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

Comparing libev/ev.pod (file contents):
Revision 1.13 by root, Mon Nov 12 08:35:36 2007 UTC vs.
Revision 1.26 by root, Tue Nov 13 03:11:57 2007 UTC

… …
39F<README.embed> in the libev distribution. If libev was configured without 39F<README.embed> in the libev distribution. If libev was configured without
40support for multiple event loops, then all functions taking an initial 40support for multiple event loops, then all functions taking an initial
41argument 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 *>)
42will not have this argument. 42will not have this argument.
43 43
44=head1 TIME AND OTHER GLOBAL FUNCTIONS 44=head1 TIME REPRESENTATION
45 45
46Libev represents time as a single floating point number, representing the 46Libev represents time as a single floating point number, representing the
47(fractional) number of seconds since the (POSIX) epoch (somewhere near 47(fractional) number of seconds since the (POSIX) epoch (somewhere near
48the 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
49called 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
50to the double type in C. 50to the double type in C.
51 51
52=head1 GLOBAL FUNCTIONS
53
54These functions can be called anytime, even before initialising the
55library in any way.
56
52=over 4 57=over 4
53 58
54=item ev_tstamp ev_time () 59=item ev_tstamp ev_time ()
55 60
56Returns 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.
57 64
58=item int ev_version_major () 65=item int ev_version_major ()
59 66
60=item int ev_version_minor () 67=item int ev_version_minor ()
61 68
… …
99An 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
100types 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
101events, and dynamically created loops which do not. 108events, and dynamically created loops which do not.
102 109
103If 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
104in 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
105create, you also create another event loop. Libev itself does no locking 112create, you also create another event loop. Libev itself does no locking
106whatsoever, so if you mix calls to the same event loop in different 113whatsoever, so if you mix calls to the same event loop in different
107threads, make sure you lock (this is usually a bad idea, though, even if 114threads, make sure you lock (this is usually a bad idea, though, even if
108done correctly, because it's hideous and inefficient). 115done correctly, because it's hideous and inefficient).
109 116
… …
236 243
237=item ev_unloop (loop, how) 244=item ev_unloop (loop, how)
238 245
239Can be used to make a call to C<ev_loop> return early (but only after it 246Can be used to make a call to C<ev_loop> return early (but only after it
240has processed all outstanding events). The C<how> argument must be either 247has processed all outstanding events). The C<how> argument must be either
241C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> call return, or 248C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
242C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return. 249C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
243 250
244=item ev_ref (loop) 251=item ev_ref (loop)
245 252
246=item ev_unref (loop) 253=item ev_unref (loop)
… …
299 306
300As long as your watcher is active (has been started but not stopped) you 307As long as your watcher is active (has been started but not stopped) you
301must not touch the values stored in it. Most specifically you must never 308must not touch the values stored in it. Most specifically you must never
302reinitialise it or call its set method. 309reinitialise it or call its set method.
303 310
304You cna check whether an event is active by calling the C<ev_is_active 311You can check whether an event is active by calling the C<ev_is_active
305(watcher *)> macro. To see whether an event is outstanding (but the 312(watcher *)> macro. To see whether an event is outstanding (but the
306callback for it has not been called yet) you cna use the C<ev_is_pending 313callback for it has not been called yet) you can use the C<ev_is_pending
307(watcher *)> macro. 314(watcher *)> macro.
308 315
309Each and every callback receives the event loop pointer as first, the 316Each and every callback receives the event loop pointer as first, the
310registered watcher structure as second, and a bitset of received events as 317registered watcher structure as second, and a bitset of received events as
311third argument. 318third argument.
312 319
313The rceeived events usually include a single bit per event type received 320The received events usually include a single bit per event type received
314(you can receive multiple events at the same time). The possible bit masks 321(you can receive multiple events at the same time). The possible bit masks
315are: 322are:
316 323
317=over 4 324=over 4
318 325
… …
372=back 379=back
373 380
374=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER 381=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
375 382
376Each watcher has, by default, a member C<void *data> that you can change 383Each watcher has, by default, a member C<void *data> that you can change
377and read at any time, libev will completely ignore it. This cna be used 384and read at any time, libev will completely ignore it. This can be used
378to associate arbitrary data with your watcher. If you need more data and 385to associate arbitrary data with your watcher. If you need more data and
379don't want to allocate memory and store a pointer to it in that data 386don't want to allocate memory and store a pointer to it in that data
380member, you can also "subclass" the watcher type and provide your own 387member, you can also "subclass" the watcher type and provide your own
381data: 388data:
382 389
… …
409=head2 C<ev_io> - is this file descriptor readable or writable 416=head2 C<ev_io> - is this file descriptor readable or writable
410 417
411I/O watchers check whether a file descriptor is readable or writable 418I/O watchers check whether a file descriptor is readable or writable
412in each iteration of the event loop (This behaviour is called 419in each iteration of the event loop (This behaviour is called
413level-triggering because you keep receiving events as long as the 420level-triggering because you keep receiving events as long as the
414condition persists. Remember you cna stop the watcher if you don't want to 421condition persists. Remember you can stop the watcher if you don't want to
415act on the event and neither want to receive future events). 422act on the event and neither want to receive future events).
416 423
417In general you can register as many read and/or write event watchers oer 424In general you can register as many read and/or write event watchers per
418fd as you want (as long as you don't confuse yourself). Setting all file 425fd as you want (as long as you don't confuse yourself). Setting all file
419descriptors to non-blocking mode is also usually a good idea (but not 426descriptors to non-blocking mode is also usually a good idea (but not
420required if you know what you are doing). 427required if you know what you are doing).
421 428
422You have to be careful with dup'ed file descriptors, though. Some backends 429You have to be careful with dup'ed file descriptors, though. Some backends
423(the linux epoll backend is a notable example) cannot handle dup'ed file 430(the linux epoll backend is a notable example) cannot handle dup'ed file
424descriptors correctly if you register interest in two or more fds pointing 431descriptors correctly if you register interest in two or more fds pointing
425to the same file/socket etc. description. 432to the same underlying file/socket etc. description (that is, they share
433the same underlying "file open").
426 434
427If you must do this, then force the use of a known-to-be-good backend 435If you must do this, then force the use of a known-to-be-good backend
428(at the time of this writing, this includes only EVMETHOD_SELECT and 436(at the time of this writing, this includes only EVMETHOD_SELECT and
429EVMETHOD_POLL). 437EVMETHOD_POLL).
430 438
… …
444 452
445Timer watchers are simple relative timers that generate an event after a 453Timer watchers are simple relative timers that generate an event after a
446given time, and optionally repeating in regular intervals after that. 454given time, and optionally repeating in regular intervals after that.
447 455
448The timers are based on real time, that is, if you register an event that 456The timers are based on real time, that is, if you register an event that
449times out after an hour and youreset your system clock to last years 457times out after an hour and you reset your system clock to last years
450time, it will still time out after (roughly) and hour. "Roughly" because 458time, it will still time out after (roughly) and hour. "Roughly" because
451detecting time jumps is hard, and soem inaccuracies are unavoidable (the 459detecting time jumps is hard, and soem inaccuracies are unavoidable (the
452monotonic clock option helps a lot here). 460monotonic clock option helps a lot here).
453 461
454The relative timeouts are calculated relative to the C<ev_now ()> 462The relative timeouts are calculated relative to the C<ev_now ()>
455time. This is usually the right thing as this timestamp refers to the time 463time. This is usually the right thing as this timestamp refers to the time
456of the event triggering whatever timeout you are modifying/starting. If 464of the event triggering whatever timeout you are modifying/starting. If
457you suspect event processing to be delayed and you *need* to base the timeout 465you suspect event processing to be delayed and you *need* to base the timeout
458ion the current time, use something like this to adjust for this: 466on the current time, use something like this to adjust for this:
459 467
460 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.); 468 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
461 469
462=over 4 470=over 4
463 471
… …
471later, again, and again, until stopped manually. 479later, again, and again, until stopped manually.
472 480
473The timer itself will do a best-effort at avoiding drift, that is, if you 481The timer itself will do a best-effort at avoiding drift, that is, if you
474configure a timer to trigger every 10 seconds, then it will trigger at 482configure a timer to trigger every 10 seconds, then it will trigger at
475exactly 10 second intervals. If, however, your program cannot keep up with 483exactly 10 second intervals. If, however, your program cannot keep up with
476the timer (ecause it takes longer than those 10 seconds to do stuff) the 484the timer (because it takes longer than those 10 seconds to do stuff) the
477timer will not fire more than once per event loop iteration. 485timer will not fire more than once per event loop iteration.
478 486
479=item ev_timer_again (loop) 487=item ev_timer_again (loop)
480 488
481This will act as if the timer timed out and restart it again if it is 489This will act as if the timer timed out and restart it again if it is
… …
495state where you do not expect data to travel on the socket, you can stop 503state where you do not expect data to travel on the socket, you can stop
496the timer, and again will automatically restart it if need be. 504the timer, and again will automatically restart it if need be.
497 505
498=back 506=back
499 507
500=head2 C<ev_periodic> - to cron or not to cron it 508=head2 C<ev_periodic> - to cron or not to cron
501 509
502Periodic watchers are also timers of a kind, but they are very versatile 510Periodic watchers are also timers of a kind, but they are very versatile
503(and unfortunately a bit complex). 511(and unfortunately a bit complex).
504 512
505Unlike C<ev_timer>'s, they are not based on real time (or relative time) 513Unlike C<ev_timer>'s, they are not based on real time (or relative time)
… …
558In this mode the values for C<interval> and C<at> are both being 566In this mode the values for C<interval> and C<at> are both being
559ignored. Instead, each time the periodic watcher gets scheduled, the 567ignored. Instead, each time the periodic watcher gets scheduled, the
560reschedule callback will be called with the watcher as first, and the 568reschedule callback will be called with the watcher as first, and the
561current time as second argument. 569current time as second argument.
562 570
563NOTE: I<This callback MUST NOT stop or destroy the periodic or any other 571NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
564periodic watcher, ever, or make any event loop modifications>. If you need 572ever, or make any event loop modifications>. If you need to stop it,
565to stop it, return C<now + 1e30> (or so, fudge fudge) and stop it afterwards. 573return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
566 574starting a prepare watcher).
567Also, I<< this callback must always return a time that is later than the
568passed C<now> value >>. Not even C<now> itself will be ok.
569 575
570Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 576Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
571ev_tstamp now)>, e.g.: 577ev_tstamp now)>, e.g.:
572 578
573 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) 579 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
… …
578It must return the next time to trigger, based on the passed time value 584It must return the next time to trigger, based on the passed time value
579(that is, the lowest time value larger than to the second argument). It 585(that is, the lowest time value larger than to the second argument). It
580will usually be called just before the callback will be triggered, but 586will usually be called just before the callback will be triggered, but
581might be called at other times, too. 587might be called at other times, too.
582 588
589NOTE: I<< This callback must always return a time that is later than the
590passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.
591
583This can be used to create very complex timers, such as a timer that 592This can be used to create very complex timers, such as a timer that
584triggers on each midnight, local time. To do this, you would calculate the 593triggers on each midnight, local time. To do this, you would calculate the
585next midnight after C<now> and return the timestamp value for this. How you do this 594next midnight after C<now> and return the timestamp value for this. How
586is, again, up to you (but it is not trivial). 595you do this is, again, up to you (but it is not trivial, which is the main
596reason I omitted it as an example).
587 597
588=back 598=back
589 599
590=item ev_periodic_again (loop, ev_periodic *) 600=item ev_periodic_again (loop, ev_periodic *)
591 601
… …
601Signal watchers will trigger an event when the process receives a specific 611Signal watchers will trigger an event when the process receives a specific
602signal one or more times. Even though signals are very asynchronous, libev 612signal one or more times. Even though signals are very asynchronous, libev
603will try it's best to deliver signals synchronously, i.e. as part of the 613will try it's best to deliver signals synchronously, i.e. as part of the
604normal event processing, like any other event. 614normal event processing, like any other event.
605 615
606You cna configure as many watchers as you like per signal. Only when the 616You can configure as many watchers as you like per signal. Only when the
607first watcher gets started will libev actually register a signal watcher 617first watcher gets started will libev actually register a signal watcher
608with the kernel (thus it coexists with your own signal handlers as long 618with the kernel (thus it coexists with your own signal handlers as long
609as you don't register any with libev). Similarly, when the last signal 619as you don't register any with libev). Similarly, when the last signal
610watcher for a signal is stopped libev will reset the signal handler to 620watcher for a signal is stopped libev will reset the signal handler to
611SIG_DFL (regardless of what it was set to before). 621SIG_DFL (regardless of what it was set to before).
… …
633=item ev_child_set (ev_child *, int pid) 643=item ev_child_set (ev_child *, int pid)
634 644
635Configures the watcher to wait for status changes of process C<pid> (or 645Configures the watcher to wait for status changes of process C<pid> (or
636I<any> process if C<pid> is specified as C<0>). The callback can look 646I<any> process if C<pid> is specified as C<0>). The callback can look
637at the C<rstatus> member of the C<ev_child> watcher structure to see 647at the C<rstatus> member of the C<ev_child> watcher structure to see
638the status word (use the macros from C<sys/wait.h>). The C<rpid> member 648the status word (use the macros from C<sys/wait.h> and see your systems
639contains the pid of the process causing the status change. 649C<waitpid> documentation). The C<rpid> member contains the pid of the
650process causing the status change.
640 651
641=back 652=back
642 653
643=head2 C<ev_idle> - when you've got nothing better to do 654=head2 C<ev_idle> - when you've got nothing better to do
644 655
645Idle watchers trigger events when there are no other I/O or timer (or 656Idle watchers trigger events when there are no other events are pending
646periodic) events pending. That is, as long as your process is busy 657(prepare, check and other idle watchers do not count). That is, as long
647handling sockets or timeouts it will not be called. But when your process 658as your process is busy handling sockets or timeouts (or even signals,
648is idle all idle watchers are being called again and again - until 659imagine) it will not be triggered. But when your process is idle all idle
660watchers are being called again and again, once per event loop iteration -
649stopped, that is, or your process receives more events. 661until stopped, that is, or your process receives more events and becomes
662busy.
650 663
651The most noteworthy effect is that as long as any idle watchers are 664The most noteworthy effect is that as long as any idle watchers are
652active, the process will not block when waiting for new events. 665active, the process will not block when waiting for new events.
653 666
654Apart from keeping your process non-blocking (which is a useful 667Apart from keeping your process non-blocking (which is a useful
… …
664kind. There is a C<ev_idle_set> macro, but using it is utterly pointless, 677kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
665believe me. 678believe me.
666 679
667=back 680=back
668 681
669=head2 prepare and check - your hooks into the event loop 682=head2 C<ev_prepare> and C<ev_check> - customise your event loop
670 683
671Prepare and check watchers usually (but not always) are used in 684Prepare and check watchers are usually (but not always) used in tandem:
672tandom. Prepare watchers get invoked before the process blocks and check 685prepare watchers get invoked before the process blocks and check watchers
673watchers afterwards. 686afterwards.
674 687
675Their main purpose is to integrate other event mechanisms into libev. This 688Their main purpose is to integrate other event mechanisms into libev. This
676could be used, for example, to track variable changes, implement your own 689could be used, for example, to track variable changes, implement your own
677watchers, integrate net-snmp or a coroutine library and lots more. 690watchers, integrate net-snmp or a coroutine library and lots more.
678 691
679This is done by examining in each prepare call which file descriptors need 692This is done by examining in each prepare call which file descriptors need
680to be watched by the other library, registering C<ev_io> watchers for them 693to be watched by the other library, registering C<ev_io> watchers for
681and starting an C<ev_timer> watcher for any timeouts (many libraries provide 694them and starting an C<ev_timer> watcher for any timeouts (many libraries
682just this functionality). Then, in the check watcher you check for any 695provide just this functionality). Then, in the check watcher you check for
683events that occured (by making your callbacks set soem flags for example) 696any events that occured (by checking the pending status of all watchers
684and call back into the library. 697and stopping them) and call back into the library. The I/O and timer
698callbacks will never actually be called (but must be valid nevertheless,
699because you never know, you know?).
685 700
686As another example, the perl Coro module uses these hooks to integrate 701As another example, the Perl Coro module uses these hooks to integrate
687coroutines into libev programs, by yielding to other active coroutines 702coroutines into libev programs, by yielding to other active coroutines
688during each prepare and only letting the process block if no coroutines 703during each prepare and only letting the process block if no coroutines
689are ready to run. 704are ready to run (it's actually more complicated: it only runs coroutines
705with priority higher than or equal to the event loop and one coroutine
706of lower priority, but only once, using idle watchers to keep the event
707loop from blocking if lower-priority coroutines are active, thus mapping
708low-priority coroutines to idle/background tasks).
690 709
691=over 4 710=over 4
692 711
693=item ev_prepare_init (ev_prepare *, callback) 712=item ev_prepare_init (ev_prepare *, callback)
694 713
695=item ev_check_init (ev_check *, callback) 714=item ev_check_init (ev_check *, callback)
696 715
697Initialises and configures the prepare or check watcher - they have no 716Initialises and configures the prepare or check watcher - they have no
698parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set> 717parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
699macros, but using them is utterly, utterly pointless. 718macros, but using them is utterly, utterly and completely pointless.
700 719
701=back 720=back
702 721
703=head1 OTHER FUNCTIONS 722=head1 OTHER FUNCTIONS
704 723
705There are some other fucntions of possible interest. Described. Here. Now. 724There are some other functions of possible interest. Described. Here. Now.
706 725
707=over 4 726=over 4
708 727
709=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback) 728=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
710 729
711This function combines a simple timer and an I/O watcher, calls your 730This function combines a simple timer and an I/O watcher, calls your
712callback on whichever event happens first and automatically stop both 731callback on whichever event happens first and automatically stop both
713watchers. This is useful if you want to wait for a single event on an fd 732watchers. This is useful if you want to wait for a single event on an fd
714or timeout without havign to allocate/configure/start/stop/free one or 733or timeout without having to allocate/configure/start/stop/free one or
715more watchers yourself. 734more watchers yourself.
716 735
717If C<fd> is less than 0, then no I/O watcher will be started and events is 736If C<fd> is less than 0, then no I/O watcher will be started and events
718ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and C<events> set 737is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and
719will be craeted and started. 738C<events> set will be craeted and started.
720 739
721If C<timeout> is less than 0, then no timeout watcher will be 740If C<timeout> is less than 0, then no timeout watcher will be
722started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and repeat 741started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
723= 0) will be started. 742repeat = 0) will be started. While C<0> is a valid timeout, it is of
743dubious value.
724 744
725The callback has the type C<void (*cb)(int revents, void *arg)> and 745The callback has the type C<void (*cb)(int revents, void *arg)> and gets
726gets passed an events set (normally a combination of C<EV_ERROR>, C<EV_READ>, 746passed an C<revents> set like normal event callbacks (a combination of
727C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg> value passed to C<ev_once>: 747C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
748value passed to C<ev_once>:
728 749
729 static void stdin_ready (int revents, void *arg) 750 static void stdin_ready (int revents, void *arg)
730 { 751 {
731 if (revents & EV_TIMEOUT) 752 if (revents & EV_TIMEOUT)
732 /* doh, nothing entered */ 753 /* doh, nothing entered */;
733 else if (revents & EV_READ) 754 else if (revents & EV_READ)
734 /* stdin might have data for us, joy! */ 755 /* stdin might have data for us, joy! */;
735 } 756 }
736 757
737 ev_once (STDIN_FILENO, EV_READm 10., stdin_ready, 0); 758 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
738 759
739=item ev_feed_event (loop, watcher, int events) 760=item ev_feed_event (loop, watcher, int events)
740 761
741Feeds the given event set into the event loop, as if the specified event 762Feeds the given event set into the event loop, as if the specified event
742has happened for the specified watcher (which must be a pointer to an 763had happened for the specified watcher (which must be a pointer to an
743initialised but not necessarily active event watcher). 764initialised but not necessarily started event watcher).
744 765
745=item ev_feed_fd_event (loop, int fd, int revents) 766=item ev_feed_fd_event (loop, int fd, int revents)
746 767
747Feed an event on the given fd, as if a file descriptor backend detected it. 768Feed an event on the given fd, as if a file descriptor backend detected
769the given events it.
748 770
749=item ev_feed_signal_event (loop, int signum) 771=item ev_feed_signal_event (loop, int signum)
750 772
751Feed an event as if the given signal occured (loop must be the default loop!). 773Feed an event as if the given signal occured (loop must be the default loop!).
752 774
753=back 775=back
754 776
777=head1 LIBEVENT EMULATION
778
779Libev offers a compatibility emulation layer for libevent. It cannot
780emulate the internals of libevent, so here are some usage hints:
781
782=over 4
783
784=item * Use it by including <event.h>, as usual.
785
786=item * The following members are fully supported: ev_base, ev_callback,
787ev_arg, ev_fd, ev_res, ev_events.
788
789=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
790maintained by libev, it does not work exactly the same way as in libevent (consider
791it a private API).
792
793=item * Priorities are not currently supported. Initialising priorities
794will fail and all watchers will have the same priority, even though there
795is an ev_pri field.
796
797=item * Other members are not supported.
798
799=item * The libev emulation is I<not> ABI compatible to libevent, you need
800to use the libev header file and library.
801
802=back
803
804=head1 C++ SUPPORT
805
806TBD.
807
755=head1 AUTHOR 808=head1 AUTHOR
756 809
757Marc Lehmann <libev@schmorp.de>. 810Marc Lehmann <libev@schmorp.de>.
758 811

Diff Legend

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