ViewVC Help
View File | Revision Log | Show Annotations | Download File
Revision: 1.121
Committed: Mon Jan 28 12:13:54 2008 UTC (16 years, 5 months ago) by root
Branch: MAIN
CVS Tags: rel-3_0
Changes since 1.120: +10 -12 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 #include <ev.h>
13 ev_io stdin_watcher;
14 ev_timer timeout_watcher;
16 /* called when data readable on stdin */
17 static void
18 stdin_cb (EV_P_ struct ev_io *w, int revents)
19 {
20 /* puts ("stdin ready"); */
21 ev_io_stop (EV_A_ w); /* just a syntax example */
22 ev_unloop (EV_A_ EVUNLOOP_ALL); /* leave all loop calls */
23 }
25 static void
26 timeout_cb (EV_P_ struct ev_timer *w, int revents)
27 {
28 /* puts ("timeout"); */
29 ev_unloop (EV_A_ EVUNLOOP_ONE); /* leave one loop call */
30 }
32 int
33 main (void)
34 {
35 struct ev_loop *loop = ev_default_loop (0);
37 /* initialise an io watcher, then start it */
38 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
39 ev_io_start (loop, &stdin_watcher);
41 /* simple non-repeating 5.5 second timeout */
42 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
43 ev_timer_start (loop, &timeout_watcher);
45 /* loop till timeout or data ready */
46 ev_loop (loop, 0);
48 return 0;
49 }
53 The newest version of this document is also available as a html-formatted
54 web page you might find easier to navigate when reading it for the first
55 time: L<>.
57 Libev is an event loop: you register interest in certain events (such as a
58 file descriptor being readable or a timeout occurring), and it will manage
59 these event sources and provide your program with events.
61 To do this, it must take more or less complete control over your process
62 (or thread) by executing the I<event loop> handler, and will then
63 communicate events via a callback mechanism.
65 You register interest in certain events by registering so-called I<event
66 watchers>, which are relatively small C structures you initialise with the
67 details of the event, and then hand it over to libev by I<starting> the
68 watcher.
70 =head2 FEATURES
72 Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
73 BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
74 for file descriptor events (C<ev_io>), the Linux C<inotify> interface
75 (for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
76 with customised rescheduling (C<ev_periodic>), synchronous signals
77 (C<ev_signal>), process status change events (C<ev_child>), and event
78 watchers dealing with the event loop mechanism itself (C<ev_idle>,
79 C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as
80 file watchers (C<ev_stat>) and even limited support for fork events
81 (C<ev_fork>).
83 It also is quite fast (see this
84 L<benchmark|> comparing it to libevent
85 for example).
89 Libev is very configurable. In this manual the default configuration will
90 be described, which supports multiple event loops. For more info about
91 various configuration options please have a look at B<EMBED> section in
92 this manual. If libev was configured without support for multiple event
93 loops, then all functions taking an initial argument of name C<loop>
94 (which is always of type C<struct ev_loop *>) will not have this argument.
98 Libev represents time as a single floating point number, representing the
99 (fractional) number of seconds since the (POSIX) epoch (somewhere near
100 the beginning of 1970, details are complicated, don't ask). This type is
101 called C<ev_tstamp>, which is what you should use too. It usually aliases
102 to the C<double> type in C, and when you need to do any calculations on
103 it, you should treat it as some floatingpoint value. Unlike the name
104 component C<stamp> might indicate, it is also used for time differences
105 throughout libev.
109 These functions can be called anytime, even before initialising the
110 library in any way.
112 =over 4
114 =item ev_tstamp ev_time ()
116 Returns the current time as libev would use it. Please note that the
117 C<ev_now> function is usually faster and also often returns the timestamp
118 you actually want to know.
120 =item ev_sleep (ev_tstamp interval)
122 Sleep for the given interval: The current thread will be blocked until
123 either it is interrupted or the given time interval has passed. Basically
124 this is a subsecond-resolution C<sleep ()>.
126 =item int ev_version_major ()
128 =item int ev_version_minor ()
130 You can find out the major and minor ABI version numbers of the library
131 you linked against by calling the functions C<ev_version_major> and
132 C<ev_version_minor>. If you want, you can compare against the global
133 symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
134 version of the library your program was compiled against.
136 These version numbers refer to the ABI version of the library, not the
137 release version.
139 Usually, it's a good idea to terminate if the major versions mismatch,
140 as this indicates an incompatible change. Minor versions are usually
141 compatible to older versions, so a larger minor version alone is usually
142 not a problem.
144 Example: Make sure we haven't accidentally been linked against the wrong
145 version.
147 assert (("libev version mismatch",
148 ev_version_major () == EV_VERSION_MAJOR
149 && ev_version_minor () >= EV_VERSION_MINOR));
151 =item unsigned int ev_supported_backends ()
153 Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
154 value) compiled into this binary of libev (independent of their
155 availability on the system you are running on). See C<ev_default_loop> for
156 a description of the set values.
158 Example: make sure we have the epoll method, because yeah this is cool and
159 a must have and can we have a torrent of it please!!!11
161 assert (("sorry, no epoll, no sex",
162 ev_supported_backends () & EVBACKEND_EPOLL));
164 =item unsigned int ev_recommended_backends ()
166 Return the set of all backends compiled into this binary of libev and also
167 recommended for this platform. This set is often smaller than the one
168 returned by C<ev_supported_backends>, as for example kqueue is broken on
169 most BSDs and will not be autodetected unless you explicitly request it
170 (assuming you know what you are doing). This is the set of backends that
171 libev will probe for if you specify no backends explicitly.
173 =item unsigned int ev_embeddable_backends ()
175 Returns the set of backends that are embeddable in other event loops. This
176 is the theoretical, all-platform, value. To find which backends
177 might be supported on the current system, you would need to look at
178 C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
179 recommended ones.
181 See the description of C<ev_embed> watchers for more info.
183 =item ev_set_allocator (void *(*cb)(void *ptr, long size))
185 Sets the allocation function to use (the prototype is similar - the
186 semantics is identical - to the realloc C function). It is used to
187 allocate and free memory (no surprises here). If it returns zero when
188 memory needs to be allocated, the library might abort or take some
189 potentially destructive action. The default is your system realloc
190 function.
192 You could override this function in high-availability programs to, say,
193 free some memory if it cannot allocate memory, to use a special allocator,
194 or even to sleep a while and retry until some memory is available.
196 Example: Replace the libev allocator with one that waits a bit and then
197 retries).
199 static void *
200 persistent_realloc (void *ptr, size_t size)
201 {
202 for (;;)
203 {
204 void *newptr = realloc (ptr, size);
206 if (newptr)
207 return newptr;
209 sleep (60);
210 }
211 }
213 ...
214 ev_set_allocator (persistent_realloc);
216 =item ev_set_syserr_cb (void (*cb)(const char *msg));
218 Set the callback function to call on a retryable syscall error (such
219 as failed select, poll, epoll_wait). The message is a printable string
220 indicating the system call or subsystem causing the problem. If this
221 callback is set, then libev will expect it to remedy the sitution, no
222 matter what, when it returns. That is, libev will generally retry the
223 requested operation, or, if the condition doesn't go away, do bad stuff
224 (such as abort).
226 Example: This is basically the same thing that libev does internally, too.
228 static void
229 fatal_error (const char *msg)
230 {
231 perror (msg);
232 abort ();
233 }
235 ...
236 ev_set_syserr_cb (fatal_error);
238 =back
242 An event loop is described by a C<struct ev_loop *>. The library knows two
243 types of such loops, the I<default> loop, which supports signals and child
244 events, and dynamically created loops which do not.
246 If you use threads, a common model is to run the default event loop
247 in your main thread (or in a separate thread) and for each thread you
248 create, you also create another event loop. Libev itself does no locking
249 whatsoever, so if you mix calls to the same event loop in different
250 threads, make sure you lock (this is usually a bad idea, though, even if
251 done correctly, because it's hideous and inefficient).
253 =over 4
255 =item struct ev_loop *ev_default_loop (unsigned int flags)
257 This will initialise the default event loop if it hasn't been initialised
258 yet and return it. If the default loop could not be initialised, returns
259 false. If it already was initialised it simply returns it (and ignores the
260 flags. If that is troubling you, check C<ev_backend ()> afterwards).
262 If you don't know what event loop to use, use the one returned from this
263 function.
265 The default loop is the only loop that can handle C<ev_signal> and
266 C<ev_child> watchers, and to do this, it always registers a handler
267 for C<SIGCHLD>. If this is a problem for your app you can either
268 create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
269 can simply overwrite the C<SIGCHLD> signal handler I<after> calling
270 C<ev_default_init>.
272 The flags argument can be used to specify special behaviour or specific
273 backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
275 The following flags are supported:
277 =over 4
279 =item C<EVFLAG_AUTO>
281 The default flags value. Use this if you have no clue (it's the right
282 thing, believe me).
284 =item C<EVFLAG_NOENV>
286 If this flag bit is ored into the flag value (or the program runs setuid
287 or setgid) then libev will I<not> look at the environment variable
288 C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
289 override the flags completely if it is found in the environment. This is
290 useful to try out specific backends to test their performance, or to work
291 around bugs.
295 Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
296 a fork, you can also make libev check for a fork in each iteration by
297 enabling this flag.
299 This works by calling C<getpid ()> on every iteration of the loop,
300 and thus this might slow down your event loop if you do a lot of loop
301 iterations and little real work, but is usually not noticeable (on my
302 Linux system for example, C<getpid> is actually a simple 5-insn sequence
303 without a syscall and thus I<very> fast, but my Linux system also has
304 C<pthread_atfork> which is even faster).
306 The big advantage of this flag is that you can forget about fork (and
307 forget about forgetting to tell libev about forking) when you use this
308 flag.
310 This flag setting cannot be overriden or specified in the C<LIBEV_FLAGS>
311 environment variable.
313 =item C<EVBACKEND_SELECT> (value 1, portable select backend)
315 This is your standard select(2) backend. Not I<completely> standard, as
316 libev tries to roll its own fd_set with no limits on the number of fds,
317 but if that fails, expect a fairly low limit on the number of fds when
318 using this backend. It doesn't scale too well (O(highest_fd)), but its
319 usually the fastest backend for a low number of (low-numbered :) fds.
321 To get good performance out of this backend you need a high amount of
322 parallelity (most of the file descriptors should be busy). If you are
323 writing a server, you should C<accept ()> in a loop to accept as many
324 connections as possible during one iteration. You might also want to have
325 a look at C<ev_set_io_collect_interval ()> to increase the amount of
326 readyness notifications you get per iteration.
328 =item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
330 And this is your standard poll(2) backend. It's more complicated
331 than select, but handles sparse fds better and has no artificial
332 limit on the number of fds you can use (except it will slow down
333 considerably with a lot of inactive fds). It scales similarly to select,
334 i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
335 performance tips.
337 =item C<EVBACKEND_EPOLL> (value 4, Linux)
339 For few fds, this backend is a bit little slower than poll and select,
340 but it scales phenomenally better. While poll and select usually scale
341 like O(total_fds) where n is the total number of fds (or the highest fd),
342 epoll scales either O(1) or O(active_fds). The epoll design has a number
343 of shortcomings, such as silently dropping events in some hard-to-detect
344 cases and rewiring a syscall per fd change, no fork support and bad
345 support for dup.
347 While stopping, setting and starting an I/O watcher in the same iteration
348 will result in some caching, there is still a syscall per such incident
349 (because the fd could point to a different file description now), so its
350 best to avoid that. Also, C<dup ()>'ed file descriptors might not work
351 very well if you register events for both fds.
353 Please note that epoll sometimes generates spurious notifications, so you
354 need to use non-blocking I/O or other means to avoid blocking when no data
355 (or space) is available.
357 Best performance from this backend is achieved by not unregistering all
358 watchers for a file descriptor until it has been closed, if possible, i.e.
359 keep at least one watcher active per fd at all times.
361 While nominally embeddeble in other event loops, this feature is broken in
362 all kernel versions tested so far.
364 =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
366 Kqueue deserves special mention, as at the time of this writing, it
367 was broken on all BSDs except NetBSD (usually it doesn't work reliably
368 with anything but sockets and pipes, except on Darwin, where of course
369 it's completely useless). For this reason it's not being "autodetected"
370 unless you explicitly specify it explicitly in the flags (i.e. using
371 C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
372 system like NetBSD.
374 You still can embed kqueue into a normal poll or select backend and use it
375 only for sockets (after having made sure that sockets work with kqueue on
376 the target platform). See C<ev_embed> watchers for more info.
378 It scales in the same way as the epoll backend, but the interface to the
379 kernel is more efficient (which says nothing about its actual speed, of
380 course). While stopping, setting and starting an I/O watcher does never
381 cause an extra syscall as with C<EVBACKEND_EPOLL>, it still adds up to
382 two event changes per incident, support for C<fork ()> is very bad and it
383 drops fds silently in similarly hard-to-detect cases.
385 This backend usually performs well under most conditions.
387 While nominally embeddable in other event loops, this doesn't work
388 everywhere, so you might need to test for this. And since it is broken
389 almost everywhere, you should only use it when you have a lot of sockets
390 (for which it usually works), by embedding it into another event loop
391 (e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and using it only for
392 sockets.
394 =item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
396 This is not implemented yet (and might never be, unless you send me an
397 implementation). According to reports, C</dev/poll> only supports sockets
398 and is not embeddable, which would limit the usefulness of this backend
399 immensely.
401 =item C<EVBACKEND_PORT> (value 32, Solaris 10)
403 This uses the Solaris 10 event port mechanism. As with everything on Solaris,
404 it's really slow, but it still scales very well (O(active_fds)).
406 Please note that solaris event ports can deliver a lot of spurious
407 notifications, so you need to use non-blocking I/O or other means to avoid
408 blocking when no data (or space) is available.
410 While this backend scales well, it requires one system call per active
411 file descriptor per loop iteration. For small and medium numbers of file
412 descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
413 might perform better.
415 On the positive side, ignoring the spurious readyness notifications, this
416 backend actually performed to specification in all tests and is fully
417 embeddable, which is a rare feat among the OS-specific backends.
419 =item C<EVBACKEND_ALL>
421 Try all backends (even potentially broken ones that wouldn't be tried
422 with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
425 It is definitely not recommended to use this flag.
427 =back
429 If one or more of these are ored into the flags value, then only these
430 backends will be tried (in the reverse order as listed here). If none are
431 specified, all backends in C<ev_recommended_backends ()> will be tried.
433 The most typical usage is like this:
435 if (!ev_default_loop (0))
436 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
438 Restrict libev to the select and poll backends, and do not allow
439 environment settings to be taken into account:
443 Use whatever libev has to offer, but make sure that kqueue is used if
444 available (warning, breaks stuff, best use only with your own private
445 event loop and only if you know the OS supports your types of fds):
447 ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
449 =item struct ev_loop *ev_loop_new (unsigned int flags)
451 Similar to C<ev_default_loop>, but always creates a new event loop that is
452 always distinct from the default loop. Unlike the default loop, it cannot
453 handle signal and child watchers, and attempts to do so will be greeted by
454 undefined behaviour (or a failed assertion if assertions are enabled).
456 Example: Try to create a event loop that uses epoll and nothing else.
458 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
459 if (!epoller)
460 fatal ("no epoll found here, maybe it hides under your chair");
462 =item ev_default_destroy ()
464 Destroys the default loop again (frees all memory and kernel state
465 etc.). None of the active event watchers will be stopped in the normal
466 sense, so e.g. C<ev_is_active> might still return true. It is your
467 responsibility to either stop all watchers cleanly yoursef I<before>
468 calling this function, or cope with the fact afterwards (which is usually
469 the easiest thing, you can just ignore the watchers and/or C<free ()> them
470 for example).
472 Note that certain global state, such as signal state, will not be freed by
473 this function, and related watchers (such as signal and child watchers)
474 would need to be stopped manually.
476 In general it is not advisable to call this function except in the
477 rare occasion where you really need to free e.g. the signal handling
478 pipe fds. If you need dynamically allocated loops it is better to use
479 C<ev_loop_new> and C<ev_loop_destroy>).
481 =item ev_loop_destroy (loop)
483 Like C<ev_default_destroy>, but destroys an event loop created by an
484 earlier call to C<ev_loop_new>.
486 =item ev_default_fork ()
488 This function sets a flag that causes subsequent C<ev_loop> iterations
489 to reinitialise the kernel state for backends that have one. Despite the
490 name, you can call it anytime, but it makes most sense after forking, in
491 the child process (or both child and parent, but that again makes little
492 sense). You I<must> call it in the child before using any of the libev
493 functions, and it will only take effect at the next C<ev_loop> iteration.
495 On the other hand, you only need to call this function in the child
496 process if and only if you want to use the event library in the child. If
497 you just fork+exec, you don't have to call it at all.
499 The function itself is quite fast and it's usually not a problem to call
500 it just in case after a fork. To make this easy, the function will fit in
501 quite nicely into a call to C<pthread_atfork>:
503 pthread_atfork (0, 0, ev_default_fork);
505 =item ev_loop_fork (loop)
507 Like C<ev_default_fork>, but acts on an event loop created by
508 C<ev_loop_new>. Yes, you have to call this on every allocated event loop
509 after fork, and how you do this is entirely your own problem.
511 =item unsigned int ev_loop_count (loop)
513 Returns the count of loop iterations for the loop, which is identical to
514 the number of times libev did poll for new events. It starts at C<0> and
515 happily wraps around with enough iterations.
517 This value can sometimes be useful as a generation counter of sorts (it
518 "ticks" the number of loop iterations), as it roughly corresponds with
519 C<ev_prepare> and C<ev_check> calls.
521 =item unsigned int ev_backend (loop)
523 Returns one of the C<EVBACKEND_*> flags indicating the event backend in
524 use.
526 =item ev_tstamp ev_now (loop)
528 Returns the current "event loop time", which is the time the event loop
529 received events and started processing them. This timestamp does not
530 change as long as callbacks are being processed, and this is also the base
531 time used for relative timers. You can treat it as the timestamp of the
532 event occurring (or more correctly, libev finding out about it).
534 =item ev_loop (loop, int flags)
536 Finally, this is it, the event handler. This function usually is called
537 after you initialised all your watchers and you want to start handling
538 events.
540 If the flags argument is specified as C<0>, it will not return until
541 either no event watchers are active anymore or C<ev_unloop> was called.
543 Please note that an explicit C<ev_unloop> is usually better than
544 relying on all watchers to be stopped when deciding when a program has
545 finished (especially in interactive programs), but having a program that
546 automatically loops as long as it has to and no longer by virtue of
547 relying on its watchers stopping correctly is a thing of beauty.
549 A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
550 those events and any outstanding ones, but will not block your process in
551 case there are no events and will return after one iteration of the loop.
553 A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
554 neccessary) and will handle those and any outstanding ones. It will block
555 your process until at least one new event arrives, and will return after
556 one iteration of the loop. This is useful if you are waiting for some
557 external event in conjunction with something not expressible using other
558 libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is
559 usually a better approach for this kind of thing.
561 Here are the gory details of what C<ev_loop> does:
563 - Before the first iteration, call any pending watchers.
564 * If EVFLAG_FORKCHECK was used, check for a fork.
565 - If a fork was detected, queue and call all fork watchers.
566 - Queue and call all prepare watchers.
567 - If we have been forked, recreate the kernel state.
568 - Update the kernel state with all outstanding changes.
569 - Update the "event loop time".
570 - Calculate for how long to sleep or block, if at all
571 (active idle watchers, EVLOOP_NONBLOCK or not having
572 any active watchers at all will result in not sleeping).
573 - Sleep if the I/O and timer collect interval say so.
574 - Block the process, waiting for any events.
575 - Queue all outstanding I/O (fd) events.
576 - Update the "event loop time" and do time jump handling.
577 - Queue all outstanding timers.
578 - Queue all outstanding periodics.
579 - If no events are pending now, queue all idle watchers.
580 - Queue all check watchers.
581 - Call all queued watchers in reverse order (i.e. check watchers first).
582 Signals and child watchers are implemented as I/O watchers, and will
583 be handled here by queueing them when their watcher gets executed.
584 - If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
585 were used, or there are no active watchers, return, otherwise
586 continue with step *.
588 Example: Queue some jobs and then loop until no events are outstanding
589 anymore.
591 ... queue jobs here, make sure they register event watchers as long
592 ... as they still have work to do (even an idle watcher will do..)
593 ev_loop (my_loop, 0);
594 ... jobs done. yeah!
596 =item ev_unloop (loop, how)
598 Can be used to make a call to C<ev_loop> return early (but only after it
599 has processed all outstanding events). The C<how> argument must be either
600 C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
601 C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
603 This "unloop state" will be cleared when entering C<ev_loop> again.
605 =item ev_ref (loop)
607 =item ev_unref (loop)
609 Ref/unref can be used to add or remove a reference count on the event
610 loop: Every watcher keeps one reference, and as long as the reference
611 count is nonzero, C<ev_loop> will not return on its own. If you have
612 a watcher you never unregister that should not keep C<ev_loop> from
613 returning, ev_unref() after starting, and ev_ref() before stopping it. For
614 example, libev itself uses this for its internal signal pipe: It is not
615 visible to the libev user and should not keep C<ev_loop> from exiting if
616 no event watchers registered by it are active. It is also an excellent
617 way to do this for generic recurring timers or from within third-party
618 libraries. Just remember to I<unref after start> and I<ref before stop>
619 (but only if the watcher wasn't active before, or was active before,
620 respectively).
622 Example: Create a signal watcher, but keep it from keeping C<ev_loop>
623 running when nothing else is active.
625 struct ev_signal exitsig;
626 ev_signal_init (&exitsig, sig_cb, SIGINT);
627 ev_signal_start (loop, &exitsig);
628 evf_unref (loop);
630 Example: For some weird reason, unregister the above signal handler again.
632 ev_ref (loop);
633 ev_signal_stop (loop, &exitsig);
635 =item ev_set_io_collect_interval (loop, ev_tstamp interval)
637 =item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
639 These advanced functions influence the time that libev will spend waiting
640 for events. Both are by default C<0>, meaning that libev will try to
641 invoke timer/periodic callbacks and I/O callbacks with minimum latency.
643 Setting these to a higher value (the C<interval> I<must> be >= C<0>)
644 allows libev to delay invocation of I/O and timer/periodic callbacks to
645 increase efficiency of loop iterations.
647 The background is that sometimes your program runs just fast enough to
648 handle one (or very few) event(s) per loop iteration. While this makes
649 the program responsive, it also wastes a lot of CPU time to poll for new
650 events, especially with backends like C<select ()> which have a high
651 overhead for the actual polling but can deliver many events at once.
653 By setting a higher I<io collect interval> you allow libev to spend more
654 time collecting I/O events, so you can handle more events per iteration,
655 at the cost of increasing latency. Timeouts (both C<ev_periodic> and
656 C<ev_timer>) will be not affected. Setting this to a non-null value will
657 introduce an additional C<ev_sleep ()> call into most loop iterations.
659 Likewise, by setting a higher I<timeout collect interval> you allow libev
660 to spend more time collecting timeouts, at the expense of increased
661 latency (the watcher callback will be called later). C<ev_io> watchers
662 will not be affected. Setting this to a non-null value will not introduce
663 any overhead in libev.
665 Many (busy) programs can usually benefit by setting the io collect
666 interval to a value near C<0.1> or so, which is often enough for
667 interactive servers (of course not for games), likewise for timeouts. It
668 usually doesn't make much sense to set it to a lower value than C<0.01>,
669 as this approsaches the timing granularity of most systems.
671 =back
676 A watcher is a structure that you create and register to record your
677 interest in some event. For instance, if you want to wait for STDIN to
678 become readable, you would create an C<ev_io> watcher for that:
680 static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)
681 {
682 ev_io_stop (w);
683 ev_unloop (loop, EVUNLOOP_ALL);
684 }
686 struct ev_loop *loop = ev_default_loop (0);
687 struct ev_io stdin_watcher;
688 ev_init (&stdin_watcher, my_cb);
689 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
690 ev_io_start (loop, &stdin_watcher);
691 ev_loop (loop, 0);
693 As you can see, you are responsible for allocating the memory for your
694 watcher structures (and it is usually a bad idea to do this on the stack,
695 although this can sometimes be quite valid).
697 Each watcher structure must be initialised by a call to C<ev_init
698 (watcher *, callback)>, which expects a callback to be provided. This
699 callback gets invoked each time the event occurs (or, in the case of io
700 watchers, each time the event loop detects that the file descriptor given
701 is readable and/or writable).
703 Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro
704 with arguments specific to this watcher type. There is also a macro
705 to combine initialisation and setting in one call: C<< ev_<type>_init
706 (watcher *, callback, ...) >>.
708 To make the watcher actually watch out for events, you have to start it
709 with a watcher-specific start function (C<< ev_<type>_start (loop, watcher
710 *) >>), and you can stop watching for events at any time by calling the
711 corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.
713 As long as your watcher is active (has been started but not stopped) you
714 must not touch the values stored in it. Most specifically you must never
715 reinitialise it or call its C<set> macro.
717 Each and every callback receives the event loop pointer as first, the
718 registered watcher structure as second, and a bitset of received events as
719 third argument.
721 The received events usually include a single bit per event type received
722 (you can receive multiple events at the same time). The possible bit masks
723 are:
725 =over 4
727 =item C<EV_READ>
729 =item C<EV_WRITE>
731 The file descriptor in the C<ev_io> watcher has become readable and/or
732 writable.
734 =item C<EV_TIMEOUT>
736 The C<ev_timer> watcher has timed out.
738 =item C<EV_PERIODIC>
740 The C<ev_periodic> watcher has timed out.
742 =item C<EV_SIGNAL>
744 The signal specified in the C<ev_signal> watcher has been received by a thread.
746 =item C<EV_CHILD>
748 The pid specified in the C<ev_child> watcher has received a status change.
750 =item C<EV_STAT>
752 The path specified in the C<ev_stat> watcher changed its attributes somehow.
754 =item C<EV_IDLE>
756 The C<ev_idle> watcher has determined that you have nothing better to do.
758 =item C<EV_PREPARE>
760 =item C<EV_CHECK>
762 All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts
763 to gather new events, and all C<ev_check> watchers are invoked just after
764 C<ev_loop> has gathered them, but before it invokes any callbacks for any
765 received events. Callbacks of both watcher types can start and stop as
766 many watchers as they want, and all of them will be taken into account
767 (for example, a C<ev_prepare> watcher might start an idle watcher to keep
768 C<ev_loop> from blocking).
770 =item C<EV_EMBED>
772 The embedded event loop specified in the C<ev_embed> watcher needs attention.
774 =item C<EV_FORK>
776 The event loop has been resumed in the child process after fork (see
777 C<ev_fork>).
779 =item C<EV_ERROR>
781 An unspecified error has occured, the watcher has been stopped. This might
782 happen because the watcher could not be properly started because libev
783 ran out of memory, a file descriptor was found to be closed or any other
784 problem. You best act on it by reporting the problem and somehow coping
785 with the watcher being stopped.
787 Libev will usually signal a few "dummy" events together with an error,
788 for example it might indicate that a fd is readable or writable, and if
789 your callbacks is well-written it can just attempt the operation and cope
790 with the error from read() or write(). This will not work in multithreaded
791 programs, though, so beware.
793 =back
797 In the following description, C<TYPE> stands for the watcher type,
798 e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
800 =over 4
802 =item C<ev_init> (ev_TYPE *watcher, callback)
804 This macro initialises the generic portion of a watcher. The contents
805 of the watcher object can be arbitrary (so C<malloc> will do). Only
806 the generic parts of the watcher are initialised, you I<need> to call
807 the type-specific C<ev_TYPE_set> macro afterwards to initialise the
808 type-specific parts. For each type there is also a C<ev_TYPE_init> macro
809 which rolls both calls into one.
811 You can reinitialise a watcher at any time as long as it has been stopped
812 (or never started) and there are no pending events outstanding.
814 The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,
815 int revents)>.
817 =item C<ev_TYPE_set> (ev_TYPE *, [args])
819 This macro initialises the type-specific parts of a watcher. You need to
820 call C<ev_init> at least once before you call this macro, but you can
821 call C<ev_TYPE_set> any number of times. You must not, however, call this
822 macro on a watcher that is active (it can be pending, however, which is a
823 difference to the C<ev_init> macro).
825 Although some watcher types do not have type-specific arguments
826 (e.g. C<ev_prepare>) you still need to call its C<set> macro.
828 =item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
830 This convinience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
831 calls into a single call. This is the most convinient method to initialise
832 a watcher. The same limitations apply, of course.
834 =item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
836 Starts (activates) the given watcher. Only active watchers will receive
837 events. If the watcher is already active nothing will happen.
839 =item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
841 Stops the given watcher again (if active) and clears the pending
842 status. It is possible that stopped watchers are pending (for example,
843 non-repeating timers are being stopped when they become pending), but
844 C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If
845 you want to free or reuse the memory used by the watcher it is therefore a
846 good idea to always call its C<ev_TYPE_stop> function.
848 =item bool ev_is_active (ev_TYPE *watcher)
850 Returns a true value iff the watcher is active (i.e. it has been started
851 and not yet been stopped). As long as a watcher is active you must not modify
852 it.
854 =item bool ev_is_pending (ev_TYPE *watcher)
856 Returns a true value iff the watcher is pending, (i.e. it has outstanding
857 events but its callback has not yet been invoked). As long as a watcher
858 is pending (but not active) you must not call an init function on it (but
859 C<ev_TYPE_set> is safe), you must not change its priority, and you must
860 make sure the watcher is available to libev (e.g. you cannot C<free ()>
861 it).
863 =item callback ev_cb (ev_TYPE *watcher)
865 Returns the callback currently set on the watcher.
867 =item ev_cb_set (ev_TYPE *watcher, callback)
869 Change the callback. You can change the callback at virtually any time
870 (modulo threads).
872 =item ev_set_priority (ev_TYPE *watcher, priority)
874 =item int ev_priority (ev_TYPE *watcher)
876 Set and query the priority of the watcher. The priority is a small
877 integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
878 (default: C<-2>). Pending watchers with higher priority will be invoked
879 before watchers with lower priority, but priority will not keep watchers
880 from being executed (except for C<ev_idle> watchers).
882 This means that priorities are I<only> used for ordering callback
883 invocation after new events have been received. This is useful, for
884 example, to reduce latency after idling, or more often, to bind two
885 watchers on the same event and make sure one is called first.
887 If you need to suppress invocation when higher priority events are pending
888 you need to look at C<ev_idle> watchers, which provide this functionality.
890 You I<must not> change the priority of a watcher as long as it is active or
891 pending.
893 The default priority used by watchers when no priority has been set is
894 always C<0>, which is supposed to not be too high and not be too low :).
896 Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
897 fine, as long as you do not mind that the priority value you query might
898 or might not have been adjusted to be within valid range.
900 =item ev_invoke (loop, ev_TYPE *watcher, int revents)
902 Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
903 C<loop> nor C<revents> need to be valid as long as the watcher callback
904 can deal with that fact.
906 =item int ev_clear_pending (loop, ev_TYPE *watcher)
908 If the watcher is pending, this function returns clears its pending status
909 and returns its C<revents> bitset (as if its callback was invoked). If the
910 watcher isn't pending it does nothing and returns C<0>.
912 =back
917 Each watcher has, by default, a member C<void *data> that you can change
918 and read at any time, libev will completely ignore it. This can be used
919 to associate arbitrary data with your watcher. If you need more data and
920 don't want to allocate memory and store a pointer to it in that data
921 member, you can also "subclass" the watcher type and provide your own
922 data:
924 struct my_io
925 {
926 struct ev_io io;
927 int otherfd;
928 void *somedata;
929 struct whatever *mostinteresting;
930 }
932 And since your callback will be called with a pointer to the watcher, you
933 can cast it back to your own type:
935 static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
936 {
937 struct my_io *w = (struct my_io *)w_;
938 ...
939 }
941 More interesting and less C-conformant ways of casting your callback type
942 instead have been omitted.
944 Another common scenario is having some data structure with multiple
945 watchers:
947 struct my_biggy
948 {
949 int some_data;
950 ev_timer t1;
951 ev_timer t2;
952 }
954 In this case getting the pointer to C<my_biggy> is a bit more complicated,
955 you need to use C<offsetof>:
957 #include <stddef.h>
959 static void
960 t1_cb (EV_P_ struct ev_timer *w, int revents)
961 {
962 struct my_biggy big = (struct my_biggy *
963 (((char *)w) - offsetof (struct my_biggy, t1));
964 }
966 static void
967 t2_cb (EV_P_ struct ev_timer *w, int revents)
968 {
969 struct my_biggy big = (struct my_biggy *
970 (((char *)w) - offsetof (struct my_biggy, t2));
971 }
974 =head1 WATCHER TYPES
976 This section describes each watcher in detail, but will not repeat
977 information given in the last section. Any initialisation/set macros,
978 functions and members specific to the watcher type are explained.
980 Members are additionally marked with either I<[read-only]>, meaning that,
981 while the watcher is active, you can look at the member and expect some
982 sensible content, but you must not modify it (you can modify it while the
983 watcher is stopped to your hearts content), or I<[read-write]>, which
984 means you can expect it to have some sensible content while the watcher
985 is active, but you can also modify it. Modifying it may not do something
986 sensible or take immediate effect (or do anything at all), but libev will
987 not crash or malfunction in any way.
990 =head2 C<ev_io> - is this file descriptor readable or writable?
992 I/O watchers check whether a file descriptor is readable or writable
993 in each iteration of the event loop, or, more precisely, when reading
994 would not block the process and writing would at least be able to write
995 some data. This behaviour is called level-triggering because you keep
996 receiving events as long as the condition persists. Remember you can stop
997 the watcher if you don't want to act on the event and neither want to
998 receive future events.
1000 In general you can register as many read and/or write event watchers per
1001 fd as you want (as long as you don't confuse yourself). Setting all file
1002 descriptors to non-blocking mode is also usually a good idea (but not
1003 required if you know what you are doing).
1005 If you must do this, then force the use of a known-to-be-good backend
1006 (at the time of this writing, this includes only C<EVBACKEND_SELECT> and
1009 Another thing you have to watch out for is that it is quite easy to
1010 receive "spurious" readyness notifications, that is your callback might
1011 be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1012 because there is no data. Not only are some backends known to create a
1013 lot of those (for example solaris ports), it is very easy to get into
1014 this situation even with a relatively standard program structure. Thus
1015 it is best to always use non-blocking I/O: An extra C<read>(2) returning
1016 C<EAGAIN> is far preferable to a program hanging until some data arrives.
1018 If you cannot run the fd in non-blocking mode (for example you should not
1019 play around with an Xlib connection), then you have to seperately re-test
1020 whether a file descriptor is really ready with a known-to-be good interface
1021 such as poll (fortunately in our Xlib example, Xlib already does this on
1022 its own, so its quite safe to use).
1024 =head3 The special problem of disappearing file descriptors
1026 Some backends (e.g. kqueue, epoll) need to be told about closing a file
1027 descriptor (either by calling C<close> explicitly or by any other means,
1028 such as C<dup>). The reason is that you register interest in some file
1029 descriptor, but when it goes away, the operating system will silently drop
1030 this interest. If another file descriptor with the same number then is
1031 registered with libev, there is no efficient way to see that this is, in
1032 fact, a different file descriptor.
1034 To avoid having to explicitly tell libev about such cases, libev follows
1035 the following policy: Each time C<ev_io_set> is being called, libev
1036 will assume that this is potentially a new file descriptor, otherwise
1037 it is assumed that the file descriptor stays the same. That means that
1038 you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1039 descriptor even if the file descriptor number itself did not change.
1041 This is how one would do it normally anyway, the important point is that
1042 the libev application should not optimise around libev but should leave
1043 optimisations to libev.
1045 =head3 The special problem of dup'ed file descriptors
1047 Some backends (e.g. epoll), cannot register events for file descriptors,
1048 but only events for the underlying file descriptions. That means when you
1049 have C<dup ()>'ed file descriptors or weirder constellations, and register
1050 events for them, only one file descriptor might actually receive events.
1052 There is no workaround possible except not registering events
1053 for potentially C<dup ()>'ed file descriptors, or to resort to
1056 =head3 The special problem of fork
1058 Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1059 useless behaviour. Libev fully supports fork, but needs to be told about
1060 it in the child.
1062 To support fork in your programs, you either have to call
1063 C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
1064 enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1068 =head3 Watcher-Specific Functions
1070 =over 4
1072 =item ev_io_init (ev_io *, callback, int fd, int events)
1074 =item ev_io_set (ev_io *, int fd, int events)
1076 Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1077 rceeive events for and events is either C<EV_READ>, C<EV_WRITE> or
1078 C<EV_READ | EV_WRITE> to receive the given events.
1080 =item int fd [read-only]
1082 The file descriptor being watched.
1084 =item int events [read-only]
1086 The events being watched.
1088 =back
1090 =head3 Examples
1092 Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1093 readable, but only once. Since it is likely line-buffered, you could
1094 attempt to read a whole line in the callback.
1096 static void
1097 stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1098 {
1099 ev_io_stop (loop, w);
1100 .. read from stdin here (or from w->fd) and haqndle any I/O errors
1101 }
1103 ...
1104 struct ev_loop *loop = ev_default_init (0);
1105 struct ev_io stdin_readable;
1106 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1107 ev_io_start (loop, &stdin_readable);
1108 ev_loop (loop, 0);
1111 =head2 C<ev_timer> - relative and optionally repeating timeouts
1113 Timer watchers are simple relative timers that generate an event after a
1114 given time, and optionally repeating in regular intervals after that.
1116 The timers are based on real time, that is, if you register an event that
1117 times out after an hour and you reset your system clock to last years
1118 time, it will still time out after (roughly) and hour. "Roughly" because
1119 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1120 monotonic clock option helps a lot here).
1122 The relative timeouts are calculated relative to the C<ev_now ()>
1123 time. This is usually the right thing as this timestamp refers to the time
1124 of the event triggering whatever timeout you are modifying/starting. If
1125 you suspect event processing to be delayed and you I<need> to base the timeout
1126 on the current time, use something like this to adjust for this:
1128 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1130 The callback is guarenteed to be invoked only when its timeout has passed,
1131 but if multiple timers become ready during the same loop iteration then
1132 order of execution is undefined.
1134 =head3 Watcher-Specific Functions and Data Members
1136 =over 4
1138 =item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1140 =item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1142 Configure the timer to trigger after C<after> seconds. If C<repeat> is
1143 C<0.>, then it will automatically be stopped. If it is positive, then the
1144 timer will automatically be configured to trigger again C<repeat> seconds
1145 later, again, and again, until stopped manually.
1147 The timer itself will do a best-effort at avoiding drift, that is, if you
1148 configure a timer to trigger every 10 seconds, then it will trigger at
1149 exactly 10 second intervals. If, however, your program cannot keep up with
1150 the timer (because it takes longer than those 10 seconds to do stuff) the
1151 timer will not fire more than once per event loop iteration.
1153 =item ev_timer_again (loop)
1155 This will act as if the timer timed out and restart it again if it is
1156 repeating. The exact semantics are:
1158 If the timer is pending, its pending status is cleared.
1160 If the timer is started but nonrepeating, stop it (as if it timed out).
1162 If the timer is repeating, either start it if necessary (with the
1163 C<repeat> value), or reset the running timer to the C<repeat> value.
1165 This sounds a bit complicated, but here is a useful and typical
1166 example: Imagine you have a tcp connection and you want a so-called idle
1167 timeout, that is, you want to be called when there have been, say, 60
1168 seconds of inactivity on the socket. The easiest way to do this is to
1169 configure an C<ev_timer> with a C<repeat> value of C<60> and then call
1170 C<ev_timer_again> each time you successfully read or write some data. If
1171 you go into an idle state where you do not expect data to travel on the
1172 socket, you can C<ev_timer_stop> the timer, and C<ev_timer_again> will
1173 automatically restart it if need be.
1175 That means you can ignore the C<after> value and C<ev_timer_start>
1176 altogether and only ever use the C<repeat> value and C<ev_timer_again>:
1178 ev_timer_init (timer, callback, 0., 5.);
1179 ev_timer_again (loop, timer);
1180 ...
1181 timer->again = 17.;
1182 ev_timer_again (loop, timer);
1183 ...
1184 timer->again = 10.;
1185 ev_timer_again (loop, timer);
1187 This is more slightly efficient then stopping/starting the timer each time
1188 you want to modify its timeout value.
1190 =item ev_tstamp repeat [read-write]
1192 The current C<repeat> value. Will be used each time the watcher times out
1193 or C<ev_timer_again> is called and determines the next timeout (if any),
1194 which is also when any modifications are taken into account.
1196 =back
1198 =head3 Examples
1200 Example: Create a timer that fires after 60 seconds.
1202 static void
1203 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1204 {
1205 .. one minute over, w is actually stopped right here
1206 }
1208 struct ev_timer mytimer;
1209 ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1210 ev_timer_start (loop, &mytimer);
1212 Example: Create a timeout timer that times out after 10 seconds of
1213 inactivity.
1215 static void
1216 timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1217 {
1218 .. ten seconds without any activity
1219 }
1221 struct ev_timer mytimer;
1222 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1223 ev_timer_again (&mytimer); /* start timer */
1224 ev_loop (loop, 0);
1226 // and in some piece of code that gets executed on any "activity":
1227 // reset the timeout to start ticking again at 10 seconds
1228 ev_timer_again (&mytimer);
1231 =head2 C<ev_periodic> - to cron or not to cron?
1233 Periodic watchers are also timers of a kind, but they are very versatile
1234 (and unfortunately a bit complex).
1236 Unlike C<ev_timer>'s, they are not based on real time (or relative time)
1237 but on wallclock time (absolute time). You can tell a periodic watcher
1238 to trigger "at" some specific point in time. For example, if you tell a
1239 periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()
1240 + 10.>) and then reset your system clock to the last year, then it will
1241 take a year to trigger the event (unlike an C<ev_timer>, which would trigger
1242 roughly 10 seconds later).
1244 They can also be used to implement vastly more complex timers, such as
1245 triggering an event on each midnight, local time or other, complicated,
1246 rules.
1248 As with timers, the callback is guarenteed to be invoked only when the
1249 time (C<at>) has been passed, but if multiple periodic timers become ready
1250 during the same loop iteration then order of execution is undefined.
1252 =head3 Watcher-Specific Functions and Data Members
1254 =over 4
1256 =item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
1258 =item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
1260 Lots of arguments, lets sort it out... There are basically three modes of
1261 operation, and we will explain them from simplest to complex:
1263 =over 4
1265 =item * absolute timer (at = time, interval = reschedule_cb = 0)
1267 In this configuration the watcher triggers an event at the wallclock time
1268 C<at> and doesn't repeat. It will not adjust when a time jump occurs,
1269 that is, if it is to be run at January 1st 2011 then it will run when the
1270 system time reaches or surpasses this time.
1272 =item * non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1274 In this mode the watcher will always be scheduled to time out at the next
1275 C<at + N * interval> time (for some integer N, which can also be negative)
1276 and then repeat, regardless of any time jumps.
1278 This can be used to create timers that do not drift with respect to system
1279 time:
1281 ev_periodic_set (&periodic, 0., 3600., 0);
1283 This doesn't mean there will always be 3600 seconds in between triggers,
1284 but only that the the callback will be called when the system time shows a
1285 full hour (UTC), or more correctly, when the system time is evenly divisible
1286 by 3600.
1288 Another way to think about it (for the mathematically inclined) is that
1289 C<ev_periodic> will try to run the callback in this mode at the next possible
1290 time where C<time = at (mod interval)>, regardless of any time jumps.
1292 For numerical stability it is preferable that the C<at> value is near
1293 C<ev_now ()> (the current time), but there is no range requirement for
1294 this value.
1296 =item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1298 In this mode the values for C<interval> and C<at> are both being
1299 ignored. Instead, each time the periodic watcher gets scheduled, the
1300 reschedule callback will be called with the watcher as first, and the
1301 current time as second argument.
1303 NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
1304 ever, or make any event loop modifications>. If you need to stop it,
1305 return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
1306 starting an C<ev_prepare> watcher, which is legal).
1308 Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
1309 ev_tstamp now)>, e.g.:
1311 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
1312 {
1313 return now + 60.;
1314 }
1316 It must return the next time to trigger, based on the passed time value
1317 (that is, the lowest time value larger than to the second argument). It
1318 will usually be called just before the callback will be triggered, but
1319 might be called at other times, too.
1321 NOTE: I<< This callback must always return a time that is later than the
1322 passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.
1324 This can be used to create very complex timers, such as a timer that
1325 triggers on each midnight, local time. To do this, you would calculate the
1326 next midnight after C<now> and return the timestamp value for this. How
1327 you do this is, again, up to you (but it is not trivial, which is the main
1328 reason I omitted it as an example).
1330 =back
1332 =item ev_periodic_again (loop, ev_periodic *)
1334 Simply stops and restarts the periodic watcher again. This is only useful
1335 when you changed some parameters or the reschedule callback would return
1336 a different time than the last time it was called (e.g. in a crond like
1337 program when the crontabs have changed).
1339 =item ev_tstamp offset [read-write]
1341 When repeating, this contains the offset value, otherwise this is the
1342 absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1344 Can be modified any time, but changes only take effect when the periodic
1345 timer fires or C<ev_periodic_again> is being called.
1347 =item ev_tstamp interval [read-write]
1349 The current interval value. Can be modified any time, but changes only
1350 take effect when the periodic timer fires or C<ev_periodic_again> is being
1351 called.
1353 =item ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]
1355 The current reschedule callback, or C<0>, if this functionality is
1356 switched off. Can be changed any time, but changes only take effect when
1357 the periodic timer fires or C<ev_periodic_again> is being called.
1359 =item ev_tstamp at [read-only]
1361 When active, contains the absolute time that the watcher is supposed to
1362 trigger next.
1364 =back
1366 =head3 Examples
1368 Example: Call a callback every hour, or, more precisely, whenever the
1369 system clock is divisible by 3600. The callback invocation times have
1370 potentially a lot of jittering, but good long-term stability.
1372 static void
1373 clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1374 {
1375 ... its now a full hour (UTC, or TAI or whatever your clock follows)
1376 }
1378 struct ev_periodic hourly_tick;
1379 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1380 ev_periodic_start (loop, &hourly_tick);
1382 Example: The same as above, but use a reschedule callback to do it:
1384 #include <math.h>
1386 static ev_tstamp
1387 my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)
1388 {
1389 return fmod (now, 3600.) + 3600.;
1390 }
1392 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1394 Example: Call a callback every hour, starting now:
1396 struct ev_periodic hourly_tick;
1397 ev_periodic_init (&hourly_tick, clock_cb,
1398 fmod (ev_now (loop), 3600.), 3600., 0);
1399 ev_periodic_start (loop, &hourly_tick);
1402 =head2 C<ev_signal> - signal me when a signal gets signalled!
1404 Signal watchers will trigger an event when the process receives a specific
1405 signal one or more times. Even though signals are very asynchronous, libev
1406 will try it's best to deliver signals synchronously, i.e. as part of the
1407 normal event processing, like any other event.
1409 You can configure as many watchers as you like per signal. Only when the
1410 first watcher gets started will libev actually register a signal watcher
1411 with the kernel (thus it coexists with your own signal handlers as long
1412 as you don't register any with libev). Similarly, when the last signal
1413 watcher for a signal is stopped libev will reset the signal handler to
1414 SIG_DFL (regardless of what it was set to before).
1416 =head3 Watcher-Specific Functions and Data Members
1418 =over 4
1420 =item ev_signal_init (ev_signal *, callback, int signum)
1422 =item ev_signal_set (ev_signal *, int signum)
1424 Configures the watcher to trigger on the given signal number (usually one
1425 of the C<SIGxxx> constants).
1427 =item int signum [read-only]
1429 The signal the watcher watches out for.
1431 =back
1434 =head2 C<ev_child> - watch out for process status changes
1436 Child watchers trigger when your process receives a SIGCHLD in response to
1437 some child status changes (most typically when a child of yours dies).
1439 =head3 Watcher-Specific Functions and Data Members
1441 =over 4
1443 =item ev_child_init (ev_child *, callback, int pid, int trace)
1445 =item ev_child_set (ev_child *, int pid, int trace)
1447 Configures the watcher to wait for status changes of process C<pid> (or
1448 I<any> process if C<pid> is specified as C<0>). The callback can look
1449 at the C<rstatus> member of the C<ev_child> watcher structure to see
1450 the status word (use the macros from C<sys/wait.h> and see your systems
1451 C<waitpid> documentation). The C<rpid> member contains the pid of the
1452 process causing the status change. C<trace> must be either C<0> (only
1453 activate the watcher when the process terminates) or C<1> (additionally
1454 activate the watcher when the process is stopped or continued).
1456 =item int pid [read-only]
1458 The process id this watcher watches out for, or C<0>, meaning any process id.
1460 =item int rpid [read-write]
1462 The process id that detected a status change.
1464 =item int rstatus [read-write]
1466 The process exit/trace status caused by C<rpid> (see your systems
1467 C<waitpid> and C<sys/wait.h> documentation for details).
1469 =back
1471 =head3 Examples
1473 Example: Try to exit cleanly on SIGINT and SIGTERM.
1475 static void
1476 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
1477 {
1478 ev_unloop (loop, EVUNLOOP_ALL);
1479 }
1481 struct ev_signal signal_watcher;
1482 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1483 ev_signal_start (loop, &sigint_cb);
1486 =head2 C<ev_stat> - did the file attributes just change?
1488 This watches a filesystem path for attribute changes. That is, it calls
1489 C<stat> regularly (or when the OS says it changed) and sees if it changed
1490 compared to the last time, invoking the callback if it did.
1492 The path does not need to exist: changing from "path exists" to "path does
1493 not exist" is a status change like any other. The condition "path does
1494 not exist" is signified by the C<st_nlink> field being zero (which is
1495 otherwise always forced to be at least one) and all the other fields of
1496 the stat buffer having unspecified contents.
1498 The path I<should> be absolute and I<must not> end in a slash. If it is
1499 relative and your working directory changes, the behaviour is undefined.
1501 Since there is no standard to do this, the portable implementation simply
1502 calls C<stat (2)> regularly on the path to see if it changed somehow. You
1503 can specify a recommended polling interval for this case. If you specify
1504 a polling interval of C<0> (highly recommended!) then a I<suitable,
1505 unspecified default> value will be used (which you can expect to be around
1506 five seconds, although this might change dynamically). Libev will also
1507 impose a minimum interval which is currently around C<0.1>, but thats
1508 usually overkill.
1510 This watcher type is not meant for massive numbers of stat watchers,
1511 as even with OS-supported change notifications, this can be
1512 resource-intensive.
1514 At the time of this writing, only the Linux inotify interface is
1515 implemented (implementing kqueue support is left as an exercise for the
1516 reader). Inotify will be used to give hints only and should not change the
1517 semantics of C<ev_stat> watchers, which means that libev sometimes needs
1518 to fall back to regular polling again even with inotify, but changes are
1519 usually detected immediately, and if the file exists there will be no
1520 polling.
1522 =head3 Inotify
1524 When C<inotify (7)> support has been compiled into libev (generally only
1525 available on Linux) and present at runtime, it will be used to speed up
1526 change detection where possible. The inotify descriptor will be created lazily
1527 when the first C<ev_stat> watcher is being started.
1529 Inotify presense does not change the semantics of C<ev_stat> watchers
1530 except that changes might be detected earlier, and in some cases, to avoid
1531 making regular C<stat> calls. Even in the presense of inotify support
1532 there are many cases where libev has to resort to regular C<stat> polling.
1534 (There is no support for kqueue, as apparently it cannot be used to
1535 implement this functionality, due to the requirement of having a file
1536 descriptor open on the object at all times).
1538 =head3 The special problem of stat time resolution
1540 The C<stat ()> syscall only supports full-second resolution portably, and
1541 even on systems where the resolution is higher, many filesystems still
1542 only support whole seconds.
1544 That means that, if the time is the only thing that changes, you might
1545 miss updates: on the first update, C<ev_stat> detects a change and calls
1546 your callback, which does something. When there is another update within
1547 the same second, C<ev_stat> will be unable to detect it.
1549 The solution to this is to delay acting on a change for a second (or till
1550 the next second boundary), using a roughly one-second delay C<ev_timer>
1551 (C<ev_timer_set (w, 0., 1.01); ev_timer_again (loop, w)>). The C<.01>
1552 is added to work around small timing inconsistencies of some operating
1553 systems.
1555 =head3 Watcher-Specific Functions and Data Members
1557 =over 4
1559 =item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
1561 =item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
1563 Configures the watcher to wait for status changes of the given
1564 C<path>. The C<interval> is a hint on how quickly a change is expected to
1565 be detected and should normally be specified as C<0> to let libev choose
1566 a suitable value. The memory pointed to by C<path> must point to the same
1567 path for as long as the watcher is active.
1569 The callback will be receive C<EV_STAT> when a change was detected,
1570 relative to the attributes at the time the watcher was started (or the
1571 last change was detected).
1573 =item ev_stat_stat (ev_stat *)
1575 Updates the stat buffer immediately with new values. If you change the
1576 watched path in your callback, you could call this fucntion to avoid
1577 detecting this change (while introducing a race condition). Can also be
1578 useful simply to find out the new values.
1580 =item ev_statdata attr [read-only]
1582 The most-recently detected attributes of the file. Although the type is of
1583 C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
1584 suitable for your system. If the C<st_nlink> member is C<0>, then there
1585 was some error while C<stat>ing the file.
1587 =item ev_statdata prev [read-only]
1589 The previous attributes of the file. The callback gets invoked whenever
1590 C<prev> != C<attr>.
1592 =item ev_tstamp interval [read-only]
1594 The specified interval.
1596 =item const char *path [read-only]
1598 The filesystem path that is being watched.
1600 =back
1602 =head3 Examples
1604 Example: Watch C</etc/passwd> for attribute changes.
1606 static void
1607 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
1608 {
1609 /* /etc/passwd changed in some way */
1610 if (w->attr.st_nlink)
1611 {
1612 printf ("passwd current size %ld\n", (long)w->attr.st_size);
1613 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
1614 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
1615 }
1616 else
1617 /* you shalt not abuse printf for puts */
1618 puts ("wow, /etc/passwd is not there, expect problems. "
1619 "if this is windows, they already arrived\n");
1620 }
1622 ...
1623 ev_stat passwd;
1625 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1626 ev_stat_start (loop, &passwd);
1628 Example: Like above, but additionally use a one-second delay so we do not
1629 miss updates (however, frequent updates will delay processing, too, so
1630 one might do the work both on C<ev_stat> callback invocation I<and> on
1631 C<ev_timer> callback invocation).
1633 static ev_stat passwd;
1634 static ev_timer timer;
1636 static void
1637 timer_cb (EV_P_ ev_timer *w, int revents)
1638 {
1639 ev_timer_stop (EV_A_ w);
1641 /* now it's one second after the most recent passwd change */
1642 }
1644 static void
1645 stat_cb (EV_P_ ev_stat *w, int revents)
1646 {
1647 /* reset the one-second timer */
1648 ev_timer_again (EV_A_ &timer);
1649 }
1651 ...
1652 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
1653 ev_stat_start (loop, &passwd);
1654 ev_timer_init (&timer, timer_cb, 0., 1.01);
1657 =head2 C<ev_idle> - when you've got nothing better to do...
1659 Idle watchers trigger events when no other events of the same or higher
1660 priority are pending (prepare, check and other idle watchers do not
1661 count).
1663 That is, as long as your process is busy handling sockets or timeouts
1664 (or even signals, imagine) of the same or higher priority it will not be
1665 triggered. But when your process is idle (or only lower-priority watchers
1666 are pending), the idle watchers are being called once per event loop
1667 iteration - until stopped, that is, or your process receives more events
1668 and becomes busy again with higher priority stuff.
1670 The most noteworthy effect is that as long as any idle watchers are
1671 active, the process will not block when waiting for new events.
1673 Apart from keeping your process non-blocking (which is a useful
1674 effect on its own sometimes), idle watchers are a good place to do
1675 "pseudo-background processing", or delay processing stuff to after the
1676 event loop has handled all outstanding events.
1678 =head3 Watcher-Specific Functions and Data Members
1680 =over 4
1682 =item ev_idle_init (ev_signal *, callback)
1684 Initialises and configures the idle watcher - it has no parameters of any
1685 kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1686 believe me.
1688 =back
1690 =head3 Examples
1692 Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1693 callback, free it. Also, use no error checking, as usual.
1695 static void
1696 idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)
1697 {
1698 free (w);
1699 // now do something you wanted to do when the program has
1700 // no longer anything immediate to do.
1701 }
1703 struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));
1704 ev_idle_init (idle_watcher, idle_cb);
1705 ev_idle_start (loop, idle_cb);
1708 =head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1710 Prepare and check watchers are usually (but not always) used in tandem:
1711 prepare watchers get invoked before the process blocks and check watchers
1712 afterwards.
1714 You I<must not> call C<ev_loop> or similar functions that enter
1715 the current event loop from either C<ev_prepare> or C<ev_check>
1716 watchers. Other loops than the current one are fine, however. The
1717 rationale behind this is that you do not need to check for recursion in
1718 those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
1719 C<ev_check> so if you have one watcher of each kind they will always be
1720 called in pairs bracketing the blocking call.
1722 Their main purpose is to integrate other event mechanisms into libev and
1723 their use is somewhat advanced. This could be used, for example, to track
1724 variable changes, implement your own watchers, integrate net-snmp or a
1725 coroutine library and lots more. They are also occasionally useful if
1726 you cache some data and want to flush it before blocking (for example,
1727 in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
1728 watcher).
1730 This is done by examining in each prepare call which file descriptors need
1731 to be watched by the other library, registering C<ev_io> watchers for
1732 them and starting an C<ev_timer> watcher for any timeouts (many libraries
1733 provide just this functionality). Then, in the check watcher you check for
1734 any events that occured (by checking the pending status of all watchers
1735 and stopping them) and call back into the library. The I/O and timer
1736 callbacks will never actually be called (but must be valid nevertheless,
1737 because you never know, you know?).
1739 As another example, the Perl Coro module uses these hooks to integrate
1740 coroutines into libev programs, by yielding to other active coroutines
1741 during each prepare and only letting the process block if no coroutines
1742 are ready to run (it's actually more complicated: it only runs coroutines
1743 with priority higher than or equal to the event loop and one coroutine
1744 of lower priority, but only once, using idle watchers to keep the event
1745 loop from blocking if lower-priority coroutines are active, thus mapping
1746 low-priority coroutines to idle/background tasks).
1748 It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1749 priority, to ensure that they are being run before any other watchers
1750 after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
1751 too) should not activate ("feed") events into libev. While libev fully
1752 supports this, they will be called before other C<ev_check> watchers
1753 did their job. As C<ev_check> watchers are often used to embed other
1754 (non-libev) event loops those other event loops might be in an unusable
1755 state until their C<ev_check> watcher ran (always remind yourself to
1756 coexist peacefully with others).
1758 =head3 Watcher-Specific Functions and Data Members
1760 =over 4
1762 =item ev_prepare_init (ev_prepare *, callback)
1764 =item ev_check_init (ev_check *, callback)
1766 Initialises and configures the prepare or check watcher - they have no
1767 parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1768 macros, but using them is utterly, utterly and completely pointless.
1770 =back
1772 =head3 Examples
1774 There are a number of principal ways to embed other event loops or modules
1775 into libev. Here are some ideas on how to include libadns into libev
1776 (there is a Perl module named C<EV::ADNS> that does this, which you could
1777 use for an actually working example. Another Perl module named C<EV::Glib>
1778 embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV
1779 into the Glib event loop).
1781 Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1782 and in a check watcher, destroy them and call into libadns. What follows
1783 is pseudo-code only of course. This requires you to either use a low
1784 priority for the check watcher or use C<ev_clear_pending> explicitly, as
1785 the callbacks for the IO/timeout watchers might not have been called yet.
1787 static ev_io iow [nfd];
1788 static ev_timer tw;
1790 static void
1791 io_cb (ev_loop *loop, ev_io *w, int revents)
1792 {
1793 }
1795 // create io watchers for each fd and a timer before blocking
1796 static void
1797 adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
1798 {
1799 int timeout = 3600000;
1800 struct pollfd fds [nfd];
1801 // actual code will need to loop here and realloc etc.
1802 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1804 /* the callback is illegal, but won't be called as we stop during check */
1805 ev_timer_init (&tw, 0, timeout * 1e-3);
1806 ev_timer_start (loop, &tw);
1808 // create one ev_io per pollfd
1809 for (int i = 0; i < nfd; ++i)
1810 {
1811 ev_io_init (iow + i, io_cb, fds [i].fd,
1812 ((fds [i].events & POLLIN ? EV_READ : 0)
1813 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1815 fds [i].revents = 0;
1816 ev_io_start (loop, iow + i);
1817 }
1818 }
1820 // stop all watchers after blocking
1821 static void
1822 adns_check_cb (ev_loop *loop, ev_check *w, int revents)
1823 {
1824 ev_timer_stop (loop, &tw);
1826 for (int i = 0; i < nfd; ++i)
1827 {
1828 // set the relevant poll flags
1829 // could also call adns_processreadable etc. here
1830 struct pollfd *fd = fds + i;
1831 int revents = ev_clear_pending (iow + i);
1832 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1833 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1835 // now stop the watcher
1836 ev_io_stop (loop, iow + i);
1837 }
1839 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1840 }
1842 Method 2: This would be just like method 1, but you run C<adns_afterpoll>
1843 in the prepare watcher and would dispose of the check watcher.
1845 Method 3: If the module to be embedded supports explicit event
1846 notification (adns does), you can also make use of the actual watcher
1847 callbacks, and only destroy/create the watchers in the prepare watcher.
1849 static void
1850 timer_cb (EV_P_ ev_timer *w, int revents)
1851 {
1852 adns_state ads = (adns_state)w->data;
1853 update_now (EV_A);
1855 adns_processtimeouts (ads, &tv_now);
1856 }
1858 static void
1859 io_cb (EV_P_ ev_io *w, int revents)
1860 {
1861 adns_state ads = (adns_state)w->data;
1862 update_now (EV_A);
1864 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1865 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1866 }
1868 // do not ever call adns_afterpoll
1870 Method 4: Do not use a prepare or check watcher because the module you
1871 want to embed is too inflexible to support it. Instead, youc na override
1872 their poll function. The drawback with this solution is that the main
1873 loop is now no longer controllable by EV. The C<Glib::EV> module does
1874 this.
1876 static gint
1877 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1878 {
1879 int got_events = 0;
1881 for (n = 0; n < nfds; ++n)
1882 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1884 if (timeout >= 0)
1885 // create/start timer
1887 // poll
1888 ev_loop (EV_A_ 0);
1890 // stop timer again
1891 if (timeout >= 0)
1892 ev_timer_stop (EV_A_ &to);
1894 // stop io watchers again - their callbacks should have set
1895 for (n = 0; n < nfds; ++n)
1896 ev_io_stop (EV_A_ iow [n]);
1898 return got_events;
1899 }
1902 =head2 C<ev_embed> - when one backend isn't enough...
1904 This is a rather advanced watcher type that lets you embed one event loop
1905 into another (currently only C<ev_io> events are supported in the embedded
1906 loop, other types of watchers might be handled in a delayed or incorrect
1907 fashion and must not be used).
1909 There are primarily two reasons you would want that: work around bugs and
1910 prioritise I/O.
1912 As an example for a bug workaround, the kqueue backend might only support
1913 sockets on some platform, so it is unusable as generic backend, but you
1914 still want to make use of it because you have many sockets and it scales
1915 so nicely. In this case, you would create a kqueue-based loop and embed it
1916 into your default loop (which might use e.g. poll). Overall operation will
1917 be a bit slower because first libev has to poll and then call kevent, but
1918 at least you can use both at what they are best.
1920 As for prioritising I/O: rarely you have the case where some fds have
1921 to be watched and handled very quickly (with low latency), and even
1922 priorities and idle watchers might have too much overhead. In this case
1923 you would put all the high priority stuff in one loop and all the rest in
1924 a second one, and embed the second one in the first.
1926 As long as the watcher is active, the callback will be invoked every time
1927 there might be events pending in the embedded loop. The callback must then
1928 call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
1929 their callbacks (you could also start an idle watcher to give the embedded
1930 loop strictly lower priority for example). You can also set the callback
1931 to C<0>, in which case the embed watcher will automatically execute the
1932 embedded loop sweep.
1934 As long as the watcher is started it will automatically handle events. The
1935 callback will be invoked whenever some events have been handled. You can
1936 set the callback to C<0> to avoid having to specify one if you are not
1937 interested in that.
1939 Also, there have not currently been made special provisions for forking:
1940 when you fork, you not only have to call C<ev_loop_fork> on both loops,
1941 but you will also have to stop and restart any C<ev_embed> watchers
1942 yourself.
1944 Unfortunately, not all backends are embeddable, only the ones returned by
1945 C<ev_embeddable_backends> are, which, unfortunately, does not include any
1946 portable one.
1948 So when you want to use this feature you will always have to be prepared
1949 that you cannot get an embeddable loop. The recommended way to get around
1950 this is to have a separate variables for your embeddable loop, try to
1951 create it, and if that fails, use the normal loop for everything.
1953 =head3 Watcher-Specific Functions and Data Members
1955 =over 4
1957 =item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
1959 =item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
1961 Configures the watcher to embed the given loop, which must be
1962 embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1963 invoked automatically, otherwise it is the responsibility of the callback
1964 to invoke it (it will continue to be called until the sweep has been done,
1965 if you do not want thta, you need to temporarily stop the embed watcher).
1967 =item ev_embed_sweep (loop, ev_embed *)
1969 Make a single, non-blocking sweep over the embedded loop. This works
1970 similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1971 apropriate way for embedded loops.
1973 =item struct ev_loop *other [read-only]
1975 The embedded event loop.
1977 =back
1979 =head3 Examples
1981 Example: Try to get an embeddable event loop and embed it into the default
1982 event loop. If that is not possible, use the default loop. The default
1983 loop is stored in C<loop_hi>, while the mebeddable loop is stored in
1984 C<loop_lo> (which is C<loop_hi> in the acse no embeddable loop can be
1985 used).
1987 struct ev_loop *loop_hi = ev_default_init (0);
1988 struct ev_loop *loop_lo = 0;
1989 struct ev_embed embed;
1991 // see if there is a chance of getting one that works
1992 // (remember that a flags value of 0 means autodetection)
1993 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
1994 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
1995 : 0;
1997 // if we got one, then embed it, otherwise default to loop_hi
1998 if (loop_lo)
1999 {
2000 ev_embed_init (&embed, 0, loop_lo);
2001 ev_embed_start (loop_hi, &embed);
2002 }
2003 else
2004 loop_lo = loop_hi;
2006 Example: Check if kqueue is available but not recommended and create
2007 a kqueue backend for use with sockets (which usually work with any
2008 kqueue implementation). Store the kqueue/socket-only event loop in
2009 C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
2011 struct ev_loop *loop = ev_default_init (0);
2012 struct ev_loop *loop_socket = 0;
2013 struct ev_embed embed;
2015 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2016 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2017 {
2018 ev_embed_init (&embed, 0, loop_socket);
2019 ev_embed_start (loop, &embed);
2020 }
2022 if (!loop_socket)
2023 loop_socket = loop;
2025 // now use loop_socket for all sockets, and loop for everything else
2028 =head2 C<ev_fork> - the audacity to resume the event loop after a fork
2030 Fork watchers are called when a C<fork ()> was detected (usually because
2031 whoever is a good citizen cared to tell libev about it by calling
2032 C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
2033 event loop blocks next and before C<ev_check> watchers are being called,
2034 and only in the child after the fork. If whoever good citizen calling
2035 C<ev_default_fork> cheats and calls it in the wrong process, the fork
2036 handlers will be invoked, too, of course.
2038 =head3 Watcher-Specific Functions and Data Members
2040 =over 4
2042 =item ev_fork_init (ev_signal *, callback)
2044 Initialises and configures the fork watcher - it has no parameters of any
2045 kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
2046 believe me.
2048 =back
2053 There are some other functions of possible interest. Described. Here. Now.
2055 =over 4
2057 =item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
2059 This function combines a simple timer and an I/O watcher, calls your
2060 callback on whichever event happens first and automatically stop both
2061 watchers. This is useful if you want to wait for a single event on an fd
2062 or timeout without having to allocate/configure/start/stop/free one or
2063 more watchers yourself.
2065 If C<fd> is less than 0, then no I/O watcher will be started and events
2066 is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and
2067 C<events> set will be craeted and started.
2069 If C<timeout> is less than 0, then no timeout watcher will be
2070 started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
2071 repeat = 0) will be started. While C<0> is a valid timeout, it is of
2072 dubious value.
2074 The callback has the type C<void (*cb)(int revents, void *arg)> and gets
2075 passed an C<revents> set like normal event callbacks (a combination of
2076 C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
2077 value passed to C<ev_once>:
2079 static void stdin_ready (int revents, void *arg)
2080 {
2081 if (revents & EV_TIMEOUT)
2082 /* doh, nothing entered */;
2083 else if (revents & EV_READ)
2084 /* stdin might have data for us, joy! */;
2085 }
2087 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2089 =item ev_feed_event (ev_loop *, watcher *, int revents)
2091 Feeds the given event set into the event loop, as if the specified event
2092 had happened for the specified watcher (which must be a pointer to an
2093 initialised but not necessarily started event watcher).
2095 =item ev_feed_fd_event (ev_loop *, int fd, int revents)
2097 Feed an event on the given fd, as if a file descriptor backend detected
2098 the given events it.
2100 =item ev_feed_signal_event (ev_loop *loop, int signum)
2102 Feed an event as if the given signal occured (C<loop> must be the default
2103 loop!).
2105 =back
2110 Libev offers a compatibility emulation layer for libevent. It cannot
2111 emulate the internals of libevent, so here are some usage hints:
2113 =over 4
2115 =item * Use it by including <event.h>, as usual.
2117 =item * The following members are fully supported: ev_base, ev_callback,
2118 ev_arg, ev_fd, ev_res, ev_events.
2120 =item * Avoid using ev_flags and the EVLIST_*-macros, while it is
2121 maintained by libev, it does not work exactly the same way as in libevent (consider
2122 it a private API).
2124 =item * Priorities are not currently supported. Initialising priorities
2125 will fail and all watchers will have the same priority, even though there
2126 is an ev_pri field.
2128 =item * Other members are not supported.
2130 =item * The libev emulation is I<not> ABI compatible to libevent, you need
2131 to use the libev header file and library.
2133 =back
2135 =head1 C++ SUPPORT
2137 Libev comes with some simplistic wrapper classes for C++ that mainly allow
2138 you to use some convinience methods to start/stop watchers and also change
2139 the callback model to a model using method callbacks on objects.
2141 To use it,
2143 #include <ev++.h>
2145 This automatically includes F<ev.h> and puts all of its definitions (many
2146 of them macros) into the global namespace. All C++ specific things are
2147 put into the C<ev> namespace. It should support all the same embedding
2148 options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
2150 Care has been taken to keep the overhead low. The only data member the C++
2151 classes add (compared to plain C-style watchers) is the event loop pointer
2152 that the watcher is associated with (or no additional members at all if
2153 you disable C<EV_MULTIPLICITY> when embedding libev).
2155 Currently, functions, and static and non-static member functions can be
2156 used as callbacks. Other types should be easy to add as long as they only
2157 need one additional pointer for context. If you need support for other
2158 types of functors please contact the author (preferably after implementing
2159 it).
2161 Here is a list of things available in the C<ev> namespace:
2163 =over 4
2165 =item C<ev::READ>, C<ev::WRITE> etc.
2167 These are just enum values with the same values as the C<EV_READ> etc.
2168 macros from F<ev.h>.
2170 =item C<ev::tstamp>, C<ev::now>
2172 Aliases to the same types/functions as with the C<ev_> prefix.
2174 =item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
2176 For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
2177 the same name in the C<ev> namespace, with the exception of C<ev_signal>
2178 which is called C<ev::sig> to avoid clashes with the C<signal> macro
2179 defines by many implementations.
2181 All of those classes have these methods:
2183 =over 4
2185 =item ev::TYPE::TYPE ()
2187 =item ev::TYPE::TYPE (struct ev_loop *)
2189 =item ev::TYPE::~TYPE
2191 The constructor (optionally) takes an event loop to associate the watcher
2192 with. If it is omitted, it will use C<EV_DEFAULT>.
2194 The constructor calls C<ev_init> for you, which means you have to call the
2195 C<set> method before starting it.
2197 It will not set a callback, however: You have to call the templated C<set>
2198 method to set a callback before you can start the watcher.
2200 (The reason why you have to use a method is a limitation in C++ which does
2201 not allow explicit template arguments for constructors).
2203 The destructor automatically stops the watcher if it is active.
2205 =item w->set<class, &class::method> (object *)
2207 This method sets the callback method to call. The method has to have a
2208 signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
2209 first argument and the C<revents> as second. The object must be given as
2210 parameter and is stored in the C<data> member of the watcher.
2212 This method synthesizes efficient thunking code to call your method from
2213 the C callback that libev requires. If your compiler can inline your
2214 callback (i.e. it is visible to it at the place of the C<set> call and
2215 your compiler is good :), then the method will be fully inlined into the
2216 thunking function, making it as fast as a direct C callback.
2218 Example: simple class declaration and watcher initialisation
2220 struct myclass
2221 {
2222 void io_cb (ev::io &w, int revents) { }
2223 }
2225 myclass obj;
2226 ev::io iow;
2227 iow.set <myclass, &myclass::io_cb> (&obj);
2229 =item w->set<function> (void *data = 0)
2231 Also sets a callback, but uses a static method or plain function as
2232 callback. The optional C<data> argument will be stored in the watcher's
2233 C<data> member and is free for you to use.
2235 The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2237 See the method-C<set> above for more details.
2239 Example:
2241 static void io_cb (ev::io &w, int revents) { }
2242 iow.set <io_cb> ();
2244 =item w->set (struct ev_loop *)
2246 Associates a different C<struct ev_loop> with this watcher. You can only
2247 do this when the watcher is inactive (and not pending either).
2249 =item w->set ([args])
2251 Basically the same as C<ev_TYPE_set>, with the same args. Must be
2252 called at least once. Unlike the C counterpart, an active watcher gets
2253 automatically stopped and restarted when reconfiguring it with this
2254 method.
2256 =item w->start ()
2258 Starts the watcher. Note that there is no C<loop> argument, as the
2259 constructor already stores the event loop.
2261 =item w->stop ()
2263 Stops the watcher if it is active. Again, no C<loop> argument.
2265 =item w->again () (C<ev::timer>, C<ev::periodic> only)
2267 For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
2268 C<ev_TYPE_again> function.
2270 =item w->sweep () (C<ev::embed> only)
2272 Invokes C<ev_embed_sweep>.
2274 =item w->update () (C<ev::stat> only)
2276 Invokes C<ev_stat_stat>.
2278 =back
2280 =back
2282 Example: Define a class with an IO and idle watcher, start one of them in
2283 the constructor.
2285 class myclass
2286 {
2287 ev::io io; void io_cb (ev::io &w, int revents);
2288 ev:idle idle void idle_cb (ev::idle &w, int revents);
2290 myclass (int fd)
2291 {
2292 io .set <myclass, &myclass::io_cb > (this);
2293 idle.set <myclass, &myclass::idle_cb> (this);
2295 io.start (fd, ev::READ);
2296 }
2297 };
2300 =head1 MACRO MAGIC
2302 Libev can be compiled with a variety of options, the most fundamantal
2303 of which is C<EV_MULTIPLICITY>. This option determines whether (most)
2304 functions and callbacks have an initial C<struct ev_loop *> argument.
2306 To make it easier to write programs that cope with either variant, the
2307 following macros are defined:
2309 =over 4
2311 =item C<EV_A>, C<EV_A_>
2313 This provides the loop I<argument> for functions, if one is required ("ev
2314 loop argument"). The C<EV_A> form is used when this is the sole argument,
2315 C<EV_A_> is used when other arguments are following. Example:
2317 ev_unref (EV_A);
2318 ev_timer_add (EV_A_ watcher);
2319 ev_loop (EV_A_ 0);
2321 It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
2322 which is often provided by the following macro.
2324 =item C<EV_P>, C<EV_P_>
2326 This provides the loop I<parameter> for functions, if one is required ("ev
2327 loop parameter"). The C<EV_P> form is used when this is the sole parameter,
2328 C<EV_P_> is used when other parameters are following. Example:
2330 // this is how ev_unref is being declared
2331 static void ev_unref (EV_P);
2333 // this is how you can declare your typical callback
2334 static void cb (EV_P_ ev_timer *w, int revents)
2336 It declares a parameter C<loop> of type C<struct ev_loop *>, quite
2337 suitable for use with C<EV_A>.
2339 =item C<EV_DEFAULT>, C<EV_DEFAULT_>
2341 Similar to the other two macros, this gives you the value of the default
2342 loop, if multiple loops are supported ("ev loop default").
2344 =back
2346 Example: Declare and initialise a check watcher, utilising the above
2347 macros so it will work regardless of whether multiple loops are supported
2348 or not.
2350 static void
2351 check_cb (EV_P_ ev_timer *w, int revents)
2352 {
2353 ev_check_stop (EV_A_ w);
2354 }
2356 ev_check check;
2357 ev_check_init (&check, check_cb);
2358 ev_check_start (EV_DEFAULT_ &check);
2359 ev_loop (EV_DEFAULT_ 0);
2361 =head1 EMBEDDING
2363 Libev can (and often is) directly embedded into host
2364 applications. Examples of applications that embed it include the Deliantra
2365 Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
2366 and rxvt-unicode.
2368 The goal is to enable you to just copy the necessary files into your
2369 source directory without having to change even a single line in them, so
2370 you can easily upgrade by simply copying (or having a checked-out copy of
2371 libev somewhere in your source tree).
2373 =head2 FILESETS
2375 Depending on what features you need you need to include one or more sets of files
2376 in your app.
2378 =head3 CORE EVENT LOOP
2380 To include only the libev core (all the C<ev_*> functions), with manual
2381 configuration (no autoconf):
2383 #define EV_STANDALONE 1
2384 #include "ev.c"
2386 This will automatically include F<ev.h>, too, and should be done in a
2387 single C source file only to provide the function implementations. To use
2388 it, do the same for F<ev.h> in all files wishing to use this API (best
2389 done by writing a wrapper around F<ev.h> that you can include instead and
2390 where you can put other configuration options):
2392 #define EV_STANDALONE 1
2393 #include "ev.h"
2395 Both header files and implementation files can be compiled with a C++
2396 compiler (at least, thats a stated goal, and breakage will be treated
2397 as a bug).
2399 You need the following files in your source tree, or in a directory
2400 in your include path (e.g. in libev/ when using -Ilibev):
2402 ev.h
2403 ev.c
2404 ev_vars.h
2405 ev_wrap.h
2407 ev_win32.c required on win32 platforms only
2409 ev_select.c only when select backend is enabled (which is enabled by default)
2410 ev_poll.c only when poll backend is enabled (disabled by default)
2411 ev_epoll.c only when the epoll backend is enabled (disabled by default)
2412 ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
2413 ev_port.c only when the solaris port backend is enabled (disabled by default)
2415 F<ev.c> includes the backend files directly when enabled, so you only need
2416 to compile this single file.
2420 To include the libevent compatibility API, also include:
2422 #include "event.c"
2424 in the file including F<ev.c>, and:
2426 #include "event.h"
2428 in the files that want to use the libevent API. This also includes F<ev.h>.
2430 You need the following additional files for this:
2432 event.h
2433 event.c
2437 Instead of using C<EV_STANDALONE=1> and providing your config in
2438 whatever way you want, you can also C<m4_include([libev.m4])> in your
2439 F<> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
2440 include F<config.h> and configure itself accordingly.
2442 For this of course you need the m4 file:
2444 libev.m4
2448 Libev can be configured via a variety of preprocessor symbols you have to define
2449 before including any of its files. The default is not to build for multiplicity
2450 and only include the select backend.
2452 =over 4
2454 =item EV_STANDALONE
2456 Must always be C<1> if you do not use autoconf configuration, which
2457 keeps libev from including F<config.h>, and it also defines dummy
2458 implementations for some libevent functions (such as logging, which is not
2459 supported). It will also not define any of the structs usually found in
2460 F<event.h> that are not directly supported by the libev core alone.
2464 If defined to be C<1>, libev will try to detect the availability of the
2465 monotonic clock option at both compiletime and runtime. Otherwise no use
2466 of the monotonic clock option will be attempted. If you enable this, you
2467 usually have to link against librt or something similar. Enabling it when
2468 the functionality isn't available is safe, though, although you have
2469 to make sure you link against any libraries where the C<clock_gettime>
2470 function is hiding in (often F<-lrt>).
2472 =item EV_USE_REALTIME
2474 If defined to be C<1>, libev will try to detect the availability of the
2475 realtime clock option at compiletime (and assume its availability at
2476 runtime if successful). Otherwise no use of the realtime clock option will
2477 be attempted. This effectively replaces C<gettimeofday> by C<clock_get
2478 (CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
2479 note about libraries in the description of C<EV_USE_MONOTONIC>, though.
2483 If defined to be C<1>, libev will assume that C<nanosleep ()> is available
2484 and will use it for delays. Otherwise it will use C<select ()>.
2486 =item EV_USE_SELECT
2488 If undefined or defined to be C<1>, libev will compile in support for the
2489 C<select>(2) backend. No attempt at autodetection will be done: if no
2490 other method takes over, select will be it. Otherwise the select backend
2491 will not be compiled in.
2495 If defined to C<1>, then the select backend will use the system C<fd_set>
2496 structure. This is useful if libev doesn't compile due to a missing
2497 C<NFDBITS> or C<fd_mask> definition or it misguesses the bitset layout on
2498 exotic systems. This usually limits the range of file descriptors to some
2499 low limit such as 1024 or might have other limitations (winsocket only
2500 allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
2501 influence the size of the C<fd_set> used.
2505 When defined to C<1>, the select backend will assume that
2506 select/socket/connect etc. don't understand file descriptors but
2507 wants osf handles on win32 (this is the case when the select to
2508 be used is the winsock select). This means that it will call
2509 C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2510 it is assumed that all these functions actually work on fds, even
2511 on win32. Should not be defined on non-win32 platforms.
2513 =item EV_FD_TO_WIN32_HANDLE
2515 If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
2516 file descriptors to socket handles. When not defining this symbol (the
2517 default), then libev will call C<_get_osfhandle>, which is usually
2518 correct. In some cases, programs use their own file descriptor management,
2519 in which case they can provide this function to map fds to socket handles.
2521 =item EV_USE_POLL
2523 If defined to be C<1>, libev will compile in support for the C<poll>(2)
2524 backend. Otherwise it will be enabled on non-win32 platforms. It
2525 takes precedence over select.
2527 =item EV_USE_EPOLL
2529 If defined to be C<1>, libev will compile in support for the Linux
2530 C<epoll>(7) backend. Its availability will be detected at runtime,
2531 otherwise another method will be used as fallback. This is the
2532 preferred backend for GNU/Linux systems.
2534 =item EV_USE_KQUEUE
2536 If defined to be C<1>, libev will compile in support for the BSD style
2537 C<kqueue>(2) backend. Its actual availability will be detected at runtime,
2538 otherwise another method will be used as fallback. This is the preferred
2539 backend for BSD and BSD-like systems, although on most BSDs kqueue only
2540 supports some types of fds correctly (the only platform we found that
2541 supports ptys for example was NetBSD), so kqueue might be compiled in, but
2542 not be used unless explicitly requested. The best way to use it is to find
2543 out whether kqueue supports your type of fd properly and use an embedded
2544 kqueue loop.
2546 =item EV_USE_PORT
2548 If defined to be C<1>, libev will compile in support for the Solaris
2549 10 port style backend. Its availability will be detected at runtime,
2550 otherwise another method will be used as fallback. This is the preferred
2551 backend for Solaris 10 systems.
2553 =item EV_USE_DEVPOLL
2555 reserved for future expansion, works like the USE symbols above.
2557 =item EV_USE_INOTIFY
2559 If defined to be C<1>, libev will compile in support for the Linux inotify
2560 interface to speed up C<ev_stat> watchers. Its actual availability will
2561 be detected at runtime.
2563 =item EV_H
2565 The name of the F<ev.h> header file used to include it. The default if
2566 undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2567 used to virtually rename the F<ev.h> header file in case of conflicts.
2569 =item EV_CONFIG_H
2571 If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2572 F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2573 C<EV_H>, above.
2575 =item EV_EVENT_H
2577 Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2578 of how the F<event.h> header can be found, the default is C<"event.h">.
2580 =item EV_PROTOTYPES
2582 If defined to be C<0>, then F<ev.h> will not define any function
2583 prototypes, but still define all the structs and other symbols. This is
2584 occasionally useful if you want to provide your own wrapper functions
2585 around libev functions.
2589 If undefined or defined to C<1>, then all event-loop-specific functions
2590 will have the C<struct ev_loop *> as first argument, and you can create
2591 additional independent event loops. Otherwise there will be no support
2592 for multiple event loops and there is no first event loop pointer
2593 argument. Instead, all functions act on the single default loop.
2595 =item EV_MINPRI
2597 =item EV_MAXPRI
2599 The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
2600 C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
2601 provide for more priorities by overriding those symbols (usually defined
2602 to be C<-2> and C<2>, respectively).
2604 When doing priority-based operations, libev usually has to linearly search
2605 all the priorities, so having many of them (hundreds) uses a lot of space
2606 and time, so using the defaults of five priorities (-2 .. +2) is usually
2607 fine.
2609 If your embedding app does not need any priorities, defining these both to
2610 C<0> will save some memory and cpu.
2614 If undefined or defined to be C<1>, then periodic timers are supported. If
2615 defined to be C<0>, then they are not. Disabling them saves a few kB of
2616 code.
2618 =item EV_IDLE_ENABLE
2620 If undefined or defined to be C<1>, then idle watchers are supported. If
2621 defined to be C<0>, then they are not. Disabling them saves a few kB of
2622 code.
2624 =item EV_EMBED_ENABLE
2626 If undefined or defined to be C<1>, then embed watchers are supported. If
2627 defined to be C<0>, then they are not.
2629 =item EV_STAT_ENABLE
2631 If undefined or defined to be C<1>, then stat watchers are supported. If
2632 defined to be C<0>, then they are not.
2634 =item EV_FORK_ENABLE
2636 If undefined or defined to be C<1>, then fork watchers are supported. If
2637 defined to be C<0>, then they are not.
2639 =item EV_MINIMAL
2641 If you need to shave off some kilobytes of code at the expense of some
2642 speed, define this symbol to C<1>. Currently only used for gcc to override
2643 some inlining decisions, saves roughly 30% codesize of amd64.
2645 =item EV_PID_HASHSIZE
2647 C<ev_child> watchers use a small hash table to distribute workload by
2648 pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
2649 than enough. If you need to manage thousands of children you might want to
2650 increase this value (I<must> be a power of two).
2654 C<ev_stat> watchers use a small hash table to distribute workload by
2655 inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2656 usually more than enough. If you need to manage thousands of C<ev_stat>
2657 watchers you might want to increase this value (I<must> be a power of
2658 two).
2660 =item EV_COMMON
2662 By default, all watchers have a C<void *data> member. By redefining
2663 this macro to a something else you can include more and other types of
2664 members. You have to define it each time you include one of the files,
2665 though, and it must be identical each time.
2667 For example, the perl EV module uses something like this:
2669 #define EV_COMMON \
2670 SV *self; /* contains this struct */ \
2671 SV *cb_sv, *fh /* note no trailing ";" */
2673 =item EV_CB_DECLARE (type)
2675 =item EV_CB_INVOKE (watcher, revents)
2677 =item ev_set_cb (ev, cb)
2679 Can be used to change the callback member declaration in each watcher,
2680 and the way callbacks are invoked and set. Must expand to a struct member
2681 definition and a statement, respectively. See the F<ev.h> header file for
2682 their default definitions. One possible use for overriding these is to
2683 avoid the C<struct ev_loop *> as first argument in all cases, or to use
2684 method calls instead of plain function calls in C++.
2688 If you need to re-export the API (e.g. via a dll) and you need a list of
2689 exported symbols, you can use the provided F<Symbol.*> files which list
2690 all public symbols, one per line:
2692 Symbols.ev for libev proper
2693 Symbols.event for the libevent emulation
2695 This can also be used to rename all public symbols to avoid clashes with
2696 multiple versions of libev linked together (which is obviously bad in
2697 itself, but sometimes it is inconvinient to avoid this).
2699 A sed command like this will create wrapper C<#define>'s that you need to
2700 include before including F<ev.h>:
2702 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
2704 This would create a file F<wrap.h> which essentially looks like this:
2706 #define ev_backend myprefix_ev_backend
2707 #define ev_check_start myprefix_ev_check_start
2708 #define ev_check_stop myprefix_ev_check_stop
2709 ...
2711 =head2 EXAMPLES
2713 For a real-world example of a program the includes libev
2714 verbatim, you can have a look at the EV perl module
2715 (L<>). It has the libev files in
2716 the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
2717 interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
2718 will be compiled. It is pretty complex because it provides its own header
2719 file.
2721 The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
2722 that everybody includes and which overrides some configure choices:
2724 #define EV_MINIMAL 1
2725 #define EV_USE_POLL 0
2726 #define EV_MULTIPLICITY 0
2727 #define EV_PERIODIC_ENABLE 0
2728 #define EV_STAT_ENABLE 0
2729 #define EV_FORK_ENABLE 0
2730 #define EV_CONFIG_H <config.h>
2731 #define EV_MINPRI 0
2732 #define EV_MAXPRI 0
2734 #include "ev++.h"
2736 And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
2738 #include "ev_cpp.h"
2739 #include "ev.c"
2742 =head1 COMPLEXITIES
2744 In this section the complexities of (many of) the algorithms used inside
2745 libev will be explained. For complexity discussions about backends see the
2746 documentation for C<ev_default_init>.
2748 All of the following are about amortised time: If an array needs to be
2749 extended, libev needs to realloc and move the whole array, but this
2750 happens asymptotically never with higher number of elements, so O(1) might
2751 mean it might do a lengthy realloc operation in rare cases, but on average
2752 it is much faster and asymptotically approaches constant time.
2754 =over 4
2756 =item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
2758 This means that, when you have a watcher that triggers in one hour and
2759 there are 100 watchers that would trigger before that then inserting will
2760 have to skip roughly seven (C<ld 100>) of these watchers.
2762 =item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2764 That means that changing a timer costs less than removing/adding them
2765 as only the relative motion in the event queue has to be paid for.
2767 =item Starting io/check/prepare/idle/signal/child watchers: O(1)
2769 These just add the watcher into an array or at the head of a list.
2771 =item Stopping check/prepare/idle watchers: O(1)
2773 =item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2775 These watchers are stored in lists then need to be walked to find the
2776 correct watcher to remove. The lists are usually short (you don't usually
2777 have many watchers waiting for the same fd or signal).
2779 =item Finding the next timer in each loop iteration: O(1)
2781 By virtue of using a binary heap, the next timer is always found at the
2782 beginning of the storage array.
2784 =item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
2786 A change means an I/O watcher gets started or stopped, which requires
2787 libev to recalculate its status (and possibly tell the kernel, depending
2788 on backend and wether C<ev_io_set> was used).
2790 =item Activating one watcher (putting it into the pending state): O(1)
2792 =item Priority handling: O(number_of_priorities)
2794 Priorities are implemented by allocating some space for each
2795 priority. When doing priority-based operations, libev usually has to
2796 linearly search all the priorities, but starting/stopping and activating
2797 watchers becomes O(1) w.r.t. prioritiy handling.
2799 =back
2802 =head1 Win32 platform limitations and workarounds
2804 Win32 doesn't support any of the standards (e.g. POSIX) that libev
2805 requires, and its I/O model is fundamentally incompatible with the POSIX
2806 model. Libev still offers limited functionality on this platform in
2807 the form of the C<EVBACKEND_SELECT> backend, and only supports socket
2808 descriptors. This only applies when using Win32 natively, not when using
2809 e.g. cygwin.
2811 There is no supported compilation method available on windows except
2812 embedding it into other applications.
2814 Due to the many, low, and arbitrary limits on the win32 platform and the
2815 abysmal performance of winsockets, using a large number of sockets is not
2816 recommended (and not reasonable). If your program needs to use more than
2817 a hundred or so sockets, then likely it needs to use a totally different
2818 implementation for windows, as libev offers the POSIX model, which cannot
2819 be implemented efficiently on windows (microsoft monopoly games).
2821 =over 4
2823 =item The winsocket select function
2825 The winsocket C<select> function doesn't follow POSIX in that it requires
2826 socket I<handles> and not socket I<file descriptors>. This makes select
2827 very inefficient, and also requires a mapping from file descriptors
2828 to socket handles. See the discussion of the C<EV_SELECT_USE_FD_SET>,
2829 C<EV_SELECT_IS_WINSOCKET> and C<EV_FD_TO_WIN32_HANDLE> preprocessor
2830 symbols for more info.
2832 The configuration for a "naked" win32 using the microsoft runtime
2833 libraries and raw winsocket select is:
2835 #define EV_USE_SELECT 1
2836 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
2838 Note that winsockets handling of fd sets is O(n), so you can easily get a
2839 complexity in the O(n²) range when using win32.
2841 =item Limited number of file descriptors
2843 Windows has numerous arbitrary (and low) limits on things. Early versions
2844 of winsocket's select only supported waiting for a max. of C<64> handles
2845 (probably owning to the fact that all windows kernels can only wait for
2846 C<64> things at the same time internally; microsoft recommends spawning a
2847 chain of threads and wait for 63 handles and the previous thread in each).
2849 Newer versions support more handles, but you need to define C<FD_SETSIZE>
2850 to some high number (e.g. C<2048>) before compiling the winsocket select
2851 call (which might be in libev or elsewhere, for example, perl does its own
2852 select emulation on windows).
2854 Another limit is the number of file descriptors in the microsoft runtime
2855 libraries, which by default is C<64> (there must be a hidden I<64> fetish
2856 or something like this inside microsoft). You can increase this by calling
2857 C<_setmaxstdio>, which can increase this limit to C<2048> (another
2858 arbitrary limit), but is broken in many versions of the microsoft runtime
2859 libraries.
2861 This might get you to about C<512> or C<2048> sockets (depending on
2862 windows version and/or the phase of the moon). To get more, you need to
2863 wrap all I/O functions and provide your own fd management, but the cost of
2864 calling select (O(n²)) will likely make this unworkable.
2866 =back
2869 =head1 AUTHOR
2871 Marc Lehmann <>.