1 |
=head1 NAME |
2 |
|
3 |
Coro - the only real threads in perl |
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 coro |
17 |
print "3\n"; |
18 |
cede; # and again |
19 |
|
20 |
# use locking |
21 |
use Coro::Semaphore; |
22 |
my $lock = new Coro::Semaphore; |
23 |
my $locked; |
24 |
|
25 |
$lock->down; |
26 |
$locked = 1; |
27 |
$lock->up; |
28 |
|
29 |
=head1 DESCRIPTION |
30 |
|
31 |
For a tutorial-style introduction, please read the L<Coro::Intro> |
32 |
manpage. This manpage mainly contains reference information. |
33 |
|
34 |
This module collection manages continuations in general, most often in |
35 |
the form of cooperative threads (also called coros, or simply "coro" |
36 |
in 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 that |
39 |
it will not switch between threads unless necessary, at easily-identified |
40 |
points in your program, so locking and parallel access are rarely an |
41 |
issue, making thread programming much safer and easier than using other |
42 |
thread models. |
43 |
|
44 |
Unlike the so-called "Perl threads" (which are not actually real threads |
45 |
but only the windows process emulation (see section of same name for more |
46 |
details) ported to unix, and as such act as processes), Coro provides |
47 |
a full shared address space, which makes communication between threads |
48 |
very easy. And Coro's threads are fast, too: disabling the Windows |
49 |
process emulation code in your perl and using Coro can easily result in |
50 |
a two to four times speed increase for your programs. A parallel matrix |
51 |
multiplication benchmark runs over 300 times faster on a single core than |
52 |
perl's pseudo-threads on a quad core using all four cores. |
53 |
|
54 |
Coro achieves that by supporting multiple running interpreters that share |
55 |
data, which is especially useful to code pseudo-parallel processes and |
56 |
for event-based programming, such as multiple HTTP-GET requests running |
57 |
concurrently. See L<Coro::AnyEvent> to learn more on how to integrate Coro |
58 |
into an event-based environment. |
59 |
|
60 |
In this module, a thread is defined as "callchain + lexical variables + |
61 |
some package variables + C stack), that is, a thread has its own callchain, |
62 |
its own set of lexicals and its own set of perls most important global |
63 |
variables (see L<Coro::State> for more configuration and background info). |
64 |
|
65 |
See also the C<SEE ALSO> section at the end of this document - the Coro |
66 |
module family is quite large. |
67 |
|
68 |
=cut |
69 |
|
70 |
package Coro; |
71 |
|
72 |
use common::sense; |
73 |
|
74 |
use Carp (); |
75 |
|
76 |
use Guard (); |
77 |
|
78 |
use Coro::State; |
79 |
|
80 |
use base qw(Coro::State Exporter); |
81 |
|
82 |
our $idle; # idle handler |
83 |
our $main; # main coro |
84 |
our $current; # current coro |
85 |
|
86 |
our $VERSION = 5.25; |
87 |
|
88 |
our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub rouse_cb rouse_wait); |
89 |
our %EXPORT_TAGS = ( |
90 |
prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)], |
91 |
); |
92 |
our @EXPORT_OK = (@{$EXPORT_TAGS{prio}}, qw(nready)); |
93 |
|
94 |
=head1 GLOBAL VARIABLES |
95 |
|
96 |
=over 4 |
97 |
|
98 |
=item $Coro::main |
99 |
|
100 |
This variable stores the Coro object that represents the main |
101 |
program. While you cna C<ready> it and do most other things you can do to |
102 |
coro, it is mainly useful to compare again C<$Coro::current>, to see |
103 |
whether you are running in the main program or not. |
104 |
|
105 |
=cut |
106 |
|
107 |
# $main is now being initialised by Coro::State |
108 |
|
109 |
=item $Coro::current |
110 |
|
111 |
The Coro object representing the current coro (the last |
112 |
coro that the Coro scheduler switched to). The initial value is |
113 |
C<$Coro::main> (of course). |
114 |
|
115 |
This variable is B<strictly> I<read-only>. You can take copies of the |
116 |
value stored in it and use it as any other Coro object, but you must |
117 |
not otherwise modify the variable itself. |
118 |
|
119 |
=cut |
120 |
|
121 |
sub current() { $current } # [DEPRECATED] |
122 |
|
123 |
=item $Coro::idle |
124 |
|
125 |
This variable is mainly useful to integrate Coro into event loops. It is |
126 |
usually better to rely on L<Coro::AnyEvent> or L<Coro::EV>, as this is |
127 |
pretty low-level functionality. |
128 |
|
129 |
This variable stores a Coro object that is put into the ready queue when |
130 |
there are no other ready threads (without invoking any ready hooks). |
131 |
|
132 |
The default implementation dies with "FATAL: deadlock detected.", followed |
133 |
by a thread listing, because the program has no other way to continue. |
134 |
|
135 |
This hook is overwritten by modules such as C<Coro::EV> and |
136 |
C<Coro::AnyEvent> to wait on an external event that hopefully wake up a |
137 |
coro so the scheduler can run it. |
138 |
|
139 |
See L<Coro::EV> or L<Coro::AnyEvent> for examples of using this technique. |
140 |
|
141 |
=cut |
142 |
|
143 |
# ||= because other modules could have provided their own by now |
144 |
$idle ||= new Coro sub { |
145 |
require Coro::Debug; |
146 |
die "FATAL: deadlock detected.\n" |
147 |
. Coro::Debug::ps_listing (); |
148 |
}; |
149 |
|
150 |
# this coro is necessary because a coro |
151 |
# cannot destroy itself. |
152 |
our @destroy; |
153 |
our $manager; |
154 |
|
155 |
$manager = new Coro sub { |
156 |
while () { |
157 |
Coro::State::cancel shift @destroy |
158 |
while @destroy; |
159 |
|
160 |
&schedule; |
161 |
} |
162 |
}; |
163 |
$manager->{desc} = "[coro manager]"; |
164 |
$manager->prio (PRIO_MAX); |
165 |
|
166 |
=back |
167 |
|
168 |
=head1 SIMPLE CORO CREATION |
169 |
|
170 |
=over 4 |
171 |
|
172 |
=item async { ... } [@args...] |
173 |
|
174 |
Create a new coro and return its Coro object (usually |
175 |
unused). The coro will be put into the ready queue, so |
176 |
it will start running automatically on the next scheduler run. |
177 |
|
178 |
The first argument is a codeblock/closure that should be executed in the |
179 |
coro. When it returns argument returns the coro is automatically |
180 |
terminated. |
181 |
|
182 |
The remaining arguments are passed as arguments to the closure. |
183 |
|
184 |
See the C<Coro::State::new> constructor for info about the coro |
185 |
environment in which coro are executed. |
186 |
|
187 |
Calling C<exit> in a coro will do the same as calling exit outside |
188 |
the coro. Likewise, when the coro dies, the program will exit, |
189 |
just as it would in the main program. |
190 |
|
191 |
If you do not want that, you can provide a default C<die> handler, or |
192 |
simply avoid dieing (by use of C<eval>). |
193 |
|
194 |
Example: Create a new coro that just prints its arguments. |
195 |
|
196 |
async { |
197 |
print "@_\n"; |
198 |
} 1,2,3,4; |
199 |
|
200 |
=item async_pool { ... } [@args...] |
201 |
|
202 |
Similar to C<async>, but uses a coro pool, so you should not call |
203 |
terminate or join on it (although you are allowed to), and you get a |
204 |
coro that might have executed other code already (which can be good |
205 |
or bad :). |
206 |
|
207 |
On the plus side, this function is about twice as fast as creating (and |
208 |
destroying) a completely new coro, so if you need a lot of generic |
209 |
coros in quick successsion, use C<async_pool>, not C<async>. |
210 |
|
211 |
The code block is executed in an C<eval> context and a warning will be |
212 |
issued in case of an exception instead of terminating the program, as |
213 |
C<async> does. As the coro is being reused, stuff like C<on_destroy> |
214 |
will not work in the expected way, unless you call terminate or cancel, |
215 |
which somehow defeats the purpose of pooling (but is fine in the |
216 |
exceptional case). |
217 |
|
218 |
The priority will be reset to C<0> after each run, tracing will be |
219 |
disabled, the description will be reset and the default output filehandle |
220 |
gets restored, so you can change all these. Otherwise the coro will |
221 |
be re-used "as-is": most notably if you change other per-coro global |
222 |
stuff such as C<$/> you I<must needs> revert that change, which is most |
223 |
simply done by using local as in: C<< local $/ >>. |
224 |
|
225 |
The idle pool size is limited to C<8> idle coros (this can be |
226 |
adjusted by changing $Coro::POOL_SIZE), but there can be as many non-idle |
227 |
coros as required. |
228 |
|
229 |
If you are concerned about pooled coros growing a lot because a |
230 |
single C<async_pool> used a lot of stackspace you can e.g. C<async_pool |
231 |
{ terminate }> once per second or so to slowly replenish the pool. In |
232 |
addition to that, when the stacks used by a handler grows larger than 32kb |
233 |
(adjustable via $Coro::POOL_RSS) it will also be destroyed. |
234 |
|
235 |
=cut |
236 |
|
237 |
our $POOL_SIZE = 8; |
238 |
our $POOL_RSS = 32 * 1024; |
239 |
our @async_pool; |
240 |
|
241 |
sub pool_handler { |
242 |
while () { |
243 |
eval { |
244 |
&{&_pool_handler} while 1; |
245 |
}; |
246 |
|
247 |
warn $@ if $@; |
248 |
} |
249 |
} |
250 |
|
251 |
=back |
252 |
|
253 |
=head1 STATIC METHODS |
254 |
|
255 |
Static methods are actually functions that implicitly operate on the |
256 |
current coro. |
257 |
|
258 |
=over 4 |
259 |
|
260 |
=item schedule |
261 |
|
262 |
Calls the scheduler. The scheduler will find the next coro that is |
263 |
to be run from the ready queue and switches to it. The next coro |
264 |
to be run is simply the one with the highest priority that is longest |
265 |
in its ready queue. If there is no coro ready, it will call the |
266 |
C<$Coro::idle> hook. |
267 |
|
268 |
Please note that the current coro will I<not> be put into the ready |
269 |
queue, so calling this function usually means you will never be called |
270 |
again unless something else (e.g. an event handler) calls C<< ->ready >>, |
271 |
thus waking you up. |
272 |
|
273 |
This makes C<schedule> I<the> generic method to use to block the current |
274 |
coro and wait for events: first you remember the current coro in |
275 |
a variable, then arrange for some callback of yours to call C<< ->ready |
276 |
>> on that once some event happens, and last you call C<schedule> to put |
277 |
yourself to sleep. Note that a lot of things can wake your coro up, |
278 |
so you need to check whether the event indeed happened, e.g. by storing the |
279 |
status in a variable. |
280 |
|
281 |
See B<HOW TO WAIT FOR A CALLBACK>, below, for some ways to wait for callbacks. |
282 |
|
283 |
=item cede |
284 |
|
285 |
"Cede" to other coros. This function puts the current coro into |
286 |
the ready queue and calls C<schedule>, which has the effect of giving |
287 |
up the current "timeslice" to other coros of the same or higher |
288 |
priority. Once your coro gets its turn again it will automatically be |
289 |
resumed. |
290 |
|
291 |
This function is often called C<yield> in other languages. |
292 |
|
293 |
=item Coro::cede_notself |
294 |
|
295 |
Works like cede, but is not exported by default and will cede to I<any> |
296 |
coro, regardless of priority. This is useful sometimes to ensure |
297 |
progress is made. |
298 |
|
299 |
=item terminate [arg...] |
300 |
|
301 |
Terminates the current coro with the given status values (see L<cancel>). |
302 |
|
303 |
=item Coro::on_enter BLOCK, Coro::on_leave BLOCK |
304 |
|
305 |
These function install enter and leave winders in the current scope. The |
306 |
enter block will be executed when on_enter is called and whenever the |
307 |
current coro is re-entered by the scheduler, while the leave block is |
308 |
executed whenever the current coro is blocked by the scheduler, and |
309 |
also when the containing scope is exited (by whatever means, be it exit, |
310 |
die, last etc.). |
311 |
|
312 |
I<Neither invoking the scheduler, nor exceptions, are allowed within those |
313 |
BLOCKs>. That means: do not even think about calling C<die> without an |
314 |
eval, and do not even think of entering the scheduler in any way. |
315 |
|
316 |
Since both BLOCKs are tied to the current scope, they will automatically |
317 |
be removed when the current scope exits. |
318 |
|
319 |
These functions implement the same concept as C<dynamic-wind> in scheme |
320 |
does, and are useful when you want to localise some resource to a specific |
321 |
coro. |
322 |
|
323 |
They slow down thread switching considerably for coros that use them |
324 |
(about 40% for a BLOCK with a single assignment, so thread switching is |
325 |
still reasonably fast if the handlers are fast). |
326 |
|
327 |
These functions are best understood by an example: The following function |
328 |
will change the current timezone to "Antarctica/South_Pole", which |
329 |
requires a call to C<tzset>, but by using C<on_enter> and C<on_leave>, |
330 |
which remember/change the current timezone and restore the previous |
331 |
value, respectively, the timezone is only changed for the coro that |
332 |
installed those handlers. |
333 |
|
334 |
use POSIX qw(tzset); |
335 |
|
336 |
async { |
337 |
my $old_tz; # store outside TZ value here |
338 |
|
339 |
Coro::on_enter { |
340 |
$old_tz = $ENV{TZ}; # remember the old value |
341 |
|
342 |
$ENV{TZ} = "Antarctica/South_Pole"; |
343 |
tzset; # enable new value |
344 |
}; |
345 |
|
346 |
Coro::on_leave { |
347 |
$ENV{TZ} = $old_tz; |
348 |
tzset; # restore old value |
349 |
}; |
350 |
|
351 |
# at this place, the timezone is Antarctica/South_Pole, |
352 |
# without disturbing the TZ of any other coro. |
353 |
}; |
354 |
|
355 |
This can be used to localise about any resource (locale, uid, current |
356 |
working directory etc.) to a block, despite the existance of other |
357 |
coros. |
358 |
|
359 |
Another interesting example implements time-sliced multitasking using |
360 |
interval timers (this could obviously be optimised, but does the job): |
361 |
|
362 |
# "timeslice" the given block |
363 |
sub timeslice(&) { |
364 |
use Time::HiRes (); |
365 |
|
366 |
Coro::on_enter { |
367 |
# on entering the thread, we set an VTALRM handler to cede |
368 |
$SIG{VTALRM} = sub { cede }; |
369 |
# and then start the interval timer |
370 |
Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0.01, 0.01; |
371 |
}; |
372 |
Coro::on_leave { |
373 |
# on leaving the thread, we stop the interval timer again |
374 |
Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0, 0; |
375 |
}; |
376 |
|
377 |
&{+shift}; |
378 |
} |
379 |
|
380 |
# use like this: |
381 |
timeslice { |
382 |
# The following is an endless loop that would normally |
383 |
# monopolise the process. Since it runs in a timesliced |
384 |
# environment, it will regularly cede to other threads. |
385 |
while () { } |
386 |
}; |
387 |
|
388 |
|
389 |
=item killall |
390 |
|
391 |
Kills/terminates/cancels all coros except the currently running one. |
392 |
|
393 |
Note that while this will try to free some of the main interpreter |
394 |
resources if the calling coro isn't the main coro, but one |
395 |
cannot free all of them, so if a coro that is not the main coro |
396 |
calls this function, there will be some one-time resource leak. |
397 |
|
398 |
=cut |
399 |
|
400 |
sub killall { |
401 |
for (Coro::State::list) { |
402 |
$_->cancel |
403 |
if $_ != $current && UNIVERSAL::isa $_, "Coro"; |
404 |
} |
405 |
} |
406 |
|
407 |
=back |
408 |
|
409 |
=head1 CORO OBJECT METHODS |
410 |
|
411 |
These are the methods you can call on coro objects (or to create |
412 |
them). |
413 |
|
414 |
=over 4 |
415 |
|
416 |
=item new Coro \&sub [, @args...] |
417 |
|
418 |
Create a new coro and return it. When the sub returns, the coro |
419 |
automatically terminates as if C<terminate> with the returned values were |
420 |
called. To make the coro run you must first put it into the ready |
421 |
queue by calling the ready method. |
422 |
|
423 |
See C<async> and C<Coro::State::new> for additional info about the |
424 |
coro environment. |
425 |
|
426 |
=cut |
427 |
|
428 |
sub _coro_run { |
429 |
terminate &{+shift}; |
430 |
} |
431 |
|
432 |
=item $success = $coro->ready |
433 |
|
434 |
Put the given coro into the end of its ready queue (there is one |
435 |
queue for each priority) and return true. If the coro is already in |
436 |
the ready queue, do nothing and return false. |
437 |
|
438 |
This ensures that the scheduler will resume this coro automatically |
439 |
once all the coro of higher priority and all coro of the same |
440 |
priority that were put into the ready queue earlier have been resumed. |
441 |
|
442 |
=item $coro->suspend |
443 |
|
444 |
Suspends the specified coro. A suspended coro works just like any other |
445 |
coro, except that the scheduler will not select a suspended coro for |
446 |
execution. |
447 |
|
448 |
Suspending a coro can be useful when you want to keep the coro from |
449 |
running, but you don't want to destroy it, or when you want to temporarily |
450 |
freeze a coro (e.g. for debugging) to resume it later. |
451 |
|
452 |
A scenario for the former would be to suspend all (other) coros after a |
453 |
fork and keep them alive, so their destructors aren't called, but new |
454 |
coros can be created. |
455 |
|
456 |
=item $coro->resume |
457 |
|
458 |
If the specified coro was suspended, it will be resumed. Note that when |
459 |
the coro was in the ready queue when it was suspended, it might have been |
460 |
unreadied by the scheduler, so an activation might have been lost. |
461 |
|
462 |
To avoid this, it is best to put a suspended coro into the ready queue |
463 |
unconditionally, as every synchronisation mechanism must protect itself |
464 |
against spurious wakeups, and the one in the Coro family certainly do |
465 |
that. |
466 |
|
467 |
=item $is_ready = $coro->is_ready |
468 |
|
469 |
Returns true iff the Coro object is in the ready queue. Unless the Coro |
470 |
object gets destroyed, it will eventually be scheduled by the scheduler. |
471 |
|
472 |
=item $is_running = $coro->is_running |
473 |
|
474 |
Returns true iff the Coro object is currently running. Only one Coro object |
475 |
can ever be in the running state (but it currently is possible to have |
476 |
multiple running Coro::States). |
477 |
|
478 |
=item $is_suspended = $coro->is_suspended |
479 |
|
480 |
Returns true iff this Coro object has been suspended. Suspended Coros will |
481 |
not ever be scheduled. |
482 |
|
483 |
=item $coro->cancel (arg...) |
484 |
|
485 |
Terminates the given Coro and makes it return the given arguments as |
486 |
status (default: the empty list). Never returns if the Coro is the |
487 |
current Coro. |
488 |
|
489 |
=cut |
490 |
|
491 |
sub cancel { |
492 |
my $self = shift; |
493 |
|
494 |
if ($current == $self) { |
495 |
terminate @_; |
496 |
} else { |
497 |
$self->{_status} = [@_]; |
498 |
Coro::State::cancel $self; |
499 |
} |
500 |
} |
501 |
|
502 |
=item $coro->schedule_to |
503 |
|
504 |
Puts the current coro to sleep (like C<Coro::schedule>), but instead |
505 |
of continuing with the next coro from the ready queue, always switch to |
506 |
the given coro object (regardless of priority etc.). The readyness |
507 |
state of that coro isn't changed. |
508 |
|
509 |
This is an advanced method for special cases - I'd love to hear about any |
510 |
uses for this one. |
511 |
|
512 |
=item $coro->cede_to |
513 |
|
514 |
Like C<schedule_to>, but puts the current coro into the ready |
515 |
queue. This has the effect of temporarily switching to the given |
516 |
coro, and continuing some time later. |
517 |
|
518 |
This is an advanced method for special cases - I'd love to hear about any |
519 |
uses for this one. |
520 |
|
521 |
=item $coro->throw ([$scalar]) |
522 |
|
523 |
If C<$throw> is specified and defined, it will be thrown as an exception |
524 |
inside the coro at the next convenient point in time. Otherwise |
525 |
clears the exception object. |
526 |
|
527 |
Coro will check for the exception each time a schedule-like-function |
528 |
returns, i.e. after each C<schedule>, C<cede>, C<< Coro::Semaphore->down |
529 |
>>, C<< Coro::Handle->readable >> and so on. Most of these functions |
530 |
detect this case and return early in case an exception is pending. |
531 |
|
532 |
The exception object will be thrown "as is" with the specified scalar in |
533 |
C<$@>, i.e. if it is a string, no line number or newline will be appended |
534 |
(unlike with C<die>). |
535 |
|
536 |
This can be used as a softer means than C<cancel> to ask a coro to |
537 |
end itself, although there is no guarantee that the exception will lead to |
538 |
termination, and if the exception isn't caught it might well end the whole |
539 |
program. |
540 |
|
541 |
You might also think of C<throw> as being the moral equivalent of |
542 |
C<kill>ing a coro with a signal (in this case, a scalar). |
543 |
|
544 |
=item $coro->join |
545 |
|
546 |
Wait until the coro terminates and return any values given to the |
547 |
C<terminate> or C<cancel> functions. C<join> can be called concurrently |
548 |
from multiple coro, and all will be resumed and given the status |
549 |
return once the C<$coro> terminates. |
550 |
|
551 |
=cut |
552 |
|
553 |
sub join { |
554 |
my $self = shift; |
555 |
|
556 |
unless ($self->{_status}) { |
557 |
my $current = $current; |
558 |
|
559 |
push @{$self->{_on_destroy}}, sub { |
560 |
$current->ready; |
561 |
undef $current; |
562 |
}; |
563 |
|
564 |
&schedule while $current; |
565 |
} |
566 |
|
567 |
wantarray ? @{$self->{_status}} : $self->{_status}[0]; |
568 |
} |
569 |
|
570 |
=item $coro->on_destroy (\&cb) |
571 |
|
572 |
Registers a callback that is called when this coro gets destroyed, |
573 |
but before it is joined. The callback gets passed the terminate arguments, |
574 |
if any, and I<must not> die, under any circumstances. |
575 |
|
576 |
=cut |
577 |
|
578 |
sub on_destroy { |
579 |
my ($self, $cb) = @_; |
580 |
|
581 |
push @{ $self->{_on_destroy} }, $cb; |
582 |
} |
583 |
|
584 |
=item $oldprio = $coro->prio ($newprio) |
585 |
|
586 |
Sets (or gets, if the argument is missing) the priority of the |
587 |
coro. Higher priority coro get run before lower priority |
588 |
coro. Priorities are small signed integers (currently -4 .. +3), |
589 |
that you can refer to using PRIO_xxx constants (use the import tag :prio |
590 |
to get then): |
591 |
|
592 |
PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN |
593 |
3 > 1 > 0 > -1 > -3 > -4 |
594 |
|
595 |
# set priority to HIGH |
596 |
current->prio (PRIO_HIGH); |
597 |
|
598 |
The idle coro ($Coro::idle) always has a lower priority than any |
599 |
existing coro. |
600 |
|
601 |
Changing the priority of the current coro will take effect immediately, |
602 |
but changing the priority of coro in the ready queue (but not |
603 |
running) will only take effect after the next schedule (of that |
604 |
coro). This is a bug that will be fixed in some future version. |
605 |
|
606 |
=item $newprio = $coro->nice ($change) |
607 |
|
608 |
Similar to C<prio>, but subtract the given value from the priority (i.e. |
609 |
higher values mean lower priority, just as in unix). |
610 |
|
611 |
=item $olddesc = $coro->desc ($newdesc) |
612 |
|
613 |
Sets (or gets in case the argument is missing) the description for this |
614 |
coro. This is just a free-form string you can associate with a |
615 |
coro. |
616 |
|
617 |
This method simply sets the C<< $coro->{desc} >> member to the given |
618 |
string. You can modify this member directly if you wish, and in fact, this |
619 |
is often preferred to indicate major processing states that cna then be |
620 |
seen for example in a L<Coro::Debug> session: |
621 |
|
622 |
sub my_long_function { |
623 |
local $Coro::current->{desc} = "now in my_long_function"; |
624 |
... |
625 |
$Coro::current->{desc} = "my_long_function: phase 1"; |
626 |
... |
627 |
$Coro::current->{desc} = "my_long_function: phase 2"; |
628 |
... |
629 |
} |
630 |
|
631 |
=cut |
632 |
|
633 |
sub desc { |
634 |
my $old = $_[0]{desc}; |
635 |
$_[0]{desc} = $_[1] if @_ > 1; |
636 |
$old; |
637 |
} |
638 |
|
639 |
sub transfer { |
640 |
require Carp; |
641 |
Carp::croak ("You must not call ->transfer on Coro objects. Use Coro::State objects or the ->schedule_to method. Caught"); |
642 |
} |
643 |
|
644 |
=back |
645 |
|
646 |
=head1 GLOBAL FUNCTIONS |
647 |
|
648 |
=over 4 |
649 |
|
650 |
=item Coro::nready |
651 |
|
652 |
Returns the number of coro that are currently in the ready state, |
653 |
i.e. that can be switched to by calling C<schedule> directory or |
654 |
indirectly. The value C<0> means that the only runnable coro is the |
655 |
currently running one, so C<cede> would have no effect, and C<schedule> |
656 |
would cause a deadlock unless there is an idle handler that wakes up some |
657 |
coro. |
658 |
|
659 |
=item my $guard = Coro::guard { ... } |
660 |
|
661 |
This function still exists, but is deprecated. Please use the |
662 |
C<Guard::guard> function instead. |
663 |
|
664 |
=cut |
665 |
|
666 |
BEGIN { *guard = \&Guard::guard } |
667 |
|
668 |
=item unblock_sub { ... } |
669 |
|
670 |
This utility function takes a BLOCK or code reference and "unblocks" it, |
671 |
returning a new coderef. Unblocking means that calling the new coderef |
672 |
will return immediately without blocking, returning nothing, while the |
673 |
original code ref will be called (with parameters) from within another |
674 |
coro. |
675 |
|
676 |
The reason this function exists is that many event libraries (such as the |
677 |
venerable L<Event|Event> module) are not thread-safe (a weaker form |
678 |
of reentrancy). This means you must not block within event callbacks, |
679 |
otherwise you might suffer from crashes or worse. The only event library |
680 |
currently known that is safe to use without C<unblock_sub> is L<EV>. |
681 |
|
682 |
Coro will try to catch you when you block in the event loop |
683 |
("FATAL:$Coro::IDLE blocked itself"), but this is just best effort and |
684 |
only works when you do not run your own event loop. |
685 |
|
686 |
This function allows your callbacks to block by executing them in another |
687 |
coro where it is safe to block. One example where blocking is handy |
688 |
is when you use the L<Coro::AIO|Coro::AIO> functions to save results to |
689 |
disk, for example. |
690 |
|
691 |
In short: simply use C<unblock_sub { ... }> instead of C<sub { ... }> when |
692 |
creating event callbacks that want to block. |
693 |
|
694 |
If your handler does not plan to block (e.g. simply sends a message to |
695 |
another coro, or puts some other coro into the ready queue), there is |
696 |
no reason to use C<unblock_sub>. |
697 |
|
698 |
Note that you also need to use C<unblock_sub> for any other callbacks that |
699 |
are indirectly executed by any C-based event loop. For example, when you |
700 |
use a module that uses L<AnyEvent> (and you use L<Coro::AnyEvent>) and it |
701 |
provides callbacks that are the result of some event callback, then you |
702 |
must not block either, or use C<unblock_sub>. |
703 |
|
704 |
=cut |
705 |
|
706 |
our @unblock_queue; |
707 |
|
708 |
# we create a special coro because we want to cede, |
709 |
# to reduce pressure on the coro pool (because most callbacks |
710 |
# return immediately and can be reused) and because we cannot cede |
711 |
# inside an event callback. |
712 |
our $unblock_scheduler = new Coro sub { |
713 |
while () { |
714 |
while (my $cb = pop @unblock_queue) { |
715 |
&async_pool (@$cb); |
716 |
|
717 |
# for short-lived callbacks, this reduces pressure on the coro pool |
718 |
# as the chance is very high that the async_poll coro will be back |
719 |
# in the idle state when cede returns |
720 |
cede; |
721 |
} |
722 |
schedule; # sleep well |
723 |
} |
724 |
}; |
725 |
$unblock_scheduler->{desc} = "[unblock_sub scheduler]"; |
726 |
|
727 |
sub unblock_sub(&) { |
728 |
my $cb = shift; |
729 |
|
730 |
sub { |
731 |
unshift @unblock_queue, [$cb, @_]; |
732 |
$unblock_scheduler->ready; |
733 |
} |
734 |
} |
735 |
|
736 |
=item $cb = rouse_cb |
737 |
|
738 |
Create and return a "rouse callback". That's a code reference that, |
739 |
when called, will remember a copy of its arguments and notify the owner |
740 |
coro of the callback. |
741 |
|
742 |
See the next function. |
743 |
|
744 |
=item @args = rouse_wait [$cb] |
745 |
|
746 |
Wait for the specified rouse callback (or the last one that was created in |
747 |
this coro). |
748 |
|
749 |
As soon as the callback is invoked (or when the callback was invoked |
750 |
before C<rouse_wait>), it will return the arguments originally passed to |
751 |
the rouse callback. In scalar context, that means you get the I<last> |
752 |
argument, just as if C<rouse_wait> had a C<return ($a1, $a2, $a3...)> |
753 |
statement at the end. |
754 |
|
755 |
See the section B<HOW TO WAIT FOR A CALLBACK> for an actual usage example. |
756 |
|
757 |
=back |
758 |
|
759 |
=cut |
760 |
|
761 |
1; |
762 |
|
763 |
=head1 HOW TO WAIT FOR A CALLBACK |
764 |
|
765 |
It is very common for a coro to wait for some callback to be |
766 |
called. This occurs naturally when you use coro in an otherwise |
767 |
event-based program, or when you use event-based libraries. |
768 |
|
769 |
These typically register a callback for some event, and call that callback |
770 |
when the event occured. In a coro, however, you typically want to |
771 |
just wait for the event, simplyifying things. |
772 |
|
773 |
For example C<< AnyEvent->child >> registers a callback to be called when |
774 |
a specific child has exited: |
775 |
|
776 |
my $child_watcher = AnyEvent->child (pid => $pid, cb => sub { ... }); |
777 |
|
778 |
But from within a coro, you often just want to write this: |
779 |
|
780 |
my $status = wait_for_child $pid; |
781 |
|
782 |
Coro offers two functions specifically designed to make this easy, |
783 |
C<Coro::rouse_cb> and C<Coro::rouse_wait>. |
784 |
|
785 |
The first function, C<rouse_cb>, generates and returns a callback that, |
786 |
when invoked, will save its arguments and notify the coro that |
787 |
created the callback. |
788 |
|
789 |
The second function, C<rouse_wait>, waits for the callback to be called |
790 |
(by calling C<schedule> to go to sleep) and returns the arguments |
791 |
originally passed to the callback. |
792 |
|
793 |
Using these functions, it becomes easy to write the C<wait_for_child> |
794 |
function mentioned above: |
795 |
|
796 |
sub wait_for_child($) { |
797 |
my ($pid) = @_; |
798 |
|
799 |
my $watcher = AnyEvent->child (pid => $pid, cb => Coro::rouse_cb); |
800 |
|
801 |
my ($rpid, $rstatus) = Coro::rouse_wait; |
802 |
$rstatus |
803 |
} |
804 |
|
805 |
In the case where C<rouse_cb> and C<rouse_wait> are not flexible enough, |
806 |
you can roll your own, using C<schedule>: |
807 |
|
808 |
sub wait_for_child($) { |
809 |
my ($pid) = @_; |
810 |
|
811 |
# store the current coro in $current, |
812 |
# and provide result variables for the closure passed to ->child |
813 |
my $current = $Coro::current; |
814 |
my ($done, $rstatus); |
815 |
|
816 |
# pass a closure to ->child |
817 |
my $watcher = AnyEvent->child (pid => $pid, cb => sub { |
818 |
$rstatus = $_[1]; # remember rstatus |
819 |
$done = 1; # mark $rstatus as valud |
820 |
}); |
821 |
|
822 |
# wait until the closure has been called |
823 |
schedule while !$done; |
824 |
|
825 |
$rstatus |
826 |
} |
827 |
|
828 |
|
829 |
=head1 BUGS/LIMITATIONS |
830 |
|
831 |
=over 4 |
832 |
|
833 |
=item fork with pthread backend |
834 |
|
835 |
When Coro is compiled using the pthread backend (which isn't recommended |
836 |
but required on many BSDs as their libcs are completely broken), then |
837 |
coro will not survive a fork. There is no known workaround except to |
838 |
fix your libc and use a saner backend. |
839 |
|
840 |
=item perl process emulation ("threads") |
841 |
|
842 |
This module is not perl-pseudo-thread-safe. You should only ever use this |
843 |
module from the first thread (this requirement might be removed in the |
844 |
future to allow per-thread schedulers, but Coro::State does not yet allow |
845 |
this). I recommend disabling thread support and using processes, as having |
846 |
the windows process emulation enabled under unix roughly halves perl |
847 |
performance, even when not used. |
848 |
|
849 |
=item coro switching is not signal safe |
850 |
|
851 |
You must not switch to another coro from within a signal handler (only |
852 |
relevant with %SIG - most event libraries provide safe signals), I<unless> |
853 |
you are sure you are not interrupting a Coro function. |
854 |
|
855 |
That means you I<MUST NOT> call any function that might "block" the |
856 |
current coro - C<cede>, C<schedule> C<< Coro::Semaphore->down >> or |
857 |
anything that calls those. Everything else, including calling C<ready>, |
858 |
works. |
859 |
|
860 |
=back |
861 |
|
862 |
|
863 |
=head1 WINDOWS PROCESS EMULATION |
864 |
|
865 |
A great many people seem to be confused about ithreads (for example, Chip |
866 |
Salzenberg called me unintelligent, incapable, stupid and gullible, |
867 |
while in the same mail making rather confused statements about perl |
868 |
ithreads (for example, that memory or files would be shared), showing his |
869 |
lack of understanding of this area - if it is hard to understand for Chip, |
870 |
it is probably not obvious to everybody). |
871 |
|
872 |
What follows is an ultra-condensed version of my talk about threads in |
873 |
scripting languages given onthe perl workshop 2009: |
874 |
|
875 |
The so-called "ithreads" were originally implemented for two reasons: |
876 |
first, to (badly) emulate unix processes on native win32 perls, and |
877 |
secondly, to replace the older, real thread model ("5.005-threads"). |
878 |
|
879 |
It does that by using threads instead of OS processes. The difference |
880 |
between processes and threads is that threads share memory (and other |
881 |
state, such as files) between threads within a single process, while |
882 |
processes do not share anything (at least not semantically). That |
883 |
means that modifications done by one thread are seen by others, while |
884 |
modifications by one process are not seen by other processes. |
885 |
|
886 |
The "ithreads" work exactly like that: when creating a new ithreads |
887 |
process, all state is copied (memory is copied physically, files and code |
888 |
is copied logically). Afterwards, it isolates all modifications. On UNIX, |
889 |
the same behaviour can be achieved by using operating system processes, |
890 |
except that UNIX typically uses hardware built into the system to do this |
891 |
efficiently, while the windows process emulation emulates this hardware in |
892 |
software (rather efficiently, but of course it is still much slower than |
893 |
dedicated hardware). |
894 |
|
895 |
As mentioned before, loading code, modifying code, modifying data |
896 |
structures and so on is only visible in the ithreads process doing the |
897 |
modification, not in other ithread processes within the same OS process. |
898 |
|
899 |
This is why "ithreads" do not implement threads for perl at all, only |
900 |
processes. What makes it so bad is that on non-windows platforms, you can |
901 |
actually take advantage of custom hardware for this purpose (as evidenced |
902 |
by the forks module, which gives you the (i-) threads API, just much |
903 |
faster). |
904 |
|
905 |
Sharing data is in the i-threads model is done by transfering data |
906 |
structures between threads using copying semantics, which is very slow - |
907 |
shared data simply does not exist. Benchmarks using i-threads which are |
908 |
communication-intensive show extremely bad behaviour with i-threads (in |
909 |
fact, so bad that Coro, which cannot take direct advantage of multiple |
910 |
CPUs, is often orders of magnitude faster because it shares data using |
911 |
real threads, refer to my talk for details). |
912 |
|
913 |
As summary, i-threads *use* threads to implement processes, while |
914 |
the compatible forks module *uses* processes to emulate, uhm, |
915 |
processes. I-threads slow down every perl program when enabled, and |
916 |
outside of windows, serve no (or little) practical purpose, but |
917 |
disadvantages every single-threaded Perl program. |
918 |
|
919 |
This is the reason that I try to avoid the name "ithreads", as it is |
920 |
misleading as it implies that it implements some kind of thread model for |
921 |
perl, and prefer the name "windows process emulation", which describes the |
922 |
actual use and behaviour of it much better. |
923 |
|
924 |
=head1 SEE ALSO |
925 |
|
926 |
Event-Loop integration: L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>. |
927 |
|
928 |
Debugging: L<Coro::Debug>. |
929 |
|
930 |
Support/Utility: L<Coro::Specific>, L<Coro::Util>. |
931 |
|
932 |
Locking and IPC: L<Coro::Signal>, L<Coro::Channel>, L<Coro::Semaphore>, |
933 |
L<Coro::SemaphoreSet>, L<Coro::RWLock>. |
934 |
|
935 |
I/O and Timers: L<Coro::Timer>, L<Coro::Handle>, L<Coro::Socket>, L<Coro::AIO>. |
936 |
|
937 |
Compatibility with other modules: L<Coro::LWP> (but see also L<AnyEvent::HTTP> for |
938 |
a better-working alternative), L<Coro::BDB>, L<Coro::Storable>, |
939 |
L<Coro::Select>. |
940 |
|
941 |
XS API: L<Coro::MakeMaker>. |
942 |
|
943 |
Low level Configuration, Thread Environment, Continuations: L<Coro::State>. |
944 |
|
945 |
=head1 AUTHOR |
946 |
|
947 |
Marc Lehmann <schmorp@schmorp.de> |
948 |
http://home.schmorp.de/ |
949 |
|
950 |
=cut |
951 |
|