix). |
432 |
|
433 |
=item $olddesc = $coroutine->desc ($newdesc) |
434 |
|
435 |
Sets (or gets in case the argument is missing) the description for this |
436 |
coroutine. This is just a free-form string you can associate with a coroutine. |
437 |
|
438 |
This method simply sets the C<< $coroutine->{desc} >> member to the given string. You |
439 |
can modify this member directly if you wish. |
440 |
|
441 |
=item $coroutine->throw ([$scalar]) |
442 |
|
443 |
If C<$throw> is specified and defined, it will be thrown as an exception |
444 |
inside the coroutine at the next convinient point in time (usually after |
445 |
it gains control at the next schedule/transfer/cede). Otherwise clears the |
446 |
exception object. |
447 |
|
448 |
The exception object will be thrown "as is" with the specified scalar in |
449 |
C<$@>, i.e. if it is a string, no line number or newline will be appended |
450 |
(unlike with C<die>). |
451 |
|
452 |
This can be used as a softer means than C<cancel> to ask a coroutine to |
453 |
end itself, although there is no guarentee that the exception will lead to |
454 |
termination, and if the exception isn't caught it might well end the whole |
455 |
program. |
456 |
|
457 |
=cut |
458 |
|
459 |
sub desc { |
460 |
my $old = $_[0]{desc}; |
461 |
$_[0]{desc} = $_[1] if @_ > 1; |
462 |
$old; |
463 |
} |
464 |
|
465 |
=back |
466 |
|
467 |
=head2 GLOBAL FUNCTIONS |
468 |
|
469 |
=over 4 |
470 |
|
471 |
=item Coro::nready |
472 |
|
473 |
Returns the number of coroutines that are currently in the ready state, |
474 |
i.e. that can be switched to. The value C<0> means that the only runnable |
475 |
coroutine is the currently running one, so C<cede> would have no effect, |
476 |
and C<schedule> would cause a deadlock unless there is an idle handler |
477 |
that wakes up some coroutines. |
478 |
|
479 |
=item my $guard = Coro::guard { ... } |
480 |
|
481 |
This creates and returns a guard object. Nothing happens until the object |
482 |
gets destroyed, in which case the codeblock given as argument will be |
483 |
executed. This is useful to free locks or other resources in case of a |
484 |
runtime error or when the coroutine gets canceled, as in both cases the |
485 |
guard block will be executed. The guard object supports only one method, |
486 |
C<< ->cancel >>, which will keep the codeblock from being executed. |
487 |
|
488 |
Example: set some flag and clear it again when the coroutine gets canceled |
489 |
or the function returns: |
490 |
|
491 |
sub do_something { |
492 |
my $guard = Coro::guard { $busy = 0 }; |
493 |
$busy = 1; |
494 |
|
495 |
# do something that requires $busy to be true |
496 |
} |
497 |
|
498 |
=cut |
499 |
|
500 |
sub guard(&) { |
501 |
bless \(my $cb = $_[0]), "Coro::guard" |
502 |
} |
503 |
|
504 |
sub Coro::guard::cancel { |
505 |
${$_[0]} = sub { }; |
506 |
} |
507 |
|
508 |
sub Coro::guard::DESTROY { |
509 |
${$_[0]}->(); |
510 |
} |
511 |
|
512 |
|
513 |
=item unblock_sub { ... } |
514 |
|
515 |
This utility function takes a BLOCK or code reference and "unblocks" it, |
516 |
returning the new coderef. This means that the new coderef will return |
517 |
immediately without blocking, returning nothing, while the original code |
518 |
ref will be called (with parameters) from within its own coroutine. |
519 |
|
520 |
The reason this function exists is that many event libraries (such as the |
521 |
venerable L<Event|Event> module) are not coroutine-safe (a weaker form |
522 |
of thread-safety). This means you must not block within event callbacks, |
523 |
otherwise you might suffer from crashes or worse. |
524 |
|
525 |
This function allows your callbacks to block by executing them in another |
526 |
coroutine where it is safe to block. One example where blocking is handy |
527 |
is when you use the L<Coro::AIO|Coro::AIO> functions to save results to |
528 |
disk. |
529 |
|
530 |
In short: simply use C<unblock_sub { ... }> instead of C<sub { ... }> when |
531 |
creating event callbacks that want to block. |
532 |
|
533 |
=cut |
534 |
|
535 |
our @unblock_queue; |
536 |
|
537 |
# we create a special coro because we want to cede, |
538 |
# to reduce pressure on the coro pool (because most callbacks |
539 |
# return immediately and can be reused) and because we cannot cede |
540 |
# inside an event callback. |
541 |
our $unblock_scheduler = new Coro sub { |
542 |
while () { |
543 |
while (my $cb = pop @unblock_queue) { |
544 |
# this is an inlined copy of async_pool |
545 |
my $coro = (pop @async_pool) || new Coro \&pool_handler; |
546 |
|
547 |
$coro->{_invoke} = $cb; |
548 |
$coro->ready; |
549 |
cede; # for short-lived callbacks, this reduces pressure on the coro pool |
550 |
} |
551 |
schedule; # sleep well |
552 |
} |
553 |
}; |
554 |
$unblock_scheduler->desc ("[unblock_sub scheduler]"); |
555 |
|
556 |
sub unblock_sub(&) { |
557 |
my $cb = shift; |
558 |
|
559 |
sub { |
560 |
unshift @unblock_queue, [$cb, @_]; |
561 |
$unblock_scheduler->ready; |
562 |
} |
563 |
} |
564 |
|
565 |
=back |
566 |
|
567 |
=cut |
568 |
|
569 |
1; |
570 |
|
571 |
=head1 BUGS/LIMITATIONS |
572 |
|
573 |
- you must make very sure that no coro is still active on global |
574 |
destruction. very bad things might happen otherwise (usually segfaults). |
575 |
|
576 |
- this module is not thread-safe. You should only ever use this module |
577 |
from the same thread (this requirement might be loosened in the future |
578 |
to allow per-thread schedulers, but Coro::State does not yet allow |
579 |
this). |
580 |
|
581 |
=head1 SEE ALSO |
582 |
|
583 |
Lower level Configuration, Coroutine Environment: L<Coro::State>. |
584 |
|
585 |
Debugging: L<Coro::Debug>. |
586 |
|
587 |
Support/Utility: L<Coro::Specific>, L<Coro::Util>. |
588 |
|
589 |
Locking/IPC: L<Coro::Signal>, L<Coro::Channel>, L<Coro::Semaphore>, L<Coro::SemaphoreSet>, L<Coro::RWLock>. |
590 |
|
591 |
Event/IO: L<Coro::Timer>, L<Coro::Event>, L<Coro::Handle>, L<Coro::Socket>. |
592 |
|
593 |
Compatibility: L<Coro::LWP>, L<Coro::Storable>, L<Coro::Select>. |
594 |
|
595 |
Embedding: L<Coro::MakeMaker>. |
596 |
|
597 |
=head1 AUTHOR |
598 |
|
599 |
Marc Lehmann <schmorp@schmorp.de> |
600 |
http://home.schmorp.de/ |
601 |
|
602 |
=cut |
603 |
|