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

Comparing libev/ev.pod (file contents):
Revision 1.1 by root, Mon Nov 12 07:58:13 2007 UTC vs.
Revision 1.27 by root, Wed Nov 14 05:02:07 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. This type is 46Libev represents time as a single floating point number, representing the
47(fractional) number of seconds since the (POSIX) epoch (somewhere near
48the beginning of 1970, details are complicated, don't ask). This type is
45called 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
46to the double type in C. 50to the double type in C.
47 51
52=head1 GLOBAL FUNCTIONS
53
54These functions can be called anytime, even before initialising the
55library in any way.
56
48=over 4 57=over 4
49 58
50=item ev_tstamp ev_time () 59=item ev_tstamp ev_time ()
51 60
52Returns the current time as libev would use it. 61Returns the current time as libev would use it. Please note that the
62C<ev_now> function is usually faster and also often returns the timestamp
63you actually want to know.
53 64
54=item int ev_version_major () 65=item int ev_version_major ()
55 66
56=item int ev_version_minor () 67=item int ev_version_minor ()
57 68
… …
59you linked against by calling the functions C<ev_version_major> and 70you linked against by calling the functions C<ev_version_major> and
60C<ev_version_minor>. If you want, you can compare against the global 71C<ev_version_minor>. If you want, you can compare against the global
61symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the 72symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
62version of the library your program was compiled against. 73version of the library your program was compiled against.
63 74
64Usually, its a good idea to terminate if the major versions mismatch, 75Usually, it's a good idea to terminate if the major versions mismatch,
65as this indicates an incompatible change. Minor versions are usually 76as this indicates an incompatible change. Minor versions are usually
66compatible to older versions, so a larger minor version alone is usually 77compatible to older versions, so a larger minor version alone is usually
67not a problem. 78not a problem.
68 79
69=item ev_set_allocator (void *(*cb)(void *ptr, long size)) 80=item ev_set_allocator (void *(*cb)(void *ptr, long size))
70 81
71Sets the allocation function to use (the prototype is similar to the 82Sets the allocation function to use (the prototype is similar to the
72realloc function). It is used to allocate and free memory (no surprises 83realloc C function, the semantics are identical). It is used to allocate
73here). If it returns zero when memory needs to be allocated, the library 84and free memory (no surprises here). If it returns zero when memory
74might abort or take some potentially destructive action. The default is 85needs to be allocated, the library might abort or take some potentially
75your system realloc function. 86destructive action. The default is your system realloc function.
76 87
77You could override this function in high-availability programs to, say, 88You could override this function in high-availability programs to, say,
78free some memory if it cannot allocate memory, to use a special allocator, 89free some memory if it cannot allocate memory, to use a special allocator,
79or even to sleep a while and retry until some memory is available. 90or even to sleep a while and retry until some memory is available.
80 91
… …
82 93
83Set the callback function to call on a retryable syscall error (such 94Set the callback function to call on a retryable syscall error (such
84as failed select, poll, epoll_wait). The message is a printable string 95as failed select, poll, epoll_wait). The message is a printable string
85indicating the system call or subsystem causing the problem. If this 96indicating the system call or subsystem causing the problem. If this
86callback is set, then libev will expect it to remedy the sitution, no 97callback is set, then libev will expect it to remedy the sitution, no
87matter what, when it returns. That is, libev will geenrally retry the 98matter what, when it returns. That is, libev will generally retry the
88requested operation, or, if the condition doesn't go away, do bad stuff 99requested operation, or, if the condition doesn't go away, do bad stuff
89(such as abort). 100(such as abort).
90 101
91=back 102=back
92 103
… …
95An event loop is described by a C<struct ev_loop *>. The library knows two 106An event loop is described by a C<struct ev_loop *>. The library knows two
96types of such loops, the I<default> loop, which supports signals and child 107types of such loops, the I<default> loop, which supports signals and child
97events, and dynamically created loops which do not. 108events, and dynamically created loops which do not.
98 109
99If you use threads, a common model is to run the default event loop 110If you use threads, a common model is to run the default event loop
100in your main thread (or in a separate thrad) and for each thread you 111in your main thread (or in a separate thread) and for each thread you
101create, you also create another event loop. Libev itself does no lockign 112create, you also create another event loop. Libev itself does no locking
102whatsoever, so if you mix calls to different event loops, make sure you 113whatsoever, so if you mix calls to the same event loop in different
103lock (this is usually a bad idea, though, even if done right). 114threads, make sure you lock (this is usually a bad idea, though, even if
115done correctly, because it's hideous and inefficient).
104 116
105=over 4 117=over 4
106 118
107=item struct ev_loop *ev_default_loop (unsigned int flags) 119=item struct ev_loop *ev_default_loop (unsigned int flags)
108 120
… …
113 125
114If you don't know what event loop to use, use the one returned from this 126If you don't know what event loop to use, use the one returned from this
115function. 127function.
116 128
117The flags argument can be used to specify special behaviour or specific 129The flags argument can be used to specify special behaviour or specific
118backends to use, and is usually specified as 0 (or EVFLAG_AUTO) 130backends to use, and is usually specified as 0 (or EVFLAG_AUTO).
119 131
120It supports the following flags: 132It supports the following flags:
121 133
122=over 4 134=over 4
123 135
124=item EVFLAG_AUTO 136=item C<EVFLAG_AUTO>
125 137
126The default flags value. Use this if you have no clue (its the right 138The default flags value. Use this if you have no clue (it's the right
127thing, believe me). 139thing, believe me).
128 140
129=item EVFLAG_NOENV 141=item C<EVFLAG_NOENV>
130 142
131If this flag bit is ored into the flag value then libev will I<not> look 143If this flag bit is ored into the flag value (or the program runs setuid
132at the environment variable C<LIBEV_FLAGS>. Otherwise (the default), this 144or setgid) then libev will I<not> look at the environment variable
133environment variable will override the flags completely. This is useful 145C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146override the flags completely if it is found in the environment. This is
134to try out specific backends to tets their performance, or to work around 147useful to try out specific backends to test their performance, or to work
135bugs. 148around bugs.
136 149
137=item EVMETHOD_SELECT portable select backend 150=item C<EVMETHOD_SELECT> (portable select backend)
138 151
139=item EVMETHOD_POLL poll backend (everywhere except windows) 152=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)
140 153
141=item EVMETHOD_EPOLL linux only 154=item C<EVMETHOD_EPOLL> (linux only)
142 155
143=item EVMETHOD_KQUEUE some bsds only 156=item C<EVMETHOD_KQUEUE> (some bsds only)
144 157
145=item EVMETHOD_DEVPOLL solaris 8 only 158=item C<EVMETHOD_DEVPOLL> (solaris 8 only)
146 159
147=item EVMETHOD_PORT solaris 10 only 160=item C<EVMETHOD_PORT> (solaris 10 only)
148 161
149If one or more of these are ored into the flags value, then only these 162If one or more of these are ored into the flags value, then only these
150backends will be tried (in the reverse order as given here). If one are 163backends will be tried (in the reverse order as given here). If one are
151specified, any backend will do. 164specified, any backend will do.
152 165
… …
161 174
162=item ev_default_destroy () 175=item ev_default_destroy ()
163 176
164Destroys the default loop again (frees all memory and kernel state 177Destroys the default loop again (frees all memory and kernel state
165etc.). This stops all registered event watchers (by not touching them in 178etc.). This stops all registered event watchers (by not touching them in
166any way whatsoever, although you cnanot rely on this :). 179any way whatsoever, although you cannot rely on this :).
167 180
168=item ev_loop_destroy (loop) 181=item ev_loop_destroy (loop)
169 182
170Like C<ev_default_destroy>, but destroys an event loop created by an 183Like C<ev_default_destroy>, but destroys an event loop created by an
171earlier call to C<ev_loop_new>. 184earlier call to C<ev_loop_new>.
… …
179 192
180You I<must> call this function after forking if and only if you want to 193You I<must> call this function after forking if and only if you want to
181use the event library in both processes. If you just fork+exec, you don't 194use the event library in both processes. If you just fork+exec, you don't
182have to call it. 195have to call it.
183 196
184The function itself is quite fast and its usually not a problem to call 197The function itself is quite fast and it's usually not a problem to call
185it just in case after a fork. To make this easy, the function will fit in 198it just in case after a fork. To make this easy, the function will fit in
186quite nicely into a call to C<pthread_atfork>: 199quite nicely into a call to C<pthread_atfork>:
187 200
188 pthread_atfork (0, 0, ev_default_fork); 201 pthread_atfork (0, 0, ev_default_fork);
189 202
… …
196=item unsigned int ev_method (loop) 209=item unsigned int ev_method (loop)
197 210
198Returns one of the C<EVMETHOD_*> flags indicating the event backend in 211Returns one of the C<EVMETHOD_*> flags indicating the event backend in
199use. 212use.
200 213
201=item ev_tstamp = ev_now (loop) 214=item ev_tstamp ev_now (loop)
202 215
203Returns the current "event loop time", which is the time the event loop 216Returns the current "event loop time", which is the time the event loop
204got events and started processing them. This timestamp does not change 217got events and started processing them. This timestamp does not change
205as long as callbacks are being processed, and this is also the base time 218as long as callbacks are being processed, and this is also the base time
206used for relative timers. You can treat it as the timestamp of the event 219used for relative timers. You can treat it as the timestamp of the event
… …
215If the flags argument is specified as 0, it will not return until either 228If the flags argument is specified as 0, it will not return until either
216no event watchers are active anymore or C<ev_unloop> was called. 229no event watchers are active anymore or C<ev_unloop> was called.
217 230
218A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle 231A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
219those events and any outstanding ones, but will not block your process in 232those events and any outstanding ones, but will not block your process in
220case there are no events. 233case there are no events and will return after one iteration of the loop.
221 234
222A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if 235A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
223neccessary) and will handle those and any outstanding ones. It will block 236neccessary) and will handle those and any outstanding ones. It will block
224your process until at least one new event arrives. 237your process until at least one new event arrives, and will return after
238one iteration of the loop.
225 239
226This flags value could be used to implement alternative looping 240This flags value could be used to implement alternative looping
227constructs, but the C<prepare> and C<check> watchers provide a better and 241constructs, but the C<prepare> and C<check> watchers provide a better and
228more generic mechanism. 242more generic mechanism.
229 243
244Here are the gory details of what ev_loop does:
245
246 1. If there are no active watchers (reference count is zero), return.
247 2. Queue and immediately call all prepare watchers.
248 3. If we have been forked, recreate the kernel state.
249 4. Update the kernel state with all outstanding changes.
250 5. Update the "event loop time".
251 6. Calculate for how long to block.
252 7. Block the process, waiting for events.
253 8. Update the "event loop time" and do time jump handling.
254 9. Queue all outstanding timers.
255 10. Queue all outstanding periodics.
256 11. If no events are pending now, queue all idle watchers.
257 12. Queue all check watchers.
258 13. Call all queued watchers in reverse order (i.e. check watchers first).
259 14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
260 was used, return, otherwise continue with step #1.
261
230=item ev_unloop (loop, how) 262=item ev_unloop (loop, how)
231 263
232Can be used to make a call to C<ev_loop> return early. The C<how> argument 264Can be used to make a call to C<ev_loop> return early (but only after it
265has processed all outstanding events). The C<how> argument must be either
233must be either C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> 266C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
234call return, or C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> 267C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
235calls return.
236 268
237=item ev_ref (loop) 269=item ev_ref (loop)
238 270
239=item ev_unref (loop) 271=item ev_unref (loop)
240 272
241Ref/unref can be used to add or remove a refcount on the event loop: Every 273Ref/unref can be used to add or remove a reference count on the event
242watcher keeps one reference. If you have a long-runing watcher you never 274loop: Every watcher keeps one reference, and as long as the reference
243unregister that should not keep ev_loop from running, ev_unref() after 275count is nonzero, C<ev_loop> will not return on its own. If you have
244starting, and ev_ref() before stopping it. Libev itself uses this for 276a watcher you never unregister that should not keep C<ev_loop> from
245example for its internal signal pipe: It is not visible to you as a user 277returning, ev_unref() after starting, and ev_ref() before stopping it. For
246and should not keep C<ev_loop> from exiting if the work is done. It is 278example, libev itself uses this for its internal signal pipe: It is not
247also an excellent way to do this for generic recurring timers or from 279visible to the libev user and should not keep C<ev_loop> from exiting if
248within third-party libraries. Just remember to unref after start and ref 280no event watchers registered by it are active. It is also an excellent
249before stop. 281way to do this for generic recurring timers or from within third-party
282libraries. Just remember to I<unref after start> and I<ref before stop>.
250 283
251=back 284=back
252 285
253=head1 ANATOMY OF A WATCHER 286=head1 ANATOMY OF A WATCHER
254 287
255A watcher is a structure that you create and register to record your 288A watcher is a structure that you create and register to record your
256interest in some event. For instance, if you want to wait for STDIN to 289interest in some event. For instance, if you want to wait for STDIN to
257become readable, you would create an ev_io watcher for that: 290become readable, you would create an C<ev_io> watcher for that:
258 291
259 static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents) 292 static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)
260 { 293 {
261 ev_io_stop (w); 294 ev_io_stop (w);
262 ev_unloop (loop, EVUNLOOP_ALL); 295 ev_unloop (loop, EVUNLOOP_ALL);
… …
291 324
292As long as your watcher is active (has been started but not stopped) you 325As long as your watcher is active (has been started but not stopped) you
293must not touch the values stored in it. Most specifically you must never 326must not touch the values stored in it. Most specifically you must never
294reinitialise it or call its set method. 327reinitialise it or call its set method.
295 328
296You cna check wether an event is active by calling the C<ev_is_active 329You can check whether an event is active by calling the C<ev_is_active
297(watcher *)> macro. To see wether an event is outstanding (but the 330(watcher *)> macro. To see whether an event is outstanding (but the
298callback for it has not been called yet) you cna use the C<ev_is_pending 331callback for it has not been called yet) you can use the C<ev_is_pending
299(watcher *)> macro. 332(watcher *)> macro.
300 333
301Each and every callback receives the event loop pointer as first, the 334Each and every callback receives the event loop pointer as first, the
302registered watcher structure as second, and a bitset of received events as 335registered watcher structure as second, and a bitset of received events as
303third argument. 336third argument.
304 337
305The rceeived events usually include a single bit per event type received 338The received events usually include a single bit per event type received
306(you can receive multiple events at the same time). The possible bit masks 339(you can receive multiple events at the same time). The possible bit masks
307are: 340are:
308 341
309=over 4 342=over 4
310 343
311=item EV_READ 344=item C<EV_READ>
312 345
313=item EV_WRITE 346=item C<EV_WRITE>
314 347
315The file descriptor in the ev_io watcher has become readable and/or 348The file descriptor in the C<ev_io> watcher has become readable and/or
316writable. 349writable.
317 350
318=item EV_TIMEOUT 351=item C<EV_TIMEOUT>
319 352
320The ev_timer watcher has timed out. 353The C<ev_timer> watcher has timed out.
321 354
322=item EV_PERIODIC 355=item C<EV_PERIODIC>
323 356
324The ev_periodic watcher has timed out. 357The C<ev_periodic> watcher has timed out.
325 358
326=item EV_SIGNAL 359=item C<EV_SIGNAL>
327 360
328The signal specified in the ev_signal watcher has been received by a thread. 361The signal specified in the C<ev_signal> watcher has been received by a thread.
329 362
330=item EV_CHILD 363=item C<EV_CHILD>
331 364
332The pid specified in the ev_child watcher has received a status change. 365The pid specified in the C<ev_child> watcher has received a status change.
333 366
334=item EV_IDLE 367=item C<EV_IDLE>
335 368
336The ev_idle watcher has determined that you have nothing better to do. 369The C<ev_idle> watcher has determined that you have nothing better to do.
337 370
338=item EV_PREPARE 371=item C<EV_PREPARE>
339 372
340=item EV_CHECK 373=item C<EV_CHECK>
341 374
342All ev_prepare watchers are invoked just I<before> C<ev_loop> starts 375All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts
343to gather new events, and all ev_check watchers are invoked just after 376to gather new events, and all C<ev_check> watchers are invoked just after
344C<ev_loop> has gathered them, but before it invokes any callbacks for any 377C<ev_loop> has gathered them, but before it invokes any callbacks for any
345received events. Callbacks of both watcher types can start and stop as 378received events. Callbacks of both watcher types can start and stop as
346many watchers as they want, and all of them will be taken into account 379many watchers as they want, and all of them will be taken into account
347(for example, a ev_prepare watcher might start an idle watcher to keep 380(for example, a C<ev_prepare> watcher might start an idle watcher to keep
348C<ev_loop> from blocking). 381C<ev_loop> from blocking).
349 382
350=item EV_ERROR 383=item C<EV_ERROR>
351 384
352An unspecified error has occured, the watcher has been stopped. This might 385An unspecified error has occured, the watcher has been stopped. This might
353happen because the watcher could not be properly started because libev 386happen because the watcher could not be properly started because libev
354ran out of memory, a file descriptor was found to be closed or any other 387ran out of memory, a file descriptor was found to be closed or any other
355problem. You best act on it by reporting the problem and somehow coping 388problem. You best act on it by reporting the problem and somehow coping
… …
364=back 397=back
365 398
366=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER 399=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
367 400
368Each watcher has, by default, a member C<void *data> that you can change 401Each watcher has, by default, a member C<void *data> that you can change
369and read at any time, libev will completely ignore it. This cna be used 402and read at any time, libev will completely ignore it. This can be used
370to associate arbitrary data with your watcher. If you need more data and 403to associate arbitrary data with your watcher. If you need more data and
371don't want to allocate memory and store a pointer to it in that data 404don't want to allocate memory and store a pointer to it in that data
372member, you can also "subclass" the watcher type and provide your own 405member, you can also "subclass" the watcher type and provide your own
373data: 406data:
374 407
… …
396=head1 WATCHER TYPES 429=head1 WATCHER TYPES
397 430
398This section describes each watcher in detail, but will not repeat 431This section describes each watcher in detail, but will not repeat
399information given in the last section. 432information given in the last section.
400 433
401=head2 struct ev_io - is my file descriptor readable or writable 434=head2 C<ev_io> - is this file descriptor readable or writable
402 435
403I/O watchers check wether a file descriptor is readable or writable 436I/O watchers check whether a file descriptor is readable or writable
404in each iteration of the event loop (This behaviour is called 437in each iteration of the event loop (This behaviour is called
405level-triggering because you keep receiving events as long as the 438level-triggering because you keep receiving events as long as the
406condition persists. Remember you cna stop the watcher if you don't want to 439condition persists. Remember you can stop the watcher if you don't want to
407act on the event and neither want to receive future events). 440act on the event and neither want to receive future events).
408 441
442In general you can register as many read and/or write event watchers per
443fd as you want (as long as you don't confuse yourself). Setting all file
444descriptors to non-blocking mode is also usually a good idea (but not
445required if you know what you are doing).
446
447You have to be careful with dup'ed file descriptors, though. Some backends
448(the linux epoll backend is a notable example) cannot handle dup'ed file
449descriptors correctly if you register interest in two or more fds pointing
450to the same underlying file/socket etc. description (that is, they share
451the same underlying "file open").
452
453If you must do this, then force the use of a known-to-be-good backend
454(at the time of this writing, this includes only EVMETHOD_SELECT and
455EVMETHOD_POLL).
456
409=over 4 457=over 4
410 458
411=item ev_io_init (ev_io *, callback, int fd, int events) 459=item ev_io_init (ev_io *, callback, int fd, int events)
412 460
413=item ev_io_set (ev_io *, int fd, int events) 461=item ev_io_set (ev_io *, int fd, int events)
414 462
415Configures an ev_io watcher. The fd is the file descriptor to rceeive 463Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive
416events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ | 464events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |
417EV_WRITE> to receive the given events. 465EV_WRITE> to receive the given events.
418 466
419=back 467=back
420 468
421=head2 struct ev_timer - relative and optionally recurring timeouts 469=head2 C<ev_timer> - relative and optionally recurring timeouts
422 470
423Timer watchers are simple relative timers that generate an event after a 471Timer watchers are simple relative timers that generate an event after a
424given time, and optionally repeating in regular intervals after that. 472given time, and optionally repeating in regular intervals after that.
425 473
426The timers are based on real time, that is, if you register an event that 474The timers are based on real time, that is, if you register an event that
427times out after an hour and youreset your system clock to last years 475times out after an hour and you reset your system clock to last years
428time, it will still time out after (roughly) and hour. "Roughly" because 476time, it will still time out after (roughly) and hour. "Roughly" because
429detecting time jumps is hard, and soem inaccuracies are unavoidable (the 477detecting time jumps is hard, and soem inaccuracies are unavoidable (the
430monotonic clock option helps a lot here). 478monotonic clock option helps a lot here).
479
480The relative timeouts are calculated relative to the C<ev_now ()>
481time. This is usually the right thing as this timestamp refers to the time
482of the event triggering whatever timeout you are modifying/starting. If
483you suspect event processing to be delayed and you *need* to base the timeout
484on the current time, use something like this to adjust for this:
485
486 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
431 487
432=over 4 488=over 4
433 489
434=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat) 490=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
435 491
… …
441later, again, and again, until stopped manually. 497later, again, and again, until stopped manually.
442 498
443The timer itself will do a best-effort at avoiding drift, that is, if you 499The timer itself will do a best-effort at avoiding drift, that is, if you
444configure a timer to trigger every 10 seconds, then it will trigger at 500configure a timer to trigger every 10 seconds, then it will trigger at
445exactly 10 second intervals. If, however, your program cannot keep up with 501exactly 10 second intervals. If, however, your program cannot keep up with
446the timer (ecause it takes longer than those 10 seconds to do stuff) the 502the timer (because it takes longer than those 10 seconds to do stuff) the
447timer will not fire more than once per event loop iteration. 503timer will not fire more than once per event loop iteration.
448 504
449=item ev_timer_again (loop) 505=item ev_timer_again (loop)
450 506
451This will act as if the timer timed out and restart it again if it is 507This will act as if the timer timed out and restart it again if it is
… …
458 514
459This sounds a bit complicated, but here is a useful and typical 515This sounds a bit complicated, but here is a useful and typical
460example: Imagine you have a tcp connection and you want a so-called idle 516example: Imagine you have a tcp connection and you want a so-called idle
461timeout, that is, you want to be called when there have been, say, 60 517timeout, that is, you want to be called when there have been, say, 60
462seconds of inactivity on the socket. The easiest way to do this is to 518seconds of inactivity on the socket. The easiest way to do this is to
463configure an ev_timer with after=repeat=60 and calling ev_timer_again each 519configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each
464time you successfully read or write some data. If you go into an idle 520time you successfully read or write some data. If you go into an idle
465state where you do not expect data to travel on the socket, you can stop 521state where you do not expect data to travel on the socket, you can stop
466the timer, and again will automatically restart it if need be. 522the timer, and again will automatically restart it if need be.
467 523
468=back 524=back
469 525
470=head2 ev_periodic 526=head2 C<ev_periodic> - to cron or not to cron
471 527
472Periodic watchers are also timers of a kind, but they are very versatile 528Periodic watchers are also timers of a kind, but they are very versatile
473(and unfortunately a bit complex). 529(and unfortunately a bit complex).
474 530
475Unlike ev_timer's, they are not based on real time (or relative time) 531Unlike C<ev_timer>'s, they are not based on real time (or relative time)
476but on wallclock time (absolute time). You can tell a periodic watcher 532but on wallclock time (absolute time). You can tell a periodic watcher
477to trigger "at" some specific point in time. For example, if you tell a 533to trigger "at" some specific point in time. For example, if you tell a
478periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now () 534periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()
479+ 10.>) and then reset your system clock to the last year, then it will 535+ 10.>) and then reset your system clock to the last year, then it will
480take a year to trigger the event (unlike an ev_timer, which would trigger 536take a year to trigger the event (unlike an C<ev_timer>, which would trigger
481roughly 10 seconds later and of course not if you reset your system time 537roughly 10 seconds later and of course not if you reset your system time
482again). 538again).
483 539
484They can also be used to implement vastly more complex timers, such as 540They can also be used to implement vastly more complex timers, such as
485triggering an event on eahc midnight, local time. 541triggering an event on eahc midnight, local time.
… …
514 570
515 ev_periodic_set (&periodic, 0., 3600., 0); 571 ev_periodic_set (&periodic, 0., 3600., 0);
516 572
517This doesn't mean there will always be 3600 seconds in between triggers, 573This doesn't mean there will always be 3600 seconds in between triggers,
518but only that the the callback will be called when the system time shows a 574but only that the the callback will be called when the system time shows a
519full hour (UTC), or more correct, when the system time is evenly divisible 575full hour (UTC), or more correctly, when the system time is evenly divisible
520by 3600. 576by 3600.
521 577
522Another way to think about it (for the mathematically inclined) is that 578Another way to think about it (for the mathematically inclined) is that
523ev_periodic will try to run the callback in this mode at the next possible 579C<ev_periodic> will try to run the callback in this mode at the next possible
524time where C<time = at (mod interval)>, regardless of any time jumps. 580time where C<time = at (mod interval)>, regardless of any time jumps.
525 581
526=item * manual reschedule mode (reschedule_cb = callback) 582=item * manual reschedule mode (reschedule_cb = callback)
527 583
528In this mode the values for C<interval> and C<at> are both being 584In this mode the values for C<interval> and C<at> are both being
529ignored. Instead, each time the periodic watcher gets scheduled, the 585ignored. Instead, each time the periodic watcher gets scheduled, the
530reschedule callback will be called with the watcher as first, and the 586reschedule callback will be called with the watcher as first, and the
531current time as second argument. 587current time as second argument.
532 588
533NOTE: I<This callback MUST NOT stop or destroy the periodic or any other 589NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
534periodic watcher, ever, or make any event loop modificstions>. If you need 590ever, or make any event loop modifications>. If you need to stop it,
535to stop it, return 1e30 (or so, fudge fudge) and stop it afterwards. 591return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
592starting a prepare watcher).
536 593
537Its prototype is c<ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 594Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
538ev_tstamp now)>, e.g.: 595ev_tstamp now)>, e.g.:
539 596
540 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) 597 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
541 { 598 {
542 return now + 60.; 599 return now + 60.;
… …
545It must return the next time to trigger, based on the passed time value 602It must return the next time to trigger, based on the passed time value
546(that is, the lowest time value larger than to the second argument). It 603(that is, the lowest time value larger than to the second argument). It
547will usually be called just before the callback will be triggered, but 604will usually be called just before the callback will be triggered, but
548might be called at other times, too. 605might be called at other times, too.
549 606
607NOTE: I<< This callback must always return a time that is later than the
608passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.
609
550This can be used to create very complex timers, such as a timer that 610This can be used to create very complex timers, such as a timer that
551triggers on each midnight, local time. To do this, you would calculate the 611triggers on each midnight, local time. To do this, you would calculate the
552next midnight after C<now> and return the timestamp value for this. How you do this 612next midnight after C<now> and return the timestamp value for this. How
553is, again, up to you (but it is not trivial). 613you do this is, again, up to you (but it is not trivial, which is the main
614reason I omitted it as an example).
554 615
555=back 616=back
556 617
557=item ev_periodic_again (loop, ev_periodic *) 618=item ev_periodic_again (loop, ev_periodic *)
558 619
… …
561a different time than the last time it was called (e.g. in a crond like 622a different time than the last time it was called (e.g. in a crond like
562program when the crontabs have changed). 623program when the crontabs have changed).
563 624
564=back 625=back
565 626
566=head2 ev_signal - signal me when a signal gets signalled 627=head2 C<ev_signal> - signal me when a signal gets signalled
567 628
568Signal watchers will trigger an event when the process receives a specific 629Signal watchers will trigger an event when the process receives a specific
569signal one or more times. Even though signals are very asynchronous, libev 630signal one or more times. Even though signals are very asynchronous, libev
570will try its best to deliver signals synchronously, i.e. as part of the 631will try it's best to deliver signals synchronously, i.e. as part of the
571normal event processing, like any other event. 632normal event processing, like any other event.
572 633
573You cna configure as many watchers as you like per signal. Only when the 634You can configure as many watchers as you like per signal. Only when the
574first watcher gets started will libev actually register a signal watcher 635first watcher gets started will libev actually register a signal watcher
575with the kernel (thus it coexists with your own signal handlers as long 636with the kernel (thus it coexists with your own signal handlers as long
576as you don't register any with libev). Similarly, when the last signal 637as you don't register any with libev). Similarly, when the last signal
577watcher for a signal is stopped libev will reset the signal handler to 638watcher for a signal is stopped libev will reset the signal handler to
578SIG_DFL (regardless of what it was set to before). 639SIG_DFL (regardless of what it was set to before).
… …
586Configures the watcher to trigger on the given signal number (usually one 647Configures the watcher to trigger on the given signal number (usually one
587of the C<SIGxxx> constants). 648of the C<SIGxxx> constants).
588 649
589=back 650=back
590 651
591=head2 ev_child - wait for pid status changes 652=head2 C<ev_child> - wait for pid status changes
592 653
593Child watchers trigger when your process receives a SIGCHLD in response to 654Child watchers trigger when your process receives a SIGCHLD in response to
594some child status changes (most typically when a child of yours dies). 655some child status changes (most typically when a child of yours dies).
595 656
596=over 4 657=over 4
… …
600=item ev_child_set (ev_child *, int pid) 661=item ev_child_set (ev_child *, int pid)
601 662
602Configures the watcher to wait for status changes of process C<pid> (or 663Configures the watcher to wait for status changes of process C<pid> (or
603I<any> process if C<pid> is specified as C<0>). The callback can look 664I<any> process if C<pid> is specified as C<0>). The callback can look
604at the C<rstatus> member of the C<ev_child> watcher structure to see 665at the C<rstatus> member of the C<ev_child> watcher structure to see
605the status word (use the macros from C<sys/wait.h>). The C<rpid> member 666the status word (use the macros from C<sys/wait.h> and see your systems
606contains the pid of the process causing the status change. 667C<waitpid> documentation). The C<rpid> member contains the pid of the
668process causing the status change.
607 669
608=back 670=back
609 671
610=head2 ev_idle - when you've got nothing better to do 672=head2 C<ev_idle> - when you've got nothing better to do
611 673
612Idle watchers trigger events when there are no other I/O or timer (or 674Idle watchers trigger events when there are no other events are pending
613periodic) events pending. That is, as long as your process is busy 675(prepare, check and other idle watchers do not count). That is, as long
614handling sockets or timeouts it will not be called. But when your process 676as your process is busy handling sockets or timeouts (or even signals,
615is idle all idle watchers are being called again and again - until 677imagine) it will not be triggered. But when your process is idle all idle
678watchers are being called again and again, once per event loop iteration -
616stopped, that is, or your process receives more events. 679until stopped, that is, or your process receives more events and becomes
680busy.
617 681
618The most noteworthy effect is that as long as any idle watchers are 682The most noteworthy effect is that as long as any idle watchers are
619active, the process will not block when waiting for new events. 683active, the process will not block when waiting for new events.
620 684
621Apart from keeping your process non-blocking (which is a useful 685Apart from keeping your process non-blocking (which is a useful
… …
631kind. There is a C<ev_idle_set> macro, but using it is utterly pointless, 695kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
632believe me. 696believe me.
633 697
634=back 698=back
635 699
636=head2 prepare and check - your hooks into the event loop 700=head2 C<ev_prepare> and C<ev_check> - customise your event loop
637 701
638Prepare and check watchers usually (but not always) are used in 702Prepare and check watchers are usually (but not always) used in tandem:
639tandom. Prepare watchers get invoked before the process blocks and check 703prepare watchers get invoked before the process blocks and check watchers
640watchers afterwards. 704afterwards.
641 705
642Their main purpose is to integrate other event mechanisms into libev. This 706Their main purpose is to integrate other event mechanisms into libev. This
643could be used, for example, to track variable changes, implement your own 707could be used, for example, to track variable changes, implement your own
644watchers, integrate net-snmp or a coroutine library and lots more. 708watchers, integrate net-snmp or a coroutine library and lots more.
645 709
646This is done by examining in each prepare call which file descriptors need 710This is done by examining in each prepare call which file descriptors need
647to be watched by the other library, registering ev_io watchers for them 711to be watched by the other library, registering C<ev_io> watchers for
648and starting an ev_timer watcher for any timeouts (many libraries provide 712them and starting an C<ev_timer> watcher for any timeouts (many libraries
649just this functionality). Then, in the check watcher you check for any 713provide just this functionality). Then, in the check watcher you check for
650events that occured (by making your callbacks set soem flags for example) 714any events that occured (by checking the pending status of all watchers
651and call back into the library. 715and stopping them) and call back into the library. The I/O and timer
716callbacks will never actually be called (but must be valid nevertheless,
717because you never know, you know?).
652 718
653As another example, the perl Coro module uses these hooks to integrate 719As another example, the Perl Coro module uses these hooks to integrate
654coroutines into libev programs, by yielding to other active coroutines 720coroutines into libev programs, by yielding to other active coroutines
655during each prepare and only letting the process block if no coroutines 721during each prepare and only letting the process block if no coroutines
656are ready to run. 722are ready to run (it's actually more complicated: it only runs coroutines
723with priority higher than or equal to the event loop and one coroutine
724of lower priority, but only once, using idle watchers to keep the event
725loop from blocking if lower-priority coroutines are active, thus mapping
726low-priority coroutines to idle/background tasks).
657 727
658=over 4 728=over 4
659 729
660=item ev_prepare_init (ev_prepare *, callback) 730=item ev_prepare_init (ev_prepare *, callback)
661 731
662=item ev_check_init (ev_check *, callback) 732=item ev_check_init (ev_check *, callback)
663 733
664Initialises and configures the prepare or check watcher - they have no 734Initialises and configures the prepare or check watcher - they have no
665parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set> 735parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
666macros, but using them is utterly, utterly pointless. 736macros, but using them is utterly, utterly and completely pointless.
667 737
668=back 738=back
669 739
670=head1 OTHER FUNCTIONS 740=head1 OTHER FUNCTIONS
671 741
672There are some other fucntions of possible interest. Described. Here. Now. 742There are some other functions of possible interest. Described. Here. Now.
673 743
674=over 4 744=over 4
675 745
676=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback) 746=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
677 747
678This function combines a simple timer and an I/O watcher, calls your 748This function combines a simple timer and an I/O watcher, calls your
679callback on whichever event happens first and automatically stop both 749callback on whichever event happens first and automatically stop both
680watchers. This is useful if you want to wait for a single event on an fd 750watchers. This is useful if you want to wait for a single event on an fd
681or timeout without havign to allocate/configure/start/stop/free one or 751or timeout without having to allocate/configure/start/stop/free one or
682more watchers yourself. 752more watchers yourself.
683 753
684If C<fd> is less than 0, then no I/O watcher will be started and events is 754If C<fd> is less than 0, then no I/O watcher will be started and events
685ignored. Otherwise, an ev_io watcher for the given C<fd> and C<events> set 755is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and
686will be craeted and started. 756C<events> set will be craeted and started.
687 757
688If C<timeout> is less than 0, then no timeout watcher will be 758If C<timeout> is less than 0, then no timeout watcher will be
689started. Otherwise an ev_timer watcher with after = C<timeout> (and repeat 759started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
690= 0) will be started. 760repeat = 0) will be started. While C<0> is a valid timeout, it is of
761dubious value.
691 762
692The callback has the type C<void (*cb)(int revents, void *arg)> and 763The callback has the type C<void (*cb)(int revents, void *arg)> and gets
693gets passed an events set (normally a combination of EV_ERROR, EV_READ, 764passed an C<revents> set like normal event callbacks (a combination of
694EV_WRITE or EV_TIMEOUT) and the C<arg> value passed to C<ev_once>: 765C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
766value passed to C<ev_once>:
695 767
696 static void stdin_ready (int revents, void *arg) 768 static void stdin_ready (int revents, void *arg)
697 { 769 {
698 if (revents & EV_TIMEOUT) 770 if (revents & EV_TIMEOUT)
699 /* doh, nothing entered */ 771 /* doh, nothing entered */;
700 else if (revents & EV_READ) 772 else if (revents & EV_READ)
701 /* stdin might have data for us, joy! */ 773 /* stdin might have data for us, joy! */;
702 } 774 }
703 775
704 ev_once (STDIN_FILENO, EV_READm 10., stdin_ready, 0); 776 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
705 777
706=item ev_feed_event (loop, watcher, int events) 778=item ev_feed_event (loop, watcher, int events)
707 779
708Feeds the given event set into the event loop, as if the specified event 780Feeds the given event set into the event loop, as if the specified event
709has happened for the specified watcher (which must be a pointer to an 781had happened for the specified watcher (which must be a pointer to an
710initialised but not necessarily active event watcher). 782initialised but not necessarily started event watcher).
711 783
712=item ev_feed_fd_event (loop, int fd, int revents) 784=item ev_feed_fd_event (loop, int fd, int revents)
713 785
714Feed an event on the given fd, as if a file descriptor backend detected it. 786Feed an event on the given fd, as if a file descriptor backend detected
787the given events it.
715 788
716=item ev_feed_signal_event (loop, int signum) 789=item ev_feed_signal_event (loop, int signum)
717 790
718Feed an event as if the given signal occured (loop must be the default loop!). 791Feed an event as if the given signal occured (loop must be the default loop!).
719 792
720=back 793=back
721 794
795=head1 LIBEVENT EMULATION
796
797Libev offers a compatibility emulation layer for libevent. It cannot
798emulate the internals of libevent, so here are some usage hints:
799
800=over 4
801
802=item * Use it by including <event.h>, as usual.
803
804=item * The following members are fully supported: ev_base, ev_callback,
805ev_arg, ev_fd, ev_res, ev_events.
806
807=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
808maintained by libev, it does not work exactly the same way as in libevent (consider
809it a private API).
810
811=item * Priorities are not currently supported. Initialising priorities
812will fail and all watchers will have the same priority, even though there
813is an ev_pri field.
814
815=item * Other members are not supported.
816
817=item * The libev emulation is I<not> ABI compatible to libevent, you need
818to use the libev header file and library.
819
820=back
821
822=head1 C++ SUPPORT
823
824TBD.
825
722=head1 AUTHOR 826=head1 AUTHOR
723 827
724Marc Lehmann <libev@schmorp.de>. 828Marc Lehmann <libev@schmorp.de>.
725 829

Diff Legend

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