ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/README

Comparing AnyEvent/README (file contents):
Revision 1.17 by root, Tue Apr 22 05:12:19 2008 UTC vs.
Revision 1.62 by root, Sun Jun 6 10:13:57 2010 UTC

1	NAME	1	NAME
2	AnyEvent - provide framework for multiple event loops	2	AnyEvent - the DBI of event loop programming
3		3
4	EV, Event, Coro::EV, Coro::Event, Glib, Tk, Perl - various supported	4	EV, Event, Glib, Tk, Perl, Event::Lib, Irssi, rxvt-unicode, IO::Async,
5	event loops	5	Qt and POE are various supported event loops/environments.
6		6
7	SYNOPSIS	7	SYNOPSIS
8	use AnyEvent;	8	use AnyEvent;
9		9
		10	# if you prefer function calls, look at the AE manpage for
		11	# an alternative API.
		12
		13	# file handle or descriptor readable
10	my $w = AnyEvent->io (fh => $fh, poll => "r|w", cb => sub {	14	my $w = AnyEvent->io (fh => $fh, poll => "r", cb => sub { ... });
		15
		16	# one-shot or repeating timers
		17	my $w = AnyEvent->timer (after => $seconds, cb => sub { ... });
		18	my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ...
		19
		20	print AnyEvent->now; # prints current event loop time
		21	print AnyEvent->time; # think Time::HiRes::time or simply CORE::time.
		22
		23	# POSIX signal
		24	my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... });
		25
		26	# child process exit
		27	my $w = AnyEvent->child (pid => $pid, cb => sub {
		28	my ($pid, $status) = @_;
11	...	29	...
12	});	30	});
13		31
14	my $w = AnyEvent->timer (after => $seconds, cb => sub {	32	# called when event loop idle (if applicable)
15	...	33	my $w = AnyEvent->idle (cb => sub { ... });
16	});
17		34
18	my $w = AnyEvent->condvar; # stores whether a condition was flagged	35	my $w = AnyEvent->condvar; # stores whether a condition was flagged
		36	$w->send; # wake up current and all future recv's
19	$w->wait; # enters "main loop" till $condvar gets ->broadcast	37	$w->recv; # enters "main loop" till $condvar gets ->send
20	$w->broadcast; # wake up current and all future wait's	38	# use a condvar in callback mode:
		39	$w->cb (sub { $_[0]->recv });
		40
		41	INTRODUCTION/TUTORIAL
		42	This manpage is mainly a reference manual. If you are interested in a
		43	tutorial or some gentle introduction, have a look at the AnyEvent::Intro
		44	manpage.
		45
		46	SUPPORT
		47	There is a mailinglist for discussing all things AnyEvent, and an IRC
		48	channel, too.
		49
		50	See the AnyEvent project page at the Schmorpforge Ta-Sa Software
		51	Repository, at <http://anyevent.schmorp.de>, for more info.
21		52
22	WHY YOU SHOULD USE THIS MODULE (OR NOT)	53	WHY YOU SHOULD USE THIS MODULE (OR NOT)
23	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen	54	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
24	nowadays. So what is different about AnyEvent?	55	nowadays. So what is different about AnyEvent?
25		56
26	Executive Summary: AnyEvent is *compatible*, AnyEvent is *free of	57	Executive Summary: AnyEvent is *compatible*, AnyEvent is *free of
27	policy* and AnyEvent is *small and efficient*.	58	policy* and AnyEvent is *small and efficient*.
28		59
29	First and foremost, *AnyEvent is not an event model* itself, it only	60	First and foremost, *AnyEvent is not an event model* itself, it only
30	interfaces to whatever event model the main program happens to use in a	61	interfaces to whatever event model the main program happens to use, in a
31	pragmatic way. For event models and certain classes of immortals alike,	62	pragmatic way. For event models and certain classes of immortals alike,
32	the statement "there can only be one" is a bitter reality: In general,	63	the statement "there can only be one" is a bitter reality: In general,
33	only one event loop can be active at the same time in a process.	64	only one event loop can be active at the same time in a process.
34	AnyEvent helps hiding the differences between those event loops.	65	AnyEvent cannot change this, but it can hide the differences between
		66	those event loops.
35		67
36	The goal of AnyEvent is to offer module authors the ability to do event	68	The goal of AnyEvent is to offer module authors the ability to do event
37	programming (waiting for I/O or timer events) without subscribing to a	69	programming (waiting for I/O or timer events) without subscribing to a
38	religion, a way of living, and most importantly: without forcing your	70	religion, a way of living, and most importantly: without forcing your
39	module users into the same thing by forcing them to use the same event	71	module users into the same thing by forcing them to use the same event
40	model you use.	72	model you use.
41		73
42	For modules like POE or IO::Async (which is a total misnomer as it is	74	For modules like POE or IO::Async (which is a total misnomer as it is
43	actually doing all I/O *synchronously*...), using them in your module is	75	actually doing all I/O *synchronously*...), using them in your module is
44	like joining a cult: After you joined, you are dependent on them and you	76	like joining a cult: After you joined, you are dependent on them and you
45	cannot use anything else, as it is simply incompatible to everything	77	cannot use anything else, as they are simply incompatible to everything
46	that isn't itself. What's worse, all the potential users of your module	78	that isn't them. What's worse, all the potential users of your module
47	are *also* forced to use the same event loop you use.	79	are *also* forced to use the same event loop you use.
48		80
49	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works	81	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
50	fine. AnyEvent + Tk works fine etc. etc. but none of these work together	82	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
51	with the rest: POE + IO::Async? no go. Tk + Event? no go. Again: if your	83	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if your
52	module uses one of those, every user of your module has to use it, too.	84	module uses one of those, every user of your module has to use it, too.
53	But if your module uses AnyEvent, it works transparently with all event	85	But if your module uses AnyEvent, it works transparently with all event
54	models it supports (including stuff like POE and IO::Async, as long as	86	models it supports (including stuff like IO::Async, as long as those use
55	those use one of the supported event loops. It is trivial to add new	87	one of the supported event loops. It is trivial to add new event loops
56	event loops to AnyEvent, too, so it is future-proof).	88	to AnyEvent, too, so it is future-proof).
57		89
58	In addition to being free of having to use *the one and only true event	90	In addition to being free of having to use *the one and only true event
59	model*, AnyEvent also is free of bloat and policy: with POE or similar	91	model*, AnyEvent also is free of bloat and policy: with POE or similar
60	modules, you get an enourmous amount of code and strict rules you have	92	modules, you get an enormous amount of code and strict rules you have to
61	to follow. AnyEvent, on the other hand, is lean and up to the point, by	93	follow. AnyEvent, on the other hand, is lean and up to the point, by
62	only offering the functionality that is necessary, in as thin as a	94	only offering the functionality that is necessary, in as thin as a
63	wrapper as technically possible.	95	wrapper as technically possible.
64		96
		97	Of course, AnyEvent comes with a big (and fully optional!) toolbox of
		98	useful functionality, such as an asynchronous DNS resolver, 100%
		99	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		100	such as Windows) and lots of real-world knowledge and workarounds for
		101	platform bugs and differences.
		102
65	Of course, if you want lots of policy (this can arguably be somewhat	103	Now, if you *do want* lots of policy (this can arguably be somewhat
66	useful) and you want to force your users to use the one and only event	104	useful) and you want to force your users to use the one and only event
67	model, you should *not* use this module.	105	model, you should *not* use this module.
68		106
69	DESCRIPTION	107	DESCRIPTION
70	AnyEvent provides an identical interface to multiple event loops. This	108	AnyEvent provides an identical interface to multiple event loops. This
…		…
75	The interface itself is vaguely similar, but not identical to the Event	113	The interface itself is vaguely similar, but not identical to the Event
76	module.	114	module.
77		115
78	During the first call of any watcher-creation method, the module tries	116	During the first call of any watcher-creation method, the module tries
79	to detect the currently loaded event loop by probing whether one of the	117	to detect the currently loaded event loop by probing whether one of the
80	following modules is already loaded: Coro::EV, Coro::Event, EV, Event,	118	following modules is already loaded: EV, Event, Glib,
81	Glib, Tk. The first one found is used. If none are found, the module	119	AnyEvent::Impl::Perl, Tk, Event::Lib, Qt, POE. The first one found is
82	tries to load these modules in the stated order. The first one that can	120	used. If none are found, the module tries to load these modules
		121	(excluding Tk, Event::Lib, Qt and POE as the pure perl adaptor should
		122	always succeed) in the order given. The first one that can be
83	be successfully loaded will be used. If, after this, still none could be	123	successfully loaded will be used. If, after this, still none could be
84	found, AnyEvent will fall back to a pure-perl event loop, which is not	124	found, AnyEvent will fall back to a pure-perl event loop, which is not
85	very efficient, but should work everywhere.	125	very efficient, but should work everywhere.
86		126
87	Because AnyEvent first checks for modules that are already loaded,	127	Because AnyEvent first checks for modules that are already loaded,
88	loading an event model explicitly before first using AnyEvent will	128	loading an event model explicitly before first using AnyEvent will
…		…
97	starts using it, all bets are off. Maybe you should tell their authors	137	starts using it, all bets are off. Maybe you should tell their authors
98	to use AnyEvent so their modules work together with others seamlessly...	138	to use AnyEvent so their modules work together with others seamlessly...
99		139
100	The pure-perl implementation of AnyEvent is called	140	The pure-perl implementation of AnyEvent is called
101	"AnyEvent::Impl::Perl". Like other event modules you can load it	141	"AnyEvent::Impl::Perl". Like other event modules you can load it
102	explicitly.	142	explicitly and enjoy the high availability of that event loop :)
103		143
104	WATCHERS	144	WATCHERS
105	AnyEvent has the central concept of a *watcher*, which is an object that	145	AnyEvent has the central concept of a *watcher*, which is an object that
106	stores relevant data for each kind of event you are waiting for, such as	146	stores relevant data for each kind of event you are waiting for, such as
107	the callback to call, the filehandle to watch, etc.	147	the callback to call, the file handle to watch, etc.
108		148
109	These watchers are normal Perl objects with normal Perl lifetime. After	149	These watchers are normal Perl objects with normal Perl lifetime. After
110	creating a watcher it will immediately "watch" for events and invoke the	150	creating a watcher it will immediately "watch" for events and invoke the
111	callback when the event occurs (of course, only when the event model is	151	callback when the event occurs (of course, only when the event model is
112	in control).	152	in control).
113		153
		154	Note that callbacks must not permanently change global variables
		155	potentially in use by the event loop (such as $_ or $[) and that
		156	callbacks must not "die". The former is good programming practise in
		157	Perl and the latter stems from the fact that exception handling differs
		158	widely between event loops.
		159
114	To disable the watcher you have to destroy it (e.g. by setting the	160	To disable the watcher you have to destroy it (e.g. by setting the
115	variable you store it in to "undef" or otherwise deleting all references	161	variable you store it in to "undef" or otherwise deleting all references
116	to it).	162	to it).
117		163
118	All watchers are created by calling a method on the "AnyEvent" class.	164	All watchers are created by calling a method on the "AnyEvent" class.
…		…
120	Many watchers either are used with "recursion" (repeating timers for	166	Many watchers either are used with "recursion" (repeating timers for
121	example), or need to refer to their watcher object in other ways.	167	example), or need to refer to their watcher object in other ways.
122		168
123	An any way to achieve that is this pattern:	169	An any way to achieve that is this pattern:
124		170
125	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {	171	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
126	# you can use $w here, for example to undef it	172	# you can use $w here, for example to undef it
127	undef $w;	173	undef $w;
128	});	174	});
129		175
130	Note that "my $w; $w =" combination. This is necessary because in Perl,	176	Note that "my $w; $w =" combination. This is necessary because in Perl,
131	my variables are only visible after the statement in which they are	177	my variables are only visible after the statement in which they are
132	declared.	178	declared.
133		179
134	IO WATCHERS	180	I/O WATCHERS
		181	$w = AnyEvent->io (
		182	fh => <filehandle_or_fileno>,
		183	poll => <"r" or "w">,
		184	cb => <callback>,
		185	);
		186
135	You can create an I/O watcher by calling the "AnyEvent->io" method with	187	You can create an I/O watcher by calling the "AnyEvent->io" method with
136	the following mandatory key-value pairs as arguments:	188	the following mandatory key-value pairs as arguments:
137		189
138	"fh" the Perl *file handle* (*not* file descriptor) to watch for events.	190	"fh" is the Perl *file handle* (or a naked file descriptor) to watch for
		191	events (AnyEvent might or might not keep a reference to this file
		192	handle). Note that only file handles pointing to things for which
		193	non-blocking operation makes sense are allowed. This includes sockets,
		194	most character devices, pipes, fifos and so on, but not for example
		195	files or block devices.
		196
139	"poll" must be a string that is either "r" or "w", which creates a	197	"poll" must be a string that is either "r" or "w", which creates a
140	watcher waiting for "r"eadable or "w"ritable events, respectively. "cb"	198	watcher waiting for "r"eadable or "w"ritable events, respectively.
		199
141	is the callback to invoke each time the file handle becomes ready.	200	"cb" is the callback to invoke each time the file handle becomes ready.
142		201
143	File handles will be kept alive, so as long as the watcher exists, the	202	Although the callback might get passed parameters, their value and
144	file handle exists, too.	203	presence is undefined and you cannot rely on them. Portable AnyEvent
		204	callbacks cannot use arguments passed to I/O watcher callbacks.
145		205
		206	The I/O watcher might use the underlying file descriptor or a copy of
146	It is not allowed to close a file handle as long as any watcher is	207	it. You must not close a file handle as long as any watcher is active on
147	active on the underlying file descriptor.	208	the underlying file descriptor.
148		209
149	Some event loops issue spurious readyness notifications, so you should	210	Some event loops issue spurious readyness notifications, so you should
150	always use non-blocking calls when reading/writing from/to your file	211	always use non-blocking calls when reading/writing from/to your file
151	handles.	212	handles.
152		213
153	Example:
154
155	# wait for readability of STDIN, then read a line and disable the watcher	214	Example: wait for readability of STDIN, then read a line and disable the
		215	watcher.
		216
156	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	217	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
157	chomp (my $input = <STDIN>);	218	chomp (my $input = <STDIN>);
158	warn "read: $input\n";	219	warn "read: $input\n";
159	undef $w;	220	undef $w;
160	});	221	});
161		222
162	TIME WATCHERS	223	TIME WATCHERS
		224	$w = AnyEvent->timer (after => <seconds>, cb => <callback>);
		225
		226	$w = AnyEvent->timer (
		227	after => <fractional_seconds>,
		228	interval => <fractional_seconds>,
		229	cb => <callback>,
		230	);
		231
163	You can create a time watcher by calling the "AnyEvent->timer" method	232	You can create a time watcher by calling the "AnyEvent->timer" method
164	with the following mandatory arguments:	233	with the following mandatory arguments:
165		234
166	"after" specifies after how many seconds (fractional values are	235	"after" specifies after how many seconds (fractional values are
167	supported) should the timer activate. "cb" the callback to invoke in	236	supported) the callback should be invoked. "cb" is the callback to
168	that case.	237	invoke in that case.
169		238
170	The timer callback will be invoked at most once: if you want a repeating	239	Although the callback might get passed parameters, their value and
171	timer you have to create a new watcher (this is a limitation by both Tk	240	presence is undefined and you cannot rely on them. Portable AnyEvent
172	and Glib).	241	callbacks cannot use arguments passed to time watcher callbacks.
173		242
174	Example:	243	The callback will normally be invoked once only. If you specify another
		244	parameter, "interval", as a strictly positive number (> 0), then the
		245	callback will be invoked regularly at that interval (in fractional
		246	seconds) after the first invocation. If "interval" is specified with a
		247	false value, then it is treated as if it were missing.
175		248
		249	The callback will be rescheduled before invoking the callback, but no
		250	attempt is done to avoid timer drift in most backends, so the interval
		251	is only approximate.
		252
176	# fire an event after 7.7 seconds	253	Example: fire an event after 7.7 seconds.
		254
177	my $w = AnyEvent->timer (after => 7.7, cb => sub {	255	my $w = AnyEvent->timer (after => 7.7, cb => sub {
178	warn "timeout\n";	256	warn "timeout\n";
179	});	257	});
180		258
181	# to cancel the timer:	259	# to cancel the timer:
182	undef $w;	260	undef $w;
183		261
184	Example 2:
185
186	# fire an event after 0.5 seconds, then roughly every second	262	Example 2: fire an event after 0.5 seconds, then roughly every second.
187	my $w;
188		263
189	my $cb = sub {
190	# cancel the old timer while creating a new one
191	$w = AnyEvent->timer (after => 1, cb => $cb);	264	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
		265	warn "timeout\n";
192	};	266	};
193
194	# start the "loop" by creating the first watcher
195	$w = AnyEvent->timer (after => 0.5, cb => $cb);
196		267
197	TIMING ISSUES	268	TIMING ISSUES
198	There are two ways to handle timers: based on real time (relative, "fire	269	There are two ways to handle timers: based on real time (relative, "fire
199	in 10 seconds") and based on wallclock time (absolute, "fire at 12	270	in 10 seconds") and based on wallclock time (absolute, "fire at 12
200	o'clock").	271	o'clock").
201		272
202	While most event loops expect timers to specified in a relative way,	273	While most event loops expect timers to specified in a relative way,
203	they use absolute time internally. This makes a difference when your	274	they use absolute time internally. This makes a difference when your
204	clock "jumps", for example, when ntp decides to set your clock backwards	275	clock "jumps", for example, when ntp decides to set your clock backwards
205	from the wrong 2014-01-01 to 2008-01-01, a watcher that you created to	276	from the wrong date of 2014-01-01 to 2008-01-01, a watcher that is
206	fire "after" a second might actually take six years to finally fire.	277	supposed to fire "after" a second might actually take six years to
		278	finally fire.
207		279
208	AnyEvent cannot compensate for this. The only event loop that is	280	AnyEvent cannot compensate for this. The only event loop that is
209	conscious about these issues is EV, which offers both relative	281	conscious about these issues is EV, which offers both relative
210	(ev_timer) and absolute (ev_periodic) timers.	282	(ev_timer, based on true relative time) and absolute (ev_periodic, based
		283	on wallclock time) timers.
211		284
212	AnyEvent always prefers relative timers, if available, matching the	285	AnyEvent always prefers relative timers, if available, matching the
213	AnyEvent API.	286	AnyEvent API.
214		287
		288	AnyEvent has two additional methods that return the "current time":
		289
		290	AnyEvent->time
		291	This returns the "current wallclock time" as a fractional number of
		292	seconds since the Epoch (the same thing as "time" or
		293	"Time::HiRes::time" return, and the result is guaranteed to be
		294	compatible with those).
		295
		296	It progresses independently of any event loop processing, i.e. each
		297	call will check the system clock, which usually gets updated
		298	frequently.
		299
		300	AnyEvent->now
		301	This also returns the "current wallclock time", but unlike "time",
		302	above, this value might change only once per event loop iteration,
		303	depending on the event loop (most return the same time as "time",
		304	above). This is the time that AnyEvent's timers get scheduled
		305	against.
		306
		307	*In almost all cases (in all cases if you don't care), this is the
		308	function to call when you want to know the current time.*
		309
		310	This function is also often faster then "AnyEvent->time", and thus
		311	the preferred method if you want some timestamp (for example,
		312	AnyEvent::Handle uses this to update it's activity timeouts).
		313
		314	The rest of this section is only of relevance if you try to be very
		315	exact with your timing, you can skip it without bad conscience.
		316
		317	For a practical example of when these times differ, consider
		318	Event::Lib and EV and the following set-up:
		319
		320	The event loop is running and has just invoked one of your callback
		321	at time=500 (assume no other callbacks delay processing). In your
		322	callback, you wait a second by executing "sleep 1" (blocking the
		323	process for a second) and then (at time=501) you create a relative
		324	timer that fires after three seconds.
		325
		326	With Event::Lib, "AnyEvent->time" and "AnyEvent->now" will both
		327	return 501, because that is the current time, and the timer will be
		328	scheduled to fire at time=504 (501 + 3).
		329
		330	With EV, "AnyEvent->time" returns 501 (as that is the current time),
		331	but "AnyEvent->now" returns 500, as that is the time the last event
		332	processing phase started. With EV, your timer gets scheduled to run
		333	at time=503 (500 + 3).
		334
		335	In one sense, Event::Lib is more exact, as it uses the current time
		336	regardless of any delays introduced by event processing. However,
		337	most callbacks do not expect large delays in processing, so this
		338	causes a higher drift (and a lot more system calls to get the
		339	current time).
		340
		341	In another sense, EV is more exact, as your timer will be scheduled
		342	at the same time, regardless of how long event processing actually
		343	took.
		344
		345	In either case, if you care (and in most cases, you don't), then you
		346	can get whatever behaviour you want with any event loop, by taking
		347	the difference between "AnyEvent->time" and "AnyEvent->now" into
		348	account.
		349
		350	AnyEvent->now_update
		351	Some event loops (such as EV or AnyEvent::Impl::Perl) cache the
		352	current time for each loop iteration (see the discussion of
		353	AnyEvent->now, above).
		354
		355	When a callback runs for a long time (or when the process sleeps),
		356	then this "current" time will differ substantially from the real
		357	time, which might affect timers and time-outs.
		358
		359	When this is the case, you can call this method, which will update
		360	the event loop's idea of "current time".
		361
		362	A typical example would be a script in a web server (e.g.
		363	"mod_perl") - when mod_perl executes the script, then the event loop
		364	will have the wrong idea about the "current time" (being potentially
		365	far in the past, when the script ran the last time). In that case
		366	you should arrange a call to "AnyEvent->now_update" each time the
		367	web server process wakes up again (e.g. at the start of your script,
		368	or in a handler).
		369
		370	Note that updating the time *might* cause some events to be handled.
		371
215	SIGNAL WATCHERS	372	SIGNAL WATCHERS
		373	$w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>);
		374
216	You can watch for signals using a signal watcher, "signal" is the signal	375	You can watch for signals using a signal watcher, "signal" is the signal
217	*name* without any "SIG" prefix, "cb" is the Perl callback to be invoked	376	*name* in uppercase and without any "SIG" prefix, "cb" is the Perl
218	whenever a signal occurs.	377	callback to be invoked whenever a signal occurs.
219		378
		379	Although the callback might get passed parameters, their value and
		380	presence is undefined and you cannot rely on them. Portable AnyEvent
		381	callbacks cannot use arguments passed to signal watcher callbacks.
		382
220	Multiple signals occurances can be clumped together into one callback	383	Multiple signal occurrences can be clumped together into one callback
221	invocation, and callback invocation will be synchronous. synchronous	384	invocation, and callback invocation will be synchronous. Synchronous
222	means that it might take a while until the signal gets handled by the	385	means that it might take a while until the signal gets handled by the
223	process, but it is guarenteed not to interrupt any other callbacks.	386	process, but it is guaranteed not to interrupt any other callbacks.
224		387
225	The main advantage of using these watchers is that you can share a	388	The main advantage of using these watchers is that you can share a
226	signal between multiple watchers.	389	signal between multiple watchers, and AnyEvent will ensure that signals
		390	will not interrupt your program at bad times.
227		391
228	This watcher might use %SIG, so programs overwriting those signals	392	This watcher might use %SIG (depending on the event loop used), so
229	directly will likely not work correctly.	393	programs overwriting those signals directly will likely not work
		394	correctly.
230		395
231	Example: exit on SIGINT	396	Example: exit on SIGINT
232		397
233	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });	398	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
234		399
		400	Restart Behaviour
		401	While restart behaviour is up to the event loop implementation, most
		402	will not restart syscalls (that includes Async::Interrupt and AnyEvent's
		403	pure perl implementation).
		404
		405	Safe/Unsafe Signals
		406	Perl signals can be either "safe" (synchronous to opcode handling) or
		407	"unsafe" (asynchronous) - the former might get delayed indefinitely, the
		408	latter might corrupt your memory.
		409
		410	AnyEvent signal handlers are, in addition, synchronous to the event
		411	loop, i.e. they will not interrupt your running perl program but will
		412	only be called as part of the normal event handling (just like timer,
		413	I/O etc. callbacks, too).
		414
		415	Signal Races, Delays and Workarounds
		416	Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
		417	callbacks to signals in a generic way, which is a pity, as you cannot do
		418	race-free signal handling in perl, requiring C libraries for this.
		419	AnyEvent will try to do it's best, which means in some cases, signals
		420	will be delayed. The maximum time a signal might be delayed is specified
		421	in $AnyEvent::MAX_SIGNAL_LATENCY (default: 10 seconds). This variable
		422	can be changed only before the first signal watcher is created, and
		423	should be left alone otherwise. This variable determines how often
		424	AnyEvent polls for signals (in case a wake-up was missed). Higher values
		425	will cause fewer spurious wake-ups, which is better for power and CPU
		426	saving.
		427
		428	All these problems can be avoided by installing the optional
		429	Async::Interrupt module, which works with most event loops. It will not
		430	work with inherently broken event loops such as Event or Event::Lib (and
		431	not with POE currently, as POE does it's own workaround with one-second
		432	latency). For those, you just have to suffer the delays.
		433
235	CHILD PROCESS WATCHERS	434	CHILD PROCESS WATCHERS
		435	$w = AnyEvent->child (pid => <process id>, cb => <callback>);
		436
236	You can also watch on a child process exit and catch its exit status.	437	You can also watch on a child process exit and catch its exit status.
237		438
238	The child process is specified by the "pid" argument (if set to 0, it	439	The child process is specified by the "pid" argument (one some backends,
239	watches for any child process exit). The watcher will trigger as often	440	using 0 watches for any child process exit, on others this will croak).
240	as status change for the child are received. This works by installing a	441	The watcher will be triggered only when the child process has finished
241	signal handler for "SIGCHLD". The callback will be called with the pid	442	and an exit status is available, not on any trace events
242	and exit status (as returned by waitpid).	443	(stopped/continued).
243		444
244	Example: wait for pid 1333	445	The callback will be called with the pid and exit status (as returned by
		446	waitpid), so unlike other watcher types, you *can* rely on child watcher
		447	callback arguments.
245		448
		449	This watcher type works by installing a signal handler for "SIGCHLD",
		450	and since it cannot be shared, nothing else should use SIGCHLD or reap
		451	random child processes (waiting for specific child processes, e.g.
		452	inside "system", is just fine).
		453
		454	There is a slight catch to child watchers, however: you usually start
		455	them *after* the child process was created, and this means the process
		456	could have exited already (and no SIGCHLD will be sent anymore).
		457
		458	Not all event models handle this correctly (neither POE nor IO::Async
		459	do, see their AnyEvent::Impl manpages for details), but even for event
		460	models that *do* handle this correctly, they usually need to be loaded
		461	before the process exits (i.e. before you fork in the first place).
		462	AnyEvent's pure perl event loop handles all cases correctly regardless
		463	of when you start the watcher.
		464
		465	This means you cannot create a child watcher as the very first thing in
		466	an AnyEvent program, you *have* to create at least one watcher before
		467	you "fork" the child (alternatively, you can call "AnyEvent::detect").
		468
		469	As most event loops do not support waiting for child events, they will
		470	be emulated by AnyEvent in most cases, in which the latency and race
		471	problems mentioned in the description of signal watchers apply.
		472
		473	Example: fork a process and wait for it
		474
		475	my $done = AnyEvent->condvar;
		476
		477	my $pid = fork or exit 5;
		478
246	my $w = AnyEvent->child (	479	my $w = AnyEvent->child (
247	pid => 1333,	480	pid => $pid,
248	cb => sub {	481	cb => sub {
249	my ($pid, $status) = @_;	482	my ($pid, $status) = @_;
250	warn "pid $pid exited with status $status";	483	warn "pid $pid exited with status $status";
		484	$done->send;
251	},	485	},
252	);	486	);
		487
		488	# do something else, then wait for process exit
		489	$done->recv;
		490
		491	IDLE WATCHERS
		492	$w = AnyEvent->idle (cb => <callback>);
		493
		494	Repeatedly invoke the callback after the process becomes idle, until
		495	either the watcher is destroyed or new events have been detected.
		496
		497	Idle watchers are useful when there is a need to do something, but it is
		498	not so important (or wise) to do it instantly. The callback will be
		499	invoked only when there is "nothing better to do", which is usually
		500	defined as "all outstanding events have been handled and no new events
		501	have been detected". That means that idle watchers ideally get invoked
		502	when the event loop has just polled for new events but none have been
		503	detected. Instead of blocking to wait for more events, the idle watchers
		504	will be invoked.
		505
		506	Unfortunately, most event loops do not really support idle watchers
		507	(only EV, Event and Glib do it in a usable fashion) - for the rest,
		508	AnyEvent will simply call the callback "from time to time".
		509
		510	Example: read lines from STDIN, but only process them when the program
		511	is otherwise idle:
		512
		513	my @lines; # read data
		514	my $idle_w;
		515	my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		516	push @lines, scalar <STDIN>;
		517
		518	# start an idle watcher, if not already done
		519	$idle_w ||= AnyEvent->idle (cb => sub {
		520	# handle only one line, when there are lines left
		521	if (my $line = shift @lines) {
		522	print "handled when idle: $line";
		523	} else {
		524	# otherwise disable the idle watcher again
		525	undef $idle_w;
		526	}
		527	});
		528	});
253		529
254	CONDITION VARIABLES	530	CONDITION VARIABLES
		531	$cv = AnyEvent->condvar;
		532
		533	$cv->send (<list>);
		534	my @res = $cv->recv;
		535
		536	If you are familiar with some event loops you will know that all of them
		537	require you to run some blocking "loop", "run" or similar function that
		538	will actively watch for new events and call your callbacks.
		539
		540	AnyEvent is slightly different: it expects somebody else to run the
		541	event loop and will only block when necessary (usually when told by the
		542	user).
		543
		544	The tool to do that is called a "condition variable", so called because
		545	they represent a condition that must become true.
		546
		547	Now is probably a good time to look at the examples further below.
		548
255	Condition variables can be created by calling the "AnyEvent->condvar"	549	Condition variables can be created by calling the "AnyEvent->condvar"
256	method without any arguments.	550	method, usually without arguments. The only argument pair allowed is
		551	"cb", which specifies a callback to be called when the condition
		552	variable becomes true, with the condition variable as the first argument
		553	(but not the results).
257		554
258	A condition variable waits for a condition - precisely that the	555	After creation, the condition variable is "false" until it becomes
259	"->broadcast" method has been called.	556	"true" by calling the "send" method (or calling the condition variable
		557	as if it were a callback, read about the caveats in the description for
		558	the "->send" method).
260		559
261	They are very useful to signal that a condition has been fulfilled, for	560	Since condition variables are the most complex part of the AnyEvent API,
		561	here are some different mental models of what they are - pick the ones
		562	you can connect to:
		563
		564	* Condition variables are like callbacks - you can call them (and pass
		565	them instead of callbacks). Unlike callbacks however, you can also
		566	wait for them to be called.
		567
		568	* Condition variables are signals - one side can emit or send them,
		569	the other side can wait for them, or install a handler that is
		570	called when the signal fires.
		571
		572	* Condition variables are like "Merge Points" - points in your program
		573	where you merge multiple independent results/control flows into one.
		574
		575	* Condition variables represent a transaction - function that start
		576	some kind of transaction can return them, leaving the caller the
		577	choice between waiting in a blocking fashion, or setting a callback.
		578
		579	* Condition variables represent future values, or promises to deliver
		580	some result, long before the result is available.
		581
		582	Condition variables are very useful to signal that something has
262	example, if you write a module that does asynchronous http requests,	583	finished, for example, if you write a module that does asynchronous http
263	then a condition variable would be the ideal candidate to signal the	584	requests, then a condition variable would be the ideal candidate to
264	availability of results.	585	signal the availability of results. The user can either act when the
		586	callback is called or can synchronously "->recv" for the results.
265		587
266	You can also use condition variables to block your main program until an	588	You can also use them to simulate traditional event loops - for example,
267	event occurs - for example, you could "->wait" in your main program	589	you can block your main program until an event occurs - for example, you
268	until the user clicks the Quit button in your app, which would	590	could "->recv" in your main program until the user clicks the Quit
269	"->broadcast" the "quit" event.	591	button of your app, which would "->send" the "quit" event.
270		592
271	Note that condition variables recurse into the event loop - if you have	593	Note that condition variables recurse into the event loop - if you have
272	two pirces of code that call "->wait" in a round-robbin fashion, you	594	two pieces of code that call "->recv" in a round-robin fashion, you
273	lose. Therefore, condition variables are good to export to your caller,	595	lose. Therefore, condition variables are good to export to your caller,
274	but you should avoid making a blocking wait yourself, at least in	596	but you should avoid making a blocking wait yourself, at least in
275	callbacks, as this asks for trouble.	597	callbacks, as this asks for trouble.
276		598
277	This object has two methods:	599	Condition variables are represented by hash refs in perl, and the keys
		600	used by AnyEvent itself are all named "_ae_XXX" to make subclassing easy
		601	(it is often useful to build your own transaction class on top of
		602	AnyEvent). To subclass, use "AnyEvent::CondVar" as base class and call
		603	it's "new" method in your own "new" method.
278		604
279	$cv->wait	605	There are two "sides" to a condition variable - the "producer side"
280	Wait (blocking if necessary) until the "->broadcast" method has been	606	which eventually calls "-> send", and the "consumer side", which waits
281	called on c<$cv>, while servicing other watchers normally.	607	for the send to occur.
282		608
283	You can only wait once on a condition - additional calls will return	609	Example: wait for a timer.
284	immediately.
285		610
286	Not all event models support a blocking wait - some die in that case	611	# condition: "wait till the timer is fired"
287	(programs might want to do that to stay interactive), so *if you are
288	using this from a module, never require a blocking wait*, but let
289	the caller decide whether the call will block or not (for example,
290	by coupling condition variables with some kind of request results
291	and supporting callbacks so the caller knows that getting the result
292	will not block, while still suppporting blocking waits if the caller
293	so desires).
294
295	Another reason *never* to "->wait" in a module is that you cannot
296	sensibly have two "->wait"'s in parallel, as that would require
297	multiple interpreters or coroutines/threads, none of which
298	"AnyEvent" can supply (the coroutine-aware backends
299	AnyEvent::Impl::CoroEV and AnyEvent::Impl::CoroEvent explicitly
300	support concurrent "->wait"'s from different coroutines, however).
301
302	$cv->broadcast
303	Flag the condition as ready - a running "->wait" and all further
304	calls to "wait" will (eventually) return after this method has been
305	called. If nobody is waiting the broadcast will be remembered..
306
307	Example:
308
309	# wait till the result is ready
310	my $result_ready = AnyEvent->condvar;	612	my $timer_fired = AnyEvent->condvar;
311		613
312	# do something such as adding a timer	614	# create the timer - we could wait for, say
313	# or socket watcher the calls $result_ready->broadcast	615	# a handle becomign ready, or even an
314	# when the "result" is ready.	616	# AnyEvent::HTTP request to finish, but
315	# in this case, we simply use a timer:	617	# in this case, we simply use a timer:
316	my $w = AnyEvent->timer (	618	my $w = AnyEvent->timer (
317	after => 1,	619	after => 1,
318	cb => sub { $result_ready->broadcast },	620	cb => sub { $timer_fired->send },
319	);	621	);
320		622
321	# this "blocks" (while handling events) till the watcher	623	# this "blocks" (while handling events) till the callback
322	# calls broadcast	624	# calls ->send
323	$result_ready->wait;	625	$timer_fired->recv;
		626
		627	Example: wait for a timer, but take advantage of the fact that condition
		628	variables are also callable directly.
		629
		630	my $done = AnyEvent->condvar;
		631	my $delay = AnyEvent->timer (after => 5, cb => $done);
		632	$done->recv;
		633
		634	Example: Imagine an API that returns a condvar and doesn't support
		635	callbacks. This is how you make a synchronous call, for example from the
		636	main program:
		637
		638	use AnyEvent::CouchDB;
		639
		640	...
		641
		642	my @info = $couchdb->info->recv;
		643
		644	And this is how you would just set a callback to be called whenever the
		645	results are available:
		646
		647	$couchdb->info->cb (sub {
		648	my @info = $_[0]->recv;
		649	});
		650
		651	METHODS FOR PRODUCERS
		652	These methods should only be used by the producing side, i.e. the
		653	code/module that eventually sends the signal. Note that it is also the
		654	producer side which creates the condvar in most cases, but it isn't
		655	uncommon for the consumer to create it as well.
		656
		657	$cv->send (...)
		658	Flag the condition as ready - a running "->recv" and all further
		659	calls to "recv" will (eventually) return after this method has been
		660	called. If nobody is waiting the send will be remembered.
		661
		662	If a callback has been set on the condition variable, it is called
		663	immediately from within send.
		664
		665	Any arguments passed to the "send" call will be returned by all
		666	future "->recv" calls.
		667
		668	Condition variables are overloaded so one can call them directly (as
		669	if they were a code reference). Calling them directly is the same as
		670	calling "send".
		671
		672	$cv->croak ($error)
		673	Similar to send, but causes all call's to "->recv" to invoke
		674	"Carp::croak" with the given error message/object/scalar.
		675
		676	This can be used to signal any errors to the condition variable
		677	user/consumer. Doing it this way instead of calling "croak" directly
		678	delays the error detetcion, but has the overwhelmign advantage that
		679	it diagnoses the error at the place where the result is expected,
		680	and not deep in some event clalback without connection to the actual
		681	code causing the problem.
		682
		683	$cv->begin ([group callback])
		684	$cv->end
		685	These two methods can be used to combine many transactions/events
		686	into one. For example, a function that pings many hosts in parallel
		687	might want to use a condition variable for the whole process.
		688
		689	Every call to "->begin" will increment a counter, and every call to
		690	"->end" will decrement it. If the counter reaches 0 in "->end", the
		691	(last) callback passed to "begin" will be executed, passing the
		692	condvar as first argument. That callback is *supposed* to call
		693	"->send", but that is not required. If no group callback was set,
		694	"send" will be called without any arguments.
		695
		696	You can think of "$cv->send" giving you an OR condition (one call
		697	sends), while "$cv->begin" and "$cv->end" giving you an AND
		698	condition (all "begin" calls must be "end"'ed before the condvar
		699	sends).
		700
		701	Let's start with a simple example: you have two I/O watchers (for
		702	example, STDOUT and STDERR for a program), and you want to wait for
		703	both streams to close before activating a condvar:
		704
		705	my $cv = AnyEvent->condvar;
		706
		707	$cv->begin; # first watcher
		708	my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
		709	defined sysread $fh1, my $buf, 4096
		710	or $cv->end;
		711	});
		712
		713	$cv->begin; # second watcher
		714	my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
		715	defined sysread $fh2, my $buf, 4096
		716	or $cv->end;
		717	});
		718
		719	$cv->recv;
		720
		721	This works because for every event source (EOF on file handle),
		722	there is one call to "begin", so the condvar waits for all calls to
		723	"end" before sending.
		724
		725	The ping example mentioned above is slightly more complicated, as
		726	the there are results to be passwd back, and the number of tasks
		727	that are begung can potentially be zero:
		728
		729	my $cv = AnyEvent->condvar;
		730
		731	my %result;
		732	$cv->begin (sub { shift->send (\%result) });
		733
		734	for my $host (@list_of_hosts) {
		735	$cv->begin;
		736	ping_host_then_call_callback $host, sub {
		737	$result{$host} = ...;
		738	$cv->end;
		739	};
		740	}
		741
		742	$cv->end;
		743
		744	This code fragment supposedly pings a number of hosts and calls
		745	"send" after results for all then have have been gathered - in any
		746	order. To achieve this, the code issues a call to "begin" when it
		747	starts each ping request and calls "end" when it has received some
		748	result for it. Since "begin" and "end" only maintain a counter, the
		749	order in which results arrive is not relevant.
		750
		751	There is an additional bracketing call to "begin" and "end" outside
		752	the loop, which serves two important purposes: first, it sets the
		753	callback to be called once the counter reaches 0, and second, it
		754	ensures that "send" is called even when "no" hosts are being pinged
		755	(the loop doesn't execute once).
		756
		757	This is the general pattern when you "fan out" into multiple (but
		758	potentially none) subrequests: use an outer "begin"/"end" pair to
		759	set the callback and ensure "end" is called at least once, and then,
		760	for each subrequest you start, call "begin" and for each subrequest
		761	you finish, call "end".
		762
		763	METHODS FOR CONSUMERS
		764	These methods should only be used by the consuming side, i.e. the code
		765	awaits the condition.
		766
		767	$cv->recv
		768	Wait (blocking if necessary) until the "->send" or "->croak" methods
		769	have been called on c<$cv>, while servicing other watchers normally.
		770
		771	You can only wait once on a condition - additional calls are valid
		772	but will return immediately.
		773
		774	If an error condition has been set by calling "->croak", then this
		775	function will call "croak".
		776
		777	In list context, all parameters passed to "send" will be returned,
		778	in scalar context only the first one will be returned.
		779
		780	Note that doing a blocking wait in a callback is not supported by
		781	any event loop, that is, recursive invocation of a blocking "->recv"
		782	is not allowed, and the "recv" call will "croak" if such a condition
		783	is detected. This condition can be slightly loosened by using
		784	Coro::AnyEvent, which allows you to do a blocking "->recv" from any
		785	thread that doesn't run the event loop itself.
		786
		787	Not all event models support a blocking wait - some die in that case
		788	(programs might want to do that to stay interactive), so *if you are
		789	using this from a module, never require a blocking wait*. Instead,
		790	let the caller decide whether the call will block or not (for
		791	example, by coupling condition variables with some kind of request
		792	results and supporting callbacks so the caller knows that getting
		793	the result will not block, while still supporting blocking waits if
		794	the caller so desires).
		795
		796	You can ensure that "-recv" never blocks by setting a callback and
		797	only calling "->recv" from within that callback (or at a later
		798	time). This will work even when the event loop does not support
		799	blocking waits otherwise.
		800
		801	$bool = $cv->ready
		802	Returns true when the condition is "true", i.e. whether "send" or
		803	"croak" have been called.
		804
		805	$cb = $cv->cb ($cb->($cv))
		806	This is a mutator function that returns the callback set and
		807	optionally replaces it before doing so.
		808
		809	The callback will be called when the condition becomes (or already
		810	was) "true", i.e. when "send" or "croak" are called (or were
		811	called), with the only argument being the condition variable itself.
		812	Calling "recv" inside the callback or at any later time is
		813	guaranteed not to block.
		814
		815	SUPPORTED EVENT LOOPS/BACKENDS
		816	The available backend classes are (every class has its own manpage):
		817
		818	Backends that are autoprobed when no other event loop can be found.
		819	EV is the preferred backend when no other event loop seems to be in
		820	use. If EV is not installed, then AnyEvent will fall back to its own
		821	pure-perl implementation, which is available everywhere as it comes
		822	with AnyEvent itself.
		823
		824	AnyEvent::Impl::EV based on EV (interface to libev, best choice).
		825	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		826
		827	Backends that are transparently being picked up when they are used.
		828	These will be used when they are currently loaded when the first
		829	watcher is created, in which case it is assumed that the application
		830	is using them. This means that AnyEvent will automatically pick the
		831	right backend when the main program loads an event module before
		832	anything starts to create watchers. Nothing special needs to be done
		833	by the main program.
		834
		835	AnyEvent::Impl::Event based on Event, very stable, few glitches.
		836	AnyEvent::Impl::Glib based on Glib, slow but very stable.
		837	AnyEvent::Impl::Tk based on Tk, very broken.
		838	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		839	AnyEvent::Impl::POE based on POE, very slow, some limitations.
		840	AnyEvent::Impl::Irssi used when running within irssi.
		841
		842	Backends with special needs.
		843	Qt requires the Qt::Application to be instantiated first, but will
		844	otherwise be picked up automatically. As long as the main program
		845	instantiates the application before any AnyEvent watchers are
		846	created, everything should just work.
		847
		848	AnyEvent::Impl::Qt based on Qt.
		849
		850	Support for IO::Async can only be partial, as it is too broken and
		851	architecturally limited to even support the AnyEvent API. It also is
		852	the only event loop that needs the loop to be set explicitly, so it
		853	can only be used by a main program knowing about AnyEvent. See
		854	AnyEvent::Impl::Async for the gory details.
		855
		856	AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed.
		857
		858	Event loops that are indirectly supported via other backends.
		859	Some event loops can be supported via other modules:
		860
		861	There is no direct support for WxWidgets (Wx) or Prima.
		862
		863	WxWidgets has no support for watching file handles. However, you can
		864	use WxWidgets through the POE adaptor, as POE has a Wx backend that
		865	simply polls 20 times per second, which was considered to be too
		866	horrible to even consider for AnyEvent.
		867
		868	Prima is not supported as nobody seems to be using it, but it has a
		869	POE backend, so it can be supported through POE.
		870
		871	AnyEvent knows about both Prima and Wx, however, and will try to
		872	load POE when detecting them, in the hope that POE will pick them
		873	up, in which case everything will be automatic.
324		874
325	GLOBAL VARIABLES AND FUNCTIONS	875	GLOBAL VARIABLES AND FUNCTIONS
		876	These are not normally required to use AnyEvent, but can be useful to
		877	write AnyEvent extension modules.
		878
326	$AnyEvent::MODEL	879	$AnyEvent::MODEL
327	Contains "undef" until the first watcher is being created. Then it	880	Contains "undef" until the first watcher is being created, before
		881	the backend has been autodetected.
		882
328	contains the event model that is being used, which is the name of	883	Afterwards it contains the event model that is being used, which is
329	the Perl class implementing the model. This class is usually one of	884	the name of the Perl class implementing the model. This class is
330	the "AnyEvent::Impl:xxx" modules, but can be any other class in the	885	usually one of the "AnyEvent::Impl:xxx" modules, but can be any
331	case AnyEvent has been extended at runtime (e.g. in *rxvt-unicode*).	886	other class in the case AnyEvent has been extended at runtime (e.g.
332		887	in *rxvt-unicode* it will be "urxvt::anyevent").
333	The known classes so far are:
334
335	AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
336	AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
337	AnyEvent::Impl::EV based on EV (an interface to libev, also best choice).
338	AnyEvent::Impl::Event based on Event, also second best choice :)
339	AnyEvent::Impl::Glib based on Glib, third-best choice.
340	AnyEvent::Impl::Tk based on Tk, very bad choice.
341	AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable.
342		888
343	AnyEvent::detect	889	AnyEvent::detect
344	Returns $AnyEvent::MODEL, forcing autodetection of the event model	890	Returns $AnyEvent::MODEL, forcing autodetection of the event model
345	if necessary. You should only call this function right before you	891	if necessary. You should only call this function right before you
346	would have created an AnyEvent watcher anyway, that is, as late as	892	would have created an AnyEvent watcher anyway, that is, as late as
347	possible at runtime.	893	possible at runtime, and not e.g. while initialising of your module.
		894
		895	If you need to do some initialisation before AnyEvent watchers are
		896	created, use "post_detect".
		897
		898	$guard = AnyEvent::post_detect { BLOCK }
		899	Arranges for the code block to be executed as soon as the event
		900	model is autodetected (or immediately if this has already happened).
		901
		902	The block will be executed *after* the actual backend has been
		903	detected ($AnyEvent::MODEL is set), but *before* any watchers have
		904	been created, so it is possible to e.g. patch @AnyEvent::ISA or do
		905	other initialisations - see the sources of AnyEvent::Strict or
		906	AnyEvent::AIO to see how this is used.
		907
		908	The most common usage is to create some global watchers, without
		909	forcing event module detection too early, for example, AnyEvent::AIO
		910	creates and installs the global IO::AIO watcher in a "post_detect"
		911	block to avoid autodetecting the event module at load time.
		912
		913	If called in scalar or list context, then it creates and returns an
		914	object that automatically removes the callback again when it is
		915	destroyed (or "undef" when the hook was immediately executed). See
		916	AnyEvent::AIO for a case where this is useful.
		917
		918	Example: Create a watcher for the IO::AIO module and store it in
		919	$WATCHER. Only do so after the event loop is initialised, though.
		920
		921	our WATCHER;
		922
		923	my $guard = AnyEvent::post_detect {
		924	$WATCHER = AnyEvent->io (fh => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb);
		925	};
		926
		927	# the ||= is important in case post_detect immediately runs the block,
		928	# as to not clobber the newly-created watcher. assigning both watcher and
		929	# post_detect guard to the same variable has the advantage of users being
		930	# able to just C<undef $WATCHER> if the watcher causes them grief.
		931
		932	$WATCHER ||= $guard;
		933
		934	@AnyEvent::post_detect
		935	If there are any code references in this array (you can "push" to it
		936	before or after loading AnyEvent), then they will called directly
		937	after the event loop has been chosen.
		938
		939	You should check $AnyEvent::MODEL before adding to this array,
		940	though: if it is defined then the event loop has already been
		941	detected, and the array will be ignored.
		942
		943	Best use "AnyEvent::post_detect { BLOCK }" when your application
		944	allows it, as it takes care of these details.
		945
		946	This variable is mainly useful for modules that can do something
		947	useful when AnyEvent is used and thus want to know when it is
		948	initialised, but do not need to even load it by default. This array
		949	provides the means to hook into AnyEvent passively, without loading
		950	it.
		951
		952	Example: To load Coro::AnyEvent whenever Coro and AnyEvent are used
		953	together, you could put this into Coro (this is the actual code used
		954	by Coro to accomplish this):
		955
		956	if (defined $AnyEvent::MODEL) {
		957	# AnyEvent already initialised, so load Coro::AnyEvent
		958	require Coro::AnyEvent;
		959	} else {
		960	# AnyEvent not yet initialised, so make sure to load Coro::AnyEvent
		961	# as soon as it is
		962	push @AnyEvent::post_detect, sub { require Coro::AnyEvent };
		963	}
348		964
349	WHAT TO DO IN A MODULE	965	WHAT TO DO IN A MODULE
350	As a module author, you should "use AnyEvent" and call AnyEvent methods	966	As a module author, you should "use AnyEvent" and call AnyEvent methods
351	freely, but you should not load a specific event module or rely on it.	967	freely, but you should not load a specific event module or rely on it.
352		968
353	Be careful when you create watchers in the module body - AnyEvent will	969	Be careful when you create watchers in the module body - AnyEvent will
354	decide which event module to use as soon as the first method is called,	970	decide which event module to use as soon as the first method is called,
355	so by calling AnyEvent in your module body you force the user of your	971	so by calling AnyEvent in your module body you force the user of your
356	module to load the event module first.	972	module to load the event module first.
357		973
358	Never call "->wait" on a condition variable unless you *know* that the	974	Never call "->recv" on a condition variable unless you *know* that the
359	"->broadcast" method has been called on it already. This is because it	975	"->send" method has been called on it already. This is because it will
360	will stall the whole program, and the whole point of using events is to	976	stall the whole program, and the whole point of using events is to stay
361	stay interactive.	977	interactive.
362		978
363	It is fine, however, to call "->wait" when the user of your module	979	It is fine, however, to call "->recv" when the user of your module
364	requests it (i.e. if you create a http request object ad have a method	980	requests it (i.e. if you create a http request object ad have a method
365	called "results" that returns the results, it should call "->wait"	981	called "results" that returns the results, it should call "->recv"
366	freely, as the user of your module knows what she is doing. always).	982	freely, as the user of your module knows what she is doing. always).
367		983
368	WHAT TO DO IN THE MAIN PROGRAM	984	WHAT TO DO IN THE MAIN PROGRAM
369	There will always be a single main program - the only place that should	985	There will always be a single main program - the only place that should
370	dictate which event model to use.	986	dictate which event model to use.
…		…
372	If it doesn't care, it can just "use AnyEvent" and use it itself, or not	988	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
373	do anything special (it does not need to be event-based) and let	989	do anything special (it does not need to be event-based) and let
374	AnyEvent decide which implementation to chose if some module relies on	990	AnyEvent decide which implementation to chose if some module relies on
375	it.	991	it.
376		992
377	If the main program relies on a specific event model. For example, in	993	If the main program relies on a specific event model - for example, in
378	Gtk2 programs you have to rely on the Glib module. You should load the	994	Gtk2 programs you have to rely on the Glib module - you should load the
379	event module before loading AnyEvent or any module that uses it:	995	event module before loading AnyEvent or any module that uses it:
380	generally speaking, you should load it as early as possible. The reason	996	generally speaking, you should load it as early as possible. The reason
381	is that modules might create watchers when they are loaded, and AnyEvent	997	is that modules might create watchers when they are loaded, and AnyEvent
382	will decide on the event model to use as soon as it creates watchers,	998	will decide on the event model to use as soon as it creates watchers,
383	and it might chose the wrong one unless you load the correct one	999	and it might chose the wrong one unless you load the correct one
384	yourself.	1000	yourself.
385		1001
386	You can chose to use a rather inefficient pure-perl implementation by	1002	You can chose to use a pure-perl implementation by loading the
387	loading the "AnyEvent::Impl::Perl" module, which gives you similar	1003	"AnyEvent::Impl::Perl" module, which gives you similar behaviour
388	behaviour everywhere, but letting AnyEvent chose is generally better.	1004	everywhere, but letting AnyEvent chose the model is generally better.
		1005
		1006	MAINLOOP EMULATION
		1007	Sometimes (often for short test scripts, or even standalone programs who
		1008	only want to use AnyEvent), you do not want to run a specific event
		1009	loop.
		1010
		1011	In that case, you can use a condition variable like this:
		1012
		1013	AnyEvent->condvar->recv;
		1014
		1015	This has the effect of entering the event loop and looping forever.
		1016
		1017	Note that usually your program has some exit condition, in which case it
		1018	is better to use the "traditional" approach of storing a condition
		1019	variable somewhere, waiting for it, and sending it when the program
		1020	should exit cleanly.
		1021
		1022	OTHER MODULES
		1023	The following is a non-exhaustive list of additional modules that use
		1024	AnyEvent as a client and can therefore be mixed easily with other
		1025	AnyEvent modules and other event loops in the same program. Some of the
		1026	modules come as part of AnyEvent, the others are available via CPAN.
		1027
		1028	AnyEvent::Util
		1029	Contains various utility functions that replace often-used but
		1030	blocking functions such as "inet_aton" by event-/callback-based
		1031	versions.
		1032
		1033	AnyEvent::Socket
		1034	Provides various utility functions for (internet protocol) sockets,
		1035	addresses and name resolution. Also functions to create non-blocking
		1036	tcp connections or tcp servers, with IPv6 and SRV record support and
		1037	more.
		1038
		1039	AnyEvent::Handle
		1040	Provide read and write buffers, manages watchers for reads and
		1041	writes, supports raw and formatted I/O, I/O queued and fully
		1042	transparent and non-blocking SSL/TLS (via AnyEvent::TLS.
		1043
		1044	AnyEvent::DNS
		1045	Provides rich asynchronous DNS resolver capabilities.
		1046
		1047	AnyEvent::HTTP, AnyEvent::IRC, AnyEvent::XMPP, AnyEvent::GPSD,
		1048	AnyEvent::IGS, AnyEvent::FCP
		1049	Implement event-based interfaces to the protocols of the same name
		1050	(for the curious, IGS is the International Go Server and FCP is the
		1051	Freenet Client Protocol).
		1052
		1053	AnyEvent::Handle::UDP
		1054	Here be danger!
		1055
		1056	As Pauli would put it, "Not only is it not right, it's not even
		1057	wrong!" - there are so many things wrong with AnyEvent::Handle::UDP,
		1058	most notably it's use of a stream-based API with a protocol that
		1059	isn't streamable, that the only way to improve it is to delete it.
		1060
		1061	It features data corruption (but typically only under load) and
		1062	general confusion. On top, the author is not only clueless about UDP
		1063	but also fact-resistant - some gems of his understanding: "connect
		1064	doesn't work with UDP", "UDP packets are not IP packets", "UDP only
		1065	has datagrams, not packets", "I don't need to implement proper error
		1066	checking as UDP doesn't support error checking" and so on - he
		1067	doesn't even understand what's wrong with his module when it is
		1068	explained to him.
		1069
		1070	AnyEvent::DBI
		1071	Executes DBI requests asynchronously in a proxy process for you,
		1072	notifying you in an event-bnased way when the operation is finished.
		1073
		1074	AnyEvent::AIO
		1075	Truly asynchronous (as opposed to non-blocking) I/O, should be in
		1076	the toolbox of every event programmer. AnyEvent::AIO transparently
		1077	fuses IO::AIO and AnyEvent together, giving AnyEvent access to
		1078	event-based file I/O, and much more.
		1079
		1080	AnyEvent::HTTPD
		1081	A simple embedded webserver.
		1082
		1083	AnyEvent::FastPing
		1084	The fastest ping in the west.
		1085
		1086	Coro
		1087	Has special support for AnyEvent via Coro::AnyEvent.
		1088
		1089	SIMPLIFIED AE API
		1090	Starting with version 5.0, AnyEvent officially supports a second, much
		1091	simpler, API that is designed to reduce the calling, typing and memory
		1092	overhead by using function call syntax and a fixed number of parameters.
		1093
		1094	See the AE manpage for details.
		1095
		1096	ERROR AND EXCEPTION HANDLING
		1097	In general, AnyEvent does not do any error handling - it relies on the
		1098	caller to do that if required. The AnyEvent::Strict module (see also the
		1099	"PERL_ANYEVENT_STRICT" environment variable, below) provides strict
		1100	checking of all AnyEvent methods, however, which is highly useful during
		1101	development.
		1102
		1103	As for exception handling (i.e. runtime errors and exceptions thrown
		1104	while executing a callback), this is not only highly event-loop
		1105	specific, but also not in any way wrapped by this module, as this is the
		1106	job of the main program.
		1107
		1108	The pure perl event loop simply re-throws the exception (usually within
		1109	"condvar->recv"), the Event and EV modules call "$Event/EV::DIED->()",
		1110	Glib uses "install_exception_handler" and so on.
		1111
		1112	ENVIRONMENT VARIABLES
		1113	The following environment variables are used by this module or its
		1114	submodules.
		1115
		1116	Note that AnyEvent will remove *all* environment variables starting with
		1117	"PERL_ANYEVENT_" from %ENV when it is loaded while taint mode is
		1118	enabled.
		1119
		1120	"PERL_ANYEVENT_VERBOSE"
		1121	By default, AnyEvent will be completely silent except in fatal
		1122	conditions. You can set this environment variable to make AnyEvent
		1123	more talkative.
		1124
		1125	When set to 1 or higher, causes AnyEvent to warn about unexpected
		1126	conditions, such as not being able to load the event model specified
		1127	by "PERL_ANYEVENT_MODEL".
		1128
		1129	When set to 2 or higher, cause AnyEvent to report to STDERR which
		1130	event model it chooses.
		1131
		1132	When set to 8 or higher, then AnyEvent will report extra information
		1133	on which optional modules it loads and how it implements certain
		1134	features.
		1135
		1136	"PERL_ANYEVENT_STRICT"
		1137	AnyEvent does not do much argument checking by default, as thorough
		1138	argument checking is very costly. Setting this variable to a true
		1139	value will cause AnyEvent to load "AnyEvent::Strict" and then to
		1140	thoroughly check the arguments passed to most method calls. If it
		1141	finds any problems, it will croak.
		1142
		1143	In other words, enables "strict" mode.
		1144
		1145	Unlike "use strict" (or it's modern cousin, "use common::sense", it
		1146	is definitely recommended to keep it off in production. Keeping
		1147	"PERL_ANYEVENT_STRICT=1" in your environment while developing
		1148	programs can be very useful, however.
		1149
		1150	"PERL_ANYEVENT_MODEL"
		1151	This can be used to specify the event model to be used by AnyEvent,
		1152	before auto detection and -probing kicks in. It must be a string
		1153	consisting entirely of ASCII letters. The string "AnyEvent::Impl::"
		1154	gets prepended and the resulting module name is loaded and if the
		1155	load was successful, used as event model. If it fails to load
		1156	AnyEvent will proceed with auto detection and -probing.
		1157
		1158	This functionality might change in future versions.
		1159
		1160	For example, to force the pure perl model (AnyEvent::Impl::Perl) you
		1161	could start your program like this:
		1162
		1163	PERL_ANYEVENT_MODEL=Perl perl ...
		1164
		1165	"PERL_ANYEVENT_PROTOCOLS"
		1166	Used by both AnyEvent::DNS and AnyEvent::Socket to determine
		1167	preferences for IPv4 or IPv6. The default is unspecified (and might
		1168	change, or be the result of auto probing).
		1169
		1170	Must be set to a comma-separated list of protocols or address
		1171	families, current supported: "ipv4" and "ipv6". Only protocols
		1172	mentioned will be used, and preference will be given to protocols
		1173	mentioned earlier in the list.
		1174
		1175	This variable can effectively be used for denial-of-service attacks
		1176	against local programs (e.g. when setuid), although the impact is
		1177	likely small, as the program has to handle conenction and other
		1178	failures anyways.
		1179
		1180	Examples: "PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6" - prefer IPv4 over
		1181	IPv6, but support both and try to use both.
		1182	"PERL_ANYEVENT_PROTOCOLS=ipv4" - only support IPv4, never try to
		1183	resolve or contact IPv6 addresses.
		1184	"PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4" support either IPv4 or IPv6, but
		1185	prefer IPv6 over IPv4.
		1186
		1187	"PERL_ANYEVENT_EDNS0"
		1188	Used by AnyEvent::DNS to decide whether to use the EDNS0 extension
		1189	for DNS. This extension is generally useful to reduce DNS traffic,
		1190	but some (broken) firewalls drop such DNS packets, which is why it
		1191	is off by default.
		1192
		1193	Setting this variable to 1 will cause AnyEvent::DNS to announce
		1194	EDNS0 in its DNS requests.
		1195
		1196	"PERL_ANYEVENT_MAX_FORKS"
		1197	The maximum number of child processes that
		1198	"AnyEvent::Util::fork_call" will create in parallel.
		1199
		1200	"PERL_ANYEVENT_MAX_OUTSTANDING_DNS"
		1201	The default value for the "max_outstanding" parameter for the
		1202	default DNS resolver - this is the maximum number of parallel DNS
		1203	requests that are sent to the DNS server.
		1204
		1205	"PERL_ANYEVENT_RESOLV_CONF"
		1206	The file to use instead of /etc/resolv.conf (or OS-specific
		1207	configuration) in the default resolver. When set to the empty
		1208	string, no default config will be used.
		1209
		1210	"PERL_ANYEVENT_CA_FILE", "PERL_ANYEVENT_CA_PATH".
		1211	When neither "ca_file" nor "ca_path" was specified during
		1212	AnyEvent::TLS context creation, and either of these environment
		1213	variables exist, they will be used to specify CA certificate
		1214	locations instead of a system-dependent default.
		1215
		1216	"PERL_ANYEVENT_AVOID_GUARD" and "PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT"
		1217	When these are set to 1, then the respective modules are not loaded.
		1218	Mostly good for testing AnyEvent itself.
389		1219
390	SUPPLYING YOUR OWN EVENT MODEL INTERFACE	1220	SUPPLYING YOUR OWN EVENT MODEL INTERFACE
391	This is an advanced topic that you do not normally need to use AnyEvent	1221	This is an advanced topic that you do not normally need to use AnyEvent
392	in a module. This section is only of use to event loop authors who want	1222	in a module. This section is only of use to event loop authors who want
393	to provide AnyEvent compatibility.	1223	to provide AnyEvent compatibility.
…		…
428	*rxvt-unicode* also cheats a bit by not providing blocking access to	1258	*rxvt-unicode* also cheats a bit by not providing blocking access to
429	condition variables: code blocking while waiting for a condition will	1259	condition variables: code blocking while waiting for a condition will
430	"die". This still works with most modules/usages, and blocking calls	1260	"die". This still works with most modules/usages, and blocking calls
431	must not be done in an interactive application, so it makes sense.	1261	must not be done in an interactive application, so it makes sense.
432		1262
433	ENVIRONMENT VARIABLES
434	The following environment variables are used by this module:
435
436	"PERL_ANYEVENT_VERBOSE" when set to 2 or higher, cause AnyEvent to
437	report to STDERR which event model it chooses.
438
439	EXAMPLE PROGRAM	1263	EXAMPLE PROGRAM
440	The following program uses an IO watcher to read data from STDIN, a	1264	The following program uses an I/O watcher to read data from STDIN, a
441	timer to display a message once per second, and a condition variable to	1265	timer to display a message once per second, and a condition variable to
442	quit the program when the user enters quit:	1266	quit the program when the user enters quit:
443		1267
444	use AnyEvent;	1268	use AnyEvent;
445		1269
…		…
450	poll => 'r',	1274	poll => 'r',
451	cb => sub {	1275	cb => sub {
452	warn "io event <$_[0]>\n"; # will always output <r>	1276	warn "io event <$_[0]>\n"; # will always output <r>
453	chomp (my $input = <STDIN>); # read a line	1277	chomp (my $input = <STDIN>); # read a line
454	warn "read: $input\n"; # output what has been read	1278	warn "read: $input\n"; # output what has been read
455	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1279	$cv->send if $input =~ /^q/i; # quit program if /^q/i
456	},	1280	},
457	);	1281	);
458		1282
459	my $time_watcher; # can only be used once
460
461	sub new_timer {
462	$timer = AnyEvent->timer (after => 1, cb => sub {	1283	my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub {
463	warn "timeout\n"; # print 'timeout' about every second	1284	warn "timeout\n"; # print 'timeout' at most every second
464	&new_timer; # and restart the time
465	});
466	}	1285	});
467		1286
468	new_timer; # create first timer
469
470	$cv->wait; # wait until user enters /^q/i	1287	$cv->recv; # wait until user enters /^q/i
471		1288
472	REAL-WORLD EXAMPLE	1289	REAL-WORLD EXAMPLE
473	Consider the Net::FCP module. It features (among others) the following	1290	Consider the Net::FCP module. It features (among others) the following
474	API calls, which are to freenet what HTTP GET requests are to http:	1291	API calls, which are to freenet what HTTP GET requests are to http:
475		1292
…		…
524	syswrite $txn->{fh}, $txn->{request}	1341	syswrite $txn->{fh}, $txn->{request}
525	or die "connection or write error";	1342	or die "connection or write error";
526	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	1343	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
527		1344
528	Again, "fh_ready_r" waits till all data has arrived, and then stores the	1345	Again, "fh_ready_r" waits till all data has arrived, and then stores the
529	result and signals any possible waiters that the request ahs finished:	1346	result and signals any possible waiters that the request has finished:
530		1347
531	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	1348	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
532		1349
533	if (end-of-file or data complete) {	1350	if (end-of-file or data complete) {
534	$txn->{result} = $txn->{buf};	1351	$txn->{result} = $txn->{buf};
535	$txn->{finished}->broadcast;	1352	$txn->{finished}->send;
536	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	1353	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
537	}	1354	}
538		1355
539	The "result" method, finally, just waits for the finished signal (if the	1356	The "result" method, finally, just waits for the finished signal (if the
540	request was already finished, it doesn't wait, of course, and returns	1357	request was already finished, it doesn't wait, of course, and returns
541	the data:	1358	the data:
542		1359
543	$txn->{finished}->wait;	1360	$txn->{finished}->recv;
544	return $txn->{result};	1361	return $txn->{result};
545		1362
546	The actual code goes further and collects all errors ("die"s,	1363	The actual code goes further and collects all errors ("die"s,
547	exceptions) that occured during request processing. The "result" method	1364	exceptions) that occurred during request processing. The "result" method
548	detects whether an exception as thrown (it is stored inside the $txn	1365	detects whether an exception as thrown (it is stored inside the $txn
549	object) and just throws the exception, which means connection errors and	1366	object) and just throws the exception, which means connection errors and
550	other problems get reported tot he code that tries to use the result,	1367	other problems get reported to the code that tries to use the result,
551	not in a random callback.	1368	not in a random callback.
552		1369
553	All of this enables the following usage styles:	1370	All of this enables the following usage styles:
554		1371
555	1. Blocking:	1372	1. Blocking:
…		…
583		1400
584	my $quit = AnyEvent->condvar;	1401	my $quit = AnyEvent->condvar;
585		1402
586	$fcp->txn_client_get ($url)->cb (sub {	1403	$fcp->txn_client_get ($url)->cb (sub {
587	...	1404	...
588	$quit->broadcast;	1405	$quit->send;
589	});	1406	});
590		1407
591	$quit->wait;	1408	$quit->recv;
		1409
		1410	BENCHMARKS
		1411	To give you an idea of the performance and overheads that AnyEvent adds
		1412	over the event loops themselves and to give you an impression of the
		1413	speed of various event loops I prepared some benchmarks.
		1414
		1415	BENCHMARKING ANYEVENT OVERHEAD
		1416	Here is a benchmark of various supported event models used natively and
		1417	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1418	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1419	which it is), lets them fire exactly once and destroys them again.
		1420
		1421	Source code for this benchmark is found as eg/bench in the AnyEvent
		1422	distribution. It uses the AE interface, which makes a real difference
		1423	for the EV and Perl backends only.
		1424
		1425	Explanation of the columns
		1426	*watcher* is the number of event watchers created/destroyed. Since
		1427	different event models feature vastly different performances, each event
		1428	loop was given a number of watchers so that overall runtime is
		1429	acceptable and similar between tested event loop (and keep them from
		1430	crashing): Glib would probably take thousands of years if asked to
		1431	process the same number of watchers as EV in this benchmark.
		1432
		1433	*bytes* is the number of bytes (as measured by the resident set size,
		1434	RSS) consumed by each watcher. This method of measuring captures both C
		1435	and Perl-based overheads.
		1436
		1437	*create* is the time, in microseconds (millionths of seconds), that it
		1438	takes to create a single watcher. The callback is a closure shared
		1439	between all watchers, to avoid adding memory overhead. That means
		1440	closure creation and memory usage is not included in the figures.
		1441
		1442	*invoke* is the time, in microseconds, used to invoke a simple callback.
		1443	The callback simply counts down a Perl variable and after it was invoked
		1444	"watcher" times, it would "->send" a condvar once to signal the end of
		1445	this phase.
		1446
		1447	*destroy* is the time, in microseconds, that it takes to destroy a
		1448	single watcher.
		1449
		1450	Results
		1451	name watchers bytes create invoke destroy comment
		1452	EV/EV 100000 223 0.47 0.43 0.27 EV native interface
		1453	EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers
		1454	Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal
		1455	Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation
		1456	Event/Event 16000 516 31.16 31.84 0.82 Event native interface
		1457	Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers
		1458	IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll
		1459	IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll
		1460	Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour
		1461	Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers
		1462	POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event
		1463	POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select
		1464
		1465	Discussion
		1466	The benchmark does *not* measure scalability of the event loop very
		1467	well. For example, a select-based event loop (such as the pure perl one)
		1468	can never compete with an event loop that uses epoll when the number of
		1469	file descriptors grows high. In this benchmark, all events become ready
		1470	at the same time, so select/poll-based implementations get an unnatural
		1471	speed boost.
		1472
		1473	Also, note that the number of watchers usually has a nonlinear effect on
		1474	overall speed, that is, creating twice as many watchers doesn't take
		1475	twice the time - usually it takes longer. This puts event loops tested
		1476	with a higher number of watchers at a disadvantage.
		1477
		1478	To put the range of results into perspective, consider that on the
		1479	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1480	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000
		1481	CPU cycles with POE.
		1482
		1483	"EV" is the sole leader regarding speed and memory use, which are both
		1484	maximal/minimal, respectively. When using the AE API there is zero
		1485	overhead (when going through the AnyEvent API create is about 5-6 times
		1486	slower, with other times being equal, so still uses far less memory than
		1487	any other event loop and is still faster than Event natively).
		1488
		1489	The pure perl implementation is hit in a few sweet spots (both the
		1490	constant timeout and the use of a single fd hit optimisations in the
		1491	perl interpreter and the backend itself). Nevertheless this shows that
		1492	it adds very little overhead in itself. Like any select-based backend
		1493	its performance becomes really bad with lots of file descriptors (and
		1494	few of them active), of course, but this was not subject of this
		1495	benchmark.
		1496
		1497	The "Event" module has a relatively high setup and callback invocation
		1498	cost, but overall scores in on the third place.
		1499
		1500	"IO::Async" performs admirably well, about on par with "Event", even
		1501	when using its pure perl backend.
		1502
		1503	"Glib"'s memory usage is quite a bit higher, but it features a faster
		1504	callback invocation and overall ends up in the same class as "Event".
		1505	However, Glib scales extremely badly, doubling the number of watchers
		1506	increases the processing time by more than a factor of four, making it
		1507	completely unusable when using larger numbers of watchers (note that
		1508	only a single file descriptor was used in the benchmark, so
		1509	inefficiencies of "poll" do not account for this).
		1510
		1511	The "Tk" adaptor works relatively well. The fact that it crashes with
		1512	more than 2000 watchers is a big setback, however, as correctness takes
		1513	precedence over speed. Nevertheless, its performance is surprising, as
		1514	the file descriptor is dup()ed for each watcher. This shows that the
		1515	dup() employed by some adaptors is not a big performance issue (it does
		1516	incur a hidden memory cost inside the kernel which is not reflected in
		1517	the figures above).
		1518
		1519	"POE", regardless of underlying event loop (whether using its pure perl
		1520	select-based backend or the Event module, the POE-EV backend couldn't be
		1521	tested because it wasn't working) shows abysmal performance and memory
		1522	usage with AnyEvent: Watchers use almost 30 times as much memory as EV
		1523	watchers, and 10 times as much memory as Event (the high memory
		1524	requirements are caused by requiring a session for each watcher).
		1525	Watcher invocation speed is almost 900 times slower than with AnyEvent's
		1526	pure perl implementation.
		1527
		1528	The design of the POE adaptor class in AnyEvent can not really account
		1529	for the performance issues, though, as session creation overhead is
		1530	small compared to execution of the state machine, which is coded pretty
		1531	optimally within AnyEvent::Impl::POE (and while everybody agrees that
		1532	using multiple sessions is not a good approach, especially regarding
		1533	memory usage, even the author of POE could not come up with a faster
		1534	design).
		1535
		1536	Summary
		1537	* Using EV through AnyEvent is faster than any other event loop (even
		1538	when used without AnyEvent), but most event loops have acceptable
		1539	performance with or without AnyEvent.
		1540
		1541	* The overhead AnyEvent adds is usually much smaller than the overhead
		1542	of the actual event loop, only with extremely fast event loops such
		1543	as EV adds AnyEvent significant overhead.
		1544
		1545	* You should avoid POE like the plague if you want performance or
		1546	reasonable memory usage.
		1547
		1548	BENCHMARKING THE LARGE SERVER CASE
		1549	This benchmark actually benchmarks the event loop itself. It works by
		1550	creating a number of "servers": each server consists of a socket pair, a
		1551	timeout watcher that gets reset on activity (but never fires), and an
		1552	I/O watcher waiting for input on one side of the socket. Each time the
		1553	socket watcher reads a byte it will write that byte to a random other
		1554	"server".
		1555
		1556	The effect is that there will be a lot of I/O watchers, only part of
		1557	which are active at any one point (so there is a constant number of
		1558	active fds for each loop iteration, but which fds these are is random).
		1559	The timeout is reset each time something is read because that reflects
		1560	how most timeouts work (and puts extra pressure on the event loops).
		1561
		1562	In this benchmark, we use 10000 socket pairs (20000 sockets), of which
		1563	100 (1%) are active. This mirrors the activity of large servers with
		1564	many connections, most of which are idle at any one point in time.
		1565
		1566	Source code for this benchmark is found as eg/bench2 in the AnyEvent
		1567	distribution. It uses the AE interface, which makes a real difference
		1568	for the EV and Perl backends only.
		1569
		1570	Explanation of the columns
		1571	*sockets* is the number of sockets, and twice the number of "servers"
		1572	(as each server has a read and write socket end).
		1573
		1574	*create* is the time it takes to create a socket pair (which is
		1575	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1576
		1577	*request*, the most important value, is the time it takes to handle a
		1578	single "request", that is, reading the token from the pipe and
		1579	forwarding it to another server. This includes deleting the old timeout
		1580	and creating a new one that moves the timeout into the future.
		1581
		1582	Results
		1583	name sockets create request
		1584	EV 20000 62.66 7.99
		1585	Perl 20000 68.32 32.64
		1586	IOAsync 20000 174.06 101.15 epoll
		1587	IOAsync 20000 174.67 610.84 poll
		1588	Event 20000 202.69 242.91
		1589	Glib 20000 557.01 1689.52
		1590	POE 20000 341.54 12086.32 uses POE::Loop::Event
		1591
		1592	Discussion
		1593	This benchmark *does* measure scalability and overall performance of the
		1594	particular event loop.
		1595
		1596	EV is again fastest. Since it is using epoll on my system, the setup
		1597	time is relatively high, though.
		1598
		1599	Perl surprisingly comes second. It is much faster than the C-based event
		1600	loops Event and Glib.
		1601
		1602	IO::Async performs very well when using its epoll backend, and still
		1603	quite good compared to Glib when using its pure perl backend.
		1604
		1605	Event suffers from high setup time as well (look at its code and you
		1606	will understand why). Callback invocation also has a high overhead
		1607	compared to the "$_->() for .."-style loop that the Perl event loop
		1608	uses. Event uses select or poll in basically all documented
		1609	configurations.
		1610
		1611	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1612	clearly fails to perform with many filehandles or in busy servers.
		1613
		1614	POE is still completely out of the picture, taking over 1000 times as
		1615	long as EV, and over 100 times as long as the Perl implementation, even
		1616	though it uses a C-based event loop in this case.
		1617
		1618	Summary
		1619	* The pure perl implementation performs extremely well.
		1620
		1621	* Avoid Glib or POE in large projects where performance matters.
		1622
		1623	BENCHMARKING SMALL SERVERS
		1624	While event loops should scale (and select-based ones do not...) even to
		1625	large servers, most programs we (or I :) actually write have only a few
		1626	I/O watchers.
		1627
		1628	In this benchmark, I use the same benchmark program as in the large
		1629	server case, but it uses only eight "servers", of which three are active
		1630	at any one time. This should reflect performance for a small server
		1631	relatively well.
		1632
		1633	The columns are identical to the previous table.
		1634
		1635	Results
		1636	name sockets create request
		1637	EV 16 20.00 6.54
		1638	Perl 16 25.75 12.62
		1639	Event 16 81.27 35.86
		1640	Glib 16 32.63 15.48
		1641	POE 16 261.87 276.28 uses POE::Loop::Event
		1642
		1643	Discussion
		1644	The benchmark tries to test the performance of a typical small server.
		1645	While knowing how various event loops perform is interesting, keep in
		1646	mind that their overhead in this case is usually not as important, due
		1647	to the small absolute number of watchers (that is, you need efficiency
		1648	and speed most when you have lots of watchers, not when you only have a
		1649	few of them).
		1650
		1651	EV is again fastest.
		1652
		1653	Perl again comes second. It is noticeably faster than the C-based event
		1654	loops Event and Glib, although the difference is too small to really
		1655	matter.
		1656
		1657	POE also performs much better in this case, but is is still far behind
		1658	the others.
		1659
		1660	Summary
		1661	* C-based event loops perform very well with small number of watchers,
		1662	as the management overhead dominates.
		1663
		1664	THE IO::Lambda BENCHMARK
		1665	Recently I was told about the benchmark in the IO::Lambda manpage, which
		1666	could be misinterpreted to make AnyEvent look bad. In fact, the
		1667	benchmark simply compares IO::Lambda with POE, and IO::Lambda looks
		1668	better (which shouldn't come as a surprise to anybody). As such, the
		1669	benchmark is fine, and mostly shows that the AnyEvent backend from
		1670	IO::Lambda isn't very optimal. But how would AnyEvent compare when used
		1671	without the extra baggage? To explore this, I wrote the equivalent
		1672	benchmark for AnyEvent.
		1673
		1674	The benchmark itself creates an echo-server, and then, for 500 times,
		1675	connects to the echo server, sends a line, waits for the reply, and then
		1676	creates the next connection. This is a rather bad benchmark, as it
		1677	doesn't test the efficiency of the framework or much non-blocking I/O,
		1678	but it is a benchmark nevertheless.
		1679
		1680	name runtime
		1681	Lambda/select 0.330 sec
		1682	+ optimized 0.122 sec
		1683	Lambda/AnyEvent 0.327 sec
		1684	+ optimized 0.138 sec
		1685	Raw sockets/select 0.077 sec
		1686	POE/select, components 0.662 sec
		1687	POE/select, raw sockets 0.226 sec
		1688	POE/select, optimized 0.404 sec
		1689
		1690	AnyEvent/select/nb 0.085 sec
		1691	AnyEvent/EV/nb 0.068 sec
		1692	+state machine 0.134 sec
		1693
		1694	The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
		1695	benchmarks actually make blocking connects and use 100% blocking I/O,
		1696	defeating the purpose of an event-based solution. All of the newly
		1697	written AnyEvent benchmarks use 100% non-blocking connects (using
		1698	AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
		1699	resolver), so AnyEvent is at a disadvantage here, as non-blocking
		1700	connects generally require a lot more bookkeeping and event handling
		1701	than blocking connects (which involve a single syscall only).
		1702
		1703	The last AnyEvent benchmark additionally uses AnyEvent::Handle, which
		1704	offers similar expressive power as POE and IO::Lambda, using
		1705	conventional Perl syntax. This means that both the echo server and the
		1706	client are 100% non-blocking, further placing it at a disadvantage.
		1707
		1708	As you can see, the AnyEvent + EV combination even beats the
		1709	hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
		1710	backend easily beats IO::Lambda and POE.
		1711
		1712	And even the 100% non-blocking version written using the high-level (and
		1713	slow :) AnyEvent::Handle abstraction beats both POE and IO::Lambda
		1714	higher level ("unoptimised") abstractions by a large margin, even though
		1715	it does all of DNS, tcp-connect and socket I/O in a non-blocking way.
		1716
		1717	The two AnyEvent benchmarks programs can be found as eg/ae0.pl and
		1718	eg/ae2.pl in the AnyEvent distribution, the remaining benchmarks are
		1719	part of the IO::Lambda distribution and were used without any changes.
		1720
		1721	SIGNALS
		1722	AnyEvent currently installs handlers for these signals:
		1723
		1724	SIGCHLD
		1725	A handler for "SIGCHLD" is installed by AnyEvent's child watcher
		1726	emulation for event loops that do not support them natively. Also,
		1727	some event loops install a similar handler.
		1728
		1729	Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE,
		1730	then AnyEvent will reset it to default, to avoid losing child exit
		1731	statuses.
		1732
		1733	SIGPIPE
		1734	A no-op handler is installed for "SIGPIPE" when $SIG{PIPE} is
		1735	"undef" when AnyEvent gets loaded.
		1736
		1737	The rationale for this is that AnyEvent users usually do not really
		1738	depend on SIGPIPE delivery (which is purely an optimisation for
		1739	shell use, or badly-written programs), but "SIGPIPE" can cause
		1740	spurious and rare program exits as a lot of people do not expect
		1741	"SIGPIPE" when writing to some random socket.
		1742
		1743	The rationale for installing a no-op handler as opposed to ignoring
		1744	it is that this way, the handler will be restored to defaults on
		1745	exec.
		1746
		1747	Feel free to install your own handler, or reset it to defaults.
		1748
		1749	RECOMMENDED/OPTIONAL MODULES
		1750	One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
		1751	it's built-in modules) are required to use it.
		1752
		1753	That does not mean that AnyEvent won't take advantage of some additional
		1754	modules if they are installed.
		1755
		1756	This section explains which additional modules will be used, and how
		1757	they affect AnyEvent's operation.
		1758
		1759	Async::Interrupt
		1760	This slightly arcane module is used to implement fast signal
		1761	handling: To my knowledge, there is no way to do completely
		1762	race-free and quick signal handling in pure perl. To ensure that
		1763	signals still get delivered, AnyEvent will start an interval timer
		1764	to wake up perl (and catch the signals) with some delay (default is
		1765	10 seconds, look for $AnyEvent::MAX_SIGNAL_LATENCY).
		1766
		1767	If this module is available, then it will be used to implement
		1768	signal catching, which means that signals will not be delayed, and
		1769	the event loop will not be interrupted regularly, which is more
		1770	efficient (and good for battery life on laptops).
		1771
		1772	This affects not just the pure-perl event loop, but also other event
		1773	loops that have no signal handling on their own (e.g. Glib, Tk, Qt).
		1774
		1775	Some event loops (POE, Event, Event::Lib) offer signal watchers
		1776	natively, and either employ their own workarounds (POE) or use
		1777	AnyEvent's workaround (using $AnyEvent::MAX_SIGNAL_LATENCY).
		1778	Installing Async::Interrupt does nothing for those backends.
		1779
		1780	EV This module isn't really "optional", as it is simply one of the
		1781	backend event loops that AnyEvent can use. However, it is simply the
		1782	best event loop available in terms of features, speed and stability:
		1783	It supports the AnyEvent API optimally, implements all the watcher
		1784	types in XS, does automatic timer adjustments even when no monotonic
		1785	clock is available, can take avdantage of advanced kernel interfaces
		1786	such as "epoll" and "kqueue", and is the fastest backend *by far*.
		1787	You can even embed Glib/Gtk2 in it (or vice versa, see EV::Glib and
		1788	Glib::EV).
		1789
		1790	If you only use backends that rely on another event loop (e.g.
		1791	"Tk"), then this module will do nothing for you.
		1792
		1793	Guard
		1794	The guard module, when used, will be used to implement
		1795	"AnyEvent::Util::guard". This speeds up guards considerably (and
		1796	uses a lot less memory), but otherwise doesn't affect guard
		1797	operation much. It is purely used for performance.
		1798
		1799	JSON and JSON::XS
		1800	One of these modules is required when you want to read or write JSON
		1801	data via AnyEvent::Handle. JSON is also written in pure-perl, but
		1802	can take advantage of the ultra-high-speed JSON::XS module when it
		1803	is installed.
		1804
		1805	Net::SSLeay
		1806	Implementing TLS/SSL in Perl is certainly interesting, but not very
		1807	worthwhile: If this module is installed, then AnyEvent::Handle (with
		1808	the help of AnyEvent::TLS), gains the ability to do TLS/SSL.
		1809
		1810	Time::HiRes
		1811	This module is part of perl since release 5.008. It will be used
		1812	when the chosen event library does not come with a timing source on
		1813	it's own. The pure-perl event loop (AnyEvent::Impl::Perl) will
		1814	additionally use it to try to use a monotonic clock for timing
		1815	stability.
		1816
		1817	FORK
		1818	Most event libraries are not fork-safe. The ones who are usually are
		1819	because they rely on inefficient but fork-safe "select" or "poll" calls
		1820	- higher performance APIs such as BSD's kqueue or the dreaded Linux
		1821	epoll are usually badly thought-out hacks that are incompatible with
		1822	fork in one way or another. Only EV is fully fork-aware and ensures that
		1823	you continue event-processing in both parent and child (or both, if you
		1824	know what you are doing).
		1825
		1826	This means that, in general, you cannot fork and do event processing in
		1827	the child if the event library was initialised before the fork (which
		1828	usually happens when the first AnyEvent watcher is created, or the
		1829	library is loaded).
		1830
		1831	If you have to fork, you must either do so *before* creating your first
		1832	watcher OR you must not use AnyEvent at all in the child OR you must do
		1833	something completely out of the scope of AnyEvent.
		1834
		1835	The problem of doing event processing in the parent *and* the child is
		1836	much more complicated: even for backends that *are* fork-aware or
		1837	fork-safe, their behaviour is not usually what you want: fork clones all
		1838	watchers, that means all timers, I/O watchers etc. are active in both
		1839	parent and child, which is almost never what you want. USing "exec" to
		1840	start worker children from some kind of manage rprocess is usually
		1841	preferred, because it is much easier and cleaner, at the expense of
		1842	having to have another binary.
		1843
		1844	SECURITY CONSIDERATIONS
		1845	AnyEvent can be forced to load any event model via
		1846	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used
		1847	to execute arbitrary code or directly gain access, it can easily be used
		1848	to make the program hang or malfunction in subtle ways, as AnyEvent
		1849	watchers will not be active when the program uses a different event
		1850	model than specified in the variable.
		1851
		1852	You can make AnyEvent completely ignore this variable by deleting it
		1853	before the first watcher gets created, e.g. with a "BEGIN" block:
		1854
		1855	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		1856
		1857	use AnyEvent;
		1858
		1859	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		1860	be used to probe what backend is used and gain other information (which
		1861	is probably even less useful to an attacker than PERL_ANYEVENT_MODEL),
		1862	and $ENV{PERL_ANYEVENT_STRICT}.
		1863
		1864	Note that AnyEvent will remove *all* environment variables starting with
		1865	"PERL_ANYEVENT_" from %ENV when it is loaded while taint mode is
		1866	enabled.
		1867
		1868	BUGS
		1869	Perl 5.8 has numerous memleaks that sometimes hit this module and are
		1870	hard to work around. If you suffer from memleaks, first upgrade to Perl
		1871	5.10 and check wether the leaks still show up. (Perl 5.10.0 has other
		1872	annoying memleaks, such as leaking on "map" and "grep" but it is usually
		1873	not as pronounced).
592		1874
593	SEE ALSO	1875	SEE ALSO
594	Event modules: Coro::EV, EV, EV::Glib, Glib::EV, Coro::Event, Event,	1876	Utility functions: AnyEvent::Util.
595	Glib::Event, Glib, Coro, Tk.
596		1877
597	Implementations: AnyEvent::Impl::CoroEV, AnyEvent::Impl::EV,	1878	Event modules: EV, EV::Glib, Glib::EV, Event, Glib::Event, Glib, Tk,
598	AnyEvent::Impl::CoroEvent, AnyEvent::Impl::Event, AnyEvent::Impl::Glib,	1879	Event::Lib, Qt, POE.
		1880
		1881	Implementations: AnyEvent::Impl::EV, AnyEvent::Impl::Event,
		1882	AnyEvent::Impl::Glib, AnyEvent::Impl::Tk, AnyEvent::Impl::Perl,
		1883	AnyEvent::Impl::EventLib, AnyEvent::Impl::Qt, AnyEvent::Impl::POE,
599	AnyEvent::Impl::Tk, AnyEvent::Impl::Perl.	1884	AnyEvent::Impl::IOAsync, Anyevent::Impl::Irssi.
600		1885
		1886	Non-blocking file handles, sockets, TCP clients and servers:
		1887	AnyEvent::Handle, AnyEvent::Socket, AnyEvent::TLS.
		1888
		1889	Asynchronous DNS: AnyEvent::DNS.
		1890
		1891	Coroutine support: Coro, Coro::AnyEvent, Coro::EV, Coro::Event,
		1892
601	Nontrivial usage examples: Net::FCP, Net::XMPP2.	1893	Nontrivial usage examples: AnyEvent::GPSD, AnyEvent::XMPP,
		1894	AnyEvent::HTTP.
602		1895
603	AUTHOR	1896	AUTHOR
604	Marc Lehmann <schmorp@schmorp.de>	1897	Marc Lehmann <schmorp@schmorp.de>
605	http://home.schmorp.de/	1898	http://home.schmorp.de/
606		1899

Diff Legend

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