ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev.pod
Revision: 1.127
Committed: Fri Feb 1 13:47:08 2008 UTC (16 years, 3 months ago) by root
Branch: MAIN
Changes since 1.126: +2 -1 lines
Log Message:
*** empty log message ***

File Contents

# User Rev Content
1 root 1.1 =head1 NAME
2    
3     libev - a high performance full-featured event loop written in C
4    
5     =head1 SYNOPSIS
6    
7     #include <ev.h>
8    
9 root 1.105 =head2 EXAMPLE PROGRAM
10 root 1.54
11     #include <ev.h>
12    
13 root 1.53 ev_io stdin_watcher;
14     ev_timer timeout_watcher;
15    
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     }
24    
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     }
31    
32     int
33     main (void)
34     {
35     struct ev_loop *loop = ev_default_loop (0);
36    
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);
40    
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);
44    
45     /* loop till timeout or data ready */
46     ev_loop (loop, 0);
47    
48     return 0;
49     }
50    
51 root 1.1 =head1 DESCRIPTION
52    
53 root 1.69 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<http://cvs.schmorp.de/libev/ev.html>.
56    
57 root 1.1 Libev is an event loop: you register interest in certain events (such as a
58 root 1.92 file descriptor being readable or a timeout occurring), and it will manage
59 root 1.4 these event sources and provide your program with events.
60 root 1.1
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.
64    
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.
69    
70 root 1.105 =head2 FEATURES
71 root 1.1
72 root 1.58 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 root 1.54 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>).
82    
83     It also is quite fast (see this
84     L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
85     for example).
86 root 1.1
87 root 1.105 =head2 CONVENTIONS
88 root 1.1
89 root 1.54 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.
95 root 1.1
96 root 1.105 =head2 TIME REPRESENTATION
97 root 1.1
98 root 1.2 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 root 1.1 called C<ev_tstamp>, which is what you should use too. It usually aliases
102 root 1.34 to the C<double> type in C, and when you need to do any calculations on
103 root 1.86 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.
106 root 1.34
107 root 1.17 =head1 GLOBAL FUNCTIONS
108    
109 root 1.18 These functions can be called anytime, even before initialising the
110     library in any way.
111    
112 root 1.1 =over 4
113    
114     =item ev_tstamp ev_time ()
115    
116 root 1.26 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.
119 root 1.1
120 root 1.97 =item ev_sleep (ev_tstamp interval)
121    
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 ()>.
125    
126 root 1.1 =item int ev_version_major ()
127    
128     =item int ev_version_minor ()
129    
130 root 1.80 You can find out the major and minor ABI version numbers of the library
131 root 1.1 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.
135    
136 root 1.80 These version numbers refer to the ABI version of the library, not the
137     release version.
138 root 1.79
139 root 1.9 Usually, it's a good idea to terminate if the major versions mismatch,
140 root 1.79 as this indicates an incompatible change. Minor versions are usually
141 root 1.1 compatible to older versions, so a larger minor version alone is usually
142     not a problem.
143    
144 root 1.54 Example: Make sure we haven't accidentally been linked against the wrong
145     version.
146 root 1.34
147     assert (("libev version mismatch",
148     ev_version_major () == EV_VERSION_MAJOR
149     && ev_version_minor () >= EV_VERSION_MINOR));
150    
151 root 1.31 =item unsigned int ev_supported_backends ()
152    
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.
157    
158 root 1.34 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
160    
161     assert (("sorry, no epoll, no sex",
162     ev_supported_backends () & EVBACKEND_EPOLL));
163    
164 root 1.31 =item unsigned int ev_recommended_backends ()
165    
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 root 1.33 libev will probe for if you specify no backends explicitly.
172 root 1.31
173 root 1.35 =item unsigned int ev_embeddable_backends ()
174    
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.
180    
181     See the description of C<ev_embed> watchers for more info.
182    
183 root 1.59 =item ev_set_allocator (void *(*cb)(void *ptr, long size))
184 root 1.1
185 root 1.59 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.
191 root 1.1
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.
195    
196 root 1.54 Example: Replace the libev allocator with one that waits a bit and then
197     retries).
198 root 1.34
199     static void *
200 root 1.52 persistent_realloc (void *ptr, size_t size)
201 root 1.34 {
202     for (;;)
203     {
204     void *newptr = realloc (ptr, size);
205    
206     if (newptr)
207     return newptr;
208    
209     sleep (60);
210     }
211     }
212    
213     ...
214     ev_set_allocator (persistent_realloc);
215    
216 root 1.1 =item ev_set_syserr_cb (void (*cb)(const char *msg));
217    
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 root 1.7 matter what, when it returns. That is, libev will generally retry the
223 root 1.1 requested operation, or, if the condition doesn't go away, do bad stuff
224     (such as abort).
225    
226 root 1.54 Example: This is basically the same thing that libev does internally, too.
227 root 1.34
228     static void
229     fatal_error (const char *msg)
230     {
231     perror (msg);
232     abort ();
233     }
234    
235     ...
236     ev_set_syserr_cb (fatal_error);
237    
238 root 1.1 =back
239    
240     =head1 FUNCTIONS CONTROLLING THE EVENT LOOP
241    
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.
245    
246     If you use threads, a common model is to run the default event loop
247 root 1.17 in your main thread (or in a separate thread) and for each thread you
248 root 1.7 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 root 1.9 done correctly, because it's hideous and inefficient).
252 root 1.1
253     =over 4
254    
255     =item struct ev_loop *ev_default_loop (unsigned int flags)
256    
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 root 1.31 flags. If that is troubling you, check C<ev_backend ()> afterwards).
261 root 1.1
262     If you don't know what event loop to use, use the one returned from this
263     function.
264    
265 root 1.118 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>.
271    
272 root 1.1 The flags argument can be used to specify special behaviour or specific
273 root 1.33 backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
274 root 1.1
275 root 1.33 The following flags are supported:
276 root 1.1
277     =over 4
278    
279 root 1.10 =item C<EVFLAG_AUTO>
280 root 1.1
281 root 1.9 The default flags value. Use this if you have no clue (it's the right
282 root 1.1 thing, believe me).
283    
284 root 1.10 =item C<EVFLAG_NOENV>
285 root 1.1
286 root 1.8 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.
292 root 1.1
293 root 1.62 =item C<EVFLAG_FORKCHECK>
294    
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.
298    
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 ayin 1.65 iterations and little real work, but is usually not noticeable (on my
302 root 1.62 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).
305    
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.
309    
310     This flag setting cannot be overriden or specified in the C<LIBEV_FLAGS>
311     environment variable.
312    
313 root 1.31 =item C<EVBACKEND_SELECT> (value 1, portable select backend)
314 root 1.1
315 root 1.29 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 root 1.102 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.
320    
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.
327 root 1.1
328 root 1.31 =item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
329 root 1.1
330 root 1.102 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.
336 root 1.1
337 root 1.31 =item C<EVBACKEND_EPOLL> (value 4, Linux)
338 root 1.1
339 root 1.29 For few fds, this backend is a bit little slower than poll and select,
340 root 1.94 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 ayin 1.96 cases and rewiring a syscall per fd change, no fork support and bad
345 root 1.102 support for dup.
346 root 1.1
347 root 1.94 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 root 1.29 (because the fd could point to a different file description now), so its
350 root 1.94 best to avoid that. Also, C<dup ()>'ed file descriptors might not work
351     very well if you register events for both fds.
352 root 1.29
353 root 1.32 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.
356    
357 root 1.102 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.
360    
361     While nominally embeddeble in other event loops, this feature is broken in
362     all kernel versions tested so far.
363    
364 root 1.31 =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
365 root 1.29
366     Kqueue deserves special mention, as at the time of this writing, it
367 root 1.100 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 root 1.33 unless you explicitly specify it explicitly in the flags (i.e. using
371 root 1.94 C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
372     system like NetBSD.
373 root 1.29
374 root 1.100 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.
377    
378 root 1.29 It scales in the same way as the epoll backend, but the interface to the
379 root 1.100 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.
384 root 1.29
385 root 1.102 This backend usually performs well under most conditions.
386    
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.
393    
394 root 1.31 =item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
395 root 1.29
396 root 1.102 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.
400 root 1.29
401 root 1.31 =item C<EVBACKEND_PORT> (value 32, Solaris 10)
402 root 1.29
403 root 1.94 This uses the Solaris 10 event port mechanism. As with everything on Solaris,
404 root 1.29 it's really slow, but it still scales very well (O(active_fds)).
405    
406 root 1.94 Please note that solaris event ports can deliver a lot of spurious
407 root 1.32 notifications, so you need to use non-blocking I/O or other means to avoid
408     blocking when no data (or space) is available.
409    
410 root 1.102 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.
414    
415 root 1.117 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.
418    
419 root 1.31 =item C<EVBACKEND_ALL>
420 root 1.29
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
423 root 1.31 C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
424 root 1.1
425 root 1.102 It is definitely not recommended to use this flag.
426    
427 root 1.1 =back
428    
429 root 1.29 If one or more of these are ored into the flags value, then only these
430 root 1.117 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.
432 root 1.29
433 root 1.33 The most typical usage is like this:
434    
435     if (!ev_default_loop (0))
436     fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
437    
438     Restrict libev to the select and poll backends, and do not allow
439     environment settings to be taken into account:
440    
441     ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
442    
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):
446    
447     ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
448    
449 root 1.1 =item struct ev_loop *ev_loop_new (unsigned int flags)
450    
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).
455    
456 root 1.54 Example: Try to create a event loop that uses epoll and nothing else.
457 root 1.34
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");
461    
462 root 1.1 =item ev_default_destroy ()
463    
464     Destroys the default loop again (frees all memory and kernel state
465 root 1.37 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 root 1.87 the easiest thing, you can just ignore the watchers and/or C<free ()> them
470 root 1.37 for example).
471 root 1.1
472 ayin 1.88 Note that certain global state, such as signal state, will not be freed by
473 root 1.87 this function, and related watchers (such as signal and child watchers)
474     would need to be stopped manually.
475    
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>).
480    
481 root 1.1 =item ev_loop_destroy (loop)
482    
483     Like C<ev_default_destroy>, but destroys an event loop created by an
484     earlier call to C<ev_loop_new>.
485    
486     =item ev_default_fork ()
487    
488 root 1.119 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.
494    
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.
498 root 1.1
499 root 1.9 The function itself is quite fast and it's usually not a problem to call
500 root 1.1 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>:
502    
503     pthread_atfork (0, 0, ev_default_fork);
504    
505     =item ev_loop_fork (loop)
506    
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.
510    
511 root 1.66 =item unsigned int ev_loop_count (loop)
512    
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.
516    
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.
520    
521 root 1.31 =item unsigned int ev_backend (loop)
522 root 1.1
523 root 1.31 Returns one of the C<EVBACKEND_*> flags indicating the event backend in
524 root 1.1 use.
525    
526 root 1.9 =item ev_tstamp ev_now (loop)
527 root 1.1
528     Returns the current "event loop time", which is the time the event loop
529 root 1.34 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 root 1.92 event occurring (or more correctly, libev finding out about it).
533 root 1.1
534     =item ev_loop (loop, int flags)
535    
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.
539    
540 root 1.33 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.
542 root 1.1
543 root 1.34 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.
548    
549 root 1.1 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 root 1.9 case there are no events and will return after one iteration of the loop.
552 root 1.1
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 root 1.9 your process until at least one new event arrives, and will return after
556 root 1.33 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.
560    
561     Here are the gory details of what C<ev_loop> does:
562    
563 root 1.77 - Before the first iteration, call any pending watchers.
564 root 1.113 * 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 root 1.33 - 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 root 1.113 - 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 root 1.33 - 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 root 1.113 - 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 *.
587 root 1.27
588 root 1.114 Example: Queue some jobs and then loop until no events are outstanding
589 root 1.34 anymore.
590    
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!
595    
596 root 1.1 =item ev_unloop (loop, how)
597    
598 root 1.9 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 root 1.25 C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
601 root 1.9 C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
602 root 1.1
603 root 1.115 This "unloop state" will be cleared when entering C<ev_loop> again.
604    
605 root 1.1 =item ev_ref (loop)
606    
607     =item ev_unref (loop)
608    
609 root 1.9 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 root 1.116 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).
621 root 1.1
622 root 1.54 Example: Create a signal watcher, but keep it from keeping C<ev_loop>
623 root 1.34 running when nothing else is active.
624    
625 root 1.54 struct ev_signal exitsig;
626 root 1.34 ev_signal_init (&exitsig, sig_cb, SIGINT);
627 root 1.54 ev_signal_start (loop, &exitsig);
628     evf_unref (loop);
629 root 1.34
630 root 1.54 Example: For some weird reason, unregister the above signal handler again.
631 root 1.34
632 root 1.54 ev_ref (loop);
633     ev_signal_stop (loop, &exitsig);
634 root 1.34
635 root 1.97 =item ev_set_io_collect_interval (loop, ev_tstamp interval)
636    
637     =item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
638    
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.
642    
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.
646    
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.
652    
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 ayin 1.101 C<ev_timer>) will be not affected. Setting this to a non-null value will
657 root 1.99 introduce an additional C<ev_sleep ()> call into most loop iterations.
658 root 1.97
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 root 1.99 will not be affected. Setting this to a non-null value will not introduce
663     any overhead in libev.
664 root 1.97
665 root 1.98 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.
670 root 1.97
671 root 1.1 =back
672    
673 root 1.42
674 root 1.1 =head1 ANATOMY OF A WATCHER
675    
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 root 1.10 become readable, you would create an C<ev_io> watcher for that:
679 root 1.1
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     }
685    
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);
692    
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).
696    
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).
702    
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, ...) >>.
707    
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 *) >>.
712    
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 root 1.36 reinitialise it or call its C<set> macro.
716 root 1.1
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.
720    
721 root 1.14 The received events usually include a single bit per event type received
722 root 1.1 (you can receive multiple events at the same time). The possible bit masks
723     are:
724    
725     =over 4
726    
727 root 1.10 =item C<EV_READ>
728 root 1.1
729 root 1.10 =item C<EV_WRITE>
730 root 1.1
731 root 1.10 The file descriptor in the C<ev_io> watcher has become readable and/or
732 root 1.1 writable.
733    
734 root 1.10 =item C<EV_TIMEOUT>
735 root 1.1
736 root 1.10 The C<ev_timer> watcher has timed out.
737 root 1.1
738 root 1.10 =item C<EV_PERIODIC>
739 root 1.1
740 root 1.10 The C<ev_periodic> watcher has timed out.
741 root 1.1
742 root 1.10 =item C<EV_SIGNAL>
743 root 1.1
744 root 1.10 The signal specified in the C<ev_signal> watcher has been received by a thread.
745 root 1.1
746 root 1.10 =item C<EV_CHILD>
747 root 1.1
748 root 1.10 The pid specified in the C<ev_child> watcher has received a status change.
749 root 1.1
750 root 1.48 =item C<EV_STAT>
751    
752     The path specified in the C<ev_stat> watcher changed its attributes somehow.
753    
754 root 1.10 =item C<EV_IDLE>
755 root 1.1
756 root 1.10 The C<ev_idle> watcher has determined that you have nothing better to do.
757 root 1.1
758 root 1.10 =item C<EV_PREPARE>
759 root 1.1
760 root 1.10 =item C<EV_CHECK>
761 root 1.1
762 root 1.10 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 root 1.1 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 root 1.10 (for example, a C<ev_prepare> watcher might start an idle watcher to keep
768 root 1.1 C<ev_loop> from blocking).
769    
770 root 1.50 =item C<EV_EMBED>
771    
772     The embedded event loop specified in the C<ev_embed> watcher needs attention.
773    
774     =item C<EV_FORK>
775    
776     The event loop has been resumed in the child process after fork (see
777     C<ev_fork>).
778    
779 root 1.122 =item C<EV_ASYNC>
780    
781     The given async watcher has been asynchronously notified (see C<ev_async>).
782    
783 root 1.10 =item C<EV_ERROR>
784 root 1.1
785     An unspecified error has occured, the watcher has been stopped. This might
786     happen because the watcher could not be properly started because libev
787     ran out of memory, a file descriptor was found to be closed or any other
788     problem. You best act on it by reporting the problem and somehow coping
789     with the watcher being stopped.
790    
791     Libev will usually signal a few "dummy" events together with an error,
792     for example it might indicate that a fd is readable or writable, and if
793     your callbacks is well-written it can just attempt the operation and cope
794     with the error from read() or write(). This will not work in multithreaded
795     programs, though, so beware.
796    
797     =back
798    
799 root 1.42 =head2 GENERIC WATCHER FUNCTIONS
800 root 1.36
801     In the following description, C<TYPE> stands for the watcher type,
802     e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
803    
804     =over 4
805    
806     =item C<ev_init> (ev_TYPE *watcher, callback)
807    
808     This macro initialises the generic portion of a watcher. The contents
809     of the watcher object can be arbitrary (so C<malloc> will do). Only
810     the generic parts of the watcher are initialised, you I<need> to call
811     the type-specific C<ev_TYPE_set> macro afterwards to initialise the
812     type-specific parts. For each type there is also a C<ev_TYPE_init> macro
813     which rolls both calls into one.
814    
815     You can reinitialise a watcher at any time as long as it has been stopped
816     (or never started) and there are no pending events outstanding.
817    
818 root 1.42 The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,
819 root 1.36 int revents)>.
820    
821     =item C<ev_TYPE_set> (ev_TYPE *, [args])
822    
823     This macro initialises the type-specific parts of a watcher. You need to
824     call C<ev_init> at least once before you call this macro, but you can
825     call C<ev_TYPE_set> any number of times. You must not, however, call this
826     macro on a watcher that is active (it can be pending, however, which is a
827     difference to the C<ev_init> macro).
828    
829     Although some watcher types do not have type-specific arguments
830     (e.g. C<ev_prepare>) you still need to call its C<set> macro.
831    
832     =item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
833    
834     This convinience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
835     calls into a single call. This is the most convinient method to initialise
836     a watcher. The same limitations apply, of course.
837    
838     =item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
839    
840     Starts (activates) the given watcher. Only active watchers will receive
841     events. If the watcher is already active nothing will happen.
842    
843     =item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
844    
845     Stops the given watcher again (if active) and clears the pending
846     status. It is possible that stopped watchers are pending (for example,
847     non-repeating timers are being stopped when they become pending), but
848     C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If
849     you want to free or reuse the memory used by the watcher it is therefore a
850     good idea to always call its C<ev_TYPE_stop> function.
851    
852     =item bool ev_is_active (ev_TYPE *watcher)
853    
854     Returns a true value iff the watcher is active (i.e. it has been started
855     and not yet been stopped). As long as a watcher is active you must not modify
856     it.
857    
858     =item bool ev_is_pending (ev_TYPE *watcher)
859    
860     Returns a true value iff the watcher is pending, (i.e. it has outstanding
861     events but its callback has not yet been invoked). As long as a watcher
862     is pending (but not active) you must not call an init function on it (but
863 root 1.73 C<ev_TYPE_set> is safe), you must not change its priority, and you must
864     make sure the watcher is available to libev (e.g. you cannot C<free ()>
865     it).
866 root 1.36
867 root 1.55 =item callback ev_cb (ev_TYPE *watcher)
868 root 1.36
869     Returns the callback currently set on the watcher.
870    
871     =item ev_cb_set (ev_TYPE *watcher, callback)
872    
873     Change the callback. You can change the callback at virtually any time
874     (modulo threads).
875    
876 root 1.67 =item ev_set_priority (ev_TYPE *watcher, priority)
877    
878     =item int ev_priority (ev_TYPE *watcher)
879    
880     Set and query the priority of the watcher. The priority is a small
881     integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
882     (default: C<-2>). Pending watchers with higher priority will be invoked
883     before watchers with lower priority, but priority will not keep watchers
884     from being executed (except for C<ev_idle> watchers).
885    
886     This means that priorities are I<only> used for ordering callback
887     invocation after new events have been received. This is useful, for
888     example, to reduce latency after idling, or more often, to bind two
889     watchers on the same event and make sure one is called first.
890    
891     If you need to suppress invocation when higher priority events are pending
892     you need to look at C<ev_idle> watchers, which provide this functionality.
893    
894 root 1.73 You I<must not> change the priority of a watcher as long as it is active or
895     pending.
896    
897 root 1.67 The default priority used by watchers when no priority has been set is
898     always C<0>, which is supposed to not be too high and not be too low :).
899    
900     Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
901     fine, as long as you do not mind that the priority value you query might
902     or might not have been adjusted to be within valid range.
903    
904 root 1.74 =item ev_invoke (loop, ev_TYPE *watcher, int revents)
905    
906     Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
907     C<loop> nor C<revents> need to be valid as long as the watcher callback
908     can deal with that fact.
909    
910     =item int ev_clear_pending (loop, ev_TYPE *watcher)
911    
912     If the watcher is pending, this function returns clears its pending status
913     and returns its C<revents> bitset (as if its callback was invoked). If the
914     watcher isn't pending it does nothing and returns C<0>.
915    
916 root 1.36 =back
917    
918    
919 root 1.1 =head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
920    
921     Each watcher has, by default, a member C<void *data> that you can change
922 root 1.14 and read at any time, libev will completely ignore it. This can be used
923 root 1.1 to associate arbitrary data with your watcher. If you need more data and
924     don't want to allocate memory and store a pointer to it in that data
925     member, you can also "subclass" the watcher type and provide your own
926     data:
927    
928     struct my_io
929     {
930     struct ev_io io;
931     int otherfd;
932     void *somedata;
933     struct whatever *mostinteresting;
934     }
935    
936     And since your callback will be called with a pointer to the watcher, you
937     can cast it back to your own type:
938    
939     static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
940     {
941     struct my_io *w = (struct my_io *)w_;
942     ...
943     }
944    
945 root 1.55 More interesting and less C-conformant ways of casting your callback type
946     instead have been omitted.
947    
948     Another common scenario is having some data structure with multiple
949     watchers:
950    
951     struct my_biggy
952     {
953     int some_data;
954     ev_timer t1;
955     ev_timer t2;
956     }
957    
958     In this case getting the pointer to C<my_biggy> is a bit more complicated,
959     you need to use C<offsetof>:
960    
961     #include <stddef.h>
962    
963     static void
964     t1_cb (EV_P_ struct ev_timer *w, int revents)
965     {
966     struct my_biggy big = (struct my_biggy *
967     (((char *)w) - offsetof (struct my_biggy, t1));
968     }
969    
970     static void
971     t2_cb (EV_P_ struct ev_timer *w, int revents)
972     {
973     struct my_biggy big = (struct my_biggy *
974     (((char *)w) - offsetof (struct my_biggy, t2));
975     }
976 root 1.1
977    
978     =head1 WATCHER TYPES
979    
980     This section describes each watcher in detail, but will not repeat
981 root 1.48 information given in the last section. Any initialisation/set macros,
982     functions and members specific to the watcher type are explained.
983    
984     Members are additionally marked with either I<[read-only]>, meaning that,
985     while the watcher is active, you can look at the member and expect some
986     sensible content, but you must not modify it (you can modify it while the
987     watcher is stopped to your hearts content), or I<[read-write]>, which
988     means you can expect it to have some sensible content while the watcher
989     is active, but you can also modify it. Modifying it may not do something
990     sensible or take immediate effect (or do anything at all), but libev will
991     not crash or malfunction in any way.
992 root 1.1
993 root 1.34
994 root 1.42 =head2 C<ev_io> - is this file descriptor readable or writable?
995 root 1.1
996 root 1.4 I/O watchers check whether a file descriptor is readable or writable
997 root 1.42 in each iteration of the event loop, or, more precisely, when reading
998     would not block the process and writing would at least be able to write
999     some data. This behaviour is called level-triggering because you keep
1000     receiving events as long as the condition persists. Remember you can stop
1001     the watcher if you don't want to act on the event and neither want to
1002     receive future events.
1003 root 1.1
1004 root 1.23 In general you can register as many read and/or write event watchers per
1005 root 1.8 fd as you want (as long as you don't confuse yourself). Setting all file
1006     descriptors to non-blocking mode is also usually a good idea (but not
1007     required if you know what you are doing).
1008    
1009     If you must do this, then force the use of a known-to-be-good backend
1010 root 1.31 (at the time of this writing, this includes only C<EVBACKEND_SELECT> and
1011     C<EVBACKEND_POLL>).
1012 root 1.8
1013 root 1.42 Another thing you have to watch out for is that it is quite easy to
1014     receive "spurious" readyness notifications, that is your callback might
1015     be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1016     because there is no data. Not only are some backends known to create a
1017     lot of those (for example solaris ports), it is very easy to get into
1018     this situation even with a relatively standard program structure. Thus
1019     it is best to always use non-blocking I/O: An extra C<read>(2) returning
1020     C<EAGAIN> is far preferable to a program hanging until some data arrives.
1021    
1022     If you cannot run the fd in non-blocking mode (for example you should not
1023     play around with an Xlib connection), then you have to seperately re-test
1024 root 1.68 whether a file descriptor is really ready with a known-to-be good interface
1025 root 1.42 such as poll (fortunately in our Xlib example, Xlib already does this on
1026     its own, so its quite safe to use).
1027    
1028 root 1.81 =head3 The special problem of disappearing file descriptors
1029    
1030 root 1.94 Some backends (e.g. kqueue, epoll) need to be told about closing a file
1031 root 1.81 descriptor (either by calling C<close> explicitly or by any other means,
1032     such as C<dup>). The reason is that you register interest in some file
1033     descriptor, but when it goes away, the operating system will silently drop
1034     this interest. If another file descriptor with the same number then is
1035     registered with libev, there is no efficient way to see that this is, in
1036     fact, a different file descriptor.
1037    
1038     To avoid having to explicitly tell libev about such cases, libev follows
1039     the following policy: Each time C<ev_io_set> is being called, libev
1040     will assume that this is potentially a new file descriptor, otherwise
1041     it is assumed that the file descriptor stays the same. That means that
1042     you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1043     descriptor even if the file descriptor number itself did not change.
1044    
1045     This is how one would do it normally anyway, the important point is that
1046     the libev application should not optimise around libev but should leave
1047     optimisations to libev.
1048    
1049 root 1.95 =head3 The special problem of dup'ed file descriptors
1050 root 1.94
1051     Some backends (e.g. epoll), cannot register events for file descriptors,
1052 root 1.103 but only events for the underlying file descriptions. That means when you
1053 root 1.109 have C<dup ()>'ed file descriptors or weirder constellations, and register
1054     events for them, only one file descriptor might actually receive events.
1055 root 1.94
1056 root 1.103 There is no workaround possible except not registering events
1057     for potentially C<dup ()>'ed file descriptors, or to resort to
1058 root 1.94 C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1059    
1060     =head3 The special problem of fork
1061    
1062     Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1063     useless behaviour. Libev fully supports fork, but needs to be told about
1064     it in the child.
1065    
1066     To support fork in your programs, you either have to call
1067     C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
1068     enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1069     C<EVBACKEND_POLL>.
1070    
1071 root 1.81
1072 root 1.82 =head3 Watcher-Specific Functions
1073    
1074 root 1.1 =over 4
1075    
1076     =item ev_io_init (ev_io *, callback, int fd, int events)
1077    
1078     =item ev_io_set (ev_io *, int fd, int events)
1079    
1080 root 1.42 Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1081     rceeive events for and events is either C<EV_READ>, C<EV_WRITE> or
1082     C<EV_READ | EV_WRITE> to receive the given events.
1083 root 1.32
1084 root 1.48 =item int fd [read-only]
1085    
1086     The file descriptor being watched.
1087    
1088     =item int events [read-only]
1089    
1090     The events being watched.
1091    
1092 root 1.1 =back
1093    
1094 root 1.111 =head3 Examples
1095    
1096 root 1.54 Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1097 root 1.34 readable, but only once. Since it is likely line-buffered, you could
1098 root 1.54 attempt to read a whole line in the callback.
1099 root 1.34
1100     static void
1101     stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1102     {
1103     ev_io_stop (loop, w);
1104     .. read from stdin here (or from w->fd) and haqndle any I/O errors
1105     }
1106    
1107     ...
1108     struct ev_loop *loop = ev_default_init (0);
1109     struct ev_io stdin_readable;
1110     ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1111     ev_io_start (loop, &stdin_readable);
1112     ev_loop (loop, 0);
1113    
1114    
1115 root 1.42 =head2 C<ev_timer> - relative and optionally repeating timeouts
1116 root 1.1
1117     Timer watchers are simple relative timers that generate an event after a
1118     given time, and optionally repeating in regular intervals after that.
1119    
1120     The timers are based on real time, that is, if you register an event that
1121 root 1.22 times out after an hour and you reset your system clock to last years
1122 root 1.1 time, it will still time out after (roughly) and hour. "Roughly" because
1123 root 1.28 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1124 root 1.1 monotonic clock option helps a lot here).
1125    
1126 root 1.9 The relative timeouts are calculated relative to the C<ev_now ()>
1127     time. This is usually the right thing as this timestamp refers to the time
1128 root 1.28 of the event triggering whatever timeout you are modifying/starting. If
1129     you suspect event processing to be delayed and you I<need> to base the timeout
1130 root 1.22 on the current time, use something like this to adjust for this:
1131 root 1.9
1132     ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1133    
1134 root 1.28 The callback is guarenteed to be invoked only when its timeout has passed,
1135     but if multiple timers become ready during the same loop iteration then
1136     order of execution is undefined.
1137    
1138 root 1.82 =head3 Watcher-Specific Functions and Data Members
1139    
1140 root 1.1 =over 4
1141    
1142     =item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1143    
1144     =item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1145    
1146     Configure the timer to trigger after C<after> seconds. If C<repeat> is
1147     C<0.>, then it will automatically be stopped. If it is positive, then the
1148     timer will automatically be configured to trigger again C<repeat> seconds
1149     later, again, and again, until stopped manually.
1150    
1151     The timer itself will do a best-effort at avoiding drift, that is, if you
1152     configure a timer to trigger every 10 seconds, then it will trigger at
1153     exactly 10 second intervals. If, however, your program cannot keep up with
1154 root 1.22 the timer (because it takes longer than those 10 seconds to do stuff) the
1155 root 1.1 timer will not fire more than once per event loop iteration.
1156    
1157     =item ev_timer_again (loop)
1158    
1159     This will act as if the timer timed out and restart it again if it is
1160     repeating. The exact semantics are:
1161    
1162 root 1.61 If the timer is pending, its pending status is cleared.
1163 root 1.1
1164 root 1.61 If the timer is started but nonrepeating, stop it (as if it timed out).
1165    
1166     If the timer is repeating, either start it if necessary (with the
1167     C<repeat> value), or reset the running timer to the C<repeat> value.
1168 root 1.1
1169     This sounds a bit complicated, but here is a useful and typical
1170 root 1.61 example: Imagine you have a tcp connection and you want a so-called idle
1171     timeout, that is, you want to be called when there have been, say, 60
1172     seconds of inactivity on the socket. The easiest way to do this is to
1173     configure an C<ev_timer> with a C<repeat> value of C<60> and then call
1174 root 1.48 C<ev_timer_again> each time you successfully read or write some data. If
1175     you go into an idle state where you do not expect data to travel on the
1176 root 1.61 socket, you can C<ev_timer_stop> the timer, and C<ev_timer_again> will
1177     automatically restart it if need be.
1178 root 1.48
1179 root 1.61 That means you can ignore the C<after> value and C<ev_timer_start>
1180     altogether and only ever use the C<repeat> value and C<ev_timer_again>:
1181 root 1.48
1182     ev_timer_init (timer, callback, 0., 5.);
1183     ev_timer_again (loop, timer);
1184     ...
1185     timer->again = 17.;
1186     ev_timer_again (loop, timer);
1187     ...
1188     timer->again = 10.;
1189     ev_timer_again (loop, timer);
1190    
1191 root 1.61 This is more slightly efficient then stopping/starting the timer each time
1192     you want to modify its timeout value.
1193 root 1.48
1194     =item ev_tstamp repeat [read-write]
1195    
1196     The current C<repeat> value. Will be used each time the watcher times out
1197     or C<ev_timer_again> is called and determines the next timeout (if any),
1198     which is also when any modifications are taken into account.
1199 root 1.1
1200     =back
1201    
1202 root 1.111 =head3 Examples
1203    
1204 root 1.54 Example: Create a timer that fires after 60 seconds.
1205 root 1.34
1206     static void
1207     one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1208     {
1209     .. one minute over, w is actually stopped right here
1210     }
1211    
1212     struct ev_timer mytimer;
1213     ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1214     ev_timer_start (loop, &mytimer);
1215    
1216 root 1.54 Example: Create a timeout timer that times out after 10 seconds of
1217 root 1.34 inactivity.
1218    
1219     static void
1220     timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1221     {
1222     .. ten seconds without any activity
1223     }
1224    
1225     struct ev_timer mytimer;
1226     ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1227     ev_timer_again (&mytimer); /* start timer */
1228     ev_loop (loop, 0);
1229    
1230     // and in some piece of code that gets executed on any "activity":
1231     // reset the timeout to start ticking again at 10 seconds
1232     ev_timer_again (&mytimer);
1233    
1234    
1235 root 1.42 =head2 C<ev_periodic> - to cron or not to cron?
1236 root 1.1
1237     Periodic watchers are also timers of a kind, but they are very versatile
1238     (and unfortunately a bit complex).
1239    
1240 root 1.10 Unlike C<ev_timer>'s, they are not based on real time (or relative time)
1241 root 1.1 but on wallclock time (absolute time). You can tell a periodic watcher
1242     to trigger "at" some specific point in time. For example, if you tell a
1243 root 1.38 periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()
1244 root 1.1 + 10.>) and then reset your system clock to the last year, then it will
1245 root 1.10 take a year to trigger the event (unlike an C<ev_timer>, which would trigger
1246 root 1.78 roughly 10 seconds later).
1247 root 1.1
1248     They can also be used to implement vastly more complex timers, such as
1249 root 1.78 triggering an event on each midnight, local time or other, complicated,
1250     rules.
1251 root 1.1
1252 root 1.28 As with timers, the callback is guarenteed to be invoked only when the
1253     time (C<at>) has been passed, but if multiple periodic timers become ready
1254     during the same loop iteration then order of execution is undefined.
1255    
1256 root 1.82 =head3 Watcher-Specific Functions and Data Members
1257    
1258 root 1.1 =over 4
1259    
1260     =item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
1261    
1262     =item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
1263    
1264     Lots of arguments, lets sort it out... There are basically three modes of
1265     operation, and we will explain them from simplest to complex:
1266    
1267     =over 4
1268    
1269 root 1.78 =item * absolute timer (at = time, interval = reschedule_cb = 0)
1270 root 1.1
1271     In this configuration the watcher triggers an event at the wallclock time
1272     C<at> and doesn't repeat. It will not adjust when a time jump occurs,
1273     that is, if it is to be run at January 1st 2011 then it will run when the
1274     system time reaches or surpasses this time.
1275    
1276 root 1.78 =item * non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1277 root 1.1
1278     In this mode the watcher will always be scheduled to time out at the next
1279 root 1.78 C<at + N * interval> time (for some integer N, which can also be negative)
1280     and then repeat, regardless of any time jumps.
1281 root 1.1
1282     This can be used to create timers that do not drift with respect to system
1283     time:
1284    
1285     ev_periodic_set (&periodic, 0., 3600., 0);
1286    
1287     This doesn't mean there will always be 3600 seconds in between triggers,
1288     but only that the the callback will be called when the system time shows a
1289 root 1.12 full hour (UTC), or more correctly, when the system time is evenly divisible
1290 root 1.1 by 3600.
1291    
1292     Another way to think about it (for the mathematically inclined) is that
1293 root 1.10 C<ev_periodic> will try to run the callback in this mode at the next possible
1294 root 1.1 time where C<time = at (mod interval)>, regardless of any time jumps.
1295    
1296 root 1.78 For numerical stability it is preferable that the C<at> value is near
1297     C<ev_now ()> (the current time), but there is no range requirement for
1298     this value.
1299    
1300     =item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1301 root 1.1
1302     In this mode the values for C<interval> and C<at> are both being
1303     ignored. Instead, each time the periodic watcher gets scheduled, the
1304     reschedule callback will be called with the watcher as first, and the
1305     current time as second argument.
1306    
1307 root 1.18 NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
1308     ever, or make any event loop modifications>. If you need to stop it,
1309     return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
1310 root 1.78 starting an C<ev_prepare> watcher, which is legal).
1311 root 1.1
1312 root 1.13 Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
1313 root 1.1 ev_tstamp now)>, e.g.:
1314    
1315     static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
1316     {
1317     return now + 60.;
1318     }
1319    
1320     It must return the next time to trigger, based on the passed time value
1321     (that is, the lowest time value larger than to the second argument). It
1322     will usually be called just before the callback will be triggered, but
1323     might be called at other times, too.
1324    
1325 root 1.18 NOTE: I<< This callback must always return a time that is later than the
1326 root 1.19 passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.
1327 root 1.18
1328 root 1.1 This can be used to create very complex timers, such as a timer that
1329     triggers on each midnight, local time. To do this, you would calculate the
1330 root 1.19 next midnight after C<now> and return the timestamp value for this. How
1331     you do this is, again, up to you (but it is not trivial, which is the main
1332     reason I omitted it as an example).
1333 root 1.1
1334     =back
1335    
1336     =item ev_periodic_again (loop, ev_periodic *)
1337    
1338     Simply stops and restarts the periodic watcher again. This is only useful
1339     when you changed some parameters or the reschedule callback would return
1340     a different time than the last time it was called (e.g. in a crond like
1341     program when the crontabs have changed).
1342    
1343 root 1.78 =item ev_tstamp offset [read-write]
1344    
1345     When repeating, this contains the offset value, otherwise this is the
1346     absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1347    
1348     Can be modified any time, but changes only take effect when the periodic
1349     timer fires or C<ev_periodic_again> is being called.
1350    
1351 root 1.48 =item ev_tstamp interval [read-write]
1352    
1353     The current interval value. Can be modified any time, but changes only
1354     take effect when the periodic timer fires or C<ev_periodic_again> is being
1355     called.
1356    
1357     =item ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]
1358    
1359     The current reschedule callback, or C<0>, if this functionality is
1360     switched off. Can be changed any time, but changes only take effect when
1361     the periodic timer fires or C<ev_periodic_again> is being called.
1362    
1363 root 1.85 =item ev_tstamp at [read-only]
1364    
1365     When active, contains the absolute time that the watcher is supposed to
1366     trigger next.
1367    
1368 root 1.1 =back
1369    
1370 root 1.111 =head3 Examples
1371    
1372 root 1.54 Example: Call a callback every hour, or, more precisely, whenever the
1373 root 1.34 system clock is divisible by 3600. The callback invocation times have
1374     potentially a lot of jittering, but good long-term stability.
1375    
1376     static void
1377     clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1378     {
1379     ... its now a full hour (UTC, or TAI or whatever your clock follows)
1380     }
1381    
1382     struct ev_periodic hourly_tick;
1383     ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1384     ev_periodic_start (loop, &hourly_tick);
1385    
1386 root 1.54 Example: The same as above, but use a reschedule callback to do it:
1387 root 1.34
1388     #include <math.h>
1389    
1390     static ev_tstamp
1391     my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)
1392     {
1393     return fmod (now, 3600.) + 3600.;
1394     }
1395    
1396     ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1397    
1398 root 1.54 Example: Call a callback every hour, starting now:
1399 root 1.34
1400     struct ev_periodic hourly_tick;
1401     ev_periodic_init (&hourly_tick, clock_cb,
1402     fmod (ev_now (loop), 3600.), 3600., 0);
1403     ev_periodic_start (loop, &hourly_tick);
1404    
1405    
1406 root 1.42 =head2 C<ev_signal> - signal me when a signal gets signalled!
1407 root 1.1
1408     Signal watchers will trigger an event when the process receives a specific
1409     signal one or more times. Even though signals are very asynchronous, libev
1410 root 1.9 will try it's best to deliver signals synchronously, i.e. as part of the
1411 root 1.1 normal event processing, like any other event.
1412    
1413 root 1.14 You can configure as many watchers as you like per signal. Only when the
1414 root 1.1 first watcher gets started will libev actually register a signal watcher
1415     with the kernel (thus it coexists with your own signal handlers as long
1416     as you don't register any with libev). Similarly, when the last signal
1417     watcher for a signal is stopped libev will reset the signal handler to
1418     SIG_DFL (regardless of what it was set to before).
1419    
1420 root 1.82 =head3 Watcher-Specific Functions and Data Members
1421    
1422 root 1.1 =over 4
1423    
1424     =item ev_signal_init (ev_signal *, callback, int signum)
1425    
1426     =item ev_signal_set (ev_signal *, int signum)
1427    
1428     Configures the watcher to trigger on the given signal number (usually one
1429     of the C<SIGxxx> constants).
1430    
1431 root 1.48 =item int signum [read-only]
1432    
1433     The signal the watcher watches out for.
1434    
1435 root 1.1 =back
1436    
1437 root 1.35
1438 root 1.42 =head2 C<ev_child> - watch out for process status changes
1439 root 1.1
1440     Child watchers trigger when your process receives a SIGCHLD in response to
1441     some child status changes (most typically when a child of yours dies).
1442    
1443 root 1.82 =head3 Watcher-Specific Functions and Data Members
1444    
1445 root 1.1 =over 4
1446    
1447 root 1.120 =item ev_child_init (ev_child *, callback, int pid, int trace)
1448 root 1.1
1449 root 1.120 =item ev_child_set (ev_child *, int pid, int trace)
1450 root 1.1
1451     Configures the watcher to wait for status changes of process C<pid> (or
1452     I<any> process if C<pid> is specified as C<0>). The callback can look
1453     at the C<rstatus> member of the C<ev_child> watcher structure to see
1454 root 1.14 the status word (use the macros from C<sys/wait.h> and see your systems
1455     C<waitpid> documentation). The C<rpid> member contains the pid of the
1456 root 1.120 process causing the status change. C<trace> must be either C<0> (only
1457     activate the watcher when the process terminates) or C<1> (additionally
1458     activate the watcher when the process is stopped or continued).
1459 root 1.1
1460 root 1.48 =item int pid [read-only]
1461    
1462     The process id this watcher watches out for, or C<0>, meaning any process id.
1463    
1464     =item int rpid [read-write]
1465    
1466     The process id that detected a status change.
1467    
1468     =item int rstatus [read-write]
1469    
1470     The process exit/trace status caused by C<rpid> (see your systems
1471     C<waitpid> and C<sys/wait.h> documentation for details).
1472    
1473 root 1.1 =back
1474    
1475 root 1.111 =head3 Examples
1476    
1477 root 1.54 Example: Try to exit cleanly on SIGINT and SIGTERM.
1478 root 1.34
1479     static void
1480     sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
1481     {
1482     ev_unloop (loop, EVUNLOOP_ALL);
1483     }
1484    
1485     struct ev_signal signal_watcher;
1486     ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1487     ev_signal_start (loop, &sigint_cb);
1488    
1489    
1490 root 1.48 =head2 C<ev_stat> - did the file attributes just change?
1491    
1492     This watches a filesystem path for attribute changes. That is, it calls
1493     C<stat> regularly (or when the OS says it changed) and sees if it changed
1494     compared to the last time, invoking the callback if it did.
1495    
1496     The path does not need to exist: changing from "path exists" to "path does
1497     not exist" is a status change like any other. The condition "path does
1498     not exist" is signified by the C<st_nlink> field being zero (which is
1499     otherwise always forced to be at least one) and all the other fields of
1500     the stat buffer having unspecified contents.
1501    
1502 root 1.60 The path I<should> be absolute and I<must not> end in a slash. If it is
1503     relative and your working directory changes, the behaviour is undefined.
1504    
1505 root 1.48 Since there is no standard to do this, the portable implementation simply
1506 root 1.57 calls C<stat (2)> regularly on the path to see if it changed somehow. You
1507 root 1.48 can specify a recommended polling interval for this case. If you specify
1508     a polling interval of C<0> (highly recommended!) then a I<suitable,
1509     unspecified default> value will be used (which you can expect to be around
1510     five seconds, although this might change dynamically). Libev will also
1511     impose a minimum interval which is currently around C<0.1>, but thats
1512     usually overkill.
1513    
1514     This watcher type is not meant for massive numbers of stat watchers,
1515     as even with OS-supported change notifications, this can be
1516     resource-intensive.
1517    
1518 root 1.57 At the time of this writing, only the Linux inotify interface is
1519     implemented (implementing kqueue support is left as an exercise for the
1520     reader). Inotify will be used to give hints only and should not change the
1521     semantics of C<ev_stat> watchers, which means that libev sometimes needs
1522     to fall back to regular polling again even with inotify, but changes are
1523     usually detected immediately, and if the file exists there will be no
1524     polling.
1525 root 1.48
1526 root 1.108 =head3 Inotify
1527    
1528     When C<inotify (7)> support has been compiled into libev (generally only
1529     available on Linux) and present at runtime, it will be used to speed up
1530     change detection where possible. The inotify descriptor will be created lazily
1531     when the first C<ev_stat> watcher is being started.
1532    
1533     Inotify presense does not change the semantics of C<ev_stat> watchers
1534     except that changes might be detected earlier, and in some cases, to avoid
1535     making regular C<stat> calls. Even in the presense of inotify support
1536     there are many cases where libev has to resort to regular C<stat> polling.
1537    
1538     (There is no support for kqueue, as apparently it cannot be used to
1539     implement this functionality, due to the requirement of having a file
1540     descriptor open on the object at all times).
1541    
1542 root 1.107 =head3 The special problem of stat time resolution
1543    
1544     The C<stat ()> syscall only supports full-second resolution portably, and
1545     even on systems where the resolution is higher, many filesystems still
1546     only support whole seconds.
1547    
1548     That means that, if the time is the only thing that changes, you might
1549     miss updates: on the first update, C<ev_stat> detects a change and calls
1550     your callback, which does something. When there is another update within
1551     the same second, C<ev_stat> will be unable to detect it.
1552    
1553     The solution to this is to delay acting on a change for a second (or till
1554     the next second boundary), using a roughly one-second delay C<ev_timer>
1555     (C<ev_timer_set (w, 0., 1.01); ev_timer_again (loop, w)>). The C<.01>
1556     is added to work around small timing inconsistencies of some operating
1557     systems.
1558    
1559 root 1.82 =head3 Watcher-Specific Functions and Data Members
1560    
1561 root 1.48 =over 4
1562    
1563     =item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
1564    
1565     =item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
1566    
1567     Configures the watcher to wait for status changes of the given
1568     C<path>. The C<interval> is a hint on how quickly a change is expected to
1569     be detected and should normally be specified as C<0> to let libev choose
1570     a suitable value. The memory pointed to by C<path> must point to the same
1571     path for as long as the watcher is active.
1572    
1573     The callback will be receive C<EV_STAT> when a change was detected,
1574     relative to the attributes at the time the watcher was started (or the
1575     last change was detected).
1576    
1577     =item ev_stat_stat (ev_stat *)
1578    
1579     Updates the stat buffer immediately with new values. If you change the
1580     watched path in your callback, you could call this fucntion to avoid
1581     detecting this change (while introducing a race condition). Can also be
1582     useful simply to find out the new values.
1583    
1584     =item ev_statdata attr [read-only]
1585    
1586     The most-recently detected attributes of the file. Although the type is of
1587     C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
1588     suitable for your system. If the C<st_nlink> member is C<0>, then there
1589     was some error while C<stat>ing the file.
1590    
1591     =item ev_statdata prev [read-only]
1592    
1593     The previous attributes of the file. The callback gets invoked whenever
1594     C<prev> != C<attr>.
1595    
1596     =item ev_tstamp interval [read-only]
1597    
1598     The specified interval.
1599    
1600     =item const char *path [read-only]
1601    
1602     The filesystem path that is being watched.
1603    
1604     =back
1605    
1606 root 1.108 =head3 Examples
1607    
1608 root 1.48 Example: Watch C</etc/passwd> for attribute changes.
1609    
1610     static void
1611     passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
1612     {
1613     /* /etc/passwd changed in some way */
1614     if (w->attr.st_nlink)
1615     {
1616     printf ("passwd current size %ld\n", (long)w->attr.st_size);
1617     printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
1618     printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
1619     }
1620     else
1621     /* you shalt not abuse printf for puts */
1622     puts ("wow, /etc/passwd is not there, expect problems. "
1623     "if this is windows, they already arrived\n");
1624     }
1625    
1626     ...
1627     ev_stat passwd;
1628    
1629 root 1.107 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1630     ev_stat_start (loop, &passwd);
1631    
1632     Example: Like above, but additionally use a one-second delay so we do not
1633     miss updates (however, frequent updates will delay processing, too, so
1634     one might do the work both on C<ev_stat> callback invocation I<and> on
1635     C<ev_timer> callback invocation).
1636    
1637     static ev_stat passwd;
1638     static ev_timer timer;
1639    
1640     static void
1641     timer_cb (EV_P_ ev_timer *w, int revents)
1642     {
1643     ev_timer_stop (EV_A_ w);
1644    
1645     /* now it's one second after the most recent passwd change */
1646     }
1647    
1648     static void
1649     stat_cb (EV_P_ ev_stat *w, int revents)
1650     {
1651     /* reset the one-second timer */
1652     ev_timer_again (EV_A_ &timer);
1653     }
1654    
1655     ...
1656     ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
1657 root 1.48 ev_stat_start (loop, &passwd);
1658 root 1.107 ev_timer_init (&timer, timer_cb, 0., 1.01);
1659 root 1.48
1660    
1661 root 1.42 =head2 C<ev_idle> - when you've got nothing better to do...
1662 root 1.1
1663 root 1.67 Idle watchers trigger events when no other events of the same or higher
1664     priority are pending (prepare, check and other idle watchers do not
1665     count).
1666    
1667     That is, as long as your process is busy handling sockets or timeouts
1668     (or even signals, imagine) of the same or higher priority it will not be
1669     triggered. But when your process is idle (or only lower-priority watchers
1670     are pending), the idle watchers are being called once per event loop
1671     iteration - until stopped, that is, or your process receives more events
1672     and becomes busy again with higher priority stuff.
1673 root 1.1
1674     The most noteworthy effect is that as long as any idle watchers are
1675     active, the process will not block when waiting for new events.
1676    
1677     Apart from keeping your process non-blocking (which is a useful
1678     effect on its own sometimes), idle watchers are a good place to do
1679     "pseudo-background processing", or delay processing stuff to after the
1680     event loop has handled all outstanding events.
1681    
1682 root 1.82 =head3 Watcher-Specific Functions and Data Members
1683    
1684 root 1.1 =over 4
1685    
1686     =item ev_idle_init (ev_signal *, callback)
1687    
1688     Initialises and configures the idle watcher - it has no parameters of any
1689     kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1690     believe me.
1691    
1692     =back
1693    
1694 root 1.111 =head3 Examples
1695    
1696 root 1.54 Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1697     callback, free it. Also, use no error checking, as usual.
1698 root 1.34
1699     static void
1700     idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)
1701     {
1702     free (w);
1703     // now do something you wanted to do when the program has
1704 root 1.121 // no longer anything immediate to do.
1705 root 1.34 }
1706    
1707     struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));
1708     ev_idle_init (idle_watcher, idle_cb);
1709     ev_idle_start (loop, idle_cb);
1710    
1711    
1712 root 1.42 =head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1713 root 1.1
1714 root 1.14 Prepare and check watchers are usually (but not always) used in tandem:
1715 root 1.20 prepare watchers get invoked before the process blocks and check watchers
1716 root 1.14 afterwards.
1717 root 1.1
1718 root 1.45 You I<must not> call C<ev_loop> or similar functions that enter
1719     the current event loop from either C<ev_prepare> or C<ev_check>
1720     watchers. Other loops than the current one are fine, however. The
1721     rationale behind this is that you do not need to check for recursion in
1722     those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
1723     C<ev_check> so if you have one watcher of each kind they will always be
1724     called in pairs bracketing the blocking call.
1725    
1726 root 1.35 Their main purpose is to integrate other event mechanisms into libev and
1727     their use is somewhat advanced. This could be used, for example, to track
1728     variable changes, implement your own watchers, integrate net-snmp or a
1729 root 1.45 coroutine library and lots more. They are also occasionally useful if
1730     you cache some data and want to flush it before blocking (for example,
1731     in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
1732     watcher).
1733 root 1.1
1734     This is done by examining in each prepare call which file descriptors need
1735 root 1.14 to be watched by the other library, registering C<ev_io> watchers for
1736     them and starting an C<ev_timer> watcher for any timeouts (many libraries
1737     provide just this functionality). Then, in the check watcher you check for
1738     any events that occured (by checking the pending status of all watchers
1739     and stopping them) and call back into the library. The I/O and timer
1740 root 1.20 callbacks will never actually be called (but must be valid nevertheless,
1741 root 1.14 because you never know, you know?).
1742 root 1.1
1743 root 1.14 As another example, the Perl Coro module uses these hooks to integrate
1744 root 1.1 coroutines into libev programs, by yielding to other active coroutines
1745     during each prepare and only letting the process block if no coroutines
1746 root 1.20 are ready to run (it's actually more complicated: it only runs coroutines
1747     with priority higher than or equal to the event loop and one coroutine
1748     of lower priority, but only once, using idle watchers to keep the event
1749     loop from blocking if lower-priority coroutines are active, thus mapping
1750     low-priority coroutines to idle/background tasks).
1751 root 1.1
1752 root 1.77 It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1753     priority, to ensure that they are being run before any other watchers
1754     after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
1755     too) should not activate ("feed") events into libev. While libev fully
1756 root 1.100 supports this, they will be called before other C<ev_check> watchers
1757     did their job. As C<ev_check> watchers are often used to embed other
1758     (non-libev) event loops those other event loops might be in an unusable
1759     state until their C<ev_check> watcher ran (always remind yourself to
1760     coexist peacefully with others).
1761 root 1.77
1762 root 1.82 =head3 Watcher-Specific Functions and Data Members
1763    
1764 root 1.1 =over 4
1765    
1766     =item ev_prepare_init (ev_prepare *, callback)
1767    
1768     =item ev_check_init (ev_check *, callback)
1769    
1770     Initialises and configures the prepare or check watcher - they have no
1771     parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1772 root 1.14 macros, but using them is utterly, utterly and completely pointless.
1773 root 1.1
1774     =back
1775    
1776 root 1.111 =head3 Examples
1777    
1778 root 1.76 There are a number of principal ways to embed other event loops or modules
1779     into libev. Here are some ideas on how to include libadns into libev
1780     (there is a Perl module named C<EV::ADNS> that does this, which you could
1781     use for an actually working example. Another Perl module named C<EV::Glib>
1782     embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV
1783     into the Glib event loop).
1784    
1785     Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1786     and in a check watcher, destroy them and call into libadns. What follows
1787     is pseudo-code only of course. This requires you to either use a low
1788     priority for the check watcher or use C<ev_clear_pending> explicitly, as
1789     the callbacks for the IO/timeout watchers might not have been called yet.
1790 root 1.45
1791     static ev_io iow [nfd];
1792     static ev_timer tw;
1793    
1794     static void
1795     io_cb (ev_loop *loop, ev_io *w, int revents)
1796     {
1797     }
1798    
1799     // create io watchers for each fd and a timer before blocking
1800     static void
1801     adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
1802     {
1803 root 1.64 int timeout = 3600000;
1804     struct pollfd fds [nfd];
1805 root 1.45 // actual code will need to loop here and realloc etc.
1806     adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1807    
1808     /* the callback is illegal, but won't be called as we stop during check */
1809     ev_timer_init (&tw, 0, timeout * 1e-3);
1810     ev_timer_start (loop, &tw);
1811    
1812 root 1.76 // create one ev_io per pollfd
1813 root 1.45 for (int i = 0; i < nfd; ++i)
1814     {
1815     ev_io_init (iow + i, io_cb, fds [i].fd,
1816     ((fds [i].events & POLLIN ? EV_READ : 0)
1817     | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1818    
1819     fds [i].revents = 0;
1820     ev_io_start (loop, iow + i);
1821     }
1822     }
1823    
1824     // stop all watchers after blocking
1825     static void
1826     adns_check_cb (ev_loop *loop, ev_check *w, int revents)
1827     {
1828     ev_timer_stop (loop, &tw);
1829    
1830     for (int i = 0; i < nfd; ++i)
1831 root 1.76 {
1832     // set the relevant poll flags
1833     // could also call adns_processreadable etc. here
1834     struct pollfd *fd = fds + i;
1835     int revents = ev_clear_pending (iow + i);
1836     if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1837     if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1838    
1839     // now stop the watcher
1840     ev_io_stop (loop, iow + i);
1841     }
1842 root 1.45
1843     adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1844     }
1845 root 1.34
1846 root 1.76 Method 2: This would be just like method 1, but you run C<adns_afterpoll>
1847     in the prepare watcher and would dispose of the check watcher.
1848    
1849     Method 3: If the module to be embedded supports explicit event
1850     notification (adns does), you can also make use of the actual watcher
1851     callbacks, and only destroy/create the watchers in the prepare watcher.
1852    
1853     static void
1854     timer_cb (EV_P_ ev_timer *w, int revents)
1855     {
1856     adns_state ads = (adns_state)w->data;
1857     update_now (EV_A);
1858    
1859     adns_processtimeouts (ads, &tv_now);
1860     }
1861    
1862     static void
1863     io_cb (EV_P_ ev_io *w, int revents)
1864     {
1865     adns_state ads = (adns_state)w->data;
1866     update_now (EV_A);
1867    
1868     if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1869     if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1870     }
1871    
1872     // do not ever call adns_afterpoll
1873    
1874     Method 4: Do not use a prepare or check watcher because the module you
1875     want to embed is too inflexible to support it. Instead, youc na override
1876     their poll function. The drawback with this solution is that the main
1877     loop is now no longer controllable by EV. The C<Glib::EV> module does
1878     this.
1879    
1880     static gint
1881     event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1882     {
1883     int got_events = 0;
1884    
1885     for (n = 0; n < nfds; ++n)
1886     // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1887    
1888     if (timeout >= 0)
1889     // create/start timer
1890    
1891     // poll
1892     ev_loop (EV_A_ 0);
1893    
1894     // stop timer again
1895     if (timeout >= 0)
1896     ev_timer_stop (EV_A_ &to);
1897    
1898     // stop io watchers again - their callbacks should have set
1899     for (n = 0; n < nfds; ++n)
1900     ev_io_stop (EV_A_ iow [n]);
1901    
1902     return got_events;
1903     }
1904    
1905 root 1.34
1906 root 1.42 =head2 C<ev_embed> - when one backend isn't enough...
1907 root 1.35
1908     This is a rather advanced watcher type that lets you embed one event loop
1909 root 1.36 into another (currently only C<ev_io> events are supported in the embedded
1910     loop, other types of watchers might be handled in a delayed or incorrect
1911 root 1.100 fashion and must not be used).
1912 root 1.35
1913     There are primarily two reasons you would want that: work around bugs and
1914     prioritise I/O.
1915    
1916     As an example for a bug workaround, the kqueue backend might only support
1917     sockets on some platform, so it is unusable as generic backend, but you
1918     still want to make use of it because you have many sockets and it scales
1919     so nicely. In this case, you would create a kqueue-based loop and embed it
1920     into your default loop (which might use e.g. poll). Overall operation will
1921     be a bit slower because first libev has to poll and then call kevent, but
1922     at least you can use both at what they are best.
1923    
1924     As for prioritising I/O: rarely you have the case where some fds have
1925     to be watched and handled very quickly (with low latency), and even
1926     priorities and idle watchers might have too much overhead. In this case
1927     you would put all the high priority stuff in one loop and all the rest in
1928     a second one, and embed the second one in the first.
1929    
1930 root 1.36 As long as the watcher is active, the callback will be invoked every time
1931     there might be events pending in the embedded loop. The callback must then
1932     call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
1933     their callbacks (you could also start an idle watcher to give the embedded
1934     loop strictly lower priority for example). You can also set the callback
1935     to C<0>, in which case the embed watcher will automatically execute the
1936     embedded loop sweep.
1937    
1938 root 1.35 As long as the watcher is started it will automatically handle events. The
1939     callback will be invoked whenever some events have been handled. You can
1940     set the callback to C<0> to avoid having to specify one if you are not
1941     interested in that.
1942    
1943     Also, there have not currently been made special provisions for forking:
1944     when you fork, you not only have to call C<ev_loop_fork> on both loops,
1945     but you will also have to stop and restart any C<ev_embed> watchers
1946     yourself.
1947    
1948     Unfortunately, not all backends are embeddable, only the ones returned by
1949     C<ev_embeddable_backends> are, which, unfortunately, does not include any
1950     portable one.
1951    
1952     So when you want to use this feature you will always have to be prepared
1953     that you cannot get an embeddable loop. The recommended way to get around
1954     this is to have a separate variables for your embeddable loop, try to
1955 root 1.111 create it, and if that fails, use the normal loop for everything.
1956 root 1.35
1957 root 1.82 =head3 Watcher-Specific Functions and Data Members
1958    
1959 root 1.35 =over 4
1960    
1961 root 1.36 =item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
1962    
1963     =item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
1964    
1965     Configures the watcher to embed the given loop, which must be
1966     embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1967     invoked automatically, otherwise it is the responsibility of the callback
1968     to invoke it (it will continue to be called until the sweep has been done,
1969     if you do not want thta, you need to temporarily stop the embed watcher).
1970 root 1.35
1971 root 1.36 =item ev_embed_sweep (loop, ev_embed *)
1972 root 1.35
1973 root 1.36 Make a single, non-blocking sweep over the embedded loop. This works
1974     similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1975     apropriate way for embedded loops.
1976 root 1.35
1977 root 1.91 =item struct ev_loop *other [read-only]
1978 root 1.48
1979     The embedded event loop.
1980    
1981 root 1.35 =back
1982    
1983 root 1.111 =head3 Examples
1984    
1985     Example: Try to get an embeddable event loop and embed it into the default
1986     event loop. If that is not possible, use the default loop. The default
1987     loop is stored in C<loop_hi>, while the mebeddable loop is stored in
1988     C<loop_lo> (which is C<loop_hi> in the acse no embeddable loop can be
1989     used).
1990    
1991     struct ev_loop *loop_hi = ev_default_init (0);
1992     struct ev_loop *loop_lo = 0;
1993     struct ev_embed embed;
1994    
1995     // see if there is a chance of getting one that works
1996     // (remember that a flags value of 0 means autodetection)
1997     loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
1998     ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
1999     : 0;
2000    
2001     // if we got one, then embed it, otherwise default to loop_hi
2002     if (loop_lo)
2003     {
2004     ev_embed_init (&embed, 0, loop_lo);
2005     ev_embed_start (loop_hi, &embed);
2006     }
2007     else
2008     loop_lo = loop_hi;
2009    
2010     Example: Check if kqueue is available but not recommended and create
2011     a kqueue backend for use with sockets (which usually work with any
2012     kqueue implementation). Store the kqueue/socket-only event loop in
2013     C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
2014    
2015     struct ev_loop *loop = ev_default_init (0);
2016     struct ev_loop *loop_socket = 0;
2017     struct ev_embed embed;
2018    
2019     if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2020     if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2021     {
2022     ev_embed_init (&embed, 0, loop_socket);
2023     ev_embed_start (loop, &embed);
2024     }
2025    
2026     if (!loop_socket)
2027     loop_socket = loop;
2028    
2029     // now use loop_socket for all sockets, and loop for everything else
2030    
2031 root 1.35
2032 root 1.50 =head2 C<ev_fork> - the audacity to resume the event loop after a fork
2033    
2034     Fork watchers are called when a C<fork ()> was detected (usually because
2035     whoever is a good citizen cared to tell libev about it by calling
2036     C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
2037     event loop blocks next and before C<ev_check> watchers are being called,
2038     and only in the child after the fork. If whoever good citizen calling
2039     C<ev_default_fork> cheats and calls it in the wrong process, the fork
2040     handlers will be invoked, too, of course.
2041    
2042 root 1.83 =head3 Watcher-Specific Functions and Data Members
2043    
2044 root 1.50 =over 4
2045    
2046     =item ev_fork_init (ev_signal *, callback)
2047    
2048     Initialises and configures the fork watcher - it has no parameters of any
2049     kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
2050     believe me.
2051    
2052     =back
2053    
2054    
2055 root 1.122 =head2 C<ev_async> - how to wake up another event loop
2056    
2057     In general, you cannot use an C<ev_loop> from multiple threads or other
2058     asynchronous sources such as signal handlers (as opposed to multiple event
2059     loops - those are of course safe to use in different threads).
2060    
2061     Sometimes, however, you need to wake up another event loop you do not
2062     control, for example because it belongs to another thread. This is what
2063     C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
2064     can signal it by calling C<ev_async_send>, which is thread- and signal
2065     safe.
2066    
2067     This functionality is very similar to C<ev_signal> watchers, as signals,
2068     too, are asynchronous in nature, and signals, too, will be compressed
2069     (i.e. the number of callback invocations may be less than the number of
2070     C<ev_async_sent> calls).
2071    
2072     Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
2073     just the default loop.
2074    
2075 root 1.124 =head3 Queueing
2076    
2077     C<ev_async> does not support queueing of data in any way. The reason
2078     is that the author does not know of a simple (or any) algorithm for a
2079     multiple-writer-single-reader queue that works in all cases and doesn't
2080     need elaborate support such as pthreads.
2081    
2082     That means that if you want to queue data, you have to provide your own
2083     queue. And here is how you would implement locking:
2084    
2085     =over 4
2086    
2087     =item queueing from a signal handler context
2088    
2089     To implement race-free queueing, you simply add to the queue in the signal
2090     handler but you block the signal handler in the watcher callback. Here is an example that does that for
2091     some fictitiuous SIGUSR1 handler:
2092    
2093     static ev_async mysig;
2094    
2095     static void
2096     sigusr1_handler (void)
2097     {
2098     sometype data;
2099    
2100     // no locking etc.
2101     queue_put (data);
2102 root 1.125 ev_async_send (DEFAULT_ &mysig);
2103 root 1.124 }
2104    
2105     static void
2106     mysig_cb (EV_P_ ev_async *w, int revents)
2107     {
2108     sometype data;
2109     sigset_t block, prev;
2110    
2111     sigemptyset (&block);
2112     sigaddset (&block, SIGUSR1);
2113     sigprocmask (SIG_BLOCK, &block, &prev);
2114    
2115     while (queue_get (&data))
2116     process (data);
2117    
2118     if (sigismember (&prev, SIGUSR1)
2119     sigprocmask (SIG_UNBLOCK, &block, 0);
2120     }
2121    
2122     (Note: pthreads in theory requires you to use C<pthread_setmask>
2123     instead of C<sigprocmask> when you use threads, but libev doesn't do it
2124     either...).
2125    
2126     =item queueing from a thread context
2127    
2128     The strategy for threads is different, as you cannot (easily) block
2129     threads but you can easily preempt them, so to queue safely you need to
2130     emply a traditional mutex lock, such as in this pthread example:
2131    
2132     static ev_async mysig;
2133     static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
2134    
2135     static void
2136     otherthread (void)
2137     {
2138     // only need to lock the actual queueing operation
2139     pthread_mutex_lock (&mymutex);
2140     queue_put (data);
2141     pthread_mutex_unlock (&mymutex);
2142    
2143 root 1.125 ev_async_send (DEFAULT_ &mysig);
2144 root 1.124 }
2145    
2146     static void
2147     mysig_cb (EV_P_ ev_async *w, int revents)
2148     {
2149     pthread_mutex_lock (&mymutex);
2150    
2151     while (queue_get (&data))
2152     process (data);
2153    
2154     pthread_mutex_unlock (&mymutex);
2155     }
2156    
2157     =back
2158    
2159    
2160 root 1.122 =head3 Watcher-Specific Functions and Data Members
2161    
2162     =over 4
2163    
2164     =item ev_async_init (ev_async *, callback)
2165    
2166     Initialises and configures the async watcher - it has no parameters of any
2167     kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
2168     believe me.
2169    
2170     =item ev_async_send (loop, ev_async *)
2171    
2172     Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2173     an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2174     C<ev_feed_event>, this call is safe to do in other threads, signal or
2175     similar contexts (see the dicusssion of C<EV_ATOMIC_T> in the embedding
2176     section below on what exactly this means).
2177    
2178     This call incurs the overhead of a syscall only once per loop iteration,
2179     so while the overhead might be noticable, it doesn't apply to repeated
2180     calls to C<ev_async_send>.
2181    
2182     =back
2183    
2184    
2185 root 1.1 =head1 OTHER FUNCTIONS
2186    
2187 root 1.14 There are some other functions of possible interest. Described. Here. Now.
2188 root 1.1
2189     =over 4
2190    
2191     =item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
2192    
2193     This function combines a simple timer and an I/O watcher, calls your
2194     callback on whichever event happens first and automatically stop both
2195     watchers. This is useful if you want to wait for a single event on an fd
2196 root 1.22 or timeout without having to allocate/configure/start/stop/free one or
2197 root 1.1 more watchers yourself.
2198    
2199 root 1.14 If C<fd> is less than 0, then no I/O watcher will be started and events
2200     is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and
2201     C<events> set will be craeted and started.
2202 root 1.1
2203     If C<timeout> is less than 0, then no timeout watcher will be
2204 root 1.14 started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
2205     repeat = 0) will be started. While C<0> is a valid timeout, it is of
2206     dubious value.
2207    
2208     The callback has the type C<void (*cb)(int revents, void *arg)> and gets
2209 root 1.21 passed an C<revents> set like normal event callbacks (a combination of
2210 root 1.14 C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
2211     value passed to C<ev_once>:
2212 root 1.1
2213     static void stdin_ready (int revents, void *arg)
2214     {
2215     if (revents & EV_TIMEOUT)
2216 root 1.14 /* doh, nothing entered */;
2217 root 1.1 else if (revents & EV_READ)
2218 root 1.14 /* stdin might have data for us, joy! */;
2219 root 1.1 }
2220    
2221 root 1.14 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2222 root 1.1
2223 root 1.36 =item ev_feed_event (ev_loop *, watcher *, int revents)
2224 root 1.1
2225     Feeds the given event set into the event loop, as if the specified event
2226 root 1.14 had happened for the specified watcher (which must be a pointer to an
2227     initialised but not necessarily started event watcher).
2228 root 1.1
2229 root 1.36 =item ev_feed_fd_event (ev_loop *, int fd, int revents)
2230 root 1.1
2231 root 1.14 Feed an event on the given fd, as if a file descriptor backend detected
2232     the given events it.
2233 root 1.1
2234 root 1.36 =item ev_feed_signal_event (ev_loop *loop, int signum)
2235 root 1.1
2236 root 1.36 Feed an event as if the given signal occured (C<loop> must be the default
2237     loop!).
2238 root 1.1
2239     =back
2240    
2241 root 1.34
2242 root 1.20 =head1 LIBEVENT EMULATION
2243    
2244 root 1.24 Libev offers a compatibility emulation layer for libevent. It cannot
2245     emulate the internals of libevent, so here are some usage hints:
2246    
2247     =over 4
2248    
2249     =item * Use it by including <event.h>, as usual.
2250    
2251     =item * The following members are fully supported: ev_base, ev_callback,
2252     ev_arg, ev_fd, ev_res, ev_events.
2253    
2254     =item * Avoid using ev_flags and the EVLIST_*-macros, while it is
2255     maintained by libev, it does not work exactly the same way as in libevent (consider
2256     it a private API).
2257    
2258     =item * Priorities are not currently supported. Initialising priorities
2259     will fail and all watchers will have the same priority, even though there
2260     is an ev_pri field.
2261    
2262     =item * Other members are not supported.
2263    
2264     =item * The libev emulation is I<not> ABI compatible to libevent, you need
2265     to use the libev header file and library.
2266    
2267     =back
2268 root 1.20
2269     =head1 C++ SUPPORT
2270    
2271 root 1.38 Libev comes with some simplistic wrapper classes for C++ that mainly allow
2272     you to use some convinience methods to start/stop watchers and also change
2273     the callback model to a model using method callbacks on objects.
2274    
2275     To use it,
2276    
2277     #include <ev++.h>
2278    
2279 root 1.71 This automatically includes F<ev.h> and puts all of its definitions (many
2280     of them macros) into the global namespace. All C++ specific things are
2281     put into the C<ev> namespace. It should support all the same embedding
2282     options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
2283    
2284 root 1.72 Care has been taken to keep the overhead low. The only data member the C++
2285     classes add (compared to plain C-style watchers) is the event loop pointer
2286     that the watcher is associated with (or no additional members at all if
2287     you disable C<EV_MULTIPLICITY> when embedding libev).
2288 root 1.71
2289 root 1.72 Currently, functions, and static and non-static member functions can be
2290 root 1.71 used as callbacks. Other types should be easy to add as long as they only
2291     need one additional pointer for context. If you need support for other
2292     types of functors please contact the author (preferably after implementing
2293     it).
2294 root 1.38
2295     Here is a list of things available in the C<ev> namespace:
2296    
2297     =over 4
2298    
2299     =item C<ev::READ>, C<ev::WRITE> etc.
2300    
2301     These are just enum values with the same values as the C<EV_READ> etc.
2302     macros from F<ev.h>.
2303    
2304     =item C<ev::tstamp>, C<ev::now>
2305    
2306     Aliases to the same types/functions as with the C<ev_> prefix.
2307    
2308     =item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
2309    
2310     For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
2311     the same name in the C<ev> namespace, with the exception of C<ev_signal>
2312     which is called C<ev::sig> to avoid clashes with the C<signal> macro
2313     defines by many implementations.
2314    
2315     All of those classes have these methods:
2316    
2317     =over 4
2318    
2319 root 1.71 =item ev::TYPE::TYPE ()
2320 root 1.38
2321 root 1.71 =item ev::TYPE::TYPE (struct ev_loop *)
2322 root 1.38
2323     =item ev::TYPE::~TYPE
2324    
2325 root 1.71 The constructor (optionally) takes an event loop to associate the watcher
2326     with. If it is omitted, it will use C<EV_DEFAULT>.
2327    
2328     The constructor calls C<ev_init> for you, which means you have to call the
2329     C<set> method before starting it.
2330    
2331     It will not set a callback, however: You have to call the templated C<set>
2332     method to set a callback before you can start the watcher.
2333    
2334     (The reason why you have to use a method is a limitation in C++ which does
2335     not allow explicit template arguments for constructors).
2336 root 1.38
2337     The destructor automatically stops the watcher if it is active.
2338    
2339 root 1.71 =item w->set<class, &class::method> (object *)
2340    
2341     This method sets the callback method to call. The method has to have a
2342     signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
2343     first argument and the C<revents> as second. The object must be given as
2344     parameter and is stored in the C<data> member of the watcher.
2345    
2346     This method synthesizes efficient thunking code to call your method from
2347     the C callback that libev requires. If your compiler can inline your
2348     callback (i.e. it is visible to it at the place of the C<set> call and
2349     your compiler is good :), then the method will be fully inlined into the
2350     thunking function, making it as fast as a direct C callback.
2351    
2352     Example: simple class declaration and watcher initialisation
2353    
2354     struct myclass
2355     {
2356     void io_cb (ev::io &w, int revents) { }
2357     }
2358    
2359     myclass obj;
2360     ev::io iow;
2361     iow.set <myclass, &myclass::io_cb> (&obj);
2362    
2363 root 1.75 =item w->set<function> (void *data = 0)
2364 root 1.71
2365     Also sets a callback, but uses a static method or plain function as
2366     callback. The optional C<data> argument will be stored in the watcher's
2367     C<data> member and is free for you to use.
2368    
2369 root 1.75 The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2370    
2371 root 1.71 See the method-C<set> above for more details.
2372    
2373 root 1.75 Example:
2374    
2375     static void io_cb (ev::io &w, int revents) { }
2376     iow.set <io_cb> ();
2377    
2378 root 1.38 =item w->set (struct ev_loop *)
2379    
2380     Associates a different C<struct ev_loop> with this watcher. You can only
2381     do this when the watcher is inactive (and not pending either).
2382    
2383     =item w->set ([args])
2384    
2385     Basically the same as C<ev_TYPE_set>, with the same args. Must be
2386 root 1.71 called at least once. Unlike the C counterpart, an active watcher gets
2387     automatically stopped and restarted when reconfiguring it with this
2388     method.
2389 root 1.38
2390     =item w->start ()
2391    
2392 root 1.71 Starts the watcher. Note that there is no C<loop> argument, as the
2393     constructor already stores the event loop.
2394 root 1.38
2395     =item w->stop ()
2396    
2397     Stops the watcher if it is active. Again, no C<loop> argument.
2398    
2399 root 1.84 =item w->again () (C<ev::timer>, C<ev::periodic> only)
2400 root 1.38
2401     For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
2402     C<ev_TYPE_again> function.
2403    
2404 root 1.84 =item w->sweep () (C<ev::embed> only)
2405 root 1.38
2406     Invokes C<ev_embed_sweep>.
2407    
2408 root 1.84 =item w->update () (C<ev::stat> only)
2409 root 1.49
2410     Invokes C<ev_stat_stat>.
2411    
2412 root 1.38 =back
2413    
2414     =back
2415    
2416     Example: Define a class with an IO and idle watcher, start one of them in
2417     the constructor.
2418    
2419     class myclass
2420     {
2421 root 1.121 ev::io io; void io_cb (ev::io &w, int revents);
2422     ev:idle idle void idle_cb (ev::idle &w, int revents);
2423 root 1.38
2424 root 1.121 myclass (int fd)
2425     {
2426     io .set <myclass, &myclass::io_cb > (this);
2427     idle.set <myclass, &myclass::idle_cb> (this);
2428 root 1.38
2429 root 1.121 io.start (fd, ev::READ);
2430     }
2431     };
2432 root 1.20
2433 root 1.50
2434     =head1 MACRO MAGIC
2435    
2436 root 1.84 Libev can be compiled with a variety of options, the most fundamantal
2437     of which is C<EV_MULTIPLICITY>. This option determines whether (most)
2438     functions and callbacks have an initial C<struct ev_loop *> argument.
2439 root 1.50
2440     To make it easier to write programs that cope with either variant, the
2441     following macros are defined:
2442    
2443     =over 4
2444    
2445     =item C<EV_A>, C<EV_A_>
2446    
2447     This provides the loop I<argument> for functions, if one is required ("ev
2448     loop argument"). The C<EV_A> form is used when this is the sole argument,
2449     C<EV_A_> is used when other arguments are following. Example:
2450    
2451     ev_unref (EV_A);
2452     ev_timer_add (EV_A_ watcher);
2453     ev_loop (EV_A_ 0);
2454    
2455     It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
2456     which is often provided by the following macro.
2457    
2458     =item C<EV_P>, C<EV_P_>
2459    
2460     This provides the loop I<parameter> for functions, if one is required ("ev
2461     loop parameter"). The C<EV_P> form is used when this is the sole parameter,
2462     C<EV_P_> is used when other parameters are following. Example:
2463    
2464     // this is how ev_unref is being declared
2465     static void ev_unref (EV_P);
2466    
2467     // this is how you can declare your typical callback
2468     static void cb (EV_P_ ev_timer *w, int revents)
2469    
2470     It declares a parameter C<loop> of type C<struct ev_loop *>, quite
2471     suitable for use with C<EV_A>.
2472    
2473     =item C<EV_DEFAULT>, C<EV_DEFAULT_>
2474    
2475     Similar to the other two macros, this gives you the value of the default
2476     loop, if multiple loops are supported ("ev loop default").
2477    
2478     =back
2479    
2480 root 1.63 Example: Declare and initialise a check watcher, utilising the above
2481 root 1.68 macros so it will work regardless of whether multiple loops are supported
2482 root 1.63 or not.
2483 root 1.50
2484     static void
2485     check_cb (EV_P_ ev_timer *w, int revents)
2486     {
2487     ev_check_stop (EV_A_ w);
2488     }
2489    
2490     ev_check check;
2491     ev_check_init (&check, check_cb);
2492     ev_check_start (EV_DEFAULT_ &check);
2493     ev_loop (EV_DEFAULT_ 0);
2494    
2495 root 1.39 =head1 EMBEDDING
2496    
2497     Libev can (and often is) directly embedded into host
2498     applications. Examples of applications that embed it include the Deliantra
2499     Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
2500     and rxvt-unicode.
2501    
2502 root 1.91 The goal is to enable you to just copy the necessary files into your
2503 root 1.39 source directory without having to change even a single line in them, so
2504     you can easily upgrade by simply copying (or having a checked-out copy of
2505     libev somewhere in your source tree).
2506    
2507     =head2 FILESETS
2508    
2509     Depending on what features you need you need to include one or more sets of files
2510     in your app.
2511    
2512     =head3 CORE EVENT LOOP
2513    
2514     To include only the libev core (all the C<ev_*> functions), with manual
2515     configuration (no autoconf):
2516    
2517     #define EV_STANDALONE 1
2518     #include "ev.c"
2519    
2520     This will automatically include F<ev.h>, too, and should be done in a
2521     single C source file only to provide the function implementations. To use
2522     it, do the same for F<ev.h> in all files wishing to use this API (best
2523     done by writing a wrapper around F<ev.h> that you can include instead and
2524     where you can put other configuration options):
2525    
2526     #define EV_STANDALONE 1
2527     #include "ev.h"
2528    
2529     Both header files and implementation files can be compiled with a C++
2530     compiler (at least, thats a stated goal, and breakage will be treated
2531     as a bug).
2532    
2533     You need the following files in your source tree, or in a directory
2534     in your include path (e.g. in libev/ when using -Ilibev):
2535    
2536     ev.h
2537     ev.c
2538     ev_vars.h
2539     ev_wrap.h
2540    
2541     ev_win32.c required on win32 platforms only
2542    
2543 root 1.63 ev_select.c only when select backend is enabled (which is enabled by default)
2544 root 1.39 ev_poll.c only when poll backend is enabled (disabled by default)
2545     ev_epoll.c only when the epoll backend is enabled (disabled by default)
2546     ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
2547     ev_port.c only when the solaris port backend is enabled (disabled by default)
2548    
2549     F<ev.c> includes the backend files directly when enabled, so you only need
2550 root 1.43 to compile this single file.
2551 root 1.39
2552     =head3 LIBEVENT COMPATIBILITY API
2553    
2554     To include the libevent compatibility API, also include:
2555    
2556     #include "event.c"
2557    
2558     in the file including F<ev.c>, and:
2559    
2560     #include "event.h"
2561    
2562     in the files that want to use the libevent API. This also includes F<ev.h>.
2563    
2564     You need the following additional files for this:
2565    
2566     event.h
2567     event.c
2568    
2569     =head3 AUTOCONF SUPPORT
2570    
2571     Instead of using C<EV_STANDALONE=1> and providing your config in
2572     whatever way you want, you can also C<m4_include([libev.m4])> in your
2573 root 1.43 F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
2574     include F<config.h> and configure itself accordingly.
2575 root 1.39
2576     For this of course you need the m4 file:
2577    
2578     libev.m4
2579    
2580     =head2 PREPROCESSOR SYMBOLS/MACROS
2581    
2582     Libev can be configured via a variety of preprocessor symbols you have to define
2583     before including any of its files. The default is not to build for multiplicity
2584     and only include the select backend.
2585    
2586     =over 4
2587    
2588     =item EV_STANDALONE
2589    
2590     Must always be C<1> if you do not use autoconf configuration, which
2591     keeps libev from including F<config.h>, and it also defines dummy
2592     implementations for some libevent functions (such as logging, which is not
2593     supported). It will also not define any of the structs usually found in
2594     F<event.h> that are not directly supported by the libev core alone.
2595    
2596     =item EV_USE_MONOTONIC
2597    
2598     If defined to be C<1>, libev will try to detect the availability of the
2599     monotonic clock option at both compiletime and runtime. Otherwise no use
2600     of the monotonic clock option will be attempted. If you enable this, you
2601     usually have to link against librt or something similar. Enabling it when
2602 root 1.92 the functionality isn't available is safe, though, although you have
2603 root 1.39 to make sure you link against any libraries where the C<clock_gettime>
2604     function is hiding in (often F<-lrt>).
2605    
2606     =item EV_USE_REALTIME
2607    
2608     If defined to be C<1>, libev will try to detect the availability of the
2609     realtime clock option at compiletime (and assume its availability at
2610     runtime if successful). Otherwise no use of the realtime clock option will
2611     be attempted. This effectively replaces C<gettimeofday> by C<clock_get
2612 root 1.90 (CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
2613     note about libraries in the description of C<EV_USE_MONOTONIC>, though.
2614 root 1.39
2615 root 1.97 =item EV_USE_NANOSLEEP
2616    
2617     If defined to be C<1>, libev will assume that C<nanosleep ()> is available
2618     and will use it for delays. Otherwise it will use C<select ()>.
2619    
2620 root 1.39 =item EV_USE_SELECT
2621    
2622     If undefined or defined to be C<1>, libev will compile in support for the
2623     C<select>(2) backend. No attempt at autodetection will be done: if no
2624     other method takes over, select will be it. Otherwise the select backend
2625     will not be compiled in.
2626    
2627     =item EV_SELECT_USE_FD_SET
2628    
2629     If defined to C<1>, then the select backend will use the system C<fd_set>
2630     structure. This is useful if libev doesn't compile due to a missing
2631     C<NFDBITS> or C<fd_mask> definition or it misguesses the bitset layout on
2632     exotic systems. This usually limits the range of file descriptors to some
2633     low limit such as 1024 or might have other limitations (winsocket only
2634     allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
2635     influence the size of the C<fd_set> used.
2636    
2637     =item EV_SELECT_IS_WINSOCKET
2638    
2639     When defined to C<1>, the select backend will assume that
2640     select/socket/connect etc. don't understand file descriptors but
2641     wants osf handles on win32 (this is the case when the select to
2642     be used is the winsock select). This means that it will call
2643     C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2644     it is assumed that all these functions actually work on fds, even
2645     on win32. Should not be defined on non-win32 platforms.
2646    
2647 root 1.112 =item EV_FD_TO_WIN32_HANDLE
2648    
2649     If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
2650     file descriptors to socket handles. When not defining this symbol (the
2651     default), then libev will call C<_get_osfhandle>, which is usually
2652     correct. In some cases, programs use their own file descriptor management,
2653     in which case they can provide this function to map fds to socket handles.
2654    
2655 root 1.39 =item EV_USE_POLL
2656    
2657     If defined to be C<1>, libev will compile in support for the C<poll>(2)
2658     backend. Otherwise it will be enabled on non-win32 platforms. It
2659     takes precedence over select.
2660    
2661     =item EV_USE_EPOLL
2662    
2663     If defined to be C<1>, libev will compile in support for the Linux
2664     C<epoll>(7) backend. Its availability will be detected at runtime,
2665     otherwise another method will be used as fallback. This is the
2666     preferred backend for GNU/Linux systems.
2667    
2668     =item EV_USE_KQUEUE
2669    
2670     If defined to be C<1>, libev will compile in support for the BSD style
2671     C<kqueue>(2) backend. Its actual availability will be detected at runtime,
2672     otherwise another method will be used as fallback. This is the preferred
2673     backend for BSD and BSD-like systems, although on most BSDs kqueue only
2674     supports some types of fds correctly (the only platform we found that
2675     supports ptys for example was NetBSD), so kqueue might be compiled in, but
2676     not be used unless explicitly requested. The best way to use it is to find
2677 root 1.41 out whether kqueue supports your type of fd properly and use an embedded
2678 root 1.39 kqueue loop.
2679    
2680     =item EV_USE_PORT
2681    
2682     If defined to be C<1>, libev will compile in support for the Solaris
2683     10 port style backend. Its availability will be detected at runtime,
2684     otherwise another method will be used as fallback. This is the preferred
2685     backend for Solaris 10 systems.
2686    
2687     =item EV_USE_DEVPOLL
2688    
2689     reserved for future expansion, works like the USE symbols above.
2690    
2691 root 1.56 =item EV_USE_INOTIFY
2692    
2693     If defined to be C<1>, libev will compile in support for the Linux inotify
2694     interface to speed up C<ev_stat> watchers. Its actual availability will
2695     be detected at runtime.
2696    
2697 root 1.123 =item EV_ATOMIC_T
2698    
2699     Libev requires an integer type (suitable for storing C<0> or C<1>) whose
2700 root 1.126 access is atomic with respect to other threads or signal contexts. No such
2701     type is easily found in the C language, so you can provide your own type
2702 root 1.127 that you know is safe for your purposes. It is used both for signal handler "locking"
2703     as well as for signal and thread safety in C<ev_async> watchers.
2704 root 1.123
2705     In the absense of this define, libev will use C<sig_atomic_t volatile>
2706 root 1.126 (from F<signal.h>), which is usually good enough on most platforms.
2707 root 1.123
2708 root 1.39 =item EV_H
2709    
2710     The name of the F<ev.h> header file used to include it. The default if
2711 root 1.118 undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2712     used to virtually rename the F<ev.h> header file in case of conflicts.
2713 root 1.39
2714     =item EV_CONFIG_H
2715    
2716     If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2717     F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2718     C<EV_H>, above.
2719    
2720     =item EV_EVENT_H
2721    
2722     Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2723 root 1.118 of how the F<event.h> header can be found, the default is C<"event.h">.
2724 root 1.39
2725     =item EV_PROTOTYPES
2726    
2727     If defined to be C<0>, then F<ev.h> will not define any function
2728     prototypes, but still define all the structs and other symbols. This is
2729     occasionally useful if you want to provide your own wrapper functions
2730     around libev functions.
2731    
2732     =item EV_MULTIPLICITY
2733    
2734     If undefined or defined to C<1>, then all event-loop-specific functions
2735     will have the C<struct ev_loop *> as first argument, and you can create
2736     additional independent event loops. Otherwise there will be no support
2737     for multiple event loops and there is no first event loop pointer
2738     argument. Instead, all functions act on the single default loop.
2739    
2740 root 1.69 =item EV_MINPRI
2741    
2742     =item EV_MAXPRI
2743    
2744     The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
2745     C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
2746     provide for more priorities by overriding those symbols (usually defined
2747     to be C<-2> and C<2>, respectively).
2748    
2749     When doing priority-based operations, libev usually has to linearly search
2750     all the priorities, so having many of them (hundreds) uses a lot of space
2751     and time, so using the defaults of five priorities (-2 .. +2) is usually
2752     fine.
2753    
2754     If your embedding app does not need any priorities, defining these both to
2755     C<0> will save some memory and cpu.
2756    
2757 root 1.47 =item EV_PERIODIC_ENABLE
2758 root 1.39
2759 root 1.47 If undefined or defined to be C<1>, then periodic timers are supported. If
2760     defined to be C<0>, then they are not. Disabling them saves a few kB of
2761     code.
2762    
2763 root 1.67 =item EV_IDLE_ENABLE
2764    
2765     If undefined or defined to be C<1>, then idle watchers are supported. If
2766     defined to be C<0>, then they are not. Disabling them saves a few kB of
2767     code.
2768    
2769 root 1.47 =item EV_EMBED_ENABLE
2770    
2771     If undefined or defined to be C<1>, then embed watchers are supported. If
2772     defined to be C<0>, then they are not.
2773    
2774     =item EV_STAT_ENABLE
2775    
2776     If undefined or defined to be C<1>, then stat watchers are supported. If
2777     defined to be C<0>, then they are not.
2778    
2779 root 1.50 =item EV_FORK_ENABLE
2780    
2781     If undefined or defined to be C<1>, then fork watchers are supported. If
2782     defined to be C<0>, then they are not.
2783    
2784 root 1.123 =item EV_ASYNC_ENABLE
2785    
2786     If undefined or defined to be C<1>, then async watchers are supported. If
2787     defined to be C<0>, then they are not.
2788    
2789 root 1.47 =item EV_MINIMAL
2790    
2791     If you need to shave off some kilobytes of code at the expense of some
2792     speed, define this symbol to C<1>. Currently only used for gcc to override
2793     some inlining decisions, saves roughly 30% codesize of amd64.
2794 root 1.39
2795 root 1.51 =item EV_PID_HASHSIZE
2796    
2797     C<ev_child> watchers use a small hash table to distribute workload by
2798     pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
2799     than enough. If you need to manage thousands of children you might want to
2800 root 1.56 increase this value (I<must> be a power of two).
2801    
2802     =item EV_INOTIFY_HASHSIZE
2803    
2804 root 1.104 C<ev_stat> watchers use a small hash table to distribute workload by
2805 root 1.56 inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2806     usually more than enough. If you need to manage thousands of C<ev_stat>
2807     watchers you might want to increase this value (I<must> be a power of
2808     two).
2809 root 1.51
2810 root 1.39 =item EV_COMMON
2811    
2812     By default, all watchers have a C<void *data> member. By redefining
2813     this macro to a something else you can include more and other types of
2814     members. You have to define it each time you include one of the files,
2815     though, and it must be identical each time.
2816    
2817     For example, the perl EV module uses something like this:
2818    
2819     #define EV_COMMON \
2820     SV *self; /* contains this struct */ \
2821     SV *cb_sv, *fh /* note no trailing ";" */
2822    
2823 root 1.44 =item EV_CB_DECLARE (type)
2824 root 1.39
2825 root 1.44 =item EV_CB_INVOKE (watcher, revents)
2826 root 1.39
2827 root 1.44 =item ev_set_cb (ev, cb)
2828 root 1.39
2829     Can be used to change the callback member declaration in each watcher,
2830     and the way callbacks are invoked and set. Must expand to a struct member
2831 root 1.93 definition and a statement, respectively. See the F<ev.h> header file for
2832 root 1.39 their default definitions. One possible use for overriding these is to
2833 root 1.44 avoid the C<struct ev_loop *> as first argument in all cases, or to use
2834     method calls instead of plain function calls in C++.
2835 root 1.39
2836 root 1.89 =head2 EXPORTED API SYMBOLS
2837    
2838     If you need to re-export the API (e.g. via a dll) and you need a list of
2839     exported symbols, you can use the provided F<Symbol.*> files which list
2840     all public symbols, one per line:
2841    
2842     Symbols.ev for libev proper
2843     Symbols.event for the libevent emulation
2844    
2845     This can also be used to rename all public symbols to avoid clashes with
2846     multiple versions of libev linked together (which is obviously bad in
2847     itself, but sometimes it is inconvinient to avoid this).
2848    
2849 root 1.92 A sed command like this will create wrapper C<#define>'s that you need to
2850 root 1.89 include before including F<ev.h>:
2851    
2852     <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
2853    
2854     This would create a file F<wrap.h> which essentially looks like this:
2855    
2856     #define ev_backend myprefix_ev_backend
2857     #define ev_check_start myprefix_ev_check_start
2858     #define ev_check_stop myprefix_ev_check_stop
2859     ...
2860    
2861 root 1.39 =head2 EXAMPLES
2862    
2863     For a real-world example of a program the includes libev
2864     verbatim, you can have a look at the EV perl module
2865     (L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
2866     the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
2867     interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
2868     will be compiled. It is pretty complex because it provides its own header
2869     file.
2870    
2871     The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
2872 root 1.63 that everybody includes and which overrides some configure choices:
2873 root 1.39
2874 root 1.63 #define EV_MINIMAL 1
2875 root 1.40 #define EV_USE_POLL 0
2876     #define EV_MULTIPLICITY 0
2877 root 1.63 #define EV_PERIODIC_ENABLE 0
2878     #define EV_STAT_ENABLE 0
2879     #define EV_FORK_ENABLE 0
2880 root 1.40 #define EV_CONFIG_H <config.h>
2881 root 1.63 #define EV_MINPRI 0
2882     #define EV_MAXPRI 0
2883 root 1.39
2884 root 1.40 #include "ev++.h"
2885 root 1.39
2886     And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
2887    
2888 root 1.40 #include "ev_cpp.h"
2889     #include "ev.c"
2890 root 1.39
2891 root 1.46
2892     =head1 COMPLEXITIES
2893    
2894     In this section the complexities of (many of) the algorithms used inside
2895     libev will be explained. For complexity discussions about backends see the
2896     documentation for C<ev_default_init>.
2897    
2898 root 1.70 All of the following are about amortised time: If an array needs to be
2899     extended, libev needs to realloc and move the whole array, but this
2900     happens asymptotically never with higher number of elements, so O(1) might
2901     mean it might do a lengthy realloc operation in rare cases, but on average
2902     it is much faster and asymptotically approaches constant time.
2903    
2904 root 1.46 =over 4
2905    
2906     =item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
2907    
2908 root 1.69 This means that, when you have a watcher that triggers in one hour and
2909     there are 100 watchers that would trigger before that then inserting will
2910 root 1.106 have to skip roughly seven (C<ld 100>) of these watchers.
2911 root 1.69
2912 root 1.106 =item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2913 root 1.46
2914 root 1.106 That means that changing a timer costs less than removing/adding them
2915 root 1.69 as only the relative motion in the event queue has to be paid for.
2916    
2917 root 1.46 =item Starting io/check/prepare/idle/signal/child watchers: O(1)
2918    
2919 root 1.70 These just add the watcher into an array or at the head of a list.
2920 root 1.106
2921 root 1.46 =item Stopping check/prepare/idle watchers: O(1)
2922    
2923 root 1.56 =item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2924 root 1.46
2925 root 1.69 These watchers are stored in lists then need to be walked to find the
2926     correct watcher to remove. The lists are usually short (you don't usually
2927     have many watchers waiting for the same fd or signal).
2928    
2929 root 1.106 =item Finding the next timer in each loop iteration: O(1)
2930    
2931     By virtue of using a binary heap, the next timer is always found at the
2932     beginning of the storage array.
2933 root 1.46
2934     =item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
2935    
2936 root 1.69 A change means an I/O watcher gets started or stopped, which requires
2937 root 1.106 libev to recalculate its status (and possibly tell the kernel, depending
2938     on backend and wether C<ev_io_set> was used).
2939 root 1.69
2940 root 1.106 =item Activating one watcher (putting it into the pending state): O(1)
2941 root 1.46
2942 root 1.69 =item Priority handling: O(number_of_priorities)
2943    
2944     Priorities are implemented by allocating some space for each
2945     priority. When doing priority-based operations, libev usually has to
2946 root 1.106 linearly search all the priorities, but starting/stopping and activating
2947     watchers becomes O(1) w.r.t. prioritiy handling.
2948 root 1.69
2949 root 1.46 =back
2950    
2951    
2952 root 1.112 =head1 Win32 platform limitations and workarounds
2953    
2954     Win32 doesn't support any of the standards (e.g. POSIX) that libev
2955     requires, and its I/O model is fundamentally incompatible with the POSIX
2956     model. Libev still offers limited functionality on this platform in
2957     the form of the C<EVBACKEND_SELECT> backend, and only supports socket
2958     descriptors. This only applies when using Win32 natively, not when using
2959     e.g. cygwin.
2960    
2961     There is no supported compilation method available on windows except
2962     embedding it into other applications.
2963    
2964     Due to the many, low, and arbitrary limits on the win32 platform and the
2965     abysmal performance of winsockets, using a large number of sockets is not
2966     recommended (and not reasonable). If your program needs to use more than
2967     a hundred or so sockets, then likely it needs to use a totally different
2968     implementation for windows, as libev offers the POSIX model, which cannot
2969     be implemented efficiently on windows (microsoft monopoly games).
2970    
2971     =over 4
2972    
2973     =item The winsocket select function
2974    
2975     The winsocket C<select> function doesn't follow POSIX in that it requires
2976     socket I<handles> and not socket I<file descriptors>. This makes select
2977     very inefficient, and also requires a mapping from file descriptors
2978     to socket handles. See the discussion of the C<EV_SELECT_USE_FD_SET>,
2979     C<EV_SELECT_IS_WINSOCKET> and C<EV_FD_TO_WIN32_HANDLE> preprocessor
2980     symbols for more info.
2981    
2982     The configuration for a "naked" win32 using the microsoft runtime
2983     libraries and raw winsocket select is:
2984    
2985     #define EV_USE_SELECT 1
2986     #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
2987    
2988     Note that winsockets handling of fd sets is O(n), so you can easily get a
2989     complexity in the O(n²) range when using win32.
2990    
2991     =item Limited number of file descriptors
2992    
2993     Windows has numerous arbitrary (and low) limits on things. Early versions
2994     of winsocket's select only supported waiting for a max. of C<64> handles
2995     (probably owning to the fact that all windows kernels can only wait for
2996     C<64> things at the same time internally; microsoft recommends spawning a
2997     chain of threads and wait for 63 handles and the previous thread in each).
2998    
2999     Newer versions support more handles, but you need to define C<FD_SETSIZE>
3000     to some high number (e.g. C<2048>) before compiling the winsocket select
3001     call (which might be in libev or elsewhere, for example, perl does its own
3002     select emulation on windows).
3003    
3004     Another limit is the number of file descriptors in the microsoft runtime
3005     libraries, which by default is C<64> (there must be a hidden I<64> fetish
3006     or something like this inside microsoft). You can increase this by calling
3007     C<_setmaxstdio>, which can increase this limit to C<2048> (another
3008     arbitrary limit), but is broken in many versions of the microsoft runtime
3009     libraries.
3010    
3011     This might get you to about C<512> or C<2048> sockets (depending on
3012     windows version and/or the phase of the moon). To get more, you need to
3013     wrap all I/O functions and provide your own fd management, but the cost of
3014     calling select (O(n²)) will likely make this unworkable.
3015    
3016     =back
3017    
3018    
3019 root 1.1 =head1 AUTHOR
3020    
3021     Marc Lehmann <libev@schmorp.de>.
3022