ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent/lib/AnyEvent.pm
(Generate patch)

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.179 by root, Thu Sep 4 10:58:58 2008 UTC vs.
Revision 1.365 by root, Tue Aug 16 14:47:26 2011 UTC

1=head1 NAME 1=head1 NAME
2 2
3AnyEvent - provide framework for multiple event loops 3AnyEvent - the DBI of event loop programming
4 4
5EV, Event, Glib, Tk, Perl, Event::Lib, Qt, POE - various supported event loops 5EV, Event, Glib, Tk, Perl, Event::Lib, Irssi, rxvt-unicode, IO::Async, Qt
6and POE are various supported event loops/environments.
6 7
7=head1 SYNOPSIS 8=head1 SYNOPSIS
8 9
9 use AnyEvent; 10 use AnyEvent;
10 11
12 # if you prefer function calls, look at the AE manpage for
13 # an alternative API.
14
15 # file handle or descriptor readable
11 my $w = AnyEvent->io (fh => $fh, poll => "r|w", cb => sub { ... }); 16 my $w = AnyEvent->io (fh => $fh, poll => "r", cb => sub { ... });
12 17
18 # one-shot or repeating timers
13 my $w = AnyEvent->timer (after => $seconds, cb => sub { ... }); 19 my $w = AnyEvent->timer (after => $seconds, cb => sub { ... });
14 my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ... 20 my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ...);
15 21
16 print AnyEvent->now; # prints current event loop time 22 print AnyEvent->now; # prints current event loop time
17 print AnyEvent->time; # think Time::HiRes::time or simply CORE::time. 23 print AnyEvent->time; # think Time::HiRes::time or simply CORE::time.
18 24
25 # POSIX signal
19 my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... }); 26 my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... });
20 27
28 # child process exit
21 my $w = AnyEvent->child (pid => $pid, cb => sub { 29 my $w = AnyEvent->child (pid => $pid, cb => sub {
22 my ($pid, $status) = @_; 30 my ($pid, $status) = @_;
23 ... 31 ...
24 }); 32 });
33
34 # called when event loop idle (if applicable)
35 my $w = AnyEvent->idle (cb => sub { ... });
25 36
26 my $w = AnyEvent->condvar; # stores whether a condition was flagged 37 my $w = AnyEvent->condvar; # stores whether a condition was flagged
27 $w->send; # wake up current and all future recv's 38 $w->send; # wake up current and all future recv's
28 $w->recv; # enters "main loop" till $condvar gets ->send 39 $w->recv; # enters "main loop" till $condvar gets ->send
29 # use a condvar in callback mode: 40 # use a condvar in callback mode:
32=head1 INTRODUCTION/TUTORIAL 43=head1 INTRODUCTION/TUTORIAL
33 44
34This manpage is mainly a reference manual. If you are interested 45This manpage is mainly a reference manual. If you are interested
35in a tutorial or some gentle introduction, have a look at the 46in a tutorial or some gentle introduction, have a look at the
36L<AnyEvent::Intro> manpage. 47L<AnyEvent::Intro> manpage.
48
49=head1 SUPPORT
50
51An FAQ document is available as L<AnyEvent::FAQ>.
52
53There also is a mailinglist for discussing all things AnyEvent, and an IRC
54channel, too.
55
56See the AnyEvent project page at the B<Schmorpforge Ta-Sa Software
57Repository>, at L<http://anyevent.schmorp.de>, for more info.
37 58
38=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT) 59=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
39 60
40Glib, POE, IO::Async, Event... CPAN offers event models by the dozen 61Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
41nowadays. So what is different about AnyEvent? 62nowadays. So what is different about AnyEvent?
57module users into the same thing by forcing them to use the same event 78module users into the same thing by forcing them to use the same event
58model you use. 79model you use.
59 80
60For modules like POE or IO::Async (which is a total misnomer as it is 81For modules like POE or IO::Async (which is a total misnomer as it is
61actually doing all I/O I<synchronously>...), using them in your module is 82actually doing all I/O I<synchronously>...), using them in your module is
62like joining a cult: After you joined, you are dependent on them and you 83like joining a cult: After you join, you are dependent on them and you
63cannot use anything else, as they are simply incompatible to everything 84cannot use anything else, as they are simply incompatible to everything
64that isn't them. What's worse, all the potential users of your 85that isn't them. What's worse, all the potential users of your
65module are I<also> forced to use the same event loop you use. 86module are I<also> forced to use the same event loop you use.
66 87
67AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works 88AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
68fine. AnyEvent + Tk works fine etc. etc. but none of these work together 89fine. AnyEvent + Tk works fine etc. etc. but none of these work together
69with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if 90with the rest: POE + EV? No go. Tk + Event? No go. Again: if your module
70your module uses one of those, every user of your module has to use it, 91uses one of those, every user of your module has to use it, too. But if
71too. But if your module uses AnyEvent, it works transparently with all 92your module uses AnyEvent, it works transparently with all event models it
72event models it supports (including stuff like IO::Async, as long as those 93supports (including stuff like IO::Async, as long as those use one of the
73use one of the supported event loops. It is trivial to add new event loops 94supported event loops. It is easy to add new event loops to AnyEvent, too,
74to AnyEvent, too, so it is future-proof). 95so it is future-proof).
75 96
76In addition to being free of having to use I<the one and only true event 97In addition to being free of having to use I<the one and only true event
77model>, AnyEvent also is free of bloat and policy: with POE or similar 98model>, AnyEvent also is free of bloat and policy: with POE or similar
78modules, you get an enormous amount of code and strict rules you have to 99modules, you get an enormous amount of code and strict rules you have to
79follow. AnyEvent, on the other hand, is lean and up to the point, by only 100follow. AnyEvent, on the other hand, is lean and to the point, by only
80offering the functionality that is necessary, in as thin as a wrapper as 101offering the functionality that is necessary, in as thin as a wrapper as
81technically possible. 102technically possible.
82 103
83Of course, AnyEvent comes with a big (and fully optional!) toolbox 104Of course, AnyEvent comes with a big (and fully optional!) toolbox
84of useful functionality, such as an asynchronous DNS resolver, 100% 105of useful functionality, such as an asynchronous DNS resolver, 100%
90useful) and you want to force your users to use the one and only event 111useful) and you want to force your users to use the one and only event
91model, you should I<not> use this module. 112model, you should I<not> use this module.
92 113
93=head1 DESCRIPTION 114=head1 DESCRIPTION
94 115
95L<AnyEvent> provides an identical interface to multiple event loops. This 116L<AnyEvent> provides a uniform interface to various event loops. This
96allows module authors to utilise an event loop without forcing module 117allows module authors to use event loop functionality without forcing
97users to use the same event loop (as only a single event loop can coexist 118module users to use a specific event loop implementation (since more
98peacefully at any one time). 119than one event loop cannot coexist peacefully).
99 120
100The interface itself is vaguely similar, but not identical to the L<Event> 121The interface itself is vaguely similar, but not identical to the L<Event>
101module. 122module.
102 123
103During the first call of any watcher-creation method, the module tries 124During the first call of any watcher-creation method, the module tries
104to detect the currently loaded event loop by probing whether one of the 125to detect the currently loaded event loop by probing whether one of the
105following modules is already loaded: L<EV>, 126following modules is already loaded: L<EV>, L<AnyEvent::Loop>,
106L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>, 127L<Event>, L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>. The first one
107L<POE>. The first one found is used. If none are found, the module tries 128found is used. If none are detected, the module tries to load the first
108to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl 129four modules in the order given; but note that if L<EV> is not
109adaptor should always succeed) in the order given. The first one that can 130available, the pure-perl L<AnyEvent::Loop> should always work, so
110be successfully loaded will be used. If, after this, still none could be 131the other two are not normally tried.
111found, AnyEvent will fall back to a pure-perl event loop, which is not
112very efficient, but should work everywhere.
113 132
114Because AnyEvent first checks for modules that are already loaded, loading 133Because AnyEvent first checks for modules that are already loaded, loading
115an event model explicitly before first using AnyEvent will likely make 134an event model explicitly before first using AnyEvent will likely make
116that model the default. For example: 135that model the default. For example:
117 136
119 use AnyEvent; 138 use AnyEvent;
120 139
121 # .. AnyEvent will likely default to Tk 140 # .. AnyEvent will likely default to Tk
122 141
123The I<likely> means that, if any module loads another event model and 142The I<likely> means that, if any module loads another event model and
124starts using it, all bets are off. Maybe you should tell their authors to 143starts using it, all bets are off - this case should be very rare though,
125use AnyEvent so their modules work together with others seamlessly... 144as very few modules hardcode event loops without announcing this very
145loudly.
126 146
127The pure-perl implementation of AnyEvent is called 147The pure-perl implementation of AnyEvent is called C<AnyEvent::Loop>. Like
128C<AnyEvent::Impl::Perl>. Like other event modules you can load it 148other event modules you can load it explicitly and enjoy the high
129explicitly and enjoy the high availability of that event loop :) 149availability of that event loop :)
130 150
131=head1 WATCHERS 151=head1 WATCHERS
132 152
133AnyEvent has the central concept of a I<watcher>, which is an object that 153AnyEvent has the central concept of a I<watcher>, which is an object that
134stores relevant data for each kind of event you are waiting for, such as 154stores relevant data for each kind of event you are waiting for, such as
137These watchers are normal Perl objects with normal Perl lifetime. After 157These watchers are normal Perl objects with normal Perl lifetime. After
138creating a watcher it will immediately "watch" for events and invoke the 158creating a watcher it will immediately "watch" for events and invoke the
139callback when the event occurs (of course, only when the event model 159callback when the event occurs (of course, only when the event model
140is in control). 160is in control).
141 161
162Note that B<callbacks must not permanently change global variables>
163potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
164callbacks must not C<die> >>. The former is good programming practice in
165Perl and the latter stems from the fact that exception handling differs
166widely between event loops.
167
142To disable the watcher you have to destroy it (e.g. by setting the 168To disable a watcher you have to destroy it (e.g. by setting the
143variable you store it in to C<undef> or otherwise deleting all references 169variable you store it in to C<undef> or otherwise deleting all references
144to it). 170to it).
145 171
146All watchers are created by calling a method on the C<AnyEvent> class. 172All watchers are created by calling a method on the C<AnyEvent> class.
147 173
148Many watchers either are used with "recursion" (repeating timers for 174Many watchers either are used with "recursion" (repeating timers for
149example), or need to refer to their watcher object in other ways. 175example), or need to refer to their watcher object in other ways.
150 176
151An any way to achieve that is this pattern: 177One way to achieve that is this pattern:
152 178
153 my $w; $w = AnyEvent->type (arg => value ..., cb => sub { 179 my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
154 # you can use $w here, for example to undef it 180 # you can use $w here, for example to undef it
155 undef $w; 181 undef $w;
156 }); 182 });
159my variables are only visible after the statement in which they are 185my variables are only visible after the statement in which they are
160declared. 186declared.
161 187
162=head2 I/O WATCHERS 188=head2 I/O WATCHERS
163 189
190 $w = AnyEvent->io (
191 fh => <filehandle_or_fileno>,
192 poll => <"r" or "w">,
193 cb => <callback>,
194 );
195
164You can create an I/O watcher by calling the C<< AnyEvent->io >> method 196You can create an I/O watcher by calling the C<< AnyEvent->io >> method
165with the following mandatory key-value pairs as arguments: 197with the following mandatory key-value pairs as arguments:
166 198
167C<fh> the Perl I<file handle> (I<not> file descriptor) to watch for events 199C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch
168(AnyEvent might or might not keep a reference to this file handle). C<poll> 200for events (AnyEvent might or might not keep a reference to this file
201handle). Note that only file handles pointing to things for which
202non-blocking operation makes sense are allowed. This includes sockets,
203most character devices, pipes, fifos and so on, but not for example files
204or block devices.
205
169must be a string that is either C<r> or C<w>, which creates a watcher 206C<poll> must be a string that is either C<r> or C<w>, which creates a
170waiting for "r"eadable or "w"ritable events, respectively. C<cb> is the 207watcher waiting for "r"eadable or "w"ritable events, respectively.
208
171callback to invoke each time the file handle becomes ready. 209C<cb> is the callback to invoke each time the file handle becomes ready.
172 210
173Although the callback might get passed parameters, their value and 211Although the callback might get passed parameters, their value and
174presence is undefined and you cannot rely on them. Portable AnyEvent 212presence is undefined and you cannot rely on them. Portable AnyEvent
175callbacks cannot use arguments passed to I/O watcher callbacks. 213callbacks cannot use arguments passed to I/O watcher callbacks.
176 214
177The I/O watcher might use the underlying file descriptor or a copy of it. 215The I/O watcher might use the underlying file descriptor or a copy of it.
178You must not close a file handle as long as any watcher is active on the 216You must not close a file handle as long as any watcher is active on the
179underlying file descriptor. 217underlying file descriptor.
180 218
181Some event loops issue spurious readyness notifications, so you should 219Some event loops issue spurious readiness notifications, so you should
182always use non-blocking calls when reading/writing from/to your file 220always use non-blocking calls when reading/writing from/to your file
183handles. 221handles.
184 222
185Example: wait for readability of STDIN, then read a line and disable the 223Example: wait for readability of STDIN, then read a line and disable the
186watcher. 224watcher.
191 undef $w; 229 undef $w;
192 }); 230 });
193 231
194=head2 TIME WATCHERS 232=head2 TIME WATCHERS
195 233
234 $w = AnyEvent->timer (after => <seconds>, cb => <callback>);
235
236 $w = AnyEvent->timer (
237 after => <fractional_seconds>,
238 interval => <fractional_seconds>,
239 cb => <callback>,
240 );
241
196You can create a time watcher by calling the C<< AnyEvent->timer >> 242You can create a time watcher by calling the C<< AnyEvent->timer >>
197method with the following mandatory arguments: 243method with the following mandatory arguments:
198 244
199C<after> specifies after how many seconds (fractional values are 245C<after> specifies after how many seconds (fractional values are
200supported) the callback should be invoked. C<cb> is the callback to invoke 246supported) the callback should be invoked. C<cb> is the callback to invoke
202 248
203Although the callback might get passed parameters, their value and 249Although the callback might get passed parameters, their value and
204presence is undefined and you cannot rely on them. Portable AnyEvent 250presence is undefined and you cannot rely on them. Portable AnyEvent
205callbacks cannot use arguments passed to time watcher callbacks. 251callbacks cannot use arguments passed to time watcher callbacks.
206 252
207The callback will normally be invoked once only. If you specify another 253The callback will normally be invoked only once. If you specify another
208parameter, C<interval>, as a strictly positive number (> 0), then the 254parameter, C<interval>, as a strictly positive number (> 0), then the
209callback will be invoked regularly at that interval (in fractional 255callback will be invoked regularly at that interval (in fractional
210seconds) after the first invocation. If C<interval> is specified with a 256seconds) after the first invocation. If C<interval> is specified with a
211false value, then it is treated as if it were missing. 257false value, then it is treated as if it were not specified at all.
212 258
213The callback will be rescheduled before invoking the callback, but no 259The callback will be rescheduled before invoking the callback, but no
214attempt is done to avoid timer drift in most backends, so the interval is 260attempt is made to avoid timer drift in most backends, so the interval is
215only approximate. 261only approximate.
216 262
217Example: fire an event after 7.7 seconds. 263Example: fire an event after 7.7 seconds.
218 264
219 my $w = AnyEvent->timer (after => 7.7, cb => sub { 265 my $w = AnyEvent->timer (after => 7.7, cb => sub {
237 283
238While most event loops expect timers to specified in a relative way, they 284While most event loops expect timers to specified in a relative way, they
239use absolute time internally. This makes a difference when your clock 285use absolute time internally. This makes a difference when your clock
240"jumps", for example, when ntp decides to set your clock backwards from 286"jumps", for example, when ntp decides to set your clock backwards from
241the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to 287the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
242fire "after" a second might actually take six years to finally fire. 288fire "after a second" might actually take six years to finally fire.
243 289
244AnyEvent cannot compensate for this. The only event loop that is conscious 290AnyEvent cannot compensate for this. The only event loop that is conscious
245about these issues is L<EV>, which offers both relative (ev_timer, based 291of these issues is L<EV>, which offers both relative (ev_timer, based
246on true relative time) and absolute (ev_periodic, based on wallclock time) 292on true relative time) and absolute (ev_periodic, based on wallclock time)
247timers. 293timers.
248 294
249AnyEvent always prefers relative timers, if available, matching the 295AnyEvent always prefers relative timers, if available, matching the
250AnyEvent API. 296AnyEvent API.
272I<In almost all cases (in all cases if you don't care), this is the 318I<In almost all cases (in all cases if you don't care), this is the
273function to call when you want to know the current time.> 319function to call when you want to know the current time.>
274 320
275This function is also often faster then C<< AnyEvent->time >>, and 321This function is also often faster then C<< AnyEvent->time >>, and
276thus the preferred method if you want some timestamp (for example, 322thus the preferred method if you want some timestamp (for example,
277L<AnyEvent::Handle> uses this to update it's activity timeouts). 323L<AnyEvent::Handle> uses this to update its activity timeouts).
278 324
279The rest of this section is only of relevance if you try to be very exact 325The rest of this section is only of relevance if you try to be very exact
280with your timing, you can skip it without bad conscience. 326with your timing; you can skip it without a bad conscience.
281 327
282For a practical example of when these times differ, consider L<Event::Lib> 328For a practical example of when these times differ, consider L<Event::Lib>
283and L<EV> and the following set-up: 329and L<EV> and the following set-up:
284 330
285The event loop is running and has just invoked one of your callback at 331The event loop is running and has just invoked one of your callbacks at
286time=500 (assume no other callbacks delay processing). In your callback, 332time=500 (assume no other callbacks delay processing). In your callback,
287you wait a second by executing C<sleep 1> (blocking the process for a 333you wait a second by executing C<sleep 1> (blocking the process for a
288second) and then (at time=501) you create a relative timer that fires 334second) and then (at time=501) you create a relative timer that fires
289after three seconds. 335after three seconds.
290 336
308In either case, if you care (and in most cases, you don't), then you 354In either case, if you care (and in most cases, you don't), then you
309can get whatever behaviour you want with any event loop, by taking the 355can get whatever behaviour you want with any event loop, by taking the
310difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into 356difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
311account. 357account.
312 358
359=item AnyEvent->now_update
360
361Some event loops (such as L<EV> or L<AnyEvent::Loop>) cache the current
362time for each loop iteration (see the discussion of L<< AnyEvent->now >>,
363above).
364
365When a callback runs for a long time (or when the process sleeps), then
366this "current" time will differ substantially from the real time, which
367might affect timers and time-outs.
368
369When this is the case, you can call this method, which will update the
370event loop's idea of "current time".
371
372A typical example would be a script in a web server (e.g. C<mod_perl>) -
373when mod_perl executes the script, then the event loop will have the wrong
374idea about the "current time" (being potentially far in the past, when the
375script ran the last time). In that case you should arrange a call to C<<
376AnyEvent->now_update >> each time the web server process wakes up again
377(e.g. at the start of your script, or in a handler).
378
379Note that updating the time I<might> cause some events to be handled.
380
313=back 381=back
314 382
315=head2 SIGNAL WATCHERS 383=head2 SIGNAL WATCHERS
384
385 $w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>);
316 386
317You can watch for signals using a signal watcher, C<signal> is the signal 387You can watch for signals using a signal watcher, C<signal> is the signal
318I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl 388I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
319callback to be invoked whenever a signal occurs. 389callback to be invoked whenever a signal occurs.
320 390
326invocation, and callback invocation will be synchronous. Synchronous means 396invocation, and callback invocation will be synchronous. Synchronous means
327that it might take a while until the signal gets handled by the process, 397that it might take a while until the signal gets handled by the process,
328but it is guaranteed not to interrupt any other callbacks. 398but it is guaranteed not to interrupt any other callbacks.
329 399
330The main advantage of using these watchers is that you can share a signal 400The main advantage of using these watchers is that you can share a signal
331between multiple watchers. 401between multiple watchers, and AnyEvent will ensure that signals will not
402interrupt your program at bad times.
332 403
333This watcher might use C<%SIG>, so programs overwriting those signals 404This watcher might use C<%SIG> (depending on the event loop used),
334directly will likely not work correctly. 405so programs overwriting those signals directly will likely not work
406correctly.
335 407
336Example: exit on SIGINT 408Example: exit on SIGINT
337 409
338 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); 410 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
339 411
412=head3 Restart Behaviour
413
414While restart behaviour is up to the event loop implementation, most will
415not restart syscalls (that includes L<Async::Interrupt> and AnyEvent's
416pure perl implementation).
417
418=head3 Safe/Unsafe Signals
419
420Perl signals can be either "safe" (synchronous to opcode handling) or
421"unsafe" (asynchronous) - the former might get delayed indefinitely, the
422latter might corrupt your memory.
423
424AnyEvent signal handlers are, in addition, synchronous to the event loop,
425i.e. they will not interrupt your running perl program but will only be
426called as part of the normal event handling (just like timer, I/O etc.
427callbacks, too).
428
429=head3 Signal Races, Delays and Workarounds
430
431Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
432callbacks to signals in a generic way, which is a pity, as you cannot
433do race-free signal handling in perl, requiring C libraries for
434this. AnyEvent will try to do its best, which means in some cases,
435signals will be delayed. The maximum time a signal might be delayed is
436specified in C<$AnyEvent::MAX_SIGNAL_LATENCY> (default: 10 seconds). This
437variable can be changed only before the first signal watcher is created,
438and should be left alone otherwise. This variable determines how often
439AnyEvent polls for signals (in case a wake-up was missed). Higher values
440will cause fewer spurious wake-ups, which is better for power and CPU
441saving.
442
443All these problems can be avoided by installing the optional
444L<Async::Interrupt> module, which works with most event loops. It will not
445work with inherently broken event loops such as L<Event> or L<Event::Lib>
446(and not with L<POE> currently, as POE does its own workaround with
447one-second latency). For those, you just have to suffer the delays.
448
340=head2 CHILD PROCESS WATCHERS 449=head2 CHILD PROCESS WATCHERS
341 450
451 $w = AnyEvent->child (pid => <process id>, cb => <callback>);
452
342You can also watch on a child process exit and catch its exit status. 453You can also watch for a child process exit and catch its exit status.
343 454
344The child process is specified by the C<pid> argument (if set to C<0>, it 455The child process is specified by the C<pid> argument (on some backends,
345watches for any child process exit). The watcher will trigger as often 456using C<0> watches for any child process exit, on others this will
346as status change for the child are received. This works by installing a 457croak). The watcher will be triggered only when the child process has
347signal handler for C<SIGCHLD>. The callback will be called with the pid 458finished and an exit status is available, not on any trace events
348and exit status (as returned by waitpid), so unlike other watcher types, 459(stopped/continued).
349you I<can> rely on child watcher callback arguments. 460
461The callback will be called with the pid and exit status (as returned by
462waitpid), so unlike other watcher types, you I<can> rely on child watcher
463callback arguments.
464
465This watcher type works by installing a signal handler for C<SIGCHLD>,
466and since it cannot be shared, nothing else should use SIGCHLD or reap
467random child processes (waiting for specific child processes, e.g. inside
468C<system>, is just fine).
350 469
351There is a slight catch to child watchers, however: you usually start them 470There is a slight catch to child watchers, however: you usually start them
352I<after> the child process was created, and this means the process could 471I<after> the child process was created, and this means the process could
353have exited already (and no SIGCHLD will be sent anymore). 472have exited already (and no SIGCHLD will be sent anymore).
354 473
355Not all event models handle this correctly (POE doesn't), but even for 474Not all event models handle this correctly (neither POE nor IO::Async do,
475see their AnyEvent::Impl manpages for details), but even for event models
356event models that I<do> handle this correctly, they usually need to be 476that I<do> handle this correctly, they usually need to be loaded before
357loaded before the process exits (i.e. before you fork in the first place). 477the process exits (i.e. before you fork in the first place). AnyEvent's
478pure perl event loop handles all cases correctly regardless of when you
479start the watcher.
358 480
359This means you cannot create a child watcher as the very first thing in an 481This means you cannot create a child watcher as the very first
360AnyEvent program, you I<have> to create at least one watcher before you 482thing in an AnyEvent program, you I<have> to create at least one
361C<fork> the child (alternatively, you can call C<AnyEvent::detect>). 483watcher before you C<fork> the child (alternatively, you can call
484C<AnyEvent::detect>).
485
486As most event loops do not support waiting for child events, they will be
487emulated by AnyEvent in most cases, in which case the latency and race
488problems mentioned in the description of signal watchers apply.
362 489
363Example: fork a process and wait for it 490Example: fork a process and wait for it
364 491
365 my $done = AnyEvent->condvar; 492 my $done = AnyEvent->condvar;
366 493
376 ); 503 );
377 504
378 # do something else, then wait for process exit 505 # do something else, then wait for process exit
379 $done->recv; 506 $done->recv;
380 507
508=head2 IDLE WATCHERS
509
510 $w = AnyEvent->idle (cb => <callback>);
511
512This will repeatedly invoke the callback after the process becomes idle,
513until either the watcher is destroyed or new events have been detected.
514
515Idle watchers are useful when there is a need to do something, but it
516is not so important (or wise) to do it instantly. The callback will be
517invoked only when there is "nothing better to do", which is usually
518defined as "all outstanding events have been handled and no new events
519have been detected". That means that idle watchers ideally get invoked
520when the event loop has just polled for new events but none have been
521detected. Instead of blocking to wait for more events, the idle watchers
522will be invoked.
523
524Unfortunately, most event loops do not really support idle watchers (only
525EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
526will simply call the callback "from time to time".
527
528Example: read lines from STDIN, but only process them when the
529program is otherwise idle:
530
531 my @lines; # read data
532 my $idle_w;
533 my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
534 push @lines, scalar <STDIN>;
535
536 # start an idle watcher, if not already done
537 $idle_w ||= AnyEvent->idle (cb => sub {
538 # handle only one line, when there are lines left
539 if (my $line = shift @lines) {
540 print "handled when idle: $line";
541 } else {
542 # otherwise disable the idle watcher again
543 undef $idle_w;
544 }
545 });
546 });
547
381=head2 CONDITION VARIABLES 548=head2 CONDITION VARIABLES
549
550 $cv = AnyEvent->condvar;
551
552 $cv->send (<list>);
553 my @res = $cv->recv;
382 554
383If you are familiar with some event loops you will know that all of them 555If you are familiar with some event loops you will know that all of them
384require you to run some blocking "loop", "run" or similar function that 556require you to run some blocking "loop", "run" or similar function that
385will actively watch for new events and call your callbacks. 557will actively watch for new events and call your callbacks.
386 558
387AnyEvent is different, it expects somebody else to run the event loop and 559AnyEvent is slightly different: it expects somebody else to run the event
388will only block when necessary (usually when told by the user). 560loop and will only block when necessary (usually when told by the user).
389 561
390The instrument to do that is called a "condition variable", so called 562The tool to do that is called a "condition variable", so called because
391because they represent a condition that must become true. 563they represent a condition that must become true.
564
565Now is probably a good time to look at the examples further below.
392 566
393Condition variables can be created by calling the C<< AnyEvent->condvar 567Condition variables can be created by calling the C<< AnyEvent->condvar
394>> method, usually without arguments. The only argument pair allowed is 568>> method, usually without arguments. The only argument pair allowed is
395
396C<cb>, which specifies a callback to be called when the condition variable 569C<cb>, which specifies a callback to be called when the condition variable
397becomes true, with the condition variable as the first argument (but not 570becomes true, with the condition variable as the first argument (but not
398the results). 571the results).
399 572
400After creation, the condition variable is "false" until it becomes "true" 573After creation, the condition variable is "false" until it becomes "true"
401by calling the C<send> method (or calling the condition variable as if it 574by calling the C<send> method (or calling the condition variable as if it
402were a callback, read about the caveats in the description for the C<< 575were a callback, read about the caveats in the description for the C<<
403->send >> method). 576->send >> method).
404 577
405Condition variables are similar to callbacks, except that you can 578Since condition variables are the most complex part of the AnyEvent API, here are
406optionally wait for them. They can also be called merge points - points 579some different mental models of what they are - pick the ones you can connect to:
407in time where multiple outstanding events have been processed. And yet 580
408another way to call them is transactions - each condition variable can be 581=over 4
409used to represent a transaction, which finishes at some point and delivers 582
410a result. 583=item * Condition variables are like callbacks - you can call them (and pass them instead
584of callbacks). Unlike callbacks however, you can also wait for them to be called.
585
586=item * Condition variables are signals - one side can emit or send them,
587the other side can wait for them, or install a handler that is called when
588the signal fires.
589
590=item * Condition variables are like "Merge Points" - points in your program
591where you merge multiple independent results/control flows into one.
592
593=item * Condition variables represent a transaction - functions that start
594some kind of transaction can return them, leaving the caller the choice
595between waiting in a blocking fashion, or setting a callback.
596
597=item * Condition variables represent future values, or promises to deliver
598some result, long before the result is available.
599
600=back
411 601
412Condition variables are very useful to signal that something has finished, 602Condition variables are very useful to signal that something has finished,
413for example, if you write a module that does asynchronous http requests, 603for example, if you write a module that does asynchronous http requests,
414then a condition variable would be the ideal candidate to signal the 604then a condition variable would be the ideal candidate to signal the
415availability of results. The user can either act when the callback is 605availability of results. The user can either act when the callback is
428 618
429Condition variables are represented by hash refs in perl, and the keys 619Condition variables are represented by hash refs in perl, and the keys
430used by AnyEvent itself are all named C<_ae_XXX> to make subclassing 620used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
431easy (it is often useful to build your own transaction class on top of 621easy (it is often useful to build your own transaction class on top of
432AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call 622AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
433it's C<new> method in your own C<new> method. 623its C<new> method in your own C<new> method.
434 624
435There are two "sides" to a condition variable - the "producer side" which 625There are two "sides" to a condition variable - the "producer side" which
436eventually calls C<< -> send >>, and the "consumer side", which waits 626eventually calls C<< -> send >>, and the "consumer side", which waits
437for the send to occur. 627for the send to occur.
438 628
439Example: wait for a timer. 629Example: wait for a timer.
440 630
441 # wait till the result is ready 631 # condition: "wait till the timer is fired"
442 my $result_ready = AnyEvent->condvar; 632 my $timer_fired = AnyEvent->condvar;
443 633
444 # do something such as adding a timer 634 # create the timer - we could wait for, say
445 # or socket watcher the calls $result_ready->send 635 # a handle becomign ready, or even an
446 # when the "result" is ready. 636 # AnyEvent::HTTP request to finish, but
447 # in this case, we simply use a timer: 637 # in this case, we simply use a timer:
448 my $w = AnyEvent->timer ( 638 my $w = AnyEvent->timer (
449 after => 1, 639 after => 1,
450 cb => sub { $result_ready->send }, 640 cb => sub { $timer_fired->send },
451 ); 641 );
452 642
453 # this "blocks" (while handling events) till the callback 643 # this "blocks" (while handling events) till the callback
454 # calls send 644 # calls ->send
455 $result_ready->recv; 645 $timer_fired->recv;
456 646
457Example: wait for a timer, but take advantage of the fact that 647Example: wait for a timer, but take advantage of the fact that condition
458condition variables are also code references. 648variables are also callable directly.
459 649
460 my $done = AnyEvent->condvar; 650 my $done = AnyEvent->condvar;
461 my $delay = AnyEvent->timer (after => 5, cb => $done); 651 my $delay = AnyEvent->timer (after => 5, cb => $done);
462 $done->recv; 652 $done->recv;
463 653
469 659
470 ... 660 ...
471 661
472 my @info = $couchdb->info->recv; 662 my @info = $couchdb->info->recv;
473 663
474And this is how you would just ste a callback to be called whenever the 664And this is how you would just set a callback to be called whenever the
475results are available: 665results are available:
476 666
477 $couchdb->info->cb (sub { 667 $couchdb->info->cb (sub {
478 my @info = $_[0]->recv; 668 my @info = $_[0]->recv;
479 }); 669 });
497immediately from within send. 687immediately from within send.
498 688
499Any arguments passed to the C<send> call will be returned by all 689Any arguments passed to the C<send> call will be returned by all
500future C<< ->recv >> calls. 690future C<< ->recv >> calls.
501 691
502Condition variables are overloaded so one can call them directly 692Condition variables are overloaded so one can call them directly (as if
503(as a code reference). Calling them directly is the same as calling 693they were a code reference). Calling them directly is the same as calling
504C<send>. Note, however, that many C-based event loops do not handle 694C<send>.
505overloading, so as tempting as it may be, passing a condition variable
506instead of a callback does not work. Both the pure perl and EV loops
507support overloading, however, as well as all functions that use perl to
508invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
509example).
510 695
511=item $cv->croak ($error) 696=item $cv->croak ($error)
512 697
513Similar to send, but causes all call's to C<< ->recv >> to invoke 698Similar to send, but causes all calls to C<< ->recv >> to invoke
514C<Carp::croak> with the given error message/object/scalar. 699C<Carp::croak> with the given error message/object/scalar.
515 700
516This can be used to signal any errors to the condition variable 701This can be used to signal any errors to the condition variable
517user/consumer. 702user/consumer. Doing it this way instead of calling C<croak> directly
703delays the error detection, but has the overwhelming advantage that it
704diagnoses the error at the place where the result is expected, and not
705deep in some event callback with no connection to the actual code causing
706the problem.
518 707
519=item $cv->begin ([group callback]) 708=item $cv->begin ([group callback])
520 709
521=item $cv->end 710=item $cv->end
522
523These two methods are EXPERIMENTAL and MIGHT CHANGE.
524 711
525These two methods can be used to combine many transactions/events into 712These two methods can be used to combine many transactions/events into
526one. For example, a function that pings many hosts in parallel might want 713one. For example, a function that pings many hosts in parallel might want
527to use a condition variable for the whole process. 714to use a condition variable for the whole process.
528 715
529Every call to C<< ->begin >> will increment a counter, and every call to 716Every call to C<< ->begin >> will increment a counter, and every call to
530C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end 717C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
531>>, the (last) callback passed to C<begin> will be executed. That callback 718>>, the (last) callback passed to C<begin> will be executed, passing the
532is I<supposed> to call C<< ->send >>, but that is not required. If no 719condvar as first argument. That callback is I<supposed> to call C<< ->send
533callback was set, C<send> will be called without any arguments. 720>>, but that is not required. If no group callback was set, C<send> will
721be called without any arguments.
534 722
535Let's clarify this with the ping example: 723You can think of C<< $cv->send >> giving you an OR condition (one call
724sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
725condition (all C<begin> calls must be C<end>'ed before the condvar sends).
726
727Let's start with a simple example: you have two I/O watchers (for example,
728STDOUT and STDERR for a program), and you want to wait for both streams to
729close before activating a condvar:
536 730
537 my $cv = AnyEvent->condvar; 731 my $cv = AnyEvent->condvar;
538 732
733 $cv->begin; # first watcher
734 my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
735 defined sysread $fh1, my $buf, 4096
736 or $cv->end;
737 });
738
739 $cv->begin; # second watcher
740 my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
741 defined sysread $fh2, my $buf, 4096
742 or $cv->end;
743 });
744
745 $cv->recv;
746
747This works because for every event source (EOF on file handle), there is
748one call to C<begin>, so the condvar waits for all calls to C<end> before
749sending.
750
751The ping example mentioned above is slightly more complicated, as the
752there are results to be passwd back, and the number of tasks that are
753begun can potentially be zero:
754
755 my $cv = AnyEvent->condvar;
756
539 my %result; 757 my %result;
540 $cv->begin (sub { $cv->send (\%result) }); 758 $cv->begin (sub { shift->send (\%result) });
541 759
542 for my $host (@list_of_hosts) { 760 for my $host (@list_of_hosts) {
543 $cv->begin; 761 $cv->begin;
544 ping_host_then_call_callback $host, sub { 762 ping_host_then_call_callback $host, sub {
545 $result{$host} = ...; 763 $result{$host} = ...;
560loop, which serves two important purposes: first, it sets the callback 778loop, which serves two important purposes: first, it sets the callback
561to be called once the counter reaches C<0>, and second, it ensures that 779to be called once the counter reaches C<0>, and second, it ensures that
562C<send> is called even when C<no> hosts are being pinged (the loop 780C<send> is called even when C<no> hosts are being pinged (the loop
563doesn't execute once). 781doesn't execute once).
564 782
565This is the general pattern when you "fan out" into multiple subrequests: 783This is the general pattern when you "fan out" into multiple (but
566use an outer C<begin>/C<end> pair to set the callback and ensure C<end> 784potentially zero) subrequests: use an outer C<begin>/C<end> pair to set
567is called at least once, and then, for each subrequest you start, call 785the callback and ensure C<end> is called at least once, and then, for each
568C<begin> and for each subrequest you finish, call C<end>. 786subrequest you start, call C<begin> and for each subrequest you finish,
787call C<end>.
569 788
570=back 789=back
571 790
572=head3 METHODS FOR CONSUMERS 791=head3 METHODS FOR CONSUMERS
573 792
577=over 4 796=over 4
578 797
579=item $cv->recv 798=item $cv->recv
580 799
581Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak 800Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
582>> methods have been called on c<$cv>, while servicing other watchers 801>> methods have been called on C<$cv>, while servicing other watchers
583normally. 802normally.
584 803
585You can only wait once on a condition - additional calls are valid but 804You can only wait once on a condition - additional calls are valid but
586will return immediately. 805will return immediately.
587 806
589function will call C<croak>. 808function will call C<croak>.
590 809
591In list context, all parameters passed to C<send> will be returned, 810In list context, all parameters passed to C<send> will be returned,
592in scalar context only the first one will be returned. 811in scalar context only the first one will be returned.
593 812
813Note that doing a blocking wait in a callback is not supported by any
814event loop, that is, recursive invocation of a blocking C<< ->recv
815>> is not allowed, and the C<recv> call will C<croak> if such a
816condition is detected. This condition can be slightly loosened by using
817L<Coro::AnyEvent>, which allows you to do a blocking C<< ->recv >> from
818any thread that doesn't run the event loop itself.
819
594Not all event models support a blocking wait - some die in that case 820Not all event models support a blocking wait - some die in that case
595(programs might want to do that to stay interactive), so I<if you are 821(programs might want to do that to stay interactive), so I<if you are
596using this from a module, never require a blocking wait>, but let the 822using this from a module, never require a blocking wait>. Instead, let the
597caller decide whether the call will block or not (for example, by coupling 823caller decide whether the call will block or not (for example, by coupling
598condition variables with some kind of request results and supporting 824condition variables with some kind of request results and supporting
599callbacks so the caller knows that getting the result will not block, 825callbacks so the caller knows that getting the result will not block,
600while still supporting blocking waits if the caller so desires). 826while still supporting blocking waits if the caller so desires).
601 827
602Another reason I<never> to C<< ->recv >> in a module is that you cannot
603sensibly have two C<< ->recv >>'s in parallel, as that would require
604multiple interpreters or coroutines/threads, none of which C<AnyEvent>
605can supply.
606
607The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
608fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
609versions and also integrates coroutines into AnyEvent, making blocking
610C<< ->recv >> calls perfectly safe as long as they are done from another
611coroutine (one that doesn't run the event loop).
612
613You can ensure that C<< -recv >> never blocks by setting a callback and 828You can ensure that C<< ->recv >> never blocks by setting a callback and
614only calling C<< ->recv >> from within that callback (or at a later 829only calling C<< ->recv >> from within that callback (or at a later
615time). This will work even when the event loop does not support blocking 830time). This will work even when the event loop does not support blocking
616waits otherwise. 831waits otherwise.
617 832
618=item $bool = $cv->ready 833=item $bool = $cv->ready
624 839
625This is a mutator function that returns the callback set and optionally 840This is a mutator function that returns the callback set and optionally
626replaces it before doing so. 841replaces it before doing so.
627 842
628The callback will be called when the condition becomes "true", i.e. when 843The callback will be called when the condition becomes "true", i.e. when
629C<send> or C<croak> are called, with the only argument being the condition 844C<send> or C<croak> are called, with the only argument being the
630variable itself. Calling C<recv> inside the callback or at any later time 845condition variable itself. If the condition is already true, the
631is guaranteed not to block. 846callback is called immediately when it is set. Calling C<recv> inside
847the callback or at any later time is guaranteed not to block.
632 848
633=back 849=back
634 850
851=head1 SUPPORTED EVENT LOOPS/BACKENDS
852
853The available backend classes are (every class has its own manpage):
854
855=over 4
856
857=item Backends that are autoprobed when no other event loop can be found.
858
859EV is the preferred backend when no other event loop seems to be in
860use. If EV is not installed, then AnyEvent will fall back to its own
861pure-perl implementation, which is available everywhere as it comes with
862AnyEvent itself.
863
864 AnyEvent::Impl::EV based on EV (interface to libev, best choice).
865 AnyEvent::Impl::Perl pure-perl AnyEvent::Loop, fast and portable.
866
867=item Backends that are transparently being picked up when they are used.
868
869These will be used if they are already loaded when the first watcher
870is created, in which case it is assumed that the application is using
871them. This means that AnyEvent will automatically pick the right backend
872when the main program loads an event module before anything starts to
873create watchers. Nothing special needs to be done by the main program.
874
875 AnyEvent::Impl::Event based on Event, very stable, few glitches.
876 AnyEvent::Impl::Glib based on Glib, slow but very stable.
877 AnyEvent::Impl::Tk based on Tk, very broken.
878 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
879 AnyEvent::Impl::POE based on POE, very slow, some limitations.
880 AnyEvent::Impl::Irssi used when running within irssi.
881 AnyEvent::Impl::IOAsync based on IO::Async.
882 AnyEvent::Impl::Cocoa based on Cocoa::EventLoop.
883 AnyEvent::Impl::FLTK2 based on FLTK (fltk 2 binding).
884
885=item Backends with special needs.
886
887Qt requires the Qt::Application to be instantiated first, but will
888otherwise be picked up automatically. As long as the main program
889instantiates the application before any AnyEvent watchers are created,
890everything should just work.
891
892 AnyEvent::Impl::Qt based on Qt.
893
894=item Event loops that are indirectly supported via other backends.
895
896Some event loops can be supported via other modules:
897
898There is no direct support for WxWidgets (L<Wx>) or L<Prima>.
899
900B<WxWidgets> has no support for watching file handles. However, you can
901use WxWidgets through the POE adaptor, as POE has a Wx backend that simply
902polls 20 times per second, which was considered to be too horrible to even
903consider for AnyEvent.
904
905B<Prima> is not supported as nobody seems to be using it, but it has a POE
906backend, so it can be supported through POE.
907
908AnyEvent knows about both L<Prima> and L<Wx>, however, and will try to
909load L<POE> when detecting them, in the hope that POE will pick them up,
910in which case everything will be automatic.
911
912=back
913
635=head1 GLOBAL VARIABLES AND FUNCTIONS 914=head1 GLOBAL VARIABLES AND FUNCTIONS
636 915
916These are not normally required to use AnyEvent, but can be useful to
917write AnyEvent extension modules.
918
637=over 4 919=over 4
638 920
639=item $AnyEvent::MODEL 921=item $AnyEvent::MODEL
640 922
641Contains C<undef> until the first watcher is being created. Then it 923Contains C<undef> until the first watcher is being created, before the
924backend has been autodetected.
925
642contains the event model that is being used, which is the name of the 926Afterwards it contains the event model that is being used, which is the
643Perl class implementing the model. This class is usually one of the 927name of the Perl class implementing the model. This class is usually one
644C<AnyEvent::Impl:xxx> modules, but can be any other class in the case 928of the C<AnyEvent::Impl::xxx> modules, but can be any other class in the
645AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>). 929case AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode> it
646 930will be C<urxvt::anyevent>).
647The known classes so far are:
648
649 AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
650 AnyEvent::Impl::Event based on Event, second best choice.
651 AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
652 AnyEvent::Impl::Glib based on Glib, third-best choice.
653 AnyEvent::Impl::Tk based on Tk, very bad choice.
654 AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
655 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
656 AnyEvent::Impl::POE based on POE, not generic enough for full support.
657
658There is no support for WxWidgets, as WxWidgets has no support for
659watching file handles. However, you can use WxWidgets through the
660POE Adaptor, as POE has a Wx backend that simply polls 20 times per
661second, which was considered to be too horrible to even consider for
662AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
663it's adaptor.
664
665AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
666autodetecting them.
667 931
668=item AnyEvent::detect 932=item AnyEvent::detect
669 933
670Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model 934Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
671if necessary. You should only call this function right before you would 935if necessary. You should only call this function right before you would
672have created an AnyEvent watcher anyway, that is, as late as possible at 936have created an AnyEvent watcher anyway, that is, as late as possible at
673runtime. 937runtime, and not e.g. during initialisation of your module.
938
939The effect of calling this function is as if a watcher had been created
940(specifically, actions that happen "when the first watcher is created"
941happen when calling detetc as well).
942
943If you need to do some initialisation before AnyEvent watchers are
944created, use C<post_detect>.
674 945
675=item $guard = AnyEvent::post_detect { BLOCK } 946=item $guard = AnyEvent::post_detect { BLOCK }
676 947
677Arranges for the code block to be executed as soon as the event model is 948Arranges for the code block to be executed as soon as the event model is
678autodetected (or immediately if this has already happened). 949autodetected (or immediately if that has already happened).
950
951The block will be executed I<after> the actual backend has been detected
952(C<$AnyEvent::MODEL> is set), but I<before> any watchers have been
953created, so it is possible to e.g. patch C<@AnyEvent::ISA> or do
954other initialisations - see the sources of L<AnyEvent::Strict> or
955L<AnyEvent::AIO> to see how this is used.
956
957The most common usage is to create some global watchers, without forcing
958event module detection too early, for example, L<AnyEvent::AIO> creates
959and installs the global L<IO::AIO> watcher in a C<post_detect> block to
960avoid autodetecting the event module at load time.
679 961
680If called in scalar or list context, then it creates and returns an object 962If called in scalar or list context, then it creates and returns an object
681that automatically removes the callback again when it is destroyed. See 963that automatically removes the callback again when it is destroyed (or
964C<undef> when the hook was immediately executed). See L<AnyEvent::AIO> for
682L<Coro::BDB> for a case where this is useful. 965a case where this is useful.
966
967Example: Create a watcher for the IO::AIO module and store it in
968C<$WATCHER>, but do so only do so after the event loop is initialised.
969
970 our WATCHER;
971
972 my $guard = AnyEvent::post_detect {
973 $WATCHER = AnyEvent->io (fh => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb);
974 };
975
976 # the ||= is important in case post_detect immediately runs the block,
977 # as to not clobber the newly-created watcher. assigning both watcher and
978 # post_detect guard to the same variable has the advantage of users being
979 # able to just C<undef $WATCHER> if the watcher causes them grief.
980
981 $WATCHER ||= $guard;
683 982
684=item @AnyEvent::post_detect 983=item @AnyEvent::post_detect
685 984
686If there are any code references in this array (you can C<push> to it 985If there are any code references in this array (you can C<push> to it
687before or after loading AnyEvent), then they will called directly after 986before or after loading AnyEvent), then they will be called directly
688the event loop has been chosen. 987after the event loop has been chosen.
689 988
690You should check C<$AnyEvent::MODEL> before adding to this array, though: 989You should check C<$AnyEvent::MODEL> before adding to this array, though:
691if it contains a true value then the event loop has already been detected, 990if it is defined then the event loop has already been detected, and the
692and the array will be ignored. 991array will be ignored.
693 992
694Best use C<AnyEvent::post_detect { BLOCK }> instead. 993Best use C<AnyEvent::post_detect { BLOCK }> when your application allows
994it, as it takes care of these details.
995
996This variable is mainly useful for modules that can do something useful
997when AnyEvent is used and thus want to know when it is initialised, but do
998not need to even load it by default. This array provides the means to hook
999into AnyEvent passively, without loading it.
1000
1001Example: To load Coro::AnyEvent whenever Coro and AnyEvent are used
1002together, you could put this into Coro (this is the actual code used by
1003Coro to accomplish this):
1004
1005 if (defined $AnyEvent::MODEL) {
1006 # AnyEvent already initialised, so load Coro::AnyEvent
1007 require Coro::AnyEvent;
1008 } else {
1009 # AnyEvent not yet initialised, so make sure to load Coro::AnyEvent
1010 # as soon as it is
1011 push @AnyEvent::post_detect, sub { require Coro::AnyEvent };
1012 }
1013
1014=item AnyEvent::postpone { BLOCK }
1015
1016Arranges for the block to be executed as soon as possible, but not before
1017the call itself returns. In practise, the block will be executed just
1018before the event loop polls for new events, or shortly afterwards.
1019
1020This function never returns anything (to make the C<return postpone { ...
1021}> idiom more useful.
1022
1023To understand the usefulness of this function, consider a function that
1024asynchronously does something for you and returns some transaction
1025object or guard to let you cancel the operation. For example,
1026C<AnyEvent::Socket::tcp_connect>:
1027
1028 # start a conenction attempt unless one is active
1029 $self->{connect_guard} ||= AnyEvent::Socket::tcp_connect "www.example.net", 80, sub {
1030 delete $self->{connect_guard};
1031 ...
1032 };
1033
1034Imagine that this function could instantly call the callback, for
1035example, because it detects an obvious error such as a negative port
1036number. Invoking the callback before the function returns causes problems
1037however: the callback will be called and will try to delete the guard
1038object. But since the function hasn't returned yet, there is nothing to
1039delete. When the function eventually returns it will assign the guard
1040object to C<< $self->{connect_guard} >>, where it will likely never be
1041deleted, so the program thinks it is still trying to connect.
1042
1043This is where C<AnyEvent::postpone> should be used. Instead of calling the
1044callback directly on error:
1045
1046 $cb->(undef), return # signal error to callback, BAD!
1047 if $some_error_condition;
1048
1049It should use C<postpone>:
1050
1051 AnyEvent::postpone { $cb->(undef) }, return # signal error to callback, later
1052 if $some_error_condition;
1053
1054=item AnyEvent::log $level, $msg[, @args]
1055
1056Log the given C<$msg> at the given C<$level>.
1057
1058Loads AnyEvent::Log on first use and calls C<AnyEvent::Log::log> -
1059consequently, look at the L<AnyEvent::Log> documentation for details.
695 1060
696=back 1061=back
697 1062
698=head1 WHAT TO DO IN A MODULE 1063=head1 WHAT TO DO IN A MODULE
699 1064
710because it will stall the whole program, and the whole point of using 1075because it will stall the whole program, and the whole point of using
711events is to stay interactive. 1076events is to stay interactive.
712 1077
713It is fine, however, to call C<< ->recv >> when the user of your module 1078It is fine, however, to call C<< ->recv >> when the user of your module
714requests it (i.e. if you create a http request object ad have a method 1079requests it (i.e. if you create a http request object ad have a method
715called C<results> that returns the results, it should call C<< ->recv >> 1080called C<results> that returns the results, it may call C<< ->recv >>
716freely, as the user of your module knows what she is doing. always). 1081freely, as the user of your module knows what she is doing. Always).
717 1082
718=head1 WHAT TO DO IN THE MAIN PROGRAM 1083=head1 WHAT TO DO IN THE MAIN PROGRAM
719 1084
720There will always be a single main program - the only place that should 1085There will always be a single main program - the only place that should
721dictate which event model to use. 1086dictate which event model to use.
722 1087
723If it doesn't care, it can just "use AnyEvent" and use it itself, or not 1088If the program is not event-based, it need not do anything special, even
724do anything special (it does not need to be event-based) and let AnyEvent 1089when it depends on a module that uses an AnyEvent. If the program itself
725decide which implementation to chose if some module relies on it. 1090uses AnyEvent, but does not care which event loop is used, all it needs
1091to do is C<use AnyEvent>. In either case, AnyEvent will choose the best
1092available loop implementation.
726 1093
727If the main program relies on a specific event model - for example, in 1094If the main program relies on a specific event model - for example, in
728Gtk2 programs you have to rely on the Glib module - you should load the 1095Gtk2 programs you have to rely on the Glib module - you should load the
729event module before loading AnyEvent or any module that uses it: generally 1096event module before loading AnyEvent or any module that uses it: generally
730speaking, you should load it as early as possible. The reason is that 1097speaking, you should load it as early as possible. The reason is that
731modules might create watchers when they are loaded, and AnyEvent will 1098modules might create watchers when they are loaded, and AnyEvent will
732decide on the event model to use as soon as it creates watchers, and it 1099decide on the event model to use as soon as it creates watchers, and it
733might chose the wrong one unless you load the correct one yourself. 1100might choose the wrong one unless you load the correct one yourself.
734 1101
735You can chose to use a pure-perl implementation by loading the 1102You can chose to use a pure-perl implementation by loading the
736C<AnyEvent::Impl::Perl> module, which gives you similar behaviour 1103C<AnyEvent::Loop> module, which gives you similar behaviour
737everywhere, but letting AnyEvent chose the model is generally better. 1104everywhere, but letting AnyEvent chose the model is generally better.
738 1105
739=head2 MAINLOOP EMULATION 1106=head2 MAINLOOP EMULATION
740 1107
741Sometimes (often for short test scripts, or even standalone programs who 1108Sometimes (often for short test scripts, or even standalone programs who
754 1121
755 1122
756=head1 OTHER MODULES 1123=head1 OTHER MODULES
757 1124
758The following is a non-exhaustive list of additional modules that use 1125The following is a non-exhaustive list of additional modules that use
759AnyEvent and can therefore be mixed easily with other AnyEvent modules 1126AnyEvent as a client and can therefore be mixed easily with other AnyEvent
760in the same program. Some of the modules come with AnyEvent, some are 1127modules and other event loops in the same program. Some of the modules
761available via CPAN. 1128come as part of AnyEvent, the others are available via CPAN.
762 1129
763=over 4 1130=over 4
764 1131
765=item L<AnyEvent::Util> 1132=item L<AnyEvent::Util>
766 1133
767Contains various utility functions that replace often-used but blocking 1134Contains various utility functions that replace often-used blocking
768functions such as C<inet_aton> by event-/callback-based versions. 1135functions such as C<inet_aton> with event/callback-based versions.
769 1136
770=item L<AnyEvent::Socket> 1137=item L<AnyEvent::Socket>
771 1138
772Provides various utility functions for (internet protocol) sockets, 1139Provides various utility functions for (internet protocol) sockets,
773addresses and name resolution. Also functions to create non-blocking tcp 1140addresses and name resolution. Also functions to create non-blocking tcp
775 1142
776=item L<AnyEvent::Handle> 1143=item L<AnyEvent::Handle>
777 1144
778Provide read and write buffers, manages watchers for reads and writes, 1145Provide read and write buffers, manages watchers for reads and writes,
779supports raw and formatted I/O, I/O queued and fully transparent and 1146supports raw and formatted I/O, I/O queued and fully transparent and
780non-blocking SSL/TLS. 1147non-blocking SSL/TLS (via L<AnyEvent::TLS>).
781 1148
782=item L<AnyEvent::DNS> 1149=item L<AnyEvent::DNS>
783 1150
784Provides rich asynchronous DNS resolver capabilities. 1151Provides rich asynchronous DNS resolver capabilities.
785 1152
1153=item L<AnyEvent::HTTP>, L<AnyEvent::IRC>, L<AnyEvent::XMPP>, L<AnyEvent::GPSD>, L<AnyEvent::IGS>, L<AnyEvent::FCP>
1154
1155Implement event-based interfaces to the protocols of the same name (for
1156the curious, IGS is the International Go Server and FCP is the Freenet
1157Client Protocol).
1158
1159=item L<AnyEvent::Handle::UDP>
1160
1161Here be danger!
1162
1163As Pauli would put it, "Not only is it not right, it's not even wrong!" -
1164there are so many things wrong with AnyEvent::Handle::UDP, most notably
1165its use of a stream-based API with a protocol that isn't streamable, that
1166the only way to improve it is to delete it.
1167
1168It features data corruption (but typically only under load) and general
1169confusion. On top, the author is not only clueless about UDP but also
1170fact-resistant - some gems of his understanding: "connect doesn't work
1171with UDP", "UDP packets are not IP packets", "UDP only has datagrams, not
1172packets", "I don't need to implement proper error checking as UDP doesn't
1173support error checking" and so on - he doesn't even understand what's
1174wrong with his module when it is explained to him.
1175
786=item L<AnyEvent::HTTP> 1176=item L<AnyEvent::DBI>
787 1177
788A simple-to-use HTTP library that is capable of making a lot of concurrent 1178Executes L<DBI> requests asynchronously in a proxy process for you,
789HTTP requests. 1179notifying you in an event-based way when the operation is finished.
1180
1181=item L<AnyEvent::AIO>
1182
1183Truly asynchronous (as opposed to non-blocking) I/O, should be in the
1184toolbox of every event programmer. AnyEvent::AIO transparently fuses
1185L<IO::AIO> and AnyEvent together, giving AnyEvent access to event-based
1186file I/O, and much more.
790 1187
791=item L<AnyEvent::HTTPD> 1188=item L<AnyEvent::HTTPD>
792 1189
793Provides a simple web application server framework. 1190A simple embedded webserver.
794 1191
795=item L<AnyEvent::FastPing> 1192=item L<AnyEvent::FastPing>
796 1193
797The fastest ping in the west. 1194The fastest ping in the west.
798 1195
799=item L<AnyEvent::DBI>
800
801Executes L<DBI> requests asynchronously in a proxy process.
802
803=item L<AnyEvent::AIO>
804
805Truly asynchronous I/O, should be in the toolbox of every event
806programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
807together.
808
809=item L<AnyEvent::BDB>
810
811Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
812L<BDB> and AnyEvent together.
813
814=item L<AnyEvent::GPSD>
815
816A non-blocking interface to gpsd, a daemon delivering GPS information.
817
818=item L<AnyEvent::IGS>
819
820A non-blocking interface to the Internet Go Server protocol (used by
821L<App::IGS>).
822
823=item L<Net::IRC3>
824
825AnyEvent based IRC client module family.
826
827=item L<Net::XMPP2>
828
829AnyEvent based XMPP (Jabber protocol) module family.
830
831=item L<Net::FCP>
832
833AnyEvent-based implementation of the Freenet Client Protocol, birthplace
834of AnyEvent.
835
836=item L<Event::ExecFlow>
837
838High level API for event-based execution flow control.
839
840=item L<Coro> 1196=item L<Coro>
841 1197
842Has special support for AnyEvent via L<Coro::AnyEvent>. 1198Has special support for AnyEvent via L<Coro::AnyEvent>.
843 1199
844=item L<IO::Lambda>
845
846The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
847
848=back 1200=back
849 1201
850=cut 1202=cut
851 1203
852package AnyEvent; 1204package AnyEvent;
853 1205
854no warnings; 1206# basically a tuned-down version of common::sense
855use strict; 1207sub common_sense {
1208 # from common:.sense 3.4
1209 ${^WARNING_BITS} ^= ${^WARNING_BITS} ^ "\x3c\x3f\x33\x00\x0f\xf0\x0f\xc0\xf0\xfc\x33\x00";
1210 # use strict vars subs - NO UTF-8, as Util.pm doesn't like this atm. (uts46data.pl)
1211 $^H |= 0x00000600;
1212}
856 1213
1214BEGIN { AnyEvent::common_sense }
1215
857use Carp; 1216use Carp ();
858 1217
859our $VERSION = 4.233; 1218our $VERSION = '6.01';
860our $MODEL; 1219our $MODEL;
861 1220
862our $AUTOLOAD;
863our @ISA; 1221our @ISA;
864 1222
865our @REGISTRY; 1223our @REGISTRY;
866 1224
867our $WIN32; 1225our $VERBOSE;
868 1226
869BEGIN { 1227BEGIN {
870 my $win32 = ! ! ($^O =~ /mswin32/i); 1228 require "AnyEvent/constants.pl";
871 eval "sub WIN32(){ $win32 }";
872}
873 1229
1230 eval "sub TAINT (){" . (${^TAINT}*1) . "}";
1231
1232 delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
1233 if ${^TAINT};
1234
874our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1; 1235 $VERBOSE = $ENV{PERL_ANYEVENT_VERBOSE}*1;
1236}
1237
1238our $MAX_SIGNAL_LATENCY = 10;
875 1239
876our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred 1240our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
877 1241
878{ 1242{
879 my $idx; 1243 my $idx;
880 $PROTOCOL{$_} = ++$idx 1244 $PROTOCOL{$_} = ++$idx
881 for reverse split /\s*,\s*/, 1245 for reverse split /\s*,\s*/,
882 $ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6"; 1246 $ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
883} 1247}
884 1248
1249our @post_detect;
1250
1251sub post_detect(&) {
1252 my ($cb) = @_;
1253
1254 push @post_detect, $cb;
1255
1256 defined wantarray
1257 ? bless \$cb, "AnyEvent::Util::postdetect"
1258 : ()
1259}
1260
1261sub AnyEvent::Util::postdetect::DESTROY {
1262 @post_detect = grep $_ != ${$_[0]}, @post_detect;
1263}
1264
1265our $POSTPONE_W;
1266our @POSTPONE;
1267
1268sub _postpone_exec {
1269 undef $POSTPONE_W;
1270
1271 &{ shift @POSTPONE }
1272 while @POSTPONE;
1273}
1274
1275sub postpone(&) {
1276 push @POSTPONE, shift;
1277
1278 $POSTPONE_W ||= AE::timer (0, 0, \&_postpone_exec);
1279
1280 ()
1281}
1282
1283sub log($$;@) {
1284 require AnyEvent::Log;
1285 # AnyEvent::Log overwrites this function
1286 goto &log;
1287}
1288
885my @models = ( 1289our @models = (
886 [EV:: => AnyEvent::Impl::EV::], 1290 [EV:: => AnyEvent::Impl::EV:: , 1],
887 [Event:: => AnyEvent::Impl::Event::], 1291 [AnyEvent::Loop:: => AnyEvent::Impl::Perl:: , 1],
888 [AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
889 # everything below here will not be autoprobed 1292 # everything below here will not (normally) be autoprobed
890 # as the pureperl backend should work everywhere 1293 # as the pure perl backend should work everywhere
891 # and is usually faster 1294 # and is usually faster
1295 [Event:: => AnyEvent::Impl::Event::, 1],
1296 [Glib:: => AnyEvent::Impl::Glib:: , 1], # becomes extremely slow with many watchers
1297 [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
1298 [Irssi:: => AnyEvent::Impl::Irssi::], # Irssi has a bogus "Event" package
892 [Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles 1299 [Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
893 [Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
894 [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
895 [Qt:: => AnyEvent::Impl::Qt::], # requires special main program 1300 [Qt:: => AnyEvent::Impl::Qt::], # requires special main program
896 [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza 1301 [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
897 [Wx:: => AnyEvent::Impl::POE::], 1302 [Wx:: => AnyEvent::Impl::POE::],
898 [Prima:: => AnyEvent::Impl::POE::], 1303 [Prima:: => AnyEvent::Impl::POE::],
1304 [IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # a bitch to autodetect
1305 [Cocoa::EventLoop:: => AnyEvent::Impl::Cocoa::],
1306 [FLTK:: => AnyEvent::Impl::FLTK2::],
899); 1307);
900 1308
901our %method = map +($_ => 1), qw(io timer time now signal child condvar one_event DESTROY); 1309our @isa_hook;
902 1310
903our @post_detect; 1311sub _isa_set {
1312 my @pkg = ("AnyEvent", (map $_->[0], grep defined, @isa_hook), $MODEL);
904 1313
1314 @{"$pkg[$_-1]::ISA"} = $pkg[$_]
1315 for 1 .. $#pkg;
1316
1317 grep $_ && $_->[1], @isa_hook
1318 and AE::_reset ();
1319}
1320
1321# used for hooking AnyEvent::Strict and AnyEvent::Debug::Wrap into the class hierarchy
1322sub _isa_hook($$;$) {
1323 my ($i, $pkg, $reset_ae) = @_;
1324
1325 $isa_hook[$i] = $pkg ? [$pkg, $reset_ae] : undef;
1326
1327 _isa_set;
1328}
1329
1330# all autoloaded methods reserve the complete glob, not just the method slot.
1331# due to bugs in perls method cache implementation.
1332our @methods = qw(io timer time now now_update signal child idle condvar);
1333
905sub post_detect(&) { 1334sub detect() {
906 my ($cb) = @_; 1335 return $MODEL if $MODEL; # some programs keep references to detect
907 1336
908 if ($MODEL) { 1337 local $!; # for good measure
909 $cb->(); 1338 local $SIG{__DIE__}; # we use eval
910 1339
911 1 1340 # free some memory
1341 *detect = sub () { $MODEL };
1342 # undef &func doesn't correctly update the method cache. grmbl.
1343 # so we delete the whole glob. grmbl.
1344 # otoh, perl doesn't let me undef an active usb, but it lets me free
1345 # a glob with an active sub. hrm. i hope it works, but perl is
1346 # usually buggy in this department. sigh.
1347 delete @{"AnyEvent::"}{@methods};
1348 undef @methods;
1349
1350 if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z0-9:]+)$/) {
1351 my $model = $1;
1352 $model = "AnyEvent::Impl::$model" unless $model =~ s/::$//;
1353 if (eval "require $model") {
1354 $MODEL = $model;
1355 AnyEvent::log 7 => "loaded model '$model' (forced by \$ENV{PERL_ANYEVENT_MODEL}), using it."
1356 if $VERBOSE >= 7;
912 } else { 1357 } else {
913 push @post_detect, $cb; 1358 AnyEvent::log warn => "unable to load model '$model' (from \$ENV{PERL_ANYEVENT_MODEL}):\n$@";
914 1359 }
915 defined wantarray
916 ? bless \$cb, "AnyEvent::Util::PostDetect"
917 : ()
918 } 1360 }
919}
920 1361
921sub AnyEvent::Util::PostDetect::DESTROY { 1362 # check for already loaded models
922 @post_detect = grep $_ != ${$_[0]}, @post_detect;
923}
924
925sub detect() {
926 unless ($MODEL) { 1363 unless ($MODEL) {
927 no strict 'refs'; 1364 for (@REGISTRY, @models) {
928 local $SIG{__DIE__}; 1365 my ($package, $model) = @$_;
929 1366 if (${"$package\::VERSION"} > 0) {
930 if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
931 my $model = "AnyEvent::Impl::$1";
932 if (eval "require $model") { 1367 if (eval "require $model") {
933 $MODEL = $model; 1368 $MODEL = $model;
934 warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1; 1369 AnyEvent::log 7 => "autodetected model '$model', using it."
935 } else { 1370 if $VERBOSE >= 7;
936 warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose; 1371 last;
1372 }
937 } 1373 }
938 } 1374 }
939 1375
940 # check for already loaded models
941 unless ($MODEL) { 1376 unless ($MODEL) {
1377 # try to autoload a model
942 for (@REGISTRY, @models) { 1378 for (@REGISTRY, @models) {
943 my ($package, $model) = @$_; 1379 my ($package, $model, $autoload) = @$_;
1380 if (
1381 $autoload
1382 and eval "require $package"
944 if (${"$package\::VERSION"} > 0) { 1383 and ${"$package\::VERSION"} > 0
945 if (eval "require $model") { 1384 and eval "require $model"
1385 ) {
946 $MODEL = $model; 1386 $MODEL = $model;
947 warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1; 1387 AnyEvent::log 7 => "autoloaded model '$model', using it."
1388 if $VERBOSE >= 7;
948 last; 1389 last;
949 }
950 } 1390 }
951 } 1391 }
952 1392
953 unless ($MODEL) {
954 # try to load a model
955
956 for (@REGISTRY, @models) {
957 my ($package, $model) = @$_;
958 if (eval "require $package"
959 and ${"$package\::VERSION"} > 0
960 and eval "require $model") {
961 $MODEL = $model;
962 warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1;
963 last;
964 }
965 }
966
967 $MODEL 1393 $MODEL
968 or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib."; 1394 or die "AnyEvent: backend autodetection failed - did you properly install AnyEvent?";
969 }
970 } 1395 }
971
972 push @{"$MODEL\::ISA"}, "AnyEvent::Base";
973
974 unshift @ISA, $MODEL;
975
976 require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
977
978 (shift @post_detect)->() while @post_detect;
979 } 1396 }
980 1397
1398 # free memory only needed for probing
1399 undef @models;
1400 undef @REGISTRY;
1401
1402 push @{"$MODEL\::ISA"}, "AnyEvent::Base";
1403
1404 # now nuke some methods that are overridden by the backend.
1405 # SUPER usage is not allowed in these.
1406 for (qw(time signal child idle)) {
1407 undef &{"AnyEvent::Base::$_"}
1408 if defined &{"$MODEL\::$_"};
1409 }
1410
1411 _isa_set;
1412
1413 if ($ENV{PERL_ANYEVENT_STRICT}) {
1414 require AnyEvent::Strict;
1415 }
1416
1417 if ($ENV{PERL_ANYEVENT_DEBUG_WRAP}) {
1418 require AnyEvent::Debug;
1419 AnyEvent::Debug::wrap ($ENV{PERL_ANYEVENT_DEBUG_WRAP});
1420 }
1421
1422 if (exists $ENV{PERL_ANYEVENT_DEBUG_SHELL}) {
1423 require AnyEvent::Socket;
1424 require AnyEvent::Debug;
1425
1426 my $shell = $ENV{PERL_ANYEVENT_DEBUG_SHELL};
1427 $shell =~ s/\$\$/$$/g;
1428
1429 my ($host, $service) = AnyEvent::Socket::parse_hostport ($shell);
1430 $AnyEvent::Debug::SHELL = AnyEvent::Debug::shell ($host, $service);
1431 }
1432
1433 (shift @post_detect)->() while @post_detect;
1434 undef @post_detect;
1435
1436 *post_detect = sub(&) {
1437 shift->();
1438
1439 undef
1440 };
1441
981 $MODEL 1442 $MODEL
982} 1443}
983 1444
984sub AUTOLOAD { 1445for my $name (@methods) {
985 (my $func = $AUTOLOAD) =~ s/.*://; 1446 *$name = sub {
986 1447 detect;
987 $method{$func} 1448 # we use goto because
988 or croak "$func: not a valid method for AnyEvent objects"; 1449 # a) it makes the thunk more transparent
989 1450 # b) it allows us to delete the thunk later
990 detect unless $MODEL; 1451 goto &{ UNIVERSAL::can AnyEvent => "SUPER::$name" }
991 1452 };
992 my $class = shift;
993 $class->$func (@_);
994} 1453}
995 1454
996# utility function to dup a filehandle. this is used by many backends 1455# utility function to dup a filehandle. this is used by many backends
997# to support binding more than one watcher per filehandle (they usually 1456# to support binding more than one watcher per filehandle (they usually
998# allow only one watcher per fd, so we dup it to get a different one). 1457# allow only one watcher per fd, so we dup it to get a different one).
999sub _dupfh($$$$) { 1458sub _dupfh($$;$$) {
1000 my ($poll, $fh, $r, $w) = @_; 1459 my ($poll, $fh, $r, $w) = @_;
1001 1460
1002 require Fcntl;
1003
1004 # cygwin requires the fh mode to be matching, unix doesn't 1461 # cygwin requires the fh mode to be matching, unix doesn't
1005 my ($rw, $mode) = $poll eq "r" ? ($r, "<") 1462 my ($rw, $mode) = $poll eq "r" ? ($r, "<&") : ($w, ">&");
1006 : $poll eq "w" ? ($w, ">")
1007 : Carp::croak "AnyEvent->io requires poll set to either 'r' or 'w'";
1008 1463
1009 open my $fh2, "$mode&" . fileno $fh 1464 open my $fh2, $mode, $fh
1010 or die "cannot dup() filehandle: $!"; 1465 or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,";
1011 1466
1012 # we assume CLOEXEC is already set by perl in all important cases 1467 # we assume CLOEXEC is already set by perl in all important cases
1013 1468
1014 ($fh2, $rw) 1469 ($fh2, $rw)
1015} 1470}
1016 1471
1472=head1 SIMPLIFIED AE API
1473
1474Starting with version 5.0, AnyEvent officially supports a second, much
1475simpler, API that is designed to reduce the calling, typing and memory
1476overhead by using function call syntax and a fixed number of parameters.
1477
1478See the L<AE> manpage for details.
1479
1480=cut
1481
1482package AE;
1483
1484our $VERSION = $AnyEvent::VERSION;
1485
1486sub _reset() {
1487 eval q{
1488 # fall back to the main API by default - backends and AnyEvent::Base
1489 # implementations can overwrite these.
1490
1491 sub io($$$) {
1492 AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2])
1493 }
1494
1495 sub timer($$$) {
1496 AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2])
1497 }
1498
1499 sub signal($$) {
1500 AnyEvent->signal (signal => $_[0], cb => $_[1])
1501 }
1502
1503 sub child($$) {
1504 AnyEvent->child (pid => $_[0], cb => $_[1])
1505 }
1506
1507 sub idle($) {
1508 AnyEvent->idle (cb => $_[0]);
1509 }
1510
1511 sub cv(;&) {
1512 AnyEvent->condvar (@_ ? (cb => $_[0]) : ())
1513 }
1514
1515 sub now() {
1516 AnyEvent->now
1517 }
1518
1519 sub now_update() {
1520 AnyEvent->now_update
1521 }
1522
1523 sub time() {
1524 AnyEvent->time
1525 }
1526
1527 *postpone = \&AnyEvent::postpone;
1528 *log = \&AnyEvent::log;
1529 };
1530 die if $@;
1531}
1532
1533BEGIN { _reset }
1534
1017package AnyEvent::Base; 1535package AnyEvent::Base;
1018 1536
1019# default implementation for now and time 1537# default implementations for many methods
1020 1538
1021BEGIN { 1539sub time {
1540 eval q{ # poor man's autoloading {}
1541 # probe for availability of Time::HiRes
1022 if (eval "use Time::HiRes (); time (); 1") { 1542 if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
1543 AnyEvent::log 8 => "AnyEvent: using Time::HiRes for sub-second timing accuracy."
1544 if $AnyEvent::VERBOSE >= 8;
1545 *time = sub { Time::HiRes::time () };
1023 *_time = \&Time::HiRes::time; 1546 *AE::time = \& Time::HiRes::time ;
1024 # if (eval "use POSIX (); (POSIX::times())... 1547 # if (eval "use POSIX (); (POSIX::times())...
1025 } else { 1548 } else {
1026 *_time = \&CORE::time; # epic fail 1549 AnyEvent::log critical => "using built-in time(), WARNING, no sub-second resolution!";
1550 *time = sub { CORE::time };
1551 *AE::time = sub (){ CORE::time };
1552 }
1553
1554 *now = \&time;
1555 };
1556 die if $@;
1557
1558 &time
1559}
1560
1561*now = \&time;
1562sub now_update { }
1563
1564sub _poll {
1565 Carp::croak "$AnyEvent::MODEL does not support blocking waits. Caught";
1566}
1567
1568# default implementation for ->condvar
1569# in fact, the default should not be overwritten
1570
1571sub condvar {
1572 eval q{ # poor man's autoloading {}
1573 *condvar = sub {
1574 bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
1575 };
1576
1577 *AE::cv = sub (;&) {
1578 bless { @_ ? (_ae_cb => shift) : () }, "AnyEvent::CondVar"
1579 };
1580 };
1581 die if $@;
1582
1583 &condvar
1584}
1585
1586# default implementation for ->signal
1587
1588our $HAVE_ASYNC_INTERRUPT;
1589
1590sub _have_async_interrupt() {
1591 $HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT}
1592 && eval "use Async::Interrupt 1.02 (); 1")
1593 unless defined $HAVE_ASYNC_INTERRUPT;
1594
1595 $HAVE_ASYNC_INTERRUPT
1596}
1597
1598our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
1599our (%SIG_ASY, %SIG_ASY_W);
1600our ($SIG_COUNT, $SIG_TW);
1601
1602# install a dummy wakeup watcher to reduce signal catching latency
1603# used by Impls
1604sub _sig_add() {
1605 unless ($SIG_COUNT++) {
1606 # try to align timer on a full-second boundary, if possible
1607 my $NOW = AE::now;
1608
1609 $SIG_TW = AE::timer
1610 $MAX_SIGNAL_LATENCY - ($NOW - int $NOW),
1611 $MAX_SIGNAL_LATENCY,
1612 sub { } # just for the PERL_ASYNC_CHECK
1613 ;
1027 } 1614 }
1028} 1615}
1029 1616
1030sub time { _time } 1617sub _sig_del {
1031sub now { _time } 1618 undef $SIG_TW
1032 1619 unless --$SIG_COUNT;
1033# default implementation for ->condvar
1034
1035sub condvar {
1036 bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, AnyEvent::CondVar::
1037} 1620}
1038 1621
1039# default implementation for ->signal 1622our $_sig_name_init; $_sig_name_init = sub {
1623 eval q{ # poor man's autoloading {}
1624 undef $_sig_name_init;
1040 1625
1041our %SIG_CB; 1626 if (_have_async_interrupt) {
1627 *sig2num = \&Async::Interrupt::sig2num;
1628 *sig2name = \&Async::Interrupt::sig2name;
1629 } else {
1630 require Config;
1631
1632 my %signame2num;
1633 @signame2num{ split ' ', $Config::Config{sig_name} }
1634 = split ' ', $Config::Config{sig_num};
1635
1636 my @signum2name;
1637 @signum2name[values %signame2num] = keys %signame2num;
1638
1639 *sig2num = sub($) {
1640 $_[0] > 0 ? shift : $signame2num{+shift}
1641 };
1642 *sig2name = sub ($) {
1643 $_[0] > 0 ? $signum2name[+shift] : shift
1644 };
1645 }
1646 };
1647 die if $@;
1648};
1649
1650sub sig2num ($) { &$_sig_name_init; &sig2num }
1651sub sig2name($) { &$_sig_name_init; &sig2name }
1042 1652
1043sub signal { 1653sub signal {
1654 eval q{ # poor man's autoloading {}
1655 # probe for availability of Async::Interrupt
1656 if (_have_async_interrupt) {
1657 AnyEvent::log 8 => "using Async::Interrupt for race-free signal handling."
1658 if $AnyEvent::VERBOSE >= 8;
1659
1660 $SIGPIPE_R = new Async::Interrupt::EventPipe;
1661 $SIG_IO = AE::io $SIGPIPE_R->fileno, 0, \&_signal_exec;
1662
1663 } else {
1664 AnyEvent::log 8 => "using emulated perl signal handling with latency timer."
1665 if $AnyEvent::VERBOSE >= 8;
1666
1667 if (AnyEvent::WIN32) {
1668 require AnyEvent::Util;
1669
1670 ($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
1671 AnyEvent::Util::fh_nonblocking ($SIGPIPE_R, 1) if $SIGPIPE_R;
1672 AnyEvent::Util::fh_nonblocking ($SIGPIPE_W, 1) if $SIGPIPE_W; # just in case
1673 } else {
1674 pipe $SIGPIPE_R, $SIGPIPE_W;
1675 fcntl $SIGPIPE_R, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_R;
1676 fcntl $SIGPIPE_W, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_W; # just in case
1677
1678 # not strictly required, as $^F is normally 2, but let's make sure...
1679 fcntl $SIGPIPE_R, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
1680 fcntl $SIGPIPE_W, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
1681 }
1682
1683 $SIGPIPE_R
1684 or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
1685
1686 $SIG_IO = AE::io $SIGPIPE_R, 0, \&_signal_exec;
1687 }
1688
1689 *signal = $HAVE_ASYNC_INTERRUPT
1690 ? sub {
1044 my (undef, %arg) = @_; 1691 my (undef, %arg) = @_;
1045 1692
1693 # async::interrupt
1046 my $signal = uc $arg{signal} 1694 my $signal = sig2num $arg{signal};
1047 or Carp::croak "required option 'signal' is missing";
1048
1049 $SIG_CB{$signal}{$arg{cb}} = $arg{cb}; 1695 $SIG_CB{$signal}{$arg{cb}} = $arg{cb};
1696
1697 $SIG_ASY{$signal} ||= new Async::Interrupt
1698 cb => sub { undef $SIG_EV{$signal} },
1699 signal => $signal,
1700 pipe => [$SIGPIPE_R->filenos],
1701 pipe_autodrain => 0,
1702 ;
1703
1704 bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
1705 }
1706 : sub {
1707 my (undef, %arg) = @_;
1708
1709 # pure perl
1710 my $signal = sig2name $arg{signal};
1711 $SIG_CB{$signal}{$arg{cb}} = $arg{cb};
1712
1050 $SIG{$signal} ||= sub { 1713 $SIG{$signal} ||= sub {
1714 local $!;
1715 syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
1716 undef $SIG_EV{$signal};
1717 };
1718
1719 # can't do signal processing without introducing races in pure perl,
1720 # so limit the signal latency.
1721 _sig_add;
1722
1723 bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
1724 }
1725 ;
1726
1727 *AnyEvent::Base::signal::DESTROY = sub {
1728 my ($signal, $cb) = @{$_[0]};
1729
1730 _sig_del;
1731
1732 delete $SIG_CB{$signal}{$cb};
1733
1734 $HAVE_ASYNC_INTERRUPT
1735 ? delete $SIG_ASY{$signal}
1736 : # delete doesn't work with older perls - they then
1737 # print weird messages, or just unconditionally exit
1738 # instead of getting the default action.
1739 undef $SIG{$signal}
1740 unless keys %{ $SIG_CB{$signal} };
1741 };
1742
1743 *_signal_exec = sub {
1744 $HAVE_ASYNC_INTERRUPT
1745 ? $SIGPIPE_R->drain
1746 : sysread $SIGPIPE_R, (my $dummy), 9;
1747
1748 while (%SIG_EV) {
1749 for (keys %SIG_EV) {
1750 delete $SIG_EV{$_};
1051 $_->() for values %{ $SIG_CB{$signal} || {} }; 1751 &$_ for values %{ $SIG_CB{$_} || {} };
1752 }
1753 }
1754 };
1052 }; 1755 };
1756 die if $@;
1053 1757
1054 bless [$signal, $arg{cb}], "AnyEvent::Base::Signal" 1758 &signal
1055}
1056
1057sub AnyEvent::Base::Signal::DESTROY {
1058 my ($signal, $cb) = @{$_[0]};
1059
1060 delete $SIG_CB{$signal}{$cb};
1061
1062 delete $SIG{$signal} unless keys %{ $SIG_CB{$signal} };
1063} 1759}
1064 1760
1065# default implementation for ->child 1761# default implementation for ->child
1066 1762
1067our %PID_CB; 1763our %PID_CB;
1068our $CHLD_W; 1764our $CHLD_W;
1069our $CHLD_DELAY_W; 1765our $CHLD_DELAY_W;
1070our $PID_IDLE;
1071our $WNOHANG;
1072 1766
1073sub _child_wait { 1767# used by many Impl's
1074 while (0 < (my $pid = waitpid -1, $WNOHANG)) { 1768sub _emit_childstatus($$) {
1769 my (undef, $rpid, $rstatus) = @_;
1770
1771 $_->($rpid, $rstatus)
1075 $_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }), 1772 for values %{ $PID_CB{$rpid} || {} },
1076 (values %{ $PID_CB{0} || {} }); 1773 values %{ $PID_CB{0} || {} };
1077 }
1078
1079 undef $PID_IDLE;
1080}
1081
1082sub _sigchld {
1083 # make sure we deliver these changes "synchronous" with the event loop.
1084 $CHLD_DELAY_W ||= AnyEvent->timer (after => 0, cb => sub {
1085 undef $CHLD_DELAY_W;
1086 &_child_wait;
1087 });
1088} 1774}
1089 1775
1090sub child { 1776sub child {
1777 eval q{ # poor man's autoloading {}
1778 *_sigchld = sub {
1779 my $pid;
1780
1781 AnyEvent->_emit_childstatus ($pid, $?)
1782 while ($pid = waitpid -1, WNOHANG) > 0;
1783 };
1784
1785 *child = sub {
1091 my (undef, %arg) = @_; 1786 my (undef, %arg) = @_;
1092 1787
1093 defined (my $pid = $arg{pid} + 0) 1788 my $pid = $arg{pid};
1094 or Carp::croak "required option 'pid' is missing"; 1789 my $cb = $arg{cb};
1095 1790
1096 $PID_CB{$pid}{$arg{cb}} = $arg{cb}; 1791 $PID_CB{$pid}{$cb+0} = $cb;
1097 1792
1098 unless ($WNOHANG) {
1099 $WNOHANG = eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
1100 }
1101
1102 unless ($CHLD_W) { 1793 unless ($CHLD_W) {
1103 $CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld); 1794 $CHLD_W = AE::signal CHLD => \&_sigchld;
1104 # child could be a zombie already, so make at least one round 1795 # child could be a zombie already, so make at least one round
1105 &_sigchld; 1796 &_sigchld;
1106 } 1797 }
1107 1798
1108 bless [$pid, $arg{cb}], "AnyEvent::Base::Child" 1799 bless [$pid, $cb+0], "AnyEvent::Base::child"
1109} 1800 };
1110 1801
1111sub AnyEvent::Base::Child::DESTROY { 1802 *AnyEvent::Base::child::DESTROY = sub {
1112 my ($pid, $cb) = @{$_[0]}; 1803 my ($pid, $icb) = @{$_[0]};
1113 1804
1114 delete $PID_CB{$pid}{$cb}; 1805 delete $PID_CB{$pid}{$icb};
1115 delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} }; 1806 delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
1116 1807
1117 undef $CHLD_W unless keys %PID_CB; 1808 undef $CHLD_W unless keys %PID_CB;
1809 };
1810 };
1811 die if $@;
1812
1813 &child
1814}
1815
1816# idle emulation is done by simply using a timer, regardless
1817# of whether the process is idle or not, and not letting
1818# the callback use more than 50% of the time.
1819sub idle {
1820 eval q{ # poor man's autoloading {}
1821 *idle = sub {
1822 my (undef, %arg) = @_;
1823
1824 my ($cb, $w, $rcb) = $arg{cb};
1825
1826 $rcb = sub {
1827 if ($cb) {
1828 $w = AE::time;
1829 &$cb;
1830 $w = AE::time - $w;
1831
1832 # never use more then 50% of the time for the idle watcher,
1833 # within some limits
1834 $w = 0.0001 if $w < 0.0001;
1835 $w = 5 if $w > 5;
1836
1837 $w = AE::timer $w, 0, $rcb;
1838 } else {
1839 # clean up...
1840 undef $w;
1841 undef $rcb;
1842 }
1843 };
1844
1845 $w = AE::timer 0.05, 0, $rcb;
1846
1847 bless \\$cb, "AnyEvent::Base::idle"
1848 };
1849
1850 *AnyEvent::Base::idle::DESTROY = sub {
1851 undef $${$_[0]};
1852 };
1853 };
1854 die if $@;
1855
1856 &idle
1118} 1857}
1119 1858
1120package AnyEvent::CondVar; 1859package AnyEvent::CondVar;
1121 1860
1122our @ISA = AnyEvent::CondVar::Base::; 1861our @ISA = AnyEvent::CondVar::Base::;
1123 1862
1863# only to be used for subclassing
1864sub new {
1865 my $class = shift;
1866 bless AnyEvent->condvar (@_), $class
1867}
1868
1124package AnyEvent::CondVar::Base; 1869package AnyEvent::CondVar::Base;
1125 1870
1126use overload 1871#use overload
1127 '&{}' => sub { my $self = shift; sub { $self->send (@_) } }, 1872# '&{}' => sub { my $self = shift; sub { $self->send (@_) } },
1128 fallback => 1; 1873# fallback => 1;
1874
1875# save 300+ kilobytes by dirtily hardcoding overloading
1876${"AnyEvent::CondVar::Base::OVERLOAD"}{dummy}++; # Register with magic by touching.
1877*{'AnyEvent::CondVar::Base::()'} = sub { }; # "Make it findable via fetchmethod."
1878*{'AnyEvent::CondVar::Base::(&{}'} = sub { my $self = shift; sub { $self->send (@_) } }; # &{}
1879${'AnyEvent::CondVar::Base::()'} = 1; # fallback
1880
1881our $WAITING;
1129 1882
1130sub _send { 1883sub _send {
1131 # nop 1884 # nop
1885}
1886
1887sub _wait {
1888 AnyEvent->_poll until $_[0]{_ae_sent};
1132} 1889}
1133 1890
1134sub send { 1891sub send {
1135 my $cv = shift; 1892 my $cv = shift;
1136 $cv->{_ae_sent} = [@_]; 1893 $cv->{_ae_sent} = [@_];
1145 1902
1146sub ready { 1903sub ready {
1147 $_[0]{_ae_sent} 1904 $_[0]{_ae_sent}
1148} 1905}
1149 1906
1150sub _wait {
1151 AnyEvent->one_event while !$_[0]{_ae_sent};
1152}
1153
1154sub recv { 1907sub recv {
1908 unless ($_[0]{_ae_sent}) {
1909 $WAITING
1910 and Carp::croak "AnyEvent::CondVar: recursive blocking wait attempted";
1911
1912 local $WAITING = 1;
1155 $_[0]->_wait; 1913 $_[0]->_wait;
1914 }
1156 1915
1157 Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak}; 1916 $_[0]{_ae_croak}
1158 wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0] 1917 and Carp::croak $_[0]{_ae_croak};
1918
1919 wantarray
1920 ? @{ $_[0]{_ae_sent} }
1921 : $_[0]{_ae_sent}[0]
1159} 1922}
1160 1923
1161sub cb { 1924sub cb {
1162 $_[0]{_ae_cb} = $_[1] if @_ > 1; 1925 my $cv = shift;
1926
1927 @_
1928 and $cv->{_ae_cb} = shift
1929 and $cv->{_ae_sent}
1930 and (delete $cv->{_ae_cb})->($cv);
1931
1163 $_[0]{_ae_cb} 1932 $cv->{_ae_cb}
1164} 1933}
1165 1934
1166sub begin { 1935sub begin {
1167 ++$_[0]{_ae_counter}; 1936 ++$_[0]{_ae_counter};
1168 $_[0]{_ae_end_cb} = $_[1] if @_ > 1; 1937 $_[0]{_ae_end_cb} = $_[1] if @_ > 1;
1173 &{ $_[0]{_ae_end_cb} || sub { $_[0]->send } }; 1942 &{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
1174} 1943}
1175 1944
1176# undocumented/compatibility with pre-3.4 1945# undocumented/compatibility with pre-3.4
1177*broadcast = \&send; 1946*broadcast = \&send;
1178*wait = \&_wait; 1947*wait = \&recv;
1948
1949=head1 ERROR AND EXCEPTION HANDLING
1950
1951In general, AnyEvent does not do any error handling - it relies on the
1952caller to do that if required. The L<AnyEvent::Strict> module (see also
1953the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
1954checking of all AnyEvent methods, however, which is highly useful during
1955development.
1956
1957As for exception handling (i.e. runtime errors and exceptions thrown while
1958executing a callback), this is not only highly event-loop specific, but
1959also not in any way wrapped by this module, as this is the job of the main
1960program.
1961
1962The pure perl event loop simply re-throws the exception (usually
1963within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
1964$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
1965so on.
1966
1967=head1 ENVIRONMENT VARIABLES
1968
1969The following environment variables are used by this module or its
1970submodules.
1971
1972Note that AnyEvent will remove I<all> environment variables starting with
1973C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
1974enabled.
1975
1976=over 4
1977
1978=item C<PERL_ANYEVENT_VERBOSE>
1979
1980By default, AnyEvent will be completely silent except in fatal
1981conditions. You can set this environment variable to make AnyEvent more
1982talkative.
1983
1984When set to C<5> or higher, causes AnyEvent to warn about unexpected
1985conditions, such as not being able to load the event model specified by
1986C<PERL_ANYEVENT_MODEL>.
1987
1988When set to C<7> or higher, cause AnyEvent to report to STDERR which event
1989model it chooses.
1990
1991When set to C<8> or higher, then AnyEvent will report extra information on
1992which optional modules it loads and how it implements certain features.
1993
1994=item C<PERL_ANYEVENT_STRICT>
1995
1996AnyEvent does not do much argument checking by default, as thorough
1997argument checking is very costly. Setting this variable to a true value
1998will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
1999check the arguments passed to most method calls. If it finds any problems,
2000it will croak.
2001
2002In other words, enables "strict" mode.
2003
2004Unlike C<use strict> (or its modern cousin, C<< use L<common::sense>
2005>>, it is definitely recommended to keep it off in production. Keeping
2006C<PERL_ANYEVENT_STRICT=1> in your environment while developing programs
2007can be very useful, however.
2008
2009=item C<PERL_ANYEVENT_DEBUG_SHELL>
2010
2011If this env variable is set, then its contents will be interpreted by
2012C<AnyEvent::Socket::parse_hostport> (after replacing every occurance of
2013C<$$> by the process pid) and an C<AnyEvent::Debug::shell> is bound on
2014that port. The shell object is saved in C<$AnyEvent::Debug::SHELL>.
2015
2016This takes place when the first watcher is created.
2017
2018For example, to bind a debug shell on a unix domain socket in
2019F<< /tmp/debug<pid>.sock >>, you could use this:
2020
2021 PERL_ANYEVENT_DEBUG_SHELL=/tmp/debug\$\$.sock perlprog
2022
2023Note that creating sockets in F</tmp> is very unsafe on multiuser
2024systems.
2025
2026=item C<PERL_ANYEVENT_DEBUG_WRAP>
2027
2028Can be set to C<0>, C<1> or C<2> and enables wrapping of all watchers for
2029debugging purposes. See C<AnyEvent::Debug::wrap> for details.
2030
2031=item C<PERL_ANYEVENT_MODEL>
2032
2033This can be used to specify the event model to be used by AnyEvent, before
2034auto detection and -probing kicks in.
2035
2036It normally is a string consisting entirely of ASCII letters (e.g. C<EV>
2037or C<IOAsync>). The string C<AnyEvent::Impl::> gets prepended and the
2038resulting module name is loaded and - if the load was successful - used as
2039event model backend. If it fails to load then AnyEvent will proceed with
2040auto detection and -probing.
2041
2042If the string ends with C<::> instead (e.g. C<AnyEvent::Impl::EV::>) then
2043nothing gets prepended and the module name is used as-is (hint: C<::> at
2044the end of a string designates a module name and quotes it appropriately).
2045
2046For example, to force the pure perl model (L<AnyEvent::Loop::Perl>) you
2047could start your program like this:
2048
2049 PERL_ANYEVENT_MODEL=Perl perl ...
2050
2051=item C<PERL_ANYEVENT_PROTOCOLS>
2052
2053Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
2054for IPv4 or IPv6. The default is unspecified (and might change, or be the result
2055of auto probing).
2056
2057Must be set to a comma-separated list of protocols or address families,
2058current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
2059used, and preference will be given to protocols mentioned earlier in the
2060list.
2061
2062This variable can effectively be used for denial-of-service attacks
2063against local programs (e.g. when setuid), although the impact is likely
2064small, as the program has to handle conenction and other failures anyways.
2065
2066Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
2067but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
2068- only support IPv4, never try to resolve or contact IPv6
2069addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
2070IPv6, but prefer IPv6 over IPv4.
2071
2072=item C<PERL_ANYEVENT_EDNS0>
2073
2074Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
2075for DNS. This extension is generally useful to reduce DNS traffic, but
2076some (broken) firewalls drop such DNS packets, which is why it is off by
2077default.
2078
2079Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
2080EDNS0 in its DNS requests.
2081
2082=item C<PERL_ANYEVENT_MAX_FORKS>
2083
2084The maximum number of child processes that C<AnyEvent::Util::fork_call>
2085will create in parallel.
2086
2087=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
2088
2089The default value for the C<max_outstanding> parameter for the default DNS
2090resolver - this is the maximum number of parallel DNS requests that are
2091sent to the DNS server.
2092
2093=item C<PERL_ANYEVENT_RESOLV_CONF>
2094
2095The file to use instead of F</etc/resolv.conf> (or OS-specific
2096configuration) in the default resolver. When set to the empty string, no
2097default config will be used.
2098
2099=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
2100
2101When neither C<ca_file> nor C<ca_path> was specified during
2102L<AnyEvent::TLS> context creation, and either of these environment
2103variables exist, they will be used to specify CA certificate locations
2104instead of a system-dependent default.
2105
2106=item C<PERL_ANYEVENT_AVOID_GUARD> and C<PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT>
2107
2108When these are set to C<1>, then the respective modules are not
2109loaded. Mostly good for testing AnyEvent itself.
2110
2111=back
1179 2112
1180=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE 2113=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
1181 2114
1182This is an advanced topic that you do not normally need to use AnyEvent in 2115This is an advanced topic that you do not normally need to use AnyEvent in
1183a module. This section is only of use to event loop authors who want to 2116a module. This section is only of use to event loop authors who want to
1217 2150
1218I<rxvt-unicode> also cheats a bit by not providing blocking access to 2151I<rxvt-unicode> also cheats a bit by not providing blocking access to
1219condition variables: code blocking while waiting for a condition will 2152condition variables: code blocking while waiting for a condition will
1220C<die>. This still works with most modules/usages, and blocking calls must 2153C<die>. This still works with most modules/usages, and blocking calls must
1221not be done in an interactive application, so it makes sense. 2154not be done in an interactive application, so it makes sense.
1222
1223=head1 ENVIRONMENT VARIABLES
1224
1225The following environment variables are used by this module:
1226
1227=over 4
1228
1229=item C<PERL_ANYEVENT_VERBOSE>
1230
1231By default, AnyEvent will be completely silent except in fatal
1232conditions. You can set this environment variable to make AnyEvent more
1233talkative.
1234
1235When set to C<1> or higher, causes AnyEvent to warn about unexpected
1236conditions, such as not being able to load the event model specified by
1237C<PERL_ANYEVENT_MODEL>.
1238
1239When set to C<2> or higher, cause AnyEvent to report to STDERR which event
1240model it chooses.
1241
1242=item C<PERL_ANYEVENT_STRICT>
1243
1244AnyEvent does not do much argument checking by default, as thorough
1245argument checking is very costly. Setting this variable to a true value
1246will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
1247check the arguments passed to most method calls. If it finds any problems
1248it will croak.
1249
1250In other words, enables "strict" mode.
1251
1252Unlike C<use strict> it is definitely recommended ot keep it off in
1253production.
1254
1255=item C<PERL_ANYEVENT_MODEL>
1256
1257This can be used to specify the event model to be used by AnyEvent, before
1258auto detection and -probing kicks in. It must be a string consisting
1259entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
1260and the resulting module name is loaded and if the load was successful,
1261used as event model. If it fails to load AnyEvent will proceed with
1262auto detection and -probing.
1263
1264This functionality might change in future versions.
1265
1266For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
1267could start your program like this:
1268
1269 PERL_ANYEVENT_MODEL=Perl perl ...
1270
1271=item C<PERL_ANYEVENT_PROTOCOLS>
1272
1273Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
1274for IPv4 or IPv6. The default is unspecified (and might change, or be the result
1275of auto probing).
1276
1277Must be set to a comma-separated list of protocols or address families,
1278current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
1279used, and preference will be given to protocols mentioned earlier in the
1280list.
1281
1282This variable can effectively be used for denial-of-service attacks
1283against local programs (e.g. when setuid), although the impact is likely
1284small, as the program has to handle connection errors already-
1285
1286Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
1287but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
1288- only support IPv4, never try to resolve or contact IPv6
1289addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
1290IPv6, but prefer IPv6 over IPv4.
1291
1292=item C<PERL_ANYEVENT_EDNS0>
1293
1294Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
1295for DNS. This extension is generally useful to reduce DNS traffic, but
1296some (broken) firewalls drop such DNS packets, which is why it is off by
1297default.
1298
1299Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
1300EDNS0 in its DNS requests.
1301
1302=item C<PERL_ANYEVENT_MAX_FORKS>
1303
1304The maximum number of child processes that C<AnyEvent::Util::fork_call>
1305will create in parallel.
1306
1307=back
1308 2155
1309=head1 EXAMPLE PROGRAM 2156=head1 EXAMPLE PROGRAM
1310 2157
1311The following program uses an I/O watcher to read data from STDIN, a timer 2158The following program uses an I/O watcher to read data from STDIN, a timer
1312to display a message once per second, and a condition variable to quit the 2159to display a message once per second, and a condition variable to quit the
1325 warn "read: $input\n"; # output what has been read 2172 warn "read: $input\n"; # output what has been read
1326 $cv->send if $input =~ /^q/i; # quit program if /^q/i 2173 $cv->send if $input =~ /^q/i; # quit program if /^q/i
1327 }, 2174 },
1328 ); 2175 );
1329 2176
1330 my $time_watcher; # can only be used once
1331
1332 sub new_timer {
1333 $timer = AnyEvent->timer (after => 1, cb => sub { 2177 my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub {
1334 warn "timeout\n"; # print 'timeout' about every second 2178 warn "timeout\n"; # print 'timeout' at most every second
1335 &new_timer; # and restart the time
1336 }); 2179 });
1337 }
1338
1339 new_timer; # create first timer
1340 2180
1341 $cv->recv; # wait until user enters /^q/i 2181 $cv->recv; # wait until user enters /^q/i
1342 2182
1343=head1 REAL-WORLD EXAMPLE 2183=head1 REAL-WORLD EXAMPLE
1344 2184
1417 2257
1418The actual code goes further and collects all errors (C<die>s, exceptions) 2258The actual code goes further and collects all errors (C<die>s, exceptions)
1419that occurred during request processing. The C<result> method detects 2259that occurred during request processing. The C<result> method detects
1420whether an exception as thrown (it is stored inside the $txn object) 2260whether an exception as thrown (it is stored inside the $txn object)
1421and just throws the exception, which means connection errors and other 2261and just throws the exception, which means connection errors and other
1422problems get reported tot he code that tries to use the result, not in a 2262problems get reported to the code that tries to use the result, not in a
1423random callback. 2263random callback.
1424 2264
1425All of this enables the following usage styles: 2265All of this enables the following usage styles:
1426 2266
14271. Blocking: 22671. Blocking:
1475through AnyEvent. The benchmark creates a lot of timers (with a zero 2315through AnyEvent. The benchmark creates a lot of timers (with a zero
1476timeout) and I/O watchers (watching STDOUT, a pty, to become writable, 2316timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
1477which it is), lets them fire exactly once and destroys them again. 2317which it is), lets them fire exactly once and destroys them again.
1478 2318
1479Source code for this benchmark is found as F<eg/bench> in the AnyEvent 2319Source code for this benchmark is found as F<eg/bench> in the AnyEvent
1480distribution. 2320distribution. It uses the L<AE> interface, which makes a real difference
2321for the EV and Perl backends only.
1481 2322
1482=head3 Explanation of the columns 2323=head3 Explanation of the columns
1483 2324
1484I<watcher> is the number of event watchers created/destroyed. Since 2325I<watcher> is the number of event watchers created/destroyed. Since
1485different event models feature vastly different performances, each event 2326different event models feature vastly different performances, each event
1506watcher. 2347watcher.
1507 2348
1508=head3 Results 2349=head3 Results
1509 2350
1510 name watchers bytes create invoke destroy comment 2351 name watchers bytes create invoke destroy comment
1511 EV/EV 400000 244 0.56 0.46 0.31 EV native interface 2352 EV/EV 100000 223 0.47 0.43 0.27 EV native interface
1512 EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers 2353 EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers
1513 CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal 2354 Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal
1514 Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation 2355 Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation
1515 Event/Event 16000 516 31.88 31.30 0.85 Event native interface 2356 Event/Event 16000 516 31.16 31.84 0.82 Event native interface
1516 Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers 2357 Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers
2358 IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll
2359 IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll
1517 Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour 2360 Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour
1518 Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers 2361 Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers
1519 POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event 2362 POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event
1520 POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select 2363 POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select
1521 2364
1522=head3 Discussion 2365=head3 Discussion
1523 2366
1524The benchmark does I<not> measure scalability of the event loop very 2367The benchmark does I<not> measure scalability of the event loop very
1525well. For example, a select-based event loop (such as the pure perl one) 2368well. For example, a select-based event loop (such as the pure perl one)
1537benchmark machine, handling an event takes roughly 1600 CPU cycles with 2380benchmark machine, handling an event takes roughly 1600 CPU cycles with
1538EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU 2381EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
1539cycles with POE. 2382cycles with POE.
1540 2383
1541C<EV> is the sole leader regarding speed and memory use, which are both 2384C<EV> is the sole leader regarding speed and memory use, which are both
1542maximal/minimal, respectively. Even when going through AnyEvent, it uses 2385maximal/minimal, respectively. When using the L<AE> API there is zero
2386overhead (when going through the AnyEvent API create is about 5-6 times
2387slower, with other times being equal, so still uses far less memory than
1543far less memory than any other event loop and is still faster than Event 2388any other event loop and is still faster than Event natively).
1544natively.
1545 2389
1546The pure perl implementation is hit in a few sweet spots (both the 2390The pure perl implementation is hit in a few sweet spots (both the
1547constant timeout and the use of a single fd hit optimisations in the perl 2391constant timeout and the use of a single fd hit optimisations in the perl
1548interpreter and the backend itself). Nevertheless this shows that it 2392interpreter and the backend itself). Nevertheless this shows that it
1549adds very little overhead in itself. Like any select-based backend its 2393adds very little overhead in itself. Like any select-based backend its
1550performance becomes really bad with lots of file descriptors (and few of 2394performance becomes really bad with lots of file descriptors (and few of
1551them active), of course, but this was not subject of this benchmark. 2395them active), of course, but this was not subject of this benchmark.
1552 2396
1553The C<Event> module has a relatively high setup and callback invocation 2397The C<Event> module has a relatively high setup and callback invocation
1554cost, but overall scores in on the third place. 2398cost, but overall scores in on the third place.
2399
2400C<IO::Async> performs admirably well, about on par with C<Event>, even
2401when using its pure perl backend.
1555 2402
1556C<Glib>'s memory usage is quite a bit higher, but it features a 2403C<Glib>'s memory usage is quite a bit higher, but it features a
1557faster callback invocation and overall ends up in the same class as 2404faster callback invocation and overall ends up in the same class as
1558C<Event>. However, Glib scales extremely badly, doubling the number of 2405C<Event>. However, Glib scales extremely badly, doubling the number of
1559watchers increases the processing time by more than a factor of four, 2406watchers increases the processing time by more than a factor of four,
1594(even when used without AnyEvent), but most event loops have acceptable 2441(even when used without AnyEvent), but most event loops have acceptable
1595performance with or without AnyEvent. 2442performance with or without AnyEvent.
1596 2443
1597=item * The overhead AnyEvent adds is usually much smaller than the overhead of 2444=item * The overhead AnyEvent adds is usually much smaller than the overhead of
1598the actual event loop, only with extremely fast event loops such as EV 2445the actual event loop, only with extremely fast event loops such as EV
1599adds AnyEvent significant overhead. 2446does AnyEvent add significant overhead.
1600 2447
1601=item * You should avoid POE like the plague if you want performance or 2448=item * You should avoid POE like the plague if you want performance or
1602reasonable memory usage. 2449reasonable memory usage.
1603 2450
1604=back 2451=back
1620In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100 2467In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
1621(1%) are active. This mirrors the activity of large servers with many 2468(1%) are active. This mirrors the activity of large servers with many
1622connections, most of which are idle at any one point in time. 2469connections, most of which are idle at any one point in time.
1623 2470
1624Source code for this benchmark is found as F<eg/bench2> in the AnyEvent 2471Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
1625distribution. 2472distribution. It uses the L<AE> interface, which makes a real difference
2473for the EV and Perl backends only.
1626 2474
1627=head3 Explanation of the columns 2475=head3 Explanation of the columns
1628 2476
1629I<sockets> is the number of sockets, and twice the number of "servers" (as 2477I<sockets> is the number of sockets, and twice the number of "servers" (as
1630each server has a read and write socket end). 2478each server has a read and write socket end).
1637it to another server. This includes deleting the old timeout and creating 2485it to another server. This includes deleting the old timeout and creating
1638a new one that moves the timeout into the future. 2486a new one that moves the timeout into the future.
1639 2487
1640=head3 Results 2488=head3 Results
1641 2489
1642 name sockets create request 2490 name sockets create request
1643 EV 20000 69.01 11.16 2491 EV 20000 62.66 7.99
1644 Perl 20000 73.32 35.87 2492 Perl 20000 68.32 32.64
1645 Event 20000 212.62 257.32 2493 IOAsync 20000 174.06 101.15 epoll
1646 Glib 20000 651.16 1896.30 2494 IOAsync 20000 174.67 610.84 poll
2495 Event 20000 202.69 242.91
2496 Glib 20000 557.01 1689.52
1647 POE 20000 349.67 12317.24 uses POE::Loop::Event 2497 POE 20000 341.54 12086.32 uses POE::Loop::Event
1648 2498
1649=head3 Discussion 2499=head3 Discussion
1650 2500
1651This benchmark I<does> measure scalability and overall performance of the 2501This benchmark I<does> measure scalability and overall performance of the
1652particular event loop. 2502particular event loop.
1654EV is again fastest. Since it is using epoll on my system, the setup time 2504EV is again fastest. Since it is using epoll on my system, the setup time
1655is relatively high, though. 2505is relatively high, though.
1656 2506
1657Perl surprisingly comes second. It is much faster than the C-based event 2507Perl surprisingly comes second. It is much faster than the C-based event
1658loops Event and Glib. 2508loops Event and Glib.
2509
2510IO::Async performs very well when using its epoll backend, and still quite
2511good compared to Glib when using its pure perl backend.
1659 2512
1660Event suffers from high setup time as well (look at its code and you will 2513Event suffers from high setup time as well (look at its code and you will
1661understand why). Callback invocation also has a high overhead compared to 2514understand why). Callback invocation also has a high overhead compared to
1662the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event 2515the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
1663uses select or poll in basically all documented configurations. 2516uses select or poll in basically all documented configurations.
1726=item * C-based event loops perform very well with small number of 2579=item * C-based event loops perform very well with small number of
1727watchers, as the management overhead dominates. 2580watchers, as the management overhead dominates.
1728 2581
1729=back 2582=back
1730 2583
2584=head2 THE IO::Lambda BENCHMARK
2585
2586Recently I was told about the benchmark in the IO::Lambda manpage, which
2587could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
2588simply compares IO::Lambda with POE, and IO::Lambda looks better (which
2589shouldn't come as a surprise to anybody). As such, the benchmark is
2590fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
2591very optimal. But how would AnyEvent compare when used without the extra
2592baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
2593
2594The benchmark itself creates an echo-server, and then, for 500 times,
2595connects to the echo server, sends a line, waits for the reply, and then
2596creates the next connection. This is a rather bad benchmark, as it doesn't
2597test the efficiency of the framework or much non-blocking I/O, but it is a
2598benchmark nevertheless.
2599
2600 name runtime
2601 Lambda/select 0.330 sec
2602 + optimized 0.122 sec
2603 Lambda/AnyEvent 0.327 sec
2604 + optimized 0.138 sec
2605 Raw sockets/select 0.077 sec
2606 POE/select, components 0.662 sec
2607 POE/select, raw sockets 0.226 sec
2608 POE/select, optimized 0.404 sec
2609
2610 AnyEvent/select/nb 0.085 sec
2611 AnyEvent/EV/nb 0.068 sec
2612 +state machine 0.134 sec
2613
2614The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
2615benchmarks actually make blocking connects and use 100% blocking I/O,
2616defeating the purpose of an event-based solution. All of the newly
2617written AnyEvent benchmarks use 100% non-blocking connects (using
2618AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
2619resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
2620generally require a lot more bookkeeping and event handling than blocking
2621connects (which involve a single syscall only).
2622
2623The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
2624offers similar expressive power as POE and IO::Lambda, using conventional
2625Perl syntax. This means that both the echo server and the client are 100%
2626non-blocking, further placing it at a disadvantage.
2627
2628As you can see, the AnyEvent + EV combination even beats the
2629hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
2630backend easily beats IO::Lambda and POE.
2631
2632And even the 100% non-blocking version written using the high-level (and
2633slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda
2634higher level ("unoptimised") abstractions by a large margin, even though
2635it does all of DNS, tcp-connect and socket I/O in a non-blocking way.
2636
2637The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
2638F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
2639part of the IO::Lambda distribution and were used without any changes.
2640
2641
2642=head1 SIGNALS
2643
2644AnyEvent currently installs handlers for these signals:
2645
2646=over 4
2647
2648=item SIGCHLD
2649
2650A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
2651emulation for event loops that do not support them natively. Also, some
2652event loops install a similar handler.
2653
2654Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE, then
2655AnyEvent will reset it to default, to avoid losing child exit statuses.
2656
2657=item SIGPIPE
2658
2659A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
2660when AnyEvent gets loaded.
2661
2662The rationale for this is that AnyEvent users usually do not really depend
2663on SIGPIPE delivery (which is purely an optimisation for shell use, or
2664badly-written programs), but C<SIGPIPE> can cause spurious and rare
2665program exits as a lot of people do not expect C<SIGPIPE> when writing to
2666some random socket.
2667
2668The rationale for installing a no-op handler as opposed to ignoring it is
2669that this way, the handler will be restored to defaults on exec.
2670
2671Feel free to install your own handler, or reset it to defaults.
2672
2673=back
2674
2675=cut
2676
2677undef $SIG{CHLD}
2678 if $SIG{CHLD} eq 'IGNORE';
2679
2680$SIG{PIPE} = sub { }
2681 unless defined $SIG{PIPE};
2682
2683=head1 RECOMMENDED/OPTIONAL MODULES
2684
2685One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
2686its built-in modules) are required to use it.
2687
2688That does not mean that AnyEvent won't take advantage of some additional
2689modules if they are installed.
2690
2691This section explains which additional modules will be used, and how they
2692affect AnyEvent's operation.
2693
2694=over 4
2695
2696=item L<Async::Interrupt>
2697
2698This slightly arcane module is used to implement fast signal handling: To
2699my knowledge, there is no way to do completely race-free and quick
2700signal handling in pure perl. To ensure that signals still get
2701delivered, AnyEvent will start an interval timer to wake up perl (and
2702catch the signals) with some delay (default is 10 seconds, look for
2703C<$AnyEvent::MAX_SIGNAL_LATENCY>).
2704
2705If this module is available, then it will be used to implement signal
2706catching, which means that signals will not be delayed, and the event loop
2707will not be interrupted regularly, which is more efficient (and good for
2708battery life on laptops).
2709
2710This affects not just the pure-perl event loop, but also other event loops
2711that have no signal handling on their own (e.g. Glib, Tk, Qt).
2712
2713Some event loops (POE, Event, Event::Lib) offer signal watchers natively,
2714and either employ their own workarounds (POE) or use AnyEvent's workaround
2715(using C<$AnyEvent::MAX_SIGNAL_LATENCY>). Installing L<Async::Interrupt>
2716does nothing for those backends.
2717
2718=item L<EV>
2719
2720This module isn't really "optional", as it is simply one of the backend
2721event loops that AnyEvent can use. However, it is simply the best event
2722loop available in terms of features, speed and stability: It supports
2723the AnyEvent API optimally, implements all the watcher types in XS, does
2724automatic timer adjustments even when no monotonic clock is available,
2725can take avdantage of advanced kernel interfaces such as C<epoll> and
2726C<kqueue>, and is the fastest backend I<by far>. You can even embed
2727L<Glib>/L<Gtk2> in it (or vice versa, see L<EV::Glib> and L<Glib::EV>).
2728
2729If you only use backends that rely on another event loop (e.g. C<Tk>),
2730then this module will do nothing for you.
2731
2732=item L<Guard>
2733
2734The guard module, when used, will be used to implement
2735C<AnyEvent::Util::guard>. This speeds up guards considerably (and uses a
2736lot less memory), but otherwise doesn't affect guard operation much. It is
2737purely used for performance.
2738
2739=item L<JSON> and L<JSON::XS>
2740
2741One of these modules is required when you want to read or write JSON data
2742via L<AnyEvent::Handle>. L<JSON> is also written in pure-perl, but can take
2743advantage of the ultra-high-speed L<JSON::XS> module when it is installed.
2744
2745=item L<Net::SSLeay>
2746
2747Implementing TLS/SSL in Perl is certainly interesting, but not very
2748worthwhile: If this module is installed, then L<AnyEvent::Handle> (with
2749the help of L<AnyEvent::TLS>), gains the ability to do TLS/SSL.
2750
2751=item L<Time::HiRes>
2752
2753This module is part of perl since release 5.008. It will be used when the
2754chosen event library does not come with a timing source of its own. The
2755pure-perl event loop (L<AnyEvent::Loop>) will additionally load it to
2756try to use a monotonic clock for timing stability.
2757
2758=back
2759
1731 2760
1732=head1 FORK 2761=head1 FORK
1733 2762
1734Most event libraries are not fork-safe. The ones who are usually are 2763Most event libraries are not fork-safe. The ones who are usually are
1735because they rely on inefficient but fork-safe C<select> or C<poll> 2764because they rely on inefficient but fork-safe C<select> or C<poll> calls
1736calls. Only L<EV> is fully fork-aware. 2765- higher performance APIs such as BSD's kqueue or the dreaded Linux epoll
2766are usually badly thought-out hacks that are incompatible with fork in
2767one way or another. Only L<EV> is fully fork-aware and ensures that you
2768continue event-processing in both parent and child (or both, if you know
2769what you are doing).
2770
2771This means that, in general, you cannot fork and do event processing in
2772the child if the event library was initialised before the fork (which
2773usually happens when the first AnyEvent watcher is created, or the library
2774is loaded).
1737 2775
1738If you have to fork, you must either do so I<before> creating your first 2776If you have to fork, you must either do so I<before> creating your first
1739watcher OR you must not use AnyEvent at all in the child. 2777watcher OR you must not use AnyEvent at all in the child OR you must do
2778something completely out of the scope of AnyEvent.
2779
2780The problem of doing event processing in the parent I<and> the child
2781is much more complicated: even for backends that I<are> fork-aware or
2782fork-safe, their behaviour is not usually what you want: fork clones all
2783watchers, that means all timers, I/O watchers etc. are active in both
2784parent and child, which is almost never what you want. USing C<exec>
2785to start worker children from some kind of manage rprocess is usually
2786preferred, because it is much easier and cleaner, at the expense of having
2787to have another binary.
1740 2788
1741 2789
1742=head1 SECURITY CONSIDERATIONS 2790=head1 SECURITY CONSIDERATIONS
1743 2791
1744AnyEvent can be forced to load any event model via 2792AnyEvent can be forced to load any event model via
1756 use AnyEvent; 2804 use AnyEvent;
1757 2805
1758Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can 2806Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
1759be used to probe what backend is used and gain other information (which is 2807be used to probe what backend is used and gain other information (which is
1760probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and 2808probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
1761$ENV{PERL_ANYEGENT_STRICT}. 2809$ENV{PERL_ANYEVENT_STRICT}.
2810
2811Note that AnyEvent will remove I<all> environment variables starting with
2812C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
2813enabled.
1762 2814
1763 2815
1764=head1 BUGS 2816=head1 BUGS
1765 2817
1766Perl 5.8 has numerous memleaks that sometimes hit this module and are hard 2818Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
1767to work around. If you suffer from memleaks, first upgrade to Perl 5.10 2819to work around. If you suffer from memleaks, first upgrade to Perl 5.10
1768and check wether the leaks still show up. (Perl 5.10.0 has other annoying 2820and check wether the leaks still show up. (Perl 5.10.0 has other annoying
1769mamleaks, such as leaking on C<map> and C<grep> but it is usually not as 2821memleaks, such as leaking on C<map> and C<grep> but it is usually not as
1770pronounced). 2822pronounced).
1771 2823
1772 2824
1773=head1 SEE ALSO 2825=head1 SEE ALSO
1774 2826
1775Utility functions: L<AnyEvent::Util>. 2827Tutorial/Introduction: L<AnyEvent::Intro>.
1776 2828
1777Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>, 2829FAQ: L<AnyEvent::FAQ>.
1778L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>. 2830
2831Utility functions: L<AnyEvent::Util> (misc. grab-bag), L<AnyEvent::Log>
2832(simply logging).
2833
2834Development/Debugging: L<AnyEvent::Strict> (stricter checking),
2835L<AnyEvent::Debug> (interactive shell, watcher tracing).
2836
2837Supported event modules: L<AnyEvent::Loop>, L<EV>, L<EV::Glib>,
2838L<Glib::EV>, L<Event>, L<Glib::Event>, L<Glib>, L<Tk>, L<Event::Lib>,
2839L<Qt>, L<POE>, L<FLTK>.
1779 2840
1780Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>, 2841Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
1781L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, 2842L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
1782L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>, 2843L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
2844L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>, L<Anyevent::Impl::Irssi>,
1783L<AnyEvent::Impl::POE>. 2845L<AnyEvent::Impl::FLTK>.
1784 2846
1785Non-blocking file handles, sockets, TCP clients and 2847Non-blocking handles, pipes, stream sockets, TCP clients and
1786servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>. 2848servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>.
1787 2849
1788Asynchronous DNS: L<AnyEvent::DNS>. 2850Asynchronous DNS: L<AnyEvent::DNS>.
1789 2851
1790Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>, 2852Thread support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>.
1791 2853
1792Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>. 2854Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::IRC>,
2855L<AnyEvent::HTTP>.
1793 2856
1794 2857
1795=head1 AUTHOR 2858=head1 AUTHOR
1796 2859
1797 Marc Lehmann <schmorp@schmorp.de> 2860 Marc Lehmann <schmorp@schmorp.de>

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines