ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.45 by root, Mon Apr 7 19:42:36 2008 UTC vs.
Revision 1.248 by root, Sat Jul 18 22:27:10 2009 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent - provide framework for multiple event loops	3	AnyEvent - provide framework for multiple event loops
4		4
5	Event, Coro, Glib, Tk, Perl - various supported event loops	5	EV, Event, Glib, Tk, Perl, Event::Lib, Qt and POE are various supported
		6	event loops.
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 wether 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
		42	This manpage is mainly a reference manual. If you are interested
		43	in a tutorial or some gentle introduction, have a look at the
		44	L<AnyEvent::Intro> manpage.
22		45
23	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)	46	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
24		47
25	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen	48	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
26	nowadays. So what is different about AnyEvent?	49	nowadays. So what is different about AnyEvent?
27		50
28	Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of	51	Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
29	policy> and AnyEvent is I<small and efficient>.	52	policy> and AnyEvent is I<small and efficient>.
30		53
31	First and foremost, I<AnyEvent is not an event model> itself, it only	54	First and foremost, I<AnyEvent is not an event model> itself, it only
32	interfaces to whatever event model the main program happens to use in a	55	interfaces to whatever event model the main program happens to use, in a
33	pragmatic way. For event models and certain classes of immortals alike,	56	pragmatic way. For event models and certain classes of immortals alike,
34	the statement "there can only be one" is a bitter reality, and AnyEvent	57	the statement "there can only be one" is a bitter reality: In general,
35	helps hiding the differences.	58	only one event loop can be active at the same time in a process. AnyEvent
		59	cannot change this, but it can hide the differences between those event
		60	loops.
36		61
37	The goal of AnyEvent is to offer module authors the ability to do event	62	The goal of AnyEvent is to offer module authors the ability to do event
38	programming (waiting for I/O or timer events) without subscribing to a	63	programming (waiting for I/O or timer events) without subscribing to a
39	religion, a way of living, and most importantly: without forcing your	64	religion, a way of living, and most importantly: without forcing your
40	module users into the same thing by forcing them to use the same event	65	module users into the same thing by forcing them to use the same event
41	model you use.	66	model you use.
42		67
43	For modules like POE or IO::Async (which is actually doing all I/O	68	For modules like POE or IO::Async (which is a total misnomer as it is
44	I<synchronously>...), using them in your module is like joining a	69	actually doing all I/O I<synchronously>...), using them in your module is
45	cult: After you joined, you are dependent on them and you cannot use	70	like joining a cult: After you joined, you are dependent on them and you
46	anything else, as it is simply incompatible to everything that isn't	71	cannot use anything else, as they are simply incompatible to everything
47	itself.	72	that isn't them. What's worse, all the potential users of your
		73	module are I<also> forced to use the same event loop you use.
48		74
49	AnyEvent + POE works fine. AnyEvent + Glib works fine. AnyEvent + Tk	75	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
50	works fine etc. etc. but none of these work together with the rest: POE	76	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
51	+ IO::Async? no go. Tk + Event? no go. If your module uses one of	77	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
52	those, every user of your module has to use it, too. If your module	78	your module uses one of those, every user of your module has to use it,
53	uses AnyEvent, it works transparently with all event models it supports	79	too. But if your module uses AnyEvent, it works transparently with all
54	(including stuff like POE and IO::Async).	80	event models it supports (including stuff like IO::Async, as long as those
		81	use one of the supported event loops. It is trivial to add new event loops
		82	to AnyEvent, too, so it is future-proof).
55		83
56	In addition of being free of having to use I<the one and only true event	84	In addition to being free of having to use I<the one and only true event
57	model>, AnyEvent also is free of bloat and policy: with POE or similar	85	model>, AnyEvent also is free of bloat and policy: with POE or similar
58	modules, you get an enourmous amount of code and strict rules you have	86	modules, you get an enormous amount of code and strict rules you have to
59	to follow. AnyEvent, on the other hand, is lean and to the point by only	87	follow. AnyEvent, on the other hand, is lean and up to the point, by only
60	offering the functionality that is useful, in as thin as a wrapper as	88	offering the functionality that is necessary, in as thin as a wrapper as
61	technically possible.	89	technically possible.
62		90
		91	Of course, AnyEvent comes with a big (and fully optional!) toolbox
		92	of useful functionality, such as an asynchronous DNS resolver, 100%
		93	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		94	such as Windows) and lots of real-world knowledge and workarounds for
		95	platform bugs and differences.
		96
63	Of course, if you want lots of policy (this can arguably be somewhat	97	Now, if you I<do want> lots of policy (this can arguably be somewhat
64	useful) and you want to force your users to the one and only event model	98	useful) and you want to force your users to use the one and only event
65	your module forces on them, you should I<not> use this module.	99	model, you should I<not> use this module.
66
67		100
68	=head1 DESCRIPTION	101	=head1 DESCRIPTION
69		102
70	L<AnyEvent> provides an identical interface to multiple event loops. This	103	L<AnyEvent> provides an identical interface to multiple event loops. This
71	allows module authors to utilise an event loop without forcing module	104	allows module authors to utilise an event loop without forcing module
72	users to use the same event loop (as only a single event loop can coexist	105	users to use the same event loop (as only a single event loop can coexist
73	peacefully at any one time).	106	peacefully at any one time).
74		107
75	The interface itself is vaguely similar but not identical to the Event	108	The interface itself is vaguely similar, but not identical to the L<Event>
76	module.	109	module.
77		110
78	On the first call of any method, the module tries to detect the currently	111	During the first call of any watcher-creation method, the module tries
79	loaded event loop by probing wether any of the following modules is	112	to detect the currently loaded event loop by probing whether one of the
80	loaded: L<Coro::Event>, L<Event>, L<Glib>, L<Tk>. The first one found is	113	following modules is already loaded: L<EV>,
81	used. If none is found, the module tries to load these modules in the	114	L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>,
82	order given. The first one that could be successfully loaded will be	115	L<POE>. The first one found is used. If none are found, the module tries
83	used. If still none could be found, AnyEvent will fall back to a pure-perl	116	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
84	event loop, which is also not very efficient.	117	adaptor should always succeed) in the order given. The first one that can
		118	be successfully loaded will be used. If, after this, still none could be
		119	found, AnyEvent will fall back to a pure-perl event loop, which is not
		120	very efficient, but should work everywhere.
85		121
86	Because AnyEvent first checks for modules that are already loaded, loading	122	Because AnyEvent first checks for modules that are already loaded, loading
87	an Event model explicitly before first using AnyEvent will likely make	123	an event model explicitly before first using AnyEvent will likely make
88	that model the default. For example:	124	that model the default. For example:
89		125
90	use Tk;	126	use Tk;
91	use AnyEvent;	127	use AnyEvent;
92		128
93	# .. AnyEvent will likely default to Tk	129	# .. AnyEvent will likely default to Tk
94		130
		131	The I<likely> means that, if any module loads another event model and
		132	starts using it, all bets are off. Maybe you should tell their authors to
		133	use AnyEvent so their modules work together with others seamlessly...
		134
95	The pure-perl implementation of AnyEvent is called	135	The pure-perl implementation of AnyEvent is called
96	C<AnyEvent::Impl::Perl>. Like other event modules you can load it	136	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
97	explicitly.	137	explicitly and enjoy the high availability of that event loop :)
98		138
99	=head1 WATCHERS	139	=head1 WATCHERS
100		140
101	AnyEvent has the central concept of a I<watcher>, which is an object that	141	AnyEvent has the central concept of a I<watcher>, which is an object that
102	stores relevant data for each kind of event you are waiting for, such as	142	stores relevant data for each kind of event you are waiting for, such as
103	the callback to call, the filehandle to watch, etc.	143	the callback to call, the file handle to watch, etc.
104		144
105	These watchers are normal Perl objects with normal Perl lifetime. After	145	These watchers are normal Perl objects with normal Perl lifetime. After
106	creating a watcher it will immediately "watch" for events and invoke	146	creating a watcher it will immediately "watch" for events and invoke the
		147	callback when the event occurs (of course, only when the event model
		148	is in control).
		149
		150	Note that B<callbacks must not permanently change global variables>
		151	potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
		152	callbacks must not C<die> >>. The former is good programming practise in
		153	Perl and the latter stems from the fact that exception handling differs
		154	widely between event loops.
		155
107	the callback. To disable the watcher you have to destroy it (e.g. by	156	To disable the watcher you have to destroy it (e.g. by setting the
108	setting the variable that stores it to C<undef> or otherwise deleting all	157	variable you store it in to C<undef> or otherwise deleting all references
109	references to it).	158	to it).
110		159
111	All watchers are created by calling a method on the C<AnyEvent> class.	160	All watchers are created by calling a method on the C<AnyEvent> class.
112		161
		162	Many watchers either are used with "recursion" (repeating timers for
		163	example), or need to refer to their watcher object in other ways.
		164
		165	An any way to achieve that is this pattern:
		166
		167	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
		168	# you can use $w here, for example to undef it
		169	undef $w;
		170	});
		171
		172	Note that C<my $w; $w => combination. This is necessary because in Perl,
		173	my variables are only visible after the statement in which they are
		174	declared.
		175
113	=head2 IO WATCHERS	176	=head2 I/O WATCHERS
114		177
115	You can create I/O watcher by calling the C<< AnyEvent->io >> method with	178	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
116	the following mandatory arguments:	179	with the following mandatory key-value pairs as arguments:
117		180
118	C<fh> the Perl I<filehandle> (not filedescriptor) to watch for	181	C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch
		182	for events (AnyEvent might or might not keep a reference to this file
		183	handle). Note that only file handles pointing to things for which
		184	non-blocking operation makes sense are allowed. This includes sockets,
		185	most character devices, pipes, fifos and so on, but not for example files
		186	or block devices.
		187
119	events. C<poll> must be a string that is either C<r> or C<w>, that creates	188	C<poll> must be a string that is either C<r> or C<w>, which creates a
120	a watcher waiting for "r"eadable or "w"ritable events. C<cb> the callback	189	watcher waiting for "r"eadable or "w"ritable events, respectively.
121	to invoke everytime the filehandle becomes ready.
122		190
123	Only one io watcher per C<fh> and C<poll> combination is allowed (i.e. on	191	C<cb> is the callback to invoke each time the file handle becomes ready.
124	a socket you can have one r + one w, not any more (limitation comes from
125	Tk - if you are sure you are not using Tk this limitation is gone).
126		192
127	Filehandles will be kept alive, so as long as the watcher exists, the	193	Although the callback might get passed parameters, their value and
128	filehandle exists, too.	194	presence is undefined and you cannot rely on them. Portable AnyEvent
		195	callbacks cannot use arguments passed to I/O watcher callbacks.
129		196
130	Example:	197	The I/O watcher might use the underlying file descriptor or a copy of it.
		198	You must not close a file handle as long as any watcher is active on the
		199	underlying file descriptor.
131		200
		201	Some event loops issue spurious readyness notifications, so you should
		202	always use non-blocking calls when reading/writing from/to your file
		203	handles.
		204
132	# wait for readability of STDIN, then read a line and disable the watcher	205	Example: wait for readability of STDIN, then read a line and disable the
		206	watcher.
		207
133	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	208	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
134	chomp (my $input = <STDIN>);	209	chomp (my $input = <STDIN>);
135	warn "read: $input\n";	210	warn "read: $input\n";
136	undef $w;	211	undef $w;
137	});	212	});
…		…
139	=head2 TIME WATCHERS	214	=head2 TIME WATCHERS
140		215
141	You can create a time watcher by calling the C<< AnyEvent->timer >>	216	You can create a time watcher by calling the C<< AnyEvent->timer >>
142	method with the following mandatory arguments:	217	method with the following mandatory arguments:
143		218
144	C<after> after how many seconds (fractions are supported) should the timer	219	C<after> specifies after how many seconds (fractional values are
145	activate. C<cb> the callback to invoke.	220	supported) the callback should be invoked. C<cb> is the callback to invoke
		221	in that case.
146		222
147	The timer callback will be invoked at most once: if you want a repeating	223	Although the callback might get passed parameters, their value and
148	timer you have to create a new watcher (this is a limitation by both Tk	224	presence is undefined and you cannot rely on them. Portable AnyEvent
149	and Glib).	225	callbacks cannot use arguments passed to time watcher callbacks.
150		226
151	Example:	227	The callback will normally be invoked once only. If you specify another
		228	parameter, C<interval>, as a strictly positive number (> 0), then the
		229	callback will be invoked regularly at that interval (in fractional
		230	seconds) after the first invocation. If C<interval> is specified with a
		231	false value, then it is treated as if it were missing.
152		232
		233	The callback will be rescheduled before invoking the callback, but no
		234	attempt is done to avoid timer drift in most backends, so the interval is
		235	only approximate.
		236
153	# fire an event after 7.7 seconds	237	Example: fire an event after 7.7 seconds.
		238
154	my $w = AnyEvent->timer (after => 7.7, cb => sub {	239	my $w = AnyEvent->timer (after => 7.7, cb => sub {
155	warn "timeout\n";	240	warn "timeout\n";
156	});	241	});
157		242
158	# to cancel the timer:	243	# to cancel the timer:
159	undef $w;	244	undef $w;
160		245
161	=head2 CONDITION WATCHERS	246	Example 2: fire an event after 0.5 seconds, then roughly every second.
162		247
163	Condition watchers can be created by calling the C<< AnyEvent->condvar >>	248	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
164	method without any arguments.	249	warn "timeout\n";
		250	};
165		251
166	A condition watcher watches for a condition - precisely that the C<<	252	=head3 TIMING ISSUES
167	->broadcast >> method has been called.
168		253
169	Note that condition watchers recurse into the event loop - if you have	254	There are two ways to handle timers: based on real time (relative, "fire
170	two watchers that call C<< ->wait >> in a round-robbin fashion, you	255	in 10 seconds") and based on wallclock time (absolute, "fire at 12
171	lose. Therefore, condition watchers are good to export to your caller, but	256	o'clock").
172	you should avoid making a blocking wait, at least in callbacks, as this
173	usually asks for trouble.
174		257
175	The watcher has only two methods:	258	While most event loops expect timers to specified in a relative way, they
		259	use absolute time internally. This makes a difference when your clock
		260	"jumps", for example, when ntp decides to set your clock backwards from
		261	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		262	fire "after" a second might actually take six years to finally fire.
		263
		264	AnyEvent cannot compensate for this. The only event loop that is conscious
		265	about these issues is L<EV>, which offers both relative (ev_timer, based
		266	on true relative time) and absolute (ev_periodic, based on wallclock time)
		267	timers.
		268
		269	AnyEvent always prefers relative timers, if available, matching the
		270	AnyEvent API.
		271
		272	AnyEvent has two additional methods that return the "current time":
176		273
177	=over 4	274	=over 4
178		275
179	=item $cv->wait	276	=item AnyEvent->time
180		277
181	Wait (blocking if necessary) until the C<< ->broadcast >> method has been	278	This returns the "current wallclock time" as a fractional number of
182	called on c<$cv>, while servicing other watchers normally.	279	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		280	return, and the result is guaranteed to be compatible with those).
183		281
184	Not all event models support a blocking wait - some die in that case, so	282	It progresses independently of any event loop processing, i.e. each call
185	if you are using this from a module, never require a blocking wait, but	283	will check the system clock, which usually gets updated frequently.
186	let the caller decide wether the call will block or not (for example,
187	by coupling condition variables with some kind of request results and
188	supporting callbacks so the caller knows that getting the result will not
189	block, while still suppporting blockign waits if the caller so desires).
190		284
191	You can only wait once on a condition - additional calls will return	285	=item AnyEvent->now
192	immediately.
193		286
194	=item $cv->broadcast	287	This also returns the "current wallclock time", but unlike C<time>, above,
		288	this value might change only once per event loop iteration, depending on
		289	the event loop (most return the same time as C<time>, above). This is the
		290	time that AnyEvent's timers get scheduled against.
195		291
196	Flag the condition as ready - a running C<< ->wait >> and all further	292	I<In almost all cases (in all cases if you don't care), this is the
197	calls to C<wait> will return after this method has been called. If nobody	293	function to call when you want to know the current time.>
198	is waiting the broadcast will be remembered..
199		294
200	Example:	295	This function is also often faster then C<< AnyEvent->time >>, and
		296	thus the preferred method if you want some timestamp (for example,
		297	L<AnyEvent::Handle> uses this to update it's activity timeouts).
		298
		299	The rest of this section is only of relevance if you try to be very exact
		300	with your timing, you can skip it without bad conscience.
		301
		302	For a practical example of when these times differ, consider L<Event::Lib>
		303	and L<EV> and the following set-up:
		304
		305	The event loop is running and has just invoked one of your callback at
		306	time=500 (assume no other callbacks delay processing). In your callback,
		307	you wait a second by executing C<sleep 1> (blocking the process for a
		308	second) and then (at time=501) you create a relative timer that fires
		309	after three seconds.
		310
		311	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		312	both return C<501>, because that is the current time, and the timer will
		313	be scheduled to fire at time=504 (C<501> + C<3>).
		314
		315	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		316	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		317	last event processing phase started. With L<EV>, your timer gets scheduled
		318	to run at time=503 (C<500> + C<3>).
		319
		320	In one sense, L<Event::Lib> is more exact, as it uses the current time
		321	regardless of any delays introduced by event processing. However, most
		322	callbacks do not expect large delays in processing, so this causes a
		323	higher drift (and a lot more system calls to get the current time).
		324
		325	In another sense, L<EV> is more exact, as your timer will be scheduled at
		326	the same time, regardless of how long event processing actually took.
		327
		328	In either case, if you care (and in most cases, you don't), then you
		329	can get whatever behaviour you want with any event loop, by taking the
		330	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		331	account.
		332
		333	=item AnyEvent->now_update
		334
		335	Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
		336	the current time for each loop iteration (see the discussion of L<<
		337	AnyEvent->now >>, above).
		338
		339	When a callback runs for a long time (or when the process sleeps), then
		340	this "current" time will differ substantially from the real time, which
		341	might affect timers and time-outs.
		342
		343	When this is the case, you can call this method, which will update the
		344	event loop's idea of "current time".
		345
		346	Note that updating the time I<might> cause some events to be handled.
		347
		348	=back
		349
		350	=head2 SIGNAL WATCHERS
		351
		352	You can watch for signals using a signal watcher, C<signal> is the signal
		353	I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
		354	callback to be invoked whenever a signal occurs.
		355
		356	Although the callback might get passed parameters, their value and
		357	presence is undefined and you cannot rely on them. Portable AnyEvent
		358	callbacks cannot use arguments passed to signal watcher callbacks.
		359
		360	Multiple signal occurrences can be clumped together into one callback
		361	invocation, and callback invocation will be synchronous. Synchronous means
		362	that it might take a while until the signal gets handled by the process,
		363	but it is guaranteed not to interrupt any other callbacks.
		364
		365	The main advantage of using these watchers is that you can share a signal
		366	between multiple watchers, and AnyEvent will ensure that signals will not
		367	interrupt your program at bad times.
		368
		369	This watcher might use C<%SIG> (depending on the event loop used),
		370	so programs overwriting those signals directly will likely not work
		371	correctly.
		372
		373	Example: exit on SIGINT
		374
		375	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		376
		377	=head3 Signal Races, Delays and Workarounds
		378
		379	Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
		380	callbacks to signals in a generic way, which is a pity, as you cannot do
		381	race-free signal handling in perl. AnyEvent will try to do it's best, but
		382	in some cases, signals will be delayed. The maximum time a signal might
		383	be delayed is specified in C<$AnyEvent::MAX_SIGNAL_LATENCY> (default: 10
		384	seconds). This variable can be changed only before the first signal
		385	watcher is created, and should be left alone otherwise. Higher values
		386	will cause fewer spurious wake-ups, which is better for power and CPU
		387	saving. All these problems can be avoided by installing the optional
		388	L<Async::Interrupt> module. This will not work with inherently broken
		389	event loops such as L<Event> or L<Event::Lib> (and not with L<POE>
		390	currently, as POE does it's own workaround with one-second latency). With
		391	those, you just have to suffer the delays.
		392
		393	=head2 CHILD PROCESS WATCHERS
		394
		395	You can also watch on a child process exit and catch its exit status.
		396
		397	The child process is specified by the C<pid> argument (if set to C<0>, it
		398	watches for any child process exit). The watcher will triggered only when
		399	the child process has finished and an exit status is available, not on
		400	any trace events (stopped/continued).
		401
		402	The callback will be called with the pid and exit status (as returned by
		403	waitpid), so unlike other watcher types, you I<can> rely on child watcher
		404	callback arguments.
		405
		406	This watcher type works by installing a signal handler for C<SIGCHLD>,
		407	and since it cannot be shared, nothing else should use SIGCHLD or reap
		408	random child processes (waiting for specific child processes, e.g. inside
		409	C<system>, is just fine).
		410
		411	There is a slight catch to child watchers, however: you usually start them
		412	I<after> the child process was created, and this means the process could
		413	have exited already (and no SIGCHLD will be sent anymore).
		414
		415	Not all event models handle this correctly (neither POE nor IO::Async do,
		416	see their AnyEvent::Impl manpages for details), but even for event models
		417	that I<do> handle this correctly, they usually need to be loaded before
		418	the process exits (i.e. before you fork in the first place). AnyEvent's
		419	pure perl event loop handles all cases correctly regardless of when you
		420	start the watcher.
		421
		422	This means you cannot create a child watcher as the very first
		423	thing in an AnyEvent program, you I<have> to create at least one
		424	watcher before you C<fork> the child (alternatively, you can call
		425	C<AnyEvent::detect>).
		426
		427	As most event loops do not support waiting for child events, they will be
		428	emulated by AnyEvent in most cases, in which the latency and race problems
		429	mentioned in the description of signal watchers apply.
		430
		431	Example: fork a process and wait for it
		432
		433	my $done = AnyEvent->condvar;
		434
		435	my $pid = fork or exit 5;
		436
		437	my $w = AnyEvent->child (
		438	pid => $pid,
		439	cb => sub {
		440	my ($pid, $status) = @_;
		441	warn "pid $pid exited with status $status";
		442	$done->send;
		443	},
		444	);
		445
		446	# do something else, then wait for process exit
		447	$done->recv;
		448
		449	=head2 IDLE WATCHERS
		450
		451	Sometimes there is a need to do something, but it is not so important
		452	to do it instantly, but only when there is nothing better to do. This
		453	"nothing better to do" is usually defined to be "no other events need
		454	attention by the event loop".
		455
		456	Idle watchers ideally get invoked when the event loop has nothing
		457	better to do, just before it would block the process to wait for new
		458	events. Instead of blocking, the idle watcher is invoked.
		459
		460	Most event loops unfortunately do not really support idle watchers (only
		461	EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
		462	will simply call the callback "from time to time".
		463
		464	Example: read lines from STDIN, but only process them when the
		465	program is otherwise idle:
		466
		467	my @lines; # read data
		468	my $idle_w;
		469	my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		470	push @lines, scalar <STDIN>;
		471
		472	# start an idle watcher, if not already done
		473	$idle_w ||= AnyEvent->idle (cb => sub {
		474	# handle only one line, when there are lines left
		475	if (my $line = shift @lines) {
		476	print "handled when idle: $line";
		477	} else {
		478	# otherwise disable the idle watcher again
		479	undef $idle_w;
		480	}
		481	});
		482	});
		483
		484	=head2 CONDITION VARIABLES
		485
		486	If you are familiar with some event loops you will know that all of them
		487	require you to run some blocking "loop", "run" or similar function that
		488	will actively watch for new events and call your callbacks.
		489
		490	AnyEvent is slightly different: it expects somebody else to run the event
		491	loop and will only block when necessary (usually when told by the user).
		492
		493	The instrument to do that is called a "condition variable", so called
		494	because they represent a condition that must become true.
		495
		496	Now is probably a good time to look at the examples further below.
		497
		498	Condition variables can be created by calling the C<< AnyEvent->condvar
		499	>> method, usually without arguments. The only argument pair allowed is
		500	C<cb>, which specifies a callback to be called when the condition variable
		501	becomes true, with the condition variable as the first argument (but not
		502	the results).
		503
		504	After creation, the condition variable is "false" until it becomes "true"
		505	by calling the C<send> method (or calling the condition variable as if it
		506	were a callback, read about the caveats in the description for the C<<
		507	->send >> method).
		508
		509	Condition variables are similar to callbacks, except that you can
		510	optionally wait for them. They can also be called merge points - points
		511	in time where multiple outstanding events have been processed. And yet
		512	another way to call them is transactions - each condition variable can be
		513	used to represent a transaction, which finishes at some point and delivers
		514	a result.
		515
		516	Condition variables are very useful to signal that something has finished,
		517	for example, if you write a module that does asynchronous http requests,
		518	then a condition variable would be the ideal candidate to signal the
		519	availability of results. The user can either act when the callback is
		520	called or can synchronously C<< ->recv >> for the results.
		521
		522	You can also use them to simulate traditional event loops - for example,
		523	you can block your main program until an event occurs - for example, you
		524	could C<< ->recv >> in your main program until the user clicks the Quit
		525	button of your app, which would C<< ->send >> the "quit" event.
		526
		527	Note that condition variables recurse into the event loop - if you have
		528	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		529	lose. Therefore, condition variables are good to export to your caller, but
		530	you should avoid making a blocking wait yourself, at least in callbacks,
		531	as this asks for trouble.
		532
		533	Condition variables are represented by hash refs in perl, and the keys
		534	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		535	easy (it is often useful to build your own transaction class on top of
		536	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		537	it's C<new> method in your own C<new> method.
		538
		539	There are two "sides" to a condition variable - the "producer side" which
		540	eventually calls C<< -> send >>, and the "consumer side", which waits
		541	for the send to occur.
		542
		543	Example: wait for a timer.
201		544
202	# wait till the result is ready	545	# wait till the result is ready
203	my $result_ready = AnyEvent->condvar;	546	my $result_ready = AnyEvent->condvar;
204		547
205	# do something such as adding a timer	548	# do something such as adding a timer
206	# or socket watcher the calls $result_ready->broadcast	549	# or socket watcher the calls $result_ready->send
207	# when the "result" is ready.	550	# when the "result" is ready.
		551	# in this case, we simply use a timer:
		552	my $w = AnyEvent->timer (
		553	after => 1,
		554	cb => sub { $result_ready->send },
		555	);
208		556
		557	# this "blocks" (while handling events) till the callback
		558	# calls -<send
209	$result_ready->wait;	559	$result_ready->recv;
		560
		561	Example: wait for a timer, but take advantage of the fact that condition
		562	variables are also callable directly.
		563
		564	my $done = AnyEvent->condvar;
		565	my $delay = AnyEvent->timer (after => 5, cb => $done);
		566	$done->recv;
		567
		568	Example: Imagine an API that returns a condvar and doesn't support
		569	callbacks. This is how you make a synchronous call, for example from
		570	the main program:
		571
		572	use AnyEvent::CouchDB;
		573
		574	...
		575
		576	my @info = $couchdb->info->recv;
		577
		578	And this is how you would just set a callback to be called whenever the
		579	results are available:
		580
		581	$couchdb->info->cb (sub {
		582	my @info = $_[0]->recv;
		583	});
		584
		585	=head3 METHODS FOR PRODUCERS
		586
		587	These methods should only be used by the producing side, i.e. the
		588	code/module that eventually sends the signal. Note that it is also
		589	the producer side which creates the condvar in most cases, but it isn't
		590	uncommon for the consumer to create it as well.
		591
		592	=over 4
		593
		594	=item $cv->send (...)
		595
		596	Flag the condition as ready - a running C<< ->recv >> and all further
		597	calls to C<recv> will (eventually) return after this method has been
		598	called. If nobody is waiting the send will be remembered.
		599
		600	If a callback has been set on the condition variable, it is called
		601	immediately from within send.
		602
		603	Any arguments passed to the C<send> call will be returned by all
		604	future C<< ->recv >> calls.
		605
		606	Condition variables are overloaded so one can call them directly (as if
		607	they were a code reference). Calling them directly is the same as calling
		608	C<send>.
		609
		610	=item $cv->croak ($error)
		611
		612	Similar to send, but causes all call's to C<< ->recv >> to invoke
		613	C<Carp::croak> with the given error message/object/scalar.
		614
		615	This can be used to signal any errors to the condition variable
		616	user/consumer. Doing it this way instead of calling C<croak> directly
		617	delays the error detetcion, but has the overwhelmign advantage that it
		618	diagnoses the error at the place where the result is expected, and not
		619	deep in some event clalback without connection to the actual code causing
		620	the problem.
		621
		622	=item $cv->begin ([group callback])
		623
		624	=item $cv->end
		625
		626	These two methods can be used to combine many transactions/events into
		627	one. For example, a function that pings many hosts in parallel might want
		628	to use a condition variable for the whole process.
		629
		630	Every call to C<< ->begin >> will increment a counter, and every call to
		631	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		632	>>, the (last) callback passed to C<begin> will be executed. That callback
		633	is I<supposed> to call C<< ->send >>, but that is not required. If no
		634	callback was set, C<send> will be called without any arguments.
		635
		636	You can think of C<< $cv->send >> giving you an OR condition (one call
		637	sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
		638	condition (all C<begin> calls must be C<end>'ed before the condvar sends).
		639
		640	Let's start with a simple example: you have two I/O watchers (for example,
		641	STDOUT and STDERR for a program), and you want to wait for both streams to
		642	close before activating a condvar:
		643
		644	my $cv = AnyEvent->condvar;
		645
		646	$cv->begin; # first watcher
		647	my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
		648	defined sysread $fh1, my $buf, 4096
		649	or $cv->end;
		650	});
		651
		652	$cv->begin; # second watcher
		653	my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
		654	defined sysread $fh2, my $buf, 4096
		655	or $cv->end;
		656	});
		657
		658	$cv->recv;
		659
		660	This works because for every event source (EOF on file handle), there is
		661	one call to C<begin>, so the condvar waits for all calls to C<end> before
		662	sending.
		663
		664	The ping example mentioned above is slightly more complicated, as the
		665	there are results to be passwd back, and the number of tasks that are
		666	begung can potentially be zero:
		667
		668	my $cv = AnyEvent->condvar;
		669
		670	my %result;
		671	$cv->begin (sub { $cv->send (\%result) });
		672
		673	for my $host (@list_of_hosts) {
		674	$cv->begin;
		675	ping_host_then_call_callback $host, sub {
		676	$result{$host} = ...;
		677	$cv->end;
		678	};
		679	}
		680
		681	$cv->end;
		682
		683	This code fragment supposedly pings a number of hosts and calls
		684	C<send> after results for all then have have been gathered - in any
		685	order. To achieve this, the code issues a call to C<begin> when it starts
		686	each ping request and calls C<end> when it has received some result for
		687	it. Since C<begin> and C<end> only maintain a counter, the order in which
		688	results arrive is not relevant.
		689
		690	There is an additional bracketing call to C<begin> and C<end> outside the
		691	loop, which serves two important purposes: first, it sets the callback
		692	to be called once the counter reaches C<0>, and second, it ensures that
		693	C<send> is called even when C<no> hosts are being pinged (the loop
		694	doesn't execute once).
		695
		696	This is the general pattern when you "fan out" into multiple (but
		697	potentially none) subrequests: use an outer C<begin>/C<end> pair to set
		698	the callback and ensure C<end> is called at least once, and then, for each
		699	subrequest you start, call C<begin> and for each subrequest you finish,
		700	call C<end>.
210		701
211	=back	702	=back
212		703
213	=head2 SIGNAL WATCHERS	704	=head3 METHODS FOR CONSUMERS
214		705
215	You can listen for signals using a signal watcher, C<signal> is the signal	706	These methods should only be used by the consuming side, i.e. the
216	I<name> without any C<SIG> prefix. Multiple signals events can be clumped	707	code awaits the condition.
217	together into one callback invocation, and callback invocation might or
218	might not be asynchronous.
219
220	These watchers might use C<%SIG>, so programs overwriting those signals
221	directly will likely not work correctly.
222
223	Example: exit on SIGINT
224
225	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
226
227	=head2 CHILD PROCESS WATCHERS
228
229	You can also listen for the status of a child process specified by the
230	C<pid> argument (or any child if the pid argument is 0). The watcher will
231	trigger as often as status change for the child are received. This works
232	by installing a signal handler for C<SIGCHLD>. The callback will be called with
233	the pid and exit status (as returned by waitpid).
234
235	Example: wait for pid 1333
236
237	my $w = AnyEvent->child (pid => 1333, cb => sub { warn "exit status $?" });
238
239	=head1 GLOBALS
240		708
241	=over 4	709	=over 4
242		710
		711	=item $cv->recv
		712
		713	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		714	>> methods have been called on c<$cv>, while servicing other watchers
		715	normally.
		716
		717	You can only wait once on a condition - additional calls are valid but
		718	will return immediately.
		719
		720	If an error condition has been set by calling C<< ->croak >>, then this
		721	function will call C<croak>.
		722
		723	In list context, all parameters passed to C<send> will be returned,
		724	in scalar context only the first one will be returned.
		725
		726	Note that doing a blocking wait in a callback is not supported by any
		727	event loop, that is, recursive invocation of a blocking C<< ->recv
		728	>> is not allowed, and the C<recv> call will C<croak> if such a
		729	condition is detected. This condition can be slightly loosened by using
		730	L<Coro::AnyEvent>, which allows you to do a blocking C<< ->recv >> from
		731	any thread that doesn't run the event loop itself.
		732
		733	Not all event models support a blocking wait - some die in that case
		734	(programs might want to do that to stay interactive), so I<if you are
		735	using this from a module, never require a blocking wait>. Instead, let the
		736	caller decide whether the call will block or not (for example, by coupling
		737	condition variables with some kind of request results and supporting
		738	callbacks so the caller knows that getting the result will not block,
		739	while still supporting blocking waits if the caller so desires).
		740
		741	You can ensure that C<< -recv >> never blocks by setting a callback and
		742	only calling C<< ->recv >> from within that callback (or at a later
		743	time). This will work even when the event loop does not support blocking
		744	waits otherwise.
		745
		746	=item $bool = $cv->ready
		747
		748	Returns true when the condition is "true", i.e. whether C<send> or
		749	C<croak> have been called.
		750
		751	=item $cb = $cv->cb ($cb->($cv))
		752
		753	This is a mutator function that returns the callback set and optionally
		754	replaces it before doing so.
		755
		756	The callback will be called when the condition becomes "true", i.e. when
		757	C<send> or C<croak> are called, with the only argument being the condition
		758	variable itself. Calling C<recv> inside the callback or at any later time
		759	is guaranteed not to block.
		760
		761	=back
		762
		763	=head1 SUPPORTED EVENT LOOPS/BACKENDS
		764
		765	The available backend classes are (every class has its own manpage):
		766
		767	=over 4
		768
		769	=item Backends that are autoprobed when no other event loop can be found.
		770
		771	EV is the preferred backend when no other event loop seems to be in
		772	use. If EV is not installed, then AnyEvent will try Event, and, failing
		773	that, will fall back to its own pure-perl implementation, which is
		774	available everywhere as it comes with AnyEvent itself.
		775
		776	AnyEvent::Impl::EV based on EV (interface to libev, best choice).
		777	AnyEvent::Impl::Event based on Event, very stable, few glitches.
		778	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		779
		780	=item Backends that are transparently being picked up when they are used.
		781
		782	These will be used when they are currently loaded when the first watcher
		783	is created, in which case it is assumed that the application is using
		784	them. This means that AnyEvent will automatically pick the right backend
		785	when the main program loads an event module before anything starts to
		786	create watchers. Nothing special needs to be done by the main program.
		787
		788	AnyEvent::Impl::Glib based on Glib, slow but very stable.
		789	AnyEvent::Impl::Tk based on Tk, very broken.
		790	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		791	AnyEvent::Impl::POE based on POE, very slow, some limitations.
		792
		793	=item Backends with special needs.
		794
		795	Qt requires the Qt::Application to be instantiated first, but will
		796	otherwise be picked up automatically. As long as the main program
		797	instantiates the application before any AnyEvent watchers are created,
		798	everything should just work.
		799
		800	AnyEvent::Impl::Qt based on Qt.
		801
		802	Support for IO::Async can only be partial, as it is too broken and
		803	architecturally limited to even support the AnyEvent API. It also
		804	is the only event loop that needs the loop to be set explicitly, so
		805	it can only be used by a main program knowing about AnyEvent. See
		806	L<AnyEvent::Impl::Async> for the gory details.
		807
		808	AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed.
		809
		810	=item Event loops that are indirectly supported via other backends.
		811
		812	Some event loops can be supported via other modules:
		813
		814	There is no direct support for WxWidgets (L<Wx>) or L<Prima>.
		815
		816	B<WxWidgets> has no support for watching file handles. However, you can
		817	use WxWidgets through the POE adaptor, as POE has a Wx backend that simply
		818	polls 20 times per second, which was considered to be too horrible to even
		819	consider for AnyEvent.
		820
		821	B<Prima> is not supported as nobody seems to be using it, but it has a POE
		822	backend, so it can be supported through POE.
		823
		824	AnyEvent knows about both L<Prima> and L<Wx>, however, and will try to
		825	load L<POE> when detecting them, in the hope that POE will pick them up,
		826	in which case everything will be automatic.
		827
		828	=back
		829
		830	=head1 GLOBAL VARIABLES AND FUNCTIONS
		831
		832	These are not normally required to use AnyEvent, but can be useful to
		833	write AnyEvent extension modules.
		834
		835	=over 4
		836
243	=item $AnyEvent::MODEL	837	=item $AnyEvent::MODEL
244		838
245	Contains C<undef> until the first watcher is being created. Then it	839	Contains C<undef> until the first watcher is being created, before the
		840	backend has been autodetected.
		841
246	contains the event model that is being used, which is the name of the	842	Afterwards it contains the event model that is being used, which is the
247	Perl class implementing the model. This class is usually one of the	843	name of the Perl class implementing the model. This class is usually one
248	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case	844	of the C<AnyEvent::Impl:xxx> modules, but can be any other class in the
249	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).	845	case AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode> it
250		846	will be C<urxvt::anyevent>).
251	The known classes so far are:
252
253	AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
254	AnyEvent::Impl::EV based on EV (an interface to libev, also best choice).
255	AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
256	AnyEvent::Impl::Event based on Event, also second best choice :)
257	AnyEvent::Impl::Glib based on Glib, second-best choice.
258	AnyEvent::Impl::Tk based on Tk, very bad choice.
259	AnyEvent::Impl::Perl pure-perl implementation, inefficient.
260		847
261	=item AnyEvent::detect	848	=item AnyEvent::detect
262		849
263	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model if	850	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
264	necessary. You should only call this function right before you would have	851	if necessary. You should only call this function right before you would
265	created an AnyEvent watcher anyway, that is, very late at runtime.	852	have created an AnyEvent watcher anyway, that is, as late as possible at
		853	runtime, and not e.g. while initialising of your module.
		854
		855	If you need to do some initialisation before AnyEvent watchers are
		856	created, use C<post_detect>.
		857
		858	=item $guard = AnyEvent::post_detect { BLOCK }
		859
		860	Arranges for the code block to be executed as soon as the event model is
		861	autodetected (or immediately if this has already happened).
		862
		863	The block will be executed I<after> the actual backend has been detected
		864	(C<$AnyEvent::MODEL> is set), but I<before> any watchers have been
		865	created, so it is possible to e.g. patch C<@AnyEvent::ISA> or do
		866	other initialisations - see the sources of L<AnyEvent::Strict> or
		867	L<AnyEvent::AIO> to see how this is used.
		868
		869	The most common usage is to create some global watchers, without forcing
		870	event module detection too early, for example, L<AnyEvent::AIO> creates
		871	and installs the global L<IO::AIO> watcher in a C<post_detect> block to
		872	avoid autodetecting the event module at load time.
		873
		874	If called in scalar or list context, then it creates and returns an object
		875	that automatically removes the callback again when it is destroyed. See
		876	L<Coro::BDB> for a case where this is useful.
		877
		878	=item @AnyEvent::post_detect
		879
		880	If there are any code references in this array (you can C<push> to it
		881	before or after loading AnyEvent), then they will called directly after
		882	the event loop has been chosen.
		883
		884	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		885	if it is defined then the event loop has already been detected, and the
		886	array will be ignored.
		887
		888	Best use C<AnyEvent::post_detect { BLOCK }> when your application allows
		889	it,as it takes care of these details.
		890
		891	This variable is mainly useful for modules that can do something useful
		892	when AnyEvent is used and thus want to know when it is initialised, but do
		893	not need to even load it by default. This array provides the means to hook
		894	into AnyEvent passively, without loading it.
266		895
267	=back	896	=back
268		897
269	=head1 WHAT TO DO IN A MODULE	898	=head1 WHAT TO DO IN A MODULE
270		899
271	As a module author, you should "use AnyEvent" and call AnyEvent methods	900	As a module author, you should C<use AnyEvent> and call AnyEvent methods
272	freely, but you should not load a specific event module or rely on it.	901	freely, but you should not load a specific event module or rely on it.
273		902
274	Be careful when you create watchers in the module body - Anyevent will	903	Be careful when you create watchers in the module body - AnyEvent will
275	decide which event module to use as soon as the first method is called, so	904	decide which event module to use as soon as the first method is called, so
276	by calling AnyEvent in your module body you force the user of your module	905	by calling AnyEvent in your module body you force the user of your module
277	to load the event module first.	906	to load the event module first.
278		907
		908	Never call C<< ->recv >> on a condition variable unless you I<know> that
		909	the C<< ->send >> method has been called on it already. This is
		910	because it will stall the whole program, and the whole point of using
		911	events is to stay interactive.
		912
		913	It is fine, however, to call C<< ->recv >> when the user of your module
		914	requests it (i.e. if you create a http request object ad have a method
		915	called C<results> that returns the results, it should call C<< ->recv >>
		916	freely, as the user of your module knows what she is doing. always).
		917
279	=head1 WHAT TO DO IN THE MAIN PROGRAM	918	=head1 WHAT TO DO IN THE MAIN PROGRAM
280		919
281	There will always be a single main program - the only place that should	920	There will always be a single main program - the only place that should
282	dictate which event model to use.	921	dictate which event model to use.
283		922
284	If it doesn't care, it can just "use AnyEvent" and use it itself, or not	923	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
285	do anything special and let AnyEvent decide which implementation to chose.	924	do anything special (it does not need to be event-based) and let AnyEvent
		925	decide which implementation to chose if some module relies on it.
286		926
287	If the main program relies on a specific event model (for example, in Gtk2	927	If the main program relies on a specific event model - for example, in
288	programs you have to rely on either Glib or Glib::Event), you should load	928	Gtk2 programs you have to rely on the Glib module - you should load the
289	it before loading AnyEvent or any module that uses it, generally, as early	929	event module before loading AnyEvent or any module that uses it: generally
290	as possible. The reason is that modules might create watchers when they	930	speaking, you should load it as early as possible. The reason is that
291	are loaded, and AnyEvent will decide on the event model to use as soon as	931	modules might create watchers when they are loaded, and AnyEvent will
292	it creates watchers, and it might chose the wrong one unless you load the	932	decide on the event model to use as soon as it creates watchers, and it
293	correct one yourself.	933	might chose the wrong one unless you load the correct one yourself.
294		934
295	You can chose to use a rather inefficient pure-perl implementation by	935	You can chose to use a pure-perl implementation by loading the
296	loading the C<AnyEvent::Impl::Perl> module, but letting AnyEvent chose is	936	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
297	generally better.	937	everywhere, but letting AnyEvent chose the model is generally better.
		938
		939	=head2 MAINLOOP EMULATION
		940
		941	Sometimes (often for short test scripts, or even standalone programs who
		942	only want to use AnyEvent), you do not want to run a specific event loop.
		943
		944	In that case, you can use a condition variable like this:
		945
		946	AnyEvent->condvar->recv;
		947
		948	This has the effect of entering the event loop and looping forever.
		949
		950	Note that usually your program has some exit condition, in which case
		951	it is better to use the "traditional" approach of storing a condition
		952	variable somewhere, waiting for it, and sending it when the program should
		953	exit cleanly.
		954
		955
		956	=head1 OTHER MODULES
		957
		958	The following is a non-exhaustive list of additional modules that use
		959	AnyEvent as a client and can therefore be mixed easily with other AnyEvent
		960	modules and other event loops in the same program. Some of the modules
		961	come with AnyEvent, most are available via CPAN.
		962
		963	=over 4
		964
		965	=item L<AnyEvent::Util>
		966
		967	Contains various utility functions that replace often-used but blocking
		968	functions such as C<inet_aton> by event-/callback-based versions.
		969
		970	=item L<AnyEvent::Socket>
		971
		972	Provides various utility functions for (internet protocol) sockets,
		973	addresses and name resolution. Also functions to create non-blocking tcp
		974	connections or tcp servers, with IPv6 and SRV record support and more.
		975
		976	=item L<AnyEvent::Handle>
		977
		978	Provide read and write buffers, manages watchers for reads and writes,
		979	supports raw and formatted I/O, I/O queued and fully transparent and
		980	non-blocking SSL/TLS (via L<AnyEvent::TLS>.
		981
		982	=item L<AnyEvent::DNS>
		983
		984	Provides rich asynchronous DNS resolver capabilities.
		985
		986	=item L<AnyEvent::HTTP>
		987
		988	A simple-to-use HTTP library that is capable of making a lot of concurrent
		989	HTTP requests.
		990
		991	=item L<AnyEvent::HTTPD>
		992
		993	Provides a simple web application server framework.
		994
		995	=item L<AnyEvent::FastPing>
		996
		997	The fastest ping in the west.
		998
		999	=item L<AnyEvent::DBI>
		1000
		1001	Executes L<DBI> requests asynchronously in a proxy process.
		1002
		1003	=item L<AnyEvent::AIO>
		1004
		1005	Truly asynchronous I/O, should be in the toolbox of every event
		1006	programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
		1007	together.
		1008
		1009	=item L<AnyEvent::BDB>
		1010
		1011	Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
		1012	L<BDB> and AnyEvent together.
		1013
		1014	=item L<AnyEvent::GPSD>
		1015
		1016	A non-blocking interface to gpsd, a daemon delivering GPS information.
		1017
		1018	=item L<AnyEvent::IRC>
		1019
		1020	AnyEvent based IRC client module family (replacing the older Net::IRC3).
		1021
		1022	=item L<AnyEvent::XMPP>
		1023
		1024	AnyEvent based XMPP (Jabber protocol) module family (replacing the older
		1025	Net::XMPP2>.
		1026
		1027	=item L<AnyEvent::IGS>
		1028
		1029	A non-blocking interface to the Internet Go Server protocol (used by
		1030	L<App::IGS>).
		1031
		1032	=item L<Net::FCP>
		1033
		1034	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		1035	of AnyEvent.
		1036
		1037	=item L<Event::ExecFlow>
		1038
		1039	High level API for event-based execution flow control.
		1040
		1041	=item L<Coro>
		1042
		1043	Has special support for AnyEvent via L<Coro::AnyEvent>.
		1044
		1045	=back
298		1046
299	=cut	1047	=cut
300		1048
301	package AnyEvent;	1049	package AnyEvent;
302		1050
		1051	# basically a tuned-down version of common::sense
		1052	sub common_sense {
303	no warnings;	1053	# no warnings
304	use strict;	1054	${^WARNING_BITS} ^= ${^WARNING_BITS};
		1055	# use strict vars subs
		1056	$^H |= 0x00000600;
		1057	}
305		1058
		1059	BEGIN { AnyEvent::common_sense }
		1060
306	use Carp;	1061	use Carp ();
307		1062
308	our $VERSION = '3.0';	1063	our $VERSION = 4.85;
309	our $MODEL;	1064	our $MODEL;
310		1065
311	our $AUTOLOAD;	1066	our $AUTOLOAD;
312	our @ISA;	1067	our @ISA;
313		1068
314	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
315
316	our @REGISTRY;	1069	our @REGISTRY;
317		1070
		1071	our $WIN32;
		1072
		1073	our $VERBOSE;
		1074
		1075	BEGIN {
		1076	eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }";
		1077	eval "sub TAINT(){ " . (${^TAINT}*1) . " }";
		1078
		1079	delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
		1080	if ${^TAINT};
		1081
		1082	$VERBOSE = $ENV{PERL_ANYEVENT_VERBOSE}*1;
		1083
		1084	}
		1085
		1086	our $MAX_SIGNAL_LATENCY = 10;
		1087
		1088	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		1089
		1090	{
		1091	my $idx;
		1092	$PROTOCOL{$_} = ++$idx
		1093	for reverse split /\s*,\s*/,
		1094	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		1095	}
		1096
318	my @models = (	1097	my @models = (
319	[Coro::EV:: => AnyEvent::Impl::CoroEV::],
320	[EV:: => AnyEvent::Impl::EV::],	1098	[EV:: => AnyEvent::Impl::EV::],
321	[Coro::Event:: => AnyEvent::Impl::CoroEvent::],
322	[Event:: => AnyEvent::Impl::Event::],	1099	[Event:: => AnyEvent::Impl::Event::],
323	[Glib:: => AnyEvent::Impl::Glib::],
324	[Tk:: => AnyEvent::Impl::Tk::],
325	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],	1100	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
		1101	# everything below here will not be autoprobed
		1102	# as the pureperl backend should work everywhere
		1103	# and is usually faster
		1104	[Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
		1105	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		1106	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		1107	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		1108	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		1109	[Wx:: => AnyEvent::Impl::POE::],
		1110	[Prima:: => AnyEvent::Impl::POE::],
		1111	# IO::Async is just too broken - we would need workarounds for its
		1112	# byzantine signal and broken child handling, among others.
		1113	# IO::Async is rather hard to detect, as it doesn't have any
		1114	# obvious default class.
		1115	# [IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1116	# [IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1117	# [IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program
326	);	1118	);
327		1119
328	our %method = map +($_ => 1), qw(io timer condvar broadcast wait signal one_event DESTROY);	1120	our %method = map +($_ => 1),
		1121	qw(io timer time now now_update signal child idle condvar one_event DESTROY);
		1122
		1123	our @post_detect;
		1124
		1125	sub post_detect(&) {
		1126	my ($cb) = @_;
		1127
		1128	if ($MODEL) {
		1129	$cb->();
		1130
		1131	1
		1132	} else {
		1133	push @post_detect, $cb;
		1134
		1135	defined wantarray
		1136	? bless \$cb, "AnyEvent::Util::postdetect"
		1137	: ()
		1138	}
		1139	}
		1140
		1141	sub AnyEvent::Util::postdetect::DESTROY {
		1142	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		1143	}
329		1144
330	sub detect() {	1145	sub detect() {
331	unless ($MODEL) {	1146	unless ($MODEL) {
332	no strict 'refs';	1147	local $SIG{__DIE__};
		1148
		1149	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
		1150	my $model = "AnyEvent::Impl::$1";
		1151	if (eval "require $model") {
		1152	$MODEL = $model;
		1153	warn "AnyEvent: loaded model '$model' (forced by \$ENV{PERL_ANYEVENT_MODEL}), using it.\n" if $VERBOSE >= 2;
		1154	} else {
		1155	warn "AnyEvent: unable to load model '$model' (from \$ENV{PERL_ANYEVENT_MODEL}):\n$@" if $VERBOSE;
		1156	}
		1157	}
333		1158
334	# check for already loaded models	1159	# check for already loaded models
		1160	unless ($MODEL) {
335	for (@REGISTRY, @models) {	1161	for (@REGISTRY, @models) {
336	my ($package, $model) = @$_;	1162	my ($package, $model) = @$_;
337	if (${"$package\::VERSION"} > 0) {	1163	if (${"$package\::VERSION"} > 0) {
338	if (eval "require $model") {	1164	if (eval "require $model") {
339	$MODEL = $model;	1165	$MODEL = $model;
340	warn "AnyEvent: found model '$model', using it.\n" if $verbose > 1;	1166	warn "AnyEvent: autodetected model '$model', using it.\n" if $VERBOSE >= 2;
341	last;	1167	last;
		1168	}
342	}	1169	}
343	}	1170	}
		1171
		1172	unless ($MODEL) {
		1173	# try to load a model
		1174
		1175	for (@REGISTRY, @models) {
		1176	my ($package, $model) = @$_;
		1177	if (eval "require $package"
		1178	and ${"$package\::VERSION"} > 0
		1179	and eval "require $model") {
		1180	$MODEL = $model;
		1181	warn "AnyEvent: autoprobed model '$model', using it.\n" if $VERBOSE >= 2;
		1182	last;
		1183	}
		1184	}
		1185
		1186	$MODEL
		1187	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
		1188	}
344	}	1189	}
345		1190
346	unless ($MODEL) {	1191	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
347	# try to load a model
348
349	for (@REGISTRY, @models) {
350	my ($package, $model) = @$_;
351	if (eval "require $package"
352	and ${"$package\::VERSION"} > 0
353	and eval "require $model") {
354	$MODEL = $model;
355	warn "AnyEvent: autoprobed and loaded model '$model', using it.\n" if $verbose > 1;
356	last;
357	}
358	}
359
360	$MODEL
361	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), Glib or Tk.";
362	}
363		1192
364	unshift @ISA, $MODEL;	1193	unshift @ISA, $MODEL;
365	push @{"$MODEL\::ISA"}, "AnyEvent::Base";	1194
		1195	require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
		1196
		1197	(shift @post_detect)->() while @post_detect;
366	}	1198	}
367		1199
368	$MODEL	1200	$MODEL
369	}	1201	}
370		1202
371	sub AUTOLOAD {	1203	sub AUTOLOAD {
372	(my $func = $AUTOLOAD) =~ s/.*://;	1204	(my $func = $AUTOLOAD) =~ s/.*://;
373		1205
374	$method{$func}	1206	$method{$func}
375	or croak "$func: not a valid method for AnyEvent objects";	1207	or Carp::croak "$func: not a valid method for AnyEvent objects";
376		1208
377	detect unless $MODEL;	1209	detect unless $MODEL;
378		1210
379	my $class = shift;	1211	my $class = shift;
380	$class->$func (@_);	1212	$class->$func (@_);
381	}	1213	}
382		1214
		1215	# utility function to dup a filehandle. this is used by many backends
		1216	# to support binding more than one watcher per filehandle (they usually
		1217	# allow only one watcher per fd, so we dup it to get a different one).
		1218	sub _dupfh($$;$$) {
		1219	my ($poll, $fh, $r, $w) = @_;
		1220
		1221	# cygwin requires the fh mode to be matching, unix doesn't
		1222	my ($rw, $mode) = $poll eq "r" ? ($r, "<&") : ($w, ">&");
		1223
		1224	open my $fh2, $mode, $fh
		1225	or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,";
		1226
		1227	# we assume CLOEXEC is already set by perl in all important cases
		1228
		1229	($fh2, $rw)
		1230	}
		1231
383	package AnyEvent::Base;	1232	package AnyEvent::Base;
384		1233
		1234	# default implementations for many methods
		1235
		1236	sub _time {
		1237	# probe for availability of Time::HiRes
		1238	if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
		1239	warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8;
		1240	*_time = \&Time::HiRes::time;
		1241	# if (eval "use POSIX (); (POSIX::times())...
		1242	} else {
		1243	warn "AnyEvent: using built-in time(), WARNING, no sub-second resolution!\n" if $VERBOSE;
		1244	*_time = sub { time }; # epic fail
		1245	}
		1246
		1247	&_time
		1248	}
		1249
		1250	sub time { _time }
		1251	sub now { _time }
		1252	sub now_update { }
		1253
385	# default implementation for ->condvar, ->wait, ->broadcast	1254	# default implementation for ->condvar
386		1255
387	sub condvar {	1256	sub condvar {
388	bless \my $flag, "AnyEvent::Base::CondVar"	1257	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
389	}
390
391	sub AnyEvent::Base::CondVar::broadcast {
392	${$_[0]}++;
393	}
394
395	sub AnyEvent::Base::CondVar::wait {
396	AnyEvent->one_event while !${$_[0]};
397	}	1258	}
398		1259
399	# default implementation for ->signal	1260	# default implementation for ->signal
400		1261
401	our %SIG_CB;	1262	our $HAVE_ASYNC_INTERRUPT;
		1263	our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
		1264	our (%SIG_ASY, %SIG_ASY_W);
		1265	our ($SIG_COUNT, $SIG_TW);
402		1266
		1267	sub _signal_exec {
		1268	$HAVE_ASYNC_INTERRUPT
		1269	? $SIGPIPE_R->drain
		1270	: sysread $SIGPIPE_R, my $dummy, 9;
		1271
		1272	while (%SIG_EV) {
		1273	for (keys %SIG_EV) {
		1274	delete $SIG_EV{$_};
		1275	$_->() for values %{ $SIG_CB{$_} || {} };
		1276	}
		1277	}
		1278	}
		1279
		1280	# install a dumym wakeupw atcher to reduce signal catching latency
		1281	sub _sig_add() {
		1282	unless ($SIG_COUNT++) {
		1283	# try to align timer on a full-second boundary, if possible
		1284	my $NOW = AnyEvent->now;
		1285
		1286	$SIG_TW = AnyEvent->timer (
		1287	after => $MAX_SIGNAL_LATENCY - ($NOW - int $NOW),
		1288	interval => $MAX_SIGNAL_LATENCY,
		1289	cb => sub { }, # just for the PERL_ASYNC_CHECK
		1290	);
		1291	}
		1292	}
		1293
		1294	sub _sig_del {
		1295	undef $SIG_TW
		1296	unless --$SIG_COUNT;
		1297	}
		1298
403	sub signal {	1299	sub _signal {
404	my (undef, %arg) = @_;	1300	my (undef, %arg) = @_;
405		1301
406	my $signal = uc $arg{signal}	1302	my $signal = uc $arg{signal}
407	or Carp::croak "required option 'signal' is missing";	1303	or Carp::croak "required option 'signal' is missing";
408		1304
409	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};	1305	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1306
		1307	if ($HAVE_ASYNC_INTERRUPT) {
		1308	# async::interrupt
		1309
		1310	$SIG_ASY{$signal} ||= do {
		1311	my $asy = new Async::Interrupt
		1312	cb => sub { undef $SIG_EV{$signal} },
		1313	signal => $signal,
		1314	pipe => [$SIGPIPE_R->filenos],
		1315	;
		1316	$asy->pipe_autodrain (0);
		1317
		1318	$asy
		1319	};
		1320
		1321	} else {
		1322	# pure perl
		1323
410	$SIG{$signal} ||= sub {	1324	$SIG{$signal} ||= sub {
411	$_->() for values %{ $SIG_CB{$signal} || {} };	1325	local $!;
		1326	syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
		1327	undef $SIG_EV{$signal};
		1328	};
		1329
		1330	# can't do signal processing without introducing races in pure perl,
		1331	# so limit the signal latency.
		1332	_sig_add;
412	};	1333	}
413		1334
414	bless [$signal, $arg{cb}], "AnyEvent::Base::Signal"	1335	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
415	}	1336	}
416		1337
		1338	sub signal {
		1339	# probe for availability of Async::Interrupt
		1340	if (!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT} && eval "use Async::Interrupt 0.6 (); 1") {
		1341	warn "AnyEvent: using Async::Interrupt for race-free signal handling.\n" if $VERBOSE >= 8;
		1342
		1343	$HAVE_ASYNC_INTERRUPT = 1;
		1344	$SIGPIPE_R = new Async::Interrupt::EventPipe;
		1345	$SIG_IO = AnyEvent->io (fh => $SIGPIPE_R->fileno, poll => "r", cb => \&_signal_exec);
		1346
		1347	} else {
		1348	warn "AnyEvent: using emulated perl signal handling with latency timer.\n" if $VERBOSE >= 8;
		1349
		1350	require Fcntl;
		1351
		1352	if (AnyEvent::WIN32) {
		1353	require AnyEvent::Util;
		1354
		1355	($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
		1356	AnyEvent::Util::fh_nonblocking ($SIGPIPE_R) if $SIGPIPE_R;
		1357	AnyEvent::Util::fh_nonblocking ($SIGPIPE_W) if $SIGPIPE_W; # just in case
		1358	} else {
		1359	pipe $SIGPIPE_R, $SIGPIPE_W;
		1360	fcntl $SIGPIPE_R, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_R;
		1361	fcntl $SIGPIPE_W, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_W; # just in case
		1362
		1363	# not strictly required, as $^F is normally 2, but let's make sure...
		1364	fcntl $SIGPIPE_R, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1365	fcntl $SIGPIPE_W, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1366	}
		1367
		1368	$SIGPIPE_R
		1369	or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
		1370
		1371	$SIG_IO = AnyEvent->io (fh => $SIGPIPE_R, poll => "r", cb => \&_signal_exec);
		1372	}
		1373
		1374	*signal = \&_signal;
		1375	&signal
		1376	}
		1377
417	sub AnyEvent::Base::Signal::DESTROY {	1378	sub AnyEvent::Base::signal::DESTROY {
418	my ($signal, $cb) = @{$_[0]};	1379	my ($signal, $cb) = @{$_[0]};
419		1380
		1381	_sig_del;
		1382
420	delete $SIG_CB{$signal}{$cb};	1383	delete $SIG_CB{$signal}{$cb};
421		1384
422	$SIG{$signal} = 'DEFAULT' unless keys %{ $SIG_CB{$signal} };	1385	$HAVE_ASYNC_INTERRUPT
		1386	? delete $SIG_ASY{$signal}
		1387	: # delete doesn't work with older perls - they then
		1388	# print weird messages, or just unconditionally exit
		1389	# instead of getting the default action.
		1390	undef $SIG{$signal}
		1391	unless keys %{ $SIG_CB{$signal} };
423	}	1392	}
424		1393
425	# default implementation for ->child	1394	# default implementation for ->child
426		1395
427	our %PID_CB;	1396	our %PID_CB;
428	our $CHLD_W;	1397	our $CHLD_W;
429	our $CHLD_DELAY_W;	1398	our $CHLD_DELAY_W;
430	our $PID_IDLE;
431	our $WNOHANG;	1399	our $WNOHANG;
432		1400
433	sub _child_wait {	1401	sub _sigchld {
434	while (0 < (my $pid = waitpid -1, $WNOHANG)) {	1402	while (0 < (my $pid = waitpid -1, $WNOHANG)) {
		1403	$_->($pid, $?)
435	$_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }),	1404	for values %{ $PID_CB{$pid} || {} },
436	(values %{ $PID_CB{0} || {} });	1405	values %{ $PID_CB{0} || {} };
437	}	1406	}
438
439	undef $PID_IDLE;
440	}
441
442	sub _sigchld {
443	# make sure we deliver these changes "synchronous" with the event loop.
444	$CHLD_DELAY_W ||= AnyEvent->timer (after => 0, cb => sub {
445	undef $CHLD_DELAY_W;
446	&_child_wait;
447	});
448	}	1407	}
449		1408
450	sub child {	1409	sub child {
451	my (undef, %arg) = @_;	1410	my (undef, %arg) = @_;
452		1411
453	defined (my $pid = $arg{pid} + 0)	1412	defined (my $pid = $arg{pid} + 0)
454	or Carp::croak "required option 'pid' is missing";	1413	or Carp::croak "required option 'pid' is missing";
455		1414
456	$PID_CB{$pid}{$arg{cb}} = $arg{cb};	1415	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
457		1416
458	unless ($WNOHANG) {	1417	# WNOHANG is almost cetrainly 1 everywhere
459	$WNOHANG = eval { require POSIX; &POSIX::WNOHANG } || 1;	1418	$WNOHANG ||= $^O =~ /^(?:openbsd|netbsd|linux|freebsd|cygwin|MSWin32)$/
460	}	1419	? 1
		1420	: eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
461		1421
462	unless ($CHLD_W) {	1422	unless ($CHLD_W) {
463	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);	1423	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
464	# child could be a zombie already, so make at least one round	1424	# child could be a zombie already, so make at least one round
465	&_sigchld;	1425	&_sigchld;
466	}	1426	}
467		1427
468	bless [$pid, $arg{cb}], "AnyEvent::Base::Child"	1428	bless [$pid, $arg{cb}], "AnyEvent::Base::child"
469	}	1429	}
470		1430
471	sub AnyEvent::Base::Child::DESTROY {	1431	sub AnyEvent::Base::child::DESTROY {
472	my ($pid, $cb) = @{$_[0]};	1432	my ($pid, $cb) = @{$_[0]};
473		1433
474	delete $PID_CB{$pid}{$cb};	1434	delete $PID_CB{$pid}{$cb};
475	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };	1435	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
476		1436
477	undef $CHLD_W unless keys %PID_CB;	1437	undef $CHLD_W unless keys %PID_CB;
478	}	1438	}
479		1439
		1440	# idle emulation is done by simply using a timer, regardless
		1441	# of whether the process is idle or not, and not letting
		1442	# the callback use more than 50% of the time.
		1443	sub idle {
		1444	my (undef, %arg) = @_;
		1445
		1446	my ($cb, $w, $rcb) = $arg{cb};
		1447
		1448	$rcb = sub {
		1449	if ($cb) {
		1450	$w = _time;
		1451	&$cb;
		1452	$w = _time - $w;
		1453
		1454	# never use more then 50% of the time for the idle watcher,
		1455	# within some limits
		1456	$w = 0.0001 if $w < 0.0001;
		1457	$w = 5 if $w > 5;
		1458
		1459	$w = AnyEvent->timer (after => $w, cb => $rcb);
		1460	} else {
		1461	# clean up...
		1462	undef $w;
		1463	undef $rcb;
		1464	}
		1465	};
		1466
		1467	$w = AnyEvent->timer (after => 0.05, cb => $rcb);
		1468
		1469	bless \\$cb, "AnyEvent::Base::idle"
		1470	}
		1471
		1472	sub AnyEvent::Base::idle::DESTROY {
		1473	undef $${$_[0]};
		1474	}
		1475
		1476	package AnyEvent::CondVar;
		1477
		1478	our @ISA = AnyEvent::CondVar::Base::;
		1479
		1480	package AnyEvent::CondVar::Base;
		1481
		1482	#use overload
		1483	# '&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1484	# fallback => 1;
		1485
		1486	# save 300+ kilobytes by dirtily hardcoding overloading
		1487	${"AnyEvent::CondVar::Base::OVERLOAD"}{dummy}++; # Register with magic by touching.
		1488	*{'AnyEvent::CondVar::Base::()'} = sub { }; # "Make it findable via fetchmethod."
		1489	*{'AnyEvent::CondVar::Base::(&{}'} = sub { my $self = shift; sub { $self->send (@_) } }; # &{}
		1490	${'AnyEvent::CondVar::Base::()'} = 1; # fallback
		1491
		1492	our $WAITING;
		1493
		1494	sub _send {
		1495	# nop
		1496	}
		1497
		1498	sub send {
		1499	my $cv = shift;
		1500	$cv->{_ae_sent} = [@_];
		1501	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1502	$cv->_send;
		1503	}
		1504
		1505	sub croak {
		1506	$_[0]{_ae_croak} = $_[1];
		1507	$_[0]->send;
		1508	}
		1509
		1510	sub ready {
		1511	$_[0]{_ae_sent}
		1512	}
		1513
		1514	sub _wait {
		1515	$WAITING
		1516	and !$_[0]{_ae_sent}
		1517	and Carp::croak "AnyEvent::CondVar: recursive blocking wait detected";
		1518
		1519	local $WAITING = 1;
		1520	AnyEvent->one_event while !$_[0]{_ae_sent};
		1521	}
		1522
		1523	sub recv {
		1524	$_[0]->_wait;
		1525
		1526	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1527	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1528	}
		1529
		1530	sub cb {
		1531	$_[0]{_ae_cb} = $_[1] if @_ > 1;
		1532	$_[0]{_ae_cb}
		1533	}
		1534
		1535	sub begin {
		1536	++$_[0]{_ae_counter};
		1537	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1538	}
		1539
		1540	sub end {
		1541	return if --$_[0]{_ae_counter};
		1542	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1543	}
		1544
		1545	# undocumented/compatibility with pre-3.4
		1546	*broadcast = \&send;
		1547	*wait = \&_wait;
		1548
		1549	=head1 ERROR AND EXCEPTION HANDLING
		1550
		1551	In general, AnyEvent does not do any error handling - it relies on the
		1552	caller to do that if required. The L<AnyEvent::Strict> module (see also
		1553	the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
		1554	checking of all AnyEvent methods, however, which is highly useful during
		1555	development.
		1556
		1557	As for exception handling (i.e. runtime errors and exceptions thrown while
		1558	executing a callback), this is not only highly event-loop specific, but
		1559	also not in any way wrapped by this module, as this is the job of the main
		1560	program.
		1561
		1562	The pure perl event loop simply re-throws the exception (usually
		1563	within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
		1564	$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
		1565	so on.
		1566
		1567	=head1 ENVIRONMENT VARIABLES
		1568
		1569	The following environment variables are used by this module or its
		1570	submodules.
		1571
		1572	Note that AnyEvent will remove I<all> environment variables starting with
		1573	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		1574	enabled.
		1575
		1576	=over 4
		1577
		1578	=item C<PERL_ANYEVENT_VERBOSE>
		1579
		1580	By default, AnyEvent will be completely silent except in fatal
		1581	conditions. You can set this environment variable to make AnyEvent more
		1582	talkative.
		1583
		1584	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1585	conditions, such as not being able to load the event model specified by
		1586	C<PERL_ANYEVENT_MODEL>.
		1587
		1588	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1589	model it chooses.
		1590
		1591	When set to C<8> or higher, then AnyEvent will report extra information on
		1592	which optional modules it loads and how it implements certain features.
		1593
		1594	=item C<PERL_ANYEVENT_STRICT>
		1595
		1596	AnyEvent does not do much argument checking by default, as thorough
		1597	argument checking is very costly. Setting this variable to a true value
		1598	will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
		1599	check the arguments passed to most method calls. If it finds any problems,
		1600	it will croak.
		1601
		1602	In other words, enables "strict" mode.
		1603
		1604	Unlike C<use strict> (or it's modern cousin, C<< use L<common::sense>
		1605	>>, it is definitely recommended to keep it off in production. Keeping
		1606	C<PERL_ANYEVENT_STRICT=1> in your environment while developing programs
		1607	can be very useful, however.
		1608
		1609	=item C<PERL_ANYEVENT_MODEL>
		1610
		1611	This can be used to specify the event model to be used by AnyEvent, before
		1612	auto detection and -probing kicks in. It must be a string consisting
		1613	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1614	and the resulting module name is loaded and if the load was successful,
		1615	used as event model. If it fails to load AnyEvent will proceed with
		1616	auto detection and -probing.
		1617
		1618	This functionality might change in future versions.
		1619
		1620	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1621	could start your program like this:
		1622
		1623	PERL_ANYEVENT_MODEL=Perl perl ...
		1624
		1625	=item C<PERL_ANYEVENT_PROTOCOLS>
		1626
		1627	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1628	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1629	of auto probing).
		1630
		1631	Must be set to a comma-separated list of protocols or address families,
		1632	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1633	used, and preference will be given to protocols mentioned earlier in the
		1634	list.
		1635
		1636	This variable can effectively be used for denial-of-service attacks
		1637	against local programs (e.g. when setuid), although the impact is likely
		1638	small, as the program has to handle conenction and other failures anyways.
		1639
		1640	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1641	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1642	- only support IPv4, never try to resolve or contact IPv6
		1643	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1644	IPv6, but prefer IPv6 over IPv4.
		1645
		1646	=item C<PERL_ANYEVENT_EDNS0>
		1647
		1648	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1649	for DNS. This extension is generally useful to reduce DNS traffic, but
		1650	some (broken) firewalls drop such DNS packets, which is why it is off by
		1651	default.
		1652
		1653	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1654	EDNS0 in its DNS requests.
		1655
		1656	=item C<PERL_ANYEVENT_MAX_FORKS>
		1657
		1658	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1659	will create in parallel.
		1660
		1661	=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
		1662
		1663	The default value for the C<max_outstanding> parameter for the default DNS
		1664	resolver - this is the maximum number of parallel DNS requests that are
		1665	sent to the DNS server.
		1666
		1667	=item C<PERL_ANYEVENT_RESOLV_CONF>
		1668
		1669	The file to use instead of F</etc/resolv.conf> (or OS-specific
		1670	configuration) in the default resolver. When set to the empty string, no
		1671	default config will be used.
		1672
		1673	=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
		1674
		1675	When neither C<ca_file> nor C<ca_path> was specified during
		1676	L<AnyEvent::TLS> context creation, and either of these environment
		1677	variables exist, they will be used to specify CA certificate locations
		1678	instead of a system-dependent default.
		1679
		1680	=item C<PERL_ANYEVENT_AVOID_GUARD> and C<PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT>
		1681
		1682	When these are set to C<1>, then the respective modules are not
		1683	loaded. Mostly good for testing AnyEvent itself.
		1684
		1685	=back
		1686
480	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE	1687	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1688
		1689	This is an advanced topic that you do not normally need to use AnyEvent in
		1690	a module. This section is only of use to event loop authors who want to
		1691	provide AnyEvent compatibility.
481		1692
482	If you need to support another event library which isn't directly	1693	If you need to support another event library which isn't directly
483	supported by AnyEvent, you can supply your own interface to it by	1694	supported by AnyEvent, you can supply your own interface to it by
484	pushing, before the first watcher gets created, the package name of	1695	pushing, before the first watcher gets created, the package name of
485	the event module and the package name of the interface to use onto	1696	the event module and the package name of the interface to use onto
486	C<@AnyEvent::REGISTRY>. You can do that before and even without loading	1697	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
487	AnyEvent.	1698	AnyEvent, so it is reasonably cheap.
488		1699
489	Example:	1700	Example:
490		1701
491	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];	1702	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
492		1703
493	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>	1704	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
494	package/class when it finds the C<urxvt> package/module is loaded. When	1705	package/class when it finds the C<urxvt> package/module is already loaded.
		1706
495	AnyEvent is loaded and asked to find a suitable event model, it will	1707	When AnyEvent is loaded and asked to find a suitable event model, it
496	first check for the presence of urxvt.	1708	will first check for the presence of urxvt by trying to C<use> the
		1709	C<urxvt::anyevent> module.
497		1710
498	The class should provide implementations for all watcher types (see	1711	The class should provide implementations for all watcher types. See
499	L<AnyEvent::Impl::Event> (source code), L<AnyEvent::Impl::Glib>	1712	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
500	(Source code) and so on for actual examples, use C<perldoc -m	1713	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
501	AnyEvent::Impl::Glib> to see the sources).	1714	see the sources.
502		1715
		1716	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1717	provide suitable (hopefully) replacements.
		1718
503	The above isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)	1719	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
504	uses the above line as-is. An interface isn't included in AnyEvent	1720	terminal emulator uses the above line as-is. An interface isn't included
505	because it doesn't make sense outside the embedded interpreter inside	1721	in AnyEvent because it doesn't make sense outside the embedded interpreter
506	I<rxvt-unicode>, and it is updated and maintained as part of the	1722	inside I<rxvt-unicode>, and it is updated and maintained as part of the
507	I<rxvt-unicode> distribution.	1723	I<rxvt-unicode> distribution.
508		1724
509	I<rxvt-unicode> also cheats a bit by not providing blocking access to	1725	I<rxvt-unicode> also cheats a bit by not providing blocking access to
510	condition variables: code blocking while waiting for a condition will	1726	condition variables: code blocking while waiting for a condition will
511	C<die>. This still works with most modules/usages, and blocking calls must	1727	C<die>. This still works with most modules/usages, and blocking calls must
512	not be in an interactive application, so it makes sense.	1728	not be done in an interactive application, so it makes sense.
513		1729
514	=head1 ENVIRONMENT VARIABLES
515
516	The following environment variables are used by this module:
517
518	C<PERL_ANYEVENT_VERBOSE> when set to C<2> or higher, reports which event
519	model gets used.
520
521	=head1 EXAMPLE	1730	=head1 EXAMPLE PROGRAM
522		1731
523	The following program uses an io watcher to read data from stdin, a timer	1732	The following program uses an I/O watcher to read data from STDIN, a timer
524	to display a message once per second, and a condvar to exit the program	1733	to display a message once per second, and a condition variable to quit the
525	when the user enters quit:	1734	program when the user enters quit:
526		1735
527	use AnyEvent;	1736	use AnyEvent;
528		1737
529	my $cv = AnyEvent->condvar;	1738	my $cv = AnyEvent->condvar;
530		1739
531	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	1740	my $io_watcher = AnyEvent->io (
		1741	fh => \*STDIN,
		1742	poll => 'r',
		1743	cb => sub {
532	warn "io event <$_[0]>\n"; # will always output <r>	1744	warn "io event <$_[0]>\n"; # will always output <r>
533	chomp (my $input = <STDIN>); # read a line	1745	chomp (my $input = <STDIN>); # read a line
534	warn "read: $input\n"; # output what has been read	1746	warn "read: $input\n"; # output what has been read
535	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1747	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1748	},
536	});	1749	);
537		1750
538	my $time_watcher; # can only be used once	1751	my $time_watcher; # can only be used once
539		1752
540	sub new_timer {	1753	sub new_timer {
541	$timer = AnyEvent->timer (after => 1, cb => sub {	1754	$timer = AnyEvent->timer (after => 1, cb => sub {
…		…
544	});	1757	});
545	}	1758	}
546		1759
547	new_timer; # create first timer	1760	new_timer; # create first timer
548		1761
549	$cv->wait; # wait until user enters /^q/i	1762	$cv->recv; # wait until user enters /^q/i
550		1763
551	=head1 REAL-WORLD EXAMPLE	1764	=head1 REAL-WORLD EXAMPLE
552		1765
553	Consider the L<Net::FCP> module. It features (among others) the following	1766	Consider the L<Net::FCP> module. It features (among others) the following
554	API calls, which are to freenet what HTTP GET requests are to http:	1767	API calls, which are to freenet what HTTP GET requests are to http:
…		…
604	syswrite $txn->{fh}, $txn->{request}	1817	syswrite $txn->{fh}, $txn->{request}
605	or die "connection or write error";	1818	or die "connection or write error";
606	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	1819	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
607		1820
608	Again, C<fh_ready_r> waits till all data has arrived, and then stores the	1821	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
609	result and signals any possible waiters that the request ahs finished:	1822	result and signals any possible waiters that the request has finished:
610		1823
611	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	1824	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
612		1825
613	if (end-of-file or data complete) {	1826	if (end-of-file or data complete) {
614	$txn->{result} = $txn->{buf};	1827	$txn->{result} = $txn->{buf};
615	$txn->{finished}->broadcast;	1828	$txn->{finished}->send;
616	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	1829	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
617	}	1830	}
618		1831
619	The C<result> method, finally, just waits for the finished signal (if the	1832	The C<result> method, finally, just waits for the finished signal (if the
620	request was already finished, it doesn't wait, of course, and returns the	1833	request was already finished, it doesn't wait, of course, and returns the
621	data:	1834	data:
622		1835
623	$txn->{finished}->wait;	1836	$txn->{finished}->recv;
624	return $txn->{result};	1837	return $txn->{result};
625		1838
626	The actual code goes further and collects all errors (C<die>s, exceptions)	1839	The actual code goes further and collects all errors (C<die>s, exceptions)
627	that occured during request processing. The C<result> method detects	1840	that occurred during request processing. The C<result> method detects
628	wether an exception as thrown (it is stored inside the $txn object)	1841	whether an exception as thrown (it is stored inside the $txn object)
629	and just throws the exception, which means connection errors and other	1842	and just throws the exception, which means connection errors and other
630	problems get reported tot he code that tries to use the result, not in a	1843	problems get reported tot he code that tries to use the result, not in a
631	random callback.	1844	random callback.
632		1845
633	All of this enables the following usage styles:	1846	All of this enables the following usage styles:
634		1847
635	1. Blocking:	1848	1. Blocking:
636		1849
637	my $data = $fcp->client_get ($url);	1850	my $data = $fcp->client_get ($url);
638		1851
639	2. Blocking, but parallelizing:	1852	2. Blocking, but running in parallel:
640		1853
641	my @datas = map $_->result,	1854	my @datas = map $_->result,
642	map $fcp->txn_client_get ($_),	1855	map $fcp->txn_client_get ($_),
643	@urls;	1856	@urls;
644		1857
645	Both blocking examples work without the module user having to know	1858	Both blocking examples work without the module user having to know
646	anything about events.	1859	anything about events.
647		1860
648	3a. Event-based in a main program, using any support Event module:	1861	3a. Event-based in a main program, using any supported event module:
649		1862
650	use Event;	1863	use EV;
651		1864
652	$fcp->txn_client_get ($url)->cb (sub {	1865	$fcp->txn_client_get ($url)->cb (sub {
653	my $txn = shift;	1866	my $txn = shift;
654	my $data = $txn->result;	1867	my $data = $txn->result;
655	...	1868	...
656	});	1869	});
657		1870
658	Event::loop;	1871	EV::loop;
659		1872
660	3b. The module user could use AnyEvent, too:	1873	3b. The module user could use AnyEvent, too:
661		1874
662	use AnyEvent;	1875	use AnyEvent;
663		1876
664	my $quit = AnyEvent->condvar;	1877	my $quit = AnyEvent->condvar;
665		1878
666	$fcp->txn_client_get ($url)->cb (sub {	1879	$fcp->txn_client_get ($url)->cb (sub {
667	...	1880	...
668	$quit->broadcast;	1881	$quit->send;
669	});	1882	});
670		1883
671	$quit->wait;	1884	$quit->recv;
		1885
		1886
		1887	=head1 BENCHMARKS
		1888
		1889	To give you an idea of the performance and overheads that AnyEvent adds
		1890	over the event loops themselves and to give you an impression of the speed
		1891	of various event loops I prepared some benchmarks.
		1892
		1893	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1894
		1895	Here is a benchmark of various supported event models used natively and
		1896	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1897	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1898	which it is), lets them fire exactly once and destroys them again.
		1899
		1900	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		1901	distribution.
		1902
		1903	=head3 Explanation of the columns
		1904
		1905	I<watcher> is the number of event watchers created/destroyed. Since
		1906	different event models feature vastly different performances, each event
		1907	loop was given a number of watchers so that overall runtime is acceptable
		1908	and similar between tested event loop (and keep them from crashing): Glib
		1909	would probably take thousands of years if asked to process the same number
		1910	of watchers as EV in this benchmark.
		1911
		1912	I<bytes> is the number of bytes (as measured by the resident set size,
		1913	RSS) consumed by each watcher. This method of measuring captures both C
		1914	and Perl-based overheads.
		1915
		1916	I<create> is the time, in microseconds (millionths of seconds), that it
		1917	takes to create a single watcher. The callback is a closure shared between
		1918	all watchers, to avoid adding memory overhead. That means closure creation
		1919	and memory usage is not included in the figures.
		1920
		1921	I<invoke> is the time, in microseconds, used to invoke a simple
		1922	callback. The callback simply counts down a Perl variable and after it was
		1923	invoked "watcher" times, it would C<< ->send >> a condvar once to
		1924	signal the end of this phase.
		1925
		1926	I<destroy> is the time, in microseconds, that it takes to destroy a single
		1927	watcher.
		1928
		1929	=head3 Results
		1930
		1931	name watchers bytes create invoke destroy comment
		1932	EV/EV 400000 224 0.47 0.35 0.27 EV native interface
		1933	EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers
		1934	CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal
		1935	Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation
		1936	Event/Event 16000 517 32.20 31.80 0.81 Event native interface
		1937	Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers
		1938	IOAsync/Any 16000 989 38.10 32.77 11.13 via IO::Async::Loop::IO_Poll
		1939	IOAsync/Any 16000 990 37.59 29.50 10.61 via IO::Async::Loop::Epoll
		1940	Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour
		1941	Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers
		1942	POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event
		1943	POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select
		1944
		1945	=head3 Discussion
		1946
		1947	The benchmark does I<not> measure scalability of the event loop very
		1948	well. For example, a select-based event loop (such as the pure perl one)
		1949	can never compete with an event loop that uses epoll when the number of
		1950	file descriptors grows high. In this benchmark, all events become ready at
		1951	the same time, so select/poll-based implementations get an unnatural speed
		1952	boost.
		1953
		1954	Also, note that the number of watchers usually has a nonlinear effect on
		1955	overall speed, that is, creating twice as many watchers doesn't take twice
		1956	the time - usually it takes longer. This puts event loops tested with a
		1957	higher number of watchers at a disadvantage.
		1958
		1959	To put the range of results into perspective, consider that on the
		1960	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1961	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1962	cycles with POE.
		1963
		1964	C<EV> is the sole leader regarding speed and memory use, which are both
		1965	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1966	far less memory than any other event loop and is still faster than Event
		1967	natively.
		1968
		1969	The pure perl implementation is hit in a few sweet spots (both the
		1970	constant timeout and the use of a single fd hit optimisations in the perl
		1971	interpreter and the backend itself). Nevertheless this shows that it
		1972	adds very little overhead in itself. Like any select-based backend its
		1973	performance becomes really bad with lots of file descriptors (and few of
		1974	them active), of course, but this was not subject of this benchmark.
		1975
		1976	The C<Event> module has a relatively high setup and callback invocation
		1977	cost, but overall scores in on the third place.
		1978
		1979	C<IO::Async> performs admirably well, about on par with C<Event>, even
		1980	when using its pure perl backend.
		1981
		1982	C<Glib>'s memory usage is quite a bit higher, but it features a
		1983	faster callback invocation and overall ends up in the same class as
		1984	C<Event>. However, Glib scales extremely badly, doubling the number of
		1985	watchers increases the processing time by more than a factor of four,
		1986	making it completely unusable when using larger numbers of watchers
		1987	(note that only a single file descriptor was used in the benchmark, so
		1988	inefficiencies of C<poll> do not account for this).
		1989
		1990	The C<Tk> adaptor works relatively well. The fact that it crashes with
		1991	more than 2000 watchers is a big setback, however, as correctness takes
		1992	precedence over speed. Nevertheless, its performance is surprising, as the
		1993	file descriptor is dup()ed for each watcher. This shows that the dup()
		1994	employed by some adaptors is not a big performance issue (it does incur a
		1995	hidden memory cost inside the kernel which is not reflected in the figures
		1996	above).
		1997
		1998	C<POE>, regardless of underlying event loop (whether using its pure perl
		1999	select-based backend or the Event module, the POE-EV backend couldn't
		2000	be tested because it wasn't working) shows abysmal performance and
		2001	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		2002	as EV watchers, and 10 times as much memory as Event (the high memory
		2003	requirements are caused by requiring a session for each watcher). Watcher
		2004	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		2005	implementation.
		2006
		2007	The design of the POE adaptor class in AnyEvent can not really account
		2008	for the performance issues, though, as session creation overhead is
		2009	small compared to execution of the state machine, which is coded pretty
		2010	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		2011	using multiple sessions is not a good approach, especially regarding
		2012	memory usage, even the author of POE could not come up with a faster
		2013	design).
		2014
		2015	=head3 Summary
		2016
		2017	=over 4
		2018
		2019	=item * Using EV through AnyEvent is faster than any other event loop
		2020	(even when used without AnyEvent), but most event loops have acceptable
		2021	performance with or without AnyEvent.
		2022
		2023	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		2024	the actual event loop, only with extremely fast event loops such as EV
		2025	adds AnyEvent significant overhead.
		2026
		2027	=item * You should avoid POE like the plague if you want performance or
		2028	reasonable memory usage.
		2029
		2030	=back
		2031
		2032	=head2 BENCHMARKING THE LARGE SERVER CASE
		2033
		2034	This benchmark actually benchmarks the event loop itself. It works by
		2035	creating a number of "servers": each server consists of a socket pair, a
		2036	timeout watcher that gets reset on activity (but never fires), and an I/O
		2037	watcher waiting for input on one side of the socket. Each time the socket
		2038	watcher reads a byte it will write that byte to a random other "server".
		2039
		2040	The effect is that there will be a lot of I/O watchers, only part of which
		2041	are active at any one point (so there is a constant number of active
		2042	fds for each loop iteration, but which fds these are is random). The
		2043	timeout is reset each time something is read because that reflects how
		2044	most timeouts work (and puts extra pressure on the event loops).
		2045
		2046	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		2047	(1%) are active. This mirrors the activity of large servers with many
		2048	connections, most of which are idle at any one point in time.
		2049
		2050	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		2051	distribution.
		2052
		2053	=head3 Explanation of the columns
		2054
		2055	I<sockets> is the number of sockets, and twice the number of "servers" (as
		2056	each server has a read and write socket end).
		2057
		2058	I<create> is the time it takes to create a socket pair (which is
		2059	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		2060
		2061	I<request>, the most important value, is the time it takes to handle a
		2062	single "request", that is, reading the token from the pipe and forwarding
		2063	it to another server. This includes deleting the old timeout and creating
		2064	a new one that moves the timeout into the future.
		2065
		2066	=head3 Results
		2067
		2068	name sockets create request
		2069	EV 20000 69.01 11.16
		2070	Perl 20000 73.32 35.87
		2071	IOAsync 20000 157.00 98.14 epoll
		2072	IOAsync 20000 159.31 616.06 poll
		2073	Event 20000 212.62 257.32
		2074	Glib 20000 651.16 1896.30
		2075	POE 20000 349.67 12317.24 uses POE::Loop::Event
		2076
		2077	=head3 Discussion
		2078
		2079	This benchmark I<does> measure scalability and overall performance of the
		2080	particular event loop.
		2081
		2082	EV is again fastest. Since it is using epoll on my system, the setup time
		2083	is relatively high, though.
		2084
		2085	Perl surprisingly comes second. It is much faster than the C-based event
		2086	loops Event and Glib.
		2087
		2088	IO::Async performs very well when using its epoll backend, and still quite
		2089	good compared to Glib when using its pure perl backend.
		2090
		2091	Event suffers from high setup time as well (look at its code and you will
		2092	understand why). Callback invocation also has a high overhead compared to
		2093	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		2094	uses select or poll in basically all documented configurations.
		2095
		2096	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		2097	clearly fails to perform with many filehandles or in busy servers.
		2098
		2099	POE is still completely out of the picture, taking over 1000 times as long
		2100	as EV, and over 100 times as long as the Perl implementation, even though
		2101	it uses a C-based event loop in this case.
		2102
		2103	=head3 Summary
		2104
		2105	=over 4
		2106
		2107	=item * The pure perl implementation performs extremely well.
		2108
		2109	=item * Avoid Glib or POE in large projects where performance matters.
		2110
		2111	=back
		2112
		2113	=head2 BENCHMARKING SMALL SERVERS
		2114
		2115	While event loops should scale (and select-based ones do not...) even to
		2116	large servers, most programs we (or I :) actually write have only a few
		2117	I/O watchers.
		2118
		2119	In this benchmark, I use the same benchmark program as in the large server
		2120	case, but it uses only eight "servers", of which three are active at any
		2121	one time. This should reflect performance for a small server relatively
		2122	well.
		2123
		2124	The columns are identical to the previous table.
		2125
		2126	=head3 Results
		2127
		2128	name sockets create request
		2129	EV 16 20.00 6.54
		2130	Perl 16 25.75 12.62
		2131	Event 16 81.27 35.86
		2132	Glib 16 32.63 15.48
		2133	POE 16 261.87 276.28 uses POE::Loop::Event
		2134
		2135	=head3 Discussion
		2136
		2137	The benchmark tries to test the performance of a typical small
		2138	server. While knowing how various event loops perform is interesting, keep
		2139	in mind that their overhead in this case is usually not as important, due
		2140	to the small absolute number of watchers (that is, you need efficiency and
		2141	speed most when you have lots of watchers, not when you only have a few of
		2142	them).
		2143
		2144	EV is again fastest.
		2145
		2146	Perl again comes second. It is noticeably faster than the C-based event
		2147	loops Event and Glib, although the difference is too small to really
		2148	matter.
		2149
		2150	POE also performs much better in this case, but is is still far behind the
		2151	others.
		2152
		2153	=head3 Summary
		2154
		2155	=over 4
		2156
		2157	=item * C-based event loops perform very well with small number of
		2158	watchers, as the management overhead dominates.
		2159
		2160	=back
		2161
		2162	=head2 THE IO::Lambda BENCHMARK
		2163
		2164	Recently I was told about the benchmark in the IO::Lambda manpage, which
		2165	could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
		2166	simply compares IO::Lambda with POE, and IO::Lambda looks better (which
		2167	shouldn't come as a surprise to anybody). As such, the benchmark is
		2168	fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
		2169	very optimal. But how would AnyEvent compare when used without the extra
		2170	baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
		2171
		2172	The benchmark itself creates an echo-server, and then, for 500 times,
		2173	connects to the echo server, sends a line, waits for the reply, and then
		2174	creates the next connection. This is a rather bad benchmark, as it doesn't
		2175	test the efficiency of the framework or much non-blocking I/O, but it is a
		2176	benchmark nevertheless.
		2177
		2178	name runtime
		2179	Lambda/select 0.330 sec
		2180	+ optimized 0.122 sec
		2181	Lambda/AnyEvent 0.327 sec
		2182	+ optimized 0.138 sec
		2183	Raw sockets/select 0.077 sec
		2184	POE/select, components 0.662 sec
		2185	POE/select, raw sockets 0.226 sec
		2186	POE/select, optimized 0.404 sec
		2187
		2188	AnyEvent/select/nb 0.085 sec
		2189	AnyEvent/EV/nb 0.068 sec
		2190	+state machine 0.134 sec
		2191
		2192	The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
		2193	benchmarks actually make blocking connects and use 100% blocking I/O,
		2194	defeating the purpose of an event-based solution. All of the newly
		2195	written AnyEvent benchmarks use 100% non-blocking connects (using
		2196	AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
		2197	resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
		2198	generally require a lot more bookkeeping and event handling than blocking
		2199	connects (which involve a single syscall only).
		2200
		2201	The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
		2202	offers similar expressive power as POE and IO::Lambda, using conventional
		2203	Perl syntax. This means that both the echo server and the client are 100%
		2204	non-blocking, further placing it at a disadvantage.
		2205
		2206	As you can see, the AnyEvent + EV combination even beats the
		2207	hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
		2208	backend easily beats IO::Lambda and POE.
		2209
		2210	And even the 100% non-blocking version written using the high-level (and
		2211	slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda by a
		2212	large margin, even though it does all of DNS, tcp-connect and socket I/O
		2213	in a non-blocking way.
		2214
		2215	The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
		2216	F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
		2217	part of the IO::lambda distribution and were used without any changes.
		2218
		2219
		2220	=head1 SIGNALS
		2221
		2222	AnyEvent currently installs handlers for these signals:
		2223
		2224	=over 4
		2225
		2226	=item SIGCHLD
		2227
		2228	A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
		2229	emulation for event loops that do not support them natively. Also, some
		2230	event loops install a similar handler.
		2231
		2232	Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE, then
		2233	AnyEvent will reset it to default, to avoid losing child exit statuses.
		2234
		2235	=item SIGPIPE
		2236
		2237	A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
		2238	when AnyEvent gets loaded.
		2239
		2240	The rationale for this is that AnyEvent users usually do not really depend
		2241	on SIGPIPE delivery (which is purely an optimisation for shell use, or
		2242	badly-written programs), but C<SIGPIPE> can cause spurious and rare
		2243	program exits as a lot of people do not expect C<SIGPIPE> when writing to
		2244	some random socket.
		2245
		2246	The rationale for installing a no-op handler as opposed to ignoring it is
		2247	that this way, the handler will be restored to defaults on exec.
		2248
		2249	Feel free to install your own handler, or reset it to defaults.
		2250
		2251	=back
		2252
		2253	=cut
		2254
		2255	undef $SIG{CHLD}
		2256	if $SIG{CHLD} eq 'IGNORE';
		2257
		2258	$SIG{PIPE} = sub { }
		2259	unless defined $SIG{PIPE};
		2260
		2261	=head1 RECOMMENDED/OPTIONAL MODULES
		2262
		2263	One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
		2264	it's built-in modules) are required to use it.
		2265
		2266	That does not mean that AnyEvent won't take advantage of some additional
		2267	modules if they are installed.
		2268
		2269	This section epxlains which additional modules will be used, and how they
		2270	affect AnyEvent's operetion.
		2271
		2272	=over 4
		2273
		2274	=item L<Async::Interrupt>
		2275
		2276	This slightly arcane module is used to implement fast signal handling: To
		2277	my knowledge, there is no way to do completely race-free and quick
		2278	signal handling in pure perl. To ensure that signals still get
		2279	delivered, AnyEvent will start an interval timer to wake up perl (and
		2280	catch the signals) with some delay (default is 10 seconds, look for
		2281	C<$AnyEvent::MAX_SIGNAL_LATENCY>).
		2282
		2283	If this module is available, then it will be used to implement signal
		2284	catching, which means that signals will not be delayed, and the event loop
		2285	will not be interrupted regularly, which is more efficient (And good for
		2286	battery life on laptops).
		2287
		2288	This affects not just the pure-perl event loop, but also other event loops
		2289	that have no signal handling on their own (e.g. Glib, Tk, Qt).
		2290
		2291	Some event loops (POE, Event, Event::Lib) offer signal watchers natively,
		2292	and either employ their own workarounds (POE) or use AnyEvent's workaround
		2293	(using C<$AnyEvent::MAX_SIGNAL_LATENCY>). Installing L<Async::Interrupt>
		2294	does nothing for those backends.
		2295
		2296	=item L<EV>
		2297
		2298	This module isn't really "optional", as it is simply one of the backend
		2299	event loops that AnyEvent can use. However, it is simply the best event
		2300	loop available in terms of features, speed and stability: It supports
		2301	the AnyEvent API optimally, implements all the watcher types in XS, does
		2302	automatic timer adjustments even when no monotonic clock is available,
		2303	can take avdantage of advanced kernel interfaces such as C<epoll> and
		2304	C<kqueue>, and is the fastest backend I<by far>. You can even embed
		2305	L<Glib>/L<Gtk2> in it (or vice versa, see L<EV::Glib> and L<Glib::EV>).
		2306
		2307	=item L<Guard>
		2308
		2309	The guard module, when used, will be used to implement
		2310	C<AnyEvent::Util::guard>. This speeds up guards considerably (and uses a
		2311	lot less memory), but otherwise doesn't affect guard operation much. It is
		2312	purely used for performance.
		2313
		2314	=item L<JSON> and L<JSON::XS>
		2315
		2316	This module is required when you want to read or write JSON data via
		2317	L<AnyEvent::Handle>. It is also written in pure-perl, but can take
		2318	advantage of the ultra-high-speed L<JSON::XS> module when it is installed.
		2319
		2320	In fact, L<AnyEvent::Handle> will use L<JSON::XS> by default if it is
		2321	installed.
		2322
		2323	=item L<Net::SSLeay>
		2324
		2325	Implementing TLS/SSL in Perl is certainly interesting, but not very
		2326	worthwhile: If this module is installed, then L<AnyEvent::Handle> (with
		2327	the help of L<AnyEvent::TLS>), gains the ability to do TLS/SSL.
		2328
		2329	=item L<Time::HiRes>
		2330
		2331	This module is part of perl since release 5.008. It will be used when the
		2332	chosen event library does not come with a timing source on it's own. The
		2333	pure-perl event loop (L<AnyEvent::Impl::Perl>) will additionally use it to
		2334	try to use a monotonic clock for timing stability.
		2335
		2336	=back
		2337
		2338
		2339	=head1 FORK
		2340
		2341	Most event libraries are not fork-safe. The ones who are usually are
		2342	because they rely on inefficient but fork-safe C<select> or C<poll>
		2343	calls. Only L<EV> is fully fork-aware.
		2344
		2345	If you have to fork, you must either do so I<before> creating your first
		2346	watcher OR you must not use AnyEvent at all in the child OR you must do
		2347	something completely out of the scope of AnyEvent.
		2348
		2349
		2350	=head1 SECURITY CONSIDERATIONS
		2351
		2352	AnyEvent can be forced to load any event model via
		2353	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		2354	execute arbitrary code or directly gain access, it can easily be used to
		2355	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		2356	will not be active when the program uses a different event model than
		2357	specified in the variable.
		2358
		2359	You can make AnyEvent completely ignore this variable by deleting it
		2360	before the first watcher gets created, e.g. with a C<BEGIN> block:
		2361
		2362	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		2363
		2364	use AnyEvent;
		2365
		2366	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		2367	be used to probe what backend is used and gain other information (which is
		2368	probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
		2369	$ENV{PERL_ANYEVENT_STRICT}.
		2370
		2371	Note that AnyEvent will remove I<all> environment variables starting with
		2372	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		2373	enabled.
		2374
		2375
		2376	=head1 BUGS
		2377
		2378	Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
		2379	to work around. If you suffer from memleaks, first upgrade to Perl 5.10
		2380	and check wether the leaks still show up. (Perl 5.10.0 has other annoying
		2381	memleaks, such as leaking on C<map> and C<grep> but it is usually not as
		2382	pronounced).
		2383
672		2384
673	=head1 SEE ALSO	2385	=head1 SEE ALSO
674		2386
675	Event modules: L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>.	2387	Utility functions: L<AnyEvent::Util>.
676		2388
677	Implementations: L<AnyEvent::Impl::Coro>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>.	2389	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
		2390	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
678		2391
679	Nontrivial usage example: L<Net::FCP>.	2392	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		2393	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		2394	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		2395	L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>.
680		2396
681	=head1	2397	Non-blocking file handles, sockets, TCP clients and
		2398	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>.
		2399
		2400	Asynchronous DNS: L<AnyEvent::DNS>.
		2401
		2402	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>,
		2403	L<Coro::Event>,
		2404
		2405	Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::XMPP>,
		2406	L<AnyEvent::HTTP>.
		2407
		2408
		2409	=head1 AUTHOR
		2410
		2411	Marc Lehmann <schmorp@schmorp.de>
		2412	http://home.schmorp.de/
682		2413
683	=cut	2414	=cut
684		2415
685	1	2416	1
686		2417

Diff Legend

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