… | |
… | |
55 | |
55 | |
56 | =head1 DESCRIPTION |
56 | =head1 DESCRIPTION |
57 | |
57 | |
58 | This module provides an interface to libev |
58 | This module provides an interface to libev |
59 | (L<http://software.schmorp.de/pkg/libev.html>). While the documentation |
59 | (L<http://software.schmorp.de/pkg/libev.html>). While the documentation |
60 | below is comprehensive, one might also consult the documentation of libev |
60 | below is comprehensive, one might also consult the documentation of |
61 | itself (L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>) for more |
61 | libev itself (L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod> or |
62 | subtle details on watcher semantics or some discussion on the available |
62 | F<perldoc EV::libev>) for more subtle details on watcher semantics or some |
63 | backends, or how to force a specific backend with C<LIBEV_FLAGS>, or just |
63 | discussion on the available backends, or how to force a specific backend |
64 | about in any case because it has much more detailed information. |
64 | with C<LIBEV_FLAGS>, or just about in any case because it has much more |
|
|
65 | detailed information. |
65 | |
66 | |
66 | This module is very fast and scalable. It is actually so fast that you |
67 | This module is very fast and scalable. It is actually so fast that you |
67 | can use it through the L<AnyEvent> module, stay portable to other event |
68 | can use it through the L<AnyEvent> module, stay portable to other event |
68 | loops (if you don't rely on any watcher types not available through it) |
69 | loops (if you don't rely on any watcher types not available through it) |
69 | and still be faster than with any other event loop currently supported in |
70 | and still be faster than with any other event loop currently supported in |
70 | Perl. |
71 | Perl. |
71 | |
72 | |
|
|
73 | =head2 MODULE EXPORTS |
|
|
74 | |
|
|
75 | This module does not export any symbols. |
|
|
76 | |
72 | =cut |
77 | =cut |
73 | |
78 | |
74 | package EV; |
79 | package EV; |
75 | |
80 | |
76 | use strict; |
81 | use common::sense; |
77 | |
82 | |
78 | BEGIN { |
83 | BEGIN { |
79 | our $VERSION = '3.42'; |
84 | our $VERSION = '3.9'; |
80 | use XSLoader; |
85 | use XSLoader; |
81 | XSLoader::load "EV", $VERSION; |
86 | XSLoader::load "EV", $VERSION; |
82 | } |
87 | } |
83 | |
88 | |
84 | @EV::IO::ISA = |
89 | @EV::IO::ISA = |
… | |
… | |
111 | default loop as this is fastest (perl-wise), best supported by other |
116 | default loop as this is fastest (perl-wise), best supported by other |
112 | modules (e.g. AnyEvent or Coro) and most portable event loop. |
117 | modules (e.g. AnyEvent or Coro) and most portable event loop. |
113 | |
118 | |
114 | For specific programs you can create additional event loops dynamically. |
119 | For specific programs you can create additional event loops dynamically. |
115 | |
120 | |
116 | If you want to take avdantage of kqueue (which often works properly for |
121 | If you want to take advantage of kqueue (which often works properly for |
117 | sockets only) even though the default loop doesn't enable it, you can |
122 | sockets only) even though the default loop doesn't enable it, you can |
118 | I<embed> a kqueue loop into the default loop: running the default loop |
123 | I<embed> a kqueue loop into the default loop: running the default loop |
119 | will then also service the kqueue loop to some extent. See the example in |
124 | will then also service the kqueue loop to some extent. See the example in |
120 | the section about embed watchers for an example on how to achieve that. |
125 | the section about embed watchers for an example on how to achieve that. |
121 | |
126 | |
122 | =over 4 |
127 | =over 4 |
123 | |
128 | |
124 | =item $loop = new EV::loop [$flags] |
129 | =item $loop = new EV::Loop [$flags] |
125 | |
130 | |
126 | Create a new event loop as per the specified flags. Please refer to the |
131 | Create a new event loop as per the specified flags. Please refer to |
127 | C<ev_loop_new ()> function description in the libev documentation |
132 | the C<ev_loop_new ()> function description in the libev documentation |
128 | (L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#GLOBAL_FUNCTIONS>) |
133 | (L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#GLOBAL_FUNCTIONS>, |
129 | for more info. |
134 | or locally-installed as F<EV::libev> manpage) for more info. |
130 | |
135 | |
131 | The loop will automatically be destroyed when it is no longer referenced |
136 | The loop will automatically be destroyed when it is no longer referenced |
132 | by any watcher and the loop object goes out of scope. |
137 | by any watcher and the loop object goes out of scope. |
133 | |
138 | |
134 | Using C<EV::FLAG_FORKCHECK> is recommended, as only the default event loop |
139 | If you are not embedding the loop, then Using C<EV::FLAG_FORKCHECK> |
135 | is protected by this module. |
140 | is recommended, as only the default event loop is protected by this |
|
|
141 | module. If you I<are> embedding this loop in the default loop, this is not |
|
|
142 | necessary, as C<EV::embed> automatically does the right thing on fork. |
136 | |
143 | |
137 | =item $loop->loop_fork |
144 | =item $loop->loop_fork |
138 | |
145 | |
139 | Must be called after a fork in the child, before entering or continuing |
146 | Must be called after a fork in the child, before entering or continuing |
140 | the event loop. An alternative is to use C<EV::FLAG_FORKCHECK> which calls |
147 | the event loop. An alternative is to use C<EV::FLAG_FORKCHECK> which calls |
… | |
… | |
191 | =item $time = EV::now |
198 | =item $time = EV::now |
192 | |
199 | |
193 | =item $time = $loop->now |
200 | =item $time = $loop->now |
194 | |
201 | |
195 | Returns the time the last event loop iteration has been started. This |
202 | Returns the time the last event loop iteration has been started. This |
196 | is the time that (relative) timers are based on, and refering to it is |
203 | is the time that (relative) timers are based on, and referring to it is |
197 | usually faster then calling EV::time. |
204 | usually faster then calling EV::time. |
198 | |
205 | |
|
|
206 | =item EV::now_update |
|
|
207 | |
|
|
208 | =item $loop->now_update |
|
|
209 | |
|
|
210 | Establishes the current time by querying the kernel, updating the time |
|
|
211 | returned by C<EV::now> in the progress. This is a costly operation and |
|
|
212 | is usually done automatically within C<EV::loop>. |
|
|
213 | |
|
|
214 | This function is rarely useful, but when some event callback runs for a |
|
|
215 | very long time without entering the event loop, updating libev's idea of |
|
|
216 | the current time is a good idea. |
|
|
217 | |
|
|
218 | =item EV::suspend |
|
|
219 | |
|
|
220 | =item $loop->suspend |
|
|
221 | |
|
|
222 | =item EV::resume |
|
|
223 | |
|
|
224 | =item $loop->resume |
|
|
225 | |
|
|
226 | These two functions suspend and resume a loop, for use when the loop is |
|
|
227 | not used for a while and timeouts should not be processed. |
|
|
228 | |
|
|
229 | A typical use case would be an interactive program such as a game: When |
|
|
230 | the user presses C<^Z> to suspend the game and resumes it an hour later it |
|
|
231 | would be best to handle timeouts as if no time had actually passed while |
|
|
232 | the program was suspended. This can be achieved by calling C<suspend> |
|
|
233 | in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling |
|
|
234 | C<resume> directly afterwards to resume timer processing. |
|
|
235 | |
|
|
236 | Effectively, all C<timer> watchers will be delayed by the time spend |
|
|
237 | between C<suspend> and C<resume>, and all C<periodic> watchers |
|
|
238 | will be rescheduled (that is, they will lose any events that would have |
|
|
239 | occured while suspended). |
|
|
240 | |
|
|
241 | After calling C<suspend> you B<must not> call I<any> function on the given |
|
|
242 | loop other than C<resume>, and you B<must not> call C<resume> |
|
|
243 | without a previous call to C<suspend>. |
|
|
244 | |
|
|
245 | Calling C<suspend>/C<resume> has the side effect of updating the event |
|
|
246 | loop time (see C<now_update>). |
|
|
247 | |
199 | =item $backend = EV::backend |
248 | =item $backend = EV::backend |
200 | |
249 | |
201 | =item $backend = $loop->backend |
250 | =item $backend = $loop->backend |
202 | |
251 | |
203 | Returns an integer describing the backend used by libev (EV::METHOD_SELECT |
252 | Returns an integer describing the backend used by libev (EV::BACKEND_SELECT |
204 | or EV::METHOD_EPOLL). |
253 | or EV::BACKEND_EPOLL). |
205 | |
254 | |
206 | =item EV::loop [$flags] |
255 | =item EV::loop [$flags] |
207 | |
256 | |
208 | =item $loop->loop ([$flags]) |
257 | =item $loop->loop ([$flags]) |
209 | |
258 | |
… | |
… | |
229 | =item $count = EV::loop_count |
278 | =item $count = EV::loop_count |
230 | |
279 | |
231 | =item $count = $loop->loop_count |
280 | =item $count = $loop->loop_count |
232 | |
281 | |
233 | Return the number of times the event loop has polled for new |
282 | Return the number of times the event loop has polled for new |
234 | events. Sometiems useful as a generation counter. |
283 | events. Sometimes useful as a generation counter. |
235 | |
284 | |
236 | =item EV::once $fh_or_undef, $events, $timeout, $cb->($revents) |
285 | =item EV::once $fh_or_undef, $events, $timeout, $cb->($revents) |
237 | |
286 | |
238 | =item $loop->once ($fh_or_undef, $events, $timeout, $cb->($revents)) |
287 | =item $loop->once ($fh_or_undef, $events, $timeout, $cb->($revents)) |
239 | |
288 | |
… | |
… | |
279 | |
328 | |
280 | =item $loop->set_timeout_collect_interval ($time) |
329 | =item $loop->set_timeout_collect_interval ($time) |
281 | |
330 | |
282 | These advanced functions set the minimum block interval when polling for I/O events and the minimum |
331 | These advanced functions set the minimum block interval when polling for I/O events and the minimum |
283 | wait interval for timer events. See the libev documentation at |
332 | wait interval for timer events. See the libev documentation at |
284 | L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#FUNCTIONS_CONTROLLING_THE_EVENT_LOOP> for |
333 | L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#FUNCTIONS_CONTROLLING_THE_EVENT_LOOP> |
285 | a more detailed discussion. |
334 | (locally installed as F<EV::libev>) for a more detailed discussion. |
|
|
335 | |
|
|
336 | =item $count = EV::pending_count |
|
|
337 | |
|
|
338 | =item $count = $loop->pending_count |
|
|
339 | |
|
|
340 | Returns the number of currently pending watchers. |
|
|
341 | |
|
|
342 | =item EV::invoke_pending |
|
|
343 | |
|
|
344 | =item $loop->invoke_pending |
|
|
345 | |
|
|
346 | Invoke all currently pending watchers. |
286 | |
347 | |
287 | =back |
348 | =back |
288 | |
349 | |
289 | |
350 | |
290 | =head1 WATCHER OBJECTS |
351 | =head1 WATCHER OBJECTS |
… | |
… | |
303 | called with at least two arguments: the watcher and a bitmask of received |
364 | called with at least two arguments: the watcher and a bitmask of received |
304 | events. |
365 | events. |
305 | |
366 | |
306 | Each watcher type has its associated bit in revents, so you can use the |
367 | Each watcher type has its associated bit in revents, so you can use the |
307 | same callback for multiple watchers. The event mask is named after the |
368 | same callback for multiple watchers. The event mask is named after the |
308 | type, i..e. EV::child sets EV::CHILD, EV::prepare sets EV::PREPARE, |
369 | type, i.e. EV::child sets EV::CHILD, EV::prepare sets EV::PREPARE, |
309 | EV::periodic sets EV::PERIODIC and so on, with the exception of I/O events |
370 | EV::periodic sets EV::PERIODIC and so on, with the exception of I/O events |
310 | (which can set both EV::READ and EV::WRITE bits), and EV::timer (which |
371 | (which can set both EV::READ and EV::WRITE bits), and EV::timer (which |
311 | uses EV::TIMEOUT). |
372 | uses EV::TIMEOUT). |
312 | |
373 | |
313 | In the rare case where one wants to create a watcher but not start it at |
374 | In the rare case where one wants to create a watcher but not start it at |
… | |
… | |
395 | |
456 | |
396 | =item $previous_state = $w->keepalive ($bool) |
457 | =item $previous_state = $w->keepalive ($bool) |
397 | |
458 | |
398 | Normally, C<EV::loop> will return when there are no active watchers |
459 | Normally, C<EV::loop> will return when there are no active watchers |
399 | (which is a "deadlock" because no progress can be made anymore). This is |
460 | (which is a "deadlock" because no progress can be made anymore). This is |
400 | convinient because it allows you to start your watchers (and your jobs), |
461 | convenient because it allows you to start your watchers (and your jobs), |
401 | call C<EV::loop> once and when it returns you know that all your jobs are |
462 | call C<EV::loop> once and when it returns you know that all your jobs are |
402 | finished (or they forgot to register some watchers for their task :). |
463 | finished (or they forgot to register some watchers for their task :). |
403 | |
464 | |
404 | Sometimes, however, this gets in your way, for example when the module |
465 | Sometimes, however, this gets in your way, for example when the module |
405 | that calls C<EV::loop> (usually the main program) is not the same module |
466 | that calls C<EV::loop> (usually the main program) is not the same module |
… | |
… | |
409 | because you happen to have this long-running UDP port watcher. |
470 | because you happen to have this long-running UDP port watcher. |
410 | |
471 | |
411 | In this case you can clear the keepalive status, which means that even |
472 | In this case you can clear the keepalive status, which means that even |
412 | though your watcher is active, it won't keep C<EV::loop> from returning. |
473 | though your watcher is active, it won't keep C<EV::loop> from returning. |
413 | |
474 | |
414 | The initial value for keepalive is true (enabled), and you cna change it |
475 | The initial value for keepalive is true (enabled), and you can change it |
415 | any time. |
476 | any time. |
416 | |
477 | |
417 | Example: Register an I/O watcher for some UDP socket but do not keep the |
478 | Example: Register an I/O watcher for some UDP socket but do not keep the |
418 | event loop from running just because of that watcher. |
479 | event loop from running just because of that watcher. |
419 | |
480 | |
… | |
… | |
637 | |
698 | |
638 | =item $w = EV::signal $signal, $callback |
699 | =item $w = EV::signal $signal, $callback |
639 | |
700 | |
640 | =item $w = EV::signal_ns $signal, $callback |
701 | =item $w = EV::signal_ns $signal, $callback |
641 | |
702 | |
|
|
703 | =item $w = $loop->signal ($signal, $callback) |
|
|
704 | |
|
|
705 | =item $w = $loop->signal_ns ($signal, $callback) |
|
|
706 | |
642 | Call the callback when $signal is received (the signal can be specified by |
707 | Call the callback when $signal is received (the signal can be specified by |
643 | number or by name, just as with C<kill> or C<%SIG>). |
708 | number or by name, just as with C<kill> or C<%SIG>). |
|
|
709 | |
|
|
710 | Only one event loop can grab a given signal - attempting to grab the same |
|
|
711 | signal from two EV loops will crash the program immediately or cause data |
|
|
712 | corruption. |
644 | |
713 | |
645 | EV will grab the signal for the process (the kernel only allows one |
714 | EV will grab the signal for the process (the kernel only allows one |
646 | component to receive a signal at a time) when you start a signal watcher, |
715 | component to receive a signal at a time) when you start a signal watcher, |
647 | and removes it again when you stop it. Perl does the same when you |
716 | and removes it again when you stop it. Perl does the same when you |
648 | add/remove callbacks to C<%SIG>, so watch out. |
717 | add/remove callbacks to C<%SIG>, so watch out. |
… | |
… | |
913 | # make the dispatcher handle any new stuff |
982 | # make the dispatcher handle any new stuff |
914 | ... not shown |
983 | ... not shown |
915 | }; |
984 | }; |
916 | |
985 | |
917 | The callbacks of the created watchers will not be called as the watchers |
986 | The callbacks of the created watchers will not be called as the watchers |
918 | are destroyed before this cna happen (remember EV::check gets called |
987 | are destroyed before this can happen (remember EV::check gets called |
919 | first). |
988 | first). |
920 | |
989 | |
921 | The C<check_ns> variant doesn't start (activate) the newly created watcher. |
990 | The C<check_ns> variant doesn't start (activate) the newly created watcher. |
922 | |
991 | |
923 | =back |
992 | =back |
… | |
… | |
954 | loop, other types of watchers might be handled in a delayed or incorrect |
1023 | loop, other types of watchers might be handled in a delayed or incorrect |
955 | fashion and must not be used). |
1024 | fashion and must not be used). |
956 | |
1025 | |
957 | See the libev documentation at |
1026 | See the libev documentation at |
958 | L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#code_ev_embed_code_when_one_backend_> |
1027 | L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#code_ev_embed_code_when_one_backend_> |
959 | for more details. |
1028 | (locally installed as F<EV::libev>) for more details. |
960 | |
1029 | |
961 | In short, this watcher is most useful on BSD systems without working |
1030 | In short, this watcher is most useful on BSD systems without working |
962 | kqueue to still be able to handle a large number of sockets: |
1031 | kqueue to still be able to handle a large number of sockets: |
963 | |
1032 | |
964 | my $socket_loop; |
1033 | my $socket_loop; |
… | |
… | |
975 | # use the default loop otherwise |
1044 | # use the default loop otherwise |
976 | $socket_loop ||= EV::default_loop; |
1045 | $socket_loop ||= EV::default_loop; |
977 | |
1046 | |
978 | =over 4 |
1047 | =over 4 |
979 | |
1048 | |
980 | =item $w = EV::embed $otherloop, $callback |
1049 | =item $w = EV::embed $otherloop[, $callback] |
981 | |
1050 | |
982 | =item $w = EV::embed_ns $otherloop, $callback |
1051 | =item $w = EV::embed_ns $otherloop[, $callback] |
983 | |
1052 | |
984 | =item $w = $loop->embed ($otherloop, $callback) |
1053 | =item $w = $loop->embed ($otherloop[, $callback]) |
985 | |
1054 | |
986 | =item $w = $loop->embed_ns ($otherloop, $callback) |
1055 | =item $w = $loop->embed_ns ($otherloop[, $callback]) |
987 | |
1056 | |
988 | Call the callback when the embedded event loop (C<$otherloop>) has any |
1057 | Call the callback when the embedded event loop (C<$otherloop>) has any |
989 | I/O activity. The C<$callback> should alwas be specified as C<undef> in |
1058 | I/O activity. The C<$callback> is optional: if it is missing, then the |
990 | this version of EV, which means the embedded event loop will be managed |
1059 | embedded event loop will be managed automatically (which is recommended), |
991 | automatically. |
1060 | otherwise you have to invoke C<sweep> yourself. |
992 | |
1061 | |
993 | The C<embed_ns> variant doesn't start (activate) the newly created watcher. |
1062 | The C<embed_ns> variant doesn't start (activate) the newly created watcher. |
994 | |
1063 | |
995 | =back |
1064 | =back |
996 | |
1065 | |
997 | =head3 ASYNC WATCHERS - how to wake up another event loop |
1066 | =head3 ASYNC WATCHERS - how to wake up another event loop |
998 | |
1067 | |
999 | Async watchers are provided by EV, but have little use in perl directly, as perl |
1068 | Async watchers are provided by EV, but have little use in perl directly, |
1000 | neither supports threads nor direct access to signal handlers or other |
1069 | as perl neither supports threads running in parallel nor direct access to |
1001 | contexts where they could be of value. |
1070 | signal handlers or other contexts where they could be of value. |
1002 | |
1071 | |
1003 | It is, however, possible to use them from the XS level. |
1072 | It is, however, possible to use them from the XS level. |
1004 | |
1073 | |
1005 | Please see the libev documentation for further details. |
1074 | Please see the libev documentation for further details. |
1006 | |
1075 | |
… | |
… | |
1034 | my $async_check = EV::check sub { }; |
1103 | my $async_check = EV::check sub { }; |
1035 | |
1104 | |
1036 | This ensures that perl gets into control for a short time to handle any |
1105 | This ensures that perl gets into control for a short time to handle any |
1037 | pending signals, and also ensures (slightly) slower overall operation. |
1106 | pending signals, and also ensures (slightly) slower overall operation. |
1038 | |
1107 | |
1039 | =head1 THREADS |
1108 | =head1 ITHREADS |
1040 | |
1109 | |
1041 | Threads are not supported by this module in any way. Perl pseudo-threads |
1110 | Ithreads are not supported by this module in any way. Perl pseudo-threads |
1042 | is evil stuff and must die. As soon as Perl gains real threads I will work |
1111 | is evil stuff and must die. Real threads as provided by Coro are fully |
1043 | on thread support for it. |
1112 | supported (and enhanced support is available via L<Coro::EV>). |
1044 | |
1113 | |
1045 | =head1 FORK |
1114 | =head1 FORK |
1046 | |
1115 | |
1047 | Most of the "improved" event delivering mechanisms of modern operating |
1116 | Most of the "improved" event delivering mechanisms of modern operating |
1048 | systems have quite a few problems with fork(2) (to put it bluntly: it is |
1117 | systems have quite a few problems with fork(2) (to put it bluntly: it is |