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 1.103 by root, Tue Apr 29 07:15:49 2008 UTC vs.
Revision 1.266 by root, Thu Jul 30 03:41:56 2009 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, Coro::EV, Coro::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 # file descriptor readable
11 my $w = AnyEvent->io (fh => $fh, poll => "r|w", cb => sub { 13 my $w = AnyEvent->io (fh => $fh, poll => "r", cb => sub { ... });
14
15 # one-shot or repeating timers
16 my $w = AnyEvent->timer (after => $seconds, cb => sub { ... });
17 my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ...
18
19 print AnyEvent->now; # prints current event loop time
20 print AnyEvent->time; # think Time::HiRes::time or simply CORE::time.
21
22 # POSIX signal
23 my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... });
24
25 # child process exit
26 my $w = AnyEvent->child (pid => $pid, cb => sub {
27 my ($pid, $status) = @_;
12 ... 28 ...
13 }); 29 });
14 30
15 my $w = AnyEvent->timer (after => $seconds, cb => sub { 31 # called when event loop idle (if applicable)
16 ... 32 my $w = AnyEvent->idle (cb => sub { ... });
17 });
18 33
19 my $w = AnyEvent->condvar; # stores whether a condition was flagged 34 my $w = AnyEvent->condvar; # stores whether a condition was flagged
35 $w->send; # wake up current and all future recv's
20 $w->wait; # enters "main loop" till $condvar gets ->broadcast 36 $w->recv; # enters "main loop" till $condvar gets ->send
21 $w->broadcast; # wake up current and all future wait's 37 # use a condvar in callback mode:
38 $w->cb (sub { $_[0]->recv });
39
40=head1 INTRODUCTION/TUTORIAL
41
42This manpage is mainly a reference manual. If you are interested
43in a tutorial or some gentle introduction, have a look at the
44L<AnyEvent::Intro> manpage.
45
46=head1 SUPPORT
47
48There is a mailinglist for discussing all things AnyEvent, and an IRC
49channel, too.
50
51See the AnyEvent project page at the B<Schmorpforge Ta-Sa Software
52Repository>, at L<http://anyevent.schmorp.de>, for more info.
22 53
23=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT) 54=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
24 55
25Glib, POE, IO::Async, Event... CPAN offers event models by the dozen 56Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
26nowadays. So what is different about AnyEvent? 57nowadays. So what is different about AnyEvent?
27 58
28Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of 59Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
29policy> and AnyEvent is I<small and efficient>. 60policy> and AnyEvent is I<small and efficient>.
30 61
31First and foremost, I<AnyEvent is not an event model> itself, it only 62First 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 63interfaces to whatever event model the main program happens to use, in a
33pragmatic way. For event models and certain classes of immortals alike, 64pragmatic way. For event models and certain classes of immortals alike,
34the statement "there can only be one" is a bitter reality: In general, 65the 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 66only one event loop can be active at the same time in a process. AnyEvent
36helps hiding the differences between those event loops. 67cannot change this, but it can hide the differences between those event
68loops.
37 69
38The goal of AnyEvent is to offer module authors the ability to do event 70The 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 71programming (waiting for I/O or timer events) without subscribing to a
40religion, a way of living, and most importantly: without forcing your 72religion, a way of living, and most importantly: without forcing your
41module users into the same thing by forcing them to use the same event 73module users into the same thing by forcing them to use the same event
42model you use. 74model you use.
43 75
44For modules like POE or IO::Async (which is a total misnomer as it is 76For 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 77actually 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 78like 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 79cannot use anything else, as they are simply incompatible to everything
48isn't itself. What's worse, all the potential users of your module are 80that isn't them. What's worse, all the potential users of your
49I<also> forced to use the same event loop you use. 81module are I<also> forced to use the same event loop you use.
50 82
51AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works 83AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
52fine. AnyEvent + Tk works fine etc. etc. but none of these work together 84fine. 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 85with 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, 86your 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 87too. But if your module uses AnyEvent, it works transparently with all
56event models it supports (including stuff like POE and IO::Async, as long 88event 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 89use one of the supported event loops. It is trivial to add new event loops
58event loops to AnyEvent, too, so it is future-proof). 90to AnyEvent, too, so it is future-proof).
59 91
60In addition to being free of having to use I<the one and only true event 92In 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 93model>, AnyEvent also is free of bloat and policy: with POE or similar
62modules, you get an enourmous amount of code and strict rules you have to 94modules, 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 95follow. AnyEvent, on the other hand, is lean and up to the point, by only
64offering the functionality that is necessary, in as thin as a wrapper as 96offering the functionality that is necessary, in as thin as a wrapper as
65technically possible. 97technically possible.
66 98
99Of course, AnyEvent comes with a big (and fully optional!) toolbox
100of useful functionality, such as an asynchronous DNS resolver, 100%
101non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
102such as Windows) and lots of real-world knowledge and workarounds for
103platform bugs and differences.
104
67Of course, if you want lots of policy (this can arguably be somewhat 105Now, if you I<do want> lots of policy (this can arguably be somewhat
68useful) and you want to force your users to use the one and only event 106useful) and you want to force your users to use the one and only event
69model, you should I<not> use this module. 107model, you should I<not> use this module.
70 108
71=head1 DESCRIPTION 109=head1 DESCRIPTION
72 110
78The interface itself is vaguely similar, but not identical to the L<Event> 116The interface itself is vaguely similar, but not identical to the L<Event>
79module. 117module.
80 118
81During the first call of any watcher-creation method, the module tries 119During the first call of any watcher-creation method, the module tries
82to detect the currently loaded event loop by probing whether one of the 120to detect the currently loaded event loop by probing whether one of the
83following modules is already loaded: L<Coro::EV>, L<Coro::Event>, L<EV>, 121following modules is already loaded: L<EV>,
84L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>, 122L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>,
85L<POE>. The first one found is used. If none are found, the module tries 123L<POE>. The first one found is used. If none are found, the module tries
86to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl 124to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
87adaptor should always succeed) in the order given. The first one that can 125adaptor should always succeed) in the order given. The first one that can
88be successfully loaded will be used. If, after this, still none could be 126be successfully loaded will be used. If, after this, still none could be
102starts using it, all bets are off. Maybe you should tell their authors to 140starts using it, all bets are off. Maybe you should tell their authors to
103use AnyEvent so their modules work together with others seamlessly... 141use AnyEvent so their modules work together with others seamlessly...
104 142
105The pure-perl implementation of AnyEvent is called 143The pure-perl implementation of AnyEvent is called
106C<AnyEvent::Impl::Perl>. Like other event modules you can load it 144C<AnyEvent::Impl::Perl>. Like other event modules you can load it
107explicitly. 145explicitly and enjoy the high availability of that event loop :)
108 146
109=head1 WATCHERS 147=head1 WATCHERS
110 148
111AnyEvent has the central concept of a I<watcher>, which is an object that 149AnyEvent has the central concept of a I<watcher>, which is an object that
112stores relevant data for each kind of event you are waiting for, such as 150stores relevant data for each kind of event you are waiting for, such as
113the callback to call, the filehandle to watch, etc. 151the callback to call, the file handle to watch, etc.
114 152
115These watchers are normal Perl objects with normal Perl lifetime. After 153These watchers are normal Perl objects with normal Perl lifetime. After
116creating a watcher it will immediately "watch" for events and invoke the 154creating a watcher it will immediately "watch" for events and invoke the
117callback when the event occurs (of course, only when the event model 155callback when the event occurs (of course, only when the event model
118is in control). 156is in control).
119 157
158Note that B<callbacks must not permanently change global variables>
159potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
160callbacks must not C<die> >>. The former is good programming practise in
161Perl and the latter stems from the fact that exception handling differs
162widely between event loops.
163
120To disable the watcher you have to destroy it (e.g. by setting the 164To disable the watcher you have to destroy it (e.g. by setting the
121variable you store it in to C<undef> or otherwise deleting all references 165variable you store it in to C<undef> or otherwise deleting all references
122to it). 166to it).
123 167
124All watchers are created by calling a method on the C<AnyEvent> class. 168All watchers are created by calling a method on the C<AnyEvent> class.
126Many watchers either are used with "recursion" (repeating timers for 170Many watchers either are used with "recursion" (repeating timers for
127example), or need to refer to their watcher object in other ways. 171example), or need to refer to their watcher object in other ways.
128 172
129An any way to achieve that is this pattern: 173An any way to achieve that is this pattern:
130 174
131 my $w; $w = AnyEvent->type (arg => value ..., cb => sub { 175 my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
132 # you can use $w here, for example to undef it 176 # you can use $w here, for example to undef it
133 undef $w; 177 undef $w;
134 }); 178 });
135 179
136Note that C<my $w; $w => combination. This is necessary because in Perl, 180Note that C<my $w; $w => combination. This is necessary because in Perl,
137my variables are only visible after the statement in which they are 181my variables are only visible after the statement in which they are
138declared. 182declared.
139 183
140=head2 I/O WATCHERS 184=head2 I/O WATCHERS
141 185
186 $w = AnyEvent->io (
187 fh => <filehandle_or_fileno>,
188 poll => <"r" or "w">,
189 cb => <callback>,
190 );
191
142You can create an I/O watcher by calling the C<< AnyEvent->io >> method 192You can create an I/O watcher by calling the C<< AnyEvent->io >> method
143with the following mandatory key-value pairs as arguments: 193with the following mandatory key-value pairs as arguments:
144 194
145C<fh> the Perl I<file handle> (I<not> file descriptor) to watch 195C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch
196for events (AnyEvent might or might not keep a reference to this file
197handle). Note that only file handles pointing to things for which
198non-blocking operation makes sense are allowed. This includes sockets,
199most character devices, pipes, fifos and so on, but not for example files
200or block devices.
201
146for events. C<poll> must be a string that is either C<r> or C<w>, 202C<poll> must be a string that is either C<r> or C<w>, which creates a
147which creates a watcher waiting for "r"eadable or "w"ritable events, 203watcher waiting for "r"eadable or "w"ritable events, respectively.
204
148respectively. C<cb> is the callback to invoke each time the file handle 205C<cb> is the callback to invoke each time the file handle becomes ready.
149becomes ready.
150 206
151Although the callback might get passed parameters, their value and 207Although the callback might get passed parameters, their value and
152presence is undefined and you cannot rely on them. Portable AnyEvent 208presence is undefined and you cannot rely on them. Portable AnyEvent
153callbacks cannot use arguments passed to I/O watcher callbacks. 209callbacks cannot use arguments passed to I/O watcher callbacks.
154 210
158 214
159Some event loops issue spurious readyness notifications, so you should 215Some event loops issue spurious readyness notifications, so you should
160always use non-blocking calls when reading/writing from/to your file 216always use non-blocking calls when reading/writing from/to your file
161handles. 217handles.
162 218
163Example:
164
165 # wait for readability of STDIN, then read a line and disable the watcher 219Example: wait for readability of STDIN, then read a line and disable the
220watcher.
221
166 my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub { 222 my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
167 chomp (my $input = <STDIN>); 223 chomp (my $input = <STDIN>);
168 warn "read: $input\n"; 224 warn "read: $input\n";
169 undef $w; 225 undef $w;
170 }); 226 });
171 227
172=head2 TIME WATCHERS 228=head2 TIME WATCHERS
173 229
230 $w = AnyEvent->timer (after => <seconds>, cb => <callback>);
231
232 $w = AnyEvent->timer (
233 after => <fractional_seconds>,
234 interval => <fractional_seconds>,
235 cb => <callback>,
236 );
237
174You can create a time watcher by calling the C<< AnyEvent->timer >> 238You can create a time watcher by calling the C<< AnyEvent->timer >>
175method with the following mandatory arguments: 239method with the following mandatory arguments:
176 240
177C<after> specifies after how many seconds (fractional values are 241C<after> specifies after how many seconds (fractional values are
178supported) the callback should be invoked. C<cb> is the callback to invoke 242supported) the callback should be invoked. C<cb> is the callback to invoke
180 244
181Although the callback might get passed parameters, their value and 245Although the callback might get passed parameters, their value and
182presence is undefined and you cannot rely on them. Portable AnyEvent 246presence is undefined and you cannot rely on them. Portable AnyEvent
183callbacks cannot use arguments passed to time watcher callbacks. 247callbacks cannot use arguments passed to time watcher callbacks.
184 248
185The timer callback will be invoked at most once: if you want a repeating 249The callback will normally be invoked once only. If you specify another
186timer you have to create a new watcher (this is a limitation by both Tk 250parameter, C<interval>, as a strictly positive number (> 0), then the
187and Glib). 251callback will be invoked regularly at that interval (in fractional
252seconds) after the first invocation. If C<interval> is specified with a
253false value, then it is treated as if it were missing.
188 254
189Example: 255The callback will be rescheduled before invoking the callback, but no
256attempt is done to avoid timer drift in most backends, so the interval is
257only approximate.
190 258
191 # fire an event after 7.7 seconds 259Example: fire an event after 7.7 seconds.
260
192 my $w = AnyEvent->timer (after => 7.7, cb => sub { 261 my $w = AnyEvent->timer (after => 7.7, cb => sub {
193 warn "timeout\n"; 262 warn "timeout\n";
194 }); 263 });
195 264
196 # to cancel the timer: 265 # to cancel the timer:
197 undef $w; 266 undef $w;
198 267
199Example 2:
200
201 # fire an event after 0.5 seconds, then roughly every second 268Example 2: fire an event after 0.5 seconds, then roughly every second.
202 my $w;
203 269
204 my $cb = sub {
205 # cancel the old timer while creating a new one
206 $w = AnyEvent->timer (after => 1, cb => $cb); 270 my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
271 warn "timeout\n";
207 }; 272 };
208
209 # start the "loop" by creating the first watcher
210 $w = AnyEvent->timer (after => 0.5, cb => $cb);
211 273
212=head3 TIMING ISSUES 274=head3 TIMING ISSUES
213 275
214There are two ways to handle timers: based on real time (relative, "fire 276There are two ways to handle timers: based on real time (relative, "fire
215in 10 seconds") and based on wallclock time (absolute, "fire at 12 277in 10 seconds") and based on wallclock time (absolute, "fire at 12
227timers. 289timers.
228 290
229AnyEvent always prefers relative timers, if available, matching the 291AnyEvent always prefers relative timers, if available, matching the
230AnyEvent API. 292AnyEvent API.
231 293
294AnyEvent has two additional methods that return the "current time":
295
296=over 4
297
298=item AnyEvent->time
299
300This returns the "current wallclock time" as a fractional number of
301seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
302return, and the result is guaranteed to be compatible with those).
303
304It progresses independently of any event loop processing, i.e. each call
305will check the system clock, which usually gets updated frequently.
306
307=item AnyEvent->now
308
309This also returns the "current wallclock time", but unlike C<time>, above,
310this value might change only once per event loop iteration, depending on
311the event loop (most return the same time as C<time>, above). This is the
312time that AnyEvent's timers get scheduled against.
313
314I<In almost all cases (in all cases if you don't care), this is the
315function to call when you want to know the current time.>
316
317This function is also often faster then C<< AnyEvent->time >>, and
318thus the preferred method if you want some timestamp (for example,
319L<AnyEvent::Handle> uses this to update it's activity timeouts).
320
321The rest of this section is only of relevance if you try to be very exact
322with your timing, you can skip it without bad conscience.
323
324For a practical example of when these times differ, consider L<Event::Lib>
325and L<EV> and the following set-up:
326
327The event loop is running and has just invoked one of your callback at
328time=500 (assume no other callbacks delay processing). In your callback,
329you wait a second by executing C<sleep 1> (blocking the process for a
330second) and then (at time=501) you create a relative timer that fires
331after three seconds.
332
333With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
334both return C<501>, because that is the current time, and the timer will
335be scheduled to fire at time=504 (C<501> + C<3>).
336
337With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
338time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
339last event processing phase started. With L<EV>, your timer gets scheduled
340to run at time=503 (C<500> + C<3>).
341
342In one sense, L<Event::Lib> is more exact, as it uses the current time
343regardless of any delays introduced by event processing. However, most
344callbacks do not expect large delays in processing, so this causes a
345higher drift (and a lot more system calls to get the current time).
346
347In another sense, L<EV> is more exact, as your timer will be scheduled at
348the same time, regardless of how long event processing actually took.
349
350In either case, if you care (and in most cases, you don't), then you
351can get whatever behaviour you want with any event loop, by taking the
352difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
353account.
354
355=item AnyEvent->now_update
356
357Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
358the current time for each loop iteration (see the discussion of L<<
359AnyEvent->now >>, above).
360
361When a callback runs for a long time (or when the process sleeps), then
362this "current" time will differ substantially from the real time, which
363might affect timers and time-outs.
364
365When this is the case, you can call this method, which will update the
366event loop's idea of "current time".
367
368Note that updating the time I<might> cause some events to be handled.
369
370=back
371
232=head2 SIGNAL WATCHERS 372=head2 SIGNAL WATCHERS
233 373
374 $w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>);
375
234You can watch for signals using a signal watcher, C<signal> is the signal 376You can watch for signals using a signal watcher, C<signal> is the signal
235I<name> without any C<SIG> prefix, C<cb> is the Perl callback to 377I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
236be invoked whenever a signal occurs. 378callback to be invoked whenever a signal occurs.
237 379
238Although the callback might get passed parameters, their value and 380Although the callback might get passed parameters, their value and
239presence is undefined and you cannot rely on them. Portable AnyEvent 381presence is undefined and you cannot rely on them. Portable AnyEvent
240callbacks cannot use arguments passed to signal watcher callbacks. 382callbacks cannot use arguments passed to signal watcher callbacks.
241 383
242Multiple signal occurances can be clumped together into one callback 384Multiple signal occurrences can be clumped together into one callback
243invocation, and callback invocation will be synchronous. synchronous means 385invocation, and callback invocation will be synchronous. Synchronous means
244that it might take a while until the signal gets handled by the process, 386that it might take a while until the signal gets handled by the process,
245but it is guarenteed not to interrupt any other callbacks. 387but it is guaranteed not to interrupt any other callbacks.
246 388
247The main advantage of using these watchers is that you can share a signal 389The main advantage of using these watchers is that you can share a signal
248between multiple watchers. 390between multiple watchers, and AnyEvent will ensure that signals will not
391interrupt your program at bad times.
249 392
250This watcher might use C<%SIG>, so programs overwriting those signals 393This watcher might use C<%SIG> (depending on the event loop used),
251directly will likely not work correctly. 394so programs overwriting those signals directly will likely not work
395correctly.
252 396
253Example: exit on SIGINT 397Example: exit on SIGINT
254 398
255 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); 399 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
256 400
401=head3 Signal Races, Delays and Workarounds
402
403Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
404callbacks to signals in a generic way, which is a pity, as you cannot do
405race-free signal handling in perl. AnyEvent will try to do it's best, but
406in some cases, signals will be delayed. The maximum time a signal might
407be delayed is specified in C<$AnyEvent::MAX_SIGNAL_LATENCY> (default: 10
408seconds). This variable can be changed only before the first signal
409watcher is created, and should be left alone otherwise. Higher values
410will cause fewer spurious wake-ups, which is better for power and CPU
411saving. All these problems can be avoided by installing the optional
412L<Async::Interrupt> module. This will not work with inherently broken
413event loops such as L<Event> or L<Event::Lib> (and not with L<POE>
414currently, as POE does it's own workaround with one-second latency). With
415those, you just have to suffer the delays.
416
257=head2 CHILD PROCESS WATCHERS 417=head2 CHILD PROCESS WATCHERS
258 418
419 $w = AnyEvent->child (pid => <process id>, cb => <callback>);
420
259You can also watch on a child process exit and catch its exit status. 421You can also watch on a child process exit and catch its exit status.
260 422
261The child process is specified by the C<pid> argument (if set to C<0>, it 423The child process is specified by the C<pid> argument (one some backends,
262watches for any child process exit). The watcher will trigger as often 424using C<0> watches for any child process exit, on others this will
263as status change for the child are received. This works by installing a 425croak). The watcher will be triggered only when the child process has
264signal handler for C<SIGCHLD>. The callback will be called with the pid 426finished and an exit status is available, not on any trace events
265and exit status (as returned by waitpid), so unlike other watcher types, 427(stopped/continued).
266you I<can> rely on child watcher callback arguments. 428
429The callback will be called with the pid and exit status (as returned by
430waitpid), so unlike other watcher types, you I<can> rely on child watcher
431callback arguments.
432
433This watcher type works by installing a signal handler for C<SIGCHLD>,
434and since it cannot be shared, nothing else should use SIGCHLD or reap
435random child processes (waiting for specific child processes, e.g. inside
436C<system>, is just fine).
267 437
268There is a slight catch to child watchers, however: you usually start them 438There is a slight catch to child watchers, however: you usually start them
269I<after> the child process was created, and this means the process could 439I<after> the child process was created, and this means the process could
270have exited already (and no SIGCHLD will be sent anymore). 440have exited already (and no SIGCHLD will be sent anymore).
271 441
272Not all event models handle this correctly (POE doesn't), but even for 442Not all event models handle this correctly (neither POE nor IO::Async do,
443see their AnyEvent::Impl manpages for details), but even for event models
273event models that I<do> handle this correctly, they usually need to be 444that I<do> handle this correctly, they usually need to be loaded before
274loaded before the process exits (i.e. before you fork in the first place). 445the process exits (i.e. before you fork in the first place). AnyEvent's
446pure perl event loop handles all cases correctly regardless of when you
447start the watcher.
275 448
276This means you cannot create a child watcher as the very first thing in an 449This means you cannot create a child watcher as the very first
277AnyEvent program, you I<have> to create at least one watcher before you 450thing in an AnyEvent program, you I<have> to create at least one
278C<fork> the child (alternatively, you can call C<AnyEvent::detect>). 451watcher before you C<fork> the child (alternatively, you can call
452C<AnyEvent::detect>).
453
454As most event loops do not support waiting for child events, they will be
455emulated by AnyEvent in most cases, in which the latency and race problems
456mentioned in the description of signal watchers apply.
279 457
280Example: fork a process and wait for it 458Example: fork a process and wait for it
281 459
282 my $done = AnyEvent->condvar; 460 my $done = AnyEvent->condvar;
283 461
284 AnyEvent::detect; # force event module to be initialised
285
286 my $pid = fork or exit 5; 462 my $pid = fork or exit 5;
287 463
288 my $w = AnyEvent->child ( 464 my $w = AnyEvent->child (
289 pid => $pid, 465 pid => $pid,
290 cb => sub { 466 cb => sub {
291 my ($pid, $status) = @_; 467 my ($pid, $status) = @_;
292 warn "pid $pid exited with status $status"; 468 warn "pid $pid exited with status $status";
293 $done->broadcast; 469 $done->send;
294 }, 470 },
295 ); 471 );
296 472
297 # do something else, then wait for process exit 473 # do something else, then wait for process exit
298 $done->wait; 474 $done->recv;
475
476=head2 IDLE WATCHERS
477
478 $w = AnyEvent->idle (cb => <callback>);
479
480Sometimes there is a need to do something, but it is not so important
481to do it instantly, but only when there is nothing better to do. This
482"nothing better to do" is usually defined to be "no other events need
483attention by the event loop".
484
485Idle watchers ideally get invoked when the event loop has nothing
486better to do, just before it would block the process to wait for new
487events. Instead of blocking, the idle watcher is invoked.
488
489Most event loops unfortunately do not really support idle watchers (only
490EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
491will simply call the callback "from time to time".
492
493Example: read lines from STDIN, but only process them when the
494program is otherwise idle:
495
496 my @lines; # read data
497 my $idle_w;
498 my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
499 push @lines, scalar <STDIN>;
500
501 # start an idle watcher, if not already done
502 $idle_w ||= AnyEvent->idle (cb => sub {
503 # handle only one line, when there are lines left
504 if (my $line = shift @lines) {
505 print "handled when idle: $line";
506 } else {
507 # otherwise disable the idle watcher again
508 undef $idle_w;
509 }
510 });
511 });
299 512
300=head2 CONDITION VARIABLES 513=head2 CONDITION VARIABLES
301 514
515 $cv = AnyEvent->condvar;
516
517 $cv->send (<list>);
518 my @res = $cv->recv;
519
520If you are familiar with some event loops you will know that all of them
521require you to run some blocking "loop", "run" or similar function that
522will actively watch for new events and call your callbacks.
523
524AnyEvent is slightly different: it expects somebody else to run the event
525loop and will only block when necessary (usually when told by the user).
526
527The instrument to do that is called a "condition variable", so called
528because they represent a condition that must become true.
529
530Now is probably a good time to look at the examples further below.
531
302Condition variables can be created by calling the C<< AnyEvent->condvar >> 532Condition variables can be created by calling the C<< AnyEvent->condvar
303method without any arguments. 533>> method, usually without arguments. The only argument pair allowed is
534C<cb>, which specifies a callback to be called when the condition variable
535becomes true, with the condition variable as the first argument (but not
536the results).
304 537
305A condition variable waits for a condition - precisely that the C<< 538After creation, the condition variable is "false" until it becomes "true"
306->broadcast >> method has been called. 539by calling the C<send> method (or calling the condition variable as if it
540were a callback, read about the caveats in the description for the C<<
541->send >> method).
307 542
308They are very useful to signal that a condition has been fulfilled, for 543Condition variables are similar to callbacks, except that you can
544optionally wait for them. They can also be called merge points - points
545in time where multiple outstanding events have been processed. And yet
546another way to call them is transactions - each condition variable can be
547used to represent a transaction, which finishes at some point and delivers
548a result. And yet some people know them as "futures" - a promise to
549compute/deliver something that you can wait for.
550
551Condition variables are very useful to signal that something has finished,
309example, if you write a module that does asynchronous http requests, 552for example, if you write a module that does asynchronous http requests,
310then a condition variable would be the ideal candidate to signal the 553then a condition variable would be the ideal candidate to signal the
311availability of results. 554availability of results. The user can either act when the callback is
555called or can synchronously C<< ->recv >> for the results.
312 556
313You can also use condition variables to block your main program until 557You can also use them to simulate traditional event loops - for example,
314an event occurs - for example, you could C<< ->wait >> in your main 558you can block your main program until an event occurs - for example, you
315program until the user clicks the Quit button in your app, which would C<< 559could C<< ->recv >> in your main program until the user clicks the Quit
316->broadcast >> the "quit" event. 560button of your app, which would C<< ->send >> the "quit" event.
317 561
318Note that condition variables recurse into the event loop - if you have 562Note that condition variables recurse into the event loop - if you have
319two pirces of code that call C<< ->wait >> in a round-robbin fashion, you 563two pieces of code that call C<< ->recv >> in a round-robin fashion, you
320lose. Therefore, condition variables are good to export to your caller, but 564lose. Therefore, condition variables are good to export to your caller, but
321you should avoid making a blocking wait yourself, at least in callbacks, 565you should avoid making a blocking wait yourself, at least in callbacks,
322as this asks for trouble. 566as this asks for trouble.
323 567
324This object has two methods: 568Condition variables are represented by hash refs in perl, and the keys
569used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
570easy (it is often useful to build your own transaction class on top of
571AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
572it's C<new> method in your own C<new> method.
325 573
326=over 4 574There are two "sides" to a condition variable - the "producer side" which
575eventually calls C<< -> send >>, and the "consumer side", which waits
576for the send to occur.
327 577
328=item $cv->wait 578Example: wait for a timer.
329
330Wait (blocking if necessary) until the C<< ->broadcast >> method has been
331called on c<$cv>, while servicing other watchers normally.
332
333You can only wait once on a condition - additional calls will return
334immediately.
335
336Not all event models support a blocking wait - some die in that case
337(programs might want to do that to stay interactive), so I<if you are
338using this from a module, never require a blocking wait>, but let the
339caller decide whether the call will block or not (for example, by coupling
340condition variables with some kind of request results and supporting
341callbacks so the caller knows that getting the result will not block,
342while still suppporting blocking waits if the caller so desires).
343
344Another reason I<never> to C<< ->wait >> in a module is that you cannot
345sensibly have two C<< ->wait >>'s in parallel, as that would require
346multiple interpreters or coroutines/threads, none of which C<AnyEvent>
347can supply (the coroutine-aware backends L<AnyEvent::Impl::CoroEV> and
348L<AnyEvent::Impl::CoroEvent> explicitly support concurrent C<< ->wait >>'s
349from different coroutines, however).
350
351=item $cv->broadcast
352
353Flag the condition as ready - a running C<< ->wait >> and all further
354calls to C<wait> will (eventually) return after this method has been
355called. If nobody is waiting the broadcast will be remembered..
356
357=back
358
359Example:
360 579
361 # wait till the result is ready 580 # wait till the result is ready
362 my $result_ready = AnyEvent->condvar; 581 my $result_ready = AnyEvent->condvar;
363 582
364 # do something such as adding a timer 583 # do something such as adding a timer
365 # or socket watcher the calls $result_ready->broadcast 584 # or socket watcher the calls $result_ready->send
366 # when the "result" is ready. 585 # when the "result" is ready.
367 # in this case, we simply use a timer: 586 # in this case, we simply use a timer:
368 my $w = AnyEvent->timer ( 587 my $w = AnyEvent->timer (
369 after => 1, 588 after => 1,
370 cb => sub { $result_ready->broadcast }, 589 cb => sub { $result_ready->send },
371 ); 590 );
372 591
373 # this "blocks" (while handling events) till the watcher 592 # this "blocks" (while handling events) till the callback
374 # calls broadcast 593 # calls -<send
375 $result_ready->wait; 594 $result_ready->recv;
595
596Example: wait for a timer, but take advantage of the fact that condition
597variables are also callable directly.
598
599 my $done = AnyEvent->condvar;
600 my $delay = AnyEvent->timer (after => 5, cb => $done);
601 $done->recv;
602
603Example: Imagine an API that returns a condvar and doesn't support
604callbacks. This is how you make a synchronous call, for example from
605the main program:
606
607 use AnyEvent::CouchDB;
608
609 ...
610
611 my @info = $couchdb->info->recv;
612
613And this is how you would just set a callback to be called whenever the
614results are available:
615
616 $couchdb->info->cb (sub {
617 my @info = $_[0]->recv;
618 });
619
620=head3 METHODS FOR PRODUCERS
621
622These methods should only be used by the producing side, i.e. the
623code/module that eventually sends the signal. Note that it is also
624the producer side which creates the condvar in most cases, but it isn't
625uncommon for the consumer to create it as well.
626
627=over 4
628
629=item $cv->send (...)
630
631Flag the condition as ready - a running C<< ->recv >> and all further
632calls to C<recv> will (eventually) return after this method has been
633called. If nobody is waiting the send will be remembered.
634
635If a callback has been set on the condition variable, it is called
636immediately from within send.
637
638Any arguments passed to the C<send> call will be returned by all
639future C<< ->recv >> calls.
640
641Condition variables are overloaded so one can call them directly (as if
642they were a code reference). Calling them directly is the same as calling
643C<send>.
644
645=item $cv->croak ($error)
646
647Similar to send, but causes all call's to C<< ->recv >> to invoke
648C<Carp::croak> with the given error message/object/scalar.
649
650This can be used to signal any errors to the condition variable
651user/consumer. Doing it this way instead of calling C<croak> directly
652delays the error detetcion, but has the overwhelmign advantage that it
653diagnoses the error at the place where the result is expected, and not
654deep in some event clalback without connection to the actual code causing
655the problem.
656
657=item $cv->begin ([group callback])
658
659=item $cv->end
660
661These two methods can be used to combine many transactions/events into
662one. For example, a function that pings many hosts in parallel might want
663to use a condition variable for the whole process.
664
665Every call to C<< ->begin >> will increment a counter, and every call to
666C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
667>>, the (last) callback passed to C<begin> will be executed. That callback
668is I<supposed> to call C<< ->send >>, but that is not required. If no
669callback was set, C<send> will be called without any arguments.
670
671You can think of C<< $cv->send >> giving you an OR condition (one call
672sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
673condition (all C<begin> calls must be C<end>'ed before the condvar sends).
674
675Let's start with a simple example: you have two I/O watchers (for example,
676STDOUT and STDERR for a program), and you want to wait for both streams to
677close before activating a condvar:
678
679 my $cv = AnyEvent->condvar;
680
681 $cv->begin; # first watcher
682 my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
683 defined sysread $fh1, my $buf, 4096
684 or $cv->end;
685 });
686
687 $cv->begin; # second watcher
688 my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
689 defined sysread $fh2, my $buf, 4096
690 or $cv->end;
691 });
692
693 $cv->recv;
694
695This works because for every event source (EOF on file handle), there is
696one call to C<begin>, so the condvar waits for all calls to C<end> before
697sending.
698
699The ping example mentioned above is slightly more complicated, as the
700there are results to be passwd back, and the number of tasks that are
701begung can potentially be zero:
702
703 my $cv = AnyEvent->condvar;
704
705 my %result;
706 $cv->begin (sub { $cv->send (\%result) });
707
708 for my $host (@list_of_hosts) {
709 $cv->begin;
710 ping_host_then_call_callback $host, sub {
711 $result{$host} = ...;
712 $cv->end;
713 };
714 }
715
716 $cv->end;
717
718This code fragment supposedly pings a number of hosts and calls
719C<send> after results for all then have have been gathered - in any
720order. To achieve this, the code issues a call to C<begin> when it starts
721each ping request and calls C<end> when it has received some result for
722it. Since C<begin> and C<end> only maintain a counter, the order in which
723results arrive is not relevant.
724
725There is an additional bracketing call to C<begin> and C<end> outside the
726loop, which serves two important purposes: first, it sets the callback
727to be called once the counter reaches C<0>, and second, it ensures that
728C<send> is called even when C<no> hosts are being pinged (the loop
729doesn't execute once).
730
731This is the general pattern when you "fan out" into multiple (but
732potentially none) subrequests: use an outer C<begin>/C<end> pair to set
733the callback and ensure C<end> is called at least once, and then, for each
734subrequest you start, call C<begin> and for each subrequest you finish,
735call C<end>.
736
737=back
738
739=head3 METHODS FOR CONSUMERS
740
741These methods should only be used by the consuming side, i.e. the
742code awaits the condition.
743
744=over 4
745
746=item $cv->recv
747
748Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
749>> methods have been called on c<$cv>, while servicing other watchers
750normally.
751
752You can only wait once on a condition - additional calls are valid but
753will return immediately.
754
755If an error condition has been set by calling C<< ->croak >>, then this
756function will call C<croak>.
757
758In list context, all parameters passed to C<send> will be returned,
759in scalar context only the first one will be returned.
760
761Note that doing a blocking wait in a callback is not supported by any
762event loop, that is, recursive invocation of a blocking C<< ->recv
763>> is not allowed, and the C<recv> call will C<croak> if such a
764condition is detected. This condition can be slightly loosened by using
765L<Coro::AnyEvent>, which allows you to do a blocking C<< ->recv >> from
766any thread that doesn't run the event loop itself.
767
768Not all event models support a blocking wait - some die in that case
769(programs might want to do that to stay interactive), so I<if you are
770using this from a module, never require a blocking wait>. Instead, let the
771caller decide whether the call will block or not (for example, by coupling
772condition variables with some kind of request results and supporting
773callbacks so the caller knows that getting the result will not block,
774while still supporting blocking waits if the caller so desires).
775
776You can ensure that C<< -recv >> never blocks by setting a callback and
777only calling C<< ->recv >> from within that callback (or at a later
778time). This will work even when the event loop does not support blocking
779waits otherwise.
780
781=item $bool = $cv->ready
782
783Returns true when the condition is "true", i.e. whether C<send> or
784C<croak> have been called.
785
786=item $cb = $cv->cb ($cb->($cv))
787
788This is a mutator function that returns the callback set and optionally
789replaces it before doing so.
790
791The callback will be called when the condition becomes "true", i.e. when
792C<send> or C<croak> are called, with the only argument being the condition
793variable itself. Calling C<recv> inside the callback or at any later time
794is guaranteed not to block.
795
796=back
797
798=head1 SUPPORTED EVENT LOOPS/BACKENDS
799
800The available backend classes are (every class has its own manpage):
801
802=over 4
803
804=item Backends that are autoprobed when no other event loop can be found.
805
806EV is the preferred backend when no other event loop seems to be in
807use. If EV is not installed, then AnyEvent will try Event, and, failing
808that, will fall back to its own pure-perl implementation, which is
809available everywhere as it comes with AnyEvent itself.
810
811 AnyEvent::Impl::EV based on EV (interface to libev, best choice).
812 AnyEvent::Impl::Event based on Event, very stable, few glitches.
813 AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
814
815=item Backends that are transparently being picked up when they are used.
816
817These will be used when they are currently loaded when the first watcher
818is created, in which case it is assumed that the application is using
819them. This means that AnyEvent will automatically pick the right backend
820when the main program loads an event module before anything starts to
821create watchers. Nothing special needs to be done by the main program.
822
823 AnyEvent::Impl::Glib based on Glib, slow but very stable.
824 AnyEvent::Impl::Tk based on Tk, very broken.
825 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
826 AnyEvent::Impl::POE based on POE, very slow, some limitations.
827 AnyEvent::Impl::Irssi used when running within irssi.
828
829=item Backends with special needs.
830
831Qt requires the Qt::Application to be instantiated first, but will
832otherwise be picked up automatically. As long as the main program
833instantiates the application before any AnyEvent watchers are created,
834everything should just work.
835
836 AnyEvent::Impl::Qt based on Qt.
837
838Support for IO::Async can only be partial, as it is too broken and
839architecturally limited to even support the AnyEvent API. It also
840is the only event loop that needs the loop to be set explicitly, so
841it can only be used by a main program knowing about AnyEvent. See
842L<AnyEvent::Impl::Async> for the gory details.
843
844 AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed.
845
846=item Event loops that are indirectly supported via other backends.
847
848Some event loops can be supported via other modules:
849
850There is no direct support for WxWidgets (L<Wx>) or L<Prima>.
851
852B<WxWidgets> has no support for watching file handles. However, you can
853use WxWidgets through the POE adaptor, as POE has a Wx backend that simply
854polls 20 times per second, which was considered to be too horrible to even
855consider for AnyEvent.
856
857B<Prima> is not supported as nobody seems to be using it, but it has a POE
858backend, so it can be supported through POE.
859
860AnyEvent knows about both L<Prima> and L<Wx>, however, and will try to
861load L<POE> when detecting them, in the hope that POE will pick them up,
862in which case everything will be automatic.
863
864=back
376 865
377=head1 GLOBAL VARIABLES AND FUNCTIONS 866=head1 GLOBAL VARIABLES AND FUNCTIONS
378 867
868These are not normally required to use AnyEvent, but can be useful to
869write AnyEvent extension modules.
870
379=over 4 871=over 4
380 872
381=item $AnyEvent::MODEL 873=item $AnyEvent::MODEL
382 874
383Contains C<undef> until the first watcher is being created. Then it 875Contains C<undef> until the first watcher is being created, before the
876backend has been autodetected.
877
384contains the event model that is being used, which is the name of the 878Afterwards it contains the event model that is being used, which is the
385Perl class implementing the model. This class is usually one of the 879name of the Perl class implementing the model. This class is usually one
386C<AnyEvent::Impl:xxx> modules, but can be any other class in the case 880of the C<AnyEvent::Impl:xxx> modules, but can be any other class in the
387AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>). 881case AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode> it
388 882will be C<urxvt::anyevent>).
389The known classes so far are:
390
391 AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
392 AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
393 AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
394 AnyEvent::Impl::Event based on Event, second best choice.
395 AnyEvent::Impl::Glib based on Glib, third-best choice.
396 AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable.
397 AnyEvent::Impl::Tk based on Tk, very bad choice.
398 AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
399 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
400 AnyEvent::Impl::POE based on POE, not generic enough for full support.
401
402There is no support for WxWidgets, as WxWidgets has no support for
403watching file handles. However, you can use WxWidgets through the
404POE Adaptor, as POE has a Wx backend that simply polls 20 times per
405second, which was considered to be too horrible to even consider for
406AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
407it's adaptor.
408
409AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
410autodetecting them.
411 883
412=item AnyEvent::detect 884=item AnyEvent::detect
413 885
414Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model 886Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
415if necessary. You should only call this function right before you would 887if necessary. You should only call this function right before you would
416have created an AnyEvent watcher anyway, that is, as late as possible at 888have created an AnyEvent watcher anyway, that is, as late as possible at
417runtime. 889runtime, and not e.g. while initialising of your module.
890
891If you need to do some initialisation before AnyEvent watchers are
892created, use C<post_detect>.
893
894=item $guard = AnyEvent::post_detect { BLOCK }
895
896Arranges for the code block to be executed as soon as the event model is
897autodetected (or immediately if this has already happened).
898
899The block will be executed I<after> the actual backend has been detected
900(C<$AnyEvent::MODEL> is set), but I<before> any watchers have been
901created, so it is possible to e.g. patch C<@AnyEvent::ISA> or do
902other initialisations - see the sources of L<AnyEvent::Strict> or
903L<AnyEvent::AIO> to see how this is used.
904
905The most common usage is to create some global watchers, without forcing
906event module detection too early, for example, L<AnyEvent::AIO> creates
907and installs the global L<IO::AIO> watcher in a C<post_detect> block to
908avoid autodetecting the event module at load time.
909
910If called in scalar or list context, then it creates and returns an object
911that automatically removes the callback again when it is destroyed (or
912C<undef> when the hook was immediately executed). See L<AnyEvent::AIO> for
913a case where this is useful.
914
915Example: Create a watcher for the IO::AIO module and store it in
916C<$WATCHER>. Only do so after the event loop is initialised, though.
917
918 our WATCHER;
919
920 my $guard = AnyEvent::post_detect {
921 $WATCHER = AnyEvent->io (fh => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb);
922 };
923
924 # the ||= is important in case post_detect immediately runs the block,
925 # as to not clobber the newly-created watcher. assigning both watcher and
926 # post_detect guard to the same variable has the advantage of users being
927 # able to just C<undef $WATCHER> if the watcher causes them grief.
928
929 $WATCHER ||= $guard;
930
931=item @AnyEvent::post_detect
932
933If there are any code references in this array (you can C<push> to it
934before or after loading AnyEvent), then they will called directly after
935the event loop has been chosen.
936
937You should check C<$AnyEvent::MODEL> before adding to this array, though:
938if it is defined then the event loop has already been detected, and the
939array will be ignored.
940
941Best use C<AnyEvent::post_detect { BLOCK }> when your application allows
942it,as it takes care of these details.
943
944This variable is mainly useful for modules that can do something useful
945when AnyEvent is used and thus want to know when it is initialised, but do
946not need to even load it by default. This array provides the means to hook
947into AnyEvent passively, without loading it.
418 948
419=back 949=back
420 950
421=head1 WHAT TO DO IN A MODULE 951=head1 WHAT TO DO IN A MODULE
422 952
426Be careful when you create watchers in the module body - AnyEvent will 956Be careful when you create watchers in the module body - AnyEvent will
427decide which event module to use as soon as the first method is called, so 957decide which event module to use as soon as the first method is called, so
428by calling AnyEvent in your module body you force the user of your module 958by calling AnyEvent in your module body you force the user of your module
429to load the event module first. 959to load the event module first.
430 960
431Never call C<< ->wait >> on a condition variable unless you I<know> that 961Never call C<< ->recv >> on a condition variable unless you I<know> that
432the C<< ->broadcast >> method has been called on it already. This is 962the C<< ->send >> method has been called on it already. This is
433because it will stall the whole program, and the whole point of using 963because it will stall the whole program, and the whole point of using
434events is to stay interactive. 964events is to stay interactive.
435 965
436It is fine, however, to call C<< ->wait >> when the user of your module 966It is fine, however, to call C<< ->recv >> when the user of your module
437requests it (i.e. if you create a http request object ad have a method 967requests it (i.e. if you create a http request object ad have a method
438called C<results> that returns the results, it should call C<< ->wait >> 968called C<results> that returns the results, it should call C<< ->recv >>
439freely, as the user of your module knows what she is doing. always). 969freely, as the user of your module knows what she is doing. always).
440 970
441=head1 WHAT TO DO IN THE MAIN PROGRAM 971=head1 WHAT TO DO IN THE MAIN PROGRAM
442 972
443There will always be a single main program - the only place that should 973There will always be a single main program - the only place that should
445 975
446If it doesn't care, it can just "use AnyEvent" and use it itself, or not 976If it doesn't care, it can just "use AnyEvent" and use it itself, or not
447do anything special (it does not need to be event-based) and let AnyEvent 977do anything special (it does not need to be event-based) and let AnyEvent
448decide which implementation to chose if some module relies on it. 978decide which implementation to chose if some module relies on it.
449 979
450If the main program relies on a specific event model. For example, in 980If the main program relies on a specific event model - for example, in
451Gtk2 programs you have to rely on the Glib module. You should load the 981Gtk2 programs you have to rely on the Glib module - you should load the
452event module before loading AnyEvent or any module that uses it: generally 982event module before loading AnyEvent or any module that uses it: generally
453speaking, you should load it as early as possible. The reason is that 983speaking, you should load it as early as possible. The reason is that
454modules might create watchers when they are loaded, and AnyEvent will 984modules might create watchers when they are loaded, and AnyEvent will
455decide on the event model to use as soon as it creates watchers, and it 985decide on the event model to use as soon as it creates watchers, and it
456might chose the wrong one unless you load the correct one yourself. 986might chose the wrong one unless you load the correct one yourself.
457 987
458You can chose to use a rather inefficient pure-perl implementation by 988You can chose to use a pure-perl implementation by loading the
459loading the C<AnyEvent::Impl::Perl> module, which gives you similar 989C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
460behaviour everywhere, but letting AnyEvent chose is generally better. 990everywhere, but letting AnyEvent chose the model is generally better.
991
992=head2 MAINLOOP EMULATION
993
994Sometimes (often for short test scripts, or even standalone programs who
995only want to use AnyEvent), you do not want to run a specific event loop.
996
997In that case, you can use a condition variable like this:
998
999 AnyEvent->condvar->recv;
1000
1001This has the effect of entering the event loop and looping forever.
1002
1003Note that usually your program has some exit condition, in which case
1004it is better to use the "traditional" approach of storing a condition
1005variable somewhere, waiting for it, and sending it when the program should
1006exit cleanly.
1007
461 1008
462=head1 OTHER MODULES 1009=head1 OTHER MODULES
463 1010
464The following is a non-exhaustive list of additional modules that use 1011The following is a non-exhaustive list of additional modules that use
465AnyEvent and can therefore be mixed easily with other AnyEvent modules 1012AnyEvent as a client and can therefore be mixed easily with other AnyEvent
466in the same program. Some of the modules come with AnyEvent, some are 1013modules and other event loops in the same program. Some of the modules
467available via CPAN. 1014come with AnyEvent, most are available via CPAN.
468 1015
469=over 4 1016=over 4
470 1017
471=item L<AnyEvent::Util> 1018=item L<AnyEvent::Util>
472 1019
473Contains various utility functions that replace often-used but blocking 1020Contains various utility functions that replace often-used but blocking
474functions such as C<inet_aton> by event-/callback-based versions. 1021functions such as C<inet_aton> by event-/callback-based versions.
475 1022
1023=item L<AnyEvent::Socket>
1024
1025Provides various utility functions for (internet protocol) sockets,
1026addresses and name resolution. Also functions to create non-blocking tcp
1027connections or tcp servers, with IPv6 and SRV record support and more.
1028
476=item L<AnyEvent::Handle> 1029=item L<AnyEvent::Handle>
477 1030
478Provide read and write buffers and manages watchers for reads and writes. 1031Provide read and write buffers, manages watchers for reads and writes,
1032supports raw and formatted I/O, I/O queued and fully transparent and
1033non-blocking SSL/TLS (via L<AnyEvent::TLS>.
479 1034
480=item L<AnyEvent::Socket> 1035=item L<AnyEvent::DNS>
481 1036
482Provides a means to do non-blocking connects, accepts etc. 1037Provides rich asynchronous DNS resolver capabilities.
1038
1039=item L<AnyEvent::HTTP>
1040
1041A simple-to-use HTTP library that is capable of making a lot of concurrent
1042HTTP requests.
483 1043
484=item L<AnyEvent::HTTPD> 1044=item L<AnyEvent::HTTPD>
485 1045
486Provides a simple web application server framework. 1046Provides a simple web application server framework.
487 1047
488=item L<AnyEvent::DNS>
489
490Provides asynchronous DNS resolver capabilities, beyond what
491L<AnyEvent::Util> offers.
492
493=item L<AnyEvent::FastPing> 1048=item L<AnyEvent::FastPing>
494 1049
495The fastest ping in the west. 1050The fastest ping in the west.
496 1051
1052=item L<AnyEvent::DBI>
1053
1054Executes L<DBI> requests asynchronously in a proxy process.
1055
1056=item L<AnyEvent::AIO>
1057
1058Truly asynchronous I/O, should be in the toolbox of every event
1059programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
1060together.
1061
1062=item L<AnyEvent::BDB>
1063
1064Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
1065L<BDB> and AnyEvent together.
1066
1067=item L<AnyEvent::GPSD>
1068
1069A non-blocking interface to gpsd, a daemon delivering GPS information.
1070
497=item L<Net::IRC3> 1071=item L<AnyEvent::IRC>
498 1072
499AnyEvent based IRC client module family. 1073AnyEvent based IRC client module family (replacing the older Net::IRC3).
500 1074
501=item L<Net::XMPP2> 1075=item L<AnyEvent::XMPP>
502 1076
503AnyEvent based XMPP (Jabber protocol) module family. 1077AnyEvent based XMPP (Jabber protocol) module family (replacing the older
1078Net::XMPP2>.
1079
1080=item L<AnyEvent::IGS>
1081
1082A non-blocking interface to the Internet Go Server protocol (used by
1083L<App::IGS>).
504 1084
505=item L<Net::FCP> 1085=item L<Net::FCP>
506 1086
507AnyEvent-based implementation of the Freenet Client Protocol, birthplace 1087AnyEvent-based implementation of the Freenet Client Protocol, birthplace
508of AnyEvent. 1088of AnyEvent.
511 1091
512High level API for event-based execution flow control. 1092High level API for event-based execution flow control.
513 1093
514=item L<Coro> 1094=item L<Coro>
515 1095
516Has special support for AnyEvent. 1096Has special support for AnyEvent via L<Coro::AnyEvent>.
517
518=item L<IO::Lambda>
519
520The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
521
522=item L<IO::AIO>
523
524Truly asynchronous I/O, should be in the toolbox of every event
525programmer. Can be trivially made to use AnyEvent.
526
527=item L<BDB>
528
529Truly asynchronous Berkeley DB access. Can be trivially made to use
530AnyEvent.
531 1097
532=back 1098=back
533 1099
534=cut 1100=cut
535 1101
536package AnyEvent; 1102package AnyEvent;
537 1103
1104# basically a tuned-down version of common::sense
1105sub common_sense {
538no warnings; 1106 # no warnings
539use strict; 1107 ${^WARNING_BITS} ^= ${^WARNING_BITS};
1108 # use strict vars subs
1109 $^H |= 0x00000600;
1110}
540 1111
1112BEGIN { AnyEvent::common_sense }
1113
541use Carp; 1114use Carp ();
542 1115
543our $VERSION = '3.3'; 1116our $VERSION = 4.881;
544our $MODEL; 1117our $MODEL;
545 1118
546our $AUTOLOAD; 1119our $AUTOLOAD;
547our @ISA; 1120our @ISA;
548 1121
549our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
550
551our @REGISTRY; 1122our @REGISTRY;
552 1123
1124our $WIN32;
1125
1126our $VERBOSE;
1127
1128BEGIN {
1129 eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }";
1130 eval "sub TAINT(){ " . (${^TAINT}*1) . " }";
1131
1132 delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
1133 if ${^TAINT};
1134
1135 $VERBOSE = $ENV{PERL_ANYEVENT_VERBOSE}*1;
1136
1137}
1138
1139our $MAX_SIGNAL_LATENCY = 10;
1140
1141our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
1142
1143{
1144 my $idx;
1145 $PROTOCOL{$_} = ++$idx
1146 for reverse split /\s*,\s*/,
1147 $ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
1148}
1149
553my @models = ( 1150my @models = (
554 [Coro::EV:: => AnyEvent::Impl::CoroEV::],
555 [Coro::Event:: => AnyEvent::Impl::CoroEvent::],
556 [EV:: => AnyEvent::Impl::EV::], 1151 [EV:: => AnyEvent::Impl::EV:: , 1],
557 [Event:: => AnyEvent::Impl::Event::], 1152 [Event:: => AnyEvent::Impl::Event::, 1],
558 [Glib:: => AnyEvent::Impl::Glib::], 1153 [AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl:: , 1],
1154 # everything below here will not (normally) be autoprobed
1155 # as the pureperl backend should work everywhere
1156 # and is usually faster
1157 [Glib:: => AnyEvent::Impl::Glib:: , 1], # becomes extremely slow with many watchers
1158 [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
1159 [Irssi:: => AnyEvent::Impl::Irssi::], # Irssi has a bogus "Event" package
559 [Tk:: => AnyEvent::Impl::Tk::], 1160 [Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
1161 [Qt:: => AnyEvent::Impl::Qt::], # requires special main program
1162 [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
560 [Wx:: => AnyEvent::Impl::POE::], 1163 [Wx:: => AnyEvent::Impl::POE::],
561 [Prima:: => AnyEvent::Impl::POE::], 1164 [Prima:: => AnyEvent::Impl::POE::],
562 [AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::], 1165 # IO::Async is just too broken - we would need workarounds for its
563 # everything below here will not be autoprobed as the pureperl backend should work everywhere 1166 # byzantine signal and broken child handling, among others.
564 [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy 1167 # IO::Async is rather hard to detect, as it doesn't have any
1168 # obvious default class.
565 [Qt:: => AnyEvent::Impl::Qt::], # requires special main program 1169# [0, IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program
566 [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza 1170# [0, IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program
1171# [0, IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program
567); 1172);
568 1173
569our %method = map +($_ => 1), qw(io timer signal child condvar broadcast wait one_event DESTROY); 1174our %method = map +($_ => 1),
1175 qw(io timer time now now_update signal child idle condvar one_event DESTROY);
1176
1177our @post_detect;
1178
1179sub post_detect(&) {
1180 my ($cb) = @_;
1181
1182 if ($MODEL) {
1183 $cb->();
1184
1185 undef
1186 } else {
1187 push @post_detect, $cb;
1188
1189 defined wantarray
1190 ? bless \$cb, "AnyEvent::Util::postdetect"
1191 : ()
1192 }
1193}
1194
1195sub AnyEvent::Util::postdetect::DESTROY {
1196 @post_detect = grep $_ != ${$_[0]}, @post_detect;
1197}
570 1198
571sub detect() { 1199sub detect() {
572 unless ($MODEL) { 1200 unless ($MODEL) {
573 no strict 'refs'; 1201 local $SIG{__DIE__};
574 1202
575 if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) { 1203 if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
576 my $model = "AnyEvent::Impl::$1"; 1204 my $model = "AnyEvent::Impl::$1";
577 if (eval "require $model") { 1205 if (eval "require $model") {
578 $MODEL = $model; 1206 $MODEL = $model;
579 warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1; 1207 warn "AnyEvent: loaded model '$model' (forced by \$ENV{PERL_ANYEVENT_MODEL}), using it.\n" if $VERBOSE >= 2;
580 } else { 1208 } else {
581 warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose; 1209 warn "AnyEvent: unable to load model '$model' (from \$ENV{PERL_ANYEVENT_MODEL}):\n$@" if $VERBOSE;
582 } 1210 }
583 } 1211 }
584 1212
585 # check for already loaded models 1213 # check for already loaded models
586 unless ($MODEL) { 1214 unless ($MODEL) {
587 for (@REGISTRY, @models) { 1215 for (@REGISTRY, @models) {
588 my ($package, $model) = @$_; 1216 my ($package, $model) = @$_;
589 if (${"$package\::VERSION"} > 0) { 1217 if (${"$package\::VERSION"} > 0) {
590 if (eval "require $model") { 1218 if (eval "require $model") {
591 $MODEL = $model; 1219 $MODEL = $model;
592 warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1; 1220 warn "AnyEvent: autodetected model '$model', using it.\n" if $VERBOSE >= 2;
593 last; 1221 last;
594 } 1222 }
595 } 1223 }
596 } 1224 }
597 1225
598 unless ($MODEL) { 1226 unless ($MODEL) {
599 # try to load a model 1227 # try to autoload a model
600
601 for (@REGISTRY, @models) { 1228 for (@REGISTRY, @models) {
602 my ($package, $model) = @$_; 1229 my ($package, $model, $autoload) = @$_;
1230 if (
1231 $autoload
603 if (eval "require $package" 1232 and eval "require $package"
604 and ${"$package\::VERSION"} > 0 1233 and ${"$package\::VERSION"} > 0
605 and eval "require $model") { 1234 and eval "require $model"
1235 ) {
606 $MODEL = $model; 1236 $MODEL = $model;
607 warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1; 1237 warn "AnyEvent: autoloaded model '$model', using it.\n" if $VERBOSE >= 2;
608 last; 1238 last;
609 } 1239 }
610 } 1240 }
611 1241
612 $MODEL 1242 $MODEL
613 or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV (or Coro+EV), Event (or Coro+Event) or Glib."; 1243 or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
614 } 1244 }
615 } 1245 }
616 1246
1247 push @{"$MODEL\::ISA"}, "AnyEvent::Base";
1248
617 unshift @ISA, $MODEL; 1249 unshift @ISA, $MODEL;
618 push @{"$MODEL\::ISA"}, "AnyEvent::Base"; 1250
1251 require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
1252
1253 (shift @post_detect)->() while @post_detect;
619 } 1254 }
620 1255
621 $MODEL 1256 $MODEL
622} 1257}
623 1258
624sub AUTOLOAD { 1259sub AUTOLOAD {
625 (my $func = $AUTOLOAD) =~ s/.*://; 1260 (my $func = $AUTOLOAD) =~ s/.*://;
626 1261
627 $method{$func} 1262 $method{$func}
628 or croak "$func: not a valid method for AnyEvent objects"; 1263 or Carp::croak "$func: not a valid method for AnyEvent objects";
629 1264
630 detect unless $MODEL; 1265 detect unless $MODEL;
631 1266
632 my $class = shift; 1267 my $class = shift;
633 $class->$func (@_); 1268 $class->$func (@_);
634} 1269}
635 1270
1271# utility function to dup a filehandle. this is used by many backends
1272# to support binding more than one watcher per filehandle (they usually
1273# allow only one watcher per fd, so we dup it to get a different one).
1274sub _dupfh($$;$$) {
1275 my ($poll, $fh, $r, $w) = @_;
1276
1277 # cygwin requires the fh mode to be matching, unix doesn't
1278 my ($rw, $mode) = $poll eq "r" ? ($r, "<&") : ($w, ">&");
1279
1280 open my $fh2, $mode, $fh
1281 or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,";
1282
1283 # we assume CLOEXEC is already set by perl in all important cases
1284
1285 ($fh2, $rw)
1286}
1287
636package AnyEvent::Base; 1288package AnyEvent::Base;
637 1289
1290# default implementations for many methods
1291
1292sub _time {
1293 # probe for availability of Time::HiRes
1294 if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
1295 warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8;
1296 *_time = \&Time::HiRes::time;
1297 # if (eval "use POSIX (); (POSIX::times())...
1298 } else {
1299 warn "AnyEvent: using built-in time(), WARNING, no sub-second resolution!\n" if $VERBOSE;
1300 *_time = sub { time }; # epic fail
1301 }
1302
1303 &_time
1304}
1305
1306sub time { _time }
1307sub now { _time }
1308sub now_update { }
1309
638# default implementation for ->condvar, ->wait, ->broadcast 1310# default implementation for ->condvar
639 1311
640sub condvar { 1312sub condvar {
641 bless \my $flag, "AnyEvent::Base::CondVar" 1313 bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
642}
643
644sub AnyEvent::Base::CondVar::broadcast {
645 ${$_[0]}++;
646}
647
648sub AnyEvent::Base::CondVar::wait {
649 AnyEvent->one_event while !${$_[0]};
650} 1314}
651 1315
652# default implementation for ->signal 1316# default implementation for ->signal
653 1317
654our %SIG_CB; 1318our $HAVE_ASYNC_INTERRUPT;
1319
1320sub _have_async_interrupt() {
1321 $HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT}
1322 && eval "use Async::Interrupt 1.0 (); 1")
1323 unless defined $HAVE_ASYNC_INTERRUPT;
1324
1325 $HAVE_ASYNC_INTERRUPT
1326}
1327
1328our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
1329our (%SIG_ASY, %SIG_ASY_W);
1330our ($SIG_COUNT, $SIG_TW);
1331
1332sub _signal_exec {
1333 $HAVE_ASYNC_INTERRUPT
1334 ? $SIGPIPE_R->drain
1335 : sysread $SIGPIPE_R, my $dummy, 9;
1336
1337 while (%SIG_EV) {
1338 for (keys %SIG_EV) {
1339 delete $SIG_EV{$_};
1340 $_->() for values %{ $SIG_CB{$_} || {} };
1341 }
1342 }
1343}
1344
1345# install a dummy wakeup watcher to reduce signal catching latency
1346sub _sig_add() {
1347 unless ($SIG_COUNT++) {
1348 # try to align timer on a full-second boundary, if possible
1349 my $NOW = AnyEvent->now;
1350
1351 $SIG_TW = AnyEvent->timer (
1352 after => $MAX_SIGNAL_LATENCY - ($NOW - int $NOW),
1353 interval => $MAX_SIGNAL_LATENCY,
1354 cb => sub { }, # just for the PERL_ASYNC_CHECK
1355 );
1356 }
1357}
1358
1359sub _sig_del {
1360 undef $SIG_TW
1361 unless --$SIG_COUNT;
1362}
1363
1364our $_sig_name_init; $_sig_name_init = sub {
1365 eval q{ # poor man's autoloading
1366 undef $_sig_name_init;
1367
1368 if (_have_async_interrupt) {
1369 *sig2num = \&Async::Interrupt::sig2num;
1370 *sig2name = \&Async::Interrupt::sig2name;
1371 } else {
1372 require Config;
1373
1374 my %signame2num;
1375 @signame2num{ split ' ', $Config::Config{sig_name} }
1376 = split ' ', $Config::Config{sig_num};
1377
1378 my @signum2name;
1379 @signum2name[values %signame2num] = keys %signame2num;
1380
1381 *sig2num = sub($) {
1382 $_[0] > 0 ? shift : $signame2num{+shift}
1383 };
1384 *sig2name = sub ($) {
1385 $_[0] > 0 ? $signum2name[+shift] : shift
1386 };
1387 }
1388 };
1389 die if $@;
1390};
1391
1392sub sig2num ($) { &$_sig_name_init; &sig2num }
1393sub sig2name($) { &$_sig_name_init; &sig2name }
655 1394
656sub signal { 1395sub signal {
1396 eval q{ # poor man's autoloading {}
1397 # probe for availability of Async::Interrupt
1398 if (_have_async_interrupt) {
1399 warn "AnyEvent: using Async::Interrupt for race-free signal handling.\n" if $VERBOSE >= 8;
1400
1401 $SIGPIPE_R = new Async::Interrupt::EventPipe;
1402 $SIG_IO = AnyEvent->io (fh => $SIGPIPE_R->fileno, poll => "r", cb => \&_signal_exec);
1403
1404 } else {
1405 warn "AnyEvent: using emulated perl signal handling with latency timer.\n" if $VERBOSE >= 8;
1406
1407 require Fcntl;
1408
1409 if (AnyEvent::WIN32) {
1410 require AnyEvent::Util;
1411
1412 ($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
1413 AnyEvent::Util::fh_nonblocking ($SIGPIPE_R, 1) if $SIGPIPE_R;
1414 AnyEvent::Util::fh_nonblocking ($SIGPIPE_W, 1) if $SIGPIPE_W; # just in case
1415 } else {
1416 pipe $SIGPIPE_R, $SIGPIPE_W;
1417 fcntl $SIGPIPE_R, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_R;
1418 fcntl $SIGPIPE_W, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_W; # just in case
1419
1420 # not strictly required, as $^F is normally 2, but let's make sure...
1421 fcntl $SIGPIPE_R, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
1422 fcntl $SIGPIPE_W, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
1423 }
1424
1425 $SIGPIPE_R
1426 or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
1427
1428 $SIG_IO = AnyEvent->io (fh => $SIGPIPE_R, poll => "r", cb => \&_signal_exec);
1429 }
1430
1431 *signal = sub {
657 my (undef, %arg) = @_; 1432 my (undef, %arg) = @_;
658 1433
659 my $signal = uc $arg{signal} 1434 my $signal = uc $arg{signal}
660 or Carp::croak "required option 'signal' is missing"; 1435 or Carp::croak "required option 'signal' is missing";
661 1436
1437 if ($HAVE_ASYNC_INTERRUPT) {
1438 # async::interrupt
1439
1440 $signal = sig2num $signal;
662 $SIG_CB{$signal}{$arg{cb}} = $arg{cb}; 1441 $SIG_CB{$signal}{$arg{cb}} = $arg{cb};
1442
1443 $SIG_ASY{$signal} ||= new Async::Interrupt
1444 cb => sub { undef $SIG_EV{$signal} },
1445 signal => $signal,
1446 pipe => [$SIGPIPE_R->filenos],
1447 pipe_autodrain => 0,
1448 ;
1449
1450 } else {
1451 # pure perl
1452
1453 # AE::Util has been loaded in signal
1454 $signal = sig2name $signal;
1455 $SIG_CB{$signal}{$arg{cb}} = $arg{cb};
1456
663 $SIG{$signal} ||= sub { 1457 $SIG{$signal} ||= sub {
664 $_->() for values %{ $SIG_CB{$signal} || {} }; 1458 local $!;
1459 syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
1460 undef $SIG_EV{$signal};
1461 };
1462
1463 # can't do signal processing without introducing races in pure perl,
1464 # so limit the signal latency.
1465 _sig_add;
1466 }
1467
1468 bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
1469 };
1470
1471 *AnyEvent::Base::signal::DESTROY = sub {
1472 my ($signal, $cb) = @{$_[0]};
1473
1474 _sig_del;
1475
1476 delete $SIG_CB{$signal}{$cb};
1477
1478 $HAVE_ASYNC_INTERRUPT
1479 ? delete $SIG_ASY{$signal}
1480 : # delete doesn't work with older perls - they then
1481 # print weird messages, or just unconditionally exit
1482 # instead of getting the default action.
1483 undef $SIG{$signal}
1484 unless keys %{ $SIG_CB{$signal} };
1485 };
665 }; 1486 };
666 1487 die if $@;
667 bless [$signal, $arg{cb}], "AnyEvent::Base::Signal" 1488 &signal
668}
669
670sub AnyEvent::Base::Signal::DESTROY {
671 my ($signal, $cb) = @{$_[0]};
672
673 delete $SIG_CB{$signal}{$cb};
674
675 $SIG{$signal} = 'DEFAULT' unless keys %{ $SIG_CB{$signal} };
676} 1489}
677 1490
678# default implementation for ->child 1491# default implementation for ->child
679 1492
680our %PID_CB; 1493our %PID_CB;
681our $CHLD_W; 1494our $CHLD_W;
682our $CHLD_DELAY_W; 1495our $CHLD_DELAY_W;
683our $PID_IDLE;
684our $WNOHANG; 1496our $WNOHANG;
685 1497
686sub _child_wait { 1498sub _emit_childstatus($$) {
687 while (0 < (my $pid = waitpid -1, $WNOHANG)) { 1499 my (undef, $rpid, $rstatus) = @_;
1500
1501 $_->($rpid, $rstatus)
688 $_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }), 1502 for values %{ $PID_CB{$rpid} || {} },
689 (values %{ $PID_CB{0} || {} }); 1503 values %{ $PID_CB{0} || {} };
690 }
691
692 undef $PID_IDLE;
693} 1504}
694 1505
695sub _sigchld { 1506sub _sigchld {
696 # make sure we deliver these changes "synchronous" with the event loop. 1507 my $pid;
697 $CHLD_DELAY_W ||= AnyEvent->timer (after => 0, cb => sub { 1508
698 undef $CHLD_DELAY_W; 1509 AnyEvent->_emit_childstatus ($pid, $?)
699 &_child_wait; 1510 while ($pid = waitpid -1, $WNOHANG) > 0;
700 });
701} 1511}
702 1512
703sub child { 1513sub child {
704 my (undef, %arg) = @_; 1514 my (undef, %arg) = @_;
705 1515
706 defined (my $pid = $arg{pid} + 0) 1516 defined (my $pid = $arg{pid} + 0)
707 or Carp::croak "required option 'pid' is missing"; 1517 or Carp::croak "required option 'pid' is missing";
708 1518
709 $PID_CB{$pid}{$arg{cb}} = $arg{cb}; 1519 $PID_CB{$pid}{$arg{cb}} = $arg{cb};
710 1520
711 unless ($WNOHANG) { 1521 # WNOHANG is almost cetrainly 1 everywhere
712 $WNOHANG = eval { require POSIX; &POSIX::WNOHANG } || 1; 1522 $WNOHANG ||= $^O =~ /^(?:openbsd|netbsd|linux|freebsd|cygwin|MSWin32)$/
713 } 1523 ? 1
1524 : eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
714 1525
715 unless ($CHLD_W) { 1526 unless ($CHLD_W) {
716 $CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld); 1527 $CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
717 # child could be a zombie already, so make at least one round 1528 # child could be a zombie already, so make at least one round
718 &_sigchld; 1529 &_sigchld;
719 } 1530 }
720 1531
721 bless [$pid, $arg{cb}], "AnyEvent::Base::Child" 1532 bless [$pid, $arg{cb}], "AnyEvent::Base::child"
722} 1533}
723 1534
724sub AnyEvent::Base::Child::DESTROY { 1535sub AnyEvent::Base::child::DESTROY {
725 my ($pid, $cb) = @{$_[0]}; 1536 my ($pid, $cb) = @{$_[0]};
726 1537
727 delete $PID_CB{$pid}{$cb}; 1538 delete $PID_CB{$pid}{$cb};
728 delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} }; 1539 delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
729 1540
730 undef $CHLD_W unless keys %PID_CB; 1541 undef $CHLD_W unless keys %PID_CB;
731} 1542}
1543
1544# idle emulation is done by simply using a timer, regardless
1545# of whether the process is idle or not, and not letting
1546# the callback use more than 50% of the time.
1547sub idle {
1548 my (undef, %arg) = @_;
1549
1550 my ($cb, $w, $rcb) = $arg{cb};
1551
1552 $rcb = sub {
1553 if ($cb) {
1554 $w = _time;
1555 &$cb;
1556 $w = _time - $w;
1557
1558 # never use more then 50% of the time for the idle watcher,
1559 # within some limits
1560 $w = 0.0001 if $w < 0.0001;
1561 $w = 5 if $w > 5;
1562
1563 $w = AnyEvent->timer (after => $w, cb => $rcb);
1564 } else {
1565 # clean up...
1566 undef $w;
1567 undef $rcb;
1568 }
1569 };
1570
1571 $w = AnyEvent->timer (after => 0.05, cb => $rcb);
1572
1573 bless \\$cb, "AnyEvent::Base::idle"
1574}
1575
1576sub AnyEvent::Base::idle::DESTROY {
1577 undef $${$_[0]};
1578}
1579
1580package AnyEvent::CondVar;
1581
1582our @ISA = AnyEvent::CondVar::Base::;
1583
1584package AnyEvent::CondVar::Base;
1585
1586#use overload
1587# '&{}' => sub { my $self = shift; sub { $self->send (@_) } },
1588# fallback => 1;
1589
1590# save 300+ kilobytes by dirtily hardcoding overloading
1591${"AnyEvent::CondVar::Base::OVERLOAD"}{dummy}++; # Register with magic by touching.
1592*{'AnyEvent::CondVar::Base::()'} = sub { }; # "Make it findable via fetchmethod."
1593*{'AnyEvent::CondVar::Base::(&{}'} = sub { my $self = shift; sub { $self->send (@_) } }; # &{}
1594${'AnyEvent::CondVar::Base::()'} = 1; # fallback
1595
1596our $WAITING;
1597
1598sub _send {
1599 # nop
1600}
1601
1602sub send {
1603 my $cv = shift;
1604 $cv->{_ae_sent} = [@_];
1605 (delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
1606 $cv->_send;
1607}
1608
1609sub croak {
1610 $_[0]{_ae_croak} = $_[1];
1611 $_[0]->send;
1612}
1613
1614sub ready {
1615 $_[0]{_ae_sent}
1616}
1617
1618sub _wait {
1619 $WAITING
1620 and !$_[0]{_ae_sent}
1621 and Carp::croak "AnyEvent::CondVar: recursive blocking wait detected";
1622
1623 local $WAITING = 1;
1624 AnyEvent->one_event while !$_[0]{_ae_sent};
1625}
1626
1627sub recv {
1628 $_[0]->_wait;
1629
1630 Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
1631 wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
1632}
1633
1634sub cb {
1635 $_[0]{_ae_cb} = $_[1] if @_ > 1;
1636 $_[0]{_ae_cb}
1637}
1638
1639sub begin {
1640 ++$_[0]{_ae_counter};
1641 $_[0]{_ae_end_cb} = $_[1] if @_ > 1;
1642}
1643
1644sub end {
1645 return if --$_[0]{_ae_counter};
1646 &{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
1647}
1648
1649# undocumented/compatibility with pre-3.4
1650*broadcast = \&send;
1651*wait = \&_wait;
1652
1653#############################################################################
1654# "new" API, currently only emulation of it
1655#############################################################################
1656
1657package AE;
1658
1659sub io($$$) {
1660 AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2])
1661}
1662
1663sub timer($$$) {
1664 AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2]);
1665}
1666
1667sub signal($$) {
1668 AnyEvent->signal (signal => $_[0], cb => $_[1]);
1669}
1670
1671sub child($$) {
1672 AnyEvent->child (pid => $_[0], cb => $_[1]);
1673}
1674
1675sub idle($) {
1676 AnyEvent->idle (cb => $_[0]);
1677}
1678
1679sub cv() {
1680 AnyEvent->condvar
1681}
1682
1683sub now() {
1684 AnyEvent->now
1685}
1686
1687sub now_update() {
1688 AnyEvent->now_update
1689}
1690
1691sub time() {
1692 AnyEvent->time
1693}
1694
1695=head1 ERROR AND EXCEPTION HANDLING
1696
1697In general, AnyEvent does not do any error handling - it relies on the
1698caller to do that if required. The L<AnyEvent::Strict> module (see also
1699the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
1700checking of all AnyEvent methods, however, which is highly useful during
1701development.
1702
1703As for exception handling (i.e. runtime errors and exceptions thrown while
1704executing a callback), this is not only highly event-loop specific, but
1705also not in any way wrapped by this module, as this is the job of the main
1706program.
1707
1708The pure perl event loop simply re-throws the exception (usually
1709within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
1710$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
1711so on.
1712
1713=head1 ENVIRONMENT VARIABLES
1714
1715The following environment variables are used by this module or its
1716submodules.
1717
1718Note that AnyEvent will remove I<all> environment variables starting with
1719C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
1720enabled.
1721
1722=over 4
1723
1724=item C<PERL_ANYEVENT_VERBOSE>
1725
1726By default, AnyEvent will be completely silent except in fatal
1727conditions. You can set this environment variable to make AnyEvent more
1728talkative.
1729
1730When set to C<1> or higher, causes AnyEvent to warn about unexpected
1731conditions, such as not being able to load the event model specified by
1732C<PERL_ANYEVENT_MODEL>.
1733
1734When set to C<2> or higher, cause AnyEvent to report to STDERR which event
1735model it chooses.
1736
1737When set to C<8> or higher, then AnyEvent will report extra information on
1738which optional modules it loads and how it implements certain features.
1739
1740=item C<PERL_ANYEVENT_STRICT>
1741
1742AnyEvent does not do much argument checking by default, as thorough
1743argument checking is very costly. Setting this variable to a true value
1744will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
1745check the arguments passed to most method calls. If it finds any problems,
1746it will croak.
1747
1748In other words, enables "strict" mode.
1749
1750Unlike C<use strict> (or it's modern cousin, C<< use L<common::sense>
1751>>, it is definitely recommended to keep it off in production. Keeping
1752C<PERL_ANYEVENT_STRICT=1> in your environment while developing programs
1753can be very useful, however.
1754
1755=item C<PERL_ANYEVENT_MODEL>
1756
1757This can be used to specify the event model to be used by AnyEvent, before
1758auto detection and -probing kicks in. It must be a string consisting
1759entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
1760and the resulting module name is loaded and if the load was successful,
1761used as event model. If it fails to load AnyEvent will proceed with
1762auto detection and -probing.
1763
1764This functionality might change in future versions.
1765
1766For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
1767could start your program like this:
1768
1769 PERL_ANYEVENT_MODEL=Perl perl ...
1770
1771=item C<PERL_ANYEVENT_PROTOCOLS>
1772
1773Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
1774for IPv4 or IPv6. The default is unspecified (and might change, or be the result
1775of auto probing).
1776
1777Must be set to a comma-separated list of protocols or address families,
1778current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
1779used, and preference will be given to protocols mentioned earlier in the
1780list.
1781
1782This variable can effectively be used for denial-of-service attacks
1783against local programs (e.g. when setuid), although the impact is likely
1784small, as the program has to handle conenction and other failures anyways.
1785
1786Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
1787but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
1788- only support IPv4, never try to resolve or contact IPv6
1789addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
1790IPv6, but prefer IPv6 over IPv4.
1791
1792=item C<PERL_ANYEVENT_EDNS0>
1793
1794Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
1795for DNS. This extension is generally useful to reduce DNS traffic, but
1796some (broken) firewalls drop such DNS packets, which is why it is off by
1797default.
1798
1799Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
1800EDNS0 in its DNS requests.
1801
1802=item C<PERL_ANYEVENT_MAX_FORKS>
1803
1804The maximum number of child processes that C<AnyEvent::Util::fork_call>
1805will create in parallel.
1806
1807=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
1808
1809The default value for the C<max_outstanding> parameter for the default DNS
1810resolver - this is the maximum number of parallel DNS requests that are
1811sent to the DNS server.
1812
1813=item C<PERL_ANYEVENT_RESOLV_CONF>
1814
1815The file to use instead of F</etc/resolv.conf> (or OS-specific
1816configuration) in the default resolver. When set to the empty string, no
1817default config will be used.
1818
1819=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
1820
1821When neither C<ca_file> nor C<ca_path> was specified during
1822L<AnyEvent::TLS> context creation, and either of these environment
1823variables exist, they will be used to specify CA certificate locations
1824instead of a system-dependent default.
1825
1826=item C<PERL_ANYEVENT_AVOID_GUARD> and C<PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT>
1827
1828When these are set to C<1>, then the respective modules are not
1829loaded. Mostly good for testing AnyEvent itself.
1830
1831=back
732 1832
733=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE 1833=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
734 1834
735This is an advanced topic that you do not normally need to use AnyEvent in 1835This is an advanced topic that you do not normally need to use AnyEvent in
736a module. This section is only of use to event loop authors who want to 1836a module. This section is only of use to event loop authors who want to
770 1870
771I<rxvt-unicode> also cheats a bit by not providing blocking access to 1871I<rxvt-unicode> also cheats a bit by not providing blocking access to
772condition variables: code blocking while waiting for a condition will 1872condition variables: code blocking while waiting for a condition will
773C<die>. This still works with most modules/usages, and blocking calls must 1873C<die>. This still works with most modules/usages, and blocking calls must
774not be done in an interactive application, so it makes sense. 1874not be done in an interactive application, so it makes sense.
775
776=head1 ENVIRONMENT VARIABLES
777
778The following environment variables are used by this module:
779
780=over 4
781
782=item C<PERL_ANYEVENT_VERBOSE>
783
784By default, AnyEvent will be completely silent except in fatal
785conditions. You can set this environment variable to make AnyEvent more
786talkative.
787
788When set to C<1> or higher, causes AnyEvent to warn about unexpected
789conditions, such as not being able to load the event model specified by
790C<PERL_ANYEVENT_MODEL>.
791
792When set to C<2> or higher, cause AnyEvent to report to STDERR which event
793model it chooses.
794
795=item C<PERL_ANYEVENT_MODEL>
796
797This can be used to specify the event model to be used by AnyEvent, before
798autodetection and -probing kicks in. It must be a string consisting
799entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
800and the resulting module name is loaded and if the load was successful,
801used as event model. If it fails to load AnyEvent will proceed with
802autodetection and -probing.
803
804This functionality might change in future versions.
805
806For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
807could start your program like this:
808
809 PERL_ANYEVENT_MODEL=Perl perl ...
810
811=back
812 1875
813=head1 EXAMPLE PROGRAM 1876=head1 EXAMPLE PROGRAM
814 1877
815The following program uses an I/O watcher to read data from STDIN, a timer 1878The following program uses an I/O watcher to read data from STDIN, a timer
816to display a message once per second, and a condition variable to quit the 1879to display a message once per second, and a condition variable to quit the
825 poll => 'r', 1888 poll => 'r',
826 cb => sub { 1889 cb => sub {
827 warn "io event <$_[0]>\n"; # will always output <r> 1890 warn "io event <$_[0]>\n"; # will always output <r>
828 chomp (my $input = <STDIN>); # read a line 1891 chomp (my $input = <STDIN>); # read a line
829 warn "read: $input\n"; # output what has been read 1892 warn "read: $input\n"; # output what has been read
830 $cv->broadcast if $input =~ /^q/i; # quit program if /^q/i 1893 $cv->send if $input =~ /^q/i; # quit program if /^q/i
831 }, 1894 },
832 ); 1895 );
833 1896
834 my $time_watcher; # can only be used once 1897 my $time_watcher; # can only be used once
835 1898
840 }); 1903 });
841 } 1904 }
842 1905
843 new_timer; # create first timer 1906 new_timer; # create first timer
844 1907
845 $cv->wait; # wait until user enters /^q/i 1908 $cv->recv; # wait until user enters /^q/i
846 1909
847=head1 REAL-WORLD EXAMPLE 1910=head1 REAL-WORLD EXAMPLE
848 1911
849Consider the L<Net::FCP> module. It features (among others) the following 1912Consider the L<Net::FCP> module. It features (among others) the following
850API calls, which are to freenet what HTTP GET requests are to http: 1913API calls, which are to freenet what HTTP GET requests are to http:
900 syswrite $txn->{fh}, $txn->{request} 1963 syswrite $txn->{fh}, $txn->{request}
901 or die "connection or write error"; 1964 or die "connection or write error";
902 $txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r }); 1965 $txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
903 1966
904Again, C<fh_ready_r> waits till all data has arrived, and then stores the 1967Again, C<fh_ready_r> waits till all data has arrived, and then stores the
905result and signals any possible waiters that the request ahs finished: 1968result and signals any possible waiters that the request has finished:
906 1969
907 sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf}; 1970 sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
908 1971
909 if (end-of-file or data complete) { 1972 if (end-of-file or data complete) {
910 $txn->{result} = $txn->{buf}; 1973 $txn->{result} = $txn->{buf};
911 $txn->{finished}->broadcast; 1974 $txn->{finished}->send;
912 $txb->{cb}->($txn) of $txn->{cb}; # also call callback 1975 $txb->{cb}->($txn) of $txn->{cb}; # also call callback
913 } 1976 }
914 1977
915The C<result> method, finally, just waits for the finished signal (if the 1978The C<result> method, finally, just waits for the finished signal (if the
916request was already finished, it doesn't wait, of course, and returns the 1979request was already finished, it doesn't wait, of course, and returns the
917data: 1980data:
918 1981
919 $txn->{finished}->wait; 1982 $txn->{finished}->recv;
920 return $txn->{result}; 1983 return $txn->{result};
921 1984
922The actual code goes further and collects all errors (C<die>s, exceptions) 1985The actual code goes further and collects all errors (C<die>s, exceptions)
923that occured during request processing. The C<result> method detects 1986that occurred during request processing. The C<result> method detects
924whether an exception as thrown (it is stored inside the $txn object) 1987whether an exception as thrown (it is stored inside the $txn object)
925and just throws the exception, which means connection errors and other 1988and just throws the exception, which means connection errors and other
926problems get reported tot he code that tries to use the result, not in a 1989problems get reported tot he code that tries to use the result, not in a
927random callback. 1990random callback.
928 1991
959 2022
960 my $quit = AnyEvent->condvar; 2023 my $quit = AnyEvent->condvar;
961 2024
962 $fcp->txn_client_get ($url)->cb (sub { 2025 $fcp->txn_client_get ($url)->cb (sub {
963 ... 2026 ...
964 $quit->broadcast; 2027 $quit->send;
965 }); 2028 });
966 2029
967 $quit->wait; 2030 $quit->recv;
968 2031
969 2032
970=head1 BENCHMARKS 2033=head1 BENCHMARKS
971 2034
972To give you an idea of the performance and overheads that AnyEvent adds 2035To give you an idea of the performance and overheads that AnyEvent adds
974of various event loops I prepared some benchmarks. 2037of various event loops I prepared some benchmarks.
975 2038
976=head2 BENCHMARKING ANYEVENT OVERHEAD 2039=head2 BENCHMARKING ANYEVENT OVERHEAD
977 2040
978Here is a benchmark of various supported event models used natively and 2041Here is a benchmark of various supported event models used natively and
979through anyevent. The benchmark creates a lot of timers (with a zero 2042through AnyEvent. The benchmark creates a lot of timers (with a zero
980timeout) and I/O watchers (watching STDOUT, a pty, to become writable, 2043timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
981which it is), lets them fire exactly once and destroys them again. 2044which it is), lets them fire exactly once and destroys them again.
982 2045
983Source code for this benchmark is found as F<eg/bench> in the AnyEvent 2046Source code for this benchmark is found as F<eg/bench> in the AnyEvent
984distribution. 2047distribution.
1001all watchers, to avoid adding memory overhead. That means closure creation 2064all watchers, to avoid adding memory overhead. That means closure creation
1002and memory usage is not included in the figures. 2065and memory usage is not included in the figures.
1003 2066
1004I<invoke> is the time, in microseconds, used to invoke a simple 2067I<invoke> is the time, in microseconds, used to invoke a simple
1005callback. The callback simply counts down a Perl variable and after it was 2068callback. The callback simply counts down a Perl variable and after it was
1006invoked "watcher" times, it would C<< ->broadcast >> a condvar once to 2069invoked "watcher" times, it would C<< ->send >> a condvar once to
1007signal the end of this phase. 2070signal the end of this phase.
1008 2071
1009I<destroy> is the time, in microseconds, that it takes to destroy a single 2072I<destroy> is the time, in microseconds, that it takes to destroy a single
1010watcher. 2073watcher.
1011 2074
1012=head3 Results 2075=head3 Results
1013 2076
1014 name watchers bytes create invoke destroy comment 2077 name watchers bytes create invoke destroy comment
1015 EV/EV 400000 244 0.56 0.46 0.31 EV native interface 2078 EV/EV 400000 224 0.47 0.35 0.27 EV native interface
1016 EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers 2079 EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers
1017 CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal 2080 CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal
1018 Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation 2081 Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation
1019 Event/Event 16000 516 31.88 31.30 0.85 Event native interface 2082 Event/Event 16000 517 32.20 31.80 0.81 Event native interface
1020 Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers 2083 Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers
2084 IOAsync/Any 16000 989 38.10 32.77 11.13 via IO::Async::Loop::IO_Poll
2085 IOAsync/Any 16000 990 37.59 29.50 10.61 via IO::Async::Loop::Epoll
1021 Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour 2086 Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour
1022 Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers 2087 Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers
1023 POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event 2088 POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event
1024 POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select 2089 POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select
1025 2090
1026=head3 Discussion 2091=head3 Discussion
1027 2092
1028The benchmark does I<not> measure scalability of the event loop very 2093The benchmark does I<not> measure scalability of the event loop very
1029well. For example, a select-based event loop (such as the pure perl one) 2094well. For example, a select-based event loop (such as the pure perl one)
1054performance becomes really bad with lots of file descriptors (and few of 2119performance becomes really bad with lots of file descriptors (and few of
1055them active), of course, but this was not subject of this benchmark. 2120them active), of course, but this was not subject of this benchmark.
1056 2121
1057The C<Event> module has a relatively high setup and callback invocation 2122The C<Event> module has a relatively high setup and callback invocation
1058cost, but overall scores in on the third place. 2123cost, but overall scores in on the third place.
2124
2125C<IO::Async> performs admirably well, about on par with C<Event>, even
2126when using its pure perl backend.
1059 2127
1060C<Glib>'s memory usage is quite a bit higher, but it features a 2128C<Glib>'s memory usage is quite a bit higher, but it features a
1061faster callback invocation and overall ends up in the same class as 2129faster callback invocation and overall ends up in the same class as
1062C<Event>. However, Glib scales extremely badly, doubling the number of 2130C<Event>. However, Glib scales extremely badly, doubling the number of
1063watchers increases the processing time by more than a factor of four, 2131watchers increases the processing time by more than a factor of four,
1107 2175
1108=back 2176=back
1109 2177
1110=head2 BENCHMARKING THE LARGE SERVER CASE 2178=head2 BENCHMARKING THE LARGE SERVER CASE
1111 2179
1112This benchmark atcually benchmarks the event loop itself. It works by 2180This benchmark actually benchmarks the event loop itself. It works by
1113creating a number of "servers": each server consists of a socketpair, a 2181creating a number of "servers": each server consists of a socket pair, a
1114timeout watcher that gets reset on activity (but never fires), and an I/O 2182timeout watcher that gets reset on activity (but never fires), and an I/O
1115watcher waiting for input on one side of the socket. Each time the socket 2183watcher waiting for input on one side of the socket. Each time the socket
1116watcher reads a byte it will write that byte to a random other "server". 2184watcher reads a byte it will write that byte to a random other "server".
1117 2185
1118The effect is that there will be a lot of I/O watchers, only part of which 2186The effect is that there will be a lot of I/O watchers, only part of which
1119are active at any one point (so there is a constant number of active 2187are active at any one point (so there is a constant number of active
1120fds for each loop iterstaion, but which fds these are is random). The 2188fds for each loop iteration, but which fds these are is random). The
1121timeout is reset each time something is read because that reflects how 2189timeout is reset each time something is read because that reflects how
1122most timeouts work (and puts extra pressure on the event loops). 2190most timeouts work (and puts extra pressure on the event loops).
1123 2191
1124In this benchmark, we use 10000 socketpairs (20000 sockets), of which 100 2192In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
1125(1%) are active. This mirrors the activity of large servers with many 2193(1%) are active. This mirrors the activity of large servers with many
1126connections, most of which are idle at any one point in time. 2194connections, most of which are idle at any one point in time.
1127 2195
1128Source code for this benchmark is found as F<eg/bench2> in the AnyEvent 2196Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
1129distribution. 2197distribution.
1131=head3 Explanation of the columns 2199=head3 Explanation of the columns
1132 2200
1133I<sockets> is the number of sockets, and twice the number of "servers" (as 2201I<sockets> is the number of sockets, and twice the number of "servers" (as
1134each server has a read and write socket end). 2202each server has a read and write socket end).
1135 2203
1136I<create> is the time it takes to create a socketpair (which is 2204I<create> is the time it takes to create a socket pair (which is
1137nontrivial) and two watchers: an I/O watcher and a timeout watcher. 2205nontrivial) and two watchers: an I/O watcher and a timeout watcher.
1138 2206
1139I<request>, the most important value, is the time it takes to handle a 2207I<request>, the most important value, is the time it takes to handle a
1140single "request", that is, reading the token from the pipe and forwarding 2208single "request", that is, reading the token from the pipe and forwarding
1141it to another server. This includes deleting the old timeout and creating 2209it to another server. This includes deleting the old timeout and creating
1142a new one that moves the timeout into the future. 2210a new one that moves the timeout into the future.
1143 2211
1144=head3 Results 2212=head3 Results
1145 2213
1146 name sockets create request 2214 name sockets create request
1147 EV 20000 69.01 11.16 2215 EV 20000 69.01 11.16
1148 Perl 20000 73.32 35.87 2216 Perl 20000 73.32 35.87
2217 IOAsync 20000 157.00 98.14 epoll
2218 IOAsync 20000 159.31 616.06 poll
1149 Event 20000 212.62 257.32 2219 Event 20000 212.62 257.32
1150 Glib 20000 651.16 1896.30 2220 Glib 20000 651.16 1896.30
1151 POE 20000 349.67 12317.24 uses POE::Loop::Event 2221 POE 20000 349.67 12317.24 uses POE::Loop::Event
1152 2222
1153=head3 Discussion 2223=head3 Discussion
1154 2224
1155This benchmark I<does> measure scalability and overall performance of the 2225This benchmark I<does> measure scalability and overall performance of the
1156particular event loop. 2226particular event loop.
1158EV is again fastest. Since it is using epoll on my system, the setup time 2228EV is again fastest. Since it is using epoll on my system, the setup time
1159is relatively high, though. 2229is relatively high, though.
1160 2230
1161Perl surprisingly comes second. It is much faster than the C-based event 2231Perl surprisingly comes second. It is much faster than the C-based event
1162loops Event and Glib. 2232loops Event and Glib.
2233
2234IO::Async performs very well when using its epoll backend, and still quite
2235good compared to Glib when using its pure perl backend.
1163 2236
1164Event suffers from high setup time as well (look at its code and you will 2237Event suffers from high setup time as well (look at its code and you will
1165understand why). Callback invocation also has a high overhead compared to 2238understand why). Callback invocation also has a high overhead compared to
1166the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event 2239the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
1167uses select or poll in basically all documented configurations. 2240uses select or poll in basically all documented configurations.
1214speed most when you have lots of watchers, not when you only have a few of 2287speed most when you have lots of watchers, not when you only have a few of
1215them). 2288them).
1216 2289
1217EV is again fastest. 2290EV is again fastest.
1218 2291
1219Perl again comes second. It is noticably faster than the C-based event 2292Perl again comes second. It is noticeably faster than the C-based event
1220loops Event and Glib, although the difference is too small to really 2293loops Event and Glib, although the difference is too small to really
1221matter. 2294matter.
1222 2295
1223POE also performs much better in this case, but is is still far behind the 2296POE also performs much better in this case, but is is still far behind the
1224others. 2297others.
1230=item * C-based event loops perform very well with small number of 2303=item * C-based event loops perform very well with small number of
1231watchers, as the management overhead dominates. 2304watchers, as the management overhead dominates.
1232 2305
1233=back 2306=back
1234 2307
2308=head2 THE IO::Lambda BENCHMARK
2309
2310Recently I was told about the benchmark in the IO::Lambda manpage, which
2311could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
2312simply compares IO::Lambda with POE, and IO::Lambda looks better (which
2313shouldn't come as a surprise to anybody). As such, the benchmark is
2314fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
2315very optimal. But how would AnyEvent compare when used without the extra
2316baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
2317
2318The benchmark itself creates an echo-server, and then, for 500 times,
2319connects to the echo server, sends a line, waits for the reply, and then
2320creates the next connection. This is a rather bad benchmark, as it doesn't
2321test the efficiency of the framework or much non-blocking I/O, but it is a
2322benchmark nevertheless.
2323
2324 name runtime
2325 Lambda/select 0.330 sec
2326 + optimized 0.122 sec
2327 Lambda/AnyEvent 0.327 sec
2328 + optimized 0.138 sec
2329 Raw sockets/select 0.077 sec
2330 POE/select, components 0.662 sec
2331 POE/select, raw sockets 0.226 sec
2332 POE/select, optimized 0.404 sec
2333
2334 AnyEvent/select/nb 0.085 sec
2335 AnyEvent/EV/nb 0.068 sec
2336 +state machine 0.134 sec
2337
2338The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
2339benchmarks actually make blocking connects and use 100% blocking I/O,
2340defeating the purpose of an event-based solution. All of the newly
2341written AnyEvent benchmarks use 100% non-blocking connects (using
2342AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
2343resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
2344generally require a lot more bookkeeping and event handling than blocking
2345connects (which involve a single syscall only).
2346
2347The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
2348offers similar expressive power as POE and IO::Lambda, using conventional
2349Perl syntax. This means that both the echo server and the client are 100%
2350non-blocking, further placing it at a disadvantage.
2351
2352As you can see, the AnyEvent + EV combination even beats the
2353hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
2354backend easily beats IO::Lambda and POE.
2355
2356And even the 100% non-blocking version written using the high-level (and
2357slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda by a
2358large margin, even though it does all of DNS, tcp-connect and socket I/O
2359in a non-blocking way.
2360
2361The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
2362F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
2363part of the IO::lambda distribution and were used without any changes.
2364
2365
2366=head1 SIGNALS
2367
2368AnyEvent currently installs handlers for these signals:
2369
2370=over 4
2371
2372=item SIGCHLD
2373
2374A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
2375emulation for event loops that do not support them natively. Also, some
2376event loops install a similar handler.
2377
2378Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE, then
2379AnyEvent will reset it to default, to avoid losing child exit statuses.
2380
2381=item SIGPIPE
2382
2383A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
2384when AnyEvent gets loaded.
2385
2386The rationale for this is that AnyEvent users usually do not really depend
2387on SIGPIPE delivery (which is purely an optimisation for shell use, or
2388badly-written programs), but C<SIGPIPE> can cause spurious and rare
2389program exits as a lot of people do not expect C<SIGPIPE> when writing to
2390some random socket.
2391
2392The rationale for installing a no-op handler as opposed to ignoring it is
2393that this way, the handler will be restored to defaults on exec.
2394
2395Feel free to install your own handler, or reset it to defaults.
2396
2397=back
2398
2399=cut
2400
2401undef $SIG{CHLD}
2402 if $SIG{CHLD} eq 'IGNORE';
2403
2404$SIG{PIPE} = sub { }
2405 unless defined $SIG{PIPE};
2406
2407=head1 RECOMMENDED/OPTIONAL MODULES
2408
2409One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
2410it's built-in modules) are required to use it.
2411
2412That does not mean that AnyEvent won't take advantage of some additional
2413modules if they are installed.
2414
2415This section epxlains which additional modules will be used, and how they
2416affect AnyEvent's operetion.
2417
2418=over 4
2419
2420=item L<Async::Interrupt>
2421
2422This slightly arcane module is used to implement fast signal handling: To
2423my knowledge, there is no way to do completely race-free and quick
2424signal handling in pure perl. To ensure that signals still get
2425delivered, AnyEvent will start an interval timer to wake up perl (and
2426catch the signals) with some delay (default is 10 seconds, look for
2427C<$AnyEvent::MAX_SIGNAL_LATENCY>).
2428
2429If this module is available, then it will be used to implement signal
2430catching, which means that signals will not be delayed, and the event loop
2431will not be interrupted regularly, which is more efficient (And good for
2432battery life on laptops).
2433
2434This affects not just the pure-perl event loop, but also other event loops
2435that have no signal handling on their own (e.g. Glib, Tk, Qt).
2436
2437Some event loops (POE, Event, Event::Lib) offer signal watchers natively,
2438and either employ their own workarounds (POE) or use AnyEvent's workaround
2439(using C<$AnyEvent::MAX_SIGNAL_LATENCY>). Installing L<Async::Interrupt>
2440does nothing for those backends.
2441
2442=item L<EV>
2443
2444This module isn't really "optional", as it is simply one of the backend
2445event loops that AnyEvent can use. However, it is simply the best event
2446loop available in terms of features, speed and stability: It supports
2447the AnyEvent API optimally, implements all the watcher types in XS, does
2448automatic timer adjustments even when no monotonic clock is available,
2449can take avdantage of advanced kernel interfaces such as C<epoll> and
2450C<kqueue>, and is the fastest backend I<by far>. You can even embed
2451L<Glib>/L<Gtk2> in it (or vice versa, see L<EV::Glib> and L<Glib::EV>).
2452
2453=item L<Guard>
2454
2455The guard module, when used, will be used to implement
2456C<AnyEvent::Util::guard>. This speeds up guards considerably (and uses a
2457lot less memory), but otherwise doesn't affect guard operation much. It is
2458purely used for performance.
2459
2460=item L<JSON> and L<JSON::XS>
2461
2462This module is required when you want to read or write JSON data via
2463L<AnyEvent::Handle>. It is also written in pure-perl, but can take
2464advantage of the ultra-high-speed L<JSON::XS> module when it is installed.
2465
2466In fact, L<AnyEvent::Handle> will use L<JSON::XS> by default if it is
2467installed.
2468
2469=item L<Net::SSLeay>
2470
2471Implementing TLS/SSL in Perl is certainly interesting, but not very
2472worthwhile: If this module is installed, then L<AnyEvent::Handle> (with
2473the help of L<AnyEvent::TLS>), gains the ability to do TLS/SSL.
2474
2475=item L<Time::HiRes>
2476
2477This module is part of perl since release 5.008. It will be used when the
2478chosen event library does not come with a timing source on it's own. The
2479pure-perl event loop (L<AnyEvent::Impl::Perl>) will additionally use it to
2480try to use a monotonic clock for timing stability.
2481
2482=back
2483
1235 2484
1236=head1 FORK 2485=head1 FORK
1237 2486
1238Most event libraries are not fork-safe. The ones who are usually are 2487Most event libraries are not fork-safe. The ones who are usually are
1239because they are so inefficient. Only L<EV> is fully fork-aware. 2488because they rely on inefficient but fork-safe C<select> or C<poll>
2489calls. Only L<EV> is fully fork-aware.
1240 2490
1241If you have to fork, you must either do so I<before> creating your first 2491If you have to fork, you must either do so I<before> creating your first
1242watcher OR you must not use AnyEvent at all in the child. 2492watcher OR you must not use AnyEvent at all in the child OR you must do
2493something completely out of the scope of AnyEvent.
1243 2494
1244 2495
1245=head1 SECURITY CONSIDERATIONS 2496=head1 SECURITY CONSIDERATIONS
1246 2497
1247AnyEvent can be forced to load any event model via 2498AnyEvent can be forced to load any event model via
1252specified in the variable. 2503specified in the variable.
1253 2504
1254You can make AnyEvent completely ignore this variable by deleting it 2505You can make AnyEvent completely ignore this variable by deleting it
1255before the first watcher gets created, e.g. with a C<BEGIN> block: 2506before the first watcher gets created, e.g. with a C<BEGIN> block:
1256 2507
1257 BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} } 2508 BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
1258 2509
1259 use AnyEvent; 2510 use AnyEvent;
2511
2512Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
2513be used to probe what backend is used and gain other information (which is
2514probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
2515$ENV{PERL_ANYEVENT_STRICT}.
2516
2517Note that AnyEvent will remove I<all> environment variables starting with
2518C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
2519enabled.
2520
2521
2522=head1 BUGS
2523
2524Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
2525to work around. If you suffer from memleaks, first upgrade to Perl 5.10
2526and check wether the leaks still show up. (Perl 5.10.0 has other annoying
2527memleaks, such as leaking on C<map> and C<grep> but it is usually not as
2528pronounced).
1260 2529
1261 2530
1262=head1 SEE ALSO 2531=head1 SEE ALSO
1263 2532
1264Event modules: L<Coro::EV>, L<EV>, L<EV::Glib>, L<Glib::EV>, 2533Utility functions: L<AnyEvent::Util>.
1265L<Coro::Event>, L<Event>, L<Glib::Event>, L<Glib>, L<Coro>, L<Tk>, 2534
2535Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
1266L<Event::Lib>, L<Qt>, L<POE>. 2536L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
1267 2537
1268Implementations: L<AnyEvent::Impl::CoroEV>, L<AnyEvent::Impl::EV>, 2538Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
2539L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
2540L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
1269L<AnyEvent::Impl::CoroEvent>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, 2541L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>, L<Anyevent::Impl::Irssi>.
1270L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, L<AnyEvent::Impl::EventLib>,
1271L<AnyEvent::Impl::Qt>, L<AnyEvent::Impl::POE>.
1272 2542
2543Non-blocking file handles, sockets, TCP clients and
2544servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>.
2545
2546Asynchronous DNS: L<AnyEvent::DNS>.
2547
2548Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>,
2549L<Coro::Event>,
2550
1273Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>. 2551Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::XMPP>,
2552L<AnyEvent::HTTP>.
1274 2553
1275 2554
1276=head1 AUTHOR 2555=head1 AUTHOR
1277 2556
1278 Marc Lehmann <schmorp@schmorp.de> 2557 Marc Lehmann <schmorp@schmorp.de>
1279 http://home.schmorp.de/ 2558 http://home.schmorp.de/
1280 2559
1281=cut 2560=cut
1282 2561
12831 25621
1284 2563

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines