ViewVC Help
View File | Revision Log | Show Annotations | Download File
Revision: 1.146
Committed: Fri Apr 11 00:31:19 2008 UTC (16 years, 3 months ago) by root
Branch: MAIN
CVS Tags: rel-3_31, rel-3_3
Changes since 1.145: +3 -0 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 =head1 NAME
3 libev - a high performance full-featured event loop written in C
5 =head1 SYNOPSIS
7 #include <ev.h>
11 // a single header file is required
12 #include <ev.h>
14 // every watcher type has its own typedef'd struct
15 // with the name ev_<type>
16 ev_io stdin_watcher;
17 ev_timer timeout_watcher;
19 // all watcher callbacks have a similar signature
20 // this callback is called when data is readable on stdin
21 static void
22 stdin_cb (EV_P_ struct ev_io *w, int revents)
23 {
24 puts ("stdin ready");
25 // for one-shot events, one must manually stop the watcher
26 // with its corresponding stop function.
27 ev_io_stop (EV_A_ w);
29 // this causes all nested ev_loop's to stop iterating
30 ev_unloop (EV_A_ EVUNLOOP_ALL);
31 }
33 // another callback, this time for a time-out
34 static void
35 timeout_cb (EV_P_ struct ev_timer *w, int revents)
36 {
37 puts ("timeout");
38 // this causes the innermost ev_loop to stop iterating
39 ev_unloop (EV_A_ EVUNLOOP_ONE);
40 }
42 int
43 main (void)
44 {
45 // use the default event loop unless you have special needs
46 struct ev_loop *loop = ev_default_loop (0);
48 // initialise an io watcher, then start it
49 // this one will watch for stdin to become readable
50 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
51 ev_io_start (loop, &stdin_watcher);
53 // initialise a timer watcher, then start it
54 // simple non-repeating 5.5 second timeout
55 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
56 ev_timer_start (loop, &timeout_watcher);
58 // now wait for events to arrive
59 ev_loop (loop, 0);
61 // unloop was called, so exit
62 return 0;
63 }
67 The newest version of this document is also available as an html-formatted
68 web page you might find easier to navigate when reading it for the first
69 time: L<>.
71 Libev is an event loop: you register interest in certain events (such as a
72 file descriptor being readable or a timeout occurring), and it will manage
73 these event sources and provide your program with events.
75 To do this, it must take more or less complete control over your process
76 (or thread) by executing the I<event loop> handler, and will then
77 communicate events via a callback mechanism.
79 You register interest in certain events by registering so-called I<event
80 watchers>, which are relatively small C structures you initialise with the
81 details of the event, and then hand it over to libev by I<starting> the
82 watcher.
84 =head2 FEATURES
86 Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
87 BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
88 for file descriptor events (C<ev_io>), the Linux C<inotify> interface
89 (for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
90 with customised rescheduling (C<ev_periodic>), synchronous signals
91 (C<ev_signal>), process status change events (C<ev_child>), and event
92 watchers dealing with the event loop mechanism itself (C<ev_idle>,
93 C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as
94 file watchers (C<ev_stat>) and even limited support for fork events
95 (C<ev_fork>).
97 It also is quite fast (see this
98 L<benchmark|> comparing it to libevent
99 for example).
101 =head2 CONVENTIONS
103 Libev is very configurable. In this manual the default (and most common)
104 configuration will be described, which supports multiple event loops. For
105 more info about various configuration options please have a look at
106 B<EMBED> section in this manual. If libev was configured without support
107 for multiple event loops, then all functions taking an initial argument of
108 name C<loop> (which is always of type C<struct ev_loop *>) will not have
109 this argument.
113 Libev represents time as a single floating point number, representing the
114 (fractional) number of seconds since the (POSIX) epoch (somewhere near
115 the beginning of 1970, details are complicated, don't ask). This type is
116 called C<ev_tstamp>, which is what you should use too. It usually aliases
117 to the C<double> type in C, and when you need to do any calculations on
118 it, you should treat it as some floatingpoint value. Unlike the name
119 component C<stamp> might indicate, it is also used for time differences
120 throughout libev.
124 These functions can be called anytime, even before initialising the
125 library in any way.
127 =over 4
129 =item ev_tstamp ev_time ()
131 Returns the current time as libev would use it. Please note that the
132 C<ev_now> function is usually faster and also often returns the timestamp
133 you actually want to know.
135 =item ev_sleep (ev_tstamp interval)
137 Sleep for the given interval: The current thread will be blocked until
138 either it is interrupted or the given time interval has passed. Basically
139 this is a subsecond-resolution C<sleep ()>.
141 =item int ev_version_major ()
143 =item int ev_version_minor ()
145 You can find out the major and minor ABI version numbers of the library
146 you linked against by calling the functions C<ev_version_major> and
147 C<ev_version_minor>. If you want, you can compare against the global
148 symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
149 version of the library your program was compiled against.
151 These version numbers refer to the ABI version of the library, not the
152 release version.
154 Usually, it's a good idea to terminate if the major versions mismatch,
155 as this indicates an incompatible change. Minor versions are usually
156 compatible to older versions, so a larger minor version alone is usually
157 not a problem.
159 Example: Make sure we haven't accidentally been linked against the wrong
160 version.
162 assert (("libev version mismatch",
163 ev_version_major () == EV_VERSION_MAJOR
164 && ev_version_minor () >= EV_VERSION_MINOR));
166 =item unsigned int ev_supported_backends ()
168 Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
169 value) compiled into this binary of libev (independent of their
170 availability on the system you are running on). See C<ev_default_loop> for
171 a description of the set values.
173 Example: make sure we have the epoll method, because yeah this is cool and
174 a must have and can we have a torrent of it please!!!11
176 assert (("sorry, no epoll, no sex",
177 ev_supported_backends () & EVBACKEND_EPOLL));
179 =item unsigned int ev_recommended_backends ()
181 Return the set of all backends compiled into this binary of libev and also
182 recommended for this platform. This set is often smaller than the one
183 returned by C<ev_supported_backends>, as for example kqueue is broken on
184 most BSDs and will not be autodetected unless you explicitly request it
185 (assuming you know what you are doing). This is the set of backends that
186 libev will probe for if you specify no backends explicitly.
188 =item unsigned int ev_embeddable_backends ()
190 Returns the set of backends that are embeddable in other event loops. This
191 is the theoretical, all-platform, value. To find which backends
192 might be supported on the current system, you would need to look at
193 C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
194 recommended ones.
196 See the description of C<ev_embed> watchers for more info.
198 =item ev_set_allocator (void *(*cb)(void *ptr, long size))
200 Sets the allocation function to use (the prototype is similar - the
201 semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
202 used to allocate and free memory (no surprises here). If it returns zero
203 when memory needs to be allocated (C<size != 0>), the library might abort
204 or take some potentially destructive action.
206 Since some systems (at least OpenBSD and Darwin) fail to implement
207 correct C<realloc> semantics, libev will use a wrapper around the system
208 C<realloc> and C<free> functions by default.
210 You could override this function in high-availability programs to, say,
211 free some memory if it cannot allocate memory, to use a special allocator,
212 or even to sleep a while and retry until some memory is available.
214 Example: Replace the libev allocator with one that waits a bit and then
215 retries (example requires a standards-compliant C<realloc>).
217 static void *
218 persistent_realloc (void *ptr, size_t size)
219 {
220 for (;;)
221 {
222 void *newptr = realloc (ptr, size);
224 if (newptr)
225 return newptr;
227 sleep (60);
228 }
229 }
231 ...
232 ev_set_allocator (persistent_realloc);
234 =item ev_set_syserr_cb (void (*cb)(const char *msg));
236 Set the callback function to call on a retryable syscall error (such
237 as failed select, poll, epoll_wait). The message is a printable string
238 indicating the system call or subsystem causing the problem. If this
239 callback is set, then libev will expect it to remedy the sitution, no
240 matter what, when it returns. That is, libev will generally retry the
241 requested operation, or, if the condition doesn't go away, do bad stuff
242 (such as abort).
244 Example: This is basically the same thing that libev does internally, too.
246 static void
247 fatal_error (const char *msg)
248 {
249 perror (msg);
250 abort ();
251 }
253 ...
254 ev_set_syserr_cb (fatal_error);
256 =back
260 An event loop is described by a C<struct ev_loop *>. The library knows two
261 types of such loops, the I<default> loop, which supports signals and child
262 events, and dynamically created loops which do not.
264 =over 4
266 =item struct ev_loop *ev_default_loop (unsigned int flags)
268 This will initialise the default event loop if it hasn't been initialised
269 yet and return it. If the default loop could not be initialised, returns
270 false. If it already was initialised it simply returns it (and ignores the
271 flags. If that is troubling you, check C<ev_backend ()> afterwards).
273 If you don't know what event loop to use, use the one returned from this
274 function.
276 Note that this function is I<not> thread-safe, so if you want to use it
277 from multiple threads, you have to lock (note also that this is unlikely,
278 as loops cannot bes hared easily between threads anyway).
280 The default loop is the only loop that can handle C<ev_signal> and
281 C<ev_child> watchers, and to do this, it always registers a handler
282 for C<SIGCHLD>. If this is a problem for your app you can either
283 create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
284 can simply overwrite the C<SIGCHLD> signal handler I<after> calling
285 C<ev_default_init>.
287 The flags argument can be used to specify special behaviour or specific
288 backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
290 The following flags are supported:
292 =over 4
294 =item C<EVFLAG_AUTO>
296 The default flags value. Use this if you have no clue (it's the right
297 thing, believe me).
299 =item C<EVFLAG_NOENV>
301 If this flag bit is ored into the flag value (or the program runs setuid
302 or setgid) then libev will I<not> look at the environment variable
303 C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
304 override the flags completely if it is found in the environment. This is
305 useful to try out specific backends to test their performance, or to work
306 around bugs.
310 Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
311 a fork, you can also make libev check for a fork in each iteration by
312 enabling this flag.
314 This works by calling C<getpid ()> on every iteration of the loop,
315 and thus this might slow down your event loop if you do a lot of loop
316 iterations and little real work, but is usually not noticeable (on my
317 GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
318 without a syscall and thus I<very> fast, but my GNU/Linux system also has
319 C<pthread_atfork> which is even faster).
321 The big advantage of this flag is that you can forget about fork (and
322 forget about forgetting to tell libev about forking) when you use this
323 flag.
325 This flag setting cannot be overriden or specified in the C<LIBEV_FLAGS>
326 environment variable.
328 =item C<EVBACKEND_SELECT> (value 1, portable select backend)
330 This is your standard select(2) backend. Not I<completely> standard, as
331 libev tries to roll its own fd_set with no limits on the number of fds,
332 but if that fails, expect a fairly low limit on the number of fds when
333 using this backend. It doesn't scale too well (O(highest_fd)), but its
334 usually the fastest backend for a low number of (low-numbered :) fds.
336 To get good performance out of this backend you need a high amount of
337 parallelity (most of the file descriptors should be busy). If you are
338 writing a server, you should C<accept ()> in a loop to accept as many
339 connections as possible during one iteration. You might also want to have
340 a look at C<ev_set_io_collect_interval ()> to increase the amount of
341 readyness notifications you get per iteration.
343 =item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
345 And this is your standard poll(2) backend. It's more complicated
346 than select, but handles sparse fds better and has no artificial
347 limit on the number of fds you can use (except it will slow down
348 considerably with a lot of inactive fds). It scales similarly to select,
349 i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
350 performance tips.
352 =item C<EVBACKEND_EPOLL> (value 4, Linux)
354 For few fds, this backend is a bit little slower than poll and select,
355 but it scales phenomenally better. While poll and select usually scale
356 like O(total_fds) where n is the total number of fds (or the highest fd),
357 epoll scales either O(1) or O(active_fds). The epoll design has a number
358 of shortcomings, such as silently dropping events in some hard-to-detect
359 cases and requiring a syscall per fd change, no fork support and bad
360 support for dup.
362 While stopping, setting and starting an I/O watcher in the same iteration
363 will result in some caching, there is still a syscall per such incident
364 (because the fd could point to a different file description now), so its
365 best to avoid that. Also, C<dup ()>'ed file descriptors might not work
366 very well if you register events for both fds.
368 Please note that epoll sometimes generates spurious notifications, so you
369 need to use non-blocking I/O or other means to avoid blocking when no data
370 (or space) is available.
372 Best performance from this backend is achieved by not unregistering all
373 watchers for a file descriptor until it has been closed, if possible, i.e.
374 keep at least one watcher active per fd at all times.
376 While nominally embeddeble in other event loops, this feature is broken in
377 all kernel versions tested so far.
379 =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
381 Kqueue deserves special mention, as at the time of this writing, it
382 was broken on all BSDs except NetBSD (usually it doesn't work reliably
383 with anything but sockets and pipes, except on Darwin, where of course
384 it's completely useless). For this reason it's not being "autodetected"
385 unless you explicitly specify it explicitly in the flags (i.e. using
386 C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
387 system like NetBSD.
389 You still can embed kqueue into a normal poll or select backend and use it
390 only for sockets (after having made sure that sockets work with kqueue on
391 the target platform). See C<ev_embed> watchers for more info.
393 It scales in the same way as the epoll backend, but the interface to the
394 kernel is more efficient (which says nothing about its actual speed, of
395 course). While stopping, setting and starting an I/O watcher does never
396 cause an extra syscall as with C<EVBACKEND_EPOLL>, it still adds up to
397 two event changes per incident, support for C<fork ()> is very bad and it
398 drops fds silently in similarly hard-to-detect cases.
400 This backend usually performs well under most conditions.
402 While nominally embeddable in other event loops, this doesn't work
403 everywhere, so you might need to test for this. And since it is broken
404 almost everywhere, you should only use it when you have a lot of sockets
405 (for which it usually works), by embedding it into another event loop
406 (e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and using it only for
407 sockets.
409 =item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
411 This is not implemented yet (and might never be, unless you send me an
412 implementation). According to reports, C</dev/poll> only supports sockets
413 and is not embeddable, which would limit the usefulness of this backend
414 immensely.
416 =item C<EVBACKEND_PORT> (value 32, Solaris 10)
418 This uses the Solaris 10 event port mechanism. As with everything on Solaris,
419 it's really slow, but it still scales very well (O(active_fds)).
421 Please note that solaris event ports can deliver a lot of spurious
422 notifications, so you need to use non-blocking I/O or other means to avoid
423 blocking when no data (or space) is available.
425 While this backend scales well, it requires one system call per active
426 file descriptor per loop iteration. For small and medium numbers of file
427 descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
428 might perform better.
430 On the positive side, ignoring the spurious readyness notifications, this
431 backend actually performed to specification in all tests and is fully
432 embeddable, which is a rare feat among the OS-specific backends.
434 =item C<EVBACKEND_ALL>
436 Try all backends (even potentially broken ones that wouldn't be tried
437 with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
440 It is definitely not recommended to use this flag.
442 =back
444 If one or more of these are ored into the flags value, then only these
445 backends will be tried (in the reverse order as listed here). If none are
446 specified, all backends in C<ev_recommended_backends ()> will be tried.
448 The most typical usage is like this:
450 if (!ev_default_loop (0))
451 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
453 Restrict libev to the select and poll backends, and do not allow
454 environment settings to be taken into account:
458 Use whatever libev has to offer, but make sure that kqueue is used if
459 available (warning, breaks stuff, best use only with your own private
460 event loop and only if you know the OS supports your types of fds):
462 ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
464 =item struct ev_loop *ev_loop_new (unsigned int flags)
466 Similar to C<ev_default_loop>, but always creates a new event loop that is
467 always distinct from the default loop. Unlike the default loop, it cannot
468 handle signal and child watchers, and attempts to do so will be greeted by
469 undefined behaviour (or a failed assertion if assertions are enabled).
471 Note that this function I<is> thread-safe, and the recommended way to use
472 libev with threads is indeed to create one loop per thread, and using the
473 default loop in the "main" or "initial" thread.
475 Example: Try to create a event loop that uses epoll and nothing else.
477 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
478 if (!epoller)
479 fatal ("no epoll found here, maybe it hides under your chair");
481 =item ev_default_destroy ()
483 Destroys the default loop again (frees all memory and kernel state
484 etc.). None of the active event watchers will be stopped in the normal
485 sense, so e.g. C<ev_is_active> might still return true. It is your
486 responsibility to either stop all watchers cleanly yoursef I<before>
487 calling this function, or cope with the fact afterwards (which is usually
488 the easiest thing, you can just ignore the watchers and/or C<free ()> them
489 for example).
491 Note that certain global state, such as signal state, will not be freed by
492 this function, and related watchers (such as signal and child watchers)
493 would need to be stopped manually.
495 In general it is not advisable to call this function except in the
496 rare occasion where you really need to free e.g. the signal handling
497 pipe fds. If you need dynamically allocated loops it is better to use
498 C<ev_loop_new> and C<ev_loop_destroy>).
500 =item ev_loop_destroy (loop)
502 Like C<ev_default_destroy>, but destroys an event loop created by an
503 earlier call to C<ev_loop_new>.
505 =item ev_default_fork ()
507 This function sets a flag that causes subsequent C<ev_loop> iterations
508 to reinitialise the kernel state for backends that have one. Despite the
509 name, you can call it anytime, but it makes most sense after forking, in
510 the child process (or both child and parent, but that again makes little
511 sense). You I<must> call it in the child before using any of the libev
512 functions, and it will only take effect at the next C<ev_loop> iteration.
514 On the other hand, you only need to call this function in the child
515 process if and only if you want to use the event library in the child. If
516 you just fork+exec, you don't have to call it at all.
518 The function itself is quite fast and it's usually not a problem to call
519 it just in case after a fork. To make this easy, the function will fit in
520 quite nicely into a call to C<pthread_atfork>:
522 pthread_atfork (0, 0, ev_default_fork);
524 =item ev_loop_fork (loop)
526 Like C<ev_default_fork>, but acts on an event loop created by
527 C<ev_loop_new>. Yes, you have to call this on every allocated event loop
528 after fork, and how you do this is entirely your own problem.
530 =item int ev_is_default_loop (loop)
532 Returns true when the given loop actually is the default loop, false otherwise.
534 =item unsigned int ev_loop_count (loop)
536 Returns the count of loop iterations for the loop, which is identical to
537 the number of times libev did poll for new events. It starts at C<0> and
538 happily wraps around with enough iterations.
540 This value can sometimes be useful as a generation counter of sorts (it
541 "ticks" the number of loop iterations), as it roughly corresponds with
542 C<ev_prepare> and C<ev_check> calls.
544 =item unsigned int ev_backend (loop)
546 Returns one of the C<EVBACKEND_*> flags indicating the event backend in
547 use.
549 =item ev_tstamp ev_now (loop)
551 Returns the current "event loop time", which is the time the event loop
552 received events and started processing them. This timestamp does not
553 change as long as callbacks are being processed, and this is also the base
554 time used for relative timers. You can treat it as the timestamp of the
555 event occurring (or more correctly, libev finding out about it).
557 =item ev_loop (loop, int flags)
559 Finally, this is it, the event handler. This function usually is called
560 after you initialised all your watchers and you want to start handling
561 events.
563 If the flags argument is specified as C<0>, it will not return until
564 either no event watchers are active anymore or C<ev_unloop> was called.
566 Please note that an explicit C<ev_unloop> is usually better than
567 relying on all watchers to be stopped when deciding when a program has
568 finished (especially in interactive programs), but having a program that
569 automatically loops as long as it has to and no longer by virtue of
570 relying on its watchers stopping correctly is a thing of beauty.
572 A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
573 those events and any outstanding ones, but will not block your process in
574 case there are no events and will return after one iteration of the loop.
576 A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
577 neccessary) and will handle those and any outstanding ones. It will block
578 your process until at least one new event arrives, and will return after
579 one iteration of the loop. This is useful if you are waiting for some
580 external event in conjunction with something not expressible using other
581 libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is
582 usually a better approach for this kind of thing.
584 Here are the gory details of what C<ev_loop> does:
586 - Before the first iteration, call any pending watchers.
587 * If EVFLAG_FORKCHECK was used, check for a fork.
588 - If a fork was detected, queue and call all fork watchers.
589 - Queue and call all prepare watchers.
590 - If we have been forked, recreate the kernel state.
591 - Update the kernel state with all outstanding changes.
592 - Update the "event loop time".
593 - Calculate for how long to sleep or block, if at all
594 (active idle watchers, EVLOOP_NONBLOCK or not having
595 any active watchers at all will result in not sleeping).
596 - Sleep if the I/O and timer collect interval say so.
597 - Block the process, waiting for any events.
598 - Queue all outstanding I/O (fd) events.
599 - Update the "event loop time" and do time jump handling.
600 - Queue all outstanding timers.
601 - Queue all outstanding periodics.
602 - If no events are pending now, queue all idle watchers.
603 - Queue all check watchers.
604 - Call all queued watchers in reverse order (i.e. check watchers first).
605 Signals and child watchers are implemented as I/O watchers, and will
606 be handled here by queueing them when their watcher gets executed.
607 - If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
608 were used, or there are no active watchers, return, otherwise
609 continue with step *.
611 Example: Queue some jobs and then loop until no events are outstanding
612 anymore.
614 ... queue jobs here, make sure they register event watchers as long
615 ... as they still have work to do (even an idle watcher will do..)
616 ev_loop (my_loop, 0);
617 ... jobs done. yeah!
619 =item ev_unloop (loop, how)
621 Can be used to make a call to C<ev_loop> return early (but only after it
622 has processed all outstanding events). The C<how> argument must be either
623 C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
624 C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
626 This "unloop state" will be cleared when entering C<ev_loop> again.
628 =item ev_ref (loop)
630 =item ev_unref (loop)
632 Ref/unref can be used to add or remove a reference count on the event
633 loop: Every watcher keeps one reference, and as long as the reference
634 count is nonzero, C<ev_loop> will not return on its own. If you have
635 a watcher you never unregister that should not keep C<ev_loop> from
636 returning, ev_unref() after starting, and ev_ref() before stopping it. For
637 example, libev itself uses this for its internal signal pipe: It is not
638 visible to the libev user and should not keep C<ev_loop> from exiting if
639 no event watchers registered by it are active. It is also an excellent
640 way to do this for generic recurring timers or from within third-party
641 libraries. Just remember to I<unref after start> and I<ref before stop>
642 (but only if the watcher wasn't active before, or was active before,
643 respectively).
645 Example: Create a signal watcher, but keep it from keeping C<ev_loop>
646 running when nothing else is active.
648 struct ev_signal exitsig;
649 ev_signal_init (&exitsig, sig_cb, SIGINT);
650 ev_signal_start (loop, &exitsig);
651 evf_unref (loop);
653 Example: For some weird reason, unregister the above signal handler again.
655 ev_ref (loop);
656 ev_signal_stop (loop, &exitsig);
658 =item ev_set_io_collect_interval (loop, ev_tstamp interval)
660 =item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
662 These advanced functions influence the time that libev will spend waiting
663 for events. Both are by default C<0>, meaning that libev will try to
664 invoke timer/periodic callbacks and I/O callbacks with minimum latency.
666 Setting these to a higher value (the C<interval> I<must> be >= C<0>)
667 allows libev to delay invocation of I/O and timer/periodic callbacks to
668 increase efficiency of loop iterations.
670 The background is that sometimes your program runs just fast enough to
671 handle one (or very few) event(s) per loop iteration. While this makes
672 the program responsive, it also wastes a lot of CPU time to poll for new
673 events, especially with backends like C<select ()> which have a high
674 overhead for the actual polling but can deliver many events at once.
676 By setting a higher I<io collect interval> you allow libev to spend more
677 time collecting I/O events, so you can handle more events per iteration,
678 at the cost of increasing latency. Timeouts (both C<ev_periodic> and
679 C<ev_timer>) will be not affected. Setting this to a non-null value will
680 introduce an additional C<ev_sleep ()> call into most loop iterations.
682 Likewise, by setting a higher I<timeout collect interval> you allow libev
683 to spend more time collecting timeouts, at the expense of increased
684 latency (the watcher callback will be called later). C<ev_io> watchers
685 will not be affected. Setting this to a non-null value will not introduce
686 any overhead in libev.
688 Many (busy) programs can usually benefit by setting the io collect
689 interval to a value near C<0.1> or so, which is often enough for
690 interactive servers (of course not for games), likewise for timeouts. It
691 usually doesn't make much sense to set it to a lower value than C<0.01>,
692 as this approsaches the timing granularity of most systems.
694 =back
699 A watcher is a structure that you create and register to record your
700 interest in some event. For instance, if you want to wait for STDIN to
701 become readable, you would create an C<ev_io> watcher for that:
703 static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)
704 {
705 ev_io_stop (w);
706 ev_unloop (loop, EVUNLOOP_ALL);
707 }
709 struct ev_loop *loop = ev_default_loop (0);
710 struct ev_io stdin_watcher;
711 ev_init (&stdin_watcher, my_cb);
712 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
713 ev_io_start (loop, &stdin_watcher);
714 ev_loop (loop, 0);
716 As you can see, you are responsible for allocating the memory for your
717 watcher structures (and it is usually a bad idea to do this on the stack,
718 although this can sometimes be quite valid).
720 Each watcher structure must be initialised by a call to C<ev_init
721 (watcher *, callback)>, which expects a callback to be provided. This
722 callback gets invoked each time the event occurs (or, in the case of io
723 watchers, each time the event loop detects that the file descriptor given
724 is readable and/or writable).
726 Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro
727 with arguments specific to this watcher type. There is also a macro
728 to combine initialisation and setting in one call: C<< ev_<type>_init
729 (watcher *, callback, ...) >>.
731 To make the watcher actually watch out for events, you have to start it
732 with a watcher-specific start function (C<< ev_<type>_start (loop, watcher
733 *) >>), and you can stop watching for events at any time by calling the
734 corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.
736 As long as your watcher is active (has been started but not stopped) you
737 must not touch the values stored in it. Most specifically you must never
738 reinitialise it or call its C<set> macro.
740 Each and every callback receives the event loop pointer as first, the
741 registered watcher structure as second, and a bitset of received events as
742 third argument.
744 The received events usually include a single bit per event type received
745 (you can receive multiple events at the same time). The possible bit masks
746 are:
748 =over 4
750 =item C<EV_READ>
752 =item C<EV_WRITE>
754 The file descriptor in the C<ev_io> watcher has become readable and/or
755 writable.
757 =item C<EV_TIMEOUT>
759 The C<ev_timer> watcher has timed out.
761 =item C<EV_PERIODIC>
763 The C<ev_periodic> watcher has timed out.
765 =item C<EV_SIGNAL>
767 The signal specified in the C<ev_signal> watcher has been received by a thread.
769 =item C<EV_CHILD>
771 The pid specified in the C<ev_child> watcher has received a status change.
773 =item C<EV_STAT>
775 The path specified in the C<ev_stat> watcher changed its attributes somehow.
777 =item C<EV_IDLE>
779 The C<ev_idle> watcher has determined that you have nothing better to do.
781 =item C<EV_PREPARE>
783 =item C<EV_CHECK>
785 All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts
786 to gather new events, and all C<ev_check> watchers are invoked just after
787 C<ev_loop> has gathered them, but before it invokes any callbacks for any
788 received events. Callbacks of both watcher types can start and stop as
789 many watchers as they want, and all of them will be taken into account
790 (for example, a C<ev_prepare> watcher might start an idle watcher to keep
791 C<ev_loop> from blocking).
793 =item C<EV_EMBED>
795 The embedded event loop specified in the C<ev_embed> watcher needs attention.
797 =item C<EV_FORK>
799 The event loop has been resumed in the child process after fork (see
800 C<ev_fork>).
802 =item C<EV_ASYNC>
804 The given async watcher has been asynchronously notified (see C<ev_async>).
806 =item C<EV_ERROR>
808 An unspecified error has occured, the watcher has been stopped. This might
809 happen because the watcher could not be properly started because libev
810 ran out of memory, a file descriptor was found to be closed or any other
811 problem. You best act on it by reporting the problem and somehow coping
812 with the watcher being stopped.
814 Libev will usually signal a few "dummy" events together with an error,
815 for example it might indicate that a fd is readable or writable, and if
816 your callbacks is well-written it can just attempt the operation and cope
817 with the error from read() or write(). This will not work in multithreaded
818 programs, though, so beware.
820 =back
824 In the following description, C<TYPE> stands for the watcher type,
825 e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
827 =over 4
829 =item C<ev_init> (ev_TYPE *watcher, callback)
831 This macro initialises the generic portion of a watcher. The contents
832 of the watcher object can be arbitrary (so C<malloc> will do). Only
833 the generic parts of the watcher are initialised, you I<need> to call
834 the type-specific C<ev_TYPE_set> macro afterwards to initialise the
835 type-specific parts. For each type there is also a C<ev_TYPE_init> macro
836 which rolls both calls into one.
838 You can reinitialise a watcher at any time as long as it has been stopped
839 (or never started) and there are no pending events outstanding.
841 The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,
842 int revents)>.
844 =item C<ev_TYPE_set> (ev_TYPE *, [args])
846 This macro initialises the type-specific parts of a watcher. You need to
847 call C<ev_init> at least once before you call this macro, but you can
848 call C<ev_TYPE_set> any number of times. You must not, however, call this
849 macro on a watcher that is active (it can be pending, however, which is a
850 difference to the C<ev_init> macro).
852 Although some watcher types do not have type-specific arguments
853 (e.g. C<ev_prepare>) you still need to call its C<set> macro.
855 =item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
857 This convinience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
858 calls into a single call. This is the most convinient method to initialise
859 a watcher. The same limitations apply, of course.
861 =item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
863 Starts (activates) the given watcher. Only active watchers will receive
864 events. If the watcher is already active nothing will happen.
866 =item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
868 Stops the given watcher again (if active) and clears the pending
869 status. It is possible that stopped watchers are pending (for example,
870 non-repeating timers are being stopped when they become pending), but
871 C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If
872 you want to free or reuse the memory used by the watcher it is therefore a
873 good idea to always call its C<ev_TYPE_stop> function.
875 =item bool ev_is_active (ev_TYPE *watcher)
877 Returns a true value iff the watcher is active (i.e. it has been started
878 and not yet been stopped). As long as a watcher is active you must not modify
879 it.
881 =item bool ev_is_pending (ev_TYPE *watcher)
883 Returns a true value iff the watcher is pending, (i.e. it has outstanding
884 events but its callback has not yet been invoked). As long as a watcher
885 is pending (but not active) you must not call an init function on it (but
886 C<ev_TYPE_set> is safe), you must not change its priority, and you must
887 make sure the watcher is available to libev (e.g. you cannot C<free ()>
888 it).
890 =item callback ev_cb (ev_TYPE *watcher)
892 Returns the callback currently set on the watcher.
894 =item ev_cb_set (ev_TYPE *watcher, callback)
896 Change the callback. You can change the callback at virtually any time
897 (modulo threads).
899 =item ev_set_priority (ev_TYPE *watcher, priority)
901 =item int ev_priority (ev_TYPE *watcher)
903 Set and query the priority of the watcher. The priority is a small
904 integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
905 (default: C<-2>). Pending watchers with higher priority will be invoked
906 before watchers with lower priority, but priority will not keep watchers
907 from being executed (except for C<ev_idle> watchers).
909 This means that priorities are I<only> used for ordering callback
910 invocation after new events have been received. This is useful, for
911 example, to reduce latency after idling, or more often, to bind two
912 watchers on the same event and make sure one is called first.
914 If you need to suppress invocation when higher priority events are pending
915 you need to look at C<ev_idle> watchers, which provide this functionality.
917 You I<must not> change the priority of a watcher as long as it is active or
918 pending.
920 The default priority used by watchers when no priority has been set is
921 always C<0>, which is supposed to not be too high and not be too low :).
923 Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
924 fine, as long as you do not mind that the priority value you query might
925 or might not have been adjusted to be within valid range.
927 =item ev_invoke (loop, ev_TYPE *watcher, int revents)
929 Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
930 C<loop> nor C<revents> need to be valid as long as the watcher callback
931 can deal with that fact.
933 =item int ev_clear_pending (loop, ev_TYPE *watcher)
935 If the watcher is pending, this function returns clears its pending status
936 and returns its C<revents> bitset (as if its callback was invoked). If the
937 watcher isn't pending it does nothing and returns C<0>.
939 =back
944 Each watcher has, by default, a member C<void *data> that you can change
945 and read at any time, libev will completely ignore it. This can be used
946 to associate arbitrary data with your watcher. If you need more data and
947 don't want to allocate memory and store a pointer to it in that data
948 member, you can also "subclass" the watcher type and provide your own
949 data:
951 struct my_io
952 {
953 struct ev_io io;
954 int otherfd;
955 void *somedata;
956 struct whatever *mostinteresting;
957 }
959 And since your callback will be called with a pointer to the watcher, you
960 can cast it back to your own type:
962 static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
963 {
964 struct my_io *w = (struct my_io *)w_;
965 ...
966 }
968 More interesting and less C-conformant ways of casting your callback type
969 instead have been omitted.
971 Another common scenario is having some data structure with multiple
972 watchers:
974 struct my_biggy
975 {
976 int some_data;
977 ev_timer t1;
978 ev_timer t2;
979 }
981 In this case getting the pointer to C<my_biggy> is a bit more complicated,
982 you need to use C<offsetof>:
984 #include <stddef.h>
986 static void
987 t1_cb (EV_P_ struct ev_timer *w, int revents)
988 {
989 struct my_biggy big = (struct my_biggy *
990 (((char *)w) - offsetof (struct my_biggy, t1));
991 }
993 static void
994 t2_cb (EV_P_ struct ev_timer *w, int revents)
995 {
996 struct my_biggy big = (struct my_biggy *
997 (((char *)w) - offsetof (struct my_biggy, t2));
998 }
1001 =head1 WATCHER TYPES
1003 This section describes each watcher in detail, but will not repeat
1004 information given in the last section. Any initialisation/set macros,
1005 functions and members specific to the watcher type are explained.
1007 Members are additionally marked with either I<[read-only]>, meaning that,
1008 while the watcher is active, you can look at the member and expect some
1009 sensible content, but you must not modify it (you can modify it while the
1010 watcher is stopped to your hearts content), or I<[read-write]>, which
1011 means you can expect it to have some sensible content while the watcher
1012 is active, but you can also modify it. Modifying it may not do something
1013 sensible or take immediate effect (or do anything at all), but libev will
1014 not crash or malfunction in any way.
1017 =head2 C<ev_io> - is this file descriptor readable or writable?
1019 I/O watchers check whether a file descriptor is readable or writable
1020 in each iteration of the event loop, or, more precisely, when reading
1021 would not block the process and writing would at least be able to write
1022 some data. This behaviour is called level-triggering because you keep
1023 receiving events as long as the condition persists. Remember you can stop
1024 the watcher if you don't want to act on the event and neither want to
1025 receive future events.
1027 In general you can register as many read and/or write event watchers per
1028 fd as you want (as long as you don't confuse yourself). Setting all file
1029 descriptors to non-blocking mode is also usually a good idea (but not
1030 required if you know what you are doing).
1032 If you must do this, then force the use of a known-to-be-good backend
1033 (at the time of this writing, this includes only C<EVBACKEND_SELECT> and
1036 Another thing you have to watch out for is that it is quite easy to
1037 receive "spurious" readyness notifications, that is your callback might
1038 be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1039 because there is no data. Not only are some backends known to create a
1040 lot of those (for example solaris ports), it is very easy to get into
1041 this situation even with a relatively standard program structure. Thus
1042 it is best to always use non-blocking I/O: An extra C<read>(2) returning
1043 C<EAGAIN> is far preferable to a program hanging until some data arrives.
1045 If you cannot run the fd in non-blocking mode (for example you should not
1046 play around with an Xlib connection), then you have to seperately re-test
1047 whether a file descriptor is really ready with a known-to-be good interface
1048 such as poll (fortunately in our Xlib example, Xlib already does this on
1049 its own, so its quite safe to use).
1051 =head3 The special problem of disappearing file descriptors
1053 Some backends (e.g. kqueue, epoll) need to be told about closing a file
1054 descriptor (either by calling C<close> explicitly or by any other means,
1055 such as C<dup>). The reason is that you register interest in some file
1056 descriptor, but when it goes away, the operating system will silently drop
1057 this interest. If another file descriptor with the same number then is
1058 registered with libev, there is no efficient way to see that this is, in
1059 fact, a different file descriptor.
1061 To avoid having to explicitly tell libev about such cases, libev follows
1062 the following policy: Each time C<ev_io_set> is being called, libev
1063 will assume that this is potentially a new file descriptor, otherwise
1064 it is assumed that the file descriptor stays the same. That means that
1065 you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1066 descriptor even if the file descriptor number itself did not change.
1068 This is how one would do it normally anyway, the important point is that
1069 the libev application should not optimise around libev but should leave
1070 optimisations to libev.
1072 =head3 The special problem of dup'ed file descriptors
1074 Some backends (e.g. epoll), cannot register events for file descriptors,
1075 but only events for the underlying file descriptions. That means when you
1076 have C<dup ()>'ed file descriptors or weirder constellations, and register
1077 events for them, only one file descriptor might actually receive events.
1079 There is no workaround possible except not registering events
1080 for potentially C<dup ()>'ed file descriptors, or to resort to
1083 =head3 The special problem of fork
1085 Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1086 useless behaviour. Libev fully supports fork, but needs to be told about
1087 it in the child.
1089 To support fork in your programs, you either have to call
1090 C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
1091 enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1094 =head3 The special problem of SIGPIPE
1096 While not really specific to libev, it is easy to forget about SIGPIPE:
1097 when reading from a pipe whose other end has been closed, your program
1098 gets send a SIGPIPE, which, by default, aborts your program. For most
1099 programs this is sensible behaviour, for daemons, this is usually
1100 undesirable.
1102 So when you encounter spurious, unexplained daemon exits, make sure you
1103 ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1104 somewhere, as that would have given you a big clue).
1107 =head3 Watcher-Specific Functions
1109 =over 4
1111 =item ev_io_init (ev_io *, callback, int fd, int events)
1113 =item ev_io_set (ev_io *, int fd, int events)
1115 Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1116 rceeive events for and events is either C<EV_READ>, C<EV_WRITE> or
1117 C<EV_READ | EV_WRITE> to receive the given events.
1119 =item int fd [read-only]
1121 The file descriptor being watched.
1123 =item int events [read-only]
1125 The events being watched.
1127 =back
1129 =head3 Examples
1131 Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1132 readable, but only once. Since it is likely line-buffered, you could
1133 attempt to read a whole line in the callback.
1135 static void
1136 stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1137 {
1138 ev_io_stop (loop, w);
1139 .. read from stdin here (or from w->fd) and haqndle any I/O errors
1140 }
1142 ...
1143 struct ev_loop *loop = ev_default_init (0);
1144 struct ev_io stdin_readable;
1145 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1146 ev_io_start (loop, &stdin_readable);
1147 ev_loop (loop, 0);
1150 =head2 C<ev_timer> - relative and optionally repeating timeouts
1152 Timer watchers are simple relative timers that generate an event after a
1153 given time, and optionally repeating in regular intervals after that.
1155 The timers are based on real time, that is, if you register an event that
1156 times out after an hour and you reset your system clock to last years
1157 time, it will still time out after (roughly) and hour. "Roughly" because
1158 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1159 monotonic clock option helps a lot here).
1161 The relative timeouts are calculated relative to the C<ev_now ()>
1162 time. This is usually the right thing as this timestamp refers to the time
1163 of the event triggering whatever timeout you are modifying/starting. If
1164 you suspect event processing to be delayed and you I<need> to base the timeout
1165 on the current time, use something like this to adjust for this:
1167 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1169 The callback is guarenteed to be invoked only when its timeout has passed,
1170 but if multiple timers become ready during the same loop iteration then
1171 order of execution is undefined.
1173 =head3 Watcher-Specific Functions and Data Members
1175 =over 4
1177 =item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1179 =item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1181 Configure the timer to trigger after C<after> seconds. If C<repeat> is
1182 C<0.>, then it will automatically be stopped. If it is positive, then the
1183 timer will automatically be configured to trigger again C<repeat> seconds
1184 later, again, and again, until stopped manually.
1186 The timer itself will do a best-effort at avoiding drift, that is, if you
1187 configure a timer to trigger every 10 seconds, then it will trigger at
1188 exactly 10 second intervals. If, however, your program cannot keep up with
1189 the timer (because it takes longer than those 10 seconds to do stuff) the
1190 timer will not fire more than once per event loop iteration.
1192 =item ev_timer_again (loop, ev_timer *)
1194 This will act as if the timer timed out and restart it again if it is
1195 repeating. The exact semantics are:
1197 If the timer is pending, its pending status is cleared.
1199 If the timer is started but nonrepeating, stop it (as if it timed out).
1201 If the timer is repeating, either start it if necessary (with the
1202 C<repeat> value), or reset the running timer to the C<repeat> value.
1204 This sounds a bit complicated, but here is a useful and typical
1205 example: Imagine you have a tcp connection and you want a so-called idle
1206 timeout, that is, you want to be called when there have been, say, 60
1207 seconds of inactivity on the socket. The easiest way to do this is to
1208 configure an C<ev_timer> with a C<repeat> value of C<60> and then call
1209 C<ev_timer_again> each time you successfully read or write some data. If
1210 you go into an idle state where you do not expect data to travel on the
1211 socket, you can C<ev_timer_stop> the timer, and C<ev_timer_again> will
1212 automatically restart it if need be.
1214 That means you can ignore the C<after> value and C<ev_timer_start>
1215 altogether and only ever use the C<repeat> value and C<ev_timer_again>:
1217 ev_timer_init (timer, callback, 0., 5.);
1218 ev_timer_again (loop, timer);
1219 ...
1220 timer->again = 17.;
1221 ev_timer_again (loop, timer);
1222 ...
1223 timer->again = 10.;
1224 ev_timer_again (loop, timer);
1226 This is more slightly efficient then stopping/starting the timer each time
1227 you want to modify its timeout value.
1229 =item ev_tstamp repeat [read-write]
1231 The current C<repeat> value. Will be used each time the watcher times out
1232 or C<ev_timer_again> is called and determines the next timeout (if any),
1233 which is also when any modifications are taken into account.
1235 =back
1237 =head3 Examples
1239 Example: Create a timer that fires after 60 seconds.
1241 static void
1242 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1243 {
1244 .. one minute over, w is actually stopped right here
1245 }
1247 struct ev_timer mytimer;
1248 ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1249 ev_timer_start (loop, &mytimer);
1251 Example: Create a timeout timer that times out after 10 seconds of
1252 inactivity.
1254 static void
1255 timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1256 {
1257 .. ten seconds without any activity
1258 }
1260 struct ev_timer mytimer;
1261 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1262 ev_timer_again (&mytimer); /* start timer */
1263 ev_loop (loop, 0);
1265 // and in some piece of code that gets executed on any "activity":
1266 // reset the timeout to start ticking again at 10 seconds
1267 ev_timer_again (&mytimer);
1270 =head2 C<ev_periodic> - to cron or not to cron?
1272 Periodic watchers are also timers of a kind, but they are very versatile
1273 (and unfortunately a bit complex).
1275 Unlike C<ev_timer>'s, they are not based on real time (or relative time)
1276 but on wallclock time (absolute time). You can tell a periodic watcher
1277 to trigger "at" some specific point in time. For example, if you tell a
1278 periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()
1279 + 10.>) and then reset your system clock to the last year, then it will
1280 take a year to trigger the event (unlike an C<ev_timer>, which would trigger
1281 roughly 10 seconds later).
1283 They can also be used to implement vastly more complex timers, such as
1284 triggering an event on each midnight, local time or other, complicated,
1285 rules.
1287 As with timers, the callback is guarenteed to be invoked only when the
1288 time (C<at>) has been passed, but if multiple periodic timers become ready
1289 during the same loop iteration then order of execution is undefined.
1291 =head3 Watcher-Specific Functions and Data Members
1293 =over 4
1295 =item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
1297 =item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
1299 Lots of arguments, lets sort it out... There are basically three modes of
1300 operation, and we will explain them from simplest to complex:
1302 =over 4
1304 =item * absolute timer (at = time, interval = reschedule_cb = 0)
1306 In this configuration the watcher triggers an event at the wallclock time
1307 C<at> and doesn't repeat. It will not adjust when a time jump occurs,
1308 that is, if it is to be run at January 1st 2011 then it will run when the
1309 system time reaches or surpasses this time.
1311 =item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1313 In this mode the watcher will always be scheduled to time out at the next
1314 C<at + N * interval> time (for some integer N, which can also be negative)
1315 and then repeat, regardless of any time jumps.
1317 This can be used to create timers that do not drift with respect to system
1318 time:
1320 ev_periodic_set (&periodic, 0., 3600., 0);
1322 This doesn't mean there will always be 3600 seconds in between triggers,
1323 but only that the the callback will be called when the system time shows a
1324 full hour (UTC), or more correctly, when the system time is evenly divisible
1325 by 3600.
1327 Another way to think about it (for the mathematically inclined) is that
1328 C<ev_periodic> will try to run the callback in this mode at the next possible
1329 time where C<time = at (mod interval)>, regardless of any time jumps.
1331 For numerical stability it is preferable that the C<at> value is near
1332 C<ev_now ()> (the current time), but there is no range requirement for
1333 this value.
1335 =item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1337 In this mode the values for C<interval> and C<at> are both being
1338 ignored. Instead, each time the periodic watcher gets scheduled, the
1339 reschedule callback will be called with the watcher as first, and the
1340 current time as second argument.
1342 NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
1343 ever, or make any event loop modifications>. If you need to stop it,
1344 return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
1345 starting an C<ev_prepare> watcher, which is legal).
1347 Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
1348 ev_tstamp now)>, e.g.:
1350 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
1351 {
1352 return now + 60.;
1353 }
1355 It must return the next time to trigger, based on the passed time value
1356 (that is, the lowest time value larger than to the second argument). It
1357 will usually be called just before the callback will be triggered, but
1358 might be called at other times, too.
1360 NOTE: I<< This callback must always return a time that is later than the
1361 passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.
1363 This can be used to create very complex timers, such as a timer that
1364 triggers on each midnight, local time. To do this, you would calculate the
1365 next midnight after C<now> and return the timestamp value for this. How
1366 you do this is, again, up to you (but it is not trivial, which is the main
1367 reason I omitted it as an example).
1369 =back
1371 =item ev_periodic_again (loop, ev_periodic *)
1373 Simply stops and restarts the periodic watcher again. This is only useful
1374 when you changed some parameters or the reschedule callback would return
1375 a different time than the last time it was called (e.g. in a crond like
1376 program when the crontabs have changed).
1378 =item ev_tstamp offset [read-write]
1380 When repeating, this contains the offset value, otherwise this is the
1381 absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1383 Can be modified any time, but changes only take effect when the periodic
1384 timer fires or C<ev_periodic_again> is being called.
1386 =item ev_tstamp interval [read-write]
1388 The current interval value. Can be modified any time, but changes only
1389 take effect when the periodic timer fires or C<ev_periodic_again> is being
1390 called.
1392 =item ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]
1394 The current reschedule callback, or C<0>, if this functionality is
1395 switched off. Can be changed any time, but changes only take effect when
1396 the periodic timer fires or C<ev_periodic_again> is being called.
1398 =item ev_tstamp at [read-only]
1400 When active, contains the absolute time that the watcher is supposed to
1401 trigger next.
1403 =back
1405 =head3 Examples
1407 Example: Call a callback every hour, or, more precisely, whenever the
1408 system clock is divisible by 3600. The callback invocation times have
1409 potentially a lot of jittering, but good long-term stability.
1411 static void
1412 clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1413 {
1414 ... its now a full hour (UTC, or TAI or whatever your clock follows)
1415 }
1417 struct ev_periodic hourly_tick;
1418 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1419 ev_periodic_start (loop, &hourly_tick);
1421 Example: The same as above, but use a reschedule callback to do it:
1423 #include <math.h>
1425 static ev_tstamp
1426 my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)
1427 {
1428 return fmod (now, 3600.) + 3600.;
1429 }
1431 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1433 Example: Call a callback every hour, starting now:
1435 struct ev_periodic hourly_tick;
1436 ev_periodic_init (&hourly_tick, clock_cb,
1437 fmod (ev_now (loop), 3600.), 3600., 0);
1438 ev_periodic_start (loop, &hourly_tick);
1441 =head2 C<ev_signal> - signal me when a signal gets signalled!
1443 Signal watchers will trigger an event when the process receives a specific
1444 signal one or more times. Even though signals are very asynchronous, libev
1445 will try it's best to deliver signals synchronously, i.e. as part of the
1446 normal event processing, like any other event.
1448 You can configure as many watchers as you like per signal. Only when the
1449 first watcher gets started will libev actually register a signal watcher
1450 with the kernel (thus it coexists with your own signal handlers as long
1451 as you don't register any with libev). Similarly, when the last signal
1452 watcher for a signal is stopped libev will reset the signal handler to
1453 SIG_DFL (regardless of what it was set to before).
1455 If possible and supported, libev will install its handlers with
1456 C<SA_RESTART> behaviour enabled, so syscalls should not be unduly
1457 interrupted. If you have a problem with syscalls getting interrupted by
1458 signals you can block all signals in an C<ev_check> watcher and unblock
1459 them in an C<ev_prepare> watcher.
1461 =head3 Watcher-Specific Functions and Data Members
1463 =over 4
1465 =item ev_signal_init (ev_signal *, callback, int signum)
1467 =item ev_signal_set (ev_signal *, int signum)
1469 Configures the watcher to trigger on the given signal number (usually one
1470 of the C<SIGxxx> constants).
1472 =item int signum [read-only]
1474 The signal the watcher watches out for.
1476 =back
1478 =head3 Examples
1480 Example: Try to exit cleanly on SIGINT and SIGTERM.
1482 static void
1483 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
1484 {
1485 ev_unloop (loop, EVUNLOOP_ALL);
1486 }
1488 struct ev_signal signal_watcher;
1489 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1490 ev_signal_start (loop, &sigint_cb);
1493 =head2 C<ev_child> - watch out for process status changes
1495 Child watchers trigger when your process receives a SIGCHLD in response to
1496 some child status changes (most typically when a child of yours dies). It
1497 is permissible to install a child watcher I<after> the child has been
1498 forked (which implies it might have already exited), as long as the event
1499 loop isn't entered (or is continued from a watcher).
1501 Only the default event loop is capable of handling signals, and therefore
1502 you can only rgeister child watchers in the default event loop.
1504 =head3 Process Interaction
1506 Libev grabs C<SIGCHLD> as soon as the default event loop is
1507 initialised. This is necessary to guarantee proper behaviour even if
1508 the first child watcher is started after the child exits. The occurance
1509 of C<SIGCHLD> is recorded asynchronously, but child reaping is done
1510 synchronously as part of the event loop processing. Libev always reaps all
1511 children, even ones not watched.
1513 =head3 Overriding the Built-In Processing
1515 Libev offers no special support for overriding the built-in child
1516 processing, but if your application collides with libev's default child
1517 handler, you can override it easily by installing your own handler for
1518 C<SIGCHLD> after initialising the default loop, and making sure the
1519 default loop never gets destroyed. You are encouraged, however, to use an
1520 event-based approach to child reaping and thus use libev's support for
1521 that, so other libev users can use C<ev_child> watchers freely.
1523 =head3 Watcher-Specific Functions and Data Members
1525 =over 4
1527 =item ev_child_init (ev_child *, callback, int pid, int trace)
1529 =item ev_child_set (ev_child *, int pid, int trace)
1531 Configures the watcher to wait for status changes of process C<pid> (or
1532 I<any> process if C<pid> is specified as C<0>). The callback can look
1533 at the C<rstatus> member of the C<ev_child> watcher structure to see
1534 the status word (use the macros from C<sys/wait.h> and see your systems
1535 C<waitpid> documentation). The C<rpid> member contains the pid of the
1536 process causing the status change. C<trace> must be either C<0> (only
1537 activate the watcher when the process terminates) or C<1> (additionally
1538 activate the watcher when the process is stopped or continued).
1540 =item int pid [read-only]
1542 The process id this watcher watches out for, or C<0>, meaning any process id.
1544 =item int rpid [read-write]
1546 The process id that detected a status change.
1548 =item int rstatus [read-write]
1550 The process exit/trace status caused by C<rpid> (see your systems
1551 C<waitpid> and C<sys/wait.h> documentation for details).
1553 =back
1555 =head3 Examples
1557 Example: C<fork()> a new process and install a child handler to wait for
1558 its completion.
1560 ev_child cw;
1562 static void
1563 child_cb (EV_P_ struct ev_child *w, int revents)
1564 {
1565 ev_child_stop (EV_A_ w);
1566 printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
1567 }
1569 pid_t pid = fork ();
1571 if (pid < 0)
1572 // error
1573 else if (pid == 0)
1574 {
1575 // the forked child executes here
1576 exit (1);
1577 }
1578 else
1579 {
1580 ev_child_init (&cw, child_cb, pid, 0);
1581 ev_child_start (EV_DEFAULT_ &cw);
1582 }
1585 =head2 C<ev_stat> - did the file attributes just change?
1587 This watches a filesystem path for attribute changes. That is, it calls
1588 C<stat> regularly (or when the OS says it changed) and sees if it changed
1589 compared to the last time, invoking the callback if it did.
1591 The path does not need to exist: changing from "path exists" to "path does
1592 not exist" is a status change like any other. The condition "path does
1593 not exist" is signified by the C<st_nlink> field being zero (which is
1594 otherwise always forced to be at least one) and all the other fields of
1595 the stat buffer having unspecified contents.
1597 The path I<should> be absolute and I<must not> end in a slash. If it is
1598 relative and your working directory changes, the behaviour is undefined.
1600 Since there is no standard to do this, the portable implementation simply
1601 calls C<stat (2)> regularly on the path to see if it changed somehow. You
1602 can specify a recommended polling interval for this case. If you specify
1603 a polling interval of C<0> (highly recommended!) then a I<suitable,
1604 unspecified default> value will be used (which you can expect to be around
1605 five seconds, although this might change dynamically). Libev will also
1606 impose a minimum interval which is currently around C<0.1>, but thats
1607 usually overkill.
1609 This watcher type is not meant for massive numbers of stat watchers,
1610 as even with OS-supported change notifications, this can be
1611 resource-intensive.
1613 At the time of this writing, only the Linux inotify interface is
1614 implemented (implementing kqueue support is left as an exercise for the
1615 reader). Inotify will be used to give hints only and should not change the
1616 semantics of C<ev_stat> watchers, which means that libev sometimes needs
1617 to fall back to regular polling again even with inotify, but changes are
1618 usually detected immediately, and if the file exists there will be no
1619 polling.
1621 =head3 ABI Issues (Largefile Support)
1623 Libev by default (unless the user overrides this) uses the default
1624 compilation environment, which means that on systems with optionally
1625 disabled large file support, you get the 32 bit version of the stat
1626 structure. When using the library from programs that change the ABI to
1627 use 64 bit file offsets the programs will fail. In that case you have to
1628 compile libev with the same flags to get binary compatibility. This is
1629 obviously the case with any flags that change the ABI, but the problem is
1630 most noticably with ev_stat and largefile support.
1632 =head3 Inotify
1634 When C<inotify (7)> support has been compiled into libev (generally only
1635 available on Linux) and present at runtime, it will be used to speed up
1636 change detection where possible. The inotify descriptor will be created lazily
1637 when the first C<ev_stat> watcher is being started.
1639 Inotify presense does not change the semantics of C<ev_stat> watchers
1640 except that changes might be detected earlier, and in some cases, to avoid
1641 making regular C<stat> calls. Even in the presense of inotify support
1642 there are many cases where libev has to resort to regular C<stat> polling.
1644 (There is no support for kqueue, as apparently it cannot be used to
1645 implement this functionality, due to the requirement of having a file
1646 descriptor open on the object at all times).
1648 =head3 The special problem of stat time resolution
1650 The C<stat ()> syscall only supports full-second resolution portably, and
1651 even on systems where the resolution is higher, many filesystems still
1652 only support whole seconds.
1654 That means that, if the time is the only thing that changes, you might
1655 miss updates: on the first update, C<ev_stat> detects a change and calls
1656 your callback, which does something. When there is another update within
1657 the same second, C<ev_stat> will be unable to detect it.
1659 The solution to this is to delay acting on a change for a second (or till
1660 the next second boundary), using a roughly one-second delay C<ev_timer>
1661 (C<ev_timer_set (w, 0., 1.01); ev_timer_again (loop, w)>). The C<.01>
1662 is added to work around small timing inconsistencies of some operating
1663 systems.
1665 =head3 Watcher-Specific Functions and Data Members
1667 =over 4
1669 =item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
1671 =item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
1673 Configures the watcher to wait for status changes of the given
1674 C<path>. The C<interval> is a hint on how quickly a change is expected to
1675 be detected and should normally be specified as C<0> to let libev choose
1676 a suitable value. The memory pointed to by C<path> must point to the same
1677 path for as long as the watcher is active.
1679 The callback will be receive C<EV_STAT> when a change was detected,
1680 relative to the attributes at the time the watcher was started (or the
1681 last change was detected).
1683 =item ev_stat_stat (loop, ev_stat *)
1685 Updates the stat buffer immediately with new values. If you change the
1686 watched path in your callback, you could call this fucntion to avoid
1687 detecting this change (while introducing a race condition). Can also be
1688 useful simply to find out the new values.
1690 =item ev_statdata attr [read-only]
1692 The most-recently detected attributes of the file. Although the type is of
1693 C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
1694 suitable for your system. If the C<st_nlink> member is C<0>, then there
1695 was some error while C<stat>ing the file.
1697 =item ev_statdata prev [read-only]
1699 The previous attributes of the file. The callback gets invoked whenever
1700 C<prev> != C<attr>.
1702 =item ev_tstamp interval [read-only]
1704 The specified interval.
1706 =item const char *path [read-only]
1708 The filesystem path that is being watched.
1710 =back
1712 =head3 Examples
1714 Example: Watch C</etc/passwd> for attribute changes.
1716 static void
1717 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
1718 {
1719 /* /etc/passwd changed in some way */
1720 if (w->attr.st_nlink)
1721 {
1722 printf ("passwd current size %ld\n", (long)w->attr.st_size);
1723 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
1724 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
1725 }
1726 else
1727 /* you shalt not abuse printf for puts */
1728 puts ("wow, /etc/passwd is not there, expect problems. "
1729 "if this is windows, they already arrived\n");
1730 }
1732 ...
1733 ev_stat passwd;
1735 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1736 ev_stat_start (loop, &passwd);
1738 Example: Like above, but additionally use a one-second delay so we do not
1739 miss updates (however, frequent updates will delay processing, too, so
1740 one might do the work both on C<ev_stat> callback invocation I<and> on
1741 C<ev_timer> callback invocation).
1743 static ev_stat passwd;
1744 static ev_timer timer;
1746 static void
1747 timer_cb (EV_P_ ev_timer *w, int revents)
1748 {
1749 ev_timer_stop (EV_A_ w);
1751 /* now it's one second after the most recent passwd change */
1752 }
1754 static void
1755 stat_cb (EV_P_ ev_stat *w, int revents)
1756 {
1757 /* reset the one-second timer */
1758 ev_timer_again (EV_A_ &timer);
1759 }
1761 ...
1762 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
1763 ev_stat_start (loop, &passwd);
1764 ev_timer_init (&timer, timer_cb, 0., 1.01);
1767 =head2 C<ev_idle> - when you've got nothing better to do...
1769 Idle watchers trigger events when no other events of the same or higher
1770 priority are pending (prepare, check and other idle watchers do not
1771 count).
1773 That is, as long as your process is busy handling sockets or timeouts
1774 (or even signals, imagine) of the same or higher priority it will not be
1775 triggered. But when your process is idle (or only lower-priority watchers
1776 are pending), the idle watchers are being called once per event loop
1777 iteration - until stopped, that is, or your process receives more events
1778 and becomes busy again with higher priority stuff.
1780 The most noteworthy effect is that as long as any idle watchers are
1781 active, the process will not block when waiting for new events.
1783 Apart from keeping your process non-blocking (which is a useful
1784 effect on its own sometimes), idle watchers are a good place to do
1785 "pseudo-background processing", or delay processing stuff to after the
1786 event loop has handled all outstanding events.
1788 =head3 Watcher-Specific Functions and Data Members
1790 =over 4
1792 =item ev_idle_init (ev_signal *, callback)
1794 Initialises and configures the idle watcher - it has no parameters of any
1795 kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1796 believe me.
1798 =back
1800 =head3 Examples
1802 Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1803 callback, free it. Also, use no error checking, as usual.
1805 static void
1806 idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)
1807 {
1808 free (w);
1809 // now do something you wanted to do when the program has
1810 // no longer anything immediate to do.
1811 }
1813 struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));
1814 ev_idle_init (idle_watcher, idle_cb);
1815 ev_idle_start (loop, idle_cb);
1818 =head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1820 Prepare and check watchers are usually (but not always) used in tandem:
1821 prepare watchers get invoked before the process blocks and check watchers
1822 afterwards.
1824 You I<must not> call C<ev_loop> or similar functions that enter
1825 the current event loop from either C<ev_prepare> or C<ev_check>
1826 watchers. Other loops than the current one are fine, however. The
1827 rationale behind this is that you do not need to check for recursion in
1828 those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
1829 C<ev_check> so if you have one watcher of each kind they will always be
1830 called in pairs bracketing the blocking call.
1832 Their main purpose is to integrate other event mechanisms into libev and
1833 their use is somewhat advanced. This could be used, for example, to track
1834 variable changes, implement your own watchers, integrate net-snmp or a
1835 coroutine library and lots more. They are also occasionally useful if
1836 you cache some data and want to flush it before blocking (for example,
1837 in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
1838 watcher).
1840 This is done by examining in each prepare call which file descriptors need
1841 to be watched by the other library, registering C<ev_io> watchers for
1842 them and starting an C<ev_timer> watcher for any timeouts (many libraries
1843 provide just this functionality). Then, in the check watcher you check for
1844 any events that occured (by checking the pending status of all watchers
1845 and stopping them) and call back into the library. The I/O and timer
1846 callbacks will never actually be called (but must be valid nevertheless,
1847 because you never know, you know?).
1849 As another example, the Perl Coro module uses these hooks to integrate
1850 coroutines into libev programs, by yielding to other active coroutines
1851 during each prepare and only letting the process block if no coroutines
1852 are ready to run (it's actually more complicated: it only runs coroutines
1853 with priority higher than or equal to the event loop and one coroutine
1854 of lower priority, but only once, using idle watchers to keep the event
1855 loop from blocking if lower-priority coroutines are active, thus mapping
1856 low-priority coroutines to idle/background tasks).
1858 It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1859 priority, to ensure that they are being run before any other watchers
1860 after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
1861 too) should not activate ("feed") events into libev. While libev fully
1862 supports this, they will be called before other C<ev_check> watchers
1863 did their job. As C<ev_check> watchers are often used to embed other
1864 (non-libev) event loops those other event loops might be in an unusable
1865 state until their C<ev_check> watcher ran (always remind yourself to
1866 coexist peacefully with others).
1868 =head3 Watcher-Specific Functions and Data Members
1870 =over 4
1872 =item ev_prepare_init (ev_prepare *, callback)
1874 =item ev_check_init (ev_check *, callback)
1876 Initialises and configures the prepare or check watcher - they have no
1877 parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1878 macros, but using them is utterly, utterly and completely pointless.
1880 =back
1882 =head3 Examples
1884 There are a number of principal ways to embed other event loops or modules
1885 into libev. Here are some ideas on how to include libadns into libev
1886 (there is a Perl module named C<EV::ADNS> that does this, which you could
1887 use for an actually working example. Another Perl module named C<EV::Glib>
1888 embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV
1889 into the Glib event loop).
1891 Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1892 and in a check watcher, destroy them and call into libadns. What follows
1893 is pseudo-code only of course. This requires you to either use a low
1894 priority for the check watcher or use C<ev_clear_pending> explicitly, as
1895 the callbacks for the IO/timeout watchers might not have been called yet.
1897 static ev_io iow [nfd];
1898 static ev_timer tw;
1900 static void
1901 io_cb (ev_loop *loop, ev_io *w, int revents)
1902 {
1903 }
1905 // create io watchers for each fd and a timer before blocking
1906 static void
1907 adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
1908 {
1909 int timeout = 3600000;
1910 struct pollfd fds [nfd];
1911 // actual code will need to loop here and realloc etc.
1912 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1914 /* the callback is illegal, but won't be called as we stop during check */
1915 ev_timer_init (&tw, 0, timeout * 1e-3);
1916 ev_timer_start (loop, &tw);
1918 // create one ev_io per pollfd
1919 for (int i = 0; i < nfd; ++i)
1920 {
1921 ev_io_init (iow + i, io_cb, fds [i].fd,
1922 ((fds [i].events & POLLIN ? EV_READ : 0)
1923 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1925 fds [i].revents = 0;
1926 ev_io_start (loop, iow + i);
1927 }
1928 }
1930 // stop all watchers after blocking
1931 static void
1932 adns_check_cb (ev_loop *loop, ev_check *w, int revents)
1933 {
1934 ev_timer_stop (loop, &tw);
1936 for (int i = 0; i < nfd; ++i)
1937 {
1938 // set the relevant poll flags
1939 // could also call adns_processreadable etc. here
1940 struct pollfd *fd = fds + i;
1941 int revents = ev_clear_pending (iow + i);
1942 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1943 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1945 // now stop the watcher
1946 ev_io_stop (loop, iow + i);
1947 }
1949 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1950 }
1952 Method 2: This would be just like method 1, but you run C<adns_afterpoll>
1953 in the prepare watcher and would dispose of the check watcher.
1955 Method 3: If the module to be embedded supports explicit event
1956 notification (adns does), you can also make use of the actual watcher
1957 callbacks, and only destroy/create the watchers in the prepare watcher.
1959 static void
1960 timer_cb (EV_P_ ev_timer *w, int revents)
1961 {
1962 adns_state ads = (adns_state)w->data;
1963 update_now (EV_A);
1965 adns_processtimeouts (ads, &tv_now);
1966 }
1968 static void
1969 io_cb (EV_P_ ev_io *w, int revents)
1970 {
1971 adns_state ads = (adns_state)w->data;
1972 update_now (EV_A);
1974 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1975 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1976 }
1978 // do not ever call adns_afterpoll
1980 Method 4: Do not use a prepare or check watcher because the module you
1981 want to embed is too inflexible to support it. Instead, youc na override
1982 their poll function. The drawback with this solution is that the main
1983 loop is now no longer controllable by EV. The C<Glib::EV> module does
1984 this.
1986 static gint
1987 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1988 {
1989 int got_events = 0;
1991 for (n = 0; n < nfds; ++n)
1992 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1994 if (timeout >= 0)
1995 // create/start timer
1997 // poll
1998 ev_loop (EV_A_ 0);
2000 // stop timer again
2001 if (timeout >= 0)
2002 ev_timer_stop (EV_A_ &to);
2004 // stop io watchers again - their callbacks should have set
2005 for (n = 0; n < nfds; ++n)
2006 ev_io_stop (EV_A_ iow [n]);
2008 return got_events;
2009 }
2012 =head2 C<ev_embed> - when one backend isn't enough...
2014 This is a rather advanced watcher type that lets you embed one event loop
2015 into another (currently only C<ev_io> events are supported in the embedded
2016 loop, other types of watchers might be handled in a delayed or incorrect
2017 fashion and must not be used).
2019 There are primarily two reasons you would want that: work around bugs and
2020 prioritise I/O.
2022 As an example for a bug workaround, the kqueue backend might only support
2023 sockets on some platform, so it is unusable as generic backend, but you
2024 still want to make use of it because you have many sockets and it scales
2025 so nicely. In this case, you would create a kqueue-based loop and embed it
2026 into your default loop (which might use e.g. poll). Overall operation will
2027 be a bit slower because first libev has to poll and then call kevent, but
2028 at least you can use both at what they are best.
2030 As for prioritising I/O: rarely you have the case where some fds have
2031 to be watched and handled very quickly (with low latency), and even
2032 priorities and idle watchers might have too much overhead. In this case
2033 you would put all the high priority stuff in one loop and all the rest in
2034 a second one, and embed the second one in the first.
2036 As long as the watcher is active, the callback will be invoked every time
2037 there might be events pending in the embedded loop. The callback must then
2038 call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
2039 their callbacks (you could also start an idle watcher to give the embedded
2040 loop strictly lower priority for example). You can also set the callback
2041 to C<0>, in which case the embed watcher will automatically execute the
2042 embedded loop sweep.
2044 As long as the watcher is started it will automatically handle events. The
2045 callback will be invoked whenever some events have been handled. You can
2046 set the callback to C<0> to avoid having to specify one if you are not
2047 interested in that.
2049 Also, there have not currently been made special provisions for forking:
2050 when you fork, you not only have to call C<ev_loop_fork> on both loops,
2051 but you will also have to stop and restart any C<ev_embed> watchers
2052 yourself.
2054 Unfortunately, not all backends are embeddable, only the ones returned by
2055 C<ev_embeddable_backends> are, which, unfortunately, does not include any
2056 portable one.
2058 So when you want to use this feature you will always have to be prepared
2059 that you cannot get an embeddable loop. The recommended way to get around
2060 this is to have a separate variables for your embeddable loop, try to
2061 create it, and if that fails, use the normal loop for everything.
2063 =head3 Watcher-Specific Functions and Data Members
2065 =over 4
2067 =item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
2069 =item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
2071 Configures the watcher to embed the given loop, which must be
2072 embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
2073 invoked automatically, otherwise it is the responsibility of the callback
2074 to invoke it (it will continue to be called until the sweep has been done,
2075 if you do not want thta, you need to temporarily stop the embed watcher).
2077 =item ev_embed_sweep (loop, ev_embed *)
2079 Make a single, non-blocking sweep over the embedded loop. This works
2080 similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
2081 apropriate way for embedded loops.
2083 =item struct ev_loop *other [read-only]
2085 The embedded event loop.
2087 =back
2089 =head3 Examples
2091 Example: Try to get an embeddable event loop and embed it into the default
2092 event loop. If that is not possible, use the default loop. The default
2093 loop is stored in C<loop_hi>, while the mebeddable loop is stored in
2094 C<loop_lo> (which is C<loop_hi> in the acse no embeddable loop can be
2095 used).
2097 struct ev_loop *loop_hi = ev_default_init (0);
2098 struct ev_loop *loop_lo = 0;
2099 struct ev_embed embed;
2101 // see if there is a chance of getting one that works
2102 // (remember that a flags value of 0 means autodetection)
2103 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
2104 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
2105 : 0;
2107 // if we got one, then embed it, otherwise default to loop_hi
2108 if (loop_lo)
2109 {
2110 ev_embed_init (&embed, 0, loop_lo);
2111 ev_embed_start (loop_hi, &embed);
2112 }
2113 else
2114 loop_lo = loop_hi;
2116 Example: Check if kqueue is available but not recommended and create
2117 a kqueue backend for use with sockets (which usually work with any
2118 kqueue implementation). Store the kqueue/socket-only event loop in
2119 C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
2121 struct ev_loop *loop = ev_default_init (0);
2122 struct ev_loop *loop_socket = 0;
2123 struct ev_embed embed;
2125 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2126 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2127 {
2128 ev_embed_init (&embed, 0, loop_socket);
2129 ev_embed_start (loop, &embed);
2130 }
2132 if (!loop_socket)
2133 loop_socket = loop;
2135 // now use loop_socket for all sockets, and loop for everything else
2138 =head2 C<ev_fork> - the audacity to resume the event loop after a fork
2140 Fork watchers are called when a C<fork ()> was detected (usually because
2141 whoever is a good citizen cared to tell libev about it by calling
2142 C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
2143 event loop blocks next and before C<ev_check> watchers are being called,
2144 and only in the child after the fork. If whoever good citizen calling
2145 C<ev_default_fork> cheats and calls it in the wrong process, the fork
2146 handlers will be invoked, too, of course.
2148 =head3 Watcher-Specific Functions and Data Members
2150 =over 4
2152 =item ev_fork_init (ev_signal *, callback)
2154 Initialises and configures the fork watcher - it has no parameters of any
2155 kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
2156 believe me.
2158 =back
2161 =head2 C<ev_async> - how to wake up another event loop
2163 In general, you cannot use an C<ev_loop> from multiple threads or other
2164 asynchronous sources such as signal handlers (as opposed to multiple event
2165 loops - those are of course safe to use in different threads).
2167 Sometimes, however, you need to wake up another event loop you do not
2168 control, for example because it belongs to another thread. This is what
2169 C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
2170 can signal it by calling C<ev_async_send>, which is thread- and signal
2171 safe.
2173 This functionality is very similar to C<ev_signal> watchers, as signals,
2174 too, are asynchronous in nature, and signals, too, will be compressed
2175 (i.e. the number of callback invocations may be less than the number of
2176 C<ev_async_sent> calls).
2178 Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
2179 just the default loop.
2181 =head3 Queueing
2183 C<ev_async> does not support queueing of data in any way. The reason
2184 is that the author does not know of a simple (or any) algorithm for a
2185 multiple-writer-single-reader queue that works in all cases and doesn't
2186 need elaborate support such as pthreads.
2188 That means that if you want to queue data, you have to provide your own
2189 queue. But at least I can tell you would implement locking around your
2190 queue:
2192 =over 4
2194 =item queueing from a signal handler context
2196 To implement race-free queueing, you simply add to the queue in the signal
2197 handler but you block the signal handler in the watcher callback. Here is an example that does that for
2198 some fictitiuous SIGUSR1 handler:
2200 static ev_async mysig;
2202 static void
2203 sigusr1_handler (void)
2204 {
2205 sometype data;
2207 // no locking etc.
2208 queue_put (data);
2209 ev_async_send (EV_DEFAULT_ &mysig);
2210 }
2212 static void
2213 mysig_cb (EV_P_ ev_async *w, int revents)
2214 {
2215 sometype data;
2216 sigset_t block, prev;
2218 sigemptyset (&block);
2219 sigaddset (&block, SIGUSR1);
2220 sigprocmask (SIG_BLOCK, &block, &prev);
2222 while (queue_get (&data))
2223 process (data);
2225 if (sigismember (&prev, SIGUSR1)
2226 sigprocmask (SIG_UNBLOCK, &block, 0);
2227 }
2229 (Note: pthreads in theory requires you to use C<pthread_setmask>
2230 instead of C<sigprocmask> when you use threads, but libev doesn't do it
2231 either...).
2233 =item queueing from a thread context
2235 The strategy for threads is different, as you cannot (easily) block
2236 threads but you can easily preempt them, so to queue safely you need to
2237 employ a traditional mutex lock, such as in this pthread example:
2239 static ev_async mysig;
2240 static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
2242 static void
2243 otherthread (void)
2244 {
2245 // only need to lock the actual queueing operation
2246 pthread_mutex_lock (&mymutex);
2247 queue_put (data);
2248 pthread_mutex_unlock (&mymutex);
2250 ev_async_send (EV_DEFAULT_ &mysig);
2251 }
2253 static void
2254 mysig_cb (EV_P_ ev_async *w, int revents)
2255 {
2256 pthread_mutex_lock (&mymutex);
2258 while (queue_get (&data))
2259 process (data);
2261 pthread_mutex_unlock (&mymutex);
2262 }
2264 =back
2267 =head3 Watcher-Specific Functions and Data Members
2269 =over 4
2271 =item ev_async_init (ev_async *, callback)
2273 Initialises and configures the async watcher - it has no parameters of any
2274 kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
2275 believe me.
2277 =item ev_async_send (loop, ev_async *)
2279 Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2280 an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2281 C<ev_feed_event>, this call is safe to do in other threads, signal or
2282 similar contexts (see the dicusssion of C<EV_ATOMIC_T> in the embedding
2283 section below on what exactly this means).
2285 This call incurs the overhead of a syscall only once per loop iteration,
2286 so while the overhead might be noticable, it doesn't apply to repeated
2287 calls to C<ev_async_send>.
2289 =item bool = ev_async_pending (ev_async *)
2291 Returns a non-zero value when C<ev_async_send> has been called on the
2292 watcher but the event has not yet been processed (or even noted) by the
2293 event loop.
2295 C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
2296 the loop iterates next and checks for the watcher to have become active,
2297 it will reset the flag again. C<ev_async_pending> can be used to very
2298 quickly check wether invoking the loop might be a good idea.
2300 Not that this does I<not> check wether the watcher itself is pending, only
2301 wether it has been requested to make this watcher pending.
2303 =back
2308 There are some other functions of possible interest. Described. Here. Now.
2310 =over 4
2312 =item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
2314 This function combines a simple timer and an I/O watcher, calls your
2315 callback on whichever event happens first and automatically stop both
2316 watchers. This is useful if you want to wait for a single event on an fd
2317 or timeout without having to allocate/configure/start/stop/free one or
2318 more watchers yourself.
2320 If C<fd> is less than 0, then no I/O watcher will be started and events
2321 is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and
2322 C<events> set will be craeted and started.
2324 If C<timeout> is less than 0, then no timeout watcher will be
2325 started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
2326 repeat = 0) will be started. While C<0> is a valid timeout, it is of
2327 dubious value.
2329 The callback has the type C<void (*cb)(int revents, void *arg)> and gets
2330 passed an C<revents> set like normal event callbacks (a combination of
2331 C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
2332 value passed to C<ev_once>:
2334 static void stdin_ready (int revents, void *arg)
2335 {
2336 if (revents & EV_TIMEOUT)
2337 /* doh, nothing entered */;
2338 else if (revents & EV_READ)
2339 /* stdin might have data for us, joy! */;
2340 }
2342 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2344 =item ev_feed_event (ev_loop *, watcher *, int revents)
2346 Feeds the given event set into the event loop, as if the specified event
2347 had happened for the specified watcher (which must be a pointer to an
2348 initialised but not necessarily started event watcher).
2350 =item ev_feed_fd_event (ev_loop *, int fd, int revents)
2352 Feed an event on the given fd, as if a file descriptor backend detected
2353 the given events it.
2355 =item ev_feed_signal_event (ev_loop *loop, int signum)
2357 Feed an event as if the given signal occured (C<loop> must be the default
2358 loop!).
2360 =back
2365 Libev offers a compatibility emulation layer for libevent. It cannot
2366 emulate the internals of libevent, so here are some usage hints:
2368 =over 4
2370 =item * Use it by including <event.h>, as usual.
2372 =item * The following members are fully supported: ev_base, ev_callback,
2373 ev_arg, ev_fd, ev_res, ev_events.
2375 =item * Avoid using ev_flags and the EVLIST_*-macros, while it is
2376 maintained by libev, it does not work exactly the same way as in libevent (consider
2377 it a private API).
2379 =item * Priorities are not currently supported. Initialising priorities
2380 will fail and all watchers will have the same priority, even though there
2381 is an ev_pri field.
2383 =item * In libevent, the last base created gets the signals, in libev, the
2384 first base created (== the default loop) gets the signals.
2386 =item * Other members are not supported.
2388 =item * The libev emulation is I<not> ABI compatible to libevent, you need
2389 to use the libev header file and library.
2391 =back
2393 =head1 C++ SUPPORT
2395 Libev comes with some simplistic wrapper classes for C++ that mainly allow
2396 you to use some convinience methods to start/stop watchers and also change
2397 the callback model to a model using method callbacks on objects.
2399 To use it,
2401 #include <ev++.h>
2403 This automatically includes F<ev.h> and puts all of its definitions (many
2404 of them macros) into the global namespace. All C++ specific things are
2405 put into the C<ev> namespace. It should support all the same embedding
2406 options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
2408 Care has been taken to keep the overhead low. The only data member the C++
2409 classes add (compared to plain C-style watchers) is the event loop pointer
2410 that the watcher is associated with (or no additional members at all if
2411 you disable C<EV_MULTIPLICITY> when embedding libev).
2413 Currently, functions, and static and non-static member functions can be
2414 used as callbacks. Other types should be easy to add as long as they only
2415 need one additional pointer for context. If you need support for other
2416 types of functors please contact the author (preferably after implementing
2417 it).
2419 Here is a list of things available in the C<ev> namespace:
2421 =over 4
2423 =item C<ev::READ>, C<ev::WRITE> etc.
2425 These are just enum values with the same values as the C<EV_READ> etc.
2426 macros from F<ev.h>.
2428 =item C<ev::tstamp>, C<ev::now>
2430 Aliases to the same types/functions as with the C<ev_> prefix.
2432 =item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
2434 For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
2435 the same name in the C<ev> namespace, with the exception of C<ev_signal>
2436 which is called C<ev::sig> to avoid clashes with the C<signal> macro
2437 defines by many implementations.
2439 All of those classes have these methods:
2441 =over 4
2443 =item ev::TYPE::TYPE ()
2445 =item ev::TYPE::TYPE (struct ev_loop *)
2447 =item ev::TYPE::~TYPE
2449 The constructor (optionally) takes an event loop to associate the watcher
2450 with. If it is omitted, it will use C<EV_DEFAULT>.
2452 The constructor calls C<ev_init> for you, which means you have to call the
2453 C<set> method before starting it.
2455 It will not set a callback, however: You have to call the templated C<set>
2456 method to set a callback before you can start the watcher.
2458 (The reason why you have to use a method is a limitation in C++ which does
2459 not allow explicit template arguments for constructors).
2461 The destructor automatically stops the watcher if it is active.
2463 =item w->set<class, &class::method> (object *)
2465 This method sets the callback method to call. The method has to have a
2466 signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
2467 first argument and the C<revents> as second. The object must be given as
2468 parameter and is stored in the C<data> member of the watcher.
2470 This method synthesizes efficient thunking code to call your method from
2471 the C callback that libev requires. If your compiler can inline your
2472 callback (i.e. it is visible to it at the place of the C<set> call and
2473 your compiler is good :), then the method will be fully inlined into the
2474 thunking function, making it as fast as a direct C callback.
2476 Example: simple class declaration and watcher initialisation
2478 struct myclass
2479 {
2480 void io_cb (ev::io &w, int revents) { }
2481 }
2483 myclass obj;
2484 ev::io iow;
2485 iow.set <myclass, &myclass::io_cb> (&obj);
2487 =item w->set<function> (void *data = 0)
2489 Also sets a callback, but uses a static method or plain function as
2490 callback. The optional C<data> argument will be stored in the watcher's
2491 C<data> member and is free for you to use.
2493 The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2495 See the method-C<set> above for more details.
2497 Example:
2499 static void io_cb (ev::io &w, int revents) { }
2500 iow.set <io_cb> ();
2502 =item w->set (struct ev_loop *)
2504 Associates a different C<struct ev_loop> with this watcher. You can only
2505 do this when the watcher is inactive (and not pending either).
2507 =item w->set ([args])
2509 Basically the same as C<ev_TYPE_set>, with the same args. Must be
2510 called at least once. Unlike the C counterpart, an active watcher gets
2511 automatically stopped and restarted when reconfiguring it with this
2512 method.
2514 =item w->start ()
2516 Starts the watcher. Note that there is no C<loop> argument, as the
2517 constructor already stores the event loop.
2519 =item w->stop ()
2521 Stops the watcher if it is active. Again, no C<loop> argument.
2523 =item w->again () (C<ev::timer>, C<ev::periodic> only)
2525 For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
2526 C<ev_TYPE_again> function.
2528 =item w->sweep () (C<ev::embed> only)
2530 Invokes C<ev_embed_sweep>.
2532 =item w->update () (C<ev::stat> only)
2534 Invokes C<ev_stat_stat>.
2536 =back
2538 =back
2540 Example: Define a class with an IO and idle watcher, start one of them in
2541 the constructor.
2543 class myclass
2544 {
2545 ev::io io; void io_cb (ev::io &w, int revents);
2546 ev:idle idle void idle_cb (ev::idle &w, int revents);
2548 myclass (int fd)
2549 {
2550 io .set <myclass, &myclass::io_cb > (this);
2551 idle.set <myclass, &myclass::idle_cb> (this);
2553 io.start (fd, ev::READ);
2554 }
2555 };
2560 Libev does not offer other language bindings itself, but bindings for a
2561 numbe rof languages exist in the form of third-party packages. If you know
2562 any interesting language binding in addition to the ones listed here, drop
2563 me a note.
2565 =over 4
2567 =item Perl
2569 The EV module implements the full libev API and is actually used to test
2570 libev. EV is developed together with libev. Apart from the EV core module,
2571 there are additional modules that implement libev-compatible interfaces
2572 to C<libadns> (C<EV::ADNS>), C<Net::SNMP> (C<Net::SNMP::EV>) and the
2573 C<libglib> event core (C<Glib::EV> and C<EV::Glib>).
2575 It can be found and installed via CPAN, its homepage is found at
2576 L<>.
2578 =item Ruby
2580 Tony Arcieri has written a ruby extension that offers access to a subset
2581 of the libev API and adds filehandle abstractions, asynchronous DNS and
2582 more on top of it. It can be found via gem servers. Its homepage is at
2583 L<>.
2585 =item D
2587 Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
2588 be found at L<;a=summary>.
2590 =back
2593 =head1 MACRO MAGIC
2595 Libev can be compiled with a variety of options, the most fundamantal
2596 of which is C<EV_MULTIPLICITY>. This option determines whether (most)
2597 functions and callbacks have an initial C<struct ev_loop *> argument.
2599 To make it easier to write programs that cope with either variant, the
2600 following macros are defined:
2602 =over 4
2604 =item C<EV_A>, C<EV_A_>
2606 This provides the loop I<argument> for functions, if one is required ("ev
2607 loop argument"). The C<EV_A> form is used when this is the sole argument,
2608 C<EV_A_> is used when other arguments are following. Example:
2610 ev_unref (EV_A);
2611 ev_timer_add (EV_A_ watcher);
2612 ev_loop (EV_A_ 0);
2614 It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
2615 which is often provided by the following macro.
2617 =item C<EV_P>, C<EV_P_>
2619 This provides the loop I<parameter> for functions, if one is required ("ev
2620 loop parameter"). The C<EV_P> form is used when this is the sole parameter,
2621 C<EV_P_> is used when other parameters are following. Example:
2623 // this is how ev_unref is being declared
2624 static void ev_unref (EV_P);
2626 // this is how you can declare your typical callback
2627 static void cb (EV_P_ ev_timer *w, int revents)
2629 It declares a parameter C<loop> of type C<struct ev_loop *>, quite
2630 suitable for use with C<EV_A>.
2632 =item C<EV_DEFAULT>, C<EV_DEFAULT_>
2634 Similar to the other two macros, this gives you the value of the default
2635 loop, if multiple loops are supported ("ev loop default").
2639 Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
2640 default loop has been initialised (C<UC> == unchecked). Their behaviour
2641 is undefined when the default loop has not been initialised by a previous
2642 execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
2644 It is often prudent to use C<EV_DEFAULT> when initialising the first
2645 watcher in a function but use C<EV_DEFAULT_UC> afterwards.
2647 =back
2649 Example: Declare and initialise a check watcher, utilising the above
2650 macros so it will work regardless of whether multiple loops are supported
2651 or not.
2653 static void
2654 check_cb (EV_P_ ev_timer *w, int revents)
2655 {
2656 ev_check_stop (EV_A_ w);
2657 }
2659 ev_check check;
2660 ev_check_init (&check, check_cb);
2661 ev_check_start (EV_DEFAULT_ &check);
2662 ev_loop (EV_DEFAULT_ 0);
2664 =head1 EMBEDDING
2666 Libev can (and often is) directly embedded into host
2667 applications. Examples of applications that embed it include the Deliantra
2668 Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
2669 and rxvt-unicode.
2671 The goal is to enable you to just copy the necessary files into your
2672 source directory without having to change even a single line in them, so
2673 you can easily upgrade by simply copying (or having a checked-out copy of
2674 libev somewhere in your source tree).
2676 =head2 FILESETS
2678 Depending on what features you need you need to include one or more sets of files
2679 in your app.
2681 =head3 CORE EVENT LOOP
2683 To include only the libev core (all the C<ev_*> functions), with manual
2684 configuration (no autoconf):
2686 #define EV_STANDALONE 1
2687 #include "ev.c"
2689 This will automatically include F<ev.h>, too, and should be done in a
2690 single C source file only to provide the function implementations. To use
2691 it, do the same for F<ev.h> in all files wishing to use this API (best
2692 done by writing a wrapper around F<ev.h> that you can include instead and
2693 where you can put other configuration options):
2695 #define EV_STANDALONE 1
2696 #include "ev.h"
2698 Both header files and implementation files can be compiled with a C++
2699 compiler (at least, thats a stated goal, and breakage will be treated
2700 as a bug).
2702 You need the following files in your source tree, or in a directory
2703 in your include path (e.g. in libev/ when using -Ilibev):
2705 ev.h
2706 ev.c
2707 ev_vars.h
2708 ev_wrap.h
2710 ev_win32.c required on win32 platforms only
2712 ev_select.c only when select backend is enabled (which is enabled by default)
2713 ev_poll.c only when poll backend is enabled (disabled by default)
2714 ev_epoll.c only when the epoll backend is enabled (disabled by default)
2715 ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
2716 ev_port.c only when the solaris port backend is enabled (disabled by default)
2718 F<ev.c> includes the backend files directly when enabled, so you only need
2719 to compile this single file.
2723 To include the libevent compatibility API, also include:
2725 #include "event.c"
2727 in the file including F<ev.c>, and:
2729 #include "event.h"
2731 in the files that want to use the libevent API. This also includes F<ev.h>.
2733 You need the following additional files for this:
2735 event.h
2736 event.c
2740 Instead of using C<EV_STANDALONE=1> and providing your config in
2741 whatever way you want, you can also C<m4_include([libev.m4])> in your
2742 F<> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
2743 include F<config.h> and configure itself accordingly.
2745 For this of course you need the m4 file:
2747 libev.m4
2751 Libev can be configured via a variety of preprocessor symbols you have to
2752 define before including any of its files. The default in the absense of
2753 autoconf is noted for every option.
2755 =over 4
2757 =item EV_STANDALONE
2759 Must always be C<1> if you do not use autoconf configuration, which
2760 keeps libev from including F<config.h>, and it also defines dummy
2761 implementations for some libevent functions (such as logging, which is not
2762 supported). It will also not define any of the structs usually found in
2763 F<event.h> that are not directly supported by the libev core alone.
2767 If defined to be C<1>, libev will try to detect the availability of the
2768 monotonic clock option at both compiletime and runtime. Otherwise no use
2769 of the monotonic clock option will be attempted. If you enable this, you
2770 usually have to link against librt or something similar. Enabling it when
2771 the functionality isn't available is safe, though, although you have
2772 to make sure you link against any libraries where the C<clock_gettime>
2773 function is hiding in (often F<-lrt>).
2775 =item EV_USE_REALTIME
2777 If defined to be C<1>, libev will try to detect the availability of the
2778 realtime clock option at compiletime (and assume its availability at
2779 runtime if successful). Otherwise no use of the realtime clock option will
2780 be attempted. This effectively replaces C<gettimeofday> by C<clock_get
2781 (CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
2782 note about libraries in the description of C<EV_USE_MONOTONIC>, though.
2786 If defined to be C<1>, libev will assume that C<nanosleep ()> is available
2787 and will use it for delays. Otherwise it will use C<select ()>.
2789 =item EV_USE_EVENTFD
2791 If defined to be C<1>, then libev will assume that C<eventfd ()> is
2792 available and will probe for kernel support at runtime. This will improve
2793 C<ev_signal> and C<ev_async> performance and reduce resource consumption.
2794 If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
2795 2.7 or newer, otherwise disabled.
2797 =item EV_USE_SELECT
2799 If undefined or defined to be C<1>, libev will compile in support for the
2800 C<select>(2) backend. No attempt at autodetection will be done: if no
2801 other method takes over, select will be it. Otherwise the select backend
2802 will not be compiled in.
2806 If defined to C<1>, then the select backend will use the system C<fd_set>
2807 structure. This is useful if libev doesn't compile due to a missing
2808 C<NFDBITS> or C<fd_mask> definition or it misguesses the bitset layout on
2809 exotic systems. This usually limits the range of file descriptors to some
2810 low limit such as 1024 or might have other limitations (winsocket only
2811 allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
2812 influence the size of the C<fd_set> used.
2816 When defined to C<1>, the select backend will assume that
2817 select/socket/connect etc. don't understand file descriptors but
2818 wants osf handles on win32 (this is the case when the select to
2819 be used is the winsock select). This means that it will call
2820 C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2821 it is assumed that all these functions actually work on fds, even
2822 on win32. Should not be defined on non-win32 platforms.
2824 =item EV_FD_TO_WIN32_HANDLE
2826 If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
2827 file descriptors to socket handles. When not defining this symbol (the
2828 default), then libev will call C<_get_osfhandle>, which is usually
2829 correct. In some cases, programs use their own file descriptor management,
2830 in which case they can provide this function to map fds to socket handles.
2832 =item EV_USE_POLL
2834 If defined to be C<1>, libev will compile in support for the C<poll>(2)
2835 backend. Otherwise it will be enabled on non-win32 platforms. It
2836 takes precedence over select.
2838 =item EV_USE_EPOLL
2840 If defined to be C<1>, libev will compile in support for the Linux
2841 C<epoll>(7) backend. Its availability will be detected at runtime,
2842 otherwise another method will be used as fallback. This is the preferred
2843 backend for GNU/Linux systems. If undefined, it will be enabled if the
2844 headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2846 =item EV_USE_KQUEUE
2848 If defined to be C<1>, libev will compile in support for the BSD style
2849 C<kqueue>(2) backend. Its actual availability will be detected at runtime,
2850 otherwise another method will be used as fallback. This is the preferred
2851 backend for BSD and BSD-like systems, although on most BSDs kqueue only
2852 supports some types of fds correctly (the only platform we found that
2853 supports ptys for example was NetBSD), so kqueue might be compiled in, but
2854 not be used unless explicitly requested. The best way to use it is to find
2855 out whether kqueue supports your type of fd properly and use an embedded
2856 kqueue loop.
2858 =item EV_USE_PORT
2860 If defined to be C<1>, libev will compile in support for the Solaris
2861 10 port style backend. Its availability will be detected at runtime,
2862 otherwise another method will be used as fallback. This is the preferred
2863 backend for Solaris 10 systems.
2865 =item EV_USE_DEVPOLL
2867 reserved for future expansion, works like the USE symbols above.
2869 =item EV_USE_INOTIFY
2871 If defined to be C<1>, libev will compile in support for the Linux inotify
2872 interface to speed up C<ev_stat> watchers. Its actual availability will
2873 be detected at runtime. If undefined, it will be enabled if the headers
2874 indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2876 =item EV_ATOMIC_T
2878 Libev requires an integer type (suitable for storing C<0> or C<1>) whose
2879 access is atomic with respect to other threads or signal contexts. No such
2880 type is easily found in the C language, so you can provide your own type
2881 that you know is safe for your purposes. It is used both for signal handler "locking"
2882 as well as for signal and thread safety in C<ev_async> watchers.
2884 In the absense of this define, libev will use C<sig_atomic_t volatile>
2885 (from F<signal.h>), which is usually good enough on most platforms.
2887 =item EV_H
2889 The name of the F<ev.h> header file used to include it. The default if
2890 undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2891 used to virtually rename the F<ev.h> header file in case of conflicts.
2893 =item EV_CONFIG_H
2895 If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2896 F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2897 C<EV_H>, above.
2899 =item EV_EVENT_H
2901 Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2902 of how the F<event.h> header can be found, the default is C<"event.h">.
2904 =item EV_PROTOTYPES
2906 If defined to be C<0>, then F<ev.h> will not define any function
2907 prototypes, but still define all the structs and other symbols. This is
2908 occasionally useful if you want to provide your own wrapper functions
2909 around libev functions.
2913 If undefined or defined to C<1>, then all event-loop-specific functions
2914 will have the C<struct ev_loop *> as first argument, and you can create
2915 additional independent event loops. Otherwise there will be no support
2916 for multiple event loops and there is no first event loop pointer
2917 argument. Instead, all functions act on the single default loop.
2919 =item EV_MINPRI
2921 =item EV_MAXPRI
2923 The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
2924 C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
2925 provide for more priorities by overriding those symbols (usually defined
2926 to be C<-2> and C<2>, respectively).
2928 When doing priority-based operations, libev usually has to linearly search
2929 all the priorities, so having many of them (hundreds) uses a lot of space
2930 and time, so using the defaults of five priorities (-2 .. +2) is usually
2931 fine.
2933 If your embedding app does not need any priorities, defining these both to
2934 C<0> will save some memory and cpu.
2938 If undefined or defined to be C<1>, then periodic timers are supported. If
2939 defined to be C<0>, then they are not. Disabling them saves a few kB of
2940 code.
2942 =item EV_IDLE_ENABLE
2944 If undefined or defined to be C<1>, then idle watchers are supported. If
2945 defined to be C<0>, then they are not. Disabling them saves a few kB of
2946 code.
2948 =item EV_EMBED_ENABLE
2950 If undefined or defined to be C<1>, then embed watchers are supported. If
2951 defined to be C<0>, then they are not.
2953 =item EV_STAT_ENABLE
2955 If undefined or defined to be C<1>, then stat watchers are supported. If
2956 defined to be C<0>, then they are not.
2958 =item EV_FORK_ENABLE
2960 If undefined or defined to be C<1>, then fork watchers are supported. If
2961 defined to be C<0>, then they are not.
2963 =item EV_ASYNC_ENABLE
2965 If undefined or defined to be C<1>, then async watchers are supported. If
2966 defined to be C<0>, then they are not.
2968 =item EV_MINIMAL
2970 If you need to shave off some kilobytes of code at the expense of some
2971 speed, define this symbol to C<1>. Currently only used for gcc to override
2972 some inlining decisions, saves roughly 30% codesize of amd64.
2974 =item EV_PID_HASHSIZE
2976 C<ev_child> watchers use a small hash table to distribute workload by
2977 pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
2978 than enough. If you need to manage thousands of children you might want to
2979 increase this value (I<must> be a power of two).
2983 C<ev_stat> watchers use a small hash table to distribute workload by
2984 inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2985 usually more than enough. If you need to manage thousands of C<ev_stat>
2986 watchers you might want to increase this value (I<must> be a power of
2987 two).
2989 =item EV_COMMON
2991 By default, all watchers have a C<void *data> member. By redefining
2992 this macro to a something else you can include more and other types of
2993 members. You have to define it each time you include one of the files,
2994 though, and it must be identical each time.
2996 For example, the perl EV module uses something like this:
2998 #define EV_COMMON \
2999 SV *self; /* contains this struct */ \
3000 SV *cb_sv, *fh /* note no trailing ";" */
3002 =item EV_CB_DECLARE (type)
3004 =item EV_CB_INVOKE (watcher, revents)
3006 =item ev_set_cb (ev, cb)
3008 Can be used to change the callback member declaration in each watcher,
3009 and the way callbacks are invoked and set. Must expand to a struct member
3010 definition and a statement, respectively. See the F<ev.h> header file for
3011 their default definitions. One possible use for overriding these is to
3012 avoid the C<struct ev_loop *> as first argument in all cases, or to use
3013 method calls instead of plain function calls in C++.
3017 If you need to re-export the API (e.g. via a dll) and you need a list of
3018 exported symbols, you can use the provided F<Symbol.*> files which list
3019 all public symbols, one per line:
3021 Symbols.ev for libev proper
3022 Symbols.event for the libevent emulation
3024 This can also be used to rename all public symbols to avoid clashes with
3025 multiple versions of libev linked together (which is obviously bad in
3026 itself, but sometimes it is inconvinient to avoid this).
3028 A sed command like this will create wrapper C<#define>'s that you need to
3029 include before including F<ev.h>:
3031 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
3033 This would create a file F<wrap.h> which essentially looks like this:
3035 #define ev_backend myprefix_ev_backend
3036 #define ev_check_start myprefix_ev_check_start
3037 #define ev_check_stop myprefix_ev_check_stop
3038 ...
3040 =head2 EXAMPLES
3042 For a real-world example of a program the includes libev
3043 verbatim, you can have a look at the EV perl module
3044 (L<>). It has the libev files in
3045 the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
3046 interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
3047 will be compiled. It is pretty complex because it provides its own header
3048 file.
3050 The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
3051 that everybody includes and which overrides some configure choices:
3053 #define EV_MINIMAL 1
3054 #define EV_USE_POLL 0
3055 #define EV_MULTIPLICITY 0
3056 #define EV_PERIODIC_ENABLE 0
3057 #define EV_STAT_ENABLE 0
3058 #define EV_FORK_ENABLE 0
3059 #define EV_CONFIG_H <config.h>
3060 #define EV_MINPRI 0
3061 #define EV_MAXPRI 0
3063 #include "ev++.h"
3065 And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
3067 #include "ev_cpp.h"
3068 #include "ev.c"
3073 =head2 THREADS
3075 Libev itself is completely threadsafe, but it uses no locking. This
3076 means that you can use as many loops as you want in parallel, as long as
3077 only one thread ever calls into one libev function with the same loop
3078 parameter.
3080 Or put differently: calls with different loop parameters can be done in
3081 parallel from multiple threads, calls with the same loop parameter must be
3082 done serially (but can be done from different threads, as long as only one
3083 thread ever is inside a call at any point in time, e.g. by using a mutex
3084 per loop).
3086 If you want to know which design is best for your problem, then I cannot
3087 help you but by giving some generic advice:
3089 =over 4
3091 =item * most applications have a main thread: use the default libev loop
3092 in that thread, or create a seperate thread running only the default loop.
3094 This helps integrating other libraries or software modules that use libev
3095 themselves and don't care/know about threading.
3097 =item * one loop per thread is usually a good model.
3099 Doing this is almost never wrong, sometimes a better-performance model
3100 exists, but it is always a good start.
3102 =item * other models exist, such as the leader/follower pattern, where one
3103 loop is handed through multiple threads in a kind of round-robbin fashion.
3105 Chosing a model is hard - look around, learn, know that usually you cna do
3106 better than you currently do :-)
3108 =item * often you need to talk to some other thread which blocks in the
3109 event loop - C<ev_async> watchers can be used to wake them up from other
3110 threads safely (or from signal contexts...).
3112 =back
3114 =head2 COROUTINES
3116 Libev is much more accomodating to coroutines ("cooperative threads"):
3117 libev fully supports nesting calls to it's functions from different
3118 coroutines (e.g. you can call C<ev_loop> on the same loop from two
3119 different coroutines and switch freely between both coroutines running the
3120 loop, as long as you don't confuse yourself). The only exception is that
3121 you must not do this from C<ev_periodic> reschedule callbacks.
3123 Care has been invested into making sure that libev does not keep local
3124 state inside C<ev_loop>, and other calls do not usually allow coroutine
3125 switches.
3128 =head1 COMPLEXITIES
3130 In this section the complexities of (many of) the algorithms used inside
3131 libev will be explained. For complexity discussions about backends see the
3132 documentation for C<ev_default_init>.
3134 All of the following are about amortised time: If an array needs to be
3135 extended, libev needs to realloc and move the whole array, but this
3136 happens asymptotically never with higher number of elements, so O(1) might
3137 mean it might do a lengthy realloc operation in rare cases, but on average
3138 it is much faster and asymptotically approaches constant time.
3140 =over 4
3142 =item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
3144 This means that, when you have a watcher that triggers in one hour and
3145 there are 100 watchers that would trigger before that then inserting will
3146 have to skip roughly seven (C<ld 100>) of these watchers.
3148 =item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
3150 That means that changing a timer costs less than removing/adding them
3151 as only the relative motion in the event queue has to be paid for.
3153 =item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
3155 These just add the watcher into an array or at the head of a list.
3157 =item Stopping check/prepare/idle/fork/async watchers: O(1)
3159 =item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
3161 These watchers are stored in lists then need to be walked to find the
3162 correct watcher to remove. The lists are usually short (you don't usually
3163 have many watchers waiting for the same fd or signal).
3165 =item Finding the next timer in each loop iteration: O(1)
3167 By virtue of using a binary heap, the next timer is always found at the
3168 beginning of the storage array.
3170 =item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
3172 A change means an I/O watcher gets started or stopped, which requires
3173 libev to recalculate its status (and possibly tell the kernel, depending
3174 on backend and wether C<ev_io_set> was used).
3176 =item Activating one watcher (putting it into the pending state): O(1)
3178 =item Priority handling: O(number_of_priorities)
3180 Priorities are implemented by allocating some space for each
3181 priority. When doing priority-based operations, libev usually has to
3182 linearly search all the priorities, but starting/stopping and activating
3183 watchers becomes O(1) w.r.t. priority handling.
3185 =item Sending an ev_async: O(1)
3187 =item Processing ev_async_send: O(number_of_async_watchers)
3189 =item Processing signals: O(max_signal_number)
3191 Sending involves a syscall I<iff> there were no other C<ev_async_send>
3192 calls in the current loop iteration. Checking for async and signal events
3193 involves iterating over all running async watchers or all signal numbers.
3195 =back
3198 =head1 Win32 platform limitations and workarounds
3200 Win32 doesn't support any of the standards (e.g. POSIX) that libev
3201 requires, and its I/O model is fundamentally incompatible with the POSIX
3202 model. Libev still offers limited functionality on this platform in
3203 the form of the C<EVBACKEND_SELECT> backend, and only supports socket
3204 descriptors. This only applies when using Win32 natively, not when using
3205 e.g. cygwin.
3207 There is no supported compilation method available on windows except
3208 embedding it into other applications.
3210 Due to the many, low, and arbitrary limits on the win32 platform and the
3211 abysmal performance of winsockets, using a large number of sockets is not
3212 recommended (and not reasonable). If your program needs to use more than
3213 a hundred or so sockets, then likely it needs to use a totally different
3214 implementation for windows, as libev offers the POSIX model, which cannot
3215 be implemented efficiently on windows (microsoft monopoly games).
3217 =over 4
3219 =item The winsocket select function
3221 The winsocket C<select> function doesn't follow POSIX in that it requires
3222 socket I<handles> and not socket I<file descriptors>. This makes select
3223 very inefficient, and also requires a mapping from file descriptors
3224 to socket handles. See the discussion of the C<EV_SELECT_USE_FD_SET>,
3225 C<EV_SELECT_IS_WINSOCKET> and C<EV_FD_TO_WIN32_HANDLE> preprocessor
3226 symbols for more info.
3228 The configuration for a "naked" win32 using the microsoft runtime
3229 libraries and raw winsocket select is:
3231 #define EV_USE_SELECT 1
3232 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
3234 Note that winsockets handling of fd sets is O(n), so you can easily get a
3235 complexity in the O(n²) range when using win32.
3237 =item Limited number of file descriptors
3239 Windows has numerous arbitrary (and low) limits on things. Early versions
3240 of winsocket's select only supported waiting for a max. of C<64> handles
3241 (probably owning to the fact that all windows kernels can only wait for
3242 C<64> things at the same time internally; microsoft recommends spawning a
3243 chain of threads and wait for 63 handles and the previous thread in each).
3245 Newer versions support more handles, but you need to define C<FD_SETSIZE>
3246 to some high number (e.g. C<2048>) before compiling the winsocket select
3247 call (which might be in libev or elsewhere, for example, perl does its own
3248 select emulation on windows).
3250 Another limit is the number of file descriptors in the microsoft runtime
3251 libraries, which by default is C<64> (there must be a hidden I<64> fetish
3252 or something like this inside microsoft). You can increase this by calling
3253 C<_setmaxstdio>, which can increase this limit to C<2048> (another
3254 arbitrary limit), but is broken in many versions of the microsoft runtime
3255 libraries.
3257 This might get you to about C<512> or C<2048> sockets (depending on
3258 windows version and/or the phase of the moon). To get more, you need to
3259 wrap all I/O functions and provide your own fd management, but the cost of
3260 calling select (O(n²)) will likely make this unworkable.
3262 =back
3265 =head1 AUTHOR
3267 Marc Lehmann <>.