ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev.pod
(Generate patch)

Comparing libev/ev.pod (file contents):
Revision 1.76 by root, Sat Dec 8 15:30:30 2007 UTC vs.
Revision 1.118 by root, Thu Jan 10 06:00:55 2008 UTC

… …
4 4
5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 #include <ev.h> 7 #include <ev.h>
8 8
9=head1 EXAMPLE PROGRAM 9=head2 EXAMPLE PROGRAM
10 10
11 #include <ev.h> 11 #include <ev.h>
12 12
13 ev_io stdin_watcher; 13 ev_io stdin_watcher;
14 ev_timer timeout_watcher; 14 ev_timer timeout_watcher;
… …
53The newest version of this document is also available as a html-formatted 53The newest version of this document is also available as a html-formatted
54web page you might find easier to navigate when reading it for the first 54web page you might find easier to navigate when reading it for the first
55time: L<http://cvs.schmorp.de/libev/ev.html>. 55time: L<http://cvs.schmorp.de/libev/ev.html>.
56 56
57Libev is an event loop: you register interest in certain events (such as a 57Libev is an event loop: you register interest in certain events (such as a
58file descriptor being readable or a timeout occuring), and it will manage 58file descriptor being readable or a timeout occurring), and it will manage
59these event sources and provide your program with events. 59these event sources and provide your program with events.
60 60
61To do this, it must take more or less complete control over your process 61To 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 62(or thread) by executing the I<event loop> handler, and will then
63communicate events via a callback mechanism. 63communicate events via a callback mechanism.
… …
65You register interest in certain events by registering so-called I<event 65You register interest in certain events by registering so-called I<event
66watchers>, which are relatively small C structures you initialise with the 66watchers>, which are relatively small C structures you initialise with the
67details of the event, and then hand it over to libev by I<starting> the 67details of the event, and then hand it over to libev by I<starting> the
68watcher. 68watcher.
69 69
70=head1 FEATURES 70=head2 FEATURES
71 71
72Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the 72Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
73BSD-specific C<kqueue> and the Solaris-specific event port mechanisms 73BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
74for file descriptor events (C<ev_io>), the Linux C<inotify> interface 74for file descriptor events (C<ev_io>), the Linux C<inotify> interface
75(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers 75(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
… …
82 82
83It also is quite fast (see this 83It also is quite fast (see this
84L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent 84L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
85for example). 85for example).
86 86
87=head1 CONVENTIONS 87=head2 CONVENTIONS
88 88
89Libev is very configurable. In this manual the default configuration will 89Libev is very configurable. In this manual the default configuration will
90be described, which supports multiple event loops. For more info about 90be described, which supports multiple event loops. For more info about
91various configuration options please have a look at B<EMBED> section in 91various configuration options please have a look at B<EMBED> section in
92this manual. If libev was configured without support for multiple event 92this manual. If libev was configured without support for multiple event
93loops, then all functions taking an initial argument of name C<loop> 93loops, 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. 94(which is always of type C<struct ev_loop *>) will not have this argument.
95 95
96=head1 TIME REPRESENTATION 96=head2 TIME REPRESENTATION
97 97
98Libev represents time as a single floating point number, representing the 98Libev represents time as a single floating point number, representing the
99(fractional) number of seconds since the (POSIX) epoch (somewhere near 99(fractional) number of seconds since the (POSIX) epoch (somewhere near
100the beginning of 1970, details are complicated, don't ask). This type is 100the beginning of 1970, details are complicated, don't ask). This type is
101called C<ev_tstamp>, which is what you should use too. It usually aliases 101called C<ev_tstamp>, which is what you should use too. It usually aliases
102to the C<double> type in C, and when you need to do any calculations on 102to the C<double> type in C, and when you need to do any calculations on
103it, you should treat it as such. 103it, you should treat it as some floatingpoint value. Unlike the name
104component C<stamp> might indicate, it is also used for time differences
105throughout libev.
104 106
105=head1 GLOBAL FUNCTIONS 107=head1 GLOBAL FUNCTIONS
106 108
107These functions can be called anytime, even before initialising the 109These functions can be called anytime, even before initialising the
108library in any way. 110library in any way.
… …
113 115
114Returns the current time as libev would use it. Please note that the 116Returns the current time as libev would use it. Please note that the
115C<ev_now> function is usually faster and also often returns the timestamp 117C<ev_now> function is usually faster and also often returns the timestamp
116you actually want to know. 118you actually want to know.
117 119
120=item ev_sleep (ev_tstamp interval)
121
122Sleep for the given interval: The current thread will be blocked until
123either it is interrupted or the given time interval has passed. Basically
124this is a subsecond-resolution C<sleep ()>.
125
118=item int ev_version_major () 126=item int ev_version_major ()
119 127
120=item int ev_version_minor () 128=item int ev_version_minor ()
121 129
122You can find out the major and minor version numbers of the library 130You can find out the major and minor ABI version numbers of the library
123you linked against by calling the functions C<ev_version_major> and 131you linked against by calling the functions C<ev_version_major> and
124C<ev_version_minor>. If you want, you can compare against the global 132C<ev_version_minor>. If you want, you can compare against the global
125symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the 133symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
126version of the library your program was compiled against. 134version of the library your program was compiled against.
127 135
136These version numbers refer to the ABI version of the library, not the
137release version.
138
128Usually, it's a good idea to terminate if the major versions mismatch, 139Usually, it's a good idea to terminate if the major versions mismatch,
129as this indicates an incompatible change. Minor versions are usually 140as this indicates an incompatible change. Minor versions are usually
130compatible to older versions, so a larger minor version alone is usually 141compatible to older versions, so a larger minor version alone is usually
131not a problem. 142not a problem.
132 143
133Example: Make sure we haven't accidentally been linked against the wrong 144Example: Make sure we haven't accidentally been linked against the wrong
134version. 145version.
… …
249flags. If that is troubling you, check C<ev_backend ()> afterwards). 260flags. If that is troubling you, check C<ev_backend ()> afterwards).
250 261
251If you don't know what event loop to use, use the one returned from this 262If you don't know what event loop to use, use the one returned from this
252function. 263function.
253 264
265The default loop is the only loop that can handle C<ev_signal> and
266C<ev_child> watchers, and to do this, it always registers a handler
267for C<SIGCHLD>. If this is a problem for your app you can either
268create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
269can simply overwrite the C<SIGCHLD> signal handler I<after> calling
270C<ev_default_init>.
271
254The flags argument can be used to specify special behaviour or specific 272The flags argument can be used to specify special behaviour or specific
255backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>). 273backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
256 274
257The following flags are supported: 275The following flags are supported:
258 276
… …
295=item C<EVBACKEND_SELECT> (value 1, portable select backend) 313=item C<EVBACKEND_SELECT> (value 1, portable select backend)
296 314
297This is your standard select(2) backend. Not I<completely> standard, as 315This is your standard select(2) backend. Not I<completely> standard, as
298libev tries to roll its own fd_set with no limits on the number of fds, 316libev tries to roll its own fd_set with no limits on the number of fds,
299but if that fails, expect a fairly low limit on the number of fds when 317but if that fails, expect a fairly low limit on the number of fds when
300using this backend. It doesn't scale too well (O(highest_fd)), but its usually 318using this backend. It doesn't scale too well (O(highest_fd)), but its
301the fastest backend for a low number of fds. 319usually the fastest backend for a low number of (low-numbered :) fds.
320
321To get good performance out of this backend you need a high amount of
322parallelity (most of the file descriptors should be busy). If you are
323writing a server, you should C<accept ()> in a loop to accept as many
324connections as possible during one iteration. You might also want to have
325a look at C<ev_set_io_collect_interval ()> to increase the amount of
326readyness notifications you get per iteration.
302 327
303=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows) 328=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
304 329
305And this is your standard poll(2) backend. It's more complicated than 330And this is your standard poll(2) backend. It's more complicated
306select, but handles sparse fds better and has no artificial limit on the 331than select, but handles sparse fds better and has no artificial
307number of fds you can use (except it will slow down considerably with a 332limit on the number of fds you can use (except it will slow down
308lot of inactive fds). It scales similarly to select, i.e. O(total_fds). 333considerably with a lot of inactive fds). It scales similarly to select,
334i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
335performance tips.
309 336
310=item C<EVBACKEND_EPOLL> (value 4, Linux) 337=item C<EVBACKEND_EPOLL> (value 4, Linux)
311 338
312For few fds, this backend is a bit little slower than poll and select, 339For few fds, this backend is a bit little slower than poll and select,
313but it scales phenomenally better. While poll and select usually scale like 340but it scales phenomenally better. While poll and select usually scale
314O(total_fds) where n is the total number of fds (or the highest fd), epoll scales 341like O(total_fds) where n is the total number of fds (or the highest fd),
315either O(1) or O(active_fds). 342epoll scales either O(1) or O(active_fds). The epoll design has a number
343of shortcomings, such as silently dropping events in some hard-to-detect
344cases and rewiring a syscall per fd change, no fork support and bad
345support for dup.
316 346
317While stopping and starting an I/O watcher in the same iteration will 347While stopping, setting and starting an I/O watcher in the same iteration
318result in some caching, there is still a syscall per such incident 348will result in some caching, there is still a syscall per such incident
319(because the fd could point to a different file description now), so its 349(because the fd could point to a different file description now), so its
320best to avoid that. Also, dup()ed file descriptors might not work very 350best to avoid that. Also, C<dup ()>'ed file descriptors might not work
321well if you register events for both fds. 351very well if you register events for both fds.
322 352
323Please note that epoll sometimes generates spurious notifications, so you 353Please note that epoll sometimes generates spurious notifications, so you
324need to use non-blocking I/O or other means to avoid blocking when no data 354need to use non-blocking I/O or other means to avoid blocking when no data
325(or space) is available. 355(or space) is available.
326 356
357Best performance from this backend is achieved by not unregistering all
358watchers for a file descriptor until it has been closed, if possible, i.e.
359keep at least one watcher active per fd at all times.
360
361While nominally embeddeble in other event loops, this feature is broken in
362all kernel versions tested so far.
363
327=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones) 364=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
328 365
329Kqueue deserves special mention, as at the time of this writing, it 366Kqueue deserves special mention, as at the time of this writing, it
330was broken on all BSDs except NetBSD (usually it doesn't work with 367was broken on all BSDs except NetBSD (usually it doesn't work reliably
331anything but sockets and pipes, except on Darwin, where of course its 368with anything but sockets and pipes, except on Darwin, where of course
332completely useless). For this reason its not being "autodetected" 369it's completely useless). For this reason it's not being "autodetected"
333unless you explicitly specify it explicitly in the flags (i.e. using 370unless you explicitly specify it explicitly in the flags (i.e. using
334C<EVBACKEND_KQUEUE>). 371C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
372system like NetBSD.
373
374You still can embed kqueue into a normal poll or select backend and use it
375only for sockets (after having made sure that sockets work with kqueue on
376the target platform). See C<ev_embed> watchers for more info.
335 377
336It scales in the same way as the epoll backend, but the interface to the 378It scales in the same way as the epoll backend, but the interface to the
337kernel is more efficient (which says nothing about its actual speed, of 379kernel is more efficient (which says nothing about its actual speed, of
338course). While starting and stopping an I/O watcher does not cause an 380course). While stopping, setting and starting an I/O watcher does never
339extra syscall as with epoll, it still adds up to four event changes per 381cause an extra syscall as with C<EVBACKEND_EPOLL>, it still adds up to
340incident, so its best to avoid that. 382two event changes per incident, support for C<fork ()> is very bad and it
383drops fds silently in similarly hard-to-detect cases.
384
385This backend usually performs well under most conditions.
386
387While nominally embeddable in other event loops, this doesn't work
388everywhere, so you might need to test for this. And since it is broken
389almost 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
392sockets.
341 393
342=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8) 394=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
343 395
344This is not implemented yet (and might never be). 396This is not implemented yet (and might never be, unless you send me an
397implementation). According to reports, C</dev/poll> only supports sockets
398and is not embeddable, which would limit the usefulness of this backend
399immensely.
345 400
346=item C<EVBACKEND_PORT> (value 32, Solaris 10) 401=item C<EVBACKEND_PORT> (value 32, Solaris 10)
347 402
348This uses the Solaris 10 port mechanism. As with everything on Solaris, 403This uses the Solaris 10 event port mechanism. As with everything on Solaris,
349it's really slow, but it still scales very well (O(active_fds)). 404it's really slow, but it still scales very well (O(active_fds)).
350 405
351Please note that solaris ports can result in a lot of spurious 406Please note that solaris event ports can deliver a lot of spurious
352notifications, so you need to use non-blocking I/O or other means to avoid 407notifications, so you need to use non-blocking I/O or other means to avoid
353blocking when no data (or space) is available. 408blocking when no data (or space) is available.
409
410While this backend scales well, it requires one system call per active
411file descriptor per loop iteration. For small and medium numbers of file
412descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
413might perform better.
414
415On the positive side, ignoring the spurious readyness notifications, this
416backend actually performed to specification in all tests and is fully
417embeddable, which is a rare feat among the OS-specific backends.
354 418
355=item C<EVBACKEND_ALL> 419=item C<EVBACKEND_ALL>
356 420
357Try all backends (even potentially broken ones that wouldn't be tried 421Try all backends (even potentially broken ones that wouldn't be tried
358with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as 422with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
359C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>. 423C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
360 424
425It is definitely not recommended to use this flag.
426
361=back 427=back
362 428
363If one or more of these are ored into the flags value, then only these 429If one or more of these are ored into the flags value, then only these
364backends will be tried (in the reverse order as given here). If none are 430backends will be tried (in the reverse order as listed here). If none are
365specified, most compiled-in backend will be tried, usually in reverse 431specified, all backends in C<ev_recommended_backends ()> will be tried.
366order of their flag values :)
367 432
368The most typical usage is like this: 433The most typical usage is like this:
369 434
370 if (!ev_default_loop (0)) 435 if (!ev_default_loop (0))
371 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?"); 436 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
… …
399Destroys the default loop again (frees all memory and kernel state 464Destroys the default loop again (frees all memory and kernel state
400etc.). None of the active event watchers will be stopped in the normal 465etc.). None of the active event watchers will be stopped in the normal
401sense, so e.g. C<ev_is_active> might still return true. It is your 466sense, so e.g. C<ev_is_active> might still return true. It is your
402responsibility to either stop all watchers cleanly yoursef I<before> 467responsibility to either stop all watchers cleanly yoursef I<before>
403calling this function, or cope with the fact afterwards (which is usually 468calling this function, or cope with the fact afterwards (which is usually
404the easiest thing, youc na just ignore the watchers and/or C<free ()> them 469the easiest thing, you can just ignore the watchers and/or C<free ()> them
405for example). 470for example).
471
472Note that certain global state, such as signal state, will not be freed by
473this function, and related watchers (such as signal and child watchers)
474would need to be stopped manually.
475
476In general it is not advisable to call this function except in the
477rare occasion where you really need to free e.g. the signal handling
478pipe fds. If you need dynamically allocated loops it is better to use
479C<ev_loop_new> and C<ev_loop_destroy>).
406 480
407=item ev_loop_destroy (loop) 481=item ev_loop_destroy (loop)
408 482
409Like C<ev_default_destroy>, but destroys an event loop created by an 483Like C<ev_default_destroy>, but destroys an event loop created by an
410earlier call to C<ev_loop_new>. 484earlier call to C<ev_loop_new>.
… …
455 529
456Returns the current "event loop time", which is the time the event loop 530Returns the current "event loop time", which is the time the event loop
457received events and started processing them. This timestamp does not 531received events and started processing them. This timestamp does not
458change as long as callbacks are being processed, and this is also the base 532change as long as callbacks are being processed, and this is also the base
459time used for relative timers. You can treat it as the timestamp of the 533time used for relative timers. You can treat it as the timestamp of the
460event occuring (or more correctly, libev finding out about it). 534event occurring (or more correctly, libev finding out about it).
461 535
462=item ev_loop (loop, int flags) 536=item ev_loop (loop, int flags)
463 537
464Finally, this is it, the event handler. This function usually is called 538Finally, this is it, the event handler. This function usually is called
465after you initialised all your watchers and you want to start handling 539after you initialised all your watchers and you want to start handling
… …
486libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is 560libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is
487usually a better approach for this kind of thing. 561usually a better approach for this kind of thing.
488 562
489Here are the gory details of what C<ev_loop> does: 563Here are the gory details of what C<ev_loop> does:
490 564
491 * If there are no active watchers (reference count is zero), return. 565 - Before the first iteration, call any pending watchers.
492 - Queue prepare watchers and then call all outstanding watchers. 566 * If EVFLAG_FORKCHECK was used, check for a fork.
567 - If a fork was detected, queue and call all fork watchers.
568 - Queue and call all prepare watchers.
493 - If we have been forked, recreate the kernel state. 569 - If we have been forked, recreate the kernel state.
494 - Update the kernel state with all outstanding changes. 570 - Update the kernel state with all outstanding changes.
495 - Update the "event loop time". 571 - Update the "event loop time".
496 - Calculate for how long to block. 572 - Calculate for how long to sleep or block, if at all
573 (active idle watchers, EVLOOP_NONBLOCK or not having
574 any active watchers at all will result in not sleeping).
575 - Sleep if the I/O and timer collect interval say so.
497 - Block the process, waiting for any events. 576 - Block the process, waiting for any events.
498 - Queue all outstanding I/O (fd) events. 577 - Queue all outstanding I/O (fd) events.
499 - Update the "event loop time" and do time jump handling. 578 - Update the "event loop time" and do time jump handling.
500 - Queue all outstanding timers. 579 - Queue all outstanding timers.
501 - Queue all outstanding periodics. 580 - Queue all outstanding periodics.
502 - If no events are pending now, queue all idle watchers. 581 - If no events are pending now, queue all idle watchers.
503 - Queue all check watchers. 582 - Queue all check watchers.
504 - Call all queued watchers in reverse order (i.e. check watchers first). 583 - Call all queued watchers in reverse order (i.e. check watchers first).
505 Signals and child watchers are implemented as I/O watchers, and will 584 Signals and child watchers are implemented as I/O watchers, and will
506 be handled here by queueing them when their watcher gets executed. 585 be handled here by queueing them when their watcher gets executed.
507 - If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK 586 - If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
508 were used, return, otherwise continue with step *. 587 were used, or there are no active watchers, return, otherwise
588 continue with step *.
509 589
510Example: Queue some jobs and then loop until no events are outsanding 590Example: Queue some jobs and then loop until no events are outstanding
511anymore. 591anymore.
512 592
513 ... queue jobs here, make sure they register event watchers as long 593 ... queue jobs here, make sure they register event watchers as long
514 ... as they still have work to do (even an idle watcher will do..) 594 ... as they still have work to do (even an idle watcher will do..)
515 ev_loop (my_loop, 0); 595 ev_loop (my_loop, 0);
… …
519 599
520Can be used to make a call to C<ev_loop> return early (but only after it 600Can be used to make a call to C<ev_loop> return early (but only after it
521has processed all outstanding events). The C<how> argument must be either 601has processed all outstanding events). The C<how> argument must be either
522C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or 602C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
523C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return. 603C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
604
605This "unloop state" will be cleared when entering C<ev_loop> again.
524 606
525=item ev_ref (loop) 607=item ev_ref (loop)
526 608
527=item ev_unref (loop) 609=item ev_unref (loop)
528 610
… …
533returning, ev_unref() after starting, and ev_ref() before stopping it. For 615returning, ev_unref() after starting, and ev_ref() before stopping it. For
534example, libev itself uses this for its internal signal pipe: It is not 616example, libev itself uses this for its internal signal pipe: It is not
535visible to the libev user and should not keep C<ev_loop> from exiting if 617visible to the libev user and should not keep C<ev_loop> from exiting if
536no event watchers registered by it are active. It is also an excellent 618no event watchers registered by it are active. It is also an excellent
537way to do this for generic recurring timers or from within third-party 619way to do this for generic recurring timers or from within third-party
538libraries. Just remember to I<unref after start> and I<ref before stop>. 620libraries. Just remember to I<unref after start> and I<ref before stop>
621(but only if the watcher wasn't active before, or was active before,
622respectively).
539 623
540Example: Create a signal watcher, but keep it from keeping C<ev_loop> 624Example: Create a signal watcher, but keep it from keeping C<ev_loop>
541running when nothing else is active. 625running when nothing else is active.
542 626
543 struct ev_signal exitsig; 627 struct ev_signal exitsig;
… …
547 631
548Example: For some weird reason, unregister the above signal handler again. 632Example: For some weird reason, unregister the above signal handler again.
549 633
550 ev_ref (loop); 634 ev_ref (loop);
551 ev_signal_stop (loop, &exitsig); 635 ev_signal_stop (loop, &exitsig);
636
637=item ev_set_io_collect_interval (loop, ev_tstamp interval)
638
639=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
640
641These advanced functions influence the time that libev will spend waiting
642for events. Both are by default C<0>, meaning that libev will try to
643invoke timer/periodic callbacks and I/O callbacks with minimum latency.
644
645Setting these to a higher value (the C<interval> I<must> be >= C<0>)
646allows libev to delay invocation of I/O and timer/periodic callbacks to
647increase efficiency of loop iterations.
648
649The background is that sometimes your program runs just fast enough to
650handle one (or very few) event(s) per loop iteration. While this makes
651the program responsive, it also wastes a lot of CPU time to poll for new
652events, especially with backends like C<select ()> which have a high
653overhead for the actual polling but can deliver many events at once.
654
655By setting a higher I<io collect interval> you allow libev to spend more
656time collecting I/O events, so you can handle more events per iteration,
657at the cost of increasing latency. Timeouts (both C<ev_periodic> and
658C<ev_timer>) will be not affected. Setting this to a non-null value will
659introduce an additional C<ev_sleep ()> call into most loop iterations.
660
661Likewise, by setting a higher I<timeout collect interval> you allow libev
662to spend more time collecting timeouts, at the expense of increased
663latency (the watcher callback will be called later). C<ev_io> watchers
664will not be affected. Setting this to a non-null value will not introduce
665any overhead in libev.
666
667Many (busy) programs can usually benefit by setting the io collect
668interval to a value near C<0.1> or so, which is often enough for
669interactive servers (of course not for games), likewise for timeouts. It
670usually doesn't make much sense to set it to a lower value than C<0.01>,
671as this approsaches the timing granularity of most systems.
552 672
553=back 673=back
554 674
555 675
556=head1 ANATOMY OF A WATCHER 676=head1 ANATOMY OF A WATCHER
… …
882In general you can register as many read and/or write event watchers per 1002In general you can register as many read and/or write event watchers per
883fd as you want (as long as you don't confuse yourself). Setting all file 1003fd as you want (as long as you don't confuse yourself). Setting all file
884descriptors to non-blocking mode is also usually a good idea (but not 1004descriptors to non-blocking mode is also usually a good idea (but not
885required if you know what you are doing). 1005required if you know what you are doing).
886 1006
887You have to be careful with dup'ed file descriptors, though. Some backends
888(the linux epoll backend is a notable example) cannot handle dup'ed file
889descriptors correctly if you register interest in two or more fds pointing
890to the same underlying file/socket/etc. description (that is, they share
891the same underlying "file open").
892
893If you must do this, then force the use of a known-to-be-good backend 1007If you must do this, then force the use of a known-to-be-good backend
894(at the time of this writing, this includes only C<EVBACKEND_SELECT> and 1008(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
895C<EVBACKEND_POLL>). 1009C<EVBACKEND_POLL>).
896 1010
897Another thing you have to watch out for is that it is quite easy to 1011Another thing you have to watch out for is that it is quite easy to
… …
907play around with an Xlib connection), then you have to seperately re-test 1021play around with an Xlib connection), then you have to seperately re-test
908whether a file descriptor is really ready with a known-to-be good interface 1022whether a file descriptor is really ready with a known-to-be good interface
909such as poll (fortunately in our Xlib example, Xlib already does this on 1023such as poll (fortunately in our Xlib example, Xlib already does this on
910its own, so its quite safe to use). 1024its own, so its quite safe to use).
911 1025
1026=head3 The special problem of disappearing file descriptors
1027
1028Some backends (e.g. kqueue, epoll) need to be told about closing a file
1029descriptor (either by calling C<close> explicitly or by any other means,
1030such as C<dup>). The reason is that you register interest in some file
1031descriptor, but when it goes away, the operating system will silently drop
1032this interest. If another file descriptor with the same number then is
1033registered with libev, there is no efficient way to see that this is, in
1034fact, a different file descriptor.
1035
1036To avoid having to explicitly tell libev about such cases, libev follows
1037the following policy: Each time C<ev_io_set> is being called, libev
1038will assume that this is potentially a new file descriptor, otherwise
1039it is assumed that the file descriptor stays the same. That means that
1040you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1041descriptor even if the file descriptor number itself did not change.
1042
1043This is how one would do it normally anyway, the important point is that
1044the libev application should not optimise around libev but should leave
1045optimisations to libev.
1046
1047=head3 The special problem of dup'ed file descriptors
1048
1049Some backends (e.g. epoll), cannot register events for file descriptors,
1050but only events for the underlying file descriptions. That means when you
1051have C<dup ()>'ed file descriptors or weirder constellations, and register
1052events for them, only one file descriptor might actually receive events.
1053
1054There is no workaround possible except not registering events
1055for potentially C<dup ()>'ed file descriptors, or to resort to
1056C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1057
1058=head3 The special problem of fork
1059
1060Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1061useless behaviour. Libev fully supports fork, but needs to be told about
1062it in the child.
1063
1064To support fork in your programs, you either have to call
1065C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
1066enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1067C<EVBACKEND_POLL>.
1068
1069
1070=head3 Watcher-Specific Functions
1071
912=over 4 1072=over 4
913 1073
914=item ev_io_init (ev_io *, callback, int fd, int events) 1074=item ev_io_init (ev_io *, callback, int fd, int events)
915 1075
916=item ev_io_set (ev_io *, int fd, int events) 1076=item ev_io_set (ev_io *, int fd, int events)
… …
926=item int events [read-only] 1086=item int events [read-only]
927 1087
928The events being watched. 1088The events being watched.
929 1089
930=back 1090=back
1091
1092=head3 Examples
931 1093
932Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well 1094Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
933readable, but only once. Since it is likely line-buffered, you could 1095readable, but only once. Since it is likely line-buffered, you could
934attempt to read a whole line in the callback. 1096attempt to read a whole line in the callback.
935 1097
… …
968 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.); 1130 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
969 1131
970The callback is guarenteed to be invoked only when its timeout has passed, 1132The callback is guarenteed to be invoked only when its timeout has passed,
971but if multiple timers become ready during the same loop iteration then 1133but if multiple timers become ready during the same loop iteration then
972order of execution is undefined. 1134order of execution is undefined.
1135
1136=head3 Watcher-Specific Functions and Data Members
973 1137
974=over 4 1138=over 4
975 1139
976=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat) 1140=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
977 1141
… …
1031or C<ev_timer_again> is called and determines the next timeout (if any), 1195or C<ev_timer_again> is called and determines the next timeout (if any),
1032which is also when any modifications are taken into account. 1196which is also when any modifications are taken into account.
1033 1197
1034=back 1198=back
1035 1199
1200=head3 Examples
1201
1036Example: Create a timer that fires after 60 seconds. 1202Example: Create a timer that fires after 60 seconds.
1037 1203
1038 static void 1204 static void
1039 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents) 1205 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1040 { 1206 {
… …
1073but on wallclock time (absolute time). You can tell a periodic watcher 1239but on wallclock time (absolute time). You can tell a periodic watcher
1074to trigger "at" some specific point in time. For example, if you tell a 1240to trigger "at" some specific point in time. For example, if you tell a
1075periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now () 1241periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()
1076+ 10.>) and then reset your system clock to the last year, then it will 1242+ 10.>) and then reset your system clock to the last year, then it will
1077take a year to trigger the event (unlike an C<ev_timer>, which would trigger 1243take a year to trigger the event (unlike an C<ev_timer>, which would trigger
1078roughly 10 seconds later and of course not if you reset your system time 1244roughly 10 seconds later).
1079again).
1080 1245
1081They can also be used to implement vastly more complex timers, such as 1246They can also be used to implement vastly more complex timers, such as
1082triggering an event on eahc midnight, local time. 1247triggering an event on each midnight, local time or other, complicated,
1248rules.
1083 1249
1084As with timers, the callback is guarenteed to be invoked only when the 1250As with timers, the callback is guarenteed to be invoked only when the
1085time (C<at>) has been passed, but if multiple periodic timers become ready 1251time (C<at>) has been passed, but if multiple periodic timers become ready
1086during the same loop iteration then order of execution is undefined. 1252during the same loop iteration then order of execution is undefined.
1087 1253
1254=head3 Watcher-Specific Functions and Data Members
1255
1088=over 4 1256=over 4
1089 1257
1090=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb) 1258=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
1091 1259
1092=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb) 1260=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
… …
1094Lots of arguments, lets sort it out... There are basically three modes of 1262Lots of arguments, lets sort it out... There are basically three modes of
1095operation, and we will explain them from simplest to complex: 1263operation, and we will explain them from simplest to complex:
1096 1264
1097=over 4 1265=over 4
1098 1266
1099=item * absolute timer (interval = reschedule_cb = 0) 1267=item * absolute timer (at = time, interval = reschedule_cb = 0)
1100 1268
1101In this configuration the watcher triggers an event at the wallclock time 1269In this configuration the watcher triggers an event at the wallclock time
1102C<at> and doesn't repeat. It will not adjust when a time jump occurs, 1270C<at> and doesn't repeat. It will not adjust when a time jump occurs,
1103that is, if it is to be run at January 1st 2011 then it will run when the 1271that is, if it is to be run at January 1st 2011 then it will run when the
1104system time reaches or surpasses this time. 1272system time reaches or surpasses this time.
1105 1273
1106=item * non-repeating interval timer (interval > 0, reschedule_cb = 0) 1274=item * non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1107 1275
1108In this mode the watcher will always be scheduled to time out at the next 1276In this mode the watcher will always be scheduled to time out at the next
1109C<at + N * interval> time (for some integer N) and then repeat, regardless 1277C<at + N * interval> time (for some integer N, which can also be negative)
1110of any time jumps. 1278and then repeat, regardless of any time jumps.
1111 1279
1112This can be used to create timers that do not drift with respect to system 1280This can be used to create timers that do not drift with respect to system
1113time: 1281time:
1114 1282
1115 ev_periodic_set (&periodic, 0., 3600., 0); 1283 ev_periodic_set (&periodic, 0., 3600., 0);
… …
1121 1289
1122Another way to think about it (for the mathematically inclined) is that 1290Another way to think about it (for the mathematically inclined) is that
1123C<ev_periodic> will try to run the callback in this mode at the next possible 1291C<ev_periodic> will try to run the callback in this mode at the next possible
1124time where C<time = at (mod interval)>, regardless of any time jumps. 1292time where C<time = at (mod interval)>, regardless of any time jumps.
1125 1293
1294For numerical stability it is preferable that the C<at> value is near
1295C<ev_now ()> (the current time), but there is no range requirement for
1296this value.
1297
1126=item * manual reschedule mode (reschedule_cb = callback) 1298=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1127 1299
1128In this mode the values for C<interval> and C<at> are both being 1300In this mode the values for C<interval> and C<at> are both being
1129ignored. Instead, each time the periodic watcher gets scheduled, the 1301ignored. Instead, each time the periodic watcher gets scheduled, the
1130reschedule callback will be called with the watcher as first, and the 1302reschedule callback will be called with the watcher as first, and the
1131current time as second argument. 1303current time as second argument.
1132 1304
1133NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, 1305NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
1134ever, or make any event loop modifications>. If you need to stop it, 1306ever, or make any event loop modifications>. If you need to stop it,
1135return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by 1307return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
1136starting a prepare watcher). 1308starting an C<ev_prepare> watcher, which is legal).
1137 1309
1138Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 1310Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
1139ev_tstamp now)>, e.g.: 1311ev_tstamp now)>, e.g.:
1140 1312
1141 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) 1313 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
… …
1164Simply stops and restarts the periodic watcher again. This is only useful 1336Simply stops and restarts the periodic watcher again. This is only useful
1165when you changed some parameters or the reschedule callback would return 1337when you changed some parameters or the reschedule callback would return
1166a different time than the last time it was called (e.g. in a crond like 1338a different time than the last time it was called (e.g. in a crond like
1167program when the crontabs have changed). 1339program when the crontabs have changed).
1168 1340
1341=item ev_tstamp offset [read-write]
1342
1343When repeating, this contains the offset value, otherwise this is the
1344absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1345
1346Can be modified any time, but changes only take effect when the periodic
1347timer fires or C<ev_periodic_again> is being called.
1348
1169=item ev_tstamp interval [read-write] 1349=item ev_tstamp interval [read-write]
1170 1350
1171The current interval value. Can be modified any time, but changes only 1351The current interval value. Can be modified any time, but changes only
1172take effect when the periodic timer fires or C<ev_periodic_again> is being 1352take effect when the periodic timer fires or C<ev_periodic_again> is being
1173called. 1353called.
… …
1176 1356
1177The current reschedule callback, or C<0>, if this functionality is 1357The current reschedule callback, or C<0>, if this functionality is
1178switched off. Can be changed any time, but changes only take effect when 1358switched off. Can be changed any time, but changes only take effect when
1179the periodic timer fires or C<ev_periodic_again> is being called. 1359the periodic timer fires or C<ev_periodic_again> is being called.
1180 1360
1361=item ev_tstamp at [read-only]
1362
1363When active, contains the absolute time that the watcher is supposed to
1364trigger next.
1365
1181=back 1366=back
1367
1368=head3 Examples
1182 1369
1183Example: Call a callback every hour, or, more precisely, whenever the 1370Example: Call a callback every hour, or, more precisely, whenever the
1184system clock is divisible by 3600. The callback invocation times have 1371system clock is divisible by 3600. The callback invocation times have
1185potentially a lot of jittering, but good long-term stability. 1372potentially a lot of jittering, but good long-term stability.
1186 1373
… …
1226with the kernel (thus it coexists with your own signal handlers as long 1413with the kernel (thus it coexists with your own signal handlers as long
1227as you don't register any with libev). Similarly, when the last signal 1414as you don't register any with libev). Similarly, when the last signal
1228watcher for a signal is stopped libev will reset the signal handler to 1415watcher for a signal is stopped libev will reset the signal handler to
1229SIG_DFL (regardless of what it was set to before). 1416SIG_DFL (regardless of what it was set to before).
1230 1417
1418=head3 Watcher-Specific Functions and Data Members
1419
1231=over 4 1420=over 4
1232 1421
1233=item ev_signal_init (ev_signal *, callback, int signum) 1422=item ev_signal_init (ev_signal *, callback, int signum)
1234 1423
1235=item ev_signal_set (ev_signal *, int signum) 1424=item ev_signal_set (ev_signal *, int signum)
… …
1246 1435
1247=head2 C<ev_child> - watch out for process status changes 1436=head2 C<ev_child> - watch out for process status changes
1248 1437
1249Child watchers trigger when your process receives a SIGCHLD in response to 1438Child watchers trigger when your process receives a SIGCHLD in response to
1250some child status changes (most typically when a child of yours dies). 1439some child status changes (most typically when a child of yours dies).
1440
1441=head3 Watcher-Specific Functions and Data Members
1251 1442
1252=over 4 1443=over 4
1253 1444
1254=item ev_child_init (ev_child *, callback, int pid) 1445=item ev_child_init (ev_child *, callback, int pid)
1255 1446
… …
1274 1465
1275The process exit/trace status caused by C<rpid> (see your systems 1466The process exit/trace status caused by C<rpid> (see your systems
1276C<waitpid> and C<sys/wait.h> documentation for details). 1467C<waitpid> and C<sys/wait.h> documentation for details).
1277 1468
1278=back 1469=back
1470
1471=head3 Examples
1279 1472
1280Example: Try to exit cleanly on SIGINT and SIGTERM. 1473Example: Try to exit cleanly on SIGINT and SIGTERM.
1281 1474
1282 static void 1475 static void
1283 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents) 1476 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
… …
1324semantics of C<ev_stat> watchers, which means that libev sometimes needs 1517semantics of C<ev_stat> watchers, which means that libev sometimes needs
1325to fall back to regular polling again even with inotify, but changes are 1518to fall back to regular polling again even with inotify, but changes are
1326usually detected immediately, and if the file exists there will be no 1519usually detected immediately, and if the file exists there will be no
1327polling. 1520polling.
1328 1521
1522=head3 Inotify
1523
1524When C<inotify (7)> support has been compiled into libev (generally only
1525available on Linux) and present at runtime, it will be used to speed up
1526change detection where possible. The inotify descriptor will be created lazily
1527when the first C<ev_stat> watcher is being started.
1528
1529Inotify presense does not change the semantics of C<ev_stat> watchers
1530except that changes might be detected earlier, and in some cases, to avoid
1531making regular C<stat> calls. Even in the presense of inotify support
1532there are many cases where libev has to resort to regular C<stat> polling.
1533
1534(There is no support for kqueue, as apparently it cannot be used to
1535implement this functionality, due to the requirement of having a file
1536descriptor open on the object at all times).
1537
1538=head3 The special problem of stat time resolution
1539
1540The C<stat ()> syscall only supports full-second resolution portably, and
1541even on systems where the resolution is higher, many filesystems still
1542only support whole seconds.
1543
1544That means that, if the time is the only thing that changes, you might
1545miss updates: on the first update, C<ev_stat> detects a change and calls
1546your callback, which does something. When there is another update within
1547the same second, C<ev_stat> will be unable to detect it.
1548
1549The solution to this is to delay acting on a change for a second (or till
1550the next second boundary), using a roughly one-second delay C<ev_timer>
1551(C<ev_timer_set (w, 0., 1.01); ev_timer_again (loop, w)>). The C<.01>
1552is added to work around small timing inconsistencies of some operating
1553systems.
1554
1555=head3 Watcher-Specific Functions and Data Members
1556
1329=over 4 1557=over 4
1330 1558
1331=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval) 1559=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
1332 1560
1333=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval) 1561=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
… …
1368=item const char *path [read-only] 1596=item const char *path [read-only]
1369 1597
1370The filesystem path that is being watched. 1598The filesystem path that is being watched.
1371 1599
1372=back 1600=back
1601
1602=head3 Examples
1373 1603
1374Example: Watch C</etc/passwd> for attribute changes. 1604Example: Watch C</etc/passwd> for attribute changes.
1375 1605
1376 static void 1606 static void
1377 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents) 1607 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
… …
1390 } 1620 }
1391 1621
1392 ... 1622 ...
1393 ev_stat passwd; 1623 ev_stat passwd;
1394 1624
1395 ev_stat_init (&passwd, passwd_cb, "/etc/passwd"); 1625 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1396 ev_stat_start (loop, &passwd); 1626 ev_stat_start (loop, &passwd);
1627
1628Example: Like above, but additionally use a one-second delay so we do not
1629miss updates (however, frequent updates will delay processing, too, so
1630one might do the work both on C<ev_stat> callback invocation I<and> on
1631C<ev_timer> callback invocation).
1632
1633 static ev_stat passwd;
1634 static ev_timer timer;
1635
1636 static void
1637 timer_cb (EV_P_ ev_timer *w, int revents)
1638 {
1639 ev_timer_stop (EV_A_ w);
1640
1641 /* now it's one second after the most recent passwd change */
1642 }
1643
1644 static void
1645 stat_cb (EV_P_ ev_stat *w, int revents)
1646 {
1647 /* reset the one-second timer */
1648 ev_timer_again (EV_A_ &timer);
1649 }
1650
1651 ...
1652 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
1653 ev_stat_start (loop, &passwd);
1654 ev_timer_init (&timer, timer_cb, 0., 1.01);
1397 1655
1398 1656
1399=head2 C<ev_idle> - when you've got nothing better to do... 1657=head2 C<ev_idle> - when you've got nothing better to do...
1400 1658
1401Idle watchers trigger events when no other events of the same or higher 1659Idle watchers trigger events when no other events of the same or higher
… …
1415Apart from keeping your process non-blocking (which is a useful 1673Apart from keeping your process non-blocking (which is a useful
1416effect on its own sometimes), idle watchers are a good place to do 1674effect on its own sometimes), idle watchers are a good place to do
1417"pseudo-background processing", or delay processing stuff to after the 1675"pseudo-background processing", or delay processing stuff to after the
1418event loop has handled all outstanding events. 1676event loop has handled all outstanding events.
1419 1677
1678=head3 Watcher-Specific Functions and Data Members
1679
1420=over 4 1680=over 4
1421 1681
1422=item ev_idle_init (ev_signal *, callback) 1682=item ev_idle_init (ev_signal *, callback)
1423 1683
1424Initialises and configures the idle watcher - it has no parameters of any 1684Initialises and configures the idle watcher - it has no parameters of any
1425kind. There is a C<ev_idle_set> macro, but using it is utterly pointless, 1685kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1426believe me. 1686believe me.
1427 1687
1428=back 1688=back
1689
1690=head3 Examples
1429 1691
1430Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the 1692Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1431callback, free it. Also, use no error checking, as usual. 1693callback, free it. Also, use no error checking, as usual.
1432 1694
1433 static void 1695 static void
… …
1481with priority higher than or equal to the event loop and one coroutine 1743with priority higher than or equal to the event loop and one coroutine
1482of lower priority, but only once, using idle watchers to keep the event 1744of lower priority, but only once, using idle watchers to keep the event
1483loop from blocking if lower-priority coroutines are active, thus mapping 1745loop from blocking if lower-priority coroutines are active, thus mapping
1484low-priority coroutines to idle/background tasks). 1746low-priority coroutines to idle/background tasks).
1485 1747
1748It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1749priority, to ensure that they are being run before any other watchers
1750after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
1751too) should not activate ("feed") events into libev. While libev fully
1752supports this, they will be called before other C<ev_check> watchers
1753did their job. As C<ev_check> watchers are often used to embed other
1754(non-libev) event loops those other event loops might be in an unusable
1755state until their C<ev_check> watcher ran (always remind yourself to
1756coexist peacefully with others).
1757
1758=head3 Watcher-Specific Functions and Data Members
1759
1486=over 4 1760=over 4
1487 1761
1488=item ev_prepare_init (ev_prepare *, callback) 1762=item ev_prepare_init (ev_prepare *, callback)
1489 1763
1490=item ev_check_init (ev_check *, callback) 1764=item ev_check_init (ev_check *, callback)
… …
1492Initialises and configures the prepare or check watcher - they have no 1766Initialises and configures the prepare or check watcher - they have no
1493parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set> 1767parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1494macros, but using them is utterly, utterly and completely pointless. 1768macros, but using them is utterly, utterly and completely pointless.
1495 1769
1496=back 1770=back
1771
1772=head3 Examples
1497 1773
1498There are a number of principal ways to embed other event loops or modules 1774There are a number of principal ways to embed other event loops or modules
1499into libev. Here are some ideas on how to include libadns into libev 1775into libev. Here are some ideas on how to include libadns into libev
1500(there is a Perl module named C<EV::ADNS> that does this, which you could 1776(there is a Perl module named C<EV::ADNS> that does this, which you could
1501use for an actually working example. Another Perl module named C<EV::Glib> 1777use for an actually working example. Another Perl module named C<EV::Glib>
… …
1670portable one. 1946portable one.
1671 1947
1672So when you want to use this feature you will always have to be prepared 1948So when you want to use this feature you will always have to be prepared
1673that you cannot get an embeddable loop. The recommended way to get around 1949that you cannot get an embeddable loop. The recommended way to get around
1674this is to have a separate variables for your embeddable loop, try to 1950this is to have a separate variables for your embeddable loop, try to
1675create it, and if that fails, use the normal loop for everything: 1951create it, and if that fails, use the normal loop for everything.
1952
1953=head3 Watcher-Specific Functions and Data Members
1954
1955=over 4
1956
1957=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
1958
1959=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
1960
1961Configures the watcher to embed the given loop, which must be
1962embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1963invoked automatically, otherwise it is the responsibility of the callback
1964to invoke it (it will continue to be called until the sweep has been done,
1965if you do not want thta, you need to temporarily stop the embed watcher).
1966
1967=item ev_embed_sweep (loop, ev_embed *)
1968
1969Make a single, non-blocking sweep over the embedded loop. This works
1970similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1971apropriate way for embedded loops.
1972
1973=item struct ev_loop *other [read-only]
1974
1975The embedded event loop.
1976
1977=back
1978
1979=head3 Examples
1980
1981Example: Try to get an embeddable event loop and embed it into the default
1982event loop. If that is not possible, use the default loop. The default
1983loop is stored in C<loop_hi>, while the mebeddable loop is stored in
1984C<loop_lo> (which is C<loop_hi> in the acse no embeddable loop can be
1985used).
1676 1986
1677 struct ev_loop *loop_hi = ev_default_init (0); 1987 struct ev_loop *loop_hi = ev_default_init (0);
1678 struct ev_loop *loop_lo = 0; 1988 struct ev_loop *loop_lo = 0;
1679 struct ev_embed embed; 1989 struct ev_embed embed;
1680 1990
… …
1691 ev_embed_start (loop_hi, &embed); 2001 ev_embed_start (loop_hi, &embed);
1692 } 2002 }
1693 else 2003 else
1694 loop_lo = loop_hi; 2004 loop_lo = loop_hi;
1695 2005
1696=over 4 2006Example: Check if kqueue is available but not recommended and create
2007a kqueue backend for use with sockets (which usually work with any
2008kqueue implementation). Store the kqueue/socket-only event loop in
2009C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
1697 2010
1698=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop) 2011 struct ev_loop *loop = ev_default_init (0);
2012 struct ev_loop *loop_socket = 0;
2013 struct ev_embed embed;
2014
2015 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2016 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2017 {
2018 ev_embed_init (&embed, 0, loop_socket);
2019 ev_embed_start (loop, &embed);
2020 }
1699 2021
1700=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop) 2022 if (!loop_socket)
2023 loop_socket = loop;
1701 2024
1702Configures the watcher to embed the given loop, which must be 2025 // now use loop_socket for all sockets, and loop for everything else
1703embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1704invoked automatically, otherwise it is the responsibility of the callback
1705to invoke it (it will continue to be called until the sweep has been done,
1706if you do not want thta, you need to temporarily stop the embed watcher).
1707
1708=item ev_embed_sweep (loop, ev_embed *)
1709
1710Make a single, non-blocking sweep over the embedded loop. This works
1711similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1712apropriate way for embedded loops.
1713
1714=item struct ev_loop *loop [read-only]
1715
1716The embedded event loop.
1717
1718=back
1719 2026
1720 2027
1721=head2 C<ev_fork> - the audacity to resume the event loop after a fork 2028=head2 C<ev_fork> - the audacity to resume the event loop after a fork
1722 2029
1723Fork watchers are called when a C<fork ()> was detected (usually because 2030Fork watchers are called when a C<fork ()> was detected (usually because
… …
1726event loop blocks next and before C<ev_check> watchers are being called, 2033event loop blocks next and before C<ev_check> watchers are being called,
1727and only in the child after the fork. If whoever good citizen calling 2034and only in the child after the fork. If whoever good citizen calling
1728C<ev_default_fork> cheats and calls it in the wrong process, the fork 2035C<ev_default_fork> cheats and calls it in the wrong process, the fork
1729handlers will be invoked, too, of course. 2036handlers will be invoked, too, of course.
1730 2037
2038=head3 Watcher-Specific Functions and Data Members
2039
1731=over 4 2040=over 4
1732 2041
1733=item ev_fork_init (ev_signal *, callback) 2042=item ev_fork_init (ev_signal *, callback)
1734 2043
1735Initialises and configures the fork watcher - it has no parameters of any 2044Initialises and configures the fork watcher - it has no parameters of any
… …
1951 2260
1952=item w->stop () 2261=item w->stop ()
1953 2262
1954Stops the watcher if it is active. Again, no C<loop> argument. 2263Stops the watcher if it is active. Again, no C<loop> argument.
1955 2264
1956=item w->again () C<ev::timer>, C<ev::periodic> only 2265=item w->again () (C<ev::timer>, C<ev::periodic> only)
1957 2266
1958For C<ev::timer> and C<ev::periodic>, this invokes the corresponding 2267For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
1959C<ev_TYPE_again> function. 2268C<ev_TYPE_again> function.
1960 2269
1961=item w->sweep () C<ev::embed> only 2270=item w->sweep () (C<ev::embed> only)
1962 2271
1963Invokes C<ev_embed_sweep>. 2272Invokes C<ev_embed_sweep>.
1964 2273
1965=item w->update () C<ev::stat> only 2274=item w->update () (C<ev::stat> only)
1966 2275
1967Invokes C<ev_stat_stat>. 2276Invokes C<ev_stat_stat>.
1968 2277
1969=back 2278=back
1970 2279
… …
1990 } 2299 }
1991 2300
1992 2301
1993=head1 MACRO MAGIC 2302=head1 MACRO MAGIC
1994 2303
1995Libev can be compiled with a variety of options, the most fundemantal is 2304Libev can be compiled with a variety of options, the most fundamantal
1996C<EV_MULTIPLICITY>. This option determines whether (most) functions and 2305of which is C<EV_MULTIPLICITY>. This option determines whether (most)
1997callbacks have an initial C<struct ev_loop *> argument. 2306functions and callbacks have an initial C<struct ev_loop *> argument.
1998 2307
1999To make it easier to write programs that cope with either variant, the 2308To make it easier to write programs that cope with either variant, the
2000following macros are defined: 2309following macros are defined:
2001 2310
2002=over 4 2311=over 4
… …
2056Libev can (and often is) directly embedded into host 2365Libev can (and often is) directly embedded into host
2057applications. Examples of applications that embed it include the Deliantra 2366applications. Examples of applications that embed it include the Deliantra
2058Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe) 2367Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
2059and rxvt-unicode. 2368and rxvt-unicode.
2060 2369
2061The goal is to enable you to just copy the neecssary files into your 2370The goal is to enable you to just copy the necessary files into your
2062source directory without having to change even a single line in them, so 2371source directory without having to change even a single line in them, so
2063you can easily upgrade by simply copying (or having a checked-out copy of 2372you can easily upgrade by simply copying (or having a checked-out copy of
2064libev somewhere in your source tree). 2373libev somewhere in your source tree).
2065 2374
2066=head2 FILESETS 2375=head2 FILESETS
… …
2156 2465
2157If defined to be C<1>, libev will try to detect the availability of the 2466If defined to be C<1>, libev will try to detect the availability of the
2158monotonic clock option at both compiletime and runtime. Otherwise no use 2467monotonic clock option at both compiletime and runtime. Otherwise no use
2159of the monotonic clock option will be attempted. If you enable this, you 2468of the monotonic clock option will be attempted. If you enable this, you
2160usually have to link against librt or something similar. Enabling it when 2469usually have to link against librt or something similar. Enabling it when
2161the functionality isn't available is safe, though, althoguh you have 2470the functionality isn't available is safe, though, although you have
2162to make sure you link against any libraries where the C<clock_gettime> 2471to make sure you link against any libraries where the C<clock_gettime>
2163function is hiding in (often F<-lrt>). 2472function is hiding in (often F<-lrt>).
2164 2473
2165=item EV_USE_REALTIME 2474=item EV_USE_REALTIME
2166 2475
2167If defined to be C<1>, libev will try to detect the availability of the 2476If defined to be C<1>, libev will try to detect the availability of the
2168realtime clock option at compiletime (and assume its availability at 2477realtime clock option at compiletime (and assume its availability at
2169runtime if successful). Otherwise no use of the realtime clock option will 2478runtime if successful). Otherwise no use of the realtime clock option will
2170be attempted. This effectively replaces C<gettimeofday> by C<clock_get 2479be attempted. This effectively replaces C<gettimeofday> by C<clock_get
2171(CLOCK_REALTIME, ...)> and will not normally affect correctness. See tzhe note about libraries 2480(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
2172in the description of C<EV_USE_MONOTONIC>, though. 2481note about libraries in the description of C<EV_USE_MONOTONIC>, though.
2482
2483=item EV_USE_NANOSLEEP
2484
2485If defined to be C<1>, libev will assume that C<nanosleep ()> is available
2486and will use it for delays. Otherwise it will use C<select ()>.
2173 2487
2174=item EV_USE_SELECT 2488=item EV_USE_SELECT
2175 2489
2176If undefined or defined to be C<1>, libev will compile in support for the 2490If undefined or defined to be C<1>, libev will compile in support for the
2177C<select>(2) backend. No attempt at autodetection will be done: if no 2491C<select>(2) backend. No attempt at autodetection will be done: if no
… …
2195wants osf handles on win32 (this is the case when the select to 2509wants osf handles on win32 (this is the case when the select to
2196be used is the winsock select). This means that it will call 2510be used is the winsock select). This means that it will call
2197C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise, 2511C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2198it is assumed that all these functions actually work on fds, even 2512it is assumed that all these functions actually work on fds, even
2199on win32. Should not be defined on non-win32 platforms. 2513on win32. Should not be defined on non-win32 platforms.
2514
2515=item EV_FD_TO_WIN32_HANDLE
2516
2517If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
2518file descriptors to socket handles. When not defining this symbol (the
2519default), then libev will call C<_get_osfhandle>, which is usually
2520correct. In some cases, programs use their own file descriptor management,
2521in which case they can provide this function to map fds to socket handles.
2200 2522
2201=item EV_USE_POLL 2523=item EV_USE_POLL
2202 2524
2203If defined to be C<1>, libev will compile in support for the C<poll>(2) 2525If defined to be C<1>, libev will compile in support for the C<poll>(2)
2204backend. Otherwise it will be enabled on non-win32 platforms. It 2526backend. Otherwise it will be enabled on non-win32 platforms. It
… …
2241be detected at runtime. 2563be detected at runtime.
2242 2564
2243=item EV_H 2565=item EV_H
2244 2566
2245The name of the F<ev.h> header file used to include it. The default if 2567The name of the F<ev.h> header file used to include it. The default if
2246undefined is C<< <ev.h> >> in F<event.h> and C<"ev.h"> in F<ev.c>. This 2568undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2247can be used to virtually rename the F<ev.h> header file in case of conflicts. 2569used to virtually rename the F<ev.h> header file in case of conflicts.
2248 2570
2249=item EV_CONFIG_H 2571=item EV_CONFIG_H
2250 2572
2251If C<EV_STANDALONE> isn't C<1>, this variable can be used to override 2573If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2252F<ev.c>'s idea of where to find the F<config.h> file, similarly to 2574F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2253C<EV_H>, above. 2575C<EV_H>, above.
2254 2576
2255=item EV_EVENT_H 2577=item EV_EVENT_H
2256 2578
2257Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea 2579Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2258of how the F<event.h> header can be found. 2580of how the F<event.h> header can be found, the default is C<"event.h">.
2259 2581
2260=item EV_PROTOTYPES 2582=item EV_PROTOTYPES
2261 2583
2262If defined to be C<0>, then F<ev.h> will not define any function 2584If defined to be C<0>, then F<ev.h> will not define any function
2263prototypes, but still define all the structs and other symbols. This is 2585prototypes, but still define all the structs and other symbols. This is
… …
2329than enough. If you need to manage thousands of children you might want to 2651than enough. If you need to manage thousands of children you might want to
2330increase this value (I<must> be a power of two). 2652increase this value (I<must> be a power of two).
2331 2653
2332=item EV_INOTIFY_HASHSIZE 2654=item EV_INOTIFY_HASHSIZE
2333 2655
2334C<ev_staz> watchers use a small hash table to distribute workload by 2656C<ev_stat> watchers use a small hash table to distribute workload by
2335inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>), 2657inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2336usually more than enough. If you need to manage thousands of C<ev_stat> 2658usually more than enough. If you need to manage thousands of C<ev_stat>
2337watchers you might want to increase this value (I<must> be a power of 2659watchers you might want to increase this value (I<must> be a power of
2338two). 2660two).
2339 2661
… …
2356 2678
2357=item ev_set_cb (ev, cb) 2679=item ev_set_cb (ev, cb)
2358 2680
2359Can be used to change the callback member declaration in each watcher, 2681Can be used to change the callback member declaration in each watcher,
2360and the way callbacks are invoked and set. Must expand to a struct member 2682and the way callbacks are invoked and set. Must expand to a struct member
2361definition and a statement, respectively. See the F<ev.v> header file for 2683definition and a statement, respectively. See the F<ev.h> header file for
2362their default definitions. One possible use for overriding these is to 2684their default definitions. One possible use for overriding these is to
2363avoid the C<struct ev_loop *> as first argument in all cases, or to use 2685avoid the C<struct ev_loop *> as first argument in all cases, or to use
2364method calls instead of plain function calls in C++. 2686method calls instead of plain function calls in C++.
2687
2688=head2 EXPORTED API SYMBOLS
2689
2690If you need to re-export the API (e.g. via a dll) and you need a list of
2691exported symbols, you can use the provided F<Symbol.*> files which list
2692all public symbols, one per line:
2693
2694 Symbols.ev for libev proper
2695 Symbols.event for the libevent emulation
2696
2697This can also be used to rename all public symbols to avoid clashes with
2698multiple versions of libev linked together (which is obviously bad in
2699itself, but sometimes it is inconvinient to avoid this).
2700
2701A sed command like this will create wrapper C<#define>'s that you need to
2702include before including F<ev.h>:
2703
2704 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
2705
2706This would create a file F<wrap.h> which essentially looks like this:
2707
2708 #define ev_backend myprefix_ev_backend
2709 #define ev_check_start myprefix_ev_check_start
2710 #define ev_check_stop myprefix_ev_check_stop
2711 ...
2365 2712
2366=head2 EXAMPLES 2713=head2 EXAMPLES
2367 2714
2368For a real-world example of a program the includes libev 2715For a real-world example of a program the includes libev
2369verbatim, you can have a look at the EV perl module 2716verbatim, you can have a look at the EV perl module
… …
2410 2757
2411=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers) 2758=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
2412 2759
2413This means that, when you have a watcher that triggers in one hour and 2760This means that, when you have a watcher that triggers in one hour and
2414there are 100 watchers that would trigger before that then inserting will 2761there are 100 watchers that would trigger before that then inserting will
2415have to skip those 100 watchers. 2762have to skip roughly seven (C<ld 100>) of these watchers.
2416 2763
2417=item Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers) 2764=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2418 2765
2419That means that for changing a timer costs less than removing/adding them 2766That means that changing a timer costs less than removing/adding them
2420as only the relative motion in the event queue has to be paid for. 2767as only the relative motion in the event queue has to be paid for.
2421 2768
2422=item Starting io/check/prepare/idle/signal/child watchers: O(1) 2769=item Starting io/check/prepare/idle/signal/child watchers: O(1)
2423 2770
2424These just add the watcher into an array or at the head of a list. 2771These just add the watcher into an array or at the head of a list.
2772
2425=item Stopping check/prepare/idle watchers: O(1) 2773=item Stopping check/prepare/idle watchers: O(1)
2426 2774
2427=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE)) 2775=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2428 2776
2429These watchers are stored in lists then need to be walked to find the 2777These watchers are stored in lists then need to be walked to find the
2430correct watcher to remove. The lists are usually short (you don't usually 2778correct watcher to remove. The lists are usually short (you don't usually
2431have many watchers waiting for the same fd or signal). 2779have many watchers waiting for the same fd or signal).
2432 2780
2433=item Finding the next timer per loop iteration: O(1) 2781=item Finding the next timer in each loop iteration: O(1)
2782
2783By virtue of using a binary heap, the next timer is always found at the
2784beginning of the storage array.
2434 2785
2435=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd) 2786=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
2436 2787
2437A change means an I/O watcher gets started or stopped, which requires 2788A change means an I/O watcher gets started or stopped, which requires
2438libev to recalculate its status (and possibly tell the kernel). 2789libev to recalculate its status (and possibly tell the kernel, depending
2790on backend and wether C<ev_io_set> was used).
2439 2791
2440=item Activating one watcher: O(1) 2792=item Activating one watcher (putting it into the pending state): O(1)
2441 2793
2442=item Priority handling: O(number_of_priorities) 2794=item Priority handling: O(number_of_priorities)
2443 2795
2444Priorities are implemented by allocating some space for each 2796Priorities are implemented by allocating some space for each
2445priority. When doing priority-based operations, libev usually has to 2797priority. When doing priority-based operations, libev usually has to
2446linearly search all the priorities. 2798linearly search all the priorities, but starting/stopping and activating
2799watchers becomes O(1) w.r.t. prioritiy handling.
2447 2800
2448=back 2801=back
2449 2802
2450 2803
2804=head1 Win32 platform limitations and workarounds
2805
2806Win32 doesn't support any of the standards (e.g. POSIX) that libev
2807requires, and its I/O model is fundamentally incompatible with the POSIX
2808model. Libev still offers limited functionality on this platform in
2809the form of the C<EVBACKEND_SELECT> backend, and only supports socket
2810descriptors. This only applies when using Win32 natively, not when using
2811e.g. cygwin.
2812
2813There is no supported compilation method available on windows except
2814embedding it into other applications.
2815
2816Due to the many, low, and arbitrary limits on the win32 platform and the
2817abysmal performance of winsockets, using a large number of sockets is not
2818recommended (and not reasonable). If your program needs to use more than
2819a hundred or so sockets, then likely it needs to use a totally different
2820implementation for windows, as libev offers the POSIX model, which cannot
2821be implemented efficiently on windows (microsoft monopoly games).
2822
2823=over 4
2824
2825=item The winsocket select function
2826
2827The winsocket C<select> function doesn't follow POSIX in that it requires
2828socket I<handles> and not socket I<file descriptors>. This makes select
2829very inefficient, and also requires a mapping from file descriptors
2830to socket handles. See the discussion of the C<EV_SELECT_USE_FD_SET>,
2831C<EV_SELECT_IS_WINSOCKET> and C<EV_FD_TO_WIN32_HANDLE> preprocessor
2832symbols for more info.
2833
2834The configuration for a "naked" win32 using the microsoft runtime
2835libraries and raw winsocket select is:
2836
2837 #define EV_USE_SELECT 1
2838 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
2839
2840Note that winsockets handling of fd sets is O(n), so you can easily get a
2841complexity in the O(n²) range when using win32.
2842
2843=item Limited number of file descriptors
2844
2845Windows has numerous arbitrary (and low) limits on things. Early versions
2846of winsocket's select only supported waiting for a max. of C<64> handles
2847(probably owning to the fact that all windows kernels can only wait for
2848C<64> things at the same time internally; microsoft recommends spawning a
2849chain of threads and wait for 63 handles and the previous thread in each).
2850
2851Newer versions support more handles, but you need to define C<FD_SETSIZE>
2852to some high number (e.g. C<2048>) before compiling the winsocket select
2853call (which might be in libev or elsewhere, for example, perl does its own
2854select emulation on windows).
2855
2856Another limit is the number of file descriptors in the microsoft runtime
2857libraries, which by default is C<64> (there must be a hidden I<64> fetish
2858or something like this inside microsoft). You can increase this by calling
2859C<_setmaxstdio>, which can increase this limit to C<2048> (another
2860arbitrary limit), but is broken in many versions of the microsoft runtime
2861libraries.
2862
2863This might get you to about C<512> or C<2048> sockets (depending on
2864windows version and/or the phase of the moon). To get more, you need to
2865wrap all I/O functions and provide your own fd management, but the cost of
2866calling select (O(n²)) will likely make this unworkable.
2867
2868=back
2869
2870
2451=head1 AUTHOR 2871=head1 AUTHOR
2452 2872
2453Marc Lehmann <libev@schmorp.de>. 2873Marc Lehmann <libev@schmorp.de>.
2454 2874

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines