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

Comparing libev/ev.pod (file contents):
Revision 1.3 by root, Mon Nov 12 08:03:31 2007 UTC vs.
Revision 1.24 by root, Mon Nov 12 19:20:05 2007 UTC

… …
8 8
9=head1 DESCRIPTION 9=head1 DESCRIPTION
10 10
11Libev is an event loop: you register interest in certain events (such as a 11Libev is an event loop: you register interest in certain events (such as a
12file descriptor being readable or a timeout occuring), and it will manage 12file descriptor being readable or a timeout occuring), and it will manage
13these event sources and provide your program events. 13these event sources and provide your program with events.
14 14
15To do this, it must take more or less complete control over your process 15To do this, it must take more or less complete control over your process
16(or thread) by executing the I<event loop> handler, and will then 16(or thread) by executing the I<event loop> handler, and will then
17communicate events via a callback mechanism. 17communicate events via a callback mechanism.
18 18
… …
25 25
26Libev supports select, poll, the linux-specific epoll and the bsd-specific 26Libev supports select, poll, the linux-specific epoll and the bsd-specific
27kqueue mechanisms for file descriptor events, relative timers, absolute 27kqueue mechanisms for file descriptor events, relative timers, absolute
28timers with customised rescheduling, signal events, process status change 28timers with customised rescheduling, signal events, process status change
29events (related to SIGCHLD), and event watchers dealing with the event 29events (related to SIGCHLD), and event watchers dealing with the event
30loop mechanism itself (idle, prepare and check watchers). 30loop mechanism itself (idle, prepare and check watchers). It also is quite
31fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing
32it to libevent for example).
31 33
32=head1 CONVENTIONS 34=head1 CONVENTIONS
33 35
34Libev is very configurable. In this manual the default configuration 36Libev is very configurable. In this manual the default configuration
35will be described, which supports multiple event loops. For more info 37will be described, which supports multiple event loops. For more info
36about various configuraiton options please have a look at the file 38about various configuration options please have a look at the file
37F<README.embed> in the libev distribution. If libev was configured without 39F<README.embed> in the libev distribution. If libev was configured without
38support for multiple event loops, then all functions taking an initial 40support for multiple event loops, then all functions taking an initial
39argument of name C<loop> (which is always of type C<struct ev_loop *>) 41argument of name C<loop> (which is always of type C<struct ev_loop *>)
40will not have this argument. 42will not have this argument.
41 43
42=head1 TIME AND OTHER GLOBAL FUNCTIONS 44=head1 TIME REPRESENTATION
43 45
44Libev represents time as a single floating point number, representing the 46Libev represents time as a single floating point number, representing the
45(fractional) number of seconds since the (POSIX) epoch (somewhere near 47(fractional) number of seconds since the (POSIX) epoch (somewhere near
46the beginning of 1970, details are complicated, don't ask). This type is 48the beginning of 1970, details are complicated, don't ask). This type is
47called C<ev_tstamp>, which is what you should use too. It usually aliases 49called C<ev_tstamp>, which is what you should use too. It usually aliases
48to the double type in C. 50to the double type in C.
49 51
52=head1 GLOBAL FUNCTIONS
53
54These functions can be called anytime, even before initialising the
55library in any way.
56
50=over 4 57=over 4
51 58
52=item ev_tstamp ev_time () 59=item ev_tstamp ev_time ()
53 60
54Returns the current time as libev would use it. 61Returns the current time as libev would use it.
… …
61you linked against by calling the functions C<ev_version_major> and 68you linked against by calling the functions C<ev_version_major> and
62C<ev_version_minor>. If you want, you can compare against the global 69C<ev_version_minor>. If you want, you can compare against the global
63symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the 70symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
64version of the library your program was compiled against. 71version of the library your program was compiled against.
65 72
66Usually, its a good idea to terminate if the major versions mismatch, 73Usually, it's a good idea to terminate if the major versions mismatch,
67as this indicates an incompatible change. Minor versions are usually 74as this indicates an incompatible change. Minor versions are usually
68compatible to older versions, so a larger minor version alone is usually 75compatible to older versions, so a larger minor version alone is usually
69not a problem. 76not a problem.
70 77
71=item ev_set_allocator (void *(*cb)(void *ptr, long size)) 78=item ev_set_allocator (void *(*cb)(void *ptr, long size))
72 79
73Sets the allocation function to use (the prototype is similar to the 80Sets the allocation function to use (the prototype is similar to the
74realloc function). It is used to allocate and free memory (no surprises 81realloc C function, the semantics are identical). It is used to allocate
75here). If it returns zero when memory needs to be allocated, the library 82and free memory (no surprises here). If it returns zero when memory
76might abort or take some potentially destructive action. The default is 83needs to be allocated, the library might abort or take some potentially
77your system realloc function. 84destructive action. The default is your system realloc function.
78 85
79You could override this function in high-availability programs to, say, 86You could override this function in high-availability programs to, say,
80free some memory if it cannot allocate memory, to use a special allocator, 87free some memory if it cannot allocate memory, to use a special allocator,
81or even to sleep a while and retry until some memory is available. 88or even to sleep a while and retry until some memory is available.
82 89
… …
84 91
85Set the callback function to call on a retryable syscall error (such 92Set the callback function to call on a retryable syscall error (such
86as failed select, poll, epoll_wait). The message is a printable string 93as failed select, poll, epoll_wait). The message is a printable string
87indicating the system call or subsystem causing the problem. If this 94indicating the system call or subsystem causing the problem. If this
88callback is set, then libev will expect it to remedy the sitution, no 95callback is set, then libev will expect it to remedy the sitution, no
89matter what, when it returns. That is, libev will geenrally retry the 96matter what, when it returns. That is, libev will generally retry the
90requested operation, or, if the condition doesn't go away, do bad stuff 97requested operation, or, if the condition doesn't go away, do bad stuff
91(such as abort). 98(such as abort).
92 99
93=back 100=back
94 101
… …
97An event loop is described by a C<struct ev_loop *>. The library knows two 104An event loop is described by a C<struct ev_loop *>. The library knows two
98types of such loops, the I<default> loop, which supports signals and child 105types of such loops, the I<default> loop, which supports signals and child
99events, and dynamically created loops which do not. 106events, and dynamically created loops which do not.
100 107
101If you use threads, a common model is to run the default event loop 108If you use threads, a common model is to run the default event loop
102in your main thread (or in a separate thrad) and for each thread you 109in your main thread (or in a separate thread) and for each thread you
103create, you also create another event loop. Libev itself does no lockign 110create, you also create another event loop. Libev itself does no locking
104whatsoever, so if you mix calls to different event loops, make sure you 111whatsoever, so if you mix calls to the same event loop in different
105lock (this is usually a bad idea, though, even if done right). 112threads, make sure you lock (this is usually a bad idea, though, even if
113done correctly, because it's hideous and inefficient).
106 114
107=over 4 115=over 4
108 116
109=item struct ev_loop *ev_default_loop (unsigned int flags) 117=item struct ev_loop *ev_default_loop (unsigned int flags)
110 118
… …
115 123
116If you don't know what event loop to use, use the one returned from this 124If you don't know what event loop to use, use the one returned from this
117function. 125function.
118 126
119The flags argument can be used to specify special behaviour or specific 127The flags argument can be used to specify special behaviour or specific
120backends to use, and is usually specified as 0 (or EVFLAG_AUTO) 128backends to use, and is usually specified as 0 (or EVFLAG_AUTO).
121 129
122It supports the following flags: 130It supports the following flags:
123 131
124=over 4 132=over 4
125 133
126=item EVFLAG_AUTO 134=item C<EVFLAG_AUTO>
127 135
128The default flags value. Use this if you have no clue (its the right 136The default flags value. Use this if you have no clue (it's the right
129thing, believe me). 137thing, believe me).
130 138
131=item EVFLAG_NOENV 139=item C<EVFLAG_NOENV>
132 140
133If this flag bit is ored into the flag value then libev will I<not> look 141If this flag bit is ored into the flag value (or the program runs setuid
134at the environment variable C<LIBEV_FLAGS>. Otherwise (the default), this 142or setgid) then libev will I<not> look at the environment variable
135environment variable will override the flags completely. This is useful 143C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
144override the flags completely if it is found in the environment. This is
136to try out specific backends to tets their performance, or to work around 145useful to try out specific backends to test their performance, or to work
137bugs. 146around bugs.
138 147
139=item EVMETHOD_SELECT portable select backend 148=item C<EVMETHOD_SELECT> (portable select backend)
140 149
141=item EVMETHOD_POLL poll backend (everywhere except windows) 150=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)
142 151
143=item EVMETHOD_EPOLL linux only 152=item C<EVMETHOD_EPOLL> (linux only)
144 153
145=item EVMETHOD_KQUEUE some bsds only 154=item C<EVMETHOD_KQUEUE> (some bsds only)
146 155
147=item EVMETHOD_DEVPOLL solaris 8 only 156=item C<EVMETHOD_DEVPOLL> (solaris 8 only)
148 157
149=item EVMETHOD_PORT solaris 10 only 158=item C<EVMETHOD_PORT> (solaris 10 only)
150 159
151If one or more of these are ored into the flags value, then only these 160If one or more of these are ored into the flags value, then only these
152backends will be tried (in the reverse order as given here). If one are 161backends will be tried (in the reverse order as given here). If one are
153specified, any backend will do. 162specified, any backend will do.
154 163
… …
163 172
164=item ev_default_destroy () 173=item ev_default_destroy ()
165 174
166Destroys the default loop again (frees all memory and kernel state 175Destroys the default loop again (frees all memory and kernel state
167etc.). This stops all registered event watchers (by not touching them in 176etc.). This stops all registered event watchers (by not touching them in
168any way whatsoever, although you cnanot rely on this :). 177any way whatsoever, although you cannot rely on this :).
169 178
170=item ev_loop_destroy (loop) 179=item ev_loop_destroy (loop)
171 180
172Like C<ev_default_destroy>, but destroys an event loop created by an 181Like C<ev_default_destroy>, but destroys an event loop created by an
173earlier call to C<ev_loop_new>. 182earlier call to C<ev_loop_new>.
… …
181 190
182You I<must> call this function after forking if and only if you want to 191You I<must> call this function after forking if and only if you want to
183use the event library in both processes. If you just fork+exec, you don't 192use the event library in both processes. If you just fork+exec, you don't
184have to call it. 193have to call it.
185 194
186The function itself is quite fast and its usually not a problem to call 195The function itself is quite fast and it's usually not a problem to call
187it just in case after a fork. To make this easy, the function will fit in 196it just in case after a fork. To make this easy, the function will fit in
188quite nicely into a call to C<pthread_atfork>: 197quite nicely into a call to C<pthread_atfork>:
189 198
190 pthread_atfork (0, 0, ev_default_fork); 199 pthread_atfork (0, 0, ev_default_fork);
191 200
… …
198=item unsigned int ev_method (loop) 207=item unsigned int ev_method (loop)
199 208
200Returns one of the C<EVMETHOD_*> flags indicating the event backend in 209Returns one of the C<EVMETHOD_*> flags indicating the event backend in
201use. 210use.
202 211
203=item ev_tstamp = ev_now (loop) 212=item ev_tstamp ev_now (loop)
204 213
205Returns the current "event loop time", which is the time the event loop 214Returns the current "event loop time", which is the time the event loop
206got events and started processing them. This timestamp does not change 215got events and started processing them. This timestamp does not change
207as long as callbacks are being processed, and this is also the base time 216as long as callbacks are being processed, and this is also the base time
208used for relative timers. You can treat it as the timestamp of the event 217used for relative timers. You can treat it as the timestamp of the event
… …
217If the flags argument is specified as 0, it will not return until either 226If the flags argument is specified as 0, it will not return until either
218no event watchers are active anymore or C<ev_unloop> was called. 227no event watchers are active anymore or C<ev_unloop> was called.
219 228
220A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle 229A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
221those events and any outstanding ones, but will not block your process in 230those events and any outstanding ones, but will not block your process in
222case there are no events. 231case there are no events and will return after one iteration of the loop.
223 232
224A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if 233A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
225neccessary) and will handle those and any outstanding ones. It will block 234neccessary) and will handle those and any outstanding ones. It will block
226your process until at least one new event arrives. 235your process until at least one new event arrives, and will return after
236one iteration of the loop.
227 237
228This flags value could be used to implement alternative looping 238This flags value could be used to implement alternative looping
229constructs, but the C<prepare> and C<check> watchers provide a better and 239constructs, but the C<prepare> and C<check> watchers provide a better and
230more generic mechanism. 240more generic mechanism.
231 241
232=item ev_unloop (loop, how) 242=item ev_unloop (loop, how)
233 243
234Can be used to make a call to C<ev_loop> return early. The C<how> argument 244Can be used to make a call to C<ev_loop> return early (but only after it
245has processed all outstanding events). The C<how> argument must be either
235must be either C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> 246C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> call return, or
236call return, or C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> 247C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
237calls return.
238 248
239=item ev_ref (loop) 249=item ev_ref (loop)
240 250
241=item ev_unref (loop) 251=item ev_unref (loop)
242 252
243Ref/unref can be used to add or remove a refcount on the event loop: Every 253Ref/unref can be used to add or remove a reference count on the event
244watcher keeps one reference. If you have a long-runing watcher you never 254loop: Every watcher keeps one reference, and as long as the reference
245unregister that should not keep ev_loop from running, ev_unref() after 255count is nonzero, C<ev_loop> will not return on its own. If you have
246starting, and ev_ref() before stopping it. Libev itself uses this for 256a watcher you never unregister that should not keep C<ev_loop> from
247example for its internal signal pipe: It is not visible to you as a user 257returning, ev_unref() after starting, and ev_ref() before stopping it. For
248and should not keep C<ev_loop> from exiting if the work is done. It is 258example, libev itself uses this for its internal signal pipe: It is not
249also an excellent way to do this for generic recurring timers or from 259visible to the libev user and should not keep C<ev_loop> from exiting if
250within third-party libraries. Just remember to unref after start and ref 260no event watchers registered by it are active. It is also an excellent
251before stop. 261way to do this for generic recurring timers or from within third-party
262libraries. Just remember to I<unref after start> and I<ref before stop>.
252 263
253=back 264=back
254 265
255=head1 ANATOMY OF A WATCHER 266=head1 ANATOMY OF A WATCHER
256 267
257A watcher is a structure that you create and register to record your 268A watcher is a structure that you create and register to record your
258interest in some event. For instance, if you want to wait for STDIN to 269interest in some event. For instance, if you want to wait for STDIN to
259become readable, you would create an ev_io watcher for that: 270become readable, you would create an C<ev_io> watcher for that:
260 271
261 static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents) 272 static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)
262 { 273 {
263 ev_io_stop (w); 274 ev_io_stop (w);
264 ev_unloop (loop, EVUNLOOP_ALL); 275 ev_unloop (loop, EVUNLOOP_ALL);
… …
293 304
294As long as your watcher is active (has been started but not stopped) you 305As long as your watcher is active (has been started but not stopped) you
295must not touch the values stored in it. Most specifically you must never 306must not touch the values stored in it. Most specifically you must never
296reinitialise it or call its set method. 307reinitialise it or call its set method.
297 308
298You cna check wether an event is active by calling the C<ev_is_active 309You can check whether an event is active by calling the C<ev_is_active
299(watcher *)> macro. To see wether an event is outstanding (but the 310(watcher *)> macro. To see whether an event is outstanding (but the
300callback for it has not been called yet) you cna use the C<ev_is_pending 311callback for it has not been called yet) you can use the C<ev_is_pending
301(watcher *)> macro. 312(watcher *)> macro.
302 313
303Each and every callback receives the event loop pointer as first, the 314Each and every callback receives the event loop pointer as first, the
304registered watcher structure as second, and a bitset of received events as 315registered watcher structure as second, and a bitset of received events as
305third argument. 316third argument.
306 317
307The rceeived events usually include a single bit per event type received 318The received events usually include a single bit per event type received
308(you can receive multiple events at the same time). The possible bit masks 319(you can receive multiple events at the same time). The possible bit masks
309are: 320are:
310 321
311=over 4 322=over 4
312 323
313=item EV_READ 324=item C<EV_READ>
314 325
315=item EV_WRITE 326=item C<EV_WRITE>
316 327
317The file descriptor in the ev_io watcher has become readable and/or 328The file descriptor in the C<ev_io> watcher has become readable and/or
318writable. 329writable.
319 330
320=item EV_TIMEOUT 331=item C<EV_TIMEOUT>
321 332
322The ev_timer watcher has timed out. 333The C<ev_timer> watcher has timed out.
323 334
324=item EV_PERIODIC 335=item C<EV_PERIODIC>
325 336
326The ev_periodic watcher has timed out. 337The C<ev_periodic> watcher has timed out.
327 338
328=item EV_SIGNAL 339=item C<EV_SIGNAL>
329 340
330The signal specified in the ev_signal watcher has been received by a thread. 341The signal specified in the C<ev_signal> watcher has been received by a thread.
331 342
332=item EV_CHILD 343=item C<EV_CHILD>
333 344
334The pid specified in the ev_child watcher has received a status change. 345The pid specified in the C<ev_child> watcher has received a status change.
335 346
336=item EV_IDLE 347=item C<EV_IDLE>
337 348
338The ev_idle watcher has determined that you have nothing better to do. 349The C<ev_idle> watcher has determined that you have nothing better to do.
339 350
340=item EV_PREPARE 351=item C<EV_PREPARE>
341 352
342=item EV_CHECK 353=item C<EV_CHECK>
343 354
344All ev_prepare watchers are invoked just I<before> C<ev_loop> starts 355All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts
345to gather new events, and all ev_check watchers are invoked just after 356to gather new events, and all C<ev_check> watchers are invoked just after
346C<ev_loop> has gathered them, but before it invokes any callbacks for any 357C<ev_loop> has gathered them, but before it invokes any callbacks for any
347received events. Callbacks of both watcher types can start and stop as 358received events. Callbacks of both watcher types can start and stop as
348many watchers as they want, and all of them will be taken into account 359many watchers as they want, and all of them will be taken into account
349(for example, a ev_prepare watcher might start an idle watcher to keep 360(for example, a C<ev_prepare> watcher might start an idle watcher to keep
350C<ev_loop> from blocking). 361C<ev_loop> from blocking).
351 362
352=item EV_ERROR 363=item C<EV_ERROR>
353 364
354An unspecified error has occured, the watcher has been stopped. This might 365An unspecified error has occured, the watcher has been stopped. This might
355happen because the watcher could not be properly started because libev 366happen because the watcher could not be properly started because libev
356ran out of memory, a file descriptor was found to be closed or any other 367ran out of memory, a file descriptor was found to be closed or any other
357problem. You best act on it by reporting the problem and somehow coping 368problem. You best act on it by reporting the problem and somehow coping
… …
366=back 377=back
367 378
368=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER 379=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
369 380
370Each watcher has, by default, a member C<void *data> that you can change 381Each watcher has, by default, a member C<void *data> that you can change
371and read at any time, libev will completely ignore it. This cna be used 382and read at any time, libev will completely ignore it. This can be used
372to associate arbitrary data with your watcher. If you need more data and 383to associate arbitrary data with your watcher. If you need more data and
373don't want to allocate memory and store a pointer to it in that data 384don't want to allocate memory and store a pointer to it in that data
374member, you can also "subclass" the watcher type and provide your own 385member, you can also "subclass" the watcher type and provide your own
375data: 386data:
376 387
… …
398=head1 WATCHER TYPES 409=head1 WATCHER TYPES
399 410
400This section describes each watcher in detail, but will not repeat 411This section describes each watcher in detail, but will not repeat
401information given in the last section. 412information given in the last section.
402 413
403=head2 struct ev_io - is my file descriptor readable or writable 414=head2 C<ev_io> - is this file descriptor readable or writable
404 415
405I/O watchers check wether a file descriptor is readable or writable 416I/O watchers check whether a file descriptor is readable or writable
406in each iteration of the event loop (This behaviour is called 417in each iteration of the event loop (This behaviour is called
407level-triggering because you keep receiving events as long as the 418level-triggering because you keep receiving events as long as the
408condition persists. Remember you cna stop the watcher if you don't want to 419condition persists. Remember you can stop the watcher if you don't want to
409act on the event and neither want to receive future events). 420act on the event and neither want to receive future events).
410 421
422In general you can register as many read and/or write event watchers per
423fd as you want (as long as you don't confuse yourself). Setting all file
424descriptors to non-blocking mode is also usually a good idea (but not
425required if you know what you are doing).
426
427You have to be careful with dup'ed file descriptors, though. Some backends
428(the linux epoll backend is a notable example) cannot handle dup'ed file
429descriptors correctly if you register interest in two or more fds pointing
430to the same underlying file/socket etc. description (that is, they share
431the same underlying "file open").
432
433If you must do this, then force the use of a known-to-be-good backend
434(at the time of this writing, this includes only EVMETHOD_SELECT and
435EVMETHOD_POLL).
436
411=over 4 437=over 4
412 438
413=item ev_io_init (ev_io *, callback, int fd, int events) 439=item ev_io_init (ev_io *, callback, int fd, int events)
414 440
415=item ev_io_set (ev_io *, int fd, int events) 441=item ev_io_set (ev_io *, int fd, int events)
416 442
417Configures an ev_io watcher. The fd is the file descriptor to rceeive 443Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive
418events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ | 444events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |
419EV_WRITE> to receive the given events. 445EV_WRITE> to receive the given events.
420 446
421=back 447=back
422 448
423=head2 struct ev_timer - relative and optionally recurring timeouts 449=head2 C<ev_timer> - relative and optionally recurring timeouts
424 450
425Timer watchers are simple relative timers that generate an event after a 451Timer watchers are simple relative timers that generate an event after a
426given time, and optionally repeating in regular intervals after that. 452given time, and optionally repeating in regular intervals after that.
427 453
428The timers are based on real time, that is, if you register an event that 454The timers are based on real time, that is, if you register an event that
429times out after an hour and youreset your system clock to last years 455times out after an hour and you reset your system clock to last years
430time, it will still time out after (roughly) and hour. "Roughly" because 456time, it will still time out after (roughly) and hour. "Roughly" because
431detecting time jumps is hard, and soem inaccuracies are unavoidable (the 457detecting time jumps is hard, and soem inaccuracies are unavoidable (the
432monotonic clock option helps a lot here). 458monotonic clock option helps a lot here).
459
460The relative timeouts are calculated relative to the C<ev_now ()>
461time. This is usually the right thing as this timestamp refers to the time
462of the event triggering whatever timeout you are modifying/starting. If
463you suspect event processing to be delayed and you *need* to base the timeout
464on the current time, use something like this to adjust for this:
465
466 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
433 467
434=over 4 468=over 4
435 469
436=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat) 470=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
437 471
… …
443later, again, and again, until stopped manually. 477later, again, and again, until stopped manually.
444 478
445The timer itself will do a best-effort at avoiding drift, that is, if you 479The timer itself will do a best-effort at avoiding drift, that is, if you
446configure a timer to trigger every 10 seconds, then it will trigger at 480configure a timer to trigger every 10 seconds, then it will trigger at
447exactly 10 second intervals. If, however, your program cannot keep up with 481exactly 10 second intervals. If, however, your program cannot keep up with
448the timer (ecause it takes longer than those 10 seconds to do stuff) the 482the timer (because it takes longer than those 10 seconds to do stuff) the
449timer will not fire more than once per event loop iteration. 483timer will not fire more than once per event loop iteration.
450 484
451=item ev_timer_again (loop) 485=item ev_timer_again (loop)
452 486
453This will act as if the timer timed out and restart it again if it is 487This will act as if the timer timed out and restart it again if it is
… …
460 494
461This sounds a bit complicated, but here is a useful and typical 495This sounds a bit complicated, but here is a useful and typical
462example: Imagine you have a tcp connection and you want a so-called idle 496example: Imagine you have a tcp connection and you want a so-called idle
463timeout, that is, you want to be called when there have been, say, 60 497timeout, that is, you want to be called when there have been, say, 60
464seconds of inactivity on the socket. The easiest way to do this is to 498seconds of inactivity on the socket. The easiest way to do this is to
465configure an ev_timer with after=repeat=60 and calling ev_timer_again each 499configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each
466time you successfully read or write some data. If you go into an idle 500time you successfully read or write some data. If you go into an idle
467state where you do not expect data to travel on the socket, you can stop 501state where you do not expect data to travel on the socket, you can stop
468the timer, and again will automatically restart it if need be. 502the timer, and again will automatically restart it if need be.
469 503
470=back 504=back
471 505
472=head2 ev_periodic - to cron or not to cron it 506=head2 C<ev_periodic> - to cron or not to cron
473 507
474Periodic watchers are also timers of a kind, but they are very versatile 508Periodic watchers are also timers of a kind, but they are very versatile
475(and unfortunately a bit complex). 509(and unfortunately a bit complex).
476 510
477Unlike ev_timer's, they are not based on real time (or relative time) 511Unlike C<ev_timer>'s, they are not based on real time (or relative time)
478but on wallclock time (absolute time). You can tell a periodic watcher 512but on wallclock time (absolute time). You can tell a periodic watcher
479to trigger "at" some specific point in time. For example, if you tell a 513to trigger "at" some specific point in time. For example, if you tell a
480periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now () 514periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()
481+ 10.>) and then reset your system clock to the last year, then it will 515+ 10.>) and then reset your system clock to the last year, then it will
482take a year to trigger the event (unlike an ev_timer, which would trigger 516take a year to trigger the event (unlike an C<ev_timer>, which would trigger
483roughly 10 seconds later and of course not if you reset your system time 517roughly 10 seconds later and of course not if you reset your system time
484again). 518again).
485 519
486They can also be used to implement vastly more complex timers, such as 520They can also be used to implement vastly more complex timers, such as
487triggering an event on eahc midnight, local time. 521triggering an event on eahc midnight, local time.
… …
516 550
517 ev_periodic_set (&periodic, 0., 3600., 0); 551 ev_periodic_set (&periodic, 0., 3600., 0);
518 552
519This doesn't mean there will always be 3600 seconds in between triggers, 553This doesn't mean there will always be 3600 seconds in between triggers,
520but only that the the callback will be called when the system time shows a 554but only that the the callback will be called when the system time shows a
521full hour (UTC), or more correct, when the system time is evenly divisible 555full hour (UTC), or more correctly, when the system time is evenly divisible
522by 3600. 556by 3600.
523 557
524Another way to think about it (for the mathematically inclined) is that 558Another way to think about it (for the mathematically inclined) is that
525ev_periodic will try to run the callback in this mode at the next possible 559C<ev_periodic> will try to run the callback in this mode at the next possible
526time where C<time = at (mod interval)>, regardless of any time jumps. 560time where C<time = at (mod interval)>, regardless of any time jumps.
527 561
528=item * manual reschedule mode (reschedule_cb = callback) 562=item * manual reschedule mode (reschedule_cb = callback)
529 563
530In this mode the values for C<interval> and C<at> are both being 564In this mode the values for C<interval> and C<at> are both being
531ignored. Instead, each time the periodic watcher gets scheduled, the 565ignored. Instead, each time the periodic watcher gets scheduled, the
532reschedule callback will be called with the watcher as first, and the 566reschedule callback will be called with the watcher as first, and the
533current time as second argument. 567current time as second argument.
534 568
535NOTE: I<This callback MUST NOT stop or destroy the periodic or any other 569NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
536periodic watcher, ever, or make any event loop modificstions>. If you need 570ever, or make any event loop modifications>. If you need to stop it,
537to stop it, return 1e30 (or so, fudge fudge) and stop it afterwards. 571return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
572starting a prepare watcher).
538 573
539Its prototype is c<ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 574Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
540ev_tstamp now)>, e.g.: 575ev_tstamp now)>, e.g.:
541 576
542 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) 577 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
543 { 578 {
544 return now + 60.; 579 return now + 60.;
… …
547It must return the next time to trigger, based on the passed time value 582It must return the next time to trigger, based on the passed time value
548(that is, the lowest time value larger than to the second argument). It 583(that is, the lowest time value larger than to the second argument). It
549will usually be called just before the callback will be triggered, but 584will usually be called just before the callback will be triggered, but
550might be called at other times, too. 585might be called at other times, too.
551 586
587NOTE: I<< This callback must always return a time that is later than the
588passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.
589
552This can be used to create very complex timers, such as a timer that 590This can be used to create very complex timers, such as a timer that
553triggers on each midnight, local time. To do this, you would calculate the 591triggers on each midnight, local time. To do this, you would calculate the
554next midnight after C<now> and return the timestamp value for this. How you do this 592next midnight after C<now> and return the timestamp value for this. How
555is, again, up to you (but it is not trivial). 593you do this is, again, up to you (but it is not trivial, which is the main
594reason I omitted it as an example).
556 595
557=back 596=back
558 597
559=item ev_periodic_again (loop, ev_periodic *) 598=item ev_periodic_again (loop, ev_periodic *)
560 599
… …
563a different time than the last time it was called (e.g. in a crond like 602a different time than the last time it was called (e.g. in a crond like
564program when the crontabs have changed). 603program when the crontabs have changed).
565 604
566=back 605=back
567 606
568=head2 ev_signal - signal me when a signal gets signalled 607=head2 C<ev_signal> - signal me when a signal gets signalled
569 608
570Signal watchers will trigger an event when the process receives a specific 609Signal watchers will trigger an event when the process receives a specific
571signal one or more times. Even though signals are very asynchronous, libev 610signal one or more times. Even though signals are very asynchronous, libev
572will try its best to deliver signals synchronously, i.e. as part of the 611will try it's best to deliver signals synchronously, i.e. as part of the
573normal event processing, like any other event. 612normal event processing, like any other event.
574 613
575You cna configure as many watchers as you like per signal. Only when the 614You can configure as many watchers as you like per signal. Only when the
576first watcher gets started will libev actually register a signal watcher 615first watcher gets started will libev actually register a signal watcher
577with the kernel (thus it coexists with your own signal handlers as long 616with the kernel (thus it coexists with your own signal handlers as long
578as you don't register any with libev). Similarly, when the last signal 617as you don't register any with libev). Similarly, when the last signal
579watcher for a signal is stopped libev will reset the signal handler to 618watcher for a signal is stopped libev will reset the signal handler to
580SIG_DFL (regardless of what it was set to before). 619SIG_DFL (regardless of what it was set to before).
… …
588Configures the watcher to trigger on the given signal number (usually one 627Configures the watcher to trigger on the given signal number (usually one
589of the C<SIGxxx> constants). 628of the C<SIGxxx> constants).
590 629
591=back 630=back
592 631
593=head2 ev_child - wait for pid status changes 632=head2 C<ev_child> - wait for pid status changes
594 633
595Child watchers trigger when your process receives a SIGCHLD in response to 634Child watchers trigger when your process receives a SIGCHLD in response to
596some child status changes (most typically when a child of yours dies). 635some child status changes (most typically when a child of yours dies).
597 636
598=over 4 637=over 4
… …
602=item ev_child_set (ev_child *, int pid) 641=item ev_child_set (ev_child *, int pid)
603 642
604Configures the watcher to wait for status changes of process C<pid> (or 643Configures the watcher to wait for status changes of process C<pid> (or
605I<any> process if C<pid> is specified as C<0>). The callback can look 644I<any> process if C<pid> is specified as C<0>). The callback can look
606at the C<rstatus> member of the C<ev_child> watcher structure to see 645at the C<rstatus> member of the C<ev_child> watcher structure to see
607the status word (use the macros from C<sys/wait.h>). The C<rpid> member 646the status word (use the macros from C<sys/wait.h> and see your systems
608contains the pid of the process causing the status change. 647C<waitpid> documentation). The C<rpid> member contains the pid of the
648process causing the status change.
609 649
610=back 650=back
611 651
612=head2 ev_idle - when you've got nothing better to do 652=head2 C<ev_idle> - when you've got nothing better to do
613 653
614Idle watchers trigger events when there are no other I/O or timer (or 654Idle watchers trigger events when there are no other events are pending
615periodic) events pending. That is, as long as your process is busy 655(prepare, check and other idle watchers do not count). That is, as long
616handling sockets or timeouts it will not be called. But when your process 656as your process is busy handling sockets or timeouts (or even signals,
617is idle all idle watchers are being called again and again - until 657imagine) it will not be triggered. But when your process is idle all idle
658watchers are being called again and again, once per event loop iteration -
618stopped, that is, or your process receives more events. 659until stopped, that is, or your process receives more events and becomes
660busy.
619 661
620The most noteworthy effect is that as long as any idle watchers are 662The most noteworthy effect is that as long as any idle watchers are
621active, the process will not block when waiting for new events. 663active, the process will not block when waiting for new events.
622 664
623Apart from keeping your process non-blocking (which is a useful 665Apart from keeping your process non-blocking (which is a useful
… …
633kind. There is a C<ev_idle_set> macro, but using it is utterly pointless, 675kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
634believe me. 676believe me.
635 677
636=back 678=back
637 679
638=head2 prepare and check - your hooks into the event loop 680=head2 C<ev_prepare> and C<ev_check> - customise your event loop
639 681
640Prepare and check watchers usually (but not always) are used in 682Prepare and check watchers are usually (but not always) used in tandem:
641tandom. Prepare watchers get invoked before the process blocks and check 683prepare watchers get invoked before the process blocks and check watchers
642watchers afterwards. 684afterwards.
643 685
644Their main purpose is to integrate other event mechanisms into libev. This 686Their main purpose is to integrate other event mechanisms into libev. This
645could be used, for example, to track variable changes, implement your own 687could be used, for example, to track variable changes, implement your own
646watchers, integrate net-snmp or a coroutine library and lots more. 688watchers, integrate net-snmp or a coroutine library and lots more.
647 689
648This is done by examining in each prepare call which file descriptors need 690This is done by examining in each prepare call which file descriptors need
649to be watched by the other library, registering ev_io watchers for them 691to be watched by the other library, registering C<ev_io> watchers for
650and starting an ev_timer watcher for any timeouts (many libraries provide 692them and starting an C<ev_timer> watcher for any timeouts (many libraries
651just this functionality). Then, in the check watcher you check for any 693provide just this functionality). Then, in the check watcher you check for
652events that occured (by making your callbacks set soem flags for example) 694any events that occured (by checking the pending status of all watchers
653and call back into the library. 695and stopping them) and call back into the library. The I/O and timer
696callbacks will never actually be called (but must be valid nevertheless,
697because you never know, you know?).
654 698
655As another example, the perl Coro module uses these hooks to integrate 699As another example, the Perl Coro module uses these hooks to integrate
656coroutines into libev programs, by yielding to other active coroutines 700coroutines into libev programs, by yielding to other active coroutines
657during each prepare and only letting the process block if no coroutines 701during each prepare and only letting the process block if no coroutines
658are ready to run. 702are ready to run (it's actually more complicated: it only runs coroutines
703with priority higher than or equal to the event loop and one coroutine
704of lower priority, but only once, using idle watchers to keep the event
705loop from blocking if lower-priority coroutines are active, thus mapping
706low-priority coroutines to idle/background tasks).
659 707
660=over 4 708=over 4
661 709
662=item ev_prepare_init (ev_prepare *, callback) 710=item ev_prepare_init (ev_prepare *, callback)
663 711
664=item ev_check_init (ev_check *, callback) 712=item ev_check_init (ev_check *, callback)
665 713
666Initialises and configures the prepare or check watcher - they have no 714Initialises and configures the prepare or check watcher - they have no
667parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set> 715parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
668macros, but using them is utterly, utterly pointless. 716macros, but using them is utterly, utterly and completely pointless.
669 717
670=back 718=back
671 719
672=head1 OTHER FUNCTIONS 720=head1 OTHER FUNCTIONS
673 721
674There are some other fucntions of possible interest. Described. Here. Now. 722There are some other functions of possible interest. Described. Here. Now.
675 723
676=over 4 724=over 4
677 725
678=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback) 726=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
679 727
680This function combines a simple timer and an I/O watcher, calls your 728This function combines a simple timer and an I/O watcher, calls your
681callback on whichever event happens first and automatically stop both 729callback on whichever event happens first and automatically stop both
682watchers. This is useful if you want to wait for a single event on an fd 730watchers. This is useful if you want to wait for a single event on an fd
683or timeout without havign to allocate/configure/start/stop/free one or 731or timeout without having to allocate/configure/start/stop/free one or
684more watchers yourself. 732more watchers yourself.
685 733
686If C<fd> is less than 0, then no I/O watcher will be started and events is 734If C<fd> is less than 0, then no I/O watcher will be started and events
687ignored. Otherwise, an ev_io watcher for the given C<fd> and C<events> set 735is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and
688will be craeted and started. 736C<events> set will be craeted and started.
689 737
690If C<timeout> is less than 0, then no timeout watcher will be 738If C<timeout> is less than 0, then no timeout watcher will be
691started. Otherwise an ev_timer watcher with after = C<timeout> (and repeat 739started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
692= 0) will be started. 740repeat = 0) will be started. While C<0> is a valid timeout, it is of
741dubious value.
693 742
694The callback has the type C<void (*cb)(int revents, void *arg)> and 743The callback has the type C<void (*cb)(int revents, void *arg)> and gets
695gets passed an events set (normally a combination of EV_ERROR, EV_READ, 744passed an C<revents> set like normal event callbacks (a combination of
696EV_WRITE or EV_TIMEOUT) and the C<arg> value passed to C<ev_once>: 745C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
746value passed to C<ev_once>:
697 747
698 static void stdin_ready (int revents, void *arg) 748 static void stdin_ready (int revents, void *arg)
699 { 749 {
700 if (revents & EV_TIMEOUT) 750 if (revents & EV_TIMEOUT)
701 /* doh, nothing entered */ 751 /* doh, nothing entered */;
702 else if (revents & EV_READ) 752 else if (revents & EV_READ)
703 /* stdin might have data for us, joy! */ 753 /* stdin might have data for us, joy! */;
704 } 754 }
705 755
706 ev_once (STDIN_FILENO, EV_READm 10., stdin_ready, 0); 756 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
707 757
708=item ev_feed_event (loop, watcher, int events) 758=item ev_feed_event (loop, watcher, int events)
709 759
710Feeds the given event set into the event loop, as if the specified event 760Feeds the given event set into the event loop, as if the specified event
711has happened for the specified watcher (which must be a pointer to an 761had happened for the specified watcher (which must be a pointer to an
712initialised but not necessarily active event watcher). 762initialised but not necessarily started event watcher).
713 763
714=item ev_feed_fd_event (loop, int fd, int revents) 764=item ev_feed_fd_event (loop, int fd, int revents)
715 765
716Feed an event on the given fd, as if a file descriptor backend detected it. 766Feed an event on the given fd, as if a file descriptor backend detected
767the given events it.
717 768
718=item ev_feed_signal_event (loop, int signum) 769=item ev_feed_signal_event (loop, int signum)
719 770
720Feed an event as if the given signal occured (loop must be the default loop!). 771Feed an event as if the given signal occured (loop must be the default loop!).
721 772
722=back 773=back
723 774
775=head1 LIBEVENT EMULATION
776
777Libev offers a compatibility emulation layer for libevent. It cannot
778emulate the internals of libevent, so here are some usage hints:
779
780=over 4
781
782=item * Use it by including <event.h>, as usual.
783
784=item * The following members are fully supported: ev_base, ev_callback,
785ev_arg, ev_fd, ev_res, ev_events.
786
787=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
788maintained by libev, it does not work exactly the same way as in libevent (consider
789it a private API).
790
791=item * Priorities are not currently supported. Initialising priorities
792will fail and all watchers will have the same priority, even though there
793is an ev_pri field.
794
795=item * Other members are not supported.
796
797=item * The libev emulation is I<not> ABI compatible to libevent, you need
798to use the libev header file and library.
799
800=back
801
802=head1 C++ SUPPORT
803
804TBD.
805
724=head1 AUTHOR 806=head1 AUTHOR
725 807
726Marc Lehmann <libev@schmorp.de>. 808Marc Lehmann <libev@schmorp.de>.
727 809

Diff Legend

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