ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Coro.pm
(Generate patch)

Comparing Coro/Coro.pm (file contents):
Revision 1.106 by root, Fri Jan 5 17:44:17 2007 UTC vs.
Revision 1.141 by root, Tue Oct 2 10:38:17 2007 UTC

20 20
21=head1 DESCRIPTION 21=head1 DESCRIPTION
22 22
23This module collection manages coroutines. Coroutines are similar 23This module collection manages coroutines. Coroutines are similar
24to threads but don't run in parallel at the same time even on SMP 24to threads but don't run in parallel at the same time even on SMP
25machines. The specific flavor of coroutine use din this module also 25machines. The specific flavor of coroutine used in this module also
26guarentees you that it will not switch between coroutines unless 26guarantees you that it will not switch between coroutines unless
27necessary, at easily-identified points in your program, so locking and 27necessary, at easily-identified points in your program, so locking and
28parallel access are rarely an issue, making coroutine programming much 28parallel access are rarely an issue, making coroutine programming much
29safer than threads programming. 29safer than threads programming.
30 30
31(Perl, however, does not natively support real threads but instead does a 31(Perl, however, does not natively support real threads but instead does a
50 50
51our $idle; # idle handler 51our $idle; # idle handler
52our $main; # main coroutine 52our $main; # main coroutine
53our $current; # current coroutine 53our $current; # current coroutine
54 54
55our $VERSION = '3.3'; 55our $VERSION = '3.8';
56 56
57our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub); 57our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub);
58our %EXPORT_TAGS = ( 58our %EXPORT_TAGS = (
59 prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)], 59 prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)],
60); 60);
108 108
109The current coroutine (the last coroutine switched to). The initial value 109The current coroutine (the last coroutine switched to). The initial value
110is C<$main> (of course). 110is C<$main> (of course).
111 111
112This variable is B<strictly> I<read-only>. It is provided for performance 112This variable is B<strictly> I<read-only>. It is provided for performance
113reasons. If performance is not essentiel you are encouraged to use the 113reasons. If performance is not essential you are encouraged to use the
114C<Coro::current> function instead. 114C<Coro::current> function instead.
115 115
116=cut 116=cut
117
118$main->{desc} = "[main::]";
117 119
118# maybe some other module used Coro::Specific before... 120# maybe some other module used Coro::Specific before...
119$main->{specific} = $current->{specific} 121$main->{specific} = $current->{specific}
120 if $current; 122 if $current;
121 123
166 while @destroy; 168 while @destroy;
167 169
168 &schedule; 170 &schedule;
169 } 171 }
170}; 172};
171 173$manager->desc ("[coro manager]");
172$manager->prio (PRIO_MAX); 174$manager->prio (PRIO_MAX);
173 175
174# static methods. not really. 176# static methods. not really.
175 177
176=back 178=back
185 187
186Create a new asynchronous coroutine and return it's coroutine object 188Create a new asynchronous coroutine and return it's coroutine object
187(usually unused). When the sub returns the new coroutine is automatically 189(usually unused). When the sub returns the new coroutine is automatically
188terminated. 190terminated.
189 191
190Calling C<exit> in a coroutine will not work correctly, so do not do that. 192Calling C<exit> in a coroutine will do the same as calling exit outside
191 193the coroutine. Likewise, when the coroutine dies, the program will exit,
192When the coroutine dies, the program will exit, just as in the main 194just as it would in the main program.
193program.
194 195
195 # create a new coroutine that just prints its arguments 196 # create a new coroutine that just prints its arguments
196 async { 197 async {
197 print "@_\n"; 198 print "@_\n";
198 } 1,2,3,4; 199 } 1,2,3,4;
210Similar to C<async>, but uses a coroutine pool, so you should not call 211Similar to C<async>, but uses a coroutine pool, so you should not call
211terminate or join (although you are allowed to), and you get a coroutine 212terminate or join (although you are allowed to), and you get a coroutine
212that might have executed other code already (which can be good or bad :). 213that might have executed other code already (which can be good or bad :).
213 214
214Also, the block is executed in an C<eval> context and a warning will be 215Also, the block is executed in an C<eval> context and a warning will be
215issued in case of an exception instead of terminating the program, as C<async> does. 216issued in case of an exception instead of terminating the program, as
217C<async> does. As the coroutine is being reused, stuff like C<on_destroy>
218will not work in the expected way, unless you call terminate or cancel,
219which somehow defeats the purpose of pooling.
216 220
217The priority will be reset to C<0> after each job, otherwise the coroutine 221The priority will be reset to C<0> after each job, otherwise the coroutine
218will be re-used "as-is". 222will be re-used "as-is".
219 223
220The pool size is limited to 8 idle coroutines (this can be adjusted by 224The pool size is limited to 8 idle coroutines (this can be adjusted by
221changing $Coro::POOL_SIZE), and there can be as many non-idle coros as 225changing $Coro::POOL_SIZE), and there can be as many non-idle coros as
222required. 226required.
223 227
224If you are concerned about pooled coroutines growing a lot because a 228If you are concerned about pooled coroutines growing a lot because a
225single C<async_pool> used a lot of stackspace you can e.g. C<async_pool { 229single C<async_pool> used a lot of stackspace you can e.g. C<async_pool
226terminate }> once per second or so to slowly replenish the pool. 230{ terminate }> once per second or so to slowly replenish the pool. In
231addition to that, when the stacks used by a handler grows larger than 16kb
232(adjustable with $Coro::POOL_RSS) it will also exit.
227 233
228=cut 234=cut
229 235
230our $POOL_SIZE = 8; 236our $POOL_SIZE = 8;
237our $POOL_RSS = 16 * 1024;
231our @pool; 238our @async_pool;
232 239
233sub pool_handler { 240sub pool_handler {
241 my $cb;
242
234 while () { 243 while () {
235 my ($cb, @arg) = @{ delete $current->{_invoke} };
236
237 eval { 244 eval {
238 $cb->(@arg); 245 while () {
246 _pool_1 $cb;
247 &$cb;
248 _pool_2 $cb;
249 &schedule;
250 }
239 }; 251 };
252
253 last if $@ eq "\3terminate\2\n";
240 warn $@ if $@; 254 warn $@ if $@;
241
242 last if @pool >= $POOL_SIZE;
243 push @pool, $current;
244
245 $current->prio (0);
246 schedule;
247 } 255 }
248} 256}
249 257
250sub async_pool(&@) { 258sub async_pool(&@) {
251 # this is also inlined into the unlock_scheduler 259 # this is also inlined into the unlock_scheduler
252 my $coro = (pop @pool or new Coro \&pool_handler); 260 my $coro = (pop @async_pool) || new Coro \&pool_handler;
253 261
254 $coro->{_invoke} = [@_]; 262 $coro->{_invoke} = [@_];
255 $coro->ready; 263 $coro->ready;
256 264
257 $coro 265 $coro
275 # wake up sleeping coroutine 283 # wake up sleeping coroutine
276 $current->ready; 284 $current->ready;
277 undef $current; 285 undef $current;
278 }; 286 };
279 287
280 # call schedule until event occured. 288 # call schedule until event occurred.
281 # in case we are woken up for other reasons 289 # in case we are woken up for other reasons
282 # (current still defined), loop. 290 # (current still defined), loop.
283 Coro::schedule while $current; 291 Coro::schedule while $current;
284 } 292 }
285 293
287 295
288"Cede" to other coroutines. This function puts the current coroutine into the 296"Cede" to other coroutines. This function puts the current coroutine into the
289ready queue and calls C<schedule>, which has the effect of giving up the 297ready queue and calls C<schedule>, which has the effect of giving up the
290current "timeslice" to other coroutines of the same or higher priority. 298current "timeslice" to other coroutines of the same or higher priority.
291 299
300Returns true if at least one coroutine switch has happened.
301
292=item Coro::cede_notself 302=item Coro::cede_notself
293 303
294Works like cede, but is not exported by default and will cede to any 304Works like cede, but is not exported by default and will cede to any
295coroutine, regardless of priority, once. 305coroutine, regardless of priority, once.
296 306
307Returns true if at least one coroutine switch has happened.
308
297=item terminate [arg...] 309=item terminate [arg...]
298 310
299Terminates the current coroutine with the given status values (see L<cancel>). 311Terminates the current coroutine with the given status values (see L<cancel>).
312
313=item killall
314
315Kills/terminates/cancels all coroutines except the currently running
316one. This is useful after a fork, either in the child or the parent, as
317usually only one of them should inherit the running coroutines.
300 318
301=cut 319=cut
302 320
303sub terminate { 321sub terminate {
304 $current->cancel (@_); 322 $current->cancel (@_);
323}
324
325sub killall {
326 for (Coro::State::list) {
327 $_->cancel
328 if $_ != $current && UNIVERSAL::isa $_, "Coro";
329 }
305} 330}
306 331
307=back 332=back
308 333
309# dynamic methods 334# dynamic methods
319Create a new coroutine and return it. When the sub returns the coroutine 344Create a new coroutine and return it. When the sub returns the coroutine
320automatically terminates as if C<terminate> with the returned values were 345automatically terminates as if C<terminate> with the returned values were
321called. To make the coroutine run you must first put it into the ready queue 346called. To make the coroutine run you must first put it into the ready queue
322by calling the ready method. 347by calling the ready method.
323 348
324Calling C<exit> in a coroutine will not work correctly, so do not do that. 349See C<async> for additional discussion.
325 350
326=cut 351=cut
327 352
328sub _run_coro { 353sub _run_coro {
329 terminate &{+shift}; 354 terminate &{+shift};
452=over 4 477=over 4
453 478
454=item Coro::nready 479=item Coro::nready
455 480
456Returns the number of coroutines that are currently in the ready state, 481Returns the number of coroutines that are currently in the ready state,
457i.e. that can be swicthed to. The value C<0> means that the only runnable 482i.e. that can be switched to. The value C<0> means that the only runnable
458coroutine is the currently running one, so C<cede> would have no effect, 483coroutine is the currently running one, so C<cede> would have no effect,
459and C<schedule> would cause a deadlock unless there is an idle handler 484and C<schedule> would cause a deadlock unless there is an idle handler
460that wakes up some coroutines. 485that wakes up some coroutines.
461 486
462=item my $guard = Coro::guard { ... } 487=item my $guard = Coro::guard { ... }
463 488
464This creates and returns a guard object. Nothing happens until the objetc 489This creates and returns a guard object. Nothing happens until the object
465gets destroyed, in which case the codeblock given as argument will be 490gets destroyed, in which case the codeblock given as argument will be
466executed. This is useful to free locks or other resources in case of a 491executed. This is useful to free locks or other resources in case of a
467runtime error or when the coroutine gets canceled, as in both cases the 492runtime error or when the coroutine gets canceled, as in both cases the
468guard block will be executed. The guard object supports only one method, 493guard block will be executed. The guard object supports only one method,
469C<< ->cancel >>, which will keep the codeblock from being executed. 494C<< ->cancel >>, which will keep the codeblock from being executed.
498This utility function takes a BLOCK or code reference and "unblocks" it, 523This utility function takes a BLOCK or code reference and "unblocks" it,
499returning the new coderef. This means that the new coderef will return 524returning the new coderef. This means that the new coderef will return
500immediately without blocking, returning nothing, while the original code 525immediately without blocking, returning nothing, while the original code
501ref will be called (with parameters) from within its own coroutine. 526ref will be called (with parameters) from within its own coroutine.
502 527
503The reason this fucntion exists is that many event libraries (such as the 528The reason this function exists is that many event libraries (such as the
504venerable L<Event|Event> module) are not coroutine-safe (a weaker form 529venerable L<Event|Event> module) are not coroutine-safe (a weaker form
505of thread-safety). This means you must not block within event callbacks, 530of thread-safety). This means you must not block within event callbacks,
506otherwise you might suffer from crashes or worse. 531otherwise you might suffer from crashes or worse.
507 532
508This function allows your callbacks to block by executing them in another 533This function allows your callbacks to block by executing them in another
519 544
520# we create a special coro because we want to cede, 545# we create a special coro because we want to cede,
521# to reduce pressure on the coro pool (because most callbacks 546# to reduce pressure on the coro pool (because most callbacks
522# return immediately and can be reused) and because we cannot cede 547# return immediately and can be reused) and because we cannot cede
523# inside an event callback. 548# inside an event callback.
524our $unblock_scheduler = async { 549our $unblock_scheduler = new Coro sub {
525 while () { 550 while () {
526 while (my $cb = pop @unblock_queue) { 551 while (my $cb = pop @unblock_queue) {
527 # this is an inlined copy of async_pool 552 # this is an inlined copy of async_pool
528 my $coro = (pop @pool or new Coro \&pool_handler); 553 my $coro = (pop @async_pool) || new Coro \&pool_handler;
529 554
530 $coro->{_invoke} = $cb; 555 $coro->{_invoke} = $cb;
531 $coro->ready; 556 $coro->ready;
532 cede; # for short-lived callbacks, this reduces pressure on the coro pool 557 cede; # for short-lived callbacks, this reduces pressure on the coro pool
533 } 558 }
534 schedule; # sleep well 559 schedule; # sleep well
535 } 560 }
536}; 561};
562$unblock_scheduler->desc ("[unblock_sub scheduler]");
537 563
538sub unblock_sub(&) { 564sub unblock_sub(&) {
539 my $cb = shift; 565 my $cb = shift;
540 566
541 sub { 567 sub {
554 580
555 - you must make very sure that no coro is still active on global 581 - you must make very sure that no coro is still active on global
556 destruction. very bad things might happen otherwise (usually segfaults). 582 destruction. very bad things might happen otherwise (usually segfaults).
557 583
558 - this module is not thread-safe. You should only ever use this module 584 - this module is not thread-safe. You should only ever use this module
559 from the same thread (this requirement might be losened in the future 585 from the same thread (this requirement might be loosened in the future
560 to allow per-thread schedulers, but Coro::State does not yet allow 586 to allow per-thread schedulers, but Coro::State does not yet allow
561 this). 587 this).
562 588
563=head1 SEE ALSO 589=head1 SEE ALSO
564 590

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines