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

Comparing EV/EV.pm (file contents):
Revision 1.6 by root, Sat Oct 27 07:30:10 2007 UTC vs.
Revision 1.103 by root, Thu Oct 2 07:49:09 2008 UTC

1=head1 NAME 1=head1 NAME
2 2
3EV - perl interface to libevent, monkey.org/~provos/libevent/ 3EV - perl interface to libev, a high performance full-featured event loop
4 4
5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 use EV; 7 use EV;
8 8
9 # TIMER 9 # TIMERS
10 10
11 my $w = EV::timer 2, 0, sub { 11 my $w = EV::timer 2, 0, sub {
12 warn "is called after 2s"; 12 warn "is called after 2s";
13 }; 13 };
14 14
15 my $w = EV::timer 2, 1, sub { 15 my $w = EV::timer 2, 2, sub {
16 warn "is called roughly every 2s (repeat = 1)"; 16 warn "is called roughly every 2s (repeat = 2)";
17 }; 17 };
18 18
19 undef $w; # destroy event watcher again 19 undef $w; # destroy event watcher again
20 20
21 # IO
22
23 my $w = EV::timer_abs 0, 60, sub { 21 my $w = EV::periodic 0, 60, 0, sub {
24 warn "is called every minute, on the minute, exactly"; 22 warn "is called every minute, on the minute, exactly";
25 }; 23 };
26 24
25 # IO
26
27 my $w = EV::io \*STDIN, EV::READ | EV::PERSIST, sub { 27 my $w = EV::io *STDIN, EV::READ, sub {
28 my ($w, $events) = @_; # all callbacks get the watcher object and event mask 28 my ($w, $revents) = @_; # all callbacks receive the watcher and event mask
29 if ($events & EV::TIMEOUT) {
30 warn "nothign received on stdin for 10 seconds, retrying";
31 } else {
32 warn "stdin is readable, you entered: ", <STDIN>; 29 warn "stdin is readable, you entered: ", <STDIN>;
33 } 30 };
31
32 # SIGNALS
33
34 my $w = EV::signal 'QUIT', sub {
35 warn "sigquit received\n";
34 }; 36 };
35 $w->timeout (10); 37
36 38 # CHILD/PID STATUS CHANGES
39
40 my $w = EV::child 666, 0, sub {
41 my ($w, $revents) = @_;
42 my $status = $w->rstatus;
43 };
44
45 # STAT CHANGES
46 my $w = EV::stat "/etc/passwd", 10, sub {
47 my ($w, $revents) = @_;
48 warn $w->path, " has changed somehow.\n";
49 };
50
37 # MAINLOOP 51 # MAINLOOP
38 EV::dispatch; # loop as long as watchers are active 52 EV::loop; # loop until EV::unloop is called or all watchers stop
39 EV::loop; # the same thing 53 EV::loop EV::LOOP_ONESHOT; # block until at least one event could be handled
40 EV::loop EV::LOOP_ONCE; 54 EV::loop EV::LOOP_NONBLOCK; # try to handle same events, but do not block
41 EV::loop EV::LOOP_ONSHOT;
42 55
43=head1 DESCRIPTION 56=head1 DESCRIPTION
44 57
45This module provides an interface to libevent 58This module provides an interface to libev
46(L<http://monkey.org/~provos/libevent/>). You probably should acquaint 59(L<http://software.schmorp.de/pkg/libev.html>). While the documentation
47yourself with its documentation and source code to be able to use this 60below is comprehensive, one might also consult the documentation of libev
48module fully. 61itself (L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>) for more
62subtle details on watcher semantics or some discussion on the available
63backends, or how to force a specific backend with C<LIBEV_FLAGS>, or just
64about in any case because it has much more detailed information.
49 65
50Please note thta this module disables the libevent EPOLL method by 66This module is very fast and scalable. It is actually so fast that you
51default, see BUGS, below, if you need to enable it. 67can use it through the L<AnyEvent> module, stay portable to other event
68loops (if you don't rely on any watcher types not available through it)
69and still be faster than with any other event loop currently supported in
70Perl.
52 71
53=cut 72=cut
54 73
55package EV; 74package EV;
56 75
76no warnings;
57use strict; 77use strict;
58 78
59BEGIN { 79BEGIN {
60 our $VERSION = '0.01'; 80 our $VERSION = '3.44';
61 use XSLoader; 81 use XSLoader;
62 XSLoader::load "EV", $VERSION; 82 XSLoader::load "EV", $VERSION;
63} 83}
64 84
65=head1 FUNCTIONAL INTERFACE 85@EV::IO::ISA =
86@EV::Timer::ISA =
87@EV::Periodic::ISA =
88@EV::Signal::ISA =
89@EV::Child::ISA =
90@EV::Stat::ISA =
91@EV::Idle::ISA =
92@EV::Prepare::ISA =
93@EV::Check::ISA =
94@EV::Embed::ISA =
95@EV::Fork::ISA =
96@EV::Async::ISA =
97 "EV::Watcher";
66 98
67=over 4 99@EV::Loop::Default::ISA = "EV::Loop";
68 100
101=head1 EVENT LOOPS
102
103EV supports multiple event loops: There is a single "default event loop"
104that can handle everything including signals and child watchers, and any
105number of "dynamic event loops" that can use different backends (with
106various limitations), but no child and signal watchers.
107
108You do not have to do anything to create the default event loop: When
109the module is loaded a suitable backend is selected on the premise of
110selecting a working backend (which for example rules out kqueue on most
111BSDs). Modules should, unless they have "special needs" always use the
112default loop as this is fastest (perl-wise), best supported by other
113modules (e.g. AnyEvent or Coro) and most portable event loop.
114
115For specific programs you can create additional event loops dynamically.
116
117If you want to take avdantage of kqueue (which often works properly for
118sockets only) even though the default loop doesn't enable it, you can
119I<embed> a kqueue loop into the default loop: running the default loop
120will then also service the kqueue loop to some extent. See the example in
121the section about embed watchers for an example on how to achieve that.
122
123=over 4
124
125=item $loop = new EV::loop [$flags]
126
127Create a new event loop as per the specified flags. Please refer to the
128C<ev_loop_new ()> function description in the libev documentation
129(L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#GLOBAL_FUNCTIONS>)
130for more info.
131
132The loop will automatically be destroyed when it is no longer referenced
133by any watcher and the loop object goes out of scope.
134
135If you are not embedding the loop, then Using C<EV::FLAG_FORKCHECK>
136is recommended, as only the default event loop is protected by this
137module. If you I<are> embedding this loop in the default loop, this is not
138necessary, as C<EV::embed> automatically does the right thing on fork.
139
140=item $loop->loop_fork
141
142Must be called after a fork in the child, before entering or continuing
143the event loop. An alternative is to use C<EV::FLAG_FORKCHECK> which calls
144this function automatically, at some performance loss (refer to the libev
145documentation).
146
147=item $loop->loop_verify
148
149Calls C<ev_verify> to make internal consistency checks (for debugging
150libev) and abort the program if any data structures were found to be
151corrupted.
152
153=item $loop = EV::default_loop [$flags]
154
155Return the default loop (which is a singleton object). Since this module
156already creates the default loop with default flags, specifying flags here
157will not have any effect unless you destroy the default loop first, which
158isn't supported. So in short: don't do it, and if you break it, you get to
159keep the pieces.
160
161=back
162
163
164=head1 BASIC INTERFACE
165
166=over 4
167
69=item $EV::NPRI 168=item $EV::DIED
70 169
71How many priority levels are available. 170Must contain a reference to a function that is called when a callback
171throws an exception (with $@ containing the error). The default prints an
172informative message and continues.
173
174If this callback throws an exception it will be silently ignored.
175
176=item $flags = EV::supported_backends
177
178=item $flags = EV::recommended_backends
179
180=item $flags = EV::embeddable_backends
181
182Returns the set (see C<EV::BACKEND_*> flags) of backends supported by this
183instance of EV, the set of recommended backends (supposed to be good) for
184this platform and the set of embeddable backends (see EMBED WATCHERS).
185
186=item EV::sleep $seconds
187
188Block the process for the given number of (fractional) seconds.
189
190=item $time = EV::time
191
192Returns the current time in (fractional) seconds since the epoch.
72 193
73=item $time = EV::now 194=item $time = EV::now
74 195
75Returns the time in (fractional) seconds since the epoch. 196=item $time = $loop->now
76 197
77=item $version = EV::version 198Returns the time the last event loop iteration has been started. This
199is the time that (relative) timers are based on, and refering to it is
200usually faster then calling EV::time.
78 201
79=item $method = EV::method 202=item $backend = EV::backend
80 203
81Return version string and event polling method used. 204=item $backend = $loop->backend
82 205
83=item EV::loop $flags # EV::LOOP_ONCE, EV::LOOP_ONESHOT 206Returns an integer describing the backend used by libev (EV::METHOD_SELECT
207or EV::METHOD_EPOLL).
84 208
85=item EV::loopexit $after 209=item EV::loop [$flags]
86 210
87Exit any active loop or dispatch after C<$after> seconds or immediately if 211=item $loop->loop ([$flags])
88C<$after> is missing or zero.
89 212
90=item EV::dispatch 213Begin checking for events and calling callbacks. It returns when a
214callback calls EV::unloop.
91 215
92Same as C<EV::loop 0>. 216The $flags argument can be one of the following:
93 217
94=item EV::event $callback 218 0 as above
219 EV::LOOP_ONESHOT block at most once (wait, but do not loop)
220 EV::LOOP_NONBLOCK do not block at all (fetch/handle events but do not wait)
95 221
96Creates a new event watcher waiting for nothing, calling the given callback. 222=item EV::unloop [$how]
97 223
224=item $loop->unloop ([$how])
225
226When called with no arguments or an argument of EV::UNLOOP_ONE, makes the
227innermost call to EV::loop return.
228
229When called with an argument of EV::UNLOOP_ALL, all calls to EV::loop will return as
230fast as possible.
231
232=item $count = EV::loop_count
233
234=item $count = $loop->loop_count
235
236Return the number of times the event loop has polled for new
237events. Sometiems useful as a generation counter.
238
239=item EV::once $fh_or_undef, $events, $timeout, $cb->($revents)
240
241=item $loop->once ($fh_or_undef, $events, $timeout, $cb->($revents))
242
243This function rolls together an I/O and a timer watcher for a single
244one-shot event without the need for managing a watcher object.
245
246If C<$fh_or_undef> is a filehandle or file descriptor, then C<$events>
247must be a bitset containing either C<EV::READ>, C<EV::WRITE> or C<EV::READ
248| EV::WRITE>, indicating the type of I/O event you want to wait for. If
249you do not want to wait for some I/O event, specify C<undef> for
250C<$fh_or_undef> and C<0> for C<$events>).
251
252If timeout is C<undef> or negative, then there will be no
253timeout. Otherwise a EV::timer with this value will be started.
254
255When an error occurs or either the timeout or I/O watcher triggers, then
256the callback will be called with the received event set (in general
257you can expect it to be a combination of C<EV::ERROR>, C<EV::READ>,
258C<EV::WRITE> and C<EV::TIMEOUT>).
259
260EV::once doesn't return anything: the watchers stay active till either
261of them triggers, then they will be stopped and freed, and the callback
262invoked.
263
264=item EV::feed_fd_event ($fd, $revents)
265
266=item $loop->feed_fd_event ($fd, $revents)
267
268Feed an event on a file descriptor into EV. EV will react to this call as
269if the readyness notifications specified by C<$revents> (a combination of
270C<EV::READ> and C<EV::WRITE>) happened on the file descriptor C<$fd>.
271
272=item EV::feed_signal_event ($signal)
273
274Feed a signal event into EV. EV will react to this call as if the signal
275specified by C<$signal> had occured.
276
277=item EV::set_io_collect_interval $time
278
279=item $loop->set_io_collect_interval ($time)
280
281=item EV::set_timeout_collect_interval $time
282
283=item $loop->set_timeout_collect_interval ($time)
284
285These advanced functions set the minimum block interval when polling for I/O events and the minimum
286wait interval for timer events. See the libev documentation at
287L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#FUNCTIONS_CONTROLLING_THE_EVENT_LOOP> for
288a more detailed discussion.
289
290=back
291
292
293=head1 WATCHER OBJECTS
294
295A watcher is an object that gets created to record your interest in some
296event. For instance, if you want to wait for STDIN to become readable, you
297would create an EV::io watcher for that:
298
299 my $watcher = EV::io *STDIN, EV::READ, sub {
300 my ($watcher, $revents) = @_;
301 warn "yeah, STDIN should now be readable without blocking!\n"
302 };
303
304All watchers can be active (waiting for events) or inactive (paused). Only
305active watchers will have their callbacks invoked. All callbacks will be
306called with at least two arguments: the watcher and a bitmask of received
307events.
308
309Each watcher type has its associated bit in revents, so you can use the
310same callback for multiple watchers. The event mask is named after the
311type, i..e. EV::child sets EV::CHILD, EV::prepare sets EV::PREPARE,
312EV::periodic sets EV::PERIODIC and so on, with the exception of I/O events
313(which can set both EV::READ and EV::WRITE bits), and EV::timer (which
314uses EV::TIMEOUT).
315
316In the rare case where one wants to create a watcher but not start it at
317the same time, each constructor has a variant with a trailing C<_ns> in
318its name, e.g. EV::io has a non-starting variant EV::io_ns and so on.
319
320Please note that a watcher will automatically be stopped when the watcher
321object is destroyed, so you I<need> to keep the watcher objects returned by
322the constructors.
323
324Also, all methods changing some aspect of a watcher (->set, ->priority,
325->fh and so on) automatically stop and start it again if it is active,
326which means pending events get lost.
327
328=head2 COMMON WATCHER METHODS
329
330This section lists methods common to all watchers.
331
332=over 4
333
334=item $w->start
335
336Starts a watcher if it isn't active already. Does nothing to an already
337active watcher. By default, all watchers start out in the active state
338(see the description of the C<_ns> variants if you need stopped watchers).
339
340=item $w->stop
341
342Stop a watcher if it is active. Also clear any pending events (events that
343have been received but that didn't yet result in a callback invocation),
344regardless of whether the watcher was active or not.
345
346=item $bool = $w->is_active
347
348Returns true if the watcher is active, false otherwise.
349
350=item $current_data = $w->data
351
352=item $old_data = $w->data ($new_data)
353
354Queries a freely usable data scalar on the watcher and optionally changes
355it. This is a way to associate custom data with a watcher:
356
357 my $w = EV::timer 60, 0, sub {
358 warn $_[0]->data;
359 };
360 $w->data ("print me!");
361
362=item $current_cb = $w->cb
363
364=item $old_cb = $w->cb ($new_cb)
365
366Queries the callback on the watcher and optionally changes it. You can do
367this at any time without the watcher restarting.
368
369=item $current_priority = $w->priority
370
371=item $old_priority = $w->priority ($new_priority)
372
373Queries the priority on the watcher and optionally changes it. Pending
374watchers with higher priority will be invoked first. The valid range of
375priorities lies between EV::MAXPRI (default 2) and EV::MINPRI (default
376-2). If the priority is outside this range it will automatically be
377normalised to the nearest valid priority.
378
379The default priority of any newly-created watcher is 0.
380
381Note that the priority semantics have not yet been fleshed out and are
382subject to almost certain change.
383
384=item $w->invoke ($revents)
385
386Call the callback *now* with the given event mask.
387
388=item $w->feed_event ($revents)
389
390Feed some events on this watcher into EV. EV will react to this call as if
391the watcher had received the given C<$revents> mask.
392
393=item $revents = $w->clear_pending
394
395If the watcher is pending, this function clears its pending status and
396returns its C<$revents> bitset (as if its callback was invoked). If the
397watcher isn't pending it does nothing and returns C<0>.
398
399=item $previous_state = $w->keepalive ($bool)
400
401Normally, C<EV::loop> will return when there are no active watchers
402(which is a "deadlock" because no progress can be made anymore). This is
403convinient because it allows you to start your watchers (and your jobs),
404call C<EV::loop> once and when it returns you know that all your jobs are
405finished (or they forgot to register some watchers for their task :).
406
407Sometimes, however, this gets in your way, for example when the module
408that calls C<EV::loop> (usually the main program) is not the same module
409as a long-living watcher (for example a DNS client module written by
410somebody else even). Then you might want any outstanding requests to be
411handled, but you would not want to keep C<EV::loop> from returning just
412because you happen to have this long-running UDP port watcher.
413
414In this case you can clear the keepalive status, which means that even
415though your watcher is active, it won't keep C<EV::loop> from returning.
416
417The initial value for keepalive is true (enabled), and you cna change it
418any time.
419
420Example: Register an I/O watcher for some UDP socket but do not keep the
421event loop from running just because of that watcher.
422
423 my $udp_socket = ...
424 my $udp_watcher = EV::io $udp_socket, EV::READ, sub { ... };
425 $udp_watcher->keepalive (0);
426
427=item $loop = $w->loop
428
429Return the loop that this watcher is attached to.
430
431=back
432
433
434=head1 WATCHER TYPES
435
436Each of the following subsections describes a single watcher type.
437
438=head3 I/O WATCHERS - is this file descriptor readable or writable?
439
440=over 4
441
98=item my $w = EV::io $fileno_or_fh, $eventmask, $callback 442=item $w = EV::io $fileno_or_fh, $eventmask, $callback
99 443
100=item my $w = EV::io_ns $fileno_or_fh, $eventmask, $callback 444=item $w = EV::io_ns $fileno_or_fh, $eventmask, $callback
445
446=item $w = $loop->io ($fileno_or_fh, $eventmask, $callback)
447
448=item $w = $loop->io_ns ($fileno_or_fh, $eventmask, $callback)
101 449
102As long as the returned watcher object is alive, call the C<$callback> 450As long as the returned watcher object is alive, call the C<$callback>
103when the events specified in C<$eventmask> happen. Initially, the timeout 451when at least one of events specified in C<$eventmask> occurs.
104is disabled.
105 452
106The C<io_ns> variant doesn't add/start the newly created watcher.
107
108Eventmask can be one or more of these constants ORed together: 453The $eventmask can be one or more of these constants ORed together:
109 454
110 EV::READ wait until read() wouldn't block anymore 455 EV::READ wait until read() wouldn't block anymore
111 EV::WRITE wait until write() wouldn't block anymore 456 EV::WRITE wait until write() wouldn't block anymore
112 EV::PERSIST stay active after an event occured
113 457
114=item my $w = EV::timer $after, $repeat, $callback
115
116=item my $w = EV::timer_ns $after, $repeat, $callback
117
118Calls the callback after C<$after> seconds. If C<$repeat> is true, the
119timer will be restarted after the callback returns. This means that the
120callback would be called roughly every C<$after> seconds, prolonged by the
121time the callback takes.
122
123The C<timer_ns> variant doesn't add/start the newly created watcher. 458The C<io_ns> variant doesn't start (activate) the newly created watcher.
124 459
125=item my $w = EV::timer_abs $at, $interval, $callback 460=item $w->set ($fileno_or_fh, $eventmask)
126 461
127=item my $w = EV::timer_abs_ns $at, $interval, $callback 462Reconfigures the watcher, see the constructor above for details. Can be
128 463called at any time.
129Similar to EV::timer, but the time is given as an absolute point in time
130(C<$at>), plus an optional C<$interval>.
131
132If the C<$interval> is zero, then the callback will be called at the time
133C<$at> if that is in the future, or as soon as possible if its in the
134past. It will not automatically repeat.
135
136If the C<$interval> is nonzero, then the watcher will always be scheduled
137to time out at the next C<$at + integer * $interval> time.
138
139This can be used to schedule a callback to run at very regular intervals,
140as long as the processing time is less then the interval (otherwise
141obviously events will be skipped).
142
143The C<timer_abs_ns> variant doesn't add/start the newly created watcher.
144
145=item my $w = EV::signal $signum, $callback
146
147=item my $w = EV::signal_ns $signum, $callback
148
149Call the callback when signal $signum is received.
150
151The C<signal_ns> variant doesn't add/start the newly created watcher.
152
153=back
154
155=head1 THE EV::Event CLASS
156
157All EV functions creating an event watcher (designated by C<my $w =>
158above) support the following methods on the returned watcher object:
159
160=over 4
161
162=item $w->add ($timeout)
163
164Stops and (re-)starts the event watcher, setting the optional timeout to
165the given value, or clearing the timeout if none is given.
166
167=item $w->start
168
169Stops and (re-)starts the event watcher without touching the timeout.
170
171=item $w->del
172
173=item $w->stop
174
175Stop the event watcher if it was started.
176
177=item $current_callback = $w->cb
178
179=item $old_callback = $w->cb ($new_callback)
180
181Return the previously set callback and optionally set a new one.
182 464
183=item $current_fh = $w->fh 465=item $current_fh = $w->fh
184 466
185=item $old_fh = $w->fh ($new_fh) 467=item $old_fh = $w->fh ($new_fh)
186 468
190 472
191=item $old_eventmask = $w->events ($new_eventmask) 473=item $old_eventmask = $w->events ($new_eventmask)
192 474
193Returns the previously set event mask and optionally set a new one. 475Returns the previously set event mask and optionally set a new one.
194 476
477=back
478
479
480=head3 TIMER WATCHERS - relative and optionally repeating timeouts
481
482=over 4
483
484=item $w = EV::timer $after, $repeat, $callback
485
486=item $w = EV::timer_ns $after, $repeat, $callback
487
488=item $w = $loop->timer ($after, $repeat, $callback)
489
490=item $w = $loop->timer_ns ($after, $repeat, $callback)
491
492Calls the callback after C<$after> seconds (which may be fractional). If
493C<$repeat> is non-zero, the timer will be restarted (with the $repeat
494value as $after) after the callback returns.
495
496This means that the callback would be called roughly after C<$after>
497seconds, and then every C<$repeat> seconds. The timer does his best not
498to drift, but it will not invoke the timer more often then once per event
499loop iteration, and might drift in other cases. If that isn't acceptable,
500look at EV::periodic, which can provide long-term stable timers.
501
502The timer is based on a monotonic clock, that is, if somebody is sitting
503in front of the machine while the timer is running and changes the system
504clock, the timer will nevertheless run (roughly) the same time.
505
506The C<timer_ns> variant doesn't start (activate) the newly created watcher.
507
195=item $w->timeout ($after, $repeat) 508=item $w->set ($after, $repeat)
196 509
197Resets the timeout (see C<EV::timer> for details). 510Reconfigures the watcher, see the constructor above for details. Can be called at
511any time.
198 512
199=item $w->timeout_abs ($at, $interval) 513=item $w->again
200 514
201Resets the timeout (see C<EV::timer_abs> for details). 515Similar to the C<start> method, but has special semantics for repeating timers:
202 516
203=item $w->priority_set ($priority) 517If the timer is active and non-repeating, it will be stopped.
204 518
205Set the priority of the watcher to C<$priority> (0 <= $priority < $EV::NPRI). 519If the timer is active and repeating, reset the timeout to occur
520C<$repeat> seconds after now.
206 521
207=back 522If the timer is inactive and repeating, start it using the repeat value.
208 523
209=head1 BUGS 524Otherwise do nothing.
210 525
211Lots. Libevent itself isn't well tested and rather buggy, and this module 526This behaviour is useful when you have a timeout for some IO
212is quite new at the moment. 527operation. You create a timer object with the same value for C<$after> and
528C<$repeat>, and then, in the read/write watcher, run the C<again> method
529on the timeout.
213 530
214Please note that the epoll method is not, in general, reliable in 531=back
215programs that use fork (even if no libveent calls are being made in the 532
216forked process). Since this is such a common issue, this module will 533
217force the epoll method in EV to be off *unless* the global variable 534=head3 PERIODIC WATCHERS - to cron or not to cron?
218$EV::ENABLE_EPOLL is set to 1 *before* loading this module for the first 535
536=over 4
537
538=item $w = EV::periodic $at, $interval, $reschedule_cb, $callback
539
540=item $w = EV::periodic_ns $at, $interval, $reschedule_cb, $callback
541
542=item $w = $loop->periodic ($at, $interval, $reschedule_cb, $callback)
543
544=item $w = $loop->periodic_ns ($at, $interval, $reschedule_cb, $callback)
545
546Similar to EV::timer, but is not based on relative timeouts but on
547absolute times. Apart from creating "simple" timers that trigger "at" the
548specified time, it can also be used for non-drifting absolute timers and
549more complex, cron-like, setups that are not adversely affected by time
550jumps (i.e. when the system clock is changed by explicit date -s or other
551means such as ntpd). It is also the most complex watcher type in EV.
552
553It has three distinct "modes":
554
555=over 4
556
557=item * absolute timer ($interval = $reschedule_cb = 0)
558
559This time simply fires at the wallclock time C<$at> and doesn't repeat. It
560will not adjust when a time jump occurs, that is, if it is to be run
561at January 1st 2011 then it will run when the system time reaches or
562surpasses this time.
563
564=item * repeating interval timer ($interval > 0, $reschedule_cb = 0)
565
566In this mode the watcher will always be scheduled to time out at the
567next C<$at + N * $interval> time (for some integer N) and then repeat,
568regardless of any time jumps.
569
570This can be used to create timers that do not drift with respect to system
219time. 571time:
572
573 my $hourly = EV::periodic 0, 3600, 0, sub { print "once/hour\n" };
574
575That doesn't mean there will always be 3600 seconds in between triggers,
576but only that the the clalback will be called when the system time shows a
577full hour (UTC).
578
579Another way to think about it (for the mathematically inclined) is that
580EV::periodic will try to run the callback in this mode at the next
581possible time where C<$time = $at (mod $interval)>, regardless of any time
582jumps.
583
584=item * manual reschedule mode ($reschedule_cb = coderef)
585
586In this mode $interval and $at are both being ignored. Instead, each
587time the periodic watcher gets scheduled, the reschedule callback
588($reschedule_cb) will be called with the watcher as first, and the current
589time as second argument.
590
591I<This callback MUST NOT stop or destroy this or any other periodic
592watcher, ever, and MUST NOT call any event loop functions or methods>. If
593you need to stop it, return 1e30 and stop it afterwards. You may create
594and start a C<EV::prepare> watcher for this task.
595
596It must return the next time to trigger, based on the passed time value
597(that is, the lowest time value larger than or equal to to the second
598argument). It will usually be called just before the callback will be
599triggered, but might be called at other times, too.
600
601This can be used to create very complex timers, such as a timer that
602triggers on each midnight, local time (actually 24 hours after the last
603midnight, to keep the example simple. If you know a way to do it correctly
604in about the same space (without requiring elaborate modules), drop me a
605note :):
606
607 my $daily = EV::periodic 0, 0, sub {
608 my ($w, $now) = @_;
609
610 use Time::Local ();
611 my (undef, undef, undef, $d, $m, $y) = localtime $now;
612 86400 + Time::Local::timelocal 0, 0, 0, $d, $m, $y
613 }, sub {
614 print "it's midnight or likely shortly after, now\n";
615 };
616
617=back
618
619The C<periodic_ns> variant doesn't start (activate) the newly created watcher.
620
621=item $w->set ($at, $interval, $reschedule_cb)
622
623Reconfigures the watcher, see the constructor above for details. Can be called at
624any time.
625
626=item $w->again
627
628Simply stops and starts the watcher again.
629
630=item $time = $w->at
631
632Return the time that the watcher is expected to trigger next.
633
634=back
635
636
637=head3 SIGNAL WATCHERS - signal me when a signal gets signalled!
638
639=over 4
640
641=item $w = EV::signal $signal, $callback
642
643=item $w = EV::signal_ns $signal, $callback
644
645Call the callback when $signal is received (the signal can be specified by
646number or by name, just as with C<kill> or C<%SIG>).
647
648EV will grab the signal for the process (the kernel only allows one
649component to receive a signal at a time) when you start a signal watcher,
650and removes it again when you stop it. Perl does the same when you
651add/remove callbacks to C<%SIG>, so watch out.
652
653You can have as many signal watchers per signal as you want.
654
655The C<signal_ns> variant doesn't start (activate) the newly created watcher.
656
657=item $w->set ($signal)
658
659Reconfigures the watcher, see the constructor above for details. Can be
660called at any time.
661
662=item $current_signum = $w->signal
663
664=item $old_signum = $w->signal ($new_signal)
665
666Returns the previously set signal (always as a number not name) and
667optionally set a new one.
668
669=back
670
671
672=head3 CHILD WATCHERS - watch out for process status changes
673
674=over 4
675
676=item $w = EV::child $pid, $trace, $callback
677
678=item $w = EV::child_ns $pid, $trace, $callback
679
680=item $w = $loop->child ($pid, $trace, $callback)
681
682=item $w = $loop->child_ns ($pid, $trace, $callback)
683
684Call the callback when a status change for pid C<$pid> (or any pid
685if C<$pid> is 0) has been received (a status change happens when the
686process terminates or is killed, or, when trace is true, additionally when
687it is stopped or continued). More precisely: when the process receives
688a C<SIGCHLD>, EV will fetch the outstanding exit/wait status for all
689changed/zombie children and call the callback.
690
691It is valid (and fully supported) to install a child watcher after a child
692has exited but before the event loop has started its next iteration (for
693example, first you C<fork>, then the new child process might exit, and
694only then do you install a child watcher in the parent for the new pid).
695
696You can access both exit (or tracing) status and pid by using the
697C<rstatus> and C<rpid> methods on the watcher object.
698
699You can have as many pid watchers per pid as you want, they will all be
700called.
701
702The C<child_ns> variant doesn't start (activate) the newly created watcher.
703
704=item $w->set ($pid, $trace)
705
706Reconfigures the watcher, see the constructor above for details. Can be called at
707any time.
708
709=item $current_pid = $w->pid
710
711Returns the previously set process id and optionally set a new one.
712
713=item $exit_status = $w->rstatus
714
715Return the exit/wait status (as returned by waitpid, see the waitpid entry
716in perlfunc).
717
718=item $pid = $w->rpid
719
720Return the pid of the awaited child (useful when you have installed a
721watcher for all pids).
722
723=back
724
725
726=head3 STAT WATCHERS - did the file attributes just change?
727
728=over 4
729
730=item $w = EV::stat $path, $interval, $callback
731
732=item $w = EV::stat_ns $path, $interval, $callback
733
734=item $w = $loop->stat ($path, $interval, $callback)
735
736=item $w = $loop->stat_ns ($path, $interval, $callback)
737
738Call the callback when a file status change has been detected on
739C<$path>. The C<$path> does not need to exist, changing from "path exists"
740to "path does not exist" is a status change like any other.
741
742The C<$interval> is a recommended polling interval for systems where
743OS-supported change notifications don't exist or are not supported. If
744you use C<0> then an unspecified default is used (which is highly
745recommended!), which is to be expected to be around five seconds usually.
746
747This watcher type is not meant for massive numbers of stat watchers,
748as even with OS-supported change notifications, this can be
749resource-intensive.
750
751The C<stat_ns> variant doesn't start (activate) the newly created watcher.
752
753=item ... = $w->stat
754
755This call is very similar to the perl C<stat> built-in: It stats (using
756C<lstat>) the path specified in the watcher and sets perls stat cache (as
757well as EV's idea of the current stat values) to the values found.
758
759In scalar context, a boolean is return indicating success or failure of
760the stat. In list context, the same 13-value list as with stat is returned
761(except that the blksize and blocks fields are not reliable).
762
763In the case of an error, errno is set to C<ENOENT> (regardless of the
764actual error value) and the C<nlink> value is forced to zero (if the stat
765was successful then nlink is guaranteed to be non-zero).
766
767See also the next two entries for more info.
768
769=item ... = $w->attr
770
771Just like C<< $w->stat >>, but without the initial stat'ing: this returns
772the values most recently detected by EV. See the next entry for more info.
773
774=item ... = $w->prev
775
776Just like C<< $w->stat >>, but without the initial stat'ing: this returns
777the previous set of values, before the change.
778
779That is, when the watcher callback is invoked, C<< $w->prev >> will be set
780to the values found I<before> a change was detected, while C<< $w->attr >>
781returns the values found leading to the change detection. The difference (if any)
782between C<prev> and C<attr> is what triggered the callback.
783
784If you did something to the filesystem object and do not want to trigger
785yet another change, you can call C<stat> to update EV's idea of what the
786current attributes are.
787
788=item $w->set ($path, $interval)
789
790Reconfigures the watcher, see the constructor above for details. Can be
791called at any time.
792
793=item $current_path = $w->path
794
795=item $old_path = $w->path ($new_path)
796
797Returns the previously set path and optionally set a new one.
798
799=item $current_interval = $w->interval
800
801=item $old_interval = $w->interval ($new_interval)
802
803Returns the previously set interval and optionally set a new one. Can be
804used to query the actual interval used.
805
806=back
807
808
809=head3 IDLE WATCHERS - when you've got nothing better to do...
810
811=over 4
812
813=item $w = EV::idle $callback
814
815=item $w = EV::idle_ns $callback
816
817=item $w = $loop->idle ($callback)
818
819=item $w = $loop->idle_ns ($callback)
820
821Call the callback when there are no other pending watchers of the same or
822higher priority (excluding check, prepare and other idle watchers of the
823same or lower priority, of course). They are called idle watchers because
824when the watcher is the highest priority pending event in the process, the
825process is considered to be idle at that priority.
826
827If you want a watcher that is only ever called when I<no> other events are
828outstanding you have to set the priority to C<EV::MINPRI>.
829
830The process will not block as long as any idle watchers are active, and
831they will be called repeatedly until stopped.
832
833For example, if you have idle watchers at priority C<0> and C<1>, and
834an I/O watcher at priority C<0>, then the idle watcher at priority C<1>
835and the I/O watcher will always run when ready. Only when the idle watcher
836at priority C<1> is stopped and the I/O watcher at priority C<0> is not
837pending with the C<0>-priority idle watcher be invoked.
838
839The C<idle_ns> variant doesn't start (activate) the newly created watcher.
840
841=back
842
843
844=head3 PREPARE WATCHERS - customise your event loop!
845
846=over 4
847
848=item $w = EV::prepare $callback
849
850=item $w = EV::prepare_ns $callback
851
852=item $w = $loop->prepare ($callback)
853
854=item $w = $loop->prepare_ns ($callback)
855
856Call the callback just before the process would block. You can still
857create/modify any watchers at this point.
858
859See the EV::check watcher, below, for explanations and an example.
860
861The C<prepare_ns> variant doesn't start (activate) the newly created watcher.
862
863=back
864
865
866=head3 CHECK WATCHERS - customise your event loop even more!
867
868=over 4
869
870=item $w = EV::check $callback
871
872=item $w = EV::check_ns $callback
873
874=item $w = $loop->check ($callback)
875
876=item $w = $loop->check_ns ($callback)
877
878Call the callback just after the process wakes up again (after it has
879gathered events), but before any other callbacks have been invoked.
880
881This is used to integrate other event-based software into the EV
882mainloop: You register a prepare callback and in there, you create io and
883timer watchers as required by the other software. Here is a real-world
884example of integrating Net::SNMP (with some details left out):
885
886 our @snmp_watcher;
887
888 our $snmp_prepare = EV::prepare sub {
889 # do nothing unless active
890 $dispatcher->{_event_queue_h}
891 or return;
892
893 # make the dispatcher handle any outstanding stuff
894 ... not shown
895
896 # create an I/O watcher for each and every socket
897 @snmp_watcher = (
898 (map { EV::io $_, EV::READ, sub { } }
899 keys %{ $dispatcher->{_descriptors} }),
900
901 EV::timer +($event->[Net::SNMP::Dispatcher::_ACTIVE]
902 ? $event->[Net::SNMP::Dispatcher::_TIME] - EV::now : 0),
903 0, sub { },
904 );
905 };
906
907The callbacks are irrelevant (and are not even being called), the
908only purpose of those watchers is to wake up the process as soon as
909one of those events occurs (socket readable, or timer timed out). The
910corresponding EV::check watcher will then clean up:
911
912 our $snmp_check = EV::check sub {
913 # destroy all watchers
914 @snmp_watcher = ();
915
916 # make the dispatcher handle any new stuff
917 ... not shown
918 };
919
920The callbacks of the created watchers will not be called as the watchers
921are destroyed before this cna happen (remember EV::check gets called
922first).
923
924The C<check_ns> variant doesn't start (activate) the newly created watcher.
925
926=back
927
928
929=head3 FORK WATCHERS - the audacity to resume the event loop after a fork
930
931Fork watchers are called when a C<fork ()> was detected. The invocation
932is done before the event loop blocks next and before C<check> watchers
933are being called, and only in the child after the fork.
934
935=over 4
936
937=item $w = EV::fork $callback
938
939=item $w = EV::fork_ns $callback
940
941=item $w = $loop->fork ($callback)
942
943=item $w = $loop->fork_ns ($callback)
944
945Call the callback before the event loop is resumed in the child process
946after a fork.
947
948The C<fork_ns> variant doesn't start (activate) the newly created watcher.
949
950=back
951
952
953=head3 EMBED WATCHERS - when one backend isn't enough...
954
955This is a rather advanced watcher type that lets you embed one event loop
956into another (currently only IO events are supported in the embedded
957loop, other types of watchers might be handled in a delayed or incorrect
958fashion and must not be used).
959
960See the libev documentation at
961L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod#code_ev_embed_code_when_one_backend_>
962for more details.
963
964In short, this watcher is most useful on BSD systems without working
965kqueue to still be able to handle a large number of sockets:
966
967 my $socket_loop;
968
969 # check wether we use SELECT or POLL _and_ KQUEUE is supported
970 if (
971 (EV::backend & (EV::BACKEND_POLL | EV::BACKEND_SELECT))
972 && (EV::supported_backends & EV::embeddable_backends & EV::BACKEND_KQUEUE)
973 ) {
974 # use kqueue for sockets
975 $socket_loop = new EV::Loop EV::BACKEND_KQUEUE | EV::FLAG_NOENV;
976 }
977
978 # use the default loop otherwise
979 $socket_loop ||= EV::default_loop;
980
981=over 4
982
983=item $w = EV::embed $otherloop[, $callback]
984
985=item $w = EV::embed_ns $otherloop[, $callback]
986
987=item $w = $loop->embed ($otherloop[, $callback])
988
989=item $w = $loop->embed_ns ($otherloop[, $callback])
990
991Call the callback when the embedded event loop (C<$otherloop>) has any
992I/O activity. The C<$callback> is optional: if it is missing, then the
993embedded event loop will be managed automatically (which is recommended),
994otherwise you have to invoke C<sweep> yourself.
995
996The C<embed_ns> variant doesn't start (activate) the newly created watcher.
997
998=back
999
1000=head3 ASYNC WATCHERS - how to wake up another event loop
1001
1002Async watchers are provided by EV, but have little use in perl directly, as perl
1003neither supports threads nor direct access to signal handlers or other
1004contexts where they could be of value.
1005
1006It is, however, possible to use them from the XS level.
1007
1008Please see the libev documentation for further details.
1009
1010=over 4
1011
1012=item $w = EV::async $callback
1013
1014=item $w = EV::async_ns $callback
1015
1016=item $w->send
1017
1018=item $bool = $w->async_pending
1019
1020=back
1021
1022
1023=head1 PERL SIGNALS
1024
1025While Perl signal handling (C<%SIG>) is not affected by EV, the behaviour
1026with EV is as the same as any other C library: Perl-signals will only be
1027handled when Perl runs, which means your signal handler might be invoked
1028only the next time an event callback is invoked.
1029
1030The solution is to use EV signal watchers (see C<EV::signal>), which will
1031ensure proper operations with regards to other event watchers.
1032
1033If you cannot do this for whatever reason, you can also force a watcher
1034to be called on every event loop iteration by installing a C<EV::check>
1035watcher:
1036
1037 my $async_check = EV::check sub { };
1038
1039This ensures that perl gets into control for a short time to handle any
1040pending signals, and also ensures (slightly) slower overall operation.
1041
1042=head1 THREADS
1043
1044Threads are not supported by this module in any way. Perl pseudo-threads
1045is evil stuff and must die. As soon as Perl gains real threads I will work
1046on thread support for it.
1047
1048=head1 FORK
1049
1050Most of the "improved" event delivering mechanisms of modern operating
1051systems have quite a few problems with fork(2) (to put it bluntly: it is
1052not supported and usually destructive). Libev makes it possible to work
1053around this by having a function that recreates the kernel state after
1054fork in the child.
1055
1056On non-win32 platforms, this module requires the pthread_atfork
1057functionality to do this automatically for you. This function is quite
1058buggy on most BSDs, though, so YMMV. The overhead for this is quite
1059negligible, because everything the function currently does is set a flag
1060that is checked only when the event loop gets used the next time, so when
1061you do fork but not use EV, the overhead is minimal.
1062
1063On win32, there is no notion of fork so all this doesn't apply, of course.
220 1064
221=cut 1065=cut
222 1066
223our $NPRI = 4; 1067our $DIED = sub {
224our $BASE; 1068 warn "EV: error in callback (ignoring): $@";
225our $ENABLE_EPOLL; 1069};
226 1070
227{ 1071default_loop
228 local $ENV{EVENT_NOEPOLL}; 1072 or die 'EV: cannot initialise libev backend. bad $ENV{LIBEV_FLAGS}?';
229 $ENV{EVENT_NOEPOLL} = 1 unless $ENABLE_EPOLL;
230 $BASE = init;
231}
232
233priority_init $NPRI;
234
235push @AnyEvent::REGISTRY, [EV => "EV::AnyEvent"];
236 1073
2371; 10741;
238 1075
239=head1 SEE ALSO 1076=head1 SEE ALSO
240 1077
241 L<EV::DNS>, L<event(3)>, L<event.h>, L<evdns.h>. 1078L<EV::ADNS> (asynchronous DNS), L<Glib::EV> (makes Glib/Gtk2 use EV as
242 L<EV::AnyEvent>. 1079event loop), L<EV::Glib> (embed Glib into EV), L<Coro::EV> (efficient
1080coroutines with EV), L<Net::SNMP::EV> (asynchronous SNMP), L<AnyEvent> for
1081event-loop agnostic and portable event driven programming.
243 1082
244=head1 AUTHOR 1083=head1 AUTHOR
245 1084
246 Marc Lehmann <schmorp@schmorp.de> 1085 Marc Lehmann <schmorp@schmorp.de>
247 http://home.schmorp.de/ 1086 http://home.schmorp.de/
248 1087
249=cut 1088=cut
250 1089

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines