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

Comparing libev/ev.pod (file contents):
Revision 1.63 by root, Thu Nov 29 20:05:59 2007 UTC vs.
Revision 1.119 by root, Tue Jan 15 04:07:37 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;
48 return 0; 48 return 0;
49 } 49 }
50 50
51=head1 DESCRIPTION 51=head1 DESCRIPTION
52 52
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
55time: L<http://cvs.schmorp.de/libev/ev.html>.
56
53Libev 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
54file descriptor being readable or a timeout occuring), and it will manage 58file descriptor being readable or a timeout occurring), and it will manage
55these event sources and provide your program with events. 59these event sources and provide your program with events.
56 60
57To 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
58(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
59communicate events via a callback mechanism. 63communicate events via a callback mechanism.
61You register interest in certain events by registering so-called I<event 65You register interest in certain events by registering so-called I<event
62watchers>, which are relatively small C structures you initialise with the 66watchers>, which are relatively small C structures you initialise with the
63details 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
64watcher. 68watcher.
65 69
66=head1 FEATURES 70=head2 FEATURES
67 71
68Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the 72Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
69BSD-specific C<kqueue> and the Solaris-specific event port mechanisms 73BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
70for file descriptor events (C<ev_io>), the Linux C<inotify> interface 74for file descriptor events (C<ev_io>), the Linux C<inotify> interface
71(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers 75(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
78 82
79It also is quite fast (see this 83It also is quite fast (see this
80L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent 84L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
81for example). 85for example).
82 86
83=head1 CONVENTIONS 87=head2 CONVENTIONS
84 88
85Libev is very configurable. In this manual the default configuration will 89Libev is very configurable. In this manual the default configuration will
86be described, which supports multiple event loops. For more info about 90be described, which supports multiple event loops. For more info about
87various configuration options please have a look at B<EMBED> section in 91various configuration options please have a look at B<EMBED> section in
88this manual. If libev was configured without support for multiple event 92this manual. If libev was configured without support for multiple event
89loops, then all functions taking an initial argument of name C<loop> 93loops, then all functions taking an initial argument of name C<loop>
90(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.
91 95
92=head1 TIME REPRESENTATION 96=head2 TIME REPRESENTATION
93 97
94Libev represents time as a single floating point number, representing the 98Libev represents time as a single floating point number, representing the
95(fractional) number of seconds since the (POSIX) epoch (somewhere near 99(fractional) number of seconds since the (POSIX) epoch (somewhere near
96the 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
97called 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
98to 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
99it, 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.
100 106
101=head1 GLOBAL FUNCTIONS 107=head1 GLOBAL FUNCTIONS
102 108
103These functions can be called anytime, even before initialising the 109These functions can be called anytime, even before initialising the
104library in any way. 110library in any way.
109 115
110Returns 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
111C<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
112you actually want to know. 118you actually want to know.
113 119
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
114=item int ev_version_major () 126=item int ev_version_major ()
115 127
116=item int ev_version_minor () 128=item int ev_version_minor ()
117 129
118You 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
119you linked against by calling the functions C<ev_version_major> and 131you linked against by calling the functions C<ev_version_major> and
120C<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
121symbols 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
122version of the library your program was compiled against. 134version of the library your program was compiled against.
123 135
136These version numbers refer to the ABI version of the library, not the
137release version.
138
124Usually, 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,
125as this indicates an incompatible change. Minor versions are usually 140as this indicates an incompatible change. Minor versions are usually
126compatible to older versions, so a larger minor version alone is usually 141compatible to older versions, so a larger minor version alone is usually
127not a problem. 142not a problem.
128 143
129Example: Make sure we haven't accidentally been linked against the wrong 144Example: Make sure we haven't accidentally been linked against the wrong
130version. 145version.
245flags. If that is troubling you, check C<ev_backend ()> afterwards). 260flags. If that is troubling you, check C<ev_backend ()> afterwards).
246 261
247If 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
248function. 263function.
249 264
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
250The flags argument can be used to specify special behaviour or specific 272The flags argument can be used to specify special behaviour or specific
251backends 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>).
252 274
253The following flags are supported: 275The following flags are supported:
254 276
274a fork, you can also make libev check for a fork in each iteration by 296a fork, you can also make libev check for a fork in each iteration by
275enabling this flag. 297enabling this flag.
276 298
277This works by calling C<getpid ()> on every iteration of the loop, 299This works by calling C<getpid ()> on every iteration of the loop,
278and thus this might slow down your event loop if you do a lot of loop 300and thus this might slow down your event loop if you do a lot of loop
279iterations and little real work, but is usually not noticable (on my 301iterations and little real work, but is usually not noticeable (on my
280Linux system for example, C<getpid> is actually a simple 5-insn sequence 302Linux system for example, C<getpid> is actually a simple 5-insn sequence
281without a syscall and thus I<very> fast, but my Linux system also has 303without a syscall and thus I<very> fast, but my Linux system also has
282C<pthread_atfork> which is even faster). 304C<pthread_atfork> which is even faster).
283 305
284The big advantage of this flag is that you can forget about fork (and 306The big advantage of this flag is that you can forget about fork (and
291=item C<EVBACKEND_SELECT> (value 1, portable select backend) 313=item C<EVBACKEND_SELECT> (value 1, portable select backend)
292 314
293This is your standard select(2) backend. Not I<completely> standard, as 315This is your standard select(2) backend. Not I<completely> standard, as
294libev 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,
295but 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
296using 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
297the 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.
298 327
299=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)
300 329
301And this is your standard poll(2) backend. It's more complicated than 330And this is your standard poll(2) backend. It's more complicated
302select, but handles sparse fds better and has no artificial limit on the 331than select, but handles sparse fds better and has no artificial
303number 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
304lot 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.
305 336
306=item C<EVBACKEND_EPOLL> (value 4, Linux) 337=item C<EVBACKEND_EPOLL> (value 4, Linux)
307 338
308For 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,
309but it scales phenomenally better. While poll and select usually scale like 340but it scales phenomenally better. While poll and select usually scale
310O(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),
311either 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.
312 346
313While 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
314result in some caching, there is still a syscall per such incident 348will result in some caching, there is still a syscall per such incident
315(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
316best 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
317well if you register events for both fds. 351very well if you register events for both fds.
318 352
319Please note that epoll sometimes generates spurious notifications, so you 353Please note that epoll sometimes generates spurious notifications, so you
320need 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
321(or space) is available. 355(or space) is available.
322 356
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
323=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones) 364=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
324 365
325Kqueue deserves special mention, as at the time of this writing, it 366Kqueue deserves special mention, as at the time of this writing, it
326was 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
327anything but sockets and pipes, except on Darwin, where of course its 368with anything but sockets and pipes, except on Darwin, where of course
328completely useless). For this reason its not being "autodetected" 369it's completely useless). For this reason it's not being "autodetected"
329unless you explicitly specify it explicitly in the flags (i.e. using 370unless you explicitly specify it explicitly in the flags (i.e. using
330C<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.
331 377
332It 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
333kernel is more efficient (which says nothing about its actual speed, of 379kernel is more efficient (which says nothing about its actual speed, of
334course). While starting and stopping an I/O watcher does not cause an 380course). While stopping, setting and starting an I/O watcher does never
335extra 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
336incident, 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.
337 393
338=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8) 394=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
339 395
340This 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.
341 400
342=item C<EVBACKEND_PORT> (value 32, Solaris 10) 401=item C<EVBACKEND_PORT> (value 32, Solaris 10)
343 402
344This uses the Solaris 10 port mechanism. As with everything on Solaris, 403This uses the Solaris 10 event port mechanism. As with everything on Solaris,
345it'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)).
346 405
347Please note that solaris ports can result in a lot of spurious 406Please note that solaris event ports can deliver a lot of spurious
348notifications, 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
349blocking 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.
350 418
351=item C<EVBACKEND_ALL> 419=item C<EVBACKEND_ALL>
352 420
353Try all backends (even potentially broken ones that wouldn't be tried 421Try all backends (even potentially broken ones that wouldn't be tried
354with 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
355C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>. 423C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
356 424
425It is definitely not recommended to use this flag.
426
357=back 427=back
358 428
359If 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
360backends 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
361specified, most compiled-in backend will be tried, usually in reverse 431specified, all backends in C<ev_recommended_backends ()> will be tried.
362order of their flag values :)
363 432
364The most typical usage is like this: 433The most typical usage is like this:
365 434
366 if (!ev_default_loop (0)) 435 if (!ev_default_loop (0))
367 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?"); 436 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
395Destroys the default loop again (frees all memory and kernel state 464Destroys the default loop again (frees all memory and kernel state
396etc.). 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
397sense, 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
398responsibility to either stop all watchers cleanly yoursef I<before> 467responsibility to either stop all watchers cleanly yoursef I<before>
399calling this function, or cope with the fact afterwards (which is usually 468calling this function, or cope with the fact afterwards (which is usually
400the 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
401for 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>).
402 480
403=item ev_loop_destroy (loop) 481=item ev_loop_destroy (loop)
404 482
405Like 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
406earlier call to C<ev_loop_new>. 484earlier call to C<ev_loop_new>.
407 485
408=item ev_default_fork () 486=item ev_default_fork ()
409 487
488This function sets a flag that causes subsequent C<ev_loop> iterations
410This function reinitialises the kernel state for backends that have 489to reinitialise the kernel state for backends that have one. Despite the
411one. Despite the name, you can call it anytime, but it makes most sense 490name, you can call it anytime, but it makes most sense after forking, in
412after forking, in either the parent or child process (or both, but that 491the child process (or both child and parent, but that again makes little
413again makes little sense). 492sense). You I<must> call it in the child before using any of the libev
493functions, and it will only take effect at the next C<ev_loop> iteration.
414 494
415You I<must> call this function in the child process after forking if and 495On the other hand, you only need to call this function in the child
416only if you want to use the event library in both processes. If you just 496process if and only if you want to use the event library in the child. If
417fork+exec, you don't have to call it. 497you just fork+exec, you don't have to call it at all.
418 498
419The function itself is quite fast and it's usually not a problem to call 499The function itself is quite fast and it's usually not a problem to call
420it just in case after a fork. To make this easy, the function will fit in 500it just in case after a fork. To make this easy, the function will fit in
421quite nicely into a call to C<pthread_atfork>: 501quite nicely into a call to C<pthread_atfork>:
422 502
423 pthread_atfork (0, 0, ev_default_fork); 503 pthread_atfork (0, 0, ev_default_fork);
424 504
425At the moment, C<EVBACKEND_SELECT> and C<EVBACKEND_POLL> are safe to use
426without calling this function, so if you force one of those backends you
427do not need to care.
428
429=item ev_loop_fork (loop) 505=item ev_loop_fork (loop)
430 506
431Like C<ev_default_fork>, but acts on an event loop created by 507Like C<ev_default_fork>, but acts on an event loop created by
432C<ev_loop_new>. Yes, you have to call this on every allocated event loop 508C<ev_loop_new>. Yes, you have to call this on every allocated event loop
433after fork, and how you do this is entirely your own problem. 509after fork, and how you do this is entirely your own problem.
510
511=item unsigned int ev_loop_count (loop)
512
513Returns the count of loop iterations for the loop, which is identical to
514the number of times libev did poll for new events. It starts at C<0> and
515happily wraps around with enough iterations.
516
517This value can sometimes be useful as a generation counter of sorts (it
518"ticks" the number of loop iterations), as it roughly corresponds with
519C<ev_prepare> and C<ev_check> calls.
434 520
435=item unsigned int ev_backend (loop) 521=item unsigned int ev_backend (loop)
436 522
437Returns one of the C<EVBACKEND_*> flags indicating the event backend in 523Returns one of the C<EVBACKEND_*> flags indicating the event backend in
438use. 524use.
441 527
442Returns the current "event loop time", which is the time the event loop 528Returns the current "event loop time", which is the time the event loop
443received events and started processing them. This timestamp does not 529received events and started processing them. This timestamp does not
444change as long as callbacks are being processed, and this is also the base 530change as long as callbacks are being processed, and this is also the base
445time used for relative timers. You can treat it as the timestamp of the 531time used for relative timers. You can treat it as the timestamp of the
446event occuring (or more correctly, libev finding out about it). 532event occurring (or more correctly, libev finding out about it).
447 533
448=item ev_loop (loop, int flags) 534=item ev_loop (loop, int flags)
449 535
450Finally, this is it, the event handler. This function usually is called 536Finally, this is it, the event handler. This function usually is called
451after you initialised all your watchers and you want to start handling 537after you initialised all your watchers and you want to start handling
472libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is 558libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is
473usually a better approach for this kind of thing. 559usually a better approach for this kind of thing.
474 560
475Here are the gory details of what C<ev_loop> does: 561Here are the gory details of what C<ev_loop> does:
476 562
477 * If there are no active watchers (reference count is zero), return. 563 - Before the first iteration, call any pending watchers.
478 - Queue prepare watchers and then call all outstanding watchers. 564 * If EVFLAG_FORKCHECK was used, check for a fork.
565 - If a fork was detected, queue and call all fork watchers.
566 - Queue and call all prepare watchers.
479 - If we have been forked, recreate the kernel state. 567 - If we have been forked, recreate the kernel state.
480 - Update the kernel state with all outstanding changes. 568 - Update the kernel state with all outstanding changes.
481 - Update the "event loop time". 569 - Update the "event loop time".
482 - Calculate for how long to block. 570 - Calculate for how long to sleep or block, if at all
571 (active idle watchers, EVLOOP_NONBLOCK or not having
572 any active watchers at all will result in not sleeping).
573 - Sleep if the I/O and timer collect interval say so.
483 - Block the process, waiting for any events. 574 - Block the process, waiting for any events.
484 - Queue all outstanding I/O (fd) events. 575 - Queue all outstanding I/O (fd) events.
485 - Update the "event loop time" and do time jump handling. 576 - Update the "event loop time" and do time jump handling.
486 - Queue all outstanding timers. 577 - Queue all outstanding timers.
487 - Queue all outstanding periodics. 578 - Queue all outstanding periodics.
488 - If no events are pending now, queue all idle watchers. 579 - If no events are pending now, queue all idle watchers.
489 - Queue all check watchers. 580 - Queue all check watchers.
490 - Call all queued watchers in reverse order (i.e. check watchers first). 581 - Call all queued watchers in reverse order (i.e. check watchers first).
491 Signals and child watchers are implemented as I/O watchers, and will 582 Signals and child watchers are implemented as I/O watchers, and will
492 be handled here by queueing them when their watcher gets executed. 583 be handled here by queueing them when their watcher gets executed.
493 - If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK 584 - If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
494 were used, return, otherwise continue with step *. 585 were used, or there are no active watchers, return, otherwise
586 continue with step *.
495 587
496Example: Queue some jobs and then loop until no events are outsanding 588Example: Queue some jobs and then loop until no events are outstanding
497anymore. 589anymore.
498 590
499 ... queue jobs here, make sure they register event watchers as long 591 ... queue jobs here, make sure they register event watchers as long
500 ... as they still have work to do (even an idle watcher will do..) 592 ... as they still have work to do (even an idle watcher will do..)
501 ev_loop (my_loop, 0); 593 ev_loop (my_loop, 0);
505 597
506Can be used to make a call to C<ev_loop> return early (but only after it 598Can be used to make a call to C<ev_loop> return early (but only after it
507has processed all outstanding events). The C<how> argument must be either 599has processed all outstanding events). The C<how> argument must be either
508C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or 600C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
509C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return. 601C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
602
603This "unloop state" will be cleared when entering C<ev_loop> again.
510 604
511=item ev_ref (loop) 605=item ev_ref (loop)
512 606
513=item ev_unref (loop) 607=item ev_unref (loop)
514 608
519returning, ev_unref() after starting, and ev_ref() before stopping it. For 613returning, ev_unref() after starting, and ev_ref() before stopping it. For
520example, libev itself uses this for its internal signal pipe: It is not 614example, libev itself uses this for its internal signal pipe: It is not
521visible to the libev user and should not keep C<ev_loop> from exiting if 615visible to the libev user and should not keep C<ev_loop> from exiting if
522no event watchers registered by it are active. It is also an excellent 616no event watchers registered by it are active. It is also an excellent
523way to do this for generic recurring timers or from within third-party 617way to do this for generic recurring timers or from within third-party
524libraries. Just remember to I<unref after start> and I<ref before stop>. 618libraries. 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,
620respectively).
525 621
526Example: Create a signal watcher, but keep it from keeping C<ev_loop> 622Example: Create a signal watcher, but keep it from keeping C<ev_loop>
527running when nothing else is active. 623running when nothing else is active.
528 624
529 struct ev_signal exitsig; 625 struct ev_signal exitsig;
533 629
534Example: For some weird reason, unregister the above signal handler again. 630Example: For some weird reason, unregister the above signal handler again.
535 631
536 ev_ref (loop); 632 ev_ref (loop);
537 ev_signal_stop (loop, &exitsig); 633 ev_signal_stop (loop, &exitsig);
634
635=item ev_set_io_collect_interval (loop, ev_tstamp interval)
636
637=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
638
639These advanced functions influence the time that libev will spend waiting
640for events. Both are by default C<0>, meaning that libev will try to
641invoke timer/periodic callbacks and I/O callbacks with minimum latency.
642
643Setting these to a higher value (the C<interval> I<must> be >= C<0>)
644allows libev to delay invocation of I/O and timer/periodic callbacks to
645increase efficiency of loop iterations.
646
647The background is that sometimes your program runs just fast enough to
648handle one (or very few) event(s) per loop iteration. While this makes
649the program responsive, it also wastes a lot of CPU time to poll for new
650events, especially with backends like C<select ()> which have a high
651overhead for the actual polling but can deliver many events at once.
652
653By setting a higher I<io collect interval> you allow libev to spend more
654time collecting I/O events, so you can handle more events per iteration,
655at the cost of increasing latency. Timeouts (both C<ev_periodic> and
656C<ev_timer>) will be not affected. Setting this to a non-null value will
657introduce an additional C<ev_sleep ()> call into most loop iterations.
658
659Likewise, by setting a higher I<timeout collect interval> you allow libev
660to spend more time collecting timeouts, at the expense of increased
661latency (the watcher callback will be called later). C<ev_io> watchers
662will not be affected. Setting this to a non-null value will not introduce
663any overhead in libev.
664
665Many (busy) programs can usually benefit by setting the io collect
666interval to a value near C<0.1> or so, which is often enough for
667interactive servers (of course not for games), likewise for timeouts. It
668usually doesn't make much sense to set it to a lower value than C<0.01>,
669as this approsaches the timing granularity of most systems.
538 670
539=back 671=back
540 672
541 673
542=head1 ANATOMY OF A WATCHER 674=head1 ANATOMY OF A WATCHER
722=item bool ev_is_pending (ev_TYPE *watcher) 854=item bool ev_is_pending (ev_TYPE *watcher)
723 855
724Returns a true value iff the watcher is pending, (i.e. it has outstanding 856Returns a true value iff the watcher is pending, (i.e. it has outstanding
725events but its callback has not yet been invoked). As long as a watcher 857events but its callback has not yet been invoked). As long as a watcher
726is pending (but not active) you must not call an init function on it (but 858is pending (but not active) you must not call an init function on it (but
727C<ev_TYPE_set> is safe) and you must make sure the watcher is available to 859C<ev_TYPE_set> is safe), you must not change its priority, and you must
728libev (e.g. you cnanot C<free ()> it). 860make sure the watcher is available to libev (e.g. you cannot C<free ()>
861it).
729 862
730=item callback ev_cb (ev_TYPE *watcher) 863=item callback ev_cb (ev_TYPE *watcher)
731 864
732Returns the callback currently set on the watcher. 865Returns the callback currently set on the watcher.
733 866
734=item ev_cb_set (ev_TYPE *watcher, callback) 867=item ev_cb_set (ev_TYPE *watcher, callback)
735 868
736Change the callback. You can change the callback at virtually any time 869Change the callback. You can change the callback at virtually any time
737(modulo threads). 870(modulo threads).
871
872=item ev_set_priority (ev_TYPE *watcher, priority)
873
874=item int ev_priority (ev_TYPE *watcher)
875
876Set and query the priority of the watcher. The priority is a small
877integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
878(default: C<-2>). Pending watchers with higher priority will be invoked
879before watchers with lower priority, but priority will not keep watchers
880from being executed (except for C<ev_idle> watchers).
881
882This means that priorities are I<only> used for ordering callback
883invocation after new events have been received. This is useful, for
884example, to reduce latency after idling, or more often, to bind two
885watchers on the same event and make sure one is called first.
886
887If you need to suppress invocation when higher priority events are pending
888you need to look at C<ev_idle> watchers, which provide this functionality.
889
890You I<must not> change the priority of a watcher as long as it is active or
891pending.
892
893The default priority used by watchers when no priority has been set is
894always C<0>, which is supposed to not be too high and not be too low :).
895
896Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
897fine, as long as you do not mind that the priority value you query might
898or might not have been adjusted to be within valid range.
899
900=item ev_invoke (loop, ev_TYPE *watcher, int revents)
901
902Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
903C<loop> nor C<revents> need to be valid as long as the watcher callback
904can deal with that fact.
905
906=item int ev_clear_pending (loop, ev_TYPE *watcher)
907
908If the watcher is pending, this function returns clears its pending status
909and returns its C<revents> bitset (as if its callback was invoked). If the
910watcher isn't pending it does nothing and returns C<0>.
738 911
739=back 912=back
740 913
741 914
742=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER 915=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
827In general you can register as many read and/or write event watchers per 1000In general you can register as many read and/or write event watchers per
828fd as you want (as long as you don't confuse yourself). Setting all file 1001fd as you want (as long as you don't confuse yourself). Setting all file
829descriptors to non-blocking mode is also usually a good idea (but not 1002descriptors to non-blocking mode is also usually a good idea (but not
830required if you know what you are doing). 1003required if you know what you are doing).
831 1004
832You have to be careful with dup'ed file descriptors, though. Some backends
833(the linux epoll backend is a notable example) cannot handle dup'ed file
834descriptors correctly if you register interest in two or more fds pointing
835to the same underlying file/socket/etc. description (that is, they share
836the same underlying "file open").
837
838If you must do this, then force the use of a known-to-be-good backend 1005If you must do this, then force the use of a known-to-be-good backend
839(at the time of this writing, this includes only C<EVBACKEND_SELECT> and 1006(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
840C<EVBACKEND_POLL>). 1007C<EVBACKEND_POLL>).
841 1008
842Another thing you have to watch out for is that it is quite easy to 1009Another thing you have to watch out for is that it is quite easy to
848it is best to always use non-blocking I/O: An extra C<read>(2) returning 1015it is best to always use non-blocking I/O: An extra C<read>(2) returning
849C<EAGAIN> is far preferable to a program hanging until some data arrives. 1016C<EAGAIN> is far preferable to a program hanging until some data arrives.
850 1017
851If you cannot run the fd in non-blocking mode (for example you should not 1018If you cannot run the fd in non-blocking mode (for example you should not
852play around with an Xlib connection), then you have to seperately re-test 1019play around with an Xlib connection), then you have to seperately re-test
853wether a file descriptor is really ready with a known-to-be good interface 1020whether a file descriptor is really ready with a known-to-be good interface
854such as poll (fortunately in our Xlib example, Xlib already does this on 1021such as poll (fortunately in our Xlib example, Xlib already does this on
855its own, so its quite safe to use). 1022its own, so its quite safe to use).
1023
1024=head3 The special problem of disappearing file descriptors
1025
1026Some backends (e.g. kqueue, epoll) need to be told about closing a file
1027descriptor (either by calling C<close> explicitly or by any other means,
1028such as C<dup>). The reason is that you register interest in some file
1029descriptor, but when it goes away, the operating system will silently drop
1030this interest. If another file descriptor with the same number then is
1031registered with libev, there is no efficient way to see that this is, in
1032fact, a different file descriptor.
1033
1034To avoid having to explicitly tell libev about such cases, libev follows
1035the following policy: Each time C<ev_io_set> is being called, libev
1036will assume that this is potentially a new file descriptor, otherwise
1037it is assumed that the file descriptor stays the same. That means that
1038you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1039descriptor even if the file descriptor number itself did not change.
1040
1041This is how one would do it normally anyway, the important point is that
1042the libev application should not optimise around libev but should leave
1043optimisations to libev.
1044
1045=head3 The special problem of dup'ed file descriptors
1046
1047Some backends (e.g. epoll), cannot register events for file descriptors,
1048but only events for the underlying file descriptions. That means when you
1049have C<dup ()>'ed file descriptors or weirder constellations, and register
1050events for them, only one file descriptor might actually receive events.
1051
1052There is no workaround possible except not registering events
1053for potentially C<dup ()>'ed file descriptors, or to resort to
1054C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1055
1056=head3 The special problem of fork
1057
1058Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1059useless behaviour. Libev fully supports fork, but needs to be told about
1060it in the child.
1061
1062To support fork in your programs, you either have to call
1063C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
1064enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1065C<EVBACKEND_POLL>.
1066
1067
1068=head3 Watcher-Specific Functions
856 1069
857=over 4 1070=over 4
858 1071
859=item ev_io_init (ev_io *, callback, int fd, int events) 1072=item ev_io_init (ev_io *, callback, int fd, int events)
860 1073
871=item int events [read-only] 1084=item int events [read-only]
872 1085
873The events being watched. 1086The events being watched.
874 1087
875=back 1088=back
1089
1090=head3 Examples
876 1091
877Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well 1092Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
878readable, but only once. Since it is likely line-buffered, you could 1093readable, but only once. Since it is likely line-buffered, you could
879attempt to read a whole line in the callback. 1094attempt to read a whole line in the callback.
880 1095
913 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.); 1128 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
914 1129
915The callback is guarenteed to be invoked only when its timeout has passed, 1130The callback is guarenteed to be invoked only when its timeout has passed,
916but if multiple timers become ready during the same loop iteration then 1131but if multiple timers become ready during the same loop iteration then
917order of execution is undefined. 1132order of execution is undefined.
1133
1134=head3 Watcher-Specific Functions and Data Members
918 1135
919=over 4 1136=over 4
920 1137
921=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat) 1138=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
922 1139
976or C<ev_timer_again> is called and determines the next timeout (if any), 1193or C<ev_timer_again> is called and determines the next timeout (if any),
977which is also when any modifications are taken into account. 1194which is also when any modifications are taken into account.
978 1195
979=back 1196=back
980 1197
1198=head3 Examples
1199
981Example: Create a timer that fires after 60 seconds. 1200Example: Create a timer that fires after 60 seconds.
982 1201
983 static void 1202 static void
984 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents) 1203 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
985 { 1204 {
1018but on wallclock time (absolute time). You can tell a periodic watcher 1237but on wallclock time (absolute time). You can tell a periodic watcher
1019to trigger "at" some specific point in time. For example, if you tell a 1238to trigger "at" some specific point in time. For example, if you tell a
1020periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now () 1239periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()
1021+ 10.>) and then reset your system clock to the last year, then it will 1240+ 10.>) and then reset your system clock to the last year, then it will
1022take a year to trigger the event (unlike an C<ev_timer>, which would trigger 1241take a year to trigger the event (unlike an C<ev_timer>, which would trigger
1023roughly 10 seconds later and of course not if you reset your system time 1242roughly 10 seconds later).
1024again).
1025 1243
1026They can also be used to implement vastly more complex timers, such as 1244They can also be used to implement vastly more complex timers, such as
1027triggering an event on eahc midnight, local time. 1245triggering an event on each midnight, local time or other, complicated,
1246rules.
1028 1247
1029As with timers, the callback is guarenteed to be invoked only when the 1248As with timers, the callback is guarenteed to be invoked only when the
1030time (C<at>) has been passed, but if multiple periodic timers become ready 1249time (C<at>) has been passed, but if multiple periodic timers become ready
1031during the same loop iteration then order of execution is undefined. 1250during the same loop iteration then order of execution is undefined.
1032 1251
1252=head3 Watcher-Specific Functions and Data Members
1253
1033=over 4 1254=over 4
1034 1255
1035=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb) 1256=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
1036 1257
1037=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb) 1258=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
1039Lots of arguments, lets sort it out... There are basically three modes of 1260Lots of arguments, lets sort it out... There are basically three modes of
1040operation, and we will explain them from simplest to complex: 1261operation, and we will explain them from simplest to complex:
1041 1262
1042=over 4 1263=over 4
1043 1264
1044=item * absolute timer (interval = reschedule_cb = 0) 1265=item * absolute timer (at = time, interval = reschedule_cb = 0)
1045 1266
1046In this configuration the watcher triggers an event at the wallclock time 1267In this configuration the watcher triggers an event at the wallclock time
1047C<at> and doesn't repeat. It will not adjust when a time jump occurs, 1268C<at> and doesn't repeat. It will not adjust when a time jump occurs,
1048that is, if it is to be run at January 1st 2011 then it will run when the 1269that is, if it is to be run at January 1st 2011 then it will run when the
1049system time reaches or surpasses this time. 1270system time reaches or surpasses this time.
1050 1271
1051=item * non-repeating interval timer (interval > 0, reschedule_cb = 0) 1272=item * non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1052 1273
1053In this mode the watcher will always be scheduled to time out at the next 1274In this mode the watcher will always be scheduled to time out at the next
1054C<at + N * interval> time (for some integer N) and then repeat, regardless 1275C<at + N * interval> time (for some integer N, which can also be negative)
1055of any time jumps. 1276and then repeat, regardless of any time jumps.
1056 1277
1057This can be used to create timers that do not drift with respect to system 1278This can be used to create timers that do not drift with respect to system
1058time: 1279time:
1059 1280
1060 ev_periodic_set (&periodic, 0., 3600., 0); 1281 ev_periodic_set (&periodic, 0., 3600., 0);
1066 1287
1067Another way to think about it (for the mathematically inclined) is that 1288Another way to think about it (for the mathematically inclined) is that
1068C<ev_periodic> will try to run the callback in this mode at the next possible 1289C<ev_periodic> will try to run the callback in this mode at the next possible
1069time where C<time = at (mod interval)>, regardless of any time jumps. 1290time where C<time = at (mod interval)>, regardless of any time jumps.
1070 1291
1292For numerical stability it is preferable that the C<at> value is near
1293C<ev_now ()> (the current time), but there is no range requirement for
1294this value.
1295
1071=item * manual reschedule mode (reschedule_cb = callback) 1296=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1072 1297
1073In this mode the values for C<interval> and C<at> are both being 1298In this mode the values for C<interval> and C<at> are both being
1074ignored. Instead, each time the periodic watcher gets scheduled, the 1299ignored. Instead, each time the periodic watcher gets scheduled, the
1075reschedule callback will be called with the watcher as first, and the 1300reschedule callback will be called with the watcher as first, and the
1076current time as second argument. 1301current time as second argument.
1077 1302
1078NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, 1303NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
1079ever, or make any event loop modifications>. If you need to stop it, 1304ever, or make any event loop modifications>. If you need to stop it,
1080return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by 1305return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
1081starting a prepare watcher). 1306starting an C<ev_prepare> watcher, which is legal).
1082 1307
1083Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 1308Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
1084ev_tstamp now)>, e.g.: 1309ev_tstamp now)>, e.g.:
1085 1310
1086 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) 1311 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
1109Simply stops and restarts the periodic watcher again. This is only useful 1334Simply stops and restarts the periodic watcher again. This is only useful
1110when you changed some parameters or the reschedule callback would return 1335when you changed some parameters or the reschedule callback would return
1111a different time than the last time it was called (e.g. in a crond like 1336a different time than the last time it was called (e.g. in a crond like
1112program when the crontabs have changed). 1337program when the crontabs have changed).
1113 1338
1339=item ev_tstamp offset [read-write]
1340
1341When repeating, this contains the offset value, otherwise this is the
1342absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1343
1344Can be modified any time, but changes only take effect when the periodic
1345timer fires or C<ev_periodic_again> is being called.
1346
1114=item ev_tstamp interval [read-write] 1347=item ev_tstamp interval [read-write]
1115 1348
1116The current interval value. Can be modified any time, but changes only 1349The current interval value. Can be modified any time, but changes only
1117take effect when the periodic timer fires or C<ev_periodic_again> is being 1350take effect when the periodic timer fires or C<ev_periodic_again> is being
1118called. 1351called.
1121 1354
1122The current reschedule callback, or C<0>, if this functionality is 1355The current reschedule callback, or C<0>, if this functionality is
1123switched off. Can be changed any time, but changes only take effect when 1356switched off. Can be changed any time, but changes only take effect when
1124the periodic timer fires or C<ev_periodic_again> is being called. 1357the periodic timer fires or C<ev_periodic_again> is being called.
1125 1358
1359=item ev_tstamp at [read-only]
1360
1361When active, contains the absolute time that the watcher is supposed to
1362trigger next.
1363
1126=back 1364=back
1365
1366=head3 Examples
1127 1367
1128Example: Call a callback every hour, or, more precisely, whenever the 1368Example: Call a callback every hour, or, more precisely, whenever the
1129system clock is divisible by 3600. The callback invocation times have 1369system clock is divisible by 3600. The callback invocation times have
1130potentially a lot of jittering, but good long-term stability. 1370potentially a lot of jittering, but good long-term stability.
1131 1371
1171with the kernel (thus it coexists with your own signal handlers as long 1411with the kernel (thus it coexists with your own signal handlers as long
1172as you don't register any with libev). Similarly, when the last signal 1412as you don't register any with libev). Similarly, when the last signal
1173watcher for a signal is stopped libev will reset the signal handler to 1413watcher for a signal is stopped libev will reset the signal handler to
1174SIG_DFL (regardless of what it was set to before). 1414SIG_DFL (regardless of what it was set to before).
1175 1415
1416=head3 Watcher-Specific Functions and Data Members
1417
1176=over 4 1418=over 4
1177 1419
1178=item ev_signal_init (ev_signal *, callback, int signum) 1420=item ev_signal_init (ev_signal *, callback, int signum)
1179 1421
1180=item ev_signal_set (ev_signal *, int signum) 1422=item ev_signal_set (ev_signal *, int signum)
1191 1433
1192=head2 C<ev_child> - watch out for process status changes 1434=head2 C<ev_child> - watch out for process status changes
1193 1435
1194Child watchers trigger when your process receives a SIGCHLD in response to 1436Child watchers trigger when your process receives a SIGCHLD in response to
1195some child status changes (most typically when a child of yours dies). 1437some child status changes (most typically when a child of yours dies).
1438
1439=head3 Watcher-Specific Functions and Data Members
1196 1440
1197=over 4 1441=over 4
1198 1442
1199=item ev_child_init (ev_child *, callback, int pid) 1443=item ev_child_init (ev_child *, callback, int pid)
1200 1444
1219 1463
1220The process exit/trace status caused by C<rpid> (see your systems 1464The process exit/trace status caused by C<rpid> (see your systems
1221C<waitpid> and C<sys/wait.h> documentation for details). 1465C<waitpid> and C<sys/wait.h> documentation for details).
1222 1466
1223=back 1467=back
1468
1469=head3 Examples
1224 1470
1225Example: Try to exit cleanly on SIGINT and SIGTERM. 1471Example: Try to exit cleanly on SIGINT and SIGTERM.
1226 1472
1227 static void 1473 static void
1228 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents) 1474 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
1269semantics of C<ev_stat> watchers, which means that libev sometimes needs 1515semantics of C<ev_stat> watchers, which means that libev sometimes needs
1270to fall back to regular polling again even with inotify, but changes are 1516to fall back to regular polling again even with inotify, but changes are
1271usually detected immediately, and if the file exists there will be no 1517usually detected immediately, and if the file exists there will be no
1272polling. 1518polling.
1273 1519
1520=head3 Inotify
1521
1522When C<inotify (7)> support has been compiled into libev (generally only
1523available on Linux) and present at runtime, it will be used to speed up
1524change detection where possible. The inotify descriptor will be created lazily
1525when the first C<ev_stat> watcher is being started.
1526
1527Inotify presense does not change the semantics of C<ev_stat> watchers
1528except that changes might be detected earlier, and in some cases, to avoid
1529making regular C<stat> calls. Even in the presense of inotify support
1530there are many cases where libev has to resort to regular C<stat> polling.
1531
1532(There is no support for kqueue, as apparently it cannot be used to
1533implement this functionality, due to the requirement of having a file
1534descriptor open on the object at all times).
1535
1536=head3 The special problem of stat time resolution
1537
1538The C<stat ()> syscall only supports full-second resolution portably, and
1539even on systems where the resolution is higher, many filesystems still
1540only support whole seconds.
1541
1542That means that, if the time is the only thing that changes, you might
1543miss updates: on the first update, C<ev_stat> detects a change and calls
1544your callback, which does something. When there is another update within
1545the same second, C<ev_stat> will be unable to detect it.
1546
1547The solution to this is to delay acting on a change for a second (or till
1548the next second boundary), using a roughly one-second delay C<ev_timer>
1549(C<ev_timer_set (w, 0., 1.01); ev_timer_again (loop, w)>). The C<.01>
1550is added to work around small timing inconsistencies of some operating
1551systems.
1552
1553=head3 Watcher-Specific Functions and Data Members
1554
1274=over 4 1555=over 4
1275 1556
1276=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval) 1557=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
1277 1558
1278=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval) 1559=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
1313=item const char *path [read-only] 1594=item const char *path [read-only]
1314 1595
1315The filesystem path that is being watched. 1596The filesystem path that is being watched.
1316 1597
1317=back 1598=back
1599
1600=head3 Examples
1318 1601
1319Example: Watch C</etc/passwd> for attribute changes. 1602Example: Watch C</etc/passwd> for attribute changes.
1320 1603
1321 static void 1604 static void
1322 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents) 1605 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
1335 } 1618 }
1336 1619
1337 ... 1620 ...
1338 ev_stat passwd; 1621 ev_stat passwd;
1339 1622
1340 ev_stat_init (&passwd, passwd_cb, "/etc/passwd"); 1623 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1341 ev_stat_start (loop, &passwd); 1624 ev_stat_start (loop, &passwd);
1342 1625
1626Example: Like above, but additionally use a one-second delay so we do not
1627miss updates (however, frequent updates will delay processing, too, so
1628one might do the work both on C<ev_stat> callback invocation I<and> on
1629C<ev_timer> callback invocation).
1630
1631 static ev_stat passwd;
1632 static ev_timer timer;
1633
1634 static void
1635 timer_cb (EV_P_ ev_timer *w, int revents)
1636 {
1637 ev_timer_stop (EV_A_ w);
1638
1639 /* now it's one second after the most recent passwd change */
1640 }
1641
1642 static void
1643 stat_cb (EV_P_ ev_stat *w, int revents)
1644 {
1645 /* reset the one-second timer */
1646 ev_timer_again (EV_A_ &timer);
1647 }
1648
1649 ...
1650 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
1651 ev_stat_start (loop, &passwd);
1652 ev_timer_init (&timer, timer_cb, 0., 1.01);
1653
1343 1654
1344=head2 C<ev_idle> - when you've got nothing better to do... 1655=head2 C<ev_idle> - when you've got nothing better to do...
1345 1656
1346Idle watchers trigger events when there are no other events are pending 1657Idle watchers trigger events when no other events of the same or higher
1347(prepare, check and other idle watchers do not count). That is, as long 1658priority are pending (prepare, check and other idle watchers do not
1348as your process is busy handling sockets or timeouts (or even signals, 1659count).
1349imagine) it will not be triggered. But when your process is idle all idle 1660
1350watchers are being called again and again, once per event loop iteration - 1661That is, as long as your process is busy handling sockets or timeouts
1662(or even signals, imagine) of the same or higher priority it will not be
1663triggered. But when your process is idle (or only lower-priority watchers
1664are pending), the idle watchers are being called once per event loop
1351until stopped, that is, or your process receives more events and becomes 1665iteration - until stopped, that is, or your process receives more events
1352busy. 1666and becomes busy again with higher priority stuff.
1353 1667
1354The most noteworthy effect is that as long as any idle watchers are 1668The most noteworthy effect is that as long as any idle watchers are
1355active, the process will not block when waiting for new events. 1669active, the process will not block when waiting for new events.
1356 1670
1357Apart from keeping your process non-blocking (which is a useful 1671Apart from keeping your process non-blocking (which is a useful
1358effect on its own sometimes), idle watchers are a good place to do 1672effect on its own sometimes), idle watchers are a good place to do
1359"pseudo-background processing", or delay processing stuff to after the 1673"pseudo-background processing", or delay processing stuff to after the
1360event loop has handled all outstanding events. 1674event loop has handled all outstanding events.
1361 1675
1676=head3 Watcher-Specific Functions and Data Members
1677
1362=over 4 1678=over 4
1363 1679
1364=item ev_idle_init (ev_signal *, callback) 1680=item ev_idle_init (ev_signal *, callback)
1365 1681
1366Initialises and configures the idle watcher - it has no parameters of any 1682Initialises and configures the idle watcher - it has no parameters of any
1367kind. There is a C<ev_idle_set> macro, but using it is utterly pointless, 1683kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1368believe me. 1684believe me.
1369 1685
1370=back 1686=back
1687
1688=head3 Examples
1371 1689
1372Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the 1690Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1373callback, free it. Also, use no error checking, as usual. 1691callback, free it. Also, use no error checking, as usual.
1374 1692
1375 static void 1693 static void
1423with priority higher than or equal to the event loop and one coroutine 1741with priority higher than or equal to the event loop and one coroutine
1424of lower priority, but only once, using idle watchers to keep the event 1742of lower priority, but only once, using idle watchers to keep the event
1425loop from blocking if lower-priority coroutines are active, thus mapping 1743loop from blocking if lower-priority coroutines are active, thus mapping
1426low-priority coroutines to idle/background tasks). 1744low-priority coroutines to idle/background tasks).
1427 1745
1746It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1747priority, to ensure that they are being run before any other watchers
1748after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
1749too) should not activate ("feed") events into libev. While libev fully
1750supports this, they will be called before other C<ev_check> watchers
1751did their job. As C<ev_check> watchers are often used to embed other
1752(non-libev) event loops those other event loops might be in an unusable
1753state until their C<ev_check> watcher ran (always remind yourself to
1754coexist peacefully with others).
1755
1756=head3 Watcher-Specific Functions and Data Members
1757
1428=over 4 1758=over 4
1429 1759
1430=item ev_prepare_init (ev_prepare *, callback) 1760=item ev_prepare_init (ev_prepare *, callback)
1431 1761
1432=item ev_check_init (ev_check *, callback) 1762=item ev_check_init (ev_check *, callback)
1435parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set> 1765parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1436macros, but using them is utterly, utterly and completely pointless. 1766macros, but using them is utterly, utterly and completely pointless.
1437 1767
1438=back 1768=back
1439 1769
1440Example: To include a library such as adns, you would add IO watchers 1770=head3 Examples
1441and a timeout watcher in a prepare handler, as required by libadns, and 1771
1772There are a number of principal ways to embed other event loops or modules
1773into libev. Here are some ideas on how to include libadns into libev
1774(there is a Perl module named C<EV::ADNS> that does this, which you could
1775use for an actually working example. Another Perl module named C<EV::Glib>
1776embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV
1777into the Glib event loop).
1778
1779Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1442in a check watcher, destroy them and call into libadns. What follows is 1780and in a check watcher, destroy them and call into libadns. What follows
1443pseudo-code only of course: 1781is pseudo-code only of course. This requires you to either use a low
1782priority for the check watcher or use C<ev_clear_pending> explicitly, as
1783the callbacks for the IO/timeout watchers might not have been called yet.
1444 1784
1445 static ev_io iow [nfd]; 1785 static ev_io iow [nfd];
1446 static ev_timer tw; 1786 static ev_timer tw;
1447 1787
1448 static void 1788 static void
1449 io_cb (ev_loop *loop, ev_io *w, int revents) 1789 io_cb (ev_loop *loop, ev_io *w, int revents)
1450 { 1790 {
1451 // set the relevant poll flags
1452 // could also call adns_processreadable etc. here
1453 struct pollfd *fd = (struct pollfd *)w->data;
1454 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1455 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1456 } 1791 }
1457 1792
1458 // create io watchers for each fd and a timer before blocking 1793 // create io watchers for each fd and a timer before blocking
1459 static void 1794 static void
1460 adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents) 1795 adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
1461 { 1796 {
1462 int timeout = 3600000;truct pollfd fds [nfd]; 1797 int timeout = 3600000;
1798 struct pollfd fds [nfd];
1463 // actual code will need to loop here and realloc etc. 1799 // actual code will need to loop here and realloc etc.
1464 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ())); 1800 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1465 1801
1466 /* the callback is illegal, but won't be called as we stop during check */ 1802 /* the callback is illegal, but won't be called as we stop during check */
1467 ev_timer_init (&tw, 0, timeout * 1e-3); 1803 ev_timer_init (&tw, 0, timeout * 1e-3);
1468 ev_timer_start (loop, &tw); 1804 ev_timer_start (loop, &tw);
1469 1805
1470 // create on ev_io per pollfd 1806 // create one ev_io per pollfd
1471 for (int i = 0; i < nfd; ++i) 1807 for (int i = 0; i < nfd; ++i)
1472 { 1808 {
1473 ev_io_init (iow + i, io_cb, fds [i].fd, 1809 ev_io_init (iow + i, io_cb, fds [i].fd,
1474 ((fds [i].events & POLLIN ? EV_READ : 0) 1810 ((fds [i].events & POLLIN ? EV_READ : 0)
1475 | (fds [i].events & POLLOUT ? EV_WRITE : 0))); 1811 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1476 1812
1477 fds [i].revents = 0; 1813 fds [i].revents = 0;
1478 iow [i].data = fds + i;
1479 ev_io_start (loop, iow + i); 1814 ev_io_start (loop, iow + i);
1480 } 1815 }
1481 } 1816 }
1482 1817
1483 // stop all watchers after blocking 1818 // stop all watchers after blocking
1485 adns_check_cb (ev_loop *loop, ev_check *w, int revents) 1820 adns_check_cb (ev_loop *loop, ev_check *w, int revents)
1486 { 1821 {
1487 ev_timer_stop (loop, &tw); 1822 ev_timer_stop (loop, &tw);
1488 1823
1489 for (int i = 0; i < nfd; ++i) 1824 for (int i = 0; i < nfd; ++i)
1825 {
1826 // set the relevant poll flags
1827 // could also call adns_processreadable etc. here
1828 struct pollfd *fd = fds + i;
1829 int revents = ev_clear_pending (iow + i);
1830 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1831 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1832
1833 // now stop the watcher
1490 ev_io_stop (loop, iow + i); 1834 ev_io_stop (loop, iow + i);
1835 }
1491 1836
1492 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop)); 1837 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1838 }
1839
1840Method 2: This would be just like method 1, but you run C<adns_afterpoll>
1841in the prepare watcher and would dispose of the check watcher.
1842
1843Method 3: If the module to be embedded supports explicit event
1844notification (adns does), you can also make use of the actual watcher
1845callbacks, and only destroy/create the watchers in the prepare watcher.
1846
1847 static void
1848 timer_cb (EV_P_ ev_timer *w, int revents)
1849 {
1850 adns_state ads = (adns_state)w->data;
1851 update_now (EV_A);
1852
1853 adns_processtimeouts (ads, &tv_now);
1854 }
1855
1856 static void
1857 io_cb (EV_P_ ev_io *w, int revents)
1858 {
1859 adns_state ads = (adns_state)w->data;
1860 update_now (EV_A);
1861
1862 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1863 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1864 }
1865
1866 // do not ever call adns_afterpoll
1867
1868Method 4: Do not use a prepare or check watcher because the module you
1869want to embed is too inflexible to support it. Instead, youc na override
1870their poll function. The drawback with this solution is that the main
1871loop is now no longer controllable by EV. The C<Glib::EV> module does
1872this.
1873
1874 static gint
1875 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1876 {
1877 int got_events = 0;
1878
1879 for (n = 0; n < nfds; ++n)
1880 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1881
1882 if (timeout >= 0)
1883 // create/start timer
1884
1885 // poll
1886 ev_loop (EV_A_ 0);
1887
1888 // stop timer again
1889 if (timeout >= 0)
1890 ev_timer_stop (EV_A_ &to);
1891
1892 // stop io watchers again - their callbacks should have set
1893 for (n = 0; n < nfds; ++n)
1894 ev_io_stop (EV_A_ iow [n]);
1895
1896 return got_events;
1493 } 1897 }
1494 1898
1495 1899
1496=head2 C<ev_embed> - when one backend isn't enough... 1900=head2 C<ev_embed> - when one backend isn't enough...
1497 1901
1540portable one. 1944portable one.
1541 1945
1542So when you want to use this feature you will always have to be prepared 1946So when you want to use this feature you will always have to be prepared
1543that you cannot get an embeddable loop. The recommended way to get around 1947that you cannot get an embeddable loop. The recommended way to get around
1544this is to have a separate variables for your embeddable loop, try to 1948this is to have a separate variables for your embeddable loop, try to
1545create it, and if that fails, use the normal loop for everything: 1949create it, and if that fails, use the normal loop for everything.
1950
1951=head3 Watcher-Specific Functions and Data Members
1952
1953=over 4
1954
1955=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
1956
1957=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
1958
1959Configures the watcher to embed the given loop, which must be
1960embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1961invoked automatically, otherwise it is the responsibility of the callback
1962to invoke it (it will continue to be called until the sweep has been done,
1963if you do not want thta, you need to temporarily stop the embed watcher).
1964
1965=item ev_embed_sweep (loop, ev_embed *)
1966
1967Make a single, non-blocking sweep over the embedded loop. This works
1968similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1969apropriate way for embedded loops.
1970
1971=item struct ev_loop *other [read-only]
1972
1973The embedded event loop.
1974
1975=back
1976
1977=head3 Examples
1978
1979Example: Try to get an embeddable event loop and embed it into the default
1980event loop. If that is not possible, use the default loop. The default
1981loop is stored in C<loop_hi>, while the mebeddable loop is stored in
1982C<loop_lo> (which is C<loop_hi> in the acse no embeddable loop can be
1983used).
1546 1984
1547 struct ev_loop *loop_hi = ev_default_init (0); 1985 struct ev_loop *loop_hi = ev_default_init (0);
1548 struct ev_loop *loop_lo = 0; 1986 struct ev_loop *loop_lo = 0;
1549 struct ev_embed embed; 1987 struct ev_embed embed;
1550 1988
1561 ev_embed_start (loop_hi, &embed); 1999 ev_embed_start (loop_hi, &embed);
1562 } 2000 }
1563 else 2001 else
1564 loop_lo = loop_hi; 2002 loop_lo = loop_hi;
1565 2003
1566=over 4 2004Example: Check if kqueue is available but not recommended and create
2005a kqueue backend for use with sockets (which usually work with any
2006kqueue implementation). Store the kqueue/socket-only event loop in
2007C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
1567 2008
1568=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop) 2009 struct ev_loop *loop = ev_default_init (0);
2010 struct ev_loop *loop_socket = 0;
2011 struct ev_embed embed;
2012
2013 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2014 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2015 {
2016 ev_embed_init (&embed, 0, loop_socket);
2017 ev_embed_start (loop, &embed);
2018 }
1569 2019
1570=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop) 2020 if (!loop_socket)
2021 loop_socket = loop;
1571 2022
1572Configures the watcher to embed the given loop, which must be 2023 // now use loop_socket for all sockets, and loop for everything else
1573embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1574invoked automatically, otherwise it is the responsibility of the callback
1575to invoke it (it will continue to be called until the sweep has been done,
1576if you do not want thta, you need to temporarily stop the embed watcher).
1577
1578=item ev_embed_sweep (loop, ev_embed *)
1579
1580Make a single, non-blocking sweep over the embedded loop. This works
1581similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1582apropriate way for embedded loops.
1583
1584=item struct ev_loop *loop [read-only]
1585
1586The embedded event loop.
1587
1588=back
1589 2024
1590 2025
1591=head2 C<ev_fork> - the audacity to resume the event loop after a fork 2026=head2 C<ev_fork> - the audacity to resume the event loop after a fork
1592 2027
1593Fork watchers are called when a C<fork ()> was detected (usually because 2028Fork watchers are called when a C<fork ()> was detected (usually because
1596event loop blocks next and before C<ev_check> watchers are being called, 2031event loop blocks next and before C<ev_check> watchers are being called,
1597and only in the child after the fork. If whoever good citizen calling 2032and only in the child after the fork. If whoever good citizen calling
1598C<ev_default_fork> cheats and calls it in the wrong process, the fork 2033C<ev_default_fork> cheats and calls it in the wrong process, the fork
1599handlers will be invoked, too, of course. 2034handlers will be invoked, too, of course.
1600 2035
2036=head3 Watcher-Specific Functions and Data Members
2037
1601=over 4 2038=over 4
1602 2039
1603=item ev_fork_init (ev_signal *, callback) 2040=item ev_fork_init (ev_signal *, callback)
1604 2041
1605Initialises and configures the fork watcher - it has no parameters of any 2042Initialises and configures the fork watcher - it has no parameters of any
1701 2138
1702To use it, 2139To use it,
1703 2140
1704 #include <ev++.h> 2141 #include <ev++.h>
1705 2142
1706(it is not installed by default). This automatically includes F<ev.h> 2143This automatically includes F<ev.h> and puts all of its definitions (many
1707and puts all of its definitions (many of them macros) into the global 2144of them macros) into the global namespace. All C++ specific things are
1708namespace. All C++ specific things are put into the C<ev> namespace. 2145put into the C<ev> namespace. It should support all the same embedding
2146options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
1709 2147
1710It should support all the same embedding options as F<ev.h>, most notably 2148Care has been taken to keep the overhead low. The only data member the C++
1711C<EV_MULTIPLICITY>. 2149classes add (compared to plain C-style watchers) is the event loop pointer
2150that the watcher is associated with (or no additional members at all if
2151you disable C<EV_MULTIPLICITY> when embedding libev).
2152
2153Currently, functions, and static and non-static member functions can be
2154used as callbacks. Other types should be easy to add as long as they only
2155need one additional pointer for context. If you need support for other
2156types of functors please contact the author (preferably after implementing
2157it).
1712 2158
1713Here is a list of things available in the C<ev> namespace: 2159Here is a list of things available in the C<ev> namespace:
1714 2160
1715=over 4 2161=over 4
1716 2162
1732 2178
1733All of those classes have these methods: 2179All of those classes have these methods:
1734 2180
1735=over 4 2181=over 4
1736 2182
1737=item ev::TYPE::TYPE (object *, object::method *) 2183=item ev::TYPE::TYPE ()
1738 2184
1739=item ev::TYPE::TYPE (object *, object::method *, struct ev_loop *) 2185=item ev::TYPE::TYPE (struct ev_loop *)
1740 2186
1741=item ev::TYPE::~TYPE 2187=item ev::TYPE::~TYPE
1742 2188
1743The constructor takes a pointer to an object and a method pointer to 2189The constructor (optionally) takes an event loop to associate the watcher
1744the event handler callback to call in this class. The constructor calls 2190with. If it is omitted, it will use C<EV_DEFAULT>.
1745C<ev_init> for you, which means you have to call the C<set> method 2191
1746before starting it. If you do not specify a loop then the constructor 2192The constructor calls C<ev_init> for you, which means you have to call the
1747automatically associates the default loop with this watcher. 2193C<set> method before starting it.
2194
2195It will not set a callback, however: You have to call the templated C<set>
2196method to set a callback before you can start the watcher.
2197
2198(The reason why you have to use a method is a limitation in C++ which does
2199not allow explicit template arguments for constructors).
1748 2200
1749The destructor automatically stops the watcher if it is active. 2201The destructor automatically stops the watcher if it is active.
2202
2203=item w->set<class, &class::method> (object *)
2204
2205This method sets the callback method to call. The method has to have a
2206signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
2207first argument and the C<revents> as second. The object must be given as
2208parameter and is stored in the C<data> member of the watcher.
2209
2210This method synthesizes efficient thunking code to call your method from
2211the C callback that libev requires. If your compiler can inline your
2212callback (i.e. it is visible to it at the place of the C<set> call and
2213your compiler is good :), then the method will be fully inlined into the
2214thunking function, making it as fast as a direct C callback.
2215
2216Example: simple class declaration and watcher initialisation
2217
2218 struct myclass
2219 {
2220 void io_cb (ev::io &w, int revents) { }
2221 }
2222
2223 myclass obj;
2224 ev::io iow;
2225 iow.set <myclass, &myclass::io_cb> (&obj);
2226
2227=item w->set<function> (void *data = 0)
2228
2229Also sets a callback, but uses a static method or plain function as
2230callback. The optional C<data> argument will be stored in the watcher's
2231C<data> member and is free for you to use.
2232
2233The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2234
2235See the method-C<set> above for more details.
2236
2237Example:
2238
2239 static void io_cb (ev::io &w, int revents) { }
2240 iow.set <io_cb> ();
1750 2241
1751=item w->set (struct ev_loop *) 2242=item w->set (struct ev_loop *)
1752 2243
1753Associates a different C<struct ev_loop> with this watcher. You can only 2244Associates a different C<struct ev_loop> with this watcher. You can only
1754do this when the watcher is inactive (and not pending either). 2245do this when the watcher is inactive (and not pending either).
1755 2246
1756=item w->set ([args]) 2247=item w->set ([args])
1757 2248
1758Basically the same as C<ev_TYPE_set>, with the same args. Must be 2249Basically the same as C<ev_TYPE_set>, with the same args. Must be
1759called at least once. Unlike the C counterpart, an active watcher gets 2250called at least once. Unlike the C counterpart, an active watcher gets
1760automatically stopped and restarted. 2251automatically stopped and restarted when reconfiguring it with this
2252method.
1761 2253
1762=item w->start () 2254=item w->start ()
1763 2255
1764Starts the watcher. Note that there is no C<loop> argument as the 2256Starts the watcher. Note that there is no C<loop> argument, as the
1765constructor already takes the loop. 2257constructor already stores the event loop.
1766 2258
1767=item w->stop () 2259=item w->stop ()
1768 2260
1769Stops the watcher if it is active. Again, no C<loop> argument. 2261Stops the watcher if it is active. Again, no C<loop> argument.
1770 2262
1771=item w->again () C<ev::timer>, C<ev::periodic> only 2263=item w->again () (C<ev::timer>, C<ev::periodic> only)
1772 2264
1773For C<ev::timer> and C<ev::periodic>, this invokes the corresponding 2265For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
1774C<ev_TYPE_again> function. 2266C<ev_TYPE_again> function.
1775 2267
1776=item w->sweep () C<ev::embed> only 2268=item w->sweep () (C<ev::embed> only)
1777 2269
1778Invokes C<ev_embed_sweep>. 2270Invokes C<ev_embed_sweep>.
1779 2271
1780=item w->update () C<ev::stat> only 2272=item w->update () (C<ev::stat> only)
1781 2273
1782Invokes C<ev_stat_stat>. 2274Invokes C<ev_stat_stat>.
1783 2275
1784=back 2276=back
1785 2277
1795 2287
1796 myclass (); 2288 myclass ();
1797 } 2289 }
1798 2290
1799 myclass::myclass (int fd) 2291 myclass::myclass (int fd)
1800 : io (this, &myclass::io_cb),
1801 idle (this, &myclass::idle_cb)
1802 { 2292 {
2293 io .set <myclass, &myclass::io_cb > (this);
2294 idle.set <myclass, &myclass::idle_cb> (this);
2295
1803 io.start (fd, ev::READ); 2296 io.start (fd, ev::READ);
1804 } 2297 }
1805 2298
1806 2299
1807=head1 MACRO MAGIC 2300=head1 MACRO MAGIC
1808 2301
1809Libev can be compiled with a variety of options, the most fundemantal is 2302Libev can be compiled with a variety of options, the most fundamantal
1810C<EV_MULTIPLICITY>. This option determines wether (most) functions and 2303of which is C<EV_MULTIPLICITY>. This option determines whether (most)
1811callbacks have an initial C<struct ev_loop *> argument. 2304functions and callbacks have an initial C<struct ev_loop *> argument.
1812 2305
1813To make it easier to write programs that cope with either variant, the 2306To make it easier to write programs that cope with either variant, the
1814following macros are defined: 2307following macros are defined:
1815 2308
1816=over 4 2309=over 4
1849loop, if multiple loops are supported ("ev loop default"). 2342loop, if multiple loops are supported ("ev loop default").
1850 2343
1851=back 2344=back
1852 2345
1853Example: Declare and initialise a check watcher, utilising the above 2346Example: Declare and initialise a check watcher, utilising the above
1854macros so it will work regardless of wether multiple loops are supported 2347macros so it will work regardless of whether multiple loops are supported
1855or not. 2348or not.
1856 2349
1857 static void 2350 static void
1858 check_cb (EV_P_ ev_timer *w, int revents) 2351 check_cb (EV_P_ ev_timer *w, int revents)
1859 { 2352 {
1870Libev can (and often is) directly embedded into host 2363Libev can (and often is) directly embedded into host
1871applications. Examples of applications that embed it include the Deliantra 2364applications. Examples of applications that embed it include the Deliantra
1872Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe) 2365Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
1873and rxvt-unicode. 2366and rxvt-unicode.
1874 2367
1875The goal is to enable you to just copy the neecssary files into your 2368The goal is to enable you to just copy the necessary files into your
1876source directory without having to change even a single line in them, so 2369source directory without having to change even a single line in them, so
1877you can easily upgrade by simply copying (or having a checked-out copy of 2370you can easily upgrade by simply copying (or having a checked-out copy of
1878libev somewhere in your source tree). 2371libev somewhere in your source tree).
1879 2372
1880=head2 FILESETS 2373=head2 FILESETS
1970 2463
1971If defined to be C<1>, libev will try to detect the availability of the 2464If defined to be C<1>, libev will try to detect the availability of the
1972monotonic clock option at both compiletime and runtime. Otherwise no use 2465monotonic clock option at both compiletime and runtime. Otherwise no use
1973of the monotonic clock option will be attempted. If you enable this, you 2466of the monotonic clock option will be attempted. If you enable this, you
1974usually have to link against librt or something similar. Enabling it when 2467usually have to link against librt or something similar. Enabling it when
1975the functionality isn't available is safe, though, althoguh you have 2468the functionality isn't available is safe, though, although you have
1976to make sure you link against any libraries where the C<clock_gettime> 2469to make sure you link against any libraries where the C<clock_gettime>
1977function is hiding in (often F<-lrt>). 2470function is hiding in (often F<-lrt>).
1978 2471
1979=item EV_USE_REALTIME 2472=item EV_USE_REALTIME
1980 2473
1981If defined to be C<1>, libev will try to detect the availability of the 2474If defined to be C<1>, libev will try to detect the availability of the
1982realtime clock option at compiletime (and assume its availability at 2475realtime clock option at compiletime (and assume its availability at
1983runtime if successful). Otherwise no use of the realtime clock option will 2476runtime if successful). Otherwise no use of the realtime clock option will
1984be attempted. This effectively replaces C<gettimeofday> by C<clock_get 2477be attempted. This effectively replaces C<gettimeofday> by C<clock_get
1985(CLOCK_REALTIME, ...)> and will not normally affect correctness. See tzhe note about libraries 2478(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
1986in the description of C<EV_USE_MONOTONIC>, though. 2479note about libraries in the description of C<EV_USE_MONOTONIC>, though.
2480
2481=item EV_USE_NANOSLEEP
2482
2483If defined to be C<1>, libev will assume that C<nanosleep ()> is available
2484and will use it for delays. Otherwise it will use C<select ()>.
1987 2485
1988=item EV_USE_SELECT 2486=item EV_USE_SELECT
1989 2487
1990If undefined or defined to be C<1>, libev will compile in support for the 2488If undefined or defined to be C<1>, libev will compile in support for the
1991C<select>(2) backend. No attempt at autodetection will be done: if no 2489C<select>(2) backend. No attempt at autodetection will be done: if no
2009wants osf handles on win32 (this is the case when the select to 2507wants osf handles on win32 (this is the case when the select to
2010be used is the winsock select). This means that it will call 2508be used is the winsock select). This means that it will call
2011C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise, 2509C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2012it is assumed that all these functions actually work on fds, even 2510it is assumed that all these functions actually work on fds, even
2013on win32. Should not be defined on non-win32 platforms. 2511on win32. Should not be defined on non-win32 platforms.
2512
2513=item EV_FD_TO_WIN32_HANDLE
2514
2515If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
2516file descriptors to socket handles. When not defining this symbol (the
2517default), then libev will call C<_get_osfhandle>, which is usually
2518correct. In some cases, programs use their own file descriptor management,
2519in which case they can provide this function to map fds to socket handles.
2014 2520
2015=item EV_USE_POLL 2521=item EV_USE_POLL
2016 2522
2017If defined to be C<1>, libev will compile in support for the C<poll>(2) 2523If defined to be C<1>, libev will compile in support for the C<poll>(2)
2018backend. Otherwise it will be enabled on non-win32 platforms. It 2524backend. Otherwise it will be enabled on non-win32 platforms. It
2055be detected at runtime. 2561be detected at runtime.
2056 2562
2057=item EV_H 2563=item EV_H
2058 2564
2059The name of the F<ev.h> header file used to include it. The default if 2565The name of the F<ev.h> header file used to include it. The default if
2060undefined is C<< <ev.h> >> in F<event.h> and C<"ev.h"> in F<ev.c>. This 2566undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2061can be used to virtually rename the F<ev.h> header file in case of conflicts. 2567used to virtually rename the F<ev.h> header file in case of conflicts.
2062 2568
2063=item EV_CONFIG_H 2569=item EV_CONFIG_H
2064 2570
2065If C<EV_STANDALONE> isn't C<1>, this variable can be used to override 2571If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2066F<ev.c>'s idea of where to find the F<config.h> file, similarly to 2572F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2067C<EV_H>, above. 2573C<EV_H>, above.
2068 2574
2069=item EV_EVENT_H 2575=item EV_EVENT_H
2070 2576
2071Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea 2577Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2072of how the F<event.h> header can be found. 2578of how the F<event.h> header can be found, the default is C<"event.h">.
2073 2579
2074=item EV_PROTOTYPES 2580=item EV_PROTOTYPES
2075 2581
2076If defined to be C<0>, then F<ev.h> will not define any function 2582If defined to be C<0>, then F<ev.h> will not define any function
2077prototypes, but still define all the structs and other symbols. This is 2583prototypes, but still define all the structs and other symbols. This is
2084will have the C<struct ev_loop *> as first argument, and you can create 2590will have the C<struct ev_loop *> as first argument, and you can create
2085additional independent event loops. Otherwise there will be no support 2591additional independent event loops. Otherwise there will be no support
2086for multiple event loops and there is no first event loop pointer 2592for multiple event loops and there is no first event loop pointer
2087argument. Instead, all functions act on the single default loop. 2593argument. Instead, all functions act on the single default loop.
2088 2594
2595=item EV_MINPRI
2596
2597=item EV_MAXPRI
2598
2599The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
2600C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
2601provide for more priorities by overriding those symbols (usually defined
2602to be C<-2> and C<2>, respectively).
2603
2604When doing priority-based operations, libev usually has to linearly search
2605all the priorities, so having many of them (hundreds) uses a lot of space
2606and time, so using the defaults of five priorities (-2 .. +2) is usually
2607fine.
2608
2609If your embedding app does not need any priorities, defining these both to
2610C<0> will save some memory and cpu.
2611
2089=item EV_PERIODIC_ENABLE 2612=item EV_PERIODIC_ENABLE
2090 2613
2091If undefined or defined to be C<1>, then periodic timers are supported. If 2614If undefined or defined to be C<1>, then periodic timers are supported. If
2615defined to be C<0>, then they are not. Disabling them saves a few kB of
2616code.
2617
2618=item EV_IDLE_ENABLE
2619
2620If undefined or defined to be C<1>, then idle watchers are supported. If
2092defined to be C<0>, then they are not. Disabling them saves a few kB of 2621defined to be C<0>, then they are not. Disabling them saves a few kB of
2093code. 2622code.
2094 2623
2095=item EV_EMBED_ENABLE 2624=item EV_EMBED_ENABLE
2096 2625
2120than enough. If you need to manage thousands of children you might want to 2649than enough. If you need to manage thousands of children you might want to
2121increase this value (I<must> be a power of two). 2650increase this value (I<must> be a power of two).
2122 2651
2123=item EV_INOTIFY_HASHSIZE 2652=item EV_INOTIFY_HASHSIZE
2124 2653
2125C<ev_staz> watchers use a small hash table to distribute workload by 2654C<ev_stat> watchers use a small hash table to distribute workload by
2126inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>), 2655inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2127usually more than enough. If you need to manage thousands of C<ev_stat> 2656usually more than enough. If you need to manage thousands of C<ev_stat>
2128watchers you might want to increase this value (I<must> be a power of 2657watchers you might want to increase this value (I<must> be a power of
2129two). 2658two).
2130 2659
2147 2676
2148=item ev_set_cb (ev, cb) 2677=item ev_set_cb (ev, cb)
2149 2678
2150Can be used to change the callback member declaration in each watcher, 2679Can be used to change the callback member declaration in each watcher,
2151and the way callbacks are invoked and set. Must expand to a struct member 2680and the way callbacks are invoked and set. Must expand to a struct member
2152definition and a statement, respectively. See the F<ev.v> header file for 2681definition and a statement, respectively. See the F<ev.h> header file for
2153their default definitions. One possible use for overriding these is to 2682their default definitions. One possible use for overriding these is to
2154avoid the C<struct ev_loop *> as first argument in all cases, or to use 2683avoid the C<struct ev_loop *> as first argument in all cases, or to use
2155method calls instead of plain function calls in C++. 2684method calls instead of plain function calls in C++.
2685
2686=head2 EXPORTED API SYMBOLS
2687
2688If you need to re-export the API (e.g. via a dll) and you need a list of
2689exported symbols, you can use the provided F<Symbol.*> files which list
2690all public symbols, one per line:
2691
2692 Symbols.ev for libev proper
2693 Symbols.event for the libevent emulation
2694
2695This can also be used to rename all public symbols to avoid clashes with
2696multiple versions of libev linked together (which is obviously bad in
2697itself, but sometimes it is inconvinient to avoid this).
2698
2699A sed command like this will create wrapper C<#define>'s that you need to
2700include before including F<ev.h>:
2701
2702 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
2703
2704This would create a file F<wrap.h> which essentially looks like this:
2705
2706 #define ev_backend myprefix_ev_backend
2707 #define ev_check_start myprefix_ev_check_start
2708 #define ev_check_stop myprefix_ev_check_stop
2709 ...
2156 2710
2157=head2 EXAMPLES 2711=head2 EXAMPLES
2158 2712
2159For a real-world example of a program the includes libev 2713For a real-world example of a program the includes libev
2160verbatim, you can have a look at the EV perl module 2714verbatim, you can have a look at the EV perl module
2189 2743
2190In this section the complexities of (many of) the algorithms used inside 2744In this section the complexities of (many of) the algorithms used inside
2191libev will be explained. For complexity discussions about backends see the 2745libev will be explained. For complexity discussions about backends see the
2192documentation for C<ev_default_init>. 2746documentation for C<ev_default_init>.
2193 2747
2748All of the following are about amortised time: If an array needs to be
2749extended, libev needs to realloc and move the whole array, but this
2750happens asymptotically never with higher number of elements, so O(1) might
2751mean it might do a lengthy realloc operation in rare cases, but on average
2752it is much faster and asymptotically approaches constant time.
2753
2194=over 4 2754=over 4
2195 2755
2196=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers) 2756=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
2197 2757
2758This means that, when you have a watcher that triggers in one hour and
2759there are 100 watchers that would trigger before that then inserting will
2760have to skip roughly seven (C<ld 100>) of these watchers.
2761
2198=item Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers) 2762=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2763
2764That means that changing a timer costs less than removing/adding them
2765as only the relative motion in the event queue has to be paid for.
2199 2766
2200=item Starting io/check/prepare/idle/signal/child watchers: O(1) 2767=item Starting io/check/prepare/idle/signal/child watchers: O(1)
2201 2768
2769These just add the watcher into an array or at the head of a list.
2770
2202=item Stopping check/prepare/idle watchers: O(1) 2771=item Stopping check/prepare/idle watchers: O(1)
2203 2772
2204=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE)) 2773=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2205 2774
2775These watchers are stored in lists then need to be walked to find the
2776correct watcher to remove. The lists are usually short (you don't usually
2777have many watchers waiting for the same fd or signal).
2778
2206=item Finding the next timer per loop iteration: O(1) 2779=item Finding the next timer in each loop iteration: O(1)
2780
2781By virtue of using a binary heap, the next timer is always found at the
2782beginning of the storage array.
2207 2783
2208=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd) 2784=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
2209 2785
2210=item Activating one watcher: O(1) 2786A change means an I/O watcher gets started or stopped, which requires
2787libev to recalculate its status (and possibly tell the kernel, depending
2788on backend and wether C<ev_io_set> was used).
2789
2790=item Activating one watcher (putting it into the pending state): O(1)
2791
2792=item Priority handling: O(number_of_priorities)
2793
2794Priorities are implemented by allocating some space for each
2795priority. When doing priority-based operations, libev usually has to
2796linearly search all the priorities, but starting/stopping and activating
2797watchers becomes O(1) w.r.t. prioritiy handling.
2211 2798
2212=back 2799=back
2213 2800
2214 2801
2802=head1 Win32 platform limitations and workarounds
2803
2804Win32 doesn't support any of the standards (e.g. POSIX) that libev
2805requires, and its I/O model is fundamentally incompatible with the POSIX
2806model. Libev still offers limited functionality on this platform in
2807the form of the C<EVBACKEND_SELECT> backend, and only supports socket
2808descriptors. This only applies when using Win32 natively, not when using
2809e.g. cygwin.
2810
2811There is no supported compilation method available on windows except
2812embedding it into other applications.
2813
2814Due to the many, low, and arbitrary limits on the win32 platform and the
2815abysmal performance of winsockets, using a large number of sockets is not
2816recommended (and not reasonable). If your program needs to use more than
2817a hundred or so sockets, then likely it needs to use a totally different
2818implementation for windows, as libev offers the POSIX model, which cannot
2819be implemented efficiently on windows (microsoft monopoly games).
2820
2821=over 4
2822
2823=item The winsocket select function
2824
2825The winsocket C<select> function doesn't follow POSIX in that it requires
2826socket I<handles> and not socket I<file descriptors>. This makes select
2827very inefficient, and also requires a mapping from file descriptors
2828to socket handles. See the discussion of the C<EV_SELECT_USE_FD_SET>,
2829C<EV_SELECT_IS_WINSOCKET> and C<EV_FD_TO_WIN32_HANDLE> preprocessor
2830symbols for more info.
2831
2832The configuration for a "naked" win32 using the microsoft runtime
2833libraries and raw winsocket select is:
2834
2835 #define EV_USE_SELECT 1
2836 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
2837
2838Note that winsockets handling of fd sets is O(n), so you can easily get a
2839complexity in the O(n²) range when using win32.
2840
2841=item Limited number of file descriptors
2842
2843Windows has numerous arbitrary (and low) limits on things. Early versions
2844of winsocket's select only supported waiting for a max. of C<64> handles
2845(probably owning to the fact that all windows kernels can only wait for
2846C<64> things at the same time internally; microsoft recommends spawning a
2847chain of threads and wait for 63 handles and the previous thread in each).
2848
2849Newer versions support more handles, but you need to define C<FD_SETSIZE>
2850to some high number (e.g. C<2048>) before compiling the winsocket select
2851call (which might be in libev or elsewhere, for example, perl does its own
2852select emulation on windows).
2853
2854Another limit is the number of file descriptors in the microsoft runtime
2855libraries, which by default is C<64> (there must be a hidden I<64> fetish
2856or something like this inside microsoft). You can increase this by calling
2857C<_setmaxstdio>, which can increase this limit to C<2048> (another
2858arbitrary limit), but is broken in many versions of the microsoft runtime
2859libraries.
2860
2861This might get you to about C<512> or C<2048> sockets (depending on
2862windows version and/or the phase of the moon). To get more, you need to
2863wrap all I/O functions and provide your own fd management, but the cost of
2864calling select (O(n²)) will likely make this unworkable.
2865
2866=back
2867
2868
2215=head1 AUTHOR 2869=head1 AUTHOR
2216 2870
2217Marc Lehmann <libev@schmorp.de>. 2871Marc Lehmann <libev@schmorp.de>.
2218 2872

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines