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 1.142 by root, Tue May 27 02:34:30 2008 UTC vs.
Revision 1.318 by root, Wed Mar 24 23:28:06 2010 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 L<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 { ... });
17
18 # one-shot or repeating timers
19 my $w = AnyEvent->timer (after => $seconds, cb => sub { ... });
20 my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ...
21
22 print AnyEvent->now; # prints current event loop time
23 print AnyEvent->time; # think Time::HiRes::time or simply CORE::time.
24
25 # POSIX signal
26 my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... });
27
28 # child process exit
29 my $w = AnyEvent->child (pid => $pid, cb => sub {
30 my ($pid, $status) = @_;
12 ... 31 ...
13 }); 32 });
14 33
15 my $w = AnyEvent->timer (after => $seconds, cb => sub { 34 # called when event loop idle (if applicable)
16 ... 35 my $w = AnyEvent->idle (cb => sub { ... });
17 });
18 36
19 my $w = AnyEvent->condvar; # stores whether a condition was flagged 37 my $w = AnyEvent->condvar; # stores whether a condition was flagged
20 $w->send; # wake up current and all future recv's 38 $w->send; # wake up current and all future recv's
21 $w->recv; # enters "main loop" till $condvar gets ->send 39 $w->recv; # enters "main loop" till $condvar gets ->send
40 # use a condvar in callback mode:
41 $w->cb (sub { $_[0]->recv });
42
43=head1 INTRODUCTION/TUTORIAL
44
45This manpage is mainly a reference manual. If you are interested
46in a tutorial or some gentle introduction, have a look at the
47L<AnyEvent::Intro> manpage.
48
49=head1 SUPPORT
50
51There is a mailinglist for discussing all things AnyEvent, and an IRC
52channel, too.
53
54See the AnyEvent project page at the B<Schmorpforge Ta-Sa Software
55Repository>, at L<http://anyevent.schmorp.de>, for more info.
22 56
23=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT) 57=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
24 58
25Glib, POE, IO::Async, Event... CPAN offers event models by the dozen 59Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
26nowadays. So what is different about AnyEvent? 60nowadays. So what is different about AnyEvent?
27 61
28Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of 62Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
29policy> and AnyEvent is I<small and efficient>. 63policy> and AnyEvent is I<small and efficient>.
30 64
31First and foremost, I<AnyEvent is not an event model> itself, it only 65First and foremost, I<AnyEvent is not an event model> itself, it only
32interfaces to whatever event model the main program happens to use in a 66interfaces to whatever event model the main program happens to use, in a
33pragmatic way. For event models and certain classes of immortals alike, 67pragmatic way. For event models and certain classes of immortals alike,
34the statement "there can only be one" is a bitter reality: In general, 68the statement "there can only be one" is a bitter reality: In general,
35only one event loop can be active at the same time in a process. AnyEvent 69only one event loop can be active at the same time in a process. AnyEvent
36helps hiding the differences between those event loops. 70cannot change this, but it can hide the differences between those event
71loops.
37 72
38The goal of AnyEvent is to offer module authors the ability to do event 73The goal of AnyEvent is to offer module authors the ability to do event
39programming (waiting for I/O or timer events) without subscribing to a 74programming (waiting for I/O or timer events) without subscribing to a
40religion, a way of living, and most importantly: without forcing your 75religion, a way of living, and most importantly: without forcing your
41module users into the same thing by forcing them to use the same event 76module users into the same thing by forcing them to use the same event
42model you use. 77model you use.
43 78
44For modules like POE or IO::Async (which is a total misnomer as it is 79For modules like POE or IO::Async (which is a total misnomer as it is
45actually doing all I/O I<synchronously>...), using them in your module is 80actually doing all I/O I<synchronously>...), using them in your module is
46like joining a cult: After you joined, you are dependent on them and you 81like joining a cult: After you joined, you are dependent on them and you
47cannot use anything else, as it is simply incompatible to everything that 82cannot use anything else, as they are simply incompatible to everything
48isn't itself. What's worse, all the potential users of your module are 83that isn't them. What's worse, all the potential users of your
49I<also> forced to use the same event loop you use. 84module are I<also> forced to use the same event loop you use.
50 85
51AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works 86AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
52fine. AnyEvent + Tk works fine etc. etc. but none of these work together 87fine. AnyEvent + Tk works fine etc. etc. but none of these work together
53with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if 88with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
54your module uses one of those, every user of your module has to use it, 89your module uses one of those, every user of your module has to use it,
55too. But if your module uses AnyEvent, it works transparently with all 90too. But if your module uses AnyEvent, it works transparently with all
56event models it supports (including stuff like POE and IO::Async, as long 91event models it supports (including stuff like IO::Async, as long as those
57as those use one of the supported event loops. It is trivial to add new 92use one of the supported event loops. It is trivial to add new event loops
58event loops to AnyEvent, too, so it is future-proof). 93to AnyEvent, too, so it is future-proof).
59 94
60In addition to being free of having to use I<the one and only true event 95In addition to being free of having to use I<the one and only true event
61model>, AnyEvent also is free of bloat and policy: with POE or similar 96model>, AnyEvent also is free of bloat and policy: with POE or similar
62modules, you get an enormous amount of code and strict rules you have to 97modules, you get an enormous amount of code and strict rules you have to
63follow. AnyEvent, on the other hand, is lean and up to the point, by only 98follow. AnyEvent, on the other hand, is lean and up to the point, by only
121These watchers are normal Perl objects with normal Perl lifetime. After 156These watchers are normal Perl objects with normal Perl lifetime. After
122creating a watcher it will immediately "watch" for events and invoke the 157creating a watcher it will immediately "watch" for events and invoke the
123callback when the event occurs (of course, only when the event model 158callback when the event occurs (of course, only when the event model
124is in control). 159is in control).
125 160
161Note that B<callbacks must not permanently change global variables>
162potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
163callbacks must not C<die> >>. The former is good programming practise in
164Perl and the latter stems from the fact that exception handling differs
165widely between event loops.
166
126To disable the watcher you have to destroy it (e.g. by setting the 167To disable the watcher you have to destroy it (e.g. by setting the
127variable you store it in to C<undef> or otherwise deleting all references 168variable you store it in to C<undef> or otherwise deleting all references
128to it). 169to it).
129 170
130All watchers are created by calling a method on the C<AnyEvent> class. 171All watchers are created by calling a method on the C<AnyEvent> class.
132Many watchers either are used with "recursion" (repeating timers for 173Many watchers either are used with "recursion" (repeating timers for
133example), or need to refer to their watcher object in other ways. 174example), or need to refer to their watcher object in other ways.
134 175
135An any way to achieve that is this pattern: 176An any way to achieve that is this pattern:
136 177
137 my $w; $w = AnyEvent->type (arg => value ..., cb => sub { 178 my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
138 # you can use $w here, for example to undef it 179 # you can use $w here, for example to undef it
139 undef $w; 180 undef $w;
140 }); 181 });
141 182
142Note that C<my $w; $w => combination. This is necessary because in Perl, 183Note that C<my $w; $w => combination. This is necessary because in Perl,
143my variables are only visible after the statement in which they are 184my variables are only visible after the statement in which they are
144declared. 185declared.
145 186
146=head2 I/O WATCHERS 187=head2 I/O WATCHERS
147 188
189 $w = AnyEvent->io (
190 fh => <filehandle_or_fileno>,
191 poll => <"r" or "w">,
192 cb => <callback>,
193 );
194
148You can create an I/O watcher by calling the C<< AnyEvent->io >> method 195You can create an I/O watcher by calling the C<< AnyEvent->io >> method
149with the following mandatory key-value pairs as arguments: 196with the following mandatory key-value pairs as arguments:
150 197
151C<fh> the Perl I<file handle> (I<not> file descriptor) to watch 198C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch
199for events (AnyEvent might or might not keep a reference to this file
200handle). Note that only file handles pointing to things for which
201non-blocking operation makes sense are allowed. This includes sockets,
202most character devices, pipes, fifos and so on, but not for example files
203or block devices.
204
152for events. C<poll> must be a string that is either C<r> or C<w>, 205C<poll> must be a string that is either C<r> or C<w>, which creates a
153which creates a watcher waiting for "r"eadable or "w"ritable events, 206watcher waiting for "r"eadable or "w"ritable events, respectively.
207
154respectively. C<cb> is the callback to invoke each time the file handle 208C<cb> is the callback to invoke each time the file handle becomes ready.
155becomes ready.
156 209
157Although the callback might get passed parameters, their value and 210Although the callback might get passed parameters, their value and
158presence is undefined and you cannot rely on them. Portable AnyEvent 211presence is undefined and you cannot rely on them. Portable AnyEvent
159callbacks cannot use arguments passed to I/O watcher callbacks. 212callbacks cannot use arguments passed to I/O watcher callbacks.
160 213
164 217
165Some event loops issue spurious readyness notifications, so you should 218Some event loops issue spurious readyness notifications, so you should
166always use non-blocking calls when reading/writing from/to your file 219always use non-blocking calls when reading/writing from/to your file
167handles. 220handles.
168 221
169Example:
170
171 # wait for readability of STDIN, then read a line and disable the watcher 222Example: wait for readability of STDIN, then read a line and disable the
223watcher.
224
172 my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub { 225 my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
173 chomp (my $input = <STDIN>); 226 chomp (my $input = <STDIN>);
174 warn "read: $input\n"; 227 warn "read: $input\n";
175 undef $w; 228 undef $w;
176 }); 229 });
177 230
178=head2 TIME WATCHERS 231=head2 TIME WATCHERS
179 232
233 $w = AnyEvent->timer (after => <seconds>, cb => <callback>);
234
235 $w = AnyEvent->timer (
236 after => <fractional_seconds>,
237 interval => <fractional_seconds>,
238 cb => <callback>,
239 );
240
180You can create a time watcher by calling the C<< AnyEvent->timer >> 241You can create a time watcher by calling the C<< AnyEvent->timer >>
181method with the following mandatory arguments: 242method with the following mandatory arguments:
182 243
183C<after> specifies after how many seconds (fractional values are 244C<after> specifies after how many seconds (fractional values are
184supported) the callback should be invoked. C<cb> is the callback to invoke 245supported) the callback should be invoked. C<cb> is the callback to invoke
186 247
187Although the callback might get passed parameters, their value and 248Although the callback might get passed parameters, their value and
188presence is undefined and you cannot rely on them. Portable AnyEvent 249presence is undefined and you cannot rely on them. Portable AnyEvent
189callbacks cannot use arguments passed to time watcher callbacks. 250callbacks cannot use arguments passed to time watcher callbacks.
190 251
191The timer callback will be invoked at most once: if you want a repeating 252The callback will normally be invoked once only. If you specify another
192timer you have to create a new watcher (this is a limitation by both Tk 253parameter, C<interval>, as a strictly positive number (> 0), then the
193and Glib). 254callback will be invoked regularly at that interval (in fractional
255seconds) after the first invocation. If C<interval> is specified with a
256false value, then it is treated as if it were missing.
194 257
195Example: 258The callback will be rescheduled before invoking the callback, but no
259attempt is done to avoid timer drift in most backends, so the interval is
260only approximate.
196 261
197 # fire an event after 7.7 seconds 262Example: fire an event after 7.7 seconds.
263
198 my $w = AnyEvent->timer (after => 7.7, cb => sub { 264 my $w = AnyEvent->timer (after => 7.7, cb => sub {
199 warn "timeout\n"; 265 warn "timeout\n";
200 }); 266 });
201 267
202 # to cancel the timer: 268 # to cancel the timer:
203 undef $w; 269 undef $w;
204 270
205Example 2:
206
207 # fire an event after 0.5 seconds, then roughly every second 271Example 2: fire an event after 0.5 seconds, then roughly every second.
208 my $w;
209 272
210 my $cb = sub {
211 # cancel the old timer while creating a new one
212 $w = AnyEvent->timer (after => 1, cb => $cb); 273 my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
274 warn "timeout\n";
213 }; 275 };
214
215 # start the "loop" by creating the first watcher
216 $w = AnyEvent->timer (after => 0.5, cb => $cb);
217 276
218=head3 TIMING ISSUES 277=head3 TIMING ISSUES
219 278
220There are two ways to handle timers: based on real time (relative, "fire 279There are two ways to handle timers: based on real time (relative, "fire
221in 10 seconds") and based on wallclock time (absolute, "fire at 12 280in 10 seconds") and based on wallclock time (absolute, "fire at 12
233timers. 292timers.
234 293
235AnyEvent always prefers relative timers, if available, matching the 294AnyEvent always prefers relative timers, if available, matching the
236AnyEvent API. 295AnyEvent API.
237 296
297AnyEvent has two additional methods that return the "current time":
298
299=over 4
300
301=item AnyEvent->time
302
303This returns the "current wallclock time" as a fractional number of
304seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
305return, and the result is guaranteed to be compatible with those).
306
307It progresses independently of any event loop processing, i.e. each call
308will check the system clock, which usually gets updated frequently.
309
310=item AnyEvent->now
311
312This also returns the "current wallclock time", but unlike C<time>, above,
313this value might change only once per event loop iteration, depending on
314the event loop (most return the same time as C<time>, above). This is the
315time that AnyEvent's timers get scheduled against.
316
317I<In almost all cases (in all cases if you don't care), this is the
318function to call when you want to know the current time.>
319
320This function is also often faster then C<< AnyEvent->time >>, and
321thus the preferred method if you want some timestamp (for example,
322L<AnyEvent::Handle> uses this to update it's activity timeouts).
323
324The rest of this section is only of relevance if you try to be very exact
325with your timing, you can skip it without bad conscience.
326
327For a practical example of when these times differ, consider L<Event::Lib>
328and L<EV> and the following set-up:
329
330The event loop is running and has just invoked one of your callback at
331time=500 (assume no other callbacks delay processing). In your callback,
332you wait a second by executing C<sleep 1> (blocking the process for a
333second) and then (at time=501) you create a relative timer that fires
334after three seconds.
335
336With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
337both return C<501>, because that is the current time, and the timer will
338be scheduled to fire at time=504 (C<501> + C<3>).
339
340With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
341time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
342last event processing phase started. With L<EV>, your timer gets scheduled
343to run at time=503 (C<500> + C<3>).
344
345In one sense, L<Event::Lib> is more exact, as it uses the current time
346regardless of any delays introduced by event processing. However, most
347callbacks do not expect large delays in processing, so this causes a
348higher drift (and a lot more system calls to get the current time).
349
350In another sense, L<EV> is more exact, as your timer will be scheduled at
351the same time, regardless of how long event processing actually took.
352
353In either case, if you care (and in most cases, you don't), then you
354can get whatever behaviour you want with any event loop, by taking the
355difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
356account.
357
358=item AnyEvent->now_update
359
360Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
361the current time for each loop iteration (see the discussion of L<<
362AnyEvent->now >>, above).
363
364When a callback runs for a long time (or when the process sleeps), then
365this "current" time will differ substantially from the real time, which
366might affect timers and time-outs.
367
368When this is the case, you can call this method, which will update the
369event loop's idea of "current time".
370
371A typical example would be a script in a web server (e.g. C<mod_perl>) -
372when mod_perl executes the script, then the event loop will have the wrong
373idea about the "current time" (being potentially far in the past, when the
374script ran the last time). In that case you should arrange a call to C<<
375AnyEvent->now_update >> each time the web server process wakes up again
376(e.g. at the start of your script, or in a handler).
377
378Note that updating the time I<might> cause some events to be handled.
379
380=back
381
238=head2 SIGNAL WATCHERS 382=head2 SIGNAL WATCHERS
239 383
384 $w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>);
385
240You can watch for signals using a signal watcher, C<signal> is the signal 386You can watch for signals using a signal watcher, C<signal> is the signal
241I<name> without any C<SIG> prefix, C<cb> is the Perl callback to 387I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
242be invoked whenever a signal occurs. 388callback to be invoked whenever a signal occurs.
243 389
244Although the callback might get passed parameters, their value and 390Although the callback might get passed parameters, their value and
245presence is undefined and you cannot rely on them. Portable AnyEvent 391presence is undefined and you cannot rely on them. Portable AnyEvent
246callbacks cannot use arguments passed to signal watcher callbacks. 392callbacks cannot use arguments passed to signal watcher callbacks.
247 393
249invocation, and callback invocation will be synchronous. Synchronous means 395invocation, and callback invocation will be synchronous. Synchronous means
250that it might take a while until the signal gets handled by the process, 396that it might take a while until the signal gets handled by the process,
251but it is guaranteed not to interrupt any other callbacks. 397but it is guaranteed not to interrupt any other callbacks.
252 398
253The main advantage of using these watchers is that you can share a signal 399The main advantage of using these watchers is that you can share a signal
254between multiple watchers. 400between multiple watchers, and AnyEvent will ensure that signals will not
401interrupt your program at bad times.
255 402
256This watcher might use C<%SIG>, so programs overwriting those signals 403This watcher might use C<%SIG> (depending on the event loop used),
257directly will likely not work correctly. 404so programs overwriting those signals directly will likely not work
405correctly.
258 406
259Example: exit on SIGINT 407Example: exit on SIGINT
260 408
261 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); 409 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
262 410
411=head3 Restart Behaviour
412
413While restart behaviour is up to the event loop implementation, most will
414not restart syscalls (that includes L<Async::Interrupt> and AnyEvent's
415pure perl implementation).
416
417=head3 Safe/Unsafe Signals
418
419Perl signals can be either "safe" (synchronous to opcode handling) or
420"unsafe" (asynchronous) - the former might get delayed indefinitely, the
421latter might corrupt your memory.
422
423AnyEvent signal handlers are, in addition, synchronous to the event loop,
424i.e. they will not interrupt your running perl program but will only be
425called as part of the normal event handling (just like timer, I/O etc.
426callbacks, too).
427
428=head3 Signal Races, Delays and Workarounds
429
430Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
431callbacks to signals in a generic way, which is a pity, as you cannot
432do race-free signal handling in perl, requiring C libraries for
433this. AnyEvent will try to do it's best, which means in some cases,
434signals will be delayed. The maximum time a signal might be delayed is
435specified in C<$AnyEvent::MAX_SIGNAL_LATENCY> (default: 10 seconds). This
436variable can be changed only before the first signal watcher is created,
437and should be left alone otherwise. This variable determines how often
438AnyEvent polls for signals (in case a wake-up was missed). Higher values
439will cause fewer spurious wake-ups, which is better for power and CPU
440saving.
441
442All these problems can be avoided by installing the optional
443L<Async::Interrupt> module, which works with most event loops. It will not
444work with inherently broken event loops such as L<Event> or L<Event::Lib>
445(and not with L<POE> currently, as POE does it's own workaround with
446one-second latency). For those, you just have to suffer the delays.
447
263=head2 CHILD PROCESS WATCHERS 448=head2 CHILD PROCESS WATCHERS
264 449
450 $w = AnyEvent->child (pid => <process id>, cb => <callback>);
451
265You can also watch on a child process exit and catch its exit status. 452You can also watch on a child process exit and catch its exit status.
266 453
267The child process is specified by the C<pid> argument (if set to C<0>, it 454The child process is specified by the C<pid> argument (one some backends,
268watches for any child process exit). The watcher will trigger as often 455using C<0> watches for any child process exit, on others this will
269as status change for the child are received. This works by installing a 456croak). The watcher will be triggered only when the child process has
270signal handler for C<SIGCHLD>. The callback will be called with the pid 457finished and an exit status is available, not on any trace events
271and exit status (as returned by waitpid), so unlike other watcher types, 458(stopped/continued).
272you I<can> rely on child watcher callback arguments. 459
460The callback will be called with the pid and exit status (as returned by
461waitpid), so unlike other watcher types, you I<can> rely on child watcher
462callback arguments.
463
464This watcher type works by installing a signal handler for C<SIGCHLD>,
465and since it cannot be shared, nothing else should use SIGCHLD or reap
466random child processes (waiting for specific child processes, e.g. inside
467C<system>, is just fine).
273 468
274There is a slight catch to child watchers, however: you usually start them 469There is a slight catch to child watchers, however: you usually start them
275I<after> the child process was created, and this means the process could 470I<after> the child process was created, and this means the process could
276have exited already (and no SIGCHLD will be sent anymore). 471have exited already (and no SIGCHLD will be sent anymore).
277 472
278Not all event models handle this correctly (POE doesn't), but even for 473Not all event models handle this correctly (neither POE nor IO::Async do,
474see their AnyEvent::Impl manpages for details), but even for event models
279event models that I<do> handle this correctly, they usually need to be 475that I<do> handle this correctly, they usually need to be loaded before
280loaded before the process exits (i.e. before you fork in the first place). 476the process exits (i.e. before you fork in the first place). AnyEvent's
477pure perl event loop handles all cases correctly regardless of when you
478start the watcher.
281 479
282This means you cannot create a child watcher as the very first thing in an 480This means you cannot create a child watcher as the very first
283AnyEvent program, you I<have> to create at least one watcher before you 481thing in an AnyEvent program, you I<have> to create at least one
284C<fork> the child (alternatively, you can call C<AnyEvent::detect>). 482watcher before you C<fork> the child (alternatively, you can call
483C<AnyEvent::detect>).
484
485As most event loops do not support waiting for child events, they will be
486emulated by AnyEvent in most cases, in which the latency and race problems
487mentioned in the description of signal watchers apply.
285 488
286Example: fork a process and wait for it 489Example: fork a process and wait for it
287 490
288 my $done = AnyEvent->condvar; 491 my $done = AnyEvent->condvar;
289 492
290 my $pid = fork or exit 5; 493 my $pid = fork or exit 5;
291 494
292 my $w = AnyEvent->child ( 495 my $w = AnyEvent->child (
293 pid => $pid, 496 pid => $pid,
294 cb => sub { 497 cb => sub {
295 my ($pid, $status) = @_; 498 my ($pid, $status) = @_;
296 warn "pid $pid exited with status $status"; 499 warn "pid $pid exited with status $status";
297 $done->send; 500 $done->send;
298 }, 501 },
299 ); 502 );
300 503
301 # do something else, then wait for process exit 504 # do something else, then wait for process exit
302 $done->recv; 505 $done->recv;
506
507=head2 IDLE WATCHERS
508
509 $w = AnyEvent->idle (cb => <callback>);
510
511Repeatedly invoke the callback after the process becomes idle, until
512either the watcher is destroyed or new events have been detected.
513
514Idle watchers are useful when there is a need to do something, but it
515is not so important (or wise) to do it instantly. The callback will be
516invoked only when there is "nothing better to do", which is usually
517defined as "all outstanding events have been handled and no new events
518have been detected". That means that idle watchers ideally get invoked
519when the event loop has just polled for new events but none have been
520detected. Instead of blocking to wait for more events, the idle watchers
521will be invoked.
522
523Unfortunately, most event loops do not really support idle watchers (only
524EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
525will simply call the callback "from time to time".
526
527Example: read lines from STDIN, but only process them when the
528program is otherwise idle:
529
530 my @lines; # read data
531 my $idle_w;
532 my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
533 push @lines, scalar <STDIN>;
534
535 # start an idle watcher, if not already done
536 $idle_w ||= AnyEvent->idle (cb => sub {
537 # handle only one line, when there are lines left
538 if (my $line = shift @lines) {
539 print "handled when idle: $line";
540 } else {
541 # otherwise disable the idle watcher again
542 undef $idle_w;
543 }
544 });
545 });
303 546
304=head2 CONDITION VARIABLES 547=head2 CONDITION VARIABLES
548
549 $cv = AnyEvent->condvar;
550
551 $cv->send (<list>);
552 my @res = $cv->recv;
305 553
306If you are familiar with some event loops you will know that all of them 554If you are familiar with some event loops you will know that all of them
307require you to run some blocking "loop", "run" or similar function that 555require you to run some blocking "loop", "run" or similar function that
308will actively watch for new events and call your callbacks. 556will actively watch for new events and call your callbacks.
309 557
310AnyEvent is different, it expects somebody else to run the event loop and 558AnyEvent is slightly different: it expects somebody else to run the event
311will only block when necessary (usually when told by the user). 559loop and will only block when necessary (usually when told by the user).
312 560
313The instrument to do that is called a "condition variable", so called 561The instrument to do that is called a "condition variable", so called
314because they represent a condition that must become true. 562because they represent a condition that must become true.
563
564Now is probably a good time to look at the examples further below.
315 565
316Condition variables can be created by calling the C<< AnyEvent->condvar 566Condition variables can be created by calling the C<< AnyEvent->condvar
317>> method, usually without arguments. The only argument pair allowed is 567>> method, usually without arguments. The only argument pair allowed is
318C<cb>, which specifies a callback to be called when the condition variable 568C<cb>, which specifies a callback to be called when the condition variable
319becomes true. 569becomes true, with the condition variable as the first argument (but not
570the results).
320 571
321After creation, the condition variable is "false" until it becomes "true" 572After creation, the condition variable is "false" until it becomes "true"
322by calling the C<send> method (or calling the condition variable as if it 573by calling the C<send> method (or calling the condition variable as if it
323were a callback, read about the caveats in the description for the C<< 574were a callback, read about the caveats in the description for the C<<
324->send >> method). 575->send >> method).
326Condition variables are similar to callbacks, except that you can 577Condition variables are similar to callbacks, except that you can
327optionally wait for them. They can also be called merge points - points 578optionally wait for them. They can also be called merge points - points
328in time where multiple outstanding events have been processed. And yet 579in time where multiple outstanding events have been processed. And yet
329another way to call them is transactions - each condition variable can be 580another way to call them is transactions - each condition variable can be
330used to represent a transaction, which finishes at some point and delivers 581used to represent a transaction, which finishes at some point and delivers
331a result. 582a result. And yet some people know them as "futures" - a promise to
583compute/deliver something that you can wait for.
332 584
333Condition variables are very useful to signal that something has finished, 585Condition variables are very useful to signal that something has finished,
334for example, if you write a module that does asynchronous http requests, 586for example, if you write a module that does asynchronous http requests,
335then a condition variable would be the ideal candidate to signal the 587then a condition variable would be the ideal candidate to signal the
336availability of results. The user can either act when the callback is 588availability of results. The user can either act when the callback is
370 after => 1, 622 after => 1,
371 cb => sub { $result_ready->send }, 623 cb => sub { $result_ready->send },
372 ); 624 );
373 625
374 # this "blocks" (while handling events) till the callback 626 # this "blocks" (while handling events) till the callback
375 # calls send 627 # calls ->send
376 $result_ready->recv; 628 $result_ready->recv;
377 629
378Example: wait for a timer, but take advantage of the fact that 630Example: wait for a timer, but take advantage of the fact that condition
379condition variables are also code references. 631variables are also callable directly.
380 632
381 my $done = AnyEvent->condvar; 633 my $done = AnyEvent->condvar;
382 my $delay = AnyEvent->timer (after => 5, cb => $done); 634 my $delay = AnyEvent->timer (after => 5, cb => $done);
383 $done->recv; 635 $done->recv;
636
637Example: Imagine an API that returns a condvar and doesn't support
638callbacks. This is how you make a synchronous call, for example from
639the main program:
640
641 use AnyEvent::CouchDB;
642
643 ...
644
645 my @info = $couchdb->info->recv;
646
647And this is how you would just set a callback to be called whenever the
648results are available:
649
650 $couchdb->info->cb (sub {
651 my @info = $_[0]->recv;
652 });
384 653
385=head3 METHODS FOR PRODUCERS 654=head3 METHODS FOR PRODUCERS
386 655
387These methods should only be used by the producing side, i.e. the 656These methods should only be used by the producing side, i.e. the
388code/module that eventually sends the signal. Note that it is also 657code/module that eventually sends the signal. Note that it is also
401immediately from within send. 670immediately from within send.
402 671
403Any arguments passed to the C<send> call will be returned by all 672Any arguments passed to the C<send> call will be returned by all
404future C<< ->recv >> calls. 673future C<< ->recv >> calls.
405 674
406Condition variables are overloaded so one can call them directly 675Condition variables are overloaded so one can call them directly (as if
407(as a code reference). Calling them directly is the same as calling 676they were a code reference). Calling them directly is the same as calling
408C<send>. Note, however, that many C-based event loops do not handle 677C<send>.
409overloading, so as tempting as it may be, passing a condition variable
410instead of a callback does not work. Both the pure perl and EV loops
411support overloading, however, as well as all functions that use perl to
412invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
413example).
414 678
415=item $cv->croak ($error) 679=item $cv->croak ($error)
416 680
417Similar to send, but causes all call's to C<< ->recv >> to invoke 681Similar to send, but causes all call's to C<< ->recv >> to invoke
418C<Carp::croak> with the given error message/object/scalar. 682C<Carp::croak> with the given error message/object/scalar.
419 683
420This can be used to signal any errors to the condition variable 684This can be used to signal any errors to the condition variable
421user/consumer. 685user/consumer. Doing it this way instead of calling C<croak> directly
686delays the error detetcion, but has the overwhelmign advantage that it
687diagnoses the error at the place where the result is expected, and not
688deep in some event clalback without connection to the actual code causing
689the problem.
422 690
423=item $cv->begin ([group callback]) 691=item $cv->begin ([group callback])
424 692
425=item $cv->end 693=item $cv->end
426
427These two methods are EXPERIMENTAL and MIGHT CHANGE.
428 694
429These two methods can be used to combine many transactions/events into 695These two methods can be used to combine many transactions/events into
430one. For example, a function that pings many hosts in parallel might want 696one. For example, a function that pings many hosts in parallel might want
431to use a condition variable for the whole process. 697to use a condition variable for the whole process.
432 698
433Every call to C<< ->begin >> will increment a counter, and every call to 699Every call to C<< ->begin >> will increment a counter, and every call to
434C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end 700C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
435>>, the (last) callback passed to C<begin> will be executed. That callback 701>>, the (last) callback passed to C<begin> will be executed, passing the
436is I<supposed> to call C<< ->send >>, but that is not required. If no 702condvar as first argument. That callback is I<supposed> to call C<< ->send
437callback was set, C<send> will be called without any arguments. 703>>, but that is not required. If no group callback was set, C<send> will
704be called without any arguments.
438 705
439Let's clarify this with the ping example: 706You can think of C<< $cv->send >> giving you an OR condition (one call
707sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
708condition (all C<begin> calls must be C<end>'ed before the condvar sends).
709
710Let's start with a simple example: you have two I/O watchers (for example,
711STDOUT and STDERR for a program), and you want to wait for both streams to
712close before activating a condvar:
440 713
441 my $cv = AnyEvent->condvar; 714 my $cv = AnyEvent->condvar;
442 715
716 $cv->begin; # first watcher
717 my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
718 defined sysread $fh1, my $buf, 4096
719 or $cv->end;
720 });
721
722 $cv->begin; # second watcher
723 my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
724 defined sysread $fh2, my $buf, 4096
725 or $cv->end;
726 });
727
728 $cv->recv;
729
730This works because for every event source (EOF on file handle), there is
731one call to C<begin>, so the condvar waits for all calls to C<end> before
732sending.
733
734The ping example mentioned above is slightly more complicated, as the
735there are results to be passwd back, and the number of tasks that are
736begung can potentially be zero:
737
738 my $cv = AnyEvent->condvar;
739
443 my %result; 740 my %result;
444 $cv->begin (sub { $cv->send (\%result) }); 741 $cv->begin (sub { shift->send (\%result) });
445 742
446 for my $host (@list_of_hosts) { 743 for my $host (@list_of_hosts) {
447 $cv->begin; 744 $cv->begin;
448 ping_host_then_call_callback $host, sub { 745 ping_host_then_call_callback $host, sub {
449 $result{$host} = ...; 746 $result{$host} = ...;
464loop, which serves two important purposes: first, it sets the callback 761loop, which serves two important purposes: first, it sets the callback
465to be called once the counter reaches C<0>, and second, it ensures that 762to be called once the counter reaches C<0>, and second, it ensures that
466C<send> is called even when C<no> hosts are being pinged (the loop 763C<send> is called even when C<no> hosts are being pinged (the loop
467doesn't execute once). 764doesn't execute once).
468 765
469This is the general pattern when you "fan out" into multiple subrequests: 766This is the general pattern when you "fan out" into multiple (but
470use an outer C<begin>/C<end> pair to set the callback and ensure C<end> 767potentially none) subrequests: use an outer C<begin>/C<end> pair to set
471is called at least once, and then, for each subrequest you start, call 768the callback and ensure C<end> is called at least once, and then, for each
472C<begin> and for each subrequest you finish, call C<end>. 769subrequest you start, call C<begin> and for each subrequest you finish,
770call C<end>.
473 771
474=back 772=back
475 773
476=head3 METHODS FOR CONSUMERS 774=head3 METHODS FOR CONSUMERS
477 775
493function will call C<croak>. 791function will call C<croak>.
494 792
495In list context, all parameters passed to C<send> will be returned, 793In list context, all parameters passed to C<send> will be returned,
496in scalar context only the first one will be returned. 794in scalar context only the first one will be returned.
497 795
796Note that doing a blocking wait in a callback is not supported by any
797event loop, that is, recursive invocation of a blocking C<< ->recv
798>> is not allowed, and the C<recv> call will C<croak> if such a
799condition is detected. This condition can be slightly loosened by using
800L<Coro::AnyEvent>, which allows you to do a blocking C<< ->recv >> from
801any thread that doesn't run the event loop itself.
802
498Not all event models support a blocking wait - some die in that case 803Not all event models support a blocking wait - some die in that case
499(programs might want to do that to stay interactive), so I<if you are 804(programs might want to do that to stay interactive), so I<if you are
500using this from a module, never require a blocking wait>, but let the 805using this from a module, never require a blocking wait>. Instead, let the
501caller decide whether the call will block or not (for example, by coupling 806caller decide whether the call will block or not (for example, by coupling
502condition variables with some kind of request results and supporting 807condition variables with some kind of request results and supporting
503callbacks so the caller knows that getting the result will not block, 808callbacks so the caller knows that getting the result will not block,
504while still supporting blocking waits if the caller so desires). 809while still supporting blocking waits if the caller so desires).
505 810
506Another reason I<never> to C<< ->recv >> in a module is that you cannot
507sensibly have two C<< ->recv >>'s in parallel, as that would require
508multiple interpreters or coroutines/threads, none of which C<AnyEvent>
509can supply.
510
511The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
512fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
513versions and also integrates coroutines into AnyEvent, making blocking
514C<< ->recv >> calls perfectly safe as long as they are done from another
515coroutine (one that doesn't run the event loop).
516
517You can ensure that C<< -recv >> never blocks by setting a callback and 811You can ensure that C<< -recv >> never blocks by setting a callback and
518only calling C<< ->recv >> from within that callback (or at a later 812only calling C<< ->recv >> from within that callback (or at a later
519time). This will work even when the event loop does not support blocking 813time). This will work even when the event loop does not support blocking
520waits otherwise. 814waits otherwise.
521 815
522=item $bool = $cv->ready 816=item $bool = $cv->ready
523 817
524Returns true when the condition is "true", i.e. whether C<send> or 818Returns true when the condition is "true", i.e. whether C<send> or
525C<croak> have been called. 819C<croak> have been called.
526 820
527=item $cb = $cv->cb ([new callback]) 821=item $cb = $cv->cb ($cb->($cv))
528 822
529This is a mutator function that returns the callback set and optionally 823This is a mutator function that returns the callback set and optionally
530replaces it before doing so. 824replaces it before doing so.
531 825
532The callback will be called when the condition becomes "true", i.e. when 826The callback will be called when the condition becomes (or already was)
533C<send> or C<croak> are called. Calling C<recv> inside the callback 827"true", i.e. when C<send> or C<croak> are called (or were called), with
828the only argument being the condition variable itself. Calling C<recv>
534or at any later time is guaranteed not to block. 829inside the callback or at any later time is guaranteed not to block.
535 830
536=back 831=back
537 832
833=head1 SUPPORTED EVENT LOOPS/BACKENDS
834
835The available backend classes are (every class has its own manpage):
836
837=over 4
838
839=item Backends that are autoprobed when no other event loop can be found.
840
841EV is the preferred backend when no other event loop seems to be in
842use. If EV is not installed, then AnyEvent will fall back to its own
843pure-perl implementation, which is available everywhere as it comes with
844AnyEvent itself.
845
846 AnyEvent::Impl::EV based on EV (interface to libev, best choice).
847 AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
848
849=item Backends that are transparently being picked up when they are used.
850
851These will be used when they are currently loaded when the first watcher
852is created, in which case it is assumed that the application is using
853them. This means that AnyEvent will automatically pick the right backend
854when the main program loads an event module before anything starts to
855create watchers. Nothing special needs to be done by the main program.
856
857 AnyEvent::Impl::Event based on Event, very stable, few glitches.
858 AnyEvent::Impl::Glib based on Glib, slow but very stable.
859 AnyEvent::Impl::Tk based on Tk, very broken.
860 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
861 AnyEvent::Impl::POE based on POE, very slow, some limitations.
862 AnyEvent::Impl::Irssi used when running within irssi.
863
864=item Backends with special needs.
865
866Qt requires the Qt::Application to be instantiated first, but will
867otherwise be picked up automatically. As long as the main program
868instantiates the application before any AnyEvent watchers are created,
869everything should just work.
870
871 AnyEvent::Impl::Qt based on Qt.
872
873Support for IO::Async can only be partial, as it is too broken and
874architecturally limited to even support the AnyEvent API. It also
875is the only event loop that needs the loop to be set explicitly, so
876it can only be used by a main program knowing about AnyEvent. See
877L<AnyEvent::Impl::Async> for the gory details.
878
879 AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed.
880
881=item Event loops that are indirectly supported via other backends.
882
883Some event loops can be supported via other modules:
884
885There is no direct support for WxWidgets (L<Wx>) or L<Prima>.
886
887B<WxWidgets> has no support for watching file handles. However, you can
888use WxWidgets through the POE adaptor, as POE has a Wx backend that simply
889polls 20 times per second, which was considered to be too horrible to even
890consider for AnyEvent.
891
892B<Prima> is not supported as nobody seems to be using it, but it has a POE
893backend, so it can be supported through POE.
894
895AnyEvent knows about both L<Prima> and L<Wx>, however, and will try to
896load L<POE> when detecting them, in the hope that POE will pick them up,
897in which case everything will be automatic.
898
899=back
900
538=head1 GLOBAL VARIABLES AND FUNCTIONS 901=head1 GLOBAL VARIABLES AND FUNCTIONS
539 902
903These are not normally required to use AnyEvent, but can be useful to
904write AnyEvent extension modules.
905
540=over 4 906=over 4
541 907
542=item $AnyEvent::MODEL 908=item $AnyEvent::MODEL
543 909
544Contains C<undef> until the first watcher is being created. Then it 910Contains C<undef> until the first watcher is being created, before the
911backend has been autodetected.
912
545contains the event model that is being used, which is the name of the 913Afterwards it contains the event model that is being used, which is the
546Perl class implementing the model. This class is usually one of the 914name of the Perl class implementing the model. This class is usually one
547C<AnyEvent::Impl:xxx> modules, but can be any other class in the case 915of the C<AnyEvent::Impl:xxx> modules, but can be any other class in the
548AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>). 916case AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode> it
549 917will be C<urxvt::anyevent>).
550The known classes so far are:
551
552 AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
553 AnyEvent::Impl::Event based on Event, second best choice.
554 AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
555 AnyEvent::Impl::Glib based on Glib, third-best choice.
556 AnyEvent::Impl::Tk based on Tk, very bad choice.
557 AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
558 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
559 AnyEvent::Impl::POE based on POE, not generic enough for full support.
560
561There is no support for WxWidgets, as WxWidgets has no support for
562watching file handles. However, you can use WxWidgets through the
563POE Adaptor, as POE has a Wx backend that simply polls 20 times per
564second, which was considered to be too horrible to even consider for
565AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
566it's adaptor.
567
568AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
569autodetecting them.
570 918
571=item AnyEvent::detect 919=item AnyEvent::detect
572 920
573Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model 921Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
574if necessary. You should only call this function right before you would 922if necessary. You should only call this function right before you would
575have created an AnyEvent watcher anyway, that is, as late as possible at 923have created an AnyEvent watcher anyway, that is, as late as possible at
576runtime. 924runtime, and not e.g. while initialising of your module.
925
926If you need to do some initialisation before AnyEvent watchers are
927created, use C<post_detect>.
577 928
578=item $guard = AnyEvent::post_detect { BLOCK } 929=item $guard = AnyEvent::post_detect { BLOCK }
579 930
580Arranges for the code block to be executed as soon as the event model is 931Arranges for the code block to be executed as soon as the event model is
581autodetected (or immediately if this has already happened). 932autodetected (or immediately if this has already happened).
582 933
934The block will be executed I<after> the actual backend has been detected
935(C<$AnyEvent::MODEL> is set), but I<before> any watchers have been
936created, so it is possible to e.g. patch C<@AnyEvent::ISA> or do
937other initialisations - see the sources of L<AnyEvent::Strict> or
938L<AnyEvent::AIO> to see how this is used.
939
940The most common usage is to create some global watchers, without forcing
941event module detection too early, for example, L<AnyEvent::AIO> creates
942and installs the global L<IO::AIO> watcher in a C<post_detect> block to
943avoid autodetecting the event module at load time.
944
583If called in scalar or list context, then it creates and returns an object 945If called in scalar or list context, then it creates and returns an object
584that automatically removes the callback again when it is destroyed. See 946that automatically removes the callback again when it is destroyed (or
947C<undef> when the hook was immediately executed). See L<AnyEvent::AIO> for
585L<Coro::BDB> for a case where this is useful. 948a case where this is useful.
949
950Example: Create a watcher for the IO::AIO module and store it in
951C<$WATCHER>. Only do so after the event loop is initialised, though.
952
953 our WATCHER;
954
955 my $guard = AnyEvent::post_detect {
956 $WATCHER = AnyEvent->io (fh => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb);
957 };
958
959 # the ||= is important in case post_detect immediately runs the block,
960 # as to not clobber the newly-created watcher. assigning both watcher and
961 # post_detect guard to the same variable has the advantage of users being
962 # able to just C<undef $WATCHER> if the watcher causes them grief.
963
964 $WATCHER ||= $guard;
586 965
587=item @AnyEvent::post_detect 966=item @AnyEvent::post_detect
588 967
589If there are any code references in this array (you can C<push> to it 968If there are any code references in this array (you can C<push> to it
590before or after loading AnyEvent), then they will called directly after 969before or after loading AnyEvent), then they will called directly after
591the event loop has been chosen. 970the event loop has been chosen.
592 971
593You should check C<$AnyEvent::MODEL> before adding to this array, though: 972You should check C<$AnyEvent::MODEL> before adding to this array, though:
594if it contains a true value then the event loop has already been detected, 973if it is defined then the event loop has already been detected, and the
595and the array will be ignored. 974array will be ignored.
596 975
597Best use C<AnyEvent::post_detect { BLOCK }> instead. 976Best use C<AnyEvent::post_detect { BLOCK }> when your application allows
977it, as it takes care of these details.
978
979This variable is mainly useful for modules that can do something useful
980when AnyEvent is used and thus want to know when it is initialised, but do
981not need to even load it by default. This array provides the means to hook
982into AnyEvent passively, without loading it.
983
984Example: To load Coro::AnyEvent whenever Coro and AnyEvent are used
985together, you could put this into Coro (this is the actual code used by
986Coro to accomplish this):
987
988 if (defined $AnyEvent::MODEL) {
989 # AnyEvent already initialised, so load Coro::AnyEvent
990 require Coro::AnyEvent;
991 } else {
992 # AnyEvent not yet initialised, so make sure to load Coro::AnyEvent
993 # as soon as it is
994 push @AnyEvent::post_detect, sub { require Coro::AnyEvent };
995 }
598 996
599=back 997=back
600 998
601=head1 WHAT TO DO IN A MODULE 999=head1 WHAT TO DO IN A MODULE
602 1000
657 1055
658 1056
659=head1 OTHER MODULES 1057=head1 OTHER MODULES
660 1058
661The following is a non-exhaustive list of additional modules that use 1059The following is a non-exhaustive list of additional modules that use
662AnyEvent and can therefore be mixed easily with other AnyEvent modules 1060AnyEvent as a client and can therefore be mixed easily with other AnyEvent
663in the same program. Some of the modules come with AnyEvent, some are 1061modules and other event loops in the same program. Some of the modules
664available via CPAN. 1062come with AnyEvent, most are available via CPAN.
665 1063
666=over 4 1064=over 4
667 1065
668=item L<AnyEvent::Util> 1066=item L<AnyEvent::Util>
669 1067
670Contains various utility functions that replace often-used but blocking 1068Contains various utility functions that replace often-used but blocking
671functions such as C<inet_aton> by event-/callback-based versions. 1069functions such as C<inet_aton> by event-/callback-based versions.
672
673=item L<AnyEvent::Handle>
674
675Provide read and write buffers and manages watchers for reads and writes.
676 1070
677=item L<AnyEvent::Socket> 1071=item L<AnyEvent::Socket>
678 1072
679Provides various utility functions for (internet protocol) sockets, 1073Provides various utility functions for (internet protocol) sockets,
680addresses and name resolution. Also functions to create non-blocking tcp 1074addresses and name resolution. Also functions to create non-blocking tcp
681connections or tcp servers, with IPv6 and SRV record support and more. 1075connections or tcp servers, with IPv6 and SRV record support and more.
682 1076
1077=item L<AnyEvent::Handle>
1078
1079Provide read and write buffers, manages watchers for reads and writes,
1080supports raw and formatted I/O, I/O queued and fully transparent and
1081non-blocking SSL/TLS (via L<AnyEvent::TLS>.
1082
683=item L<AnyEvent::DNS> 1083=item L<AnyEvent::DNS>
684 1084
685Provides rich asynchronous DNS resolver capabilities. 1085Provides rich asynchronous DNS resolver capabilities.
686 1086
1087=item L<AnyEvent::HTTP>
1088
1089A simple-to-use HTTP library that is capable of making a lot of concurrent
1090HTTP requests.
1091
687=item L<AnyEvent::HTTPD> 1092=item L<AnyEvent::HTTPD>
688 1093
689Provides a simple web application server framework. 1094Provides a simple web application server framework.
690 1095
691=item L<AnyEvent::FastPing> 1096=item L<AnyEvent::FastPing>
692 1097
693The fastest ping in the west. 1098The fastest ping in the west.
694 1099
1100=item L<AnyEvent::DBI>
1101
1102Executes L<DBI> requests asynchronously in a proxy process.
1103
1104=item L<AnyEvent::AIO>
1105
1106Truly asynchronous I/O, should be in the toolbox of every event
1107programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
1108together.
1109
1110=item L<AnyEvent::BDB>
1111
1112Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
1113L<BDB> and AnyEvent together.
1114
1115=item L<AnyEvent::GPSD>
1116
1117A non-blocking interface to gpsd, a daemon delivering GPS information.
1118
695=item L<Net::IRC3> 1119=item L<AnyEvent::IRC>
696 1120
697AnyEvent based IRC client module family. 1121AnyEvent based IRC client module family (replacing the older Net::IRC3).
698 1122
699=item L<Net::XMPP2> 1123=item L<AnyEvent::XMPP>
700 1124
701AnyEvent based XMPP (Jabber protocol) module family. 1125AnyEvent based XMPP (Jabber protocol) module family (replacing the older
1126Net::XMPP2>.
1127
1128=item L<AnyEvent::IGS>
1129
1130A non-blocking interface to the Internet Go Server protocol (used by
1131L<App::IGS>).
702 1132
703=item L<Net::FCP> 1133=item L<Net::FCP>
704 1134
705AnyEvent-based implementation of the Freenet Client Protocol, birthplace 1135AnyEvent-based implementation of the Freenet Client Protocol, birthplace
706of AnyEvent. 1136of AnyEvent.
711 1141
712=item L<Coro> 1142=item L<Coro>
713 1143
714Has special support for AnyEvent via L<Coro::AnyEvent>. 1144Has special support for AnyEvent via L<Coro::AnyEvent>.
715 1145
716=item L<AnyEvent::AIO>, L<IO::AIO>
717
718Truly asynchronous I/O, should be in the toolbox of every event
719programmer. AnyEvent::AIO transparently fuses IO::AIO and AnyEvent
720together.
721
722=item L<AnyEvent::BDB>, L<BDB>
723
724Truly asynchronous Berkeley DB access. AnyEvent::AIO transparently fuses
725IO::AIO and AnyEvent together.
726
727=item L<IO::Lambda>
728
729The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
730
731=back 1146=back
732 1147
733=cut 1148=cut
734 1149
735package AnyEvent; 1150package AnyEvent;
736 1151
737no warnings; 1152# basically a tuned-down version of common::sense
738use strict; 1153sub common_sense {
1154 # from common:.sense 1.0
1155 ${^WARNING_BITS} = "\xfc\x3f\x33\x00\x0f\xf3\xcf\xc0\xf3\xfc\x33\x00";
1156 # use strict vars subs - NO UTF-8, as Util.pm doesn't like this atm. (uts46data.pl)
1157 $^H |= 0x00000600;
1158}
739 1159
1160BEGIN { AnyEvent::common_sense }
1161
740use Carp; 1162use Carp ();
741 1163
742our $VERSION = '4.05'; 1164our $VERSION = '5.251';
743our $MODEL; 1165our $MODEL;
744 1166
745our $AUTOLOAD; 1167our $AUTOLOAD;
746our @ISA; 1168our @ISA;
747 1169
748our @REGISTRY; 1170our @REGISTRY;
749 1171
750our $WIN32; 1172our $VERBOSE;
751 1173
752BEGIN { 1174BEGIN {
753 my $win32 = ! ! ($^O =~ /mswin32/i); 1175 require "AnyEvent/constants.pl";
754 eval "sub WIN32(){ $win32 }";
755}
756 1176
1177 eval "sub TAINT (){" . (${^TAINT}*1) . "}";
1178
1179 delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
1180 if ${^TAINT};
1181
757our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1; 1182 $VERBOSE = $ENV{PERL_ANYEVENT_VERBOSE}*1;
1183
1184}
1185
1186our $MAX_SIGNAL_LATENCY = 10;
758 1187
759our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred 1188our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
760 1189
761{ 1190{
762 my $idx; 1191 my $idx;
764 for reverse split /\s*,\s*/, 1193 for reverse split /\s*,\s*/,
765 $ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6"; 1194 $ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
766} 1195}
767 1196
768my @models = ( 1197my @models = (
769 [EV:: => AnyEvent::Impl::EV::], 1198 [EV:: => AnyEvent::Impl::EV:: , 1],
770 [Event:: => AnyEvent::Impl::Event::],
771 [AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::], 1199 [AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl:: , 1],
772 # everything below here will not be autoprobed 1200 # everything below here will not (normally) be autoprobed
773 # as the pureperl backend should work everywhere 1201 # as the pureperl backend should work everywhere
774 # and is usually faster 1202 # and is usually faster
1203 [Event:: => AnyEvent::Impl::Event::, 1],
1204 [Glib:: => AnyEvent::Impl::Glib:: , 1], # becomes extremely slow with many watchers
1205 [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
1206 [Irssi:: => AnyEvent::Impl::Irssi::], # Irssi has a bogus "Event" package
775 [Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles 1207 [Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
776 [Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
777 [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
778 [Qt:: => AnyEvent::Impl::Qt::], # requires special main program 1208 [Qt:: => AnyEvent::Impl::Qt::], # requires special main program
779 [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza 1209 [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
780 [Wx:: => AnyEvent::Impl::POE::], 1210 [Wx:: => AnyEvent::Impl::POE::],
781 [Prima:: => AnyEvent::Impl::POE::], 1211 [Prima:: => AnyEvent::Impl::POE::],
1212 # IO::Async is just too broken - we would need workarounds for its
1213 # byzantine signal and broken child handling, among others.
1214 # IO::Async is rather hard to detect, as it doesn't have any
1215 # obvious default class.
1216 [IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program
1217 [IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program
1218 [IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program
1219 [AnyEvent::Impl::IOAsync:: => AnyEvent::Impl::IOAsync::], # requires special main program
782); 1220);
783 1221
784our %method = map +($_ => 1), qw(io timer signal child condvar one_event DESTROY); 1222our %method = map +($_ => 1),
1223 qw(io timer time now now_update signal child idle condvar one_event DESTROY);
785 1224
786our @post_detect; 1225our @post_detect;
787 1226
788sub post_detect(&) { 1227sub post_detect(&) {
789 my ($cb) = @_; 1228 my ($cb) = @_;
790 1229
791 if ($MODEL) {
792 $cb->();
793
794 1
795 } else {
796 push @post_detect, $cb; 1230 push @post_detect, $cb;
797 1231
798 defined wantarray 1232 defined wantarray
799 ? bless \$cb, "AnyEvent::Util::PostDetect" 1233 ? bless \$cb, "AnyEvent::Util::postdetect"
800 : () 1234 : ()
1235}
1236
1237sub AnyEvent::Util::postdetect::DESTROY {
1238 @post_detect = grep $_ != ${$_[0]}, @post_detect;
1239}
1240
1241sub detect() {
1242 # free some memory
1243 *detect = sub () { $MODEL };
1244
1245 local $!; # for good measure
1246 local $SIG{__DIE__};
1247
1248 if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
1249 my $model = "AnyEvent::Impl::$1";
1250 if (eval "require $model") {
1251 $MODEL = $model;
1252 warn "AnyEvent: loaded model '$model' (forced by \$ENV{PERL_ANYEVENT_MODEL}), using it.\n" if $VERBOSE >= 2;
1253 } else {
1254 warn "AnyEvent: unable to load model '$model' (from \$ENV{PERL_ANYEVENT_MODEL}):\n$@" if $VERBOSE;
1255 }
801 } 1256 }
802}
803 1257
804sub AnyEvent::Util::PostDetect::DESTROY { 1258 # check for already loaded models
805 @post_detect = grep $_ != ${$_[0]}, @post_detect;
806}
807
808sub detect() {
809 unless ($MODEL) { 1259 unless ($MODEL) {
810 no strict 'refs'; 1260 for (@REGISTRY, @models) {
811 local $SIG{__DIE__}; 1261 my ($package, $model) = @$_;
812 1262 if (${"$package\::VERSION"} > 0) {
813 if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
814 my $model = "AnyEvent::Impl::$1";
815 if (eval "require $model") { 1263 if (eval "require $model") {
816 $MODEL = $model; 1264 $MODEL = $model;
817 warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1; 1265 warn "AnyEvent: autodetected model '$model', using it.\n" if $VERBOSE >= 2;
818 } else { 1266 last;
819 warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose; 1267 }
820 } 1268 }
821 } 1269 }
822 1270
823 # check for already loaded models
824 unless ($MODEL) { 1271 unless ($MODEL) {
1272 # try to autoload a model
825 for (@REGISTRY, @models) { 1273 for (@REGISTRY, @models) {
826 my ($package, $model) = @$_; 1274 my ($package, $model, $autoload) = @$_;
1275 if (
1276 $autoload
1277 and eval "require $package"
827 if (${"$package\::VERSION"} > 0) { 1278 and ${"$package\::VERSION"} > 0
828 if (eval "require $model") { 1279 and eval "require $model"
1280 ) {
829 $MODEL = $model; 1281 $MODEL = $model;
830 warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1; 1282 warn "AnyEvent: autoloaded model '$model', using it.\n" if $VERBOSE >= 2;
831 last; 1283 last;
832 }
833 } 1284 }
834 } 1285 }
835 1286
836 unless ($MODEL) {
837 # try to load a model
838
839 for (@REGISTRY, @models) {
840 my ($package, $model) = @$_;
841 if (eval "require $package"
842 and ${"$package\::VERSION"} > 0
843 and eval "require $model") {
844 $MODEL = $model;
845 warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1;
846 last;
847 }
848 }
849
850 $MODEL 1287 $MODEL
851 or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib."; 1288 or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
852 }
853 } 1289 }
854
855 unshift @ISA, $MODEL;
856 push @{"$MODEL\::ISA"}, "AnyEvent::Base";
857
858 (shift @post_detect)->() while @post_detect;
859 } 1290 }
1291
1292 @models = (); # free probe data
1293
1294 push @{"$MODEL\::ISA"}, "AnyEvent::Base";
1295 unshift @ISA, $MODEL;
1296
1297 # now nuke some methods that are overriden by the backend.
1298 # SUPER is not allowed.
1299 for (qw(time signal child idle)) {
1300 undef &{"AnyEvent::Base::$_"}
1301 if defined &{"$MODEL\::$_"};
1302 }
1303
1304 require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
1305
1306 (shift @post_detect)->() while @post_detect;
1307
1308 *post_detect = sub(&) {
1309 shift->();
1310
1311 undef
1312 };
860 1313
861 $MODEL 1314 $MODEL
862} 1315}
863 1316
864sub AUTOLOAD { 1317sub AUTOLOAD {
865 (my $func = $AUTOLOAD) =~ s/.*://; 1318 (my $func = $AUTOLOAD) =~ s/.*://;
866 1319
867 $method{$func} 1320 $method{$func}
868 or croak "$func: not a valid method for AnyEvent objects"; 1321 or Carp::croak "$func: not a valid AnyEvent class method";
869 1322
870 detect unless $MODEL; 1323 detect;
871 1324
872 my $class = shift; 1325 my $class = shift;
873 $class->$func (@_); 1326 $class->$func (@_);
874} 1327}
875 1328
1329# utility function to dup a filehandle. this is used by many backends
1330# to support binding more than one watcher per filehandle (they usually
1331# allow only one watcher per fd, so we dup it to get a different one).
1332sub _dupfh($$;$$) {
1333 my ($poll, $fh, $r, $w) = @_;
1334
1335 # cygwin requires the fh mode to be matching, unix doesn't
1336 my ($rw, $mode) = $poll eq "r" ? ($r, "<&") : ($w, ">&");
1337
1338 open my $fh2, $mode, $fh
1339 or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,";
1340
1341 # we assume CLOEXEC is already set by perl in all important cases
1342
1343 ($fh2, $rw)
1344}
1345
1346=head1 SIMPLIFIED AE API
1347
1348Starting with version 5.0, AnyEvent officially supports a second, much
1349simpler, API that is designed to reduce the calling, typing and memory
1350overhead by using function call syntax and a fixed number of parameters.
1351
1352See the L<AE> manpage for details.
1353
1354=cut
1355
1356package AE;
1357
1358our $VERSION = $AnyEvent::VERSION;
1359
1360# fall back to the main API by default - backends and AnyEvent::Base
1361# implementations can overwrite these.
1362
1363sub io($$$) {
1364 AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2])
1365}
1366
1367sub timer($$$) {
1368 AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2])
1369}
1370
1371sub signal($$) {
1372 AnyEvent->signal (signal => $_[0], cb => $_[1])
1373}
1374
1375sub child($$) {
1376 AnyEvent->child (pid => $_[0], cb => $_[1])
1377}
1378
1379sub idle($) {
1380 AnyEvent->idle (cb => $_[0])
1381}
1382
1383sub cv(;&) {
1384 AnyEvent->condvar (@_ ? (cb => $_[0]) : ())
1385}
1386
1387sub now() {
1388 AnyEvent->now
1389}
1390
1391sub now_update() {
1392 AnyEvent->now_update
1393}
1394
1395sub time() {
1396 AnyEvent->time
1397}
1398
876package AnyEvent::Base; 1399package AnyEvent::Base;
877 1400
1401# default implementations for many methods
1402
1403sub time {
1404 eval q{ # poor man's autoloading {}
1405 # probe for availability of Time::HiRes
1406 if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
1407 warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8;
1408 *AE::time = \&Time::HiRes::time;
1409 # if (eval "use POSIX (); (POSIX::times())...
1410 } else {
1411 warn "AnyEvent: using built-in time(), WARNING, no sub-second resolution!\n" if $VERBOSE;
1412 *AE::time = sub (){ time }; # epic fail
1413 }
1414
1415 *time = sub { AE::time }; # different prototypes
1416 };
1417 die if $@;
1418
1419 &time
1420}
1421
1422*now = \&time;
1423
1424sub now_update { }
1425
878# default implementation for ->condvar 1426# default implementation for ->condvar
879 1427
880sub condvar { 1428sub condvar {
1429 eval q{ # poor man's autoloading {}
1430 *condvar = sub {
881 bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, AnyEvent::CondVar:: 1431 bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
1432 };
1433
1434 *AE::cv = sub (;&) {
1435 bless { @_ ? (_ae_cb => shift) : () }, "AnyEvent::CondVar"
1436 };
1437 };
1438 die if $@;
1439
1440 &condvar
882} 1441}
883 1442
884# default implementation for ->signal 1443# default implementation for ->signal
885 1444
886our %SIG_CB; 1445our $HAVE_ASYNC_INTERRUPT;
1446
1447sub _have_async_interrupt() {
1448 $HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT}
1449 && eval "use Async::Interrupt 1.02 (); 1")
1450 unless defined $HAVE_ASYNC_INTERRUPT;
1451
1452 $HAVE_ASYNC_INTERRUPT
1453}
1454
1455our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
1456our (%SIG_ASY, %SIG_ASY_W);
1457our ($SIG_COUNT, $SIG_TW);
1458
1459# install a dummy wakeup watcher to reduce signal catching latency
1460# used by Impls
1461sub _sig_add() {
1462 unless ($SIG_COUNT++) {
1463 # try to align timer on a full-second boundary, if possible
1464 my $NOW = AE::now;
1465
1466 $SIG_TW = AE::timer
1467 $MAX_SIGNAL_LATENCY - ($NOW - int $NOW),
1468 $MAX_SIGNAL_LATENCY,
1469 sub { } # just for the PERL_ASYNC_CHECK
1470 ;
1471 }
1472}
1473
1474sub _sig_del {
1475 undef $SIG_TW
1476 unless --$SIG_COUNT;
1477}
1478
1479our $_sig_name_init; $_sig_name_init = sub {
1480 eval q{ # poor man's autoloading {}
1481 undef $_sig_name_init;
1482
1483 if (_have_async_interrupt) {
1484 *sig2num = \&Async::Interrupt::sig2num;
1485 *sig2name = \&Async::Interrupt::sig2name;
1486 } else {
1487 require Config;
1488
1489 my %signame2num;
1490 @signame2num{ split ' ', $Config::Config{sig_name} }
1491 = split ' ', $Config::Config{sig_num};
1492
1493 my @signum2name;
1494 @signum2name[values %signame2num] = keys %signame2num;
1495
1496 *sig2num = sub($) {
1497 $_[0] > 0 ? shift : $signame2num{+shift}
1498 };
1499 *sig2name = sub ($) {
1500 $_[0] > 0 ? $signum2name[+shift] : shift
1501 };
1502 }
1503 };
1504 die if $@;
1505};
1506
1507sub sig2num ($) { &$_sig_name_init; &sig2num }
1508sub sig2name($) { &$_sig_name_init; &sig2name }
887 1509
888sub signal { 1510sub signal {
1511 eval q{ # poor man's autoloading {}
1512 # probe for availability of Async::Interrupt
1513 if (_have_async_interrupt) {
1514 warn "AnyEvent: using Async::Interrupt for race-free signal handling.\n" if $VERBOSE >= 8;
1515
1516 $SIGPIPE_R = new Async::Interrupt::EventPipe;
1517 $SIG_IO = AE::io $SIGPIPE_R->fileno, 0, \&_signal_exec;
1518
1519 } else {
1520 warn "AnyEvent: using emulated perl signal handling with latency timer.\n" if $VERBOSE >= 8;
1521
1522 if (AnyEvent::WIN32) {
1523 require AnyEvent::Util;
1524
1525 ($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
1526 AnyEvent::Util::fh_nonblocking ($SIGPIPE_R, 1) if $SIGPIPE_R;
1527 AnyEvent::Util::fh_nonblocking ($SIGPIPE_W, 1) if $SIGPIPE_W; # just in case
1528 } else {
1529 pipe $SIGPIPE_R, $SIGPIPE_W;
1530 fcntl $SIGPIPE_R, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_R;
1531 fcntl $SIGPIPE_W, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_W; # just in case
1532
1533 # not strictly required, as $^F is normally 2, but let's make sure...
1534 fcntl $SIGPIPE_R, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
1535 fcntl $SIGPIPE_W, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
1536 }
1537
1538 $SIGPIPE_R
1539 or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
1540
1541 $SIG_IO = AE::io $SIGPIPE_R, 0, \&_signal_exec;
1542 }
1543
1544 *signal = $HAVE_ASYNC_INTERRUPT
1545 ? sub {
889 my (undef, %arg) = @_; 1546 my (undef, %arg) = @_;
890 1547
1548 # async::interrupt
891 my $signal = uc $arg{signal} 1549 my $signal = sig2num $arg{signal};
892 or Carp::croak "required option 'signal' is missing";
893
894 $SIG_CB{$signal}{$arg{cb}} = $arg{cb}; 1550 $SIG_CB{$signal}{$arg{cb}} = $arg{cb};
1551
1552 $SIG_ASY{$signal} ||= new Async::Interrupt
1553 cb => sub { undef $SIG_EV{$signal} },
1554 signal => $signal,
1555 pipe => [$SIGPIPE_R->filenos],
1556 pipe_autodrain => 0,
1557 ;
1558
1559 bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
1560 }
1561 : sub {
1562 my (undef, %arg) = @_;
1563
1564 # pure perl
1565 my $signal = sig2name $arg{signal};
1566 $SIG_CB{$signal}{$arg{cb}} = $arg{cb};
1567
895 $SIG{$signal} ||= sub { 1568 $SIG{$signal} ||= sub {
1569 local $!;
1570 syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
1571 undef $SIG_EV{$signal};
1572 };
1573
1574 # can't do signal processing without introducing races in pure perl,
1575 # so limit the signal latency.
1576 _sig_add;
1577
1578 bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
1579 }
1580 ;
1581
1582 *AnyEvent::Base::signal::DESTROY = sub {
1583 my ($signal, $cb) = @{$_[0]};
1584
1585 _sig_del;
1586
1587 delete $SIG_CB{$signal}{$cb};
1588
1589 $HAVE_ASYNC_INTERRUPT
1590 ? delete $SIG_ASY{$signal}
1591 : # delete doesn't work with older perls - they then
1592 # print weird messages, or just unconditionally exit
1593 # instead of getting the default action.
1594 undef $SIG{$signal}
1595 unless keys %{ $SIG_CB{$signal} };
1596 };
1597
1598 *_signal_exec = sub {
1599 $HAVE_ASYNC_INTERRUPT
1600 ? $SIGPIPE_R->drain
1601 : sysread $SIGPIPE_R, (my $dummy), 9;
1602
1603 while (%SIG_EV) {
1604 for (keys %SIG_EV) {
1605 delete $SIG_EV{$_};
896 $_->() for values %{ $SIG_CB{$signal} || {} }; 1606 $_->() for values %{ $SIG_CB{$_} || {} };
1607 }
1608 }
1609 };
897 }; 1610 };
1611 die if $@;
898 1612
899 bless [$signal, $arg{cb}], "AnyEvent::Base::Signal" 1613 &signal
900}
901
902sub AnyEvent::Base::Signal::DESTROY {
903 my ($signal, $cb) = @{$_[0]};
904
905 delete $SIG_CB{$signal}{$cb};
906
907 $SIG{$signal} = 'DEFAULT' unless keys %{ $SIG_CB{$signal} };
908} 1614}
909 1615
910# default implementation for ->child 1616# default implementation for ->child
911 1617
912our %PID_CB; 1618our %PID_CB;
913our $CHLD_W; 1619our $CHLD_W;
914our $CHLD_DELAY_W; 1620our $CHLD_DELAY_W;
915our $PID_IDLE;
916our $WNOHANG; 1621our $WNOHANG;
917 1622
918sub _child_wait { 1623# used by many Impl's
919 while (0 < (my $pid = waitpid -1, $WNOHANG)) { 1624sub _emit_childstatus($$) {
1625 my (undef, $rpid, $rstatus) = @_;
1626
1627 $_->($rpid, $rstatus)
920 $_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }), 1628 for values %{ $PID_CB{$rpid} || {} },
921 (values %{ $PID_CB{0} || {} }); 1629 values %{ $PID_CB{0} || {} };
922 }
923
924 undef $PID_IDLE;
925}
926
927sub _sigchld {
928 # make sure we deliver these changes "synchronous" with the event loop.
929 $CHLD_DELAY_W ||= AnyEvent->timer (after => 0, cb => sub {
930 undef $CHLD_DELAY_W;
931 &_child_wait;
932 });
933} 1630}
934 1631
935sub child { 1632sub child {
1633 eval q{ # poor man's autoloading {}
1634 *_sigchld = sub {
1635 my $pid;
1636
1637 AnyEvent->_emit_childstatus ($pid, $?)
1638 while ($pid = waitpid -1, $WNOHANG) > 0;
1639 };
1640
1641 *child = sub {
936 my (undef, %arg) = @_; 1642 my (undef, %arg) = @_;
937 1643
938 defined (my $pid = $arg{pid} + 0) 1644 defined (my $pid = $arg{pid} + 0)
939 or Carp::croak "required option 'pid' is missing"; 1645 or Carp::croak "required option 'pid' is missing";
940 1646
941 $PID_CB{$pid}{$arg{cb}} = $arg{cb}; 1647 $PID_CB{$pid}{$arg{cb}} = $arg{cb};
942 1648
943 unless ($WNOHANG) { 1649 # WNOHANG is almost cetrainly 1 everywhere
1650 $WNOHANG ||= $^O =~ /^(?:openbsd|netbsd|linux|freebsd|cygwin|MSWin32)$/
1651 ? 1
944 $WNOHANG = eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1; 1652 : eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
945 }
946 1653
947 unless ($CHLD_W) { 1654 unless ($CHLD_W) {
948 $CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld); 1655 $CHLD_W = AE::signal CHLD => \&_sigchld;
949 # child could be a zombie already, so make at least one round 1656 # child could be a zombie already, so make at least one round
950 &_sigchld; 1657 &_sigchld;
951 } 1658 }
952 1659
953 bless [$pid, $arg{cb}], "AnyEvent::Base::Child" 1660 bless [$pid, $arg{cb}], "AnyEvent::Base::child"
954} 1661 };
955 1662
956sub AnyEvent::Base::Child::DESTROY { 1663 *AnyEvent::Base::child::DESTROY = sub {
957 my ($pid, $cb) = @{$_[0]}; 1664 my ($pid, $cb) = @{$_[0]};
958 1665
959 delete $PID_CB{$pid}{$cb}; 1666 delete $PID_CB{$pid}{$cb};
960 delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} }; 1667 delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
961 1668
962 undef $CHLD_W unless keys %PID_CB; 1669 undef $CHLD_W unless keys %PID_CB;
1670 };
1671 };
1672 die if $@;
1673
1674 &child
1675}
1676
1677# idle emulation is done by simply using a timer, regardless
1678# of whether the process is idle or not, and not letting
1679# the callback use more than 50% of the time.
1680sub idle {
1681 eval q{ # poor man's autoloading {}
1682 *idle = sub {
1683 my (undef, %arg) = @_;
1684
1685 my ($cb, $w, $rcb) = $arg{cb};
1686
1687 $rcb = sub {
1688 if ($cb) {
1689 $w = _time;
1690 &$cb;
1691 $w = _time - $w;
1692
1693 # never use more then 50% of the time for the idle watcher,
1694 # within some limits
1695 $w = 0.0001 if $w < 0.0001;
1696 $w = 5 if $w > 5;
1697
1698 $w = AE::timer $w, 0, $rcb;
1699 } else {
1700 # clean up...
1701 undef $w;
1702 undef $rcb;
1703 }
1704 };
1705
1706 $w = AE::timer 0.05, 0, $rcb;
1707
1708 bless \\$cb, "AnyEvent::Base::idle"
1709 };
1710
1711 *AnyEvent::Base::idle::DESTROY = sub {
1712 undef $${$_[0]};
1713 };
1714 };
1715 die if $@;
1716
1717 &idle
963} 1718}
964 1719
965package AnyEvent::CondVar; 1720package AnyEvent::CondVar;
966 1721
967our @ISA = AnyEvent::CondVar::Base::; 1722our @ISA = AnyEvent::CondVar::Base::;
968 1723
969package AnyEvent::CondVar::Base; 1724package AnyEvent::CondVar::Base;
970 1725
971use overload 1726#use overload
972 '&{}' => sub { my $self = shift; sub { $self->send (@_) } }, 1727# '&{}' => sub { my $self = shift; sub { $self->send (@_) } },
973 fallback => 1; 1728# fallback => 1;
1729
1730# save 300+ kilobytes by dirtily hardcoding overloading
1731${"AnyEvent::CondVar::Base::OVERLOAD"}{dummy}++; # Register with magic by touching.
1732*{'AnyEvent::CondVar::Base::()'} = sub { }; # "Make it findable via fetchmethod."
1733*{'AnyEvent::CondVar::Base::(&{}'} = sub { my $self = shift; sub { $self->send (@_) } }; # &{}
1734${'AnyEvent::CondVar::Base::()'} = 1; # fallback
1735
1736our $WAITING;
974 1737
975sub _send { 1738sub _send {
976 # nop 1739 # nop
977} 1740}
978 1741
991sub ready { 1754sub ready {
992 $_[0]{_ae_sent} 1755 $_[0]{_ae_sent}
993} 1756}
994 1757
995sub _wait { 1758sub _wait {
1759 $WAITING
1760 and !$_[0]{_ae_sent}
1761 and Carp::croak "AnyEvent::CondVar: recursive blocking wait detected";
1762
1763 local $WAITING = 1;
996 AnyEvent->one_event while !$_[0]{_ae_sent}; 1764 AnyEvent->one_event while !$_[0]{_ae_sent};
997} 1765}
998 1766
999sub recv { 1767sub recv {
1000 $_[0]->_wait; 1768 $_[0]->_wait;
1002 Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak}; 1770 Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
1003 wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0] 1771 wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
1004} 1772}
1005 1773
1006sub cb { 1774sub cb {
1007 $_[0]{_ae_cb} = $_[1] if @_ > 1; 1775 my $cv = shift;
1776
1777 @_
1778 and $cv->{_ae_cb} = shift
1779 and $cv->{_ae_sent}
1780 and (delete $cv->{_ae_cb})->($cv);
1781
1008 $_[0]{_ae_cb} 1782 $cv->{_ae_cb}
1009} 1783}
1010 1784
1011sub begin { 1785sub begin {
1012 ++$_[0]{_ae_counter}; 1786 ++$_[0]{_ae_counter};
1013 $_[0]{_ae_end_cb} = $_[1] if @_ > 1; 1787 $_[0]{_ae_end_cb} = $_[1] if @_ > 1;
1019} 1793}
1020 1794
1021# undocumented/compatibility with pre-3.4 1795# undocumented/compatibility with pre-3.4
1022*broadcast = \&send; 1796*broadcast = \&send;
1023*wait = \&_wait; 1797*wait = \&_wait;
1798
1799=head1 ERROR AND EXCEPTION HANDLING
1800
1801In general, AnyEvent does not do any error handling - it relies on the
1802caller to do that if required. The L<AnyEvent::Strict> module (see also
1803the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
1804checking of all AnyEvent methods, however, which is highly useful during
1805development.
1806
1807As for exception handling (i.e. runtime errors and exceptions thrown while
1808executing a callback), this is not only highly event-loop specific, but
1809also not in any way wrapped by this module, as this is the job of the main
1810program.
1811
1812The pure perl event loop simply re-throws the exception (usually
1813within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
1814$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
1815so on.
1816
1817=head1 ENVIRONMENT VARIABLES
1818
1819The following environment variables are used by this module or its
1820submodules.
1821
1822Note that AnyEvent will remove I<all> environment variables starting with
1823C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
1824enabled.
1825
1826=over 4
1827
1828=item C<PERL_ANYEVENT_VERBOSE>
1829
1830By default, AnyEvent will be completely silent except in fatal
1831conditions. You can set this environment variable to make AnyEvent more
1832talkative.
1833
1834When set to C<1> or higher, causes AnyEvent to warn about unexpected
1835conditions, such as not being able to load the event model specified by
1836C<PERL_ANYEVENT_MODEL>.
1837
1838When set to C<2> or higher, cause AnyEvent to report to STDERR which event
1839model it chooses.
1840
1841When set to C<8> or higher, then AnyEvent will report extra information on
1842which optional modules it loads and how it implements certain features.
1843
1844=item C<PERL_ANYEVENT_STRICT>
1845
1846AnyEvent does not do much argument checking by default, as thorough
1847argument checking is very costly. Setting this variable to a true value
1848will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
1849check the arguments passed to most method calls. If it finds any problems,
1850it will croak.
1851
1852In other words, enables "strict" mode.
1853
1854Unlike C<use strict> (or it's modern cousin, C<< use L<common::sense>
1855>>, it is definitely recommended to keep it off in production. Keeping
1856C<PERL_ANYEVENT_STRICT=1> in your environment while developing programs
1857can be very useful, however.
1858
1859=item C<PERL_ANYEVENT_MODEL>
1860
1861This can be used to specify the event model to be used by AnyEvent, before
1862auto detection and -probing kicks in. It must be a string consisting
1863entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
1864and the resulting module name is loaded and if the load was successful,
1865used as event model. If it fails to load AnyEvent will proceed with
1866auto detection and -probing.
1867
1868This functionality might change in future versions.
1869
1870For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
1871could start your program like this:
1872
1873 PERL_ANYEVENT_MODEL=Perl perl ...
1874
1875=item C<PERL_ANYEVENT_PROTOCOLS>
1876
1877Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
1878for IPv4 or IPv6. The default is unspecified (and might change, or be the result
1879of auto probing).
1880
1881Must be set to a comma-separated list of protocols or address families,
1882current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
1883used, and preference will be given to protocols mentioned earlier in the
1884list.
1885
1886This variable can effectively be used for denial-of-service attacks
1887against local programs (e.g. when setuid), although the impact is likely
1888small, as the program has to handle conenction and other failures anyways.
1889
1890Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
1891but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
1892- only support IPv4, never try to resolve or contact IPv6
1893addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
1894IPv6, but prefer IPv6 over IPv4.
1895
1896=item C<PERL_ANYEVENT_EDNS0>
1897
1898Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
1899for DNS. This extension is generally useful to reduce DNS traffic, but
1900some (broken) firewalls drop such DNS packets, which is why it is off by
1901default.
1902
1903Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
1904EDNS0 in its DNS requests.
1905
1906=item C<PERL_ANYEVENT_MAX_FORKS>
1907
1908The maximum number of child processes that C<AnyEvent::Util::fork_call>
1909will create in parallel.
1910
1911=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
1912
1913The default value for the C<max_outstanding> parameter for the default DNS
1914resolver - this is the maximum number of parallel DNS requests that are
1915sent to the DNS server.
1916
1917=item C<PERL_ANYEVENT_RESOLV_CONF>
1918
1919The file to use instead of F</etc/resolv.conf> (or OS-specific
1920configuration) in the default resolver. When set to the empty string, no
1921default config will be used.
1922
1923=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
1924
1925When neither C<ca_file> nor C<ca_path> was specified during
1926L<AnyEvent::TLS> context creation, and either of these environment
1927variables exist, they will be used to specify CA certificate locations
1928instead of a system-dependent default.
1929
1930=item C<PERL_ANYEVENT_AVOID_GUARD> and C<PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT>
1931
1932When these are set to C<1>, then the respective modules are not
1933loaded. Mostly good for testing AnyEvent itself.
1934
1935=back
1024 1936
1025=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE 1937=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
1026 1938
1027This is an advanced topic that you do not normally need to use AnyEvent in 1939This is an advanced topic that you do not normally need to use AnyEvent in
1028a module. This section is only of use to event loop authors who want to 1940a module. This section is only of use to event loop authors who want to
1062 1974
1063I<rxvt-unicode> also cheats a bit by not providing blocking access to 1975I<rxvt-unicode> also cheats a bit by not providing blocking access to
1064condition variables: code blocking while waiting for a condition will 1976condition variables: code blocking while waiting for a condition will
1065C<die>. This still works with most modules/usages, and blocking calls must 1977C<die>. This still works with most modules/usages, and blocking calls must
1066not be done in an interactive application, so it makes sense. 1978not be done in an interactive application, so it makes sense.
1067
1068=head1 ENVIRONMENT VARIABLES
1069
1070The following environment variables are used by this module:
1071
1072=over 4
1073
1074=item C<PERL_ANYEVENT_VERBOSE>
1075
1076By default, AnyEvent will be completely silent except in fatal
1077conditions. You can set this environment variable to make AnyEvent more
1078talkative.
1079
1080When set to C<1> or higher, causes AnyEvent to warn about unexpected
1081conditions, such as not being able to load the event model specified by
1082C<PERL_ANYEVENT_MODEL>.
1083
1084When set to C<2> or higher, cause AnyEvent to report to STDERR which event
1085model it chooses.
1086
1087=item C<PERL_ANYEVENT_MODEL>
1088
1089This can be used to specify the event model to be used by AnyEvent, before
1090auto detection and -probing kicks in. It must be a string consisting
1091entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
1092and the resulting module name is loaded and if the load was successful,
1093used as event model. If it fails to load AnyEvent will proceed with
1094auto detection and -probing.
1095
1096This functionality might change in future versions.
1097
1098For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
1099could start your program like this:
1100
1101 PERL_ANYEVENT_MODEL=Perl perl ...
1102
1103=item C<PERL_ANYEVENT_PROTOCOLS>
1104
1105Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
1106for IPv4 or IPv6. The default is unspecified (and might change, or be the result
1107of auto probing).
1108
1109Must be set to a comma-separated list of protocols or address families,
1110current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
1111used, and preference will be given to protocols mentioned earlier in the
1112list.
1113
1114This variable can effectively be used for denial-of-service attacks
1115against local programs (e.g. when setuid), although the impact is likely
1116small, as the program has to handle connection errors already-
1117
1118Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
1119but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
1120- only support IPv4, never try to resolve or contact IPv6
1121addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
1122IPv6, but prefer IPv6 over IPv4.
1123
1124=item C<PERL_ANYEVENT_EDNS0>
1125
1126Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
1127for DNS. This extension is generally useful to reduce DNS traffic, but
1128some (broken) firewalls drop such DNS packets, which is why it is off by
1129default.
1130
1131Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
1132EDNS0 in its DNS requests.
1133
1134=item C<PERL_ANYEVENT_MAX_FORKS>
1135
1136The maximum number of child processes that C<AnyEvent::Util::fork_call>
1137will create in parallel.
1138
1139=back
1140 1979
1141=head1 EXAMPLE PROGRAM 1980=head1 EXAMPLE PROGRAM
1142 1981
1143The following program uses an I/O watcher to read data from STDIN, a timer 1982The following program uses an I/O watcher to read data from STDIN, a timer
1144to display a message once per second, and a condition variable to quit the 1983to display a message once per second, and a condition variable to quit the
1157 warn "read: $input\n"; # output what has been read 1996 warn "read: $input\n"; # output what has been read
1158 $cv->send if $input =~ /^q/i; # quit program if /^q/i 1997 $cv->send if $input =~ /^q/i; # quit program if /^q/i
1159 }, 1998 },
1160 ); 1999 );
1161 2000
1162 my $time_watcher; # can only be used once
1163
1164 sub new_timer {
1165 $timer = AnyEvent->timer (after => 1, cb => sub { 2001 my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub {
1166 warn "timeout\n"; # print 'timeout' about every second 2002 warn "timeout\n"; # print 'timeout' at most every second
1167 &new_timer; # and restart the time
1168 }); 2003 });
1169 }
1170
1171 new_timer; # create first timer
1172 2004
1173 $cv->recv; # wait until user enters /^q/i 2005 $cv->recv; # wait until user enters /^q/i
1174 2006
1175=head1 REAL-WORLD EXAMPLE 2007=head1 REAL-WORLD EXAMPLE
1176 2008
1249 2081
1250The actual code goes further and collects all errors (C<die>s, exceptions) 2082The actual code goes further and collects all errors (C<die>s, exceptions)
1251that occurred during request processing. The C<result> method detects 2083that occurred during request processing. The C<result> method detects
1252whether an exception as thrown (it is stored inside the $txn object) 2084whether an exception as thrown (it is stored inside the $txn object)
1253and just throws the exception, which means connection errors and other 2085and just throws the exception, which means connection errors and other
1254problems get reported tot he code that tries to use the result, not in a 2086problems get reported to the code that tries to use the result, not in a
1255random callback. 2087random callback.
1256 2088
1257All of this enables the following usage styles: 2089All of this enables the following usage styles:
1258 2090
12591. Blocking: 20911. Blocking:
1307through AnyEvent. The benchmark creates a lot of timers (with a zero 2139through AnyEvent. The benchmark creates a lot of timers (with a zero
1308timeout) and I/O watchers (watching STDOUT, a pty, to become writable, 2140timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
1309which it is), lets them fire exactly once and destroys them again. 2141which it is), lets them fire exactly once and destroys them again.
1310 2142
1311Source code for this benchmark is found as F<eg/bench> in the AnyEvent 2143Source code for this benchmark is found as F<eg/bench> in the AnyEvent
1312distribution. 2144distribution. It uses the L<AE> interface, which makes a real difference
2145for the EV and Perl backends only.
1313 2146
1314=head3 Explanation of the columns 2147=head3 Explanation of the columns
1315 2148
1316I<watcher> is the number of event watchers created/destroyed. Since 2149I<watcher> is the number of event watchers created/destroyed. Since
1317different event models feature vastly different performances, each event 2150different event models feature vastly different performances, each event
1338watcher. 2171watcher.
1339 2172
1340=head3 Results 2173=head3 Results
1341 2174
1342 name watchers bytes create invoke destroy comment 2175 name watchers bytes create invoke destroy comment
1343 EV/EV 400000 244 0.56 0.46 0.31 EV native interface 2176 EV/EV 100000 223 0.47 0.43 0.27 EV native interface
1344 EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers 2177 EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers
1345 CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal 2178 Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal
1346 Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation 2179 Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation
1347 Event/Event 16000 516 31.88 31.30 0.85 Event native interface 2180 Event/Event 16000 516 31.16 31.84 0.82 Event native interface
1348 Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers 2181 Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers
2182 IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll
2183 IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll
1349 Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour 2184 Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour
1350 Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers 2185 Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers
1351 POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event 2186 POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event
1352 POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select 2187 POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select
1353 2188
1354=head3 Discussion 2189=head3 Discussion
1355 2190
1356The benchmark does I<not> measure scalability of the event loop very 2191The benchmark does I<not> measure scalability of the event loop very
1357well. For example, a select-based event loop (such as the pure perl one) 2192well. For example, a select-based event loop (such as the pure perl one)
1369benchmark machine, handling an event takes roughly 1600 CPU cycles with 2204benchmark machine, handling an event takes roughly 1600 CPU cycles with
1370EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU 2205EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
1371cycles with POE. 2206cycles with POE.
1372 2207
1373C<EV> is the sole leader regarding speed and memory use, which are both 2208C<EV> is the sole leader regarding speed and memory use, which are both
1374maximal/minimal, respectively. Even when going through AnyEvent, it uses 2209maximal/minimal, respectively. When using the L<AE> API there is zero
2210overhead (when going through the AnyEvent API create is about 5-6 times
2211slower, with other times being equal, so still uses far less memory than
1375far less memory than any other event loop and is still faster than Event 2212any other event loop and is still faster than Event natively).
1376natively.
1377 2213
1378The pure perl implementation is hit in a few sweet spots (both the 2214The pure perl implementation is hit in a few sweet spots (both the
1379constant timeout and the use of a single fd hit optimisations in the perl 2215constant timeout and the use of a single fd hit optimisations in the perl
1380interpreter and the backend itself). Nevertheless this shows that it 2216interpreter and the backend itself). Nevertheless this shows that it
1381adds very little overhead in itself. Like any select-based backend its 2217adds very little overhead in itself. Like any select-based backend its
1382performance becomes really bad with lots of file descriptors (and few of 2218performance becomes really bad with lots of file descriptors (and few of
1383them active), of course, but this was not subject of this benchmark. 2219them active), of course, but this was not subject of this benchmark.
1384 2220
1385The C<Event> module has a relatively high setup and callback invocation 2221The C<Event> module has a relatively high setup and callback invocation
1386cost, but overall scores in on the third place. 2222cost, but overall scores in on the third place.
2223
2224C<IO::Async> performs admirably well, about on par with C<Event>, even
2225when using its pure perl backend.
1387 2226
1388C<Glib>'s memory usage is quite a bit higher, but it features a 2227C<Glib>'s memory usage is quite a bit higher, but it features a
1389faster callback invocation and overall ends up in the same class as 2228faster callback invocation and overall ends up in the same class as
1390C<Event>. However, Glib scales extremely badly, doubling the number of 2229C<Event>. However, Glib scales extremely badly, doubling the number of
1391watchers increases the processing time by more than a factor of four, 2230watchers increases the processing time by more than a factor of four,
1452In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100 2291In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
1453(1%) are active. This mirrors the activity of large servers with many 2292(1%) are active. This mirrors the activity of large servers with many
1454connections, most of which are idle at any one point in time. 2293connections, most of which are idle at any one point in time.
1455 2294
1456Source code for this benchmark is found as F<eg/bench2> in the AnyEvent 2295Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
1457distribution. 2296distribution. It uses the L<AE> interface, which makes a real difference
2297for the EV and Perl backends only.
1458 2298
1459=head3 Explanation of the columns 2299=head3 Explanation of the columns
1460 2300
1461I<sockets> is the number of sockets, and twice the number of "servers" (as 2301I<sockets> is the number of sockets, and twice the number of "servers" (as
1462each server has a read and write socket end). 2302each server has a read and write socket end).
1469it to another server. This includes deleting the old timeout and creating 2309it to another server. This includes deleting the old timeout and creating
1470a new one that moves the timeout into the future. 2310a new one that moves the timeout into the future.
1471 2311
1472=head3 Results 2312=head3 Results
1473 2313
1474 name sockets create request 2314 name sockets create request
1475 EV 20000 69.01 11.16 2315 EV 20000 62.66 7.99
1476 Perl 20000 73.32 35.87 2316 Perl 20000 68.32 32.64
1477 Event 20000 212.62 257.32 2317 IOAsync 20000 174.06 101.15 epoll
1478 Glib 20000 651.16 1896.30 2318 IOAsync 20000 174.67 610.84 poll
2319 Event 20000 202.69 242.91
2320 Glib 20000 557.01 1689.52
1479 POE 20000 349.67 12317.24 uses POE::Loop::Event 2321 POE 20000 341.54 12086.32 uses POE::Loop::Event
1480 2322
1481=head3 Discussion 2323=head3 Discussion
1482 2324
1483This benchmark I<does> measure scalability and overall performance of the 2325This benchmark I<does> measure scalability and overall performance of the
1484particular event loop. 2326particular event loop.
1486EV is again fastest. Since it is using epoll on my system, the setup time 2328EV is again fastest. Since it is using epoll on my system, the setup time
1487is relatively high, though. 2329is relatively high, though.
1488 2330
1489Perl surprisingly comes second. It is much faster than the C-based event 2331Perl surprisingly comes second. It is much faster than the C-based event
1490loops Event and Glib. 2332loops Event and Glib.
2333
2334IO::Async performs very well when using its epoll backend, and still quite
2335good compared to Glib when using its pure perl backend.
1491 2336
1492Event suffers from high setup time as well (look at its code and you will 2337Event suffers from high setup time as well (look at its code and you will
1493understand why). Callback invocation also has a high overhead compared to 2338understand why). Callback invocation also has a high overhead compared to
1494the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event 2339the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
1495uses select or poll in basically all documented configurations. 2340uses select or poll in basically all documented configurations.
1558=item * C-based event loops perform very well with small number of 2403=item * C-based event loops perform very well with small number of
1559watchers, as the management overhead dominates. 2404watchers, as the management overhead dominates.
1560 2405
1561=back 2406=back
1562 2407
2408=head2 THE IO::Lambda BENCHMARK
2409
2410Recently I was told about the benchmark in the IO::Lambda manpage, which
2411could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
2412simply compares IO::Lambda with POE, and IO::Lambda looks better (which
2413shouldn't come as a surprise to anybody). As such, the benchmark is
2414fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
2415very optimal. But how would AnyEvent compare when used without the extra
2416baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
2417
2418The benchmark itself creates an echo-server, and then, for 500 times,
2419connects to the echo server, sends a line, waits for the reply, and then
2420creates the next connection. This is a rather bad benchmark, as it doesn't
2421test the efficiency of the framework or much non-blocking I/O, but it is a
2422benchmark nevertheless.
2423
2424 name runtime
2425 Lambda/select 0.330 sec
2426 + optimized 0.122 sec
2427 Lambda/AnyEvent 0.327 sec
2428 + optimized 0.138 sec
2429 Raw sockets/select 0.077 sec
2430 POE/select, components 0.662 sec
2431 POE/select, raw sockets 0.226 sec
2432 POE/select, optimized 0.404 sec
2433
2434 AnyEvent/select/nb 0.085 sec
2435 AnyEvent/EV/nb 0.068 sec
2436 +state machine 0.134 sec
2437
2438The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
2439benchmarks actually make blocking connects and use 100% blocking I/O,
2440defeating the purpose of an event-based solution. All of the newly
2441written AnyEvent benchmarks use 100% non-blocking connects (using
2442AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
2443resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
2444generally require a lot more bookkeeping and event handling than blocking
2445connects (which involve a single syscall only).
2446
2447The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
2448offers similar expressive power as POE and IO::Lambda, using conventional
2449Perl syntax. This means that both the echo server and the client are 100%
2450non-blocking, further placing it at a disadvantage.
2451
2452As you can see, the AnyEvent + EV combination even beats the
2453hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
2454backend easily beats IO::Lambda and POE.
2455
2456And even the 100% non-blocking version written using the high-level (and
2457slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda
2458higher level ("unoptimised") abstractions by a large margin, even though
2459it does all of DNS, tcp-connect and socket I/O in a non-blocking way.
2460
2461The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
2462F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
2463part of the IO::Lambda distribution and were used without any changes.
2464
2465
2466=head1 SIGNALS
2467
2468AnyEvent currently installs handlers for these signals:
2469
2470=over 4
2471
2472=item SIGCHLD
2473
2474A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
2475emulation for event loops that do not support them natively. Also, some
2476event loops install a similar handler.
2477
2478Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE, then
2479AnyEvent will reset it to default, to avoid losing child exit statuses.
2480
2481=item SIGPIPE
2482
2483A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
2484when AnyEvent gets loaded.
2485
2486The rationale for this is that AnyEvent users usually do not really depend
2487on SIGPIPE delivery (which is purely an optimisation for shell use, or
2488badly-written programs), but C<SIGPIPE> can cause spurious and rare
2489program exits as a lot of people do not expect C<SIGPIPE> when writing to
2490some random socket.
2491
2492The rationale for installing a no-op handler as opposed to ignoring it is
2493that this way, the handler will be restored to defaults on exec.
2494
2495Feel free to install your own handler, or reset it to defaults.
2496
2497=back
2498
2499=cut
2500
2501undef $SIG{CHLD}
2502 if $SIG{CHLD} eq 'IGNORE';
2503
2504$SIG{PIPE} = sub { }
2505 unless defined $SIG{PIPE};
2506
2507=head1 RECOMMENDED/OPTIONAL MODULES
2508
2509One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
2510it's built-in modules) are required to use it.
2511
2512That does not mean that AnyEvent won't take advantage of some additional
2513modules if they are installed.
2514
2515This section explains which additional modules will be used, and how they
2516affect AnyEvent's operation.
2517
2518=over 4
2519
2520=item L<Async::Interrupt>
2521
2522This slightly arcane module is used to implement fast signal handling: To
2523my knowledge, there is no way to do completely race-free and quick
2524signal handling in pure perl. To ensure that signals still get
2525delivered, AnyEvent will start an interval timer to wake up perl (and
2526catch the signals) with some delay (default is 10 seconds, look for
2527C<$AnyEvent::MAX_SIGNAL_LATENCY>).
2528
2529If this module is available, then it will be used to implement signal
2530catching, which means that signals will not be delayed, and the event loop
2531will not be interrupted regularly, which is more efficient (and good for
2532battery life on laptops).
2533
2534This affects not just the pure-perl event loop, but also other event loops
2535that have no signal handling on their own (e.g. Glib, Tk, Qt).
2536
2537Some event loops (POE, Event, Event::Lib) offer signal watchers natively,
2538and either employ their own workarounds (POE) or use AnyEvent's workaround
2539(using C<$AnyEvent::MAX_SIGNAL_LATENCY>). Installing L<Async::Interrupt>
2540does nothing for those backends.
2541
2542=item L<EV>
2543
2544This module isn't really "optional", as it is simply one of the backend
2545event loops that AnyEvent can use. However, it is simply the best event
2546loop available in terms of features, speed and stability: It supports
2547the AnyEvent API optimally, implements all the watcher types in XS, does
2548automatic timer adjustments even when no monotonic clock is available,
2549can take avdantage of advanced kernel interfaces such as C<epoll> and
2550C<kqueue>, and is the fastest backend I<by far>. You can even embed
2551L<Glib>/L<Gtk2> in it (or vice versa, see L<EV::Glib> and L<Glib::EV>).
2552
2553If you only use backends that rely on another event loop (e.g. C<Tk>),
2554then this module will do nothing for you.
2555
2556=item L<Guard>
2557
2558The guard module, when used, will be used to implement
2559C<AnyEvent::Util::guard>. This speeds up guards considerably (and uses a
2560lot less memory), but otherwise doesn't affect guard operation much. It is
2561purely used for performance.
2562
2563=item L<JSON> and L<JSON::XS>
2564
2565One of these modules is required when you want to read or write JSON data
2566via L<AnyEvent::Handle>. L<JSON> is also written in pure-perl, but can take
2567advantage of the ultra-high-speed L<JSON::XS> module when it is installed.
2568
2569=item L<Net::SSLeay>
2570
2571Implementing TLS/SSL in Perl is certainly interesting, but not very
2572worthwhile: If this module is installed, then L<AnyEvent::Handle> (with
2573the help of L<AnyEvent::TLS>), gains the ability to do TLS/SSL.
2574
2575=item L<Time::HiRes>
2576
2577This module is part of perl since release 5.008. It will be used when the
2578chosen event library does not come with a timing source on it's own. The
2579pure-perl event loop (L<AnyEvent::Impl::Perl>) will additionally use it to
2580try to use a monotonic clock for timing stability.
2581
2582=back
2583
1563 2584
1564=head1 FORK 2585=head1 FORK
1565 2586
1566Most event libraries are not fork-safe. The ones who are usually are 2587Most event libraries are not fork-safe. The ones who are usually are
1567because they rely on inefficient but fork-safe C<select> or C<poll> 2588because they rely on inefficient but fork-safe C<select> or C<poll> calls
1568calls. Only L<EV> is fully fork-aware. 2589- higher performance APIs such as BSD's kqueue or the dreaded Linux epoll
2590are usually badly thought-out hacks that are incompatible with fork in
2591one way or another. Only L<EV> is fully fork-aware and ensures that you
2592continue event-processing in both parent and child (or both, if you know
2593what you are doing).
2594
2595This means that, in general, you cannot fork and do event processing in
2596the child if the event library was initialised before the fork (which
2597usually happens when the first AnyEvent watcher is created, or the library
2598is loaded).
1569 2599
1570If you have to fork, you must either do so I<before> creating your first 2600If you have to fork, you must either do so I<before> creating your first
1571watcher OR you must not use AnyEvent at all in the child. 2601watcher OR you must not use AnyEvent at all in the child OR you must do
2602something completely out of the scope of AnyEvent.
2603
2604The problem of doing event processing in the parent I<and> the child
2605is much more complicated: even for backends that I<are> fork-aware or
2606fork-safe, their behaviour is not usually what you want: fork clones all
2607watchers, that means all timers, I/O watchers etc. are active in both
2608parent and child, which is almost never what you want. USing C<exec>
2609to start worker children from some kind of manage rprocess is usually
2610preferred, because it is much easier and cleaner, at the expense of having
2611to have another binary.
1572 2612
1573 2613
1574=head1 SECURITY CONSIDERATIONS 2614=head1 SECURITY CONSIDERATIONS
1575 2615
1576AnyEvent can be forced to load any event model via 2616AnyEvent can be forced to load any event model via
1581specified in the variable. 2621specified in the variable.
1582 2622
1583You can make AnyEvent completely ignore this variable by deleting it 2623You can make AnyEvent completely ignore this variable by deleting it
1584before the first watcher gets created, e.g. with a C<BEGIN> block: 2624before the first watcher gets created, e.g. with a C<BEGIN> block:
1585 2625
1586 BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} } 2626 BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
1587 2627
1588 use AnyEvent; 2628 use AnyEvent;
1589 2629
1590Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can 2630Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
1591be used to probe what backend is used and gain other information (which is 2631be used to probe what backend is used and gain other information (which is
1592probably even less useful to an attacker than PERL_ANYEVENT_MODEL). 2632probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
2633$ENV{PERL_ANYEVENT_STRICT}.
2634
2635Note that AnyEvent will remove I<all> environment variables starting with
2636C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
2637enabled.
2638
2639
2640=head1 BUGS
2641
2642Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
2643to work around. If you suffer from memleaks, first upgrade to Perl 5.10
2644and check wether the leaks still show up. (Perl 5.10.0 has other annoying
2645memleaks, such as leaking on C<map> and C<grep> but it is usually not as
2646pronounced).
1593 2647
1594 2648
1595=head1 SEE ALSO 2649=head1 SEE ALSO
1596 2650
1597Utility functions: L<AnyEvent::Util>. 2651Utility functions: L<AnyEvent::Util>.
1600L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>. 2654L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
1601 2655
1602Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>, 2656Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
1603L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, 2657L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
1604L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>, 2658L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
1605L<AnyEvent::Impl::POE>. 2659L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>, L<Anyevent::Impl::Irssi>.
1606 2660
1607Non-blocking file handles, sockets, TCP clients and 2661Non-blocking file handles, sockets, TCP clients and
1608servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>. 2662servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>.
1609 2663
1610Asynchronous DNS: L<AnyEvent::DNS>. 2664Asynchronous DNS: L<AnyEvent::DNS>.
1611 2665
1612Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>, 2666Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>,
2667L<Coro::Event>,
1613 2668
1614Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>. 2669Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::XMPP>,
2670L<AnyEvent::HTTP>.
1615 2671
1616 2672
1617=head1 AUTHOR 2673=head1 AUTHOR
1618 2674
1619 Marc Lehmann <schmorp@schmorp.de> 2675 Marc Lehmann <schmorp@schmorp.de>
1620 http://home.schmorp.de/ 2676 http://home.schmorp.de/
1621 2677
1622=cut 2678=cut
1623 2679
16241 26801
1625 2681

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines