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

Comparing Coro/Coro.pm (file contents):
Revision 1.165 by root, Thu Dec 20 07:29:14 2007 UTC vs.
Revision 1.245 by root, Sun Dec 14 19:52:58 2008 UTC

1=head1 NAME 1=head1 NAME
2 2
3Coro - coroutine process abstraction 3Coro - the only real threads in perl
4 4
5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 use Coro; 7 use Coro;
8 8
9 async { 9 async {
10 # some asynchronous thread of execution 10 # some asynchronous thread of execution
11 print "2\n"; 11 print "2\n";
12 cede; # yield back to main 12 cede; # yield back to main
13 print "4\n"; 13 print "4\n";
14 }; 14 };
15 print "1\n"; 15 print "1\n";
16 cede; # yield to coroutine 16 cede; # yield to coroutine
17 print "3\n"; 17 print "3\n";
18 cede; # and again 18 cede; # and again
19 19
20 # use locking 20 # use locking
21 use Coro::Semaphore;
21 my $lock = new Coro::Semaphore; 22 my $lock = new Coro::Semaphore;
22 my $locked; 23 my $locked;
23 24
24 $lock->down; 25 $lock->down;
25 $locked = 1; 26 $locked = 1;
26 $lock->up; 27 $lock->up;
27 28
28=head1 DESCRIPTION 29=head1 DESCRIPTION
29 30
30This module collection manages coroutines. Coroutines are similar 31For a tutorial-style introduction, please read the L<Coro::Intro>
31to threads but don't run in parallel at the same time even on SMP 32manpage. This manpage mainly contains reference information.
32machines. The specific flavor of coroutine used in this module also
33guarantees you that it will not switch between coroutines unless
34necessary, at easily-identified points in your program, so locking and
35parallel access are rarely an issue, making coroutine programming much
36safer than threads programming.
37 33
38(Perl, however, does not natively support real threads but instead does a 34This module collection manages continuations in general, most often
39very slow and memory-intensive emulation of processes using threads. This 35in the form of cooperative threads (also called coroutines in the
40is a performance win on Windows machines, and a loss everywhere else). 36documentation). They are similar to kernel threads but don't (in general)
37run in parallel at the same time even on SMP machines. The specific flavor
38of thread offered by this module also guarantees you that it will not
39switch between threads unless necessary, at easily-identified points in
40your program, so locking and parallel access are rarely an issue, making
41thread programming much safer and easier than using other thread models.
41 42
43Unlike the so-called "Perl threads" (which are not actually real threads
44but only the windows process emulation ported to unix), Coro provides a
45full shared address space, which makes communication between threads
46very easy. And threads are fast, too: disabling the Windows process
47emulation code in your perl and using Coro can easily result in a two to
48four times speed increase for your programs.
49
50Coro achieves that by supporting multiple running interpreters that share
51data, which is especially useful to code pseudo-parallel processes and
52for event-based programming, such as multiple HTTP-GET requests running
53concurrently. See L<Coro::AnyEvent> to learn more on how to integrate Coro
54into an event-based environment.
55
42In this module, coroutines are defined as "callchain + lexical variables + 56In this module, a thread is defined as "callchain + lexical variables +
43@_ + $_ + $@ + $/ + C stack), that is, a coroutine has its own callchain, 57@_ + $_ + $@ + $/ + C stack), that is, a thread has its own callchain,
44its own set of lexicals and its own set of perls most important global 58its own set of lexicals and its own set of perls most important global
45variables (see L<Coro::State> for more configuration). 59variables (see L<Coro::State> for more configuration and background info).
60
61See also the C<SEE ALSO> section at the end of this document - the Coro
62module family is quite large.
46 63
47=cut 64=cut
48 65
49package Coro; 66package Coro;
50 67
51use strict; 68use strict qw(vars subs);
52no warnings "uninitialized"; 69no warnings "uninitialized";
53 70
54use Coro::State; 71use Coro::State;
55 72
56use base qw(Coro::State Exporter); 73use base qw(Coro::State Exporter);
57 74
58our $idle; # idle handler 75our $idle; # idle handler
59our $main; # main coroutine 76our $main; # main coroutine
60our $current; # current coroutine 77our $current; # current coroutine
61 78
62our $VERSION = '4.34'; 79our $VERSION = 5.13;
63 80
64our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub); 81our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub);
65our %EXPORT_TAGS = ( 82our %EXPORT_TAGS = (
66 prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)], 83 prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)],
67); 84);
68our @EXPORT_OK = (@{$EXPORT_TAGS{prio}}, qw(nready)); 85our @EXPORT_OK = (@{$EXPORT_TAGS{prio}}, qw(nready));
69 86
70{ 87=head1 GLOBAL VARIABLES
71 my @async;
72 my $init;
73
74 # this way of handling attributes simply is NOT scalable ;()
75 sub import {
76 no strict 'refs';
77
78 Coro->export_to_level (1, @_);
79
80 my $old = *{(caller)[0]."::MODIFY_CODE_ATTRIBUTES"}{CODE};
81 *{(caller)[0]."::MODIFY_CODE_ATTRIBUTES"} = sub {
82 my ($package, $ref) = (shift, shift);
83 my @attrs;
84 for (@_) {
85 if ($_ eq "Coro") {
86 push @async, $ref;
87 unless ($init++) {
88 eval q{
89 sub INIT {
90 &async(pop @async) while @async;
91 }
92 };
93 }
94 } else {
95 push @attrs, $_;
96 }
97 }
98 return $old ? $old->($package, $ref, @attrs) : @attrs;
99 };
100 }
101
102}
103 88
104=over 4 89=over 4
105 90
106=item $main 91=item $Coro::main
107 92
108This coroutine represents the main program. 93This variable stores the coroutine object that represents the main
94program. While you cna C<ready> it and do most other things you can do to
95coroutines, it is mainly useful to compare again C<$Coro::current>, to see
96whether you are running in the main program or not.
109 97
110=cut 98=cut
111 99
112$main = new Coro; 100# $main is now being initialised by Coro::State
113 101
114=item $current (or as function: current) 102=item $Coro::current
115 103
116The current coroutine (the last coroutine switched to). The initial value 104The coroutine object representing the current coroutine (the last
105coroutine that the Coro scheduler switched to). The initial value is
117is C<$main> (of course). 106C<$Coro::main> (of course).
118 107
119This variable is B<strictly> I<read-only>. It is provided for performance 108This variable is B<strictly> I<read-only>. You can take copies of the
120reasons. If performance is not essential you are encouraged to use the 109value stored in it and use it as any other coroutine object, but you must
121C<Coro::current> function instead. 110not otherwise modify the variable itself.
122 111
123=cut 112=cut
124 113
125$main->{desc} = "[main::]";
126
127# maybe some other module used Coro::Specific before...
128$main->{_specific} = $current->{_specific}
129 if $current;
130
131_set_current $main;
132
133sub current() { $current } 114sub current() { $current } # [DEPRECATED]
134 115
135=item $idle 116=item $Coro::idle
136 117
118This variable is mainly useful to integrate Coro into event loops. It is
119usually better to rely on L<Coro::AnyEvent> or L<Coro::EV>, as this is
120pretty low-level functionality.
121
122This variable stores either a coroutine or a callback.
123
137A callback that is called whenever the scheduler finds no ready coroutines 124If it is a callback, the it is called whenever the scheduler finds no
138to run. The default implementation prints "FATAL: deadlock detected" and 125ready coroutines to run. The default implementation prints "FATAL:
139exits, because the program has no other way to continue. 126deadlock detected" and exits, because the program has no other way to
127continue.
140 128
129If it is a coroutine object, then this object will be readied (without
130invoking any ready hooks, however) when the scheduler finds no other ready
131coroutines to run.
132
141This hook is overwritten by modules such as C<Coro::Timer> and 133This hook is overwritten by modules such as C<Coro::EV> and
142C<Coro::Event> to wait on an external event that hopefully wake up a 134C<Coro::AnyEvent> to wait on an external event that hopefully wake up a
143coroutine so the scheduler can run it. 135coroutine so the scheduler can run it.
136
137Note that the callback I<must not>, under any circumstances, block
138the current coroutine. Normally, this is achieved by having an "idle
139coroutine" that calls the event loop and then blocks again, and then
140readying that coroutine in the idle handler, or by simply placing the idle
141coroutine in this variable.
142
143See L<Coro::Event> or L<Coro::AnyEvent> for examples of using this
144technique.
144 145
145Please note that if your callback recursively invokes perl (e.g. for event 146Please note that if your callback recursively invokes perl (e.g. for event
146handlers), then it must be prepared to be called recursively itself. 147handlers), then it must be prepared to be called recursively itself.
147 148
148=cut 149=cut
… …
150$idle = sub { 151$idle = sub {
151 require Carp; 152 require Carp;
152 Carp::croak ("FATAL: deadlock detected"); 153 Carp::croak ("FATAL: deadlock detected");
153}; 154};
154 155
155sub _cancel {
156 my ($self) = @_;
157
158 # free coroutine data and mark as destructed
159 $self->_destroy
160 or return;
161
162 # call all destruction callbacks
163 $_->(@{$self->{_status}})
164 for @{(delete $self->{_on_destroy}) || []};
165}
166
167# this coroutine is necessary because a coroutine 156# this coroutine is necessary because a coroutine
168# cannot destroy itself. 157# cannot destroy itself.
169my @destroy; 158our @destroy;
170my $manager; 159our $manager;
171 160
172$manager = new Coro sub { 161$manager = new Coro sub {
173 while () { 162 while () {
174 (shift @destroy)->_cancel 163 Coro::_cancel shift @destroy
175 while @destroy; 164 while @destroy;
176 165
177 &schedule; 166 &schedule;
178 } 167 }
179}; 168};
180$manager->desc ("[coro manager]"); 169$manager->{desc} = "[coro manager]";
181$manager->prio (PRIO_MAX); 170$manager->prio (PRIO_MAX);
182 171
183# static methods. not really.
184
185=back 172=back
186 173
187=head2 STATIC METHODS 174=head1 SIMPLE COROUTINE CREATION
188
189Static methods are actually functions that operate on the current coroutine only.
190 175
191=over 4 176=over 4
192 177
193=item async { ... } [@args...] 178=item async { ... } [@args...]
194 179
195Create a new asynchronous coroutine and return it's coroutine object 180Create a new coroutine and return its coroutine object (usually
196(usually unused). When the sub returns the new coroutine is automatically 181unused). The coroutine will be put into the ready queue, so
182it will start running automatically on the next scheduler run.
183
184The first argument is a codeblock/closure that should be executed in the
185coroutine. When it returns argument returns the coroutine is automatically
197terminated. 186terminated.
198 187
188The remaining arguments are passed as arguments to the closure.
189
199See the C<Coro::State::new> constructor for info about the coroutine 190See the C<Coro::State::new> constructor for info about the coroutine
200environment in which coroutines run. 191environment in which coroutines are executed.
201 192
202Calling C<exit> in a coroutine will do the same as calling exit outside 193Calling C<exit> in a coroutine will do the same as calling exit outside
203the coroutine. Likewise, when the coroutine dies, the program will exit, 194the coroutine. Likewise, when the coroutine dies, the program will exit,
204just as it would in the main program. 195just as it would in the main program.
205 196
197If you do not want that, you can provide a default C<die> handler, or
198simply avoid dieing (by use of C<eval>).
199
206 # create a new coroutine that just prints its arguments 200Example: Create a new coroutine that just prints its arguments.
201
207 async { 202 async {
208 print "@_\n"; 203 print "@_\n";
209 } 1,2,3,4; 204 } 1,2,3,4;
210 205
211=cut 206=cut
… …
217} 212}
218 213
219=item async_pool { ... } [@args...] 214=item async_pool { ... } [@args...]
220 215
221Similar to C<async>, but uses a coroutine pool, so you should not call 216Similar to C<async>, but uses a coroutine pool, so you should not call
222terminate or join (although you are allowed to), and you get a coroutine 217terminate or join on it (although you are allowed to), and you get a
223that might have executed other code already (which can be good or bad :). 218coroutine that might have executed other code already (which can be good
219or bad :).
224 220
221On the plus side, this function is about twice as fast as creating (and
222destroying) a completely new coroutine, so if you need a lot of generic
223coroutines in quick successsion, use C<async_pool>, not C<async>.
224
225Also, the block is executed in an C<eval> context and a warning will be 225The code block is executed in an C<eval> context and a warning will be
226issued in case of an exception instead of terminating the program, as 226issued in case of an exception instead of terminating the program, as
227C<async> does. As the coroutine is being reused, stuff like C<on_destroy> 227C<async> does. As the coroutine is being reused, stuff like C<on_destroy>
228will not work in the expected way, unless you call terminate or cancel, 228will not work in the expected way, unless you call terminate or cancel,
229which somehow defeats the purpose of pooling. 229which somehow defeats the purpose of pooling (but is fine in the
230exceptional case).
230 231
231The priority will be reset to C<0> after each job, tracing will be 232The priority will be reset to C<0> after each run, tracing will be
232disabled, the description will be reset and the default output filehandle 233disabled, the description will be reset and the default output filehandle
233gets restored, so you can change alkl these. Otherwise the coroutine will 234gets restored, so you can change all these. Otherwise the coroutine will
234be re-used "as-is": most notably if you change other per-coroutine global 235be re-used "as-is": most notably if you change other per-coroutine global
235stuff such as C<$/> you need to revert that change, which is most simply 236stuff such as C<$/> you I<must needs> revert that change, which is most
236done by using local as in C< local $/ >. 237simply done by using local as in: C<< local $/ >>.
237 238
238The pool size is limited to 8 idle coroutines (this can be adjusted by 239The idle pool size is limited to C<8> idle coroutines (this can be
239changing $Coro::POOL_SIZE), and there can be as many non-idle coros as 240adjusted by changing $Coro::POOL_SIZE), but there can be as many non-idle
240required. 241coros as required.
241 242
242If you are concerned about pooled coroutines growing a lot because a 243If you are concerned about pooled coroutines growing a lot because a
243single C<async_pool> used a lot of stackspace you can e.g. C<async_pool 244single C<async_pool> used a lot of stackspace you can e.g. C<async_pool
244{ terminate }> once per second or so to slowly replenish the pool. In 245{ terminate }> once per second or so to slowly replenish the pool. In
245addition to that, when the stacks used by a handler grows larger than 16kb 246addition to that, when the stacks used by a handler grows larger than 32kb
246(adjustable with $Coro::POOL_RSS) it will also exit. 247(adjustable via $Coro::POOL_RSS) it will also be destroyed.
247 248
248=cut 249=cut
249 250
250our $POOL_SIZE = 8; 251our $POOL_SIZE = 8;
251our $POOL_RSS = 16 * 1024; 252our $POOL_RSS = 32 * 1024;
252our @async_pool; 253our @async_pool;
253 254
254sub pool_handler { 255sub pool_handler {
255 my $cb;
256
257 while () { 256 while () {
258 eval { 257 eval {
259 while () { 258 &{&_pool_handler} while 1;
260 _pool_1 $cb;
261 &$cb;
262 _pool_2 $cb;
263 &schedule;
264 }
265 }; 259 };
266 260
267 last if $@ eq "\3async_pool terminate\2\n";
268 warn $@ if $@; 261 warn $@ if $@;
269 } 262 }
270} 263}
271 264
272sub async_pool(&@) { 265=back
273 # this is also inlined into the unlock_scheduler
274 my $coro = (pop @async_pool) || new Coro \&pool_handler;
275 266
276 $coro->{_invoke} = [@_]; 267=head1 STATIC METHODS
277 $coro->ready;
278 268
279 $coro 269Static methods are actually functions that implicitly operate on the
280} 270current coroutine.
271
272=over 4
281 273
282=item schedule 274=item schedule
283 275
284Calls the scheduler. Please note that the current coroutine will not be put 276Calls the scheduler. The scheduler will find the next coroutine that is
277to be run from the ready queue and switches to it. The next coroutine
278to be run is simply the one with the highest priority that is longest
279in its ready queue. If there is no coroutine ready, it will clal the
280C<$Coro::idle> hook.
281
282Please note that the current coroutine will I<not> be put into the ready
285into the ready queue, so calling this function usually means you will 283queue, so calling this function usually means you will never be called
286never be called again unless something else (e.g. an event handler) calls 284again unless something else (e.g. an event handler) calls C<< ->ready >>,
287ready. 285thus waking you up.
288 286
289The canonical way to wait on external events is this: 287This makes C<schedule> I<the> generic method to use to block the current
288coroutine and wait for events: first you remember the current coroutine in
289a variable, then arrange for some callback of yours to call C<< ->ready
290>> on that once some event happens, and last you call C<schedule> to put
291yourself to sleep. Note that a lot of things can wake your coroutine up,
292so you need to check whether the event indeed happened, e.g. by storing the
293status in a variable.
290 294
291 { 295See B<HOW TO WAIT FOR A CALLBACK>, below, for some ways to wait for callbacks.
292 # remember current coroutine
293 my $current = $Coro::current;
294
295 # register a hypothetical event handler
296 on_event_invoke sub {
297 # wake up sleeping coroutine
298 $current->ready;
299 undef $current;
300 };
301
302 # call schedule until event occurred.
303 # in case we are woken up for other reasons
304 # (current still defined), loop.
305 Coro::schedule while $current;
306 }
307 296
308=item cede 297=item cede
309 298
310"Cede" to other coroutines. This function puts the current coroutine into the 299"Cede" to other coroutines. This function puts the current coroutine into
311ready queue and calls C<schedule>, which has the effect of giving up the 300the ready queue and calls C<schedule>, which has the effect of giving
312current "timeslice" to other coroutines of the same or higher priority. 301up the current "timeslice" to other coroutines of the same or higher
302priority. Once your coroutine gets its turn again it will automatically be
303resumed.
304
305This function is often called C<yield> in other languages.
313 306
314=item Coro::cede_notself 307=item Coro::cede_notself
315 308
316Works like cede, but is not exported by default and will cede to any 309Works like cede, but is not exported by default and will cede to I<any>
317coroutine, regardless of priority, once. 310coroutine, regardless of priority. This is useful sometimes to ensure
311progress is made.
318 312
319=item terminate [arg...] 313=item terminate [arg...]
320 314
321Terminates the current coroutine with the given status values (see L<cancel>). 315Terminates the current coroutine with the given status values (see L<cancel>).
322 316
323=item killall 317=item killall
324 318
325Kills/terminates/cancels all coroutines except the currently running 319Kills/terminates/cancels all coroutines except the currently running
326one. This is useful after a fork, either in the child or the parent, as 320one. This can be useful after a fork, either in the child or the parent,
327usually only one of them should inherit the running coroutines. 321as usually only one of them should inherit the running coroutines.
328 322
329=cut 323Note that in the implementation, destructors run as normal, making this
324function not so useful after a fork. Future versions of this function
325might try to free resources without running any code.
330 326
331sub terminate { 327Note that while this will try to free some of the main programs resources,
332 $current->cancel (@_); 328you cannot free all of them, so if a coroutine that is not the main
333} 329program calls this function, there will be some one-time resource leak.
330
331=cut
334 332
335sub killall { 333sub killall {
336 for (Coro::State::list) { 334 for (Coro::State::list) {
337 $_->cancel 335 $_->cancel
338 if $_ != $current && UNIVERSAL::isa $_, "Coro"; 336 if $_ != $current && UNIVERSAL::isa $_, "Coro";
339 } 337 }
340} 338}
341 339
342=back 340=back
343 341
344# dynamic methods
345
346=head2 COROUTINE METHODS 342=head1 COROUTINE OBJECT METHODS
347 343
348These are the methods you can call on coroutine objects. 344These are the methods you can call on coroutine objects (or to create
345them).
349 346
350=over 4 347=over 4
351 348
352=item new Coro \&sub [, @args...] 349=item new Coro \&sub [, @args...]
353 350
354Create a new coroutine and return it. When the sub returns the coroutine 351Create a new coroutine and return it. When the sub returns, the coroutine
355automatically terminates as if C<terminate> with the returned values were 352automatically terminates as if C<terminate> with the returned values were
356called. To make the coroutine run you must first put it into the ready queue 353called. To make the coroutine run you must first put it into the ready
357by calling the ready method. 354queue by calling the ready method.
358 355
359See C<async> and C<Coro::State::new> for additional info about the 356See C<async> and C<Coro::State::new> for additional info about the
360coroutine environment. 357coroutine environment.
361 358
362=cut 359=cut
363 360
364sub _run_coro { 361sub _coro_run {
365 terminate &{+shift}; 362 terminate &{+shift};
366} 363}
367 364
368sub new {
369 my $class = shift;
370
371 $class->SUPER::new (\&_run_coro, @_)
372}
373
374=item $success = $coroutine->ready 365=item $success = $coroutine->ready
375 366
376Put the given coroutine into the ready queue (according to it's priority) 367Put the given coroutine into the end of its ready queue (there is one
377and return true. If the coroutine is already in the ready queue, do nothing 368queue for each priority) and return true. If the coroutine is already in
378and return false. 369the ready queue, do nothing and return false.
370
371This ensures that the scheduler will resume this coroutine automatically
372once all the coroutines of higher priority and all coroutines of the same
373priority that were put into the ready queue earlier have been resumed.
379 374
380=item $is_ready = $coroutine->is_ready 375=item $is_ready = $coroutine->is_ready
381 376
382Return wether the coroutine is currently the ready queue or not, 377Return whether the coroutine is currently the ready queue or not,
383 378
384=item $coroutine->cancel (arg...) 379=item $coroutine->cancel (arg...)
385 380
386Terminates the given coroutine and makes it return the given arguments as 381Terminates the given coroutine and makes it return the given arguments as
387status (default: the empty list). Never returns if the coroutine is the 382status (default: the empty list). Never returns if the coroutine is the
… …
389 384
390=cut 385=cut
391 386
392sub cancel { 387sub cancel {
393 my $self = shift; 388 my $self = shift;
394 $self->{_status} = [@_];
395 389
396 if ($current == $self) { 390 if ($current == $self) {
397 push @destroy, $self; 391 terminate @_;
398 $manager->ready;
399 &schedule while 1;
400 } else { 392 } else {
393 $self->{_status} = [@_];
401 $self->_cancel; 394 $self->_cancel;
402 } 395 }
403} 396}
397
398=item $coroutine->schedule_to
399
400Puts the current coroutine to sleep (like C<Coro::schedule>), but instead
401of continuing with the next coro from the ready queue, always switch to
402the given coroutine object (regardless of priority etc.). The readyness
403state of that coroutine isn't changed.
404
405This is an advanced method for special cases - I'd love to hear about any
406uses for this one.
407
408=item $coroutine->cede_to
409
410Like C<schedule_to>, but puts the current coroutine into the ready
411queue. This has the effect of temporarily switching to the given
412coroutine, and continuing some time later.
413
414This is an advanced method for special cases - I'd love to hear about any
415uses for this one.
416
417=item $coroutine->throw ([$scalar])
418
419If C<$throw> is specified and defined, it will be thrown as an exception
420inside the coroutine at the next convenient point in time. Otherwise
421clears the exception object.
422
423Coro will check for the exception each time a schedule-like-function
424returns, i.e. after each C<schedule>, C<cede>, C<< Coro::Semaphore->down
425>>, C<< Coro::Handle->readable >> and so on. Most of these functions
426detect this case and return early in case an exception is pending.
427
428The exception object will be thrown "as is" with the specified scalar in
429C<$@>, i.e. if it is a string, no line number or newline will be appended
430(unlike with C<die>).
431
432This can be used as a softer means than C<cancel> to ask a coroutine to
433end itself, although there is no guarantee that the exception will lead to
434termination, and if the exception isn't caught it might well end the whole
435program.
436
437You might also think of C<throw> as being the moral equivalent of
438C<kill>ing a coroutine with a signal (in this case, a scalar).
404 439
405=item $coroutine->join 440=item $coroutine->join
406 441
407Wait until the coroutine terminates and return any values given to the 442Wait until the coroutine terminates and return any values given to the
408C<terminate> or C<cancel> functions. C<join> can be called concurrently 443C<terminate> or C<cancel> functions. C<join> can be called concurrently
409from multiple coroutines. 444from multiple coroutines, and all will be resumed and given the status
445return once the C<$coroutine> terminates.
410 446
411=cut 447=cut
412 448
413sub join { 449sub join {
414 my $self = shift; 450 my $self = shift;
… …
429 465
430=item $coroutine->on_destroy (\&cb) 466=item $coroutine->on_destroy (\&cb)
431 467
432Registers a callback that is called when this coroutine gets destroyed, 468Registers a callback that is called when this coroutine gets destroyed,
433but before it is joined. The callback gets passed the terminate arguments, 469but before it is joined. The callback gets passed the terminate arguments,
434if any. 470if any, and I<must not> die, under any circumstances.
435 471
436=cut 472=cut
437 473
438sub on_destroy { 474sub on_destroy {
439 my ($self, $cb) = @_; 475 my ($self, $cb) = @_;
… …
469higher values mean lower priority, just as in unix). 505higher values mean lower priority, just as in unix).
470 506
471=item $olddesc = $coroutine->desc ($newdesc) 507=item $olddesc = $coroutine->desc ($newdesc)
472 508
473Sets (or gets in case the argument is missing) the description for this 509Sets (or gets in case the argument is missing) the description for this
474coroutine. This is just a free-form string you can associate with a coroutine. 510coroutine. This is just a free-form string you can associate with a
511coroutine.
475 512
476This method simply sets the C<< $coroutine->{desc} >> member to the given string. You 513This method simply sets the C<< $coroutine->{desc} >> member to the given
477can modify this member directly if you wish. 514string. You can modify this member directly if you wish.
478
479=item $coroutine->throw ([$scalar])
480
481If C<$throw> is specified and defined, it will be thrown as an exception
482inside the coroutine at the next convinient point in time (usually after
483it gains control at the next schedule/transfer/cede). Otherwise clears the
484exception object.
485
486The exception object will be thrown "as is" with the specified scalar in
487C<$@>, i.e. if it is a string, no line number or newline will be appended
488(unlike with C<die>).
489
490This can be used as a softer means than C<cancel> to ask a coroutine to
491end itself, although there is no guarentee that the exception will lead to
492termination, and if the exception isn't caught it might well end the whole
493program.
494 515
495=cut 516=cut
496 517
497sub desc { 518sub desc {
498 my $old = $_[0]{desc}; 519 my $old = $_[0]{desc};
499 $_[0]{desc} = $_[1] if @_ > 1; 520 $_[0]{desc} = $_[1] if @_ > 1;
500 $old; 521 $old;
501} 522}
502 523
524sub transfer {
525 require Carp;
526 Carp::croak ("You must not call ->transfer on Coro objects. Use Coro::State objects or the ->schedule_to method. Caught");
527}
528
503=back 529=back
504 530
505=head2 GLOBAL FUNCTIONS 531=head1 GLOBAL FUNCTIONS
506 532
507=over 4 533=over 4
508 534
509=item Coro::nready 535=item Coro::nready
510 536
511Returns the number of coroutines that are currently in the ready state, 537Returns the number of coroutines that are currently in the ready state,
512i.e. that can be switched to. The value C<0> means that the only runnable 538i.e. that can be switched to by calling C<schedule> directory or
539indirectly. The value C<0> means that the only runnable coroutine is the
513coroutine is the currently running one, so C<cede> would have no effect, 540currently running one, so C<cede> would have no effect, and C<schedule>
514and C<schedule> would cause a deadlock unless there is an idle handler 541would cause a deadlock unless there is an idle handler that wakes up some
515that wakes up some coroutines. 542coroutines.
516 543
517=item my $guard = Coro::guard { ... } 544=item my $guard = Coro::guard { ... }
518 545
519This creates and returns a guard object. Nothing happens until the object 546This function still exists, but is deprecated. Please use the
520gets destroyed, in which case the codeblock given as argument will be 547C<Guard::guard> function instead.
521executed. This is useful to free locks or other resources in case of a
522runtime error or when the coroutine gets canceled, as in both cases the
523guard block will be executed. The guard object supports only one method,
524C<< ->cancel >>, which will keep the codeblock from being executed.
525 548
526Example: set some flag and clear it again when the coroutine gets canceled
527or the function returns:
528
529 sub do_something {
530 my $guard = Coro::guard { $busy = 0 };
531 $busy = 1;
532
533 # do something that requires $busy to be true
534 }
535
536=cut 549=cut
537 550
538sub guard(&) { 551BEGIN { *guard = \&Guard::guard }
539 bless \(my $cb = $_[0]), "Coro::guard"
540}
541
542sub Coro::guard::cancel {
543 ${$_[0]} = sub { };
544}
545
546sub Coro::guard::DESTROY {
547 ${$_[0]}->();
548}
549
550 552
551=item unblock_sub { ... } 553=item unblock_sub { ... }
552 554
553This utility function takes a BLOCK or code reference and "unblocks" it, 555This utility function takes a BLOCK or code reference and "unblocks" it,
554returning the new coderef. This means that the new coderef will return 556returning a new coderef. Unblocking means that calling the new coderef
555immediately without blocking, returning nothing, while the original code 557will return immediately without blocking, returning nothing, while the
556ref will be called (with parameters) from within its own coroutine. 558original code ref will be called (with parameters) from within another
559coroutine.
557 560
558The reason this function exists is that many event libraries (such as the 561The reason this function exists is that many event libraries (such as the
559venerable L<Event|Event> module) are not coroutine-safe (a weaker form 562venerable L<Event|Event> module) are not coroutine-safe (a weaker form
560of thread-safety). This means you must not block within event callbacks, 563of reentrancy). This means you must not block within event callbacks,
561otherwise you might suffer from crashes or worse. 564otherwise you might suffer from crashes or worse. The only event library
565currently known that is safe to use without C<unblock_sub> is L<EV>.
562 566
563This function allows your callbacks to block by executing them in another 567This function allows your callbacks to block by executing them in another
564coroutine where it is safe to block. One example where blocking is handy 568coroutine where it is safe to block. One example where blocking is handy
565is when you use the L<Coro::AIO|Coro::AIO> functions to save results to 569is when you use the L<Coro::AIO|Coro::AIO> functions to save results to
566disk. 570disk, for example.
567 571
568In short: simply use C<unblock_sub { ... }> instead of C<sub { ... }> when 572In short: simply use C<unblock_sub { ... }> instead of C<sub { ... }> when
569creating event callbacks that want to block. 573creating event callbacks that want to block.
574
575If your handler does not plan to block (e.g. simply sends a message to
576another coroutine, or puts some other coroutine into the ready queue),
577there is no reason to use C<unblock_sub>.
578
579Note that you also need to use C<unblock_sub> for any other callbacks that
580are indirectly executed by any C-based event loop. For example, when you
581use a module that uses L<AnyEvent> (and you use L<Coro::AnyEvent>) and it
582provides callbacks that are the result of some event callback, then you
583must not block either, or use C<unblock_sub>.
570 584
571=cut 585=cut
572 586
573our @unblock_queue; 587our @unblock_queue;
574 588
… …
577# return immediately and can be reused) and because we cannot cede 591# return immediately and can be reused) and because we cannot cede
578# inside an event callback. 592# inside an event callback.
579our $unblock_scheduler = new Coro sub { 593our $unblock_scheduler = new Coro sub {
580 while () { 594 while () {
581 while (my $cb = pop @unblock_queue) { 595 while (my $cb = pop @unblock_queue) {
582 # this is an inlined copy of async_pool 596 &async_pool (@$cb);
583 my $coro = (pop @async_pool) || new Coro \&pool_handler;
584 597
585 $coro->{_invoke} = $cb;
586 $coro->ready;
587 cede; # for short-lived callbacks, this reduces pressure on the coro pool 598 # for short-lived callbacks, this reduces pressure on the coro pool
599 # as the chance is very high that the async_poll coro will be back
600 # in the idle state when cede returns
601 cede;
588 } 602 }
589 schedule; # sleep well 603 schedule; # sleep well
590 } 604 }
591}; 605};
592$unblock_scheduler->desc ("[unblock_sub scheduler]"); 606$unblock_scheduler->{desc} = "[unblock_sub scheduler]";
593 607
594sub unblock_sub(&) { 608sub unblock_sub(&) {
595 my $cb = shift; 609 my $cb = shift;
596 610
597 sub { 611 sub {
598 unshift @unblock_queue, [$cb, @_]; 612 unshift @unblock_queue, [$cb, @_];
599 $unblock_scheduler->ready; 613 $unblock_scheduler->ready;
600 } 614 }
601} 615}
602 616
617=item $cb = Coro::rouse_cb
618
619Create and return a "rouse callback". That's a code reference that,
620when called, will remember a copy of its arguments and notify the owner
621coroutine of the callback.
622
623See the next function.
624
625=item @args = Coro::rouse_wait [$cb]
626
627Wait for the specified rouse callback (or the last one that was created in
628this coroutine).
629
630As soon as the callback is invoked (or when the callback was invoked
631before C<rouse_wait>), it will return the arguments originally passed to
632the rouse callback.
633
634See the section B<HOW TO WAIT FOR A CALLBACK> for an actual usage example.
635
603=back 636=back
604 637
605=cut 638=cut
606 639
6071; 6401;
608 641
642=head1 HOW TO WAIT FOR A CALLBACK
643
644It is very common for a coroutine to wait for some callback to be
645called. This occurs naturally when you use coroutines in an otherwise
646event-based program, or when you use event-based libraries.
647
648These typically register a callback for some event, and call that callback
649when the event occured. In a coroutine, however, you typically want to
650just wait for the event, simplyifying things.
651
652For example C<< AnyEvent->child >> registers a callback to be called when
653a specific child has exited:
654
655 my $child_watcher = AnyEvent->child (pid => $pid, cb => sub { ... });
656
657But from withina coroutine, you often just want to write this:
658
659 my $status = wait_for_child $pid;
660
661Coro offers two functions specifically designed to make this easy,
662C<Coro::rouse_cb> and C<Coro::rouse_wait>.
663
664The first function, C<rouse_cb>, generates and returns a callback that,
665when invoked, will save its arguments and notify the coroutine that
666created the callback.
667
668The second function, C<rouse_wait>, waits for the callback to be called
669(by calling C<schedule> to go to sleep) and returns the arguments
670originally passed to the callback.
671
672Using these functions, it becomes easy to write the C<wait_for_child>
673function mentioned above:
674
675 sub wait_for_child($) {
676 my ($pid) = @_;
677
678 my $watcher = AnyEvent->child (pid => $pid, cb => Coro::rouse_cb);
679
680 my ($rpid, $rstatus) = Coro::rouse_wait;
681 $rstatus
682 }
683
684In the case where C<rouse_cb> and C<rouse_wait> are not flexible enough,
685you can roll your own, using C<schedule>:
686
687 sub wait_for_child($) {
688 my ($pid) = @_;
689
690 # store the current coroutine in $current,
691 # and provide result variables for the closure passed to ->child
692 my $current = $Coro::current;
693 my ($done, $rstatus);
694
695 # pass a closure to ->child
696 my $watcher = AnyEvent->child (pid => $pid, cb => sub {
697 $rstatus = $_[1]; # remember rstatus
698 $done = 1; # mark $rstatus as valud
699 });
700
701 # wait until the closure has been called
702 schedule while !$done;
703
704 $rstatus
705 }
706
707
609=head1 BUGS/LIMITATIONS 708=head1 BUGS/LIMITATIONS
610 709
611 - you must make very sure that no coro is still active on global 710=over 4
612 destruction. very bad things might happen otherwise (usually segfaults).
613 711
712=item fork with pthread backend
713
714When Coro is compiled using the pthread backend (which isn't recommended
715but required on many BSDs as their libcs are completely broken), then
716coroutines will not survive a fork. There is no known workaround except to
717fix your libc and use a saner backend.
718
719=item perl process emulation ("threads")
720
614 - this module is not thread-safe. You should only ever use this module 721This module is not perl-pseudo-thread-safe. You should only ever use this
615 from the same thread (this requirement might be loosened in the future 722module from the first thread (this requirement might be removed in the
616 to allow per-thread schedulers, but Coro::State does not yet allow 723future to allow per-thread schedulers, but Coro::State does not yet allow
617 this). 724this). I recommend disabling thread support and using processes, as having
725the windows process emulation enabled under unix roughly halves perl
726performance, even when not used.
727
728=item coroutine switching not signal safe
729
730You must not switch to another coroutine from within a signal handler
731(only relevant with %SIG - most event libraries provide safe signals).
732
733That means you I<MUST NOT> call any function that might "block" the
734current coroutine - C<cede>, C<schedule> C<< Coro::Semaphore->down >> or
735anything that calls those. Everything else, including calling C<ready>,
736works.
737
738=back
739
618 740
619=head1 SEE ALSO 741=head1 SEE ALSO
620 742
621Lower level Configuration, Coroutine Environment: L<Coro::State>. 743Event-Loop integration: L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>.
622 744
623Debugging: L<Coro::Debug>. 745Debugging: L<Coro::Debug>.
624 746
625Support/Utility: L<Coro::Specific>, L<Coro::Util>. 747Support/Utility: L<Coro::Specific>, L<Coro::Util>.
626 748
627Locking/IPC: L<Coro::Signal>, L<Coro::Channel>, L<Coro::Semaphore>, L<Coro::SemaphoreSet>, L<Coro::RWLock>. 749Locking and IPC: L<Coro::Signal>, L<Coro::Channel>, L<Coro::Semaphore>,
750L<Coro::SemaphoreSet>, L<Coro::RWLock>.
628 751
629Event/IO: L<Coro::Timer>, L<Coro::Event>, L<Coro::Handle>, L<Coro::Socket>. 752I/O and Timers: L<Coro::Timer>, L<Coro::Handle>, L<Coro::Socket>, L<Coro::AIO>.
630 753
631Compatibility: L<Coro::LWP>, L<Coro::Storable>, L<Coro::Select>. 754Compatibility with other modules: L<Coro::LWP> (but see also L<AnyEvent::HTTP> for
755a better-working alternative), L<Coro::BDB>, L<Coro::Storable>,
756L<Coro::Select>.
632 757
633Embedding: L<Coro::MakeMaker>. 758XS API: L<Coro::MakeMaker>.
759
760Low level Configuration, Thread Environment, Continuations: L<Coro::State>.
634 761
635=head1 AUTHOR 762=head1 AUTHOR
636 763
637 Marc Lehmann <schmorp@schmorp.de> 764 Marc Lehmann <schmorp@schmorp.de>
638 http://home.schmorp.de/ 765 http://home.schmorp.de/

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines