1 |
NAME |
2 |
EV - perl interface to libev, a high performance full-featured event |
3 |
loop |
4 |
|
5 |
SYNOPSIS |
6 |
use EV; |
7 |
|
8 |
# TIMERS |
9 |
|
10 |
my $w = EV::timer 2, 0, sub { |
11 |
warn "is called after 2s"; |
12 |
}; |
13 |
|
14 |
my $w = EV::timer 2, 2, sub { |
15 |
warn "is called roughly every 2s (repeat = 2)"; |
16 |
}; |
17 |
|
18 |
undef $w; # destroy event watcher again |
19 |
|
20 |
my $w = EV::periodic 0, 60, 0, sub { |
21 |
warn "is called every minute, on the minute, exactly"; |
22 |
}; |
23 |
|
24 |
# IO |
25 |
|
26 |
my $w = EV::io *STDIN, EV::READ, sub { |
27 |
my ($w, $revents) = @_; # all callbacks receive the watcher and event mask |
28 |
warn "stdin is readable, you entered: ", <STDIN>; |
29 |
}; |
30 |
|
31 |
# SIGNALS |
32 |
|
33 |
my $w = EV::signal 'QUIT', sub { |
34 |
warn "sigquit received\n"; |
35 |
}; |
36 |
|
37 |
# CHILD/PID STATUS CHANGES |
38 |
|
39 |
my $w = EV::child 666, sub { |
40 |
my ($w, $revents) = @_; |
41 |
my $status = $w->rstatus; |
42 |
}; |
43 |
|
44 |
# MAINLOOP |
45 |
EV::loop; # loop until EV::unloop is called or all watchers stop |
46 |
EV::loop EV::LOOP_ONESHOT; # block until at least one event could be handled |
47 |
EV::loop EV::LOOP_NONBLOCK; # try to handle same events, but do not block |
48 |
|
49 |
DESCRIPTION |
50 |
This module provides an interface to libev |
51 |
(<http://software.schmorp.de/pkg/libev.html>). While the documentation |
52 |
below is comprehensive, one might also consult the documentation of |
53 |
libev itself (<http://cvs.schmorp.de/libev/ev.html>) for more subtle |
54 |
details on watcher semantics or some discussion on the available |
55 |
backends, or how to force a specific backend with "LIBEV_FLAGS". |
56 |
|
57 |
BASIC INTERFACE |
58 |
$EV::DIED |
59 |
Must contain a reference to a function that is called when a |
60 |
callback throws an exception (with $@ containing thr error). The |
61 |
default prints an informative message and continues. |
62 |
|
63 |
If this callback throws an exception it will be silently ignored. |
64 |
|
65 |
$time = EV::time |
66 |
Returns the current time in (fractional) seconds since the epoch. |
67 |
|
68 |
$time = EV::now |
69 |
Returns the time the last event loop iteration has been started. |
70 |
This is the time that (relative) timers are based on, and refering |
71 |
to it is usually faster then calling EV::time. |
72 |
|
73 |
$method = EV::method |
74 |
Returns an integer describing the backend used by libev |
75 |
(EV::METHOD_SELECT or EV::METHOD_EPOLL). |
76 |
|
77 |
EV::loop [$flags] |
78 |
Begin checking for events and calling callbacks. It returns when a |
79 |
callback calls EV::unloop. |
80 |
|
81 |
The $flags argument can be one of the following: |
82 |
|
83 |
0 as above |
84 |
EV::LOOP_ONESHOT block at most once (wait, but do not loop) |
85 |
EV::LOOP_NONBLOCK do not block at all (fetch/handle events but do not wait) |
86 |
|
87 |
EV::unloop [$how] |
88 |
When called with no arguments or an argument of EV::UNLOOP_ONE, |
89 |
makes the innermost call to EV::loop return. |
90 |
|
91 |
When called with an argument of EV::UNLOOP_ALL, all calls to |
92 |
EV::loop will return as fast as possible. |
93 |
|
94 |
EV::once $fh_or_undef, $events, $timeout, $cb->($revents) |
95 |
This function rolls together an I/O and a timer watcher for a single |
96 |
one-shot event without the need for managing a watcher object. |
97 |
|
98 |
If $fh_or_undef is a filehandle or file descriptor, then $events |
99 |
must be a bitset containing either "EV::READ", "EV::WRITE" or |
100 |
"EV::READ | EV::WRITE", indicating the type of I/O event you want to |
101 |
wait for. If you do not want to wait for some I/O event, specify |
102 |
"undef" for $fh_or_undef and 0 for $events). |
103 |
|
104 |
If timeout is "undef" or negative, then there will be no timeout. |
105 |
Otherwise a EV::timer with this value will be started. |
106 |
|
107 |
When an error occurs or either the timeout or I/O watcher triggers, |
108 |
then the callback will be called with the received event set (in |
109 |
general you can expect it to be a combination of "EV:ERROR", |
110 |
"EV::READ", "EV::WRITE" and "EV::TIMEOUT"). |
111 |
|
112 |
EV::once doesn't return anything: the watchers stay active till |
113 |
either of them triggers, then they will be stopped and freed, and |
114 |
the callback invoked. |
115 |
|
116 |
WATCHER |
117 |
A watcher is an object that gets created to record your interest in some |
118 |
event. For instance, if you want to wait for STDIN to become readable, |
119 |
you would create an EV::io watcher for that: |
120 |
|
121 |
my $watcher = EV::io *STDIN, EV::READ, sub { |
122 |
my ($watcher, $revents) = @_; |
123 |
warn "yeah, STDIN should not be readable without blocking!\n" |
124 |
}; |
125 |
|
126 |
All watchers can be active (waiting for events) or inactive (paused). |
127 |
Only active watchers will have their callbacks invoked. All callbacks |
128 |
will be called with at least two arguments: the watcher and a bitmask of |
129 |
received events. |
130 |
|
131 |
Each watcher type has its associated bit in revents, so you can use the |
132 |
same callback for multiple watchers. The event mask is named after the |
133 |
type, i..e. EV::child sets EV::CHILD, EV::prepare sets EV::PREPARE, |
134 |
EV::periodic sets EV::PERIODIC and so on, with the exception of IO |
135 |
events (which can set both EV::READ and EV::WRITE bits), and EV::timer |
136 |
(which uses EV::TIMEOUT). |
137 |
|
138 |
In the rare case where one wants to create a watcher but not start it at |
139 |
the same time, each constructor has a variant with a trailing "_ns" in |
140 |
its name, e.g. EV::io has a non-starting variant EV::io_ns and so on. |
141 |
|
142 |
Please note that a watcher will automatically be stopped when the |
143 |
watcher object is destroyed, so you *need* to keep the watcher objects |
144 |
returned by the constructors. |
145 |
|
146 |
Also, all methods changing some aspect of a watcher (->set, ->priority, |
147 |
->fh and so on) automatically stop and start it again if it is active, |
148 |
which means pending events get lost. |
149 |
|
150 |
WATCHER TYPES |
151 |
Now lets move to the existing watcher types and asociated methods. |
152 |
|
153 |
The following methods are available for all watchers. Then followes a |
154 |
description of each watcher constructor (EV::io, EV::timer, |
155 |
EV::periodic, EV::signal, EV::child, EV::idle, EV::prepare and |
156 |
EV::check), followed by any type-specific methods (if any). |
157 |
|
158 |
$w->start |
159 |
Starts a watcher if it isn't active already. Does nothing to an |
160 |
already active watcher. By default, all watchers start out in the |
161 |
active state (see the description of the "_ns" variants if you need |
162 |
stopped watchers). |
163 |
|
164 |
$w->stop |
165 |
Stop a watcher if it is active. Also clear any pending events |
166 |
(events that have been received but that didn't yet result in a |
167 |
callback invocation), regardless of wether the watcher was active or |
168 |
not. |
169 |
|
170 |
$bool = $w->is_active |
171 |
Returns true if the watcher is active, false otherwise. |
172 |
|
173 |
$current_data = $w->data |
174 |
$old_data = $w->data ($new_data) |
175 |
Queries a freely usable data scalar on the watcher and optionally |
176 |
changes it. This is a way to associate custom data with a watcher: |
177 |
|
178 |
my $w = EV::timer 60, 0, sub { |
179 |
warn $_[0]->data; |
180 |
}; |
181 |
$w->data ("print me!"); |
182 |
|
183 |
$current_cb = $w->cb |
184 |
$old_cb = $w->cb ($new_cb) |
185 |
Queries the callback on the watcher and optionally changes it. You |
186 |
can do this at any time without the watcher restarting. |
187 |
|
188 |
$current_priority = $w->priority |
189 |
$old_priority = $w->priority ($new_priority) |
190 |
Queries the priority on the watcher and optionally changes it. |
191 |
Pending watchers with higher priority will be invoked first. The |
192 |
valid range of priorities lies between EV::MAXPRI (default 2) and |
193 |
EV::MINPRI (default -2). If the priority is outside this range it |
194 |
will automatically be normalised to the nearest valid priority. |
195 |
|
196 |
The default priority of any newly-created watcher is 0. |
197 |
|
198 |
Note that the priority semantics have not yet been fleshed out and |
199 |
are subject to almost certain change. |
200 |
|
201 |
$w->trigger ($revents) |
202 |
Call the callback *now* with the given event mask. |
203 |
|
204 |
$previous_state = $w->keepalive ($bool) |
205 |
Normally, "EV::loop" will return when there are no active watchers |
206 |
(which is a "deadlock" because no progress can be made anymore). |
207 |
This is convinient because it allows you to start your watchers (and |
208 |
your jobs), call "EV::loop" once and when it returns you know that |
209 |
all your jobs are finished (or they forgot to register some watchers |
210 |
for their task :). |
211 |
|
212 |
Sometimes, however, this gets in your way, for example when you the |
213 |
module that calls "EV::loop" (usually the main program) is not the |
214 |
same module as a long-living watcher (for example a DNS client |
215 |
module written by somebody else even). Then you might want any |
216 |
outstanding requests to be handled, but you would not want to keep |
217 |
"EV::loop" from returning just because you happen to have this |
218 |
long-running UDP port watcher. |
219 |
|
220 |
In this case you can clear the keepalive status, which means that |
221 |
even though your watcher is active, it won't keep "EV::loop" from |
222 |
returning. |
223 |
|
224 |
The initial value for keepalive is true (enabled), and you cna |
225 |
change it any time. |
226 |
|
227 |
Example: Register an IO watcher for some UDP socket but do not keep |
228 |
the event loop from running just because of that watcher. |
229 |
|
230 |
my $udp_socket = ... |
231 |
my $udp_watcher = EV::io $udp_socket, EV::READ, sub { ... }; |
232 |
$udp_watcher->keepalive (0); |
233 |
|
234 |
$w = EV::io $fileno_or_fh, $eventmask, $callback |
235 |
$w = EV::io_ns $fileno_or_fh, $eventmask, $callback |
236 |
As long as the returned watcher object is alive, call the $callback |
237 |
when the events specified in $eventmask. |
238 |
|
239 |
The $eventmask can be one or more of these constants ORed together: |
240 |
|
241 |
EV::READ wait until read() wouldn't block anymore |
242 |
EV::WRITE wait until write() wouldn't block anymore |
243 |
|
244 |
The "io_ns" variant doesn't start (activate) the newly created |
245 |
watcher. |
246 |
|
247 |
$w->set ($fileno_or_fh, $eventmask) |
248 |
Reconfigures the watcher, see the constructor above for details. Can |
249 |
be called at any time. |
250 |
|
251 |
$current_fh = $w->fh |
252 |
$old_fh = $w->fh ($new_fh) |
253 |
Returns the previously set filehandle and optionally set a new one. |
254 |
|
255 |
$current_eventmask = $w->events |
256 |
$old_eventmask = $w->events ($new_eventmask) |
257 |
Returns the previously set event mask and optionally set a new one. |
258 |
|
259 |
$w = EV::timer $after, $repeat, $callback |
260 |
$w = EV::timer_ns $after, $repeat, $callback |
261 |
Calls the callback after $after seconds (which may be fractional). |
262 |
If $repeat is non-zero, the timer will be restarted (with the |
263 |
$repeat value as $after) after the callback returns. |
264 |
|
265 |
This means that the callback would be called roughly after $after |
266 |
seconds, and then every $repeat seconds. The timer does his best not |
267 |
to drift, but it will not invoke the timer more often then once per |
268 |
event loop iteration, and might drift in other cases. If that isn't |
269 |
acceptable, look at EV::periodic, which can provide long-term stable |
270 |
timers. |
271 |
|
272 |
The timer is based on a monotonic clock, that is, if somebody is |
273 |
sitting in front of the machine while the timer is running and |
274 |
changes the system clock, the timer will nevertheless run (roughly) |
275 |
the same time. |
276 |
|
277 |
The "timer_ns" variant doesn't start (activate) the newly created |
278 |
watcher. |
279 |
|
280 |
$w->set ($after, $repeat) |
281 |
Reconfigures the watcher, see the constructor above for details. Can |
282 |
be at any time. |
283 |
|
284 |
$w->again |
285 |
Similar to the "start" method, but has special semantics for |
286 |
repeating timers: |
287 |
|
288 |
If the timer is active and non-repeating, it will be stopped. |
289 |
|
290 |
If the timer is active and repeating, reset the timeout to occur |
291 |
$repeat seconds after now. |
292 |
|
293 |
If the timer is inactive and repeating, start it using the repeat |
294 |
value. |
295 |
|
296 |
Otherwise do nothing. |
297 |
|
298 |
This behaviour is useful when you have a timeout for some IO |
299 |
operation. You create a timer object with the same value for $after |
300 |
and $repeat, and then, in the read/write watcher, run the "again" |
301 |
method on the timeout. |
302 |
|
303 |
$w = EV::periodic $at, $interval, $reschedule_cb, $callback |
304 |
$w = EV::periodic_ns $at, $interval, $reschedule_cb, $callback |
305 |
Similar to EV::timer, but is not based on relative timeouts but on |
306 |
absolute times. Apart from creating "simple" timers that trigger |
307 |
"at" the specified time, it can also be used for non-drifting |
308 |
absolute timers and more complex, cron-like, setups that are not |
309 |
adversely affected by time jumps (i.e. when the system clock is |
310 |
changed by explicit date -s or other means such as ntpd). It is also |
311 |
the most complex watcher type in EV. |
312 |
|
313 |
It has three distinct "modes": |
314 |
|
315 |
* absolute timer ($interval = $reschedule_cb = 0) |
316 |
This time simply fires at the wallclock time $at and doesn't |
317 |
repeat. It will not adjust when a time jump occurs, that is, if |
318 |
it is to be run at January 1st 2011 then it will run when the |
319 |
system time reaches or surpasses this time. |
320 |
|
321 |
* non-repeating interval timer ($interval > 0, $reschedule_cb = 0) |
322 |
In this mode the watcher will always be scheduled to time out at |
323 |
the next "$at + N * $interval" time (for some integer N) and |
324 |
then repeat, regardless of any time jumps. |
325 |
|
326 |
This can be used to create timers that do not drift with respect |
327 |
to system time: |
328 |
|
329 |
my $hourly = EV::periodic 0, 3600, 0, sub { print "once/hour\n" }; |
330 |
|
331 |
That doesn't mean there will always be 3600 seconds in between |
332 |
triggers, but only that the the clalback will be called when the |
333 |
system time shows a full hour (UTC). |
334 |
|
335 |
Another way to think about it (for the mathematically inclined) |
336 |
is that EV::periodic will try to run the callback in this mode |
337 |
at the next possible time where "$time = $at (mod $interval)", |
338 |
regardless of any time jumps. |
339 |
|
340 |
* manual reschedule mode ($reschedule_cb = coderef) |
341 |
In this mode $interval and $at are both being ignored. Instead, |
342 |
each time the periodic watcher gets scheduled, the reschedule |
343 |
callback ($reschedule_cb) will be called with the watcher as |
344 |
first, and the current time as second argument. |
345 |
|
346 |
*This callback MUST NOT stop or destroy this or any other |
347 |
periodic watcher, ever*. If you need to stop it, return 1e30 and |
348 |
stop it afterwards. |
349 |
|
350 |
It must return the next time to trigger, based on the passed |
351 |
time value (that is, the lowest time value larger than to the |
352 |
second argument). It will usually be called just before the |
353 |
callback will be triggered, but might be called at other times, |
354 |
too. |
355 |
|
356 |
This can be used to create very complex timers, such as a timer |
357 |
that triggers on each midnight, local time (actually 24 hours |
358 |
after the last midnight, to keep the example simple. If you know |
359 |
a way to do it correctly in about the same space (without |
360 |
requiring elaborate modules), drop me a note :): |
361 |
|
362 |
my $daily = EV::periodic 0, 0, sub { |
363 |
my ($w, $now) = @_; |
364 |
|
365 |
use Time::Local (); |
366 |
my (undef, undef, undef, $d, $m, $y) = localtime $now; |
367 |
86400 + Time::Local::timelocal 0, 0, 0, $d, $m, $y |
368 |
}, sub { |
369 |
print "it's midnight or likely shortly after, now\n"; |
370 |
}; |
371 |
|
372 |
The "periodic_ns" variant doesn't start (activate) the newly created |
373 |
watcher. |
374 |
|
375 |
$w->set ($at, $interval, $reschedule_cb) |
376 |
Reconfigures the watcher, see the constructor above for details. Can |
377 |
be at any time. |
378 |
|
379 |
$w->again |
380 |
Simply stops and starts the watcher again. |
381 |
|
382 |
$w = EV::signal $signal, $callback |
383 |
$w = EV::signal_ns $signal, $callback |
384 |
Call the callback when $signal is received (the signal can be |
385 |
specified by number or by name, just as with kill or %SIG). |
386 |
|
387 |
EV will grab the signal for the process (the kernel only allows one |
388 |
component to receive a signal at a time) when you start a signal |
389 |
watcher, and removes it again when you stop it. Perl does the same |
390 |
when you add/remove callbacks to %SIG, so watch out. |
391 |
|
392 |
You can have as many signal watchers per signal as you want. |
393 |
|
394 |
The "signal_ns" variant doesn't start (activate) the newly created |
395 |
watcher. |
396 |
|
397 |
$w->set ($signal) |
398 |
Reconfigures the watcher, see the constructor above for details. Can |
399 |
be at any time. |
400 |
|
401 |
$current_signum = $w->signal |
402 |
$old_signum = $w->signal ($new_signal) |
403 |
Returns the previously set signal (always as a number not name) and |
404 |
optionally set a new one. |
405 |
|
406 |
$w = EV::child $pid, $callback |
407 |
$w = EV::child_ns $pid, $callback |
408 |
Call the callback when a status change for pid $pid (or any pid if |
409 |
$pid is 0) has been received. More precisely: when the process |
410 |
receives a SIGCHLD, EV will fetch the outstanding exit/wait status |
411 |
for all changed/zombie children and call the callback. |
412 |
|
413 |
You can access both status and pid by using the "rstatus" and "rpid" |
414 |
methods on the watcher object. |
415 |
|
416 |
You can have as many pid watchers per pid as you want. |
417 |
|
418 |
The "child_ns" variant doesn't start (activate) the newly created |
419 |
watcher. |
420 |
|
421 |
$w->set ($pid) |
422 |
Reconfigures the watcher, see the constructor above for details. Can |
423 |
be at any time. |
424 |
|
425 |
$current_pid = $w->pid |
426 |
$old_pid = $w->pid ($new_pid) |
427 |
Returns the previously set process id and optionally set a new one. |
428 |
|
429 |
$exit_status = $w->rstatus |
430 |
Return the exit/wait status (as returned by waitpid, see the waitpid |
431 |
entry in perlfunc). |
432 |
|
433 |
$pid = $w->rpid |
434 |
Return the pid of the awaited child (useful when you have installed |
435 |
a watcher for all pids). |
436 |
|
437 |
$w = EV::idle $callback |
438 |
$w = EV::idle_ns $callback |
439 |
Call the callback when there are no pending io, timer/periodic, |
440 |
signal or child events, i.e. when the process is idle. |
441 |
|
442 |
The process will not block as long as any idle watchers are active, |
443 |
and they will be called repeatedly until stopped. |
444 |
|
445 |
The "idle_ns" variant doesn't start (activate) the newly created |
446 |
watcher. |
447 |
|
448 |
$w = EV::prepare $callback |
449 |
$w = EV::prepare_ns $callback |
450 |
Call the callback just before the process would block. You can still |
451 |
create/modify any watchers at this point. |
452 |
|
453 |
See the EV::check watcher, below, for explanations and an example. |
454 |
|
455 |
The "prepare_ns" variant doesn't start (activate) the newly created |
456 |
watcher. |
457 |
|
458 |
$w = EV::check $callback |
459 |
$w = EV::check_ns $callback |
460 |
Call the callback just after the process wakes up again (after it |
461 |
has gathered events), but before any other callbacks have been |
462 |
invoked. |
463 |
|
464 |
This is used to integrate other event-based software into the EV |
465 |
mainloop: You register a prepare callback and in there, you create |
466 |
io and timer watchers as required by the other software. Here is a |
467 |
real-world example of integrating Net::SNMP (with some details left |
468 |
out): |
469 |
|
470 |
our @snmp_watcher; |
471 |
|
472 |
our $snmp_prepare = EV::prepare sub { |
473 |
# do nothing unless active |
474 |
$dispatcher->{_event_queue_h} |
475 |
or return; |
476 |
|
477 |
# make the dispatcher handle any outstanding stuff |
478 |
... not shown |
479 |
|
480 |
# create an IO watcher for each and every socket |
481 |
@snmp_watcher = ( |
482 |
(map { EV::io $_, EV::READ, sub { } } |
483 |
keys %{ $dispatcher->{_descriptors} }), |
484 |
|
485 |
EV::timer +($event->[Net::SNMP::Dispatcher::_ACTIVE] |
486 |
? $event->[Net::SNMP::Dispatcher::_TIME] - EV::now : 0), |
487 |
0, sub { }, |
488 |
); |
489 |
}; |
490 |
|
491 |
The callbacks are irrelevant (and are not even being called), the |
492 |
only purpose of those watchers is to wake up the process as soon as |
493 |
one of those events occurs (socket readable, or timer timed out). |
494 |
The corresponding EV::check watcher will then clean up: |
495 |
|
496 |
our $snmp_check = EV::check sub { |
497 |
# destroy all watchers |
498 |
@snmp_watcher = (); |
499 |
|
500 |
# make the dispatcher handle any new stuff |
501 |
... not shown |
502 |
}; |
503 |
|
504 |
The callbacks of the created watchers will not be called as the |
505 |
watchers are destroyed before this cna happen (remember EV::check |
506 |
gets called first). |
507 |
|
508 |
The "check_ns" variant doesn't start (activate) the newly created |
509 |
watcher. |
510 |
|
511 |
THREADS |
512 |
Threads are not supported by this module in any way. Perl pseudo-threads |
513 |
is evil stuff and must die. As soon as Perl gains real threads I will |
514 |
work on thread support for it. |
515 |
|
516 |
FORK |
517 |
Most of the "improved" event delivering mechanisms of modern operating |
518 |
systems have quite a few problems with fork(2) (to put it bluntly: it is |
519 |
not supported and usually destructive). Libev makes it possible to work |
520 |
around this by having a function that recreates the kernel state after |
521 |
fork in the child. |
522 |
|
523 |
On non-win32 platforms, this module requires the pthread_atfork |
524 |
functionality to do this automatically for you. This function is quite |
525 |
buggy on most BSDs, though, so YMMV. The overhead for this is quite |
526 |
negligible, because everything the function currently does is set a flag |
527 |
that is checked only when the event loop gets used the next time, so |
528 |
when you do fork but not use EV, the overhead is minimal. |
529 |
|
530 |
On win32, there is no notion of fork so all this doesn't apply, of |
531 |
course. |
532 |
|
533 |
SEE ALSO |
534 |
L<EV::DNS>. |
535 |
|
536 |
AUTHOR |
537 |
Marc Lehmann <schmorp@schmorp.de> |
538 |
http://home.schmorp.de/ |
539 |
|