ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.4 by root, Thu Dec 1 22:04:50 2005 UTC vs.
Revision 1.332 by root, Tue Aug 31 23:32:40 2010 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent - provide framework for multiple event loops	3	AnyEvent - the DBI of event loop programming
4		4
5	Event, Coro, Glib, Tk - various supported event loops	5	EV, Event, Glib, Tk, Perl, Event::Lib, Irssi, rxvt-unicode, IO::Async, Qt
		6	and POE are various supported event loops/environments.
6		7
7	=head1 SYNOPSIS	8	=head1 SYNOPSIS
8		9
9	use AnyEvent;	10	use AnyEvent;
10		11
		12	# if you prefer function calls, look at the AE manpage for
		13	# an alternative API.
		14
		15	# file handle or descriptor readable
11	my $w = AnyEvent->timer (fh => ..., poll => "[rw]+", cb => sub {	16	my $w = AnyEvent->io (fh => $fh, poll => "r", cb => sub { ... });
12	my ($poll_got) = @_;	17
		18	# one-shot or repeating timers
		19	my $w = AnyEvent->timer (after => $seconds, cb => sub { ... });
		20	my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ...);
		21
		22	print AnyEvent->now; # prints current event loop time
		23	print AnyEvent->time; # think Time::HiRes::time or simply CORE::time.
		24
		25	# POSIX signal
		26	my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... });
		27
		28	# child process exit
		29	my $w = AnyEvent->child (pid => $pid, cb => sub {
		30	my ($pid, $status) = @_;
13	...	31	...
14	});	32	});
		33
		34	# called when event loop idle (if applicable)
		35	my $w = AnyEvent->idle (cb => sub { ... });
		36
		37	my $w = AnyEvent->condvar; # stores whether a condition was flagged
		38	$w->send; # wake up current and all future recv's
		39	$w->recv; # enters "main loop" till $condvar gets ->send
		40	# use a condvar in callback mode:
		41	$w->cb (sub { $_[0]->recv });
		42
		43	=head1 INTRODUCTION/TUTORIAL
		44
		45	This manpage is mainly a reference manual. If you are interested
		46	in a tutorial or some gentle introduction, have a look at the
		47	L<AnyEvent::Intro> manpage.
		48
		49	=head1 SUPPORT
		50
		51	There is a mailinglist for discussing all things AnyEvent, and an IRC
		52	channel, too.
		53
		54	See the AnyEvent project page at the B<Schmorpforge Ta-Sa Software
		55	Repository>, at L<http://anyevent.schmorp.de>, for more info.
		56
		57	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
		58
		59	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
		60	nowadays. So what is different about AnyEvent?
		61
		62	Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
		63	policy> and AnyEvent is I<small and efficient>.
		64
		65	First and foremost, I<AnyEvent is not an event model> itself, it only
		66	interfaces to whatever event model the main program happens to use, in a
		67	pragmatic way. For event models and certain classes of immortals alike,
		68	the statement "there can only be one" is a bitter reality: In general,
		69	only one event loop can be active at the same time in a process. AnyEvent
		70	cannot change this, but it can hide the differences between those event
		71	loops.
		72
		73	The goal of AnyEvent is to offer module authors the ability to do event
		74	programming (waiting for I/O or timer events) without subscribing to a
		75	religion, a way of living, and most importantly: without forcing your
		76	module users into the same thing by forcing them to use the same event
		77	model you use.
		78
		79	For modules like POE or IO::Async (which is a total misnomer as it is
		80	actually doing all I/O I<synchronously>...), using them in your module is
		81	like joining a cult: After you join, you are dependent on them and you
		82	cannot use anything else, as they are simply incompatible to everything
		83	that isn't them. What's worse, all the potential users of your
		84	module are I<also> forced to use the same event loop you use.
		85
		86	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
		87	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
		88	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
		89	your module uses one of those, every user of your module has to use it,
		90	too. But if your module uses AnyEvent, it works transparently with all
		91	event models it supports (including stuff like IO::Async, as long as those
		92	use one of the supported event loops. It is easy to add new event loops
		93	to AnyEvent, too, so it is future-proof).
		94
		95	In addition to being free of having to use I<the one and only true event
		96	model>, AnyEvent also is free of bloat and policy: with POE or similar
		97	modules, you get an enormous amount of code and strict rules you have to
		98	follow. AnyEvent, on the other hand, is lean and to the point, by only
		99	offering the functionality that is necessary, in as thin as a wrapper as
		100	technically possible.
		101
		102	Of course, AnyEvent comes with a big (and fully optional!) toolbox
		103	of useful functionality, such as an asynchronous DNS resolver, 100%
		104	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		105	such as Windows) and lots of real-world knowledge and workarounds for
		106	platform bugs and differences.
		107
		108	Now, if you I<do want> lots of policy (this can arguably be somewhat
		109	useful) and you want to force your users to use the one and only event
		110	model, you should I<not> use this module.
		111
		112	=head1 DESCRIPTION
		113
		114	L<AnyEvent> provides a uniform interface to various event loops. This
		115	allows module authors to use event loop functionality without forcing
		116	module users to use a specific event loop implementation (since more
		117	than one event loop cannot coexist peacefully).
		118
		119	The interface itself is vaguely similar, but not identical to the L<Event>
		120	module.
		121
		122	During the first call of any watcher-creation method, the module tries
		123	to detect the currently loaded event loop by probing whether one of the
		124	following modules is already loaded: L<EV>, L<AnyEvent::Impl::Perl>,
		125	L<Event>, L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>. The first one
		126	found is used. If none are detected, the module tries to load the first
		127	four modules in the order given; but note that if L<EV> is not
		128	available, the pure-perl L<AnyEvent::Impl::Perl> should always work, so
		129	the other two are not normally tried.
		130
		131	Because AnyEvent first checks for modules that are already loaded, loading
		132	an event model explicitly before first using AnyEvent will likely make
		133	that model the default. For example:
		134
		135	use Tk;
		136	use AnyEvent;
		137
		138	# .. AnyEvent will likely default to Tk
		139
		140	The I<likely> means that, if any module loads another event model and
		141	starts using it, all bets are off - this case should be very rare though,
		142	as very few modules hardcode event loops without announcing this very
		143	loudly.
		144
		145	The pure-perl implementation of AnyEvent is called
		146	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
		147	explicitly and enjoy the high availability of that event loop :)
		148
		149	=head1 WATCHERS
		150
		151	AnyEvent has the central concept of a I<watcher>, which is an object that
		152	stores relevant data for each kind of event you are waiting for, such as
		153	the callback to call, the file handle to watch, etc.
		154
		155	These watchers are normal Perl objects with normal Perl lifetime. After
		156	creating a watcher it will immediately "watch" for events and invoke the
		157	callback when the event occurs (of course, only when the event model
		158	is in control).
		159
		160	Note that B<callbacks must not permanently change global variables>
		161	potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
		162	callbacks must not C<die> >>. The former is good programming practice in
		163	Perl and the latter stems from the fact that exception handling differs
		164	widely between event loops.
		165
		166	To disable a watcher you have to destroy it (e.g. by setting the
		167	variable you store it in to C<undef> or otherwise deleting all references
		168	to it).
		169
		170	All watchers are created by calling a method on the C<AnyEvent> class.
		171
		172	Many watchers either are used with "recursion" (repeating timers for
		173	example), or need to refer to their watcher object in other ways.
		174
		175	One way to achieve that is this pattern:
		176
		177	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
		178	# you can use $w here, for example to undef it
		179	undef $w;
		180	});
		181
		182	Note that C<my $w; $w => combination. This is necessary because in Perl,
		183	my variables are only visible after the statement in which they are
		184	declared.
		185
		186	=head2 I/O WATCHERS
		187
		188	$w = AnyEvent->io (
		189	fh => <filehandle_or_fileno>,
		190	poll => <"r" or "w">,
		191	cb => <callback>,
		192	);
		193
		194	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
		195	with the following mandatory key-value pairs as arguments:
		196
		197	C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch
		198	for events (AnyEvent might or might not keep a reference to this file
		199	handle). Note that only file handles pointing to things for which
		200	non-blocking operation makes sense are allowed. This includes sockets,
		201	most character devices, pipes, fifos and so on, but not for example files
		202	or block devices.
		203
		204	C<poll> must be a string that is either C<r> or C<w>, which creates a
		205	watcher waiting for "r"eadable or "w"ritable events, respectively.
		206
		207	C<cb> is the callback to invoke each time the file handle becomes ready.
		208
		209	Although the callback might get passed parameters, their value and
		210	presence is undefined and you cannot rely on them. Portable AnyEvent
		211	callbacks cannot use arguments passed to I/O watcher callbacks.
		212
		213	The I/O watcher might use the underlying file descriptor or a copy of it.
		214	You must not close a file handle as long as any watcher is active on the
		215	underlying file descriptor.
		216
		217	Some event loops issue spurious readiness notifications, so you should
		218	always use non-blocking calls when reading/writing from/to your file
		219	handles.
		220
		221	Example: wait for readability of STDIN, then read a line and disable the
		222	watcher.
		223
		224	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		225	chomp (my $input = <STDIN>);
		226	warn "read: $input\n";
		227	undef $w;
		228	});
		229
		230	=head2 TIME WATCHERS
		231
		232	$w = AnyEvent->timer (after => <seconds>, cb => <callback>);
		233
		234	$w = AnyEvent->timer (
		235	after => <fractional_seconds>,
		236	interval => <fractional_seconds>,
		237	cb => <callback>,
		238	);
		239
		240	You can create a time watcher by calling the C<< AnyEvent->timer >>
		241	method with the following mandatory arguments:
		242
		243	C<after> specifies after how many seconds (fractional values are
		244	supported) the callback should be invoked. C<cb> is the callback to invoke
		245	in that case.
		246
		247	Although the callback might get passed parameters, their value and
		248	presence is undefined and you cannot rely on them. Portable AnyEvent
		249	callbacks cannot use arguments passed to time watcher callbacks.
		250
		251	The callback will normally be invoked only once. If you specify another
		252	parameter, C<interval>, as a strictly positive number (> 0), then the
		253	callback will be invoked regularly at that interval (in fractional
		254	seconds) after the first invocation. If C<interval> is specified with a
		255	false value, then it is treated as if it were not specified at all.
		256
		257	The callback will be rescheduled before invoking the callback, but no
		258	attempt is made to avoid timer drift in most backends, so the interval is
		259	only approximate.
		260
		261	Example: fire an event after 7.7 seconds.
		262
15	my $w = AnyEvent->io (after => $seconds, cb => sub {	263	my $w = AnyEvent->timer (after => 7.7, cb => sub {
		264	warn "timeout\n";
		265	});
		266
		267	# to cancel the timer:
		268	undef $w;
		269
		270	Example 2: fire an event after 0.5 seconds, then roughly every second.
		271
		272	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
		273	warn "timeout\n";
		274	};
		275
		276	=head3 TIMING ISSUES
		277
		278	There are two ways to handle timers: based on real time (relative, "fire
		279	in 10 seconds") and based on wallclock time (absolute, "fire at 12
		280	o'clock").
		281
		282	While most event loops expect timers to specified in a relative way, they
		283	use absolute time internally. This makes a difference when your clock
		284	"jumps", for example, when ntp decides to set your clock backwards from
		285	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		286	fire "after a second" might actually take six years to finally fire.
		287
		288	AnyEvent cannot compensate for this. The only event loop that is conscious
		289	of these issues is L<EV>, which offers both relative (ev_timer, based
		290	on true relative time) and absolute (ev_periodic, based on wallclock time)
		291	timers.
		292
		293	AnyEvent always prefers relative timers, if available, matching the
		294	AnyEvent API.
		295
		296	AnyEvent has two additional methods that return the "current time":
		297
		298	=over 4
		299
		300	=item AnyEvent->time
		301
		302	This returns the "current wallclock time" as a fractional number of
		303	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		304	return, and the result is guaranteed to be compatible with those).
		305
		306	It progresses independently of any event loop processing, i.e. each call
		307	will check the system clock, which usually gets updated frequently.
		308
		309	=item AnyEvent->now
		310
		311	This also returns the "current wallclock time", but unlike C<time>, above,
		312	this value might change only once per event loop iteration, depending on
		313	the event loop (most return the same time as C<time>, above). This is the
		314	time that AnyEvent's timers get scheduled against.
		315
		316	I<In almost all cases (in all cases if you don't care), this is the
		317	function to call when you want to know the current time.>
		318
		319	This function is also often faster then C<< AnyEvent->time >>, and
		320	thus the preferred method if you want some timestamp (for example,
		321	L<AnyEvent::Handle> uses this to update its activity timeouts).
		322
		323	The rest of this section is only of relevance if you try to be very exact
		324	with your timing; you can skip it without a bad conscience.
		325
		326	For a practical example of when these times differ, consider L<Event::Lib>
		327	and L<EV> and the following set-up:
		328
		329	The event loop is running and has just invoked one of your callbacks at
		330	time=500 (assume no other callbacks delay processing). In your callback,
		331	you wait a second by executing C<sleep 1> (blocking the process for a
		332	second) and then (at time=501) you create a relative timer that fires
		333	after three seconds.
		334
		335	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		336	both return C<501>, because that is the current time, and the timer will
		337	be scheduled to fire at time=504 (C<501> + C<3>).
		338
		339	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		340	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		341	last event processing phase started. With L<EV>, your timer gets scheduled
		342	to run at time=503 (C<500> + C<3>).
		343
		344	In one sense, L<Event::Lib> is more exact, as it uses the current time
		345	regardless of any delays introduced by event processing. However, most
		346	callbacks do not expect large delays in processing, so this causes a
		347	higher drift (and a lot more system calls to get the current time).
		348
		349	In another sense, L<EV> is more exact, as your timer will be scheduled at
		350	the same time, regardless of how long event processing actually took.
		351
		352	In either case, if you care (and in most cases, you don't), then you
		353	can get whatever behaviour you want with any event loop, by taking the
		354	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		355	account.
		356
		357	=item AnyEvent->now_update
		358
		359	Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
		360	the current time for each loop iteration (see the discussion of L<<
		361	AnyEvent->now >>, above).
		362
		363	When a callback runs for a long time (or when the process sleeps), then
		364	this "current" time will differ substantially from the real time, which
		365	might affect timers and time-outs.
		366
		367	When this is the case, you can call this method, which will update the
		368	event loop's idea of "current time".
		369
		370	A typical example would be a script in a web server (e.g. C<mod_perl>) -
		371	when mod_perl executes the script, then the event loop will have the wrong
		372	idea about the "current time" (being potentially far in the past, when the
		373	script ran the last time). In that case you should arrange a call to C<<
		374	AnyEvent->now_update >> each time the web server process wakes up again
		375	(e.g. at the start of your script, or in a handler).
		376
		377	Note that updating the time I<might> cause some events to be handled.
		378
		379	=back
		380
		381	=head2 SIGNAL WATCHERS
		382
		383	$w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>);
		384
		385	You can watch for signals using a signal watcher, C<signal> is the signal
		386	I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
		387	callback to be invoked whenever a signal occurs.
		388
		389	Although the callback might get passed parameters, their value and
		390	presence is undefined and you cannot rely on them. Portable AnyEvent
		391	callbacks cannot use arguments passed to signal watcher callbacks.
		392
		393	Multiple signal occurrences can be clumped together into one callback
		394	invocation, and callback invocation will be synchronous. Synchronous means
		395	that it might take a while until the signal gets handled by the process,
		396	but it is guaranteed not to interrupt any other callbacks.
		397
		398	The main advantage of using these watchers is that you can share a signal
		399	between multiple watchers, and AnyEvent will ensure that signals will not
		400	interrupt your program at bad times.
		401
		402	This watcher might use C<%SIG> (depending on the event loop used),
		403	so programs overwriting those signals directly will likely not work
		404	correctly.
		405
		406	Example: exit on SIGINT
		407
		408	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		409
		410	=head3 Restart Behaviour
		411
		412	While restart behaviour is up to the event loop implementation, most will
		413	not restart syscalls (that includes L<Async::Interrupt> and AnyEvent's
		414	pure perl implementation).
		415
		416	=head3 Safe/Unsafe Signals
		417
		418	Perl signals can be either "safe" (synchronous to opcode handling) or
		419	"unsafe" (asynchronous) - the former might get delayed indefinitely, the
		420	latter might corrupt your memory.
		421
		422	AnyEvent signal handlers are, in addition, synchronous to the event loop,
		423	i.e. they will not interrupt your running perl program but will only be
		424	called as part of the normal event handling (just like timer, I/O etc.
		425	callbacks, too).
		426
		427	=head3 Signal Races, Delays and Workarounds
		428
		429	Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
		430	callbacks to signals in a generic way, which is a pity, as you cannot
		431	do race-free signal handling in perl, requiring C libraries for
		432	this. AnyEvent will try to do its best, which means in some cases,
		433	signals will be delayed. The maximum time a signal might be delayed is
		434	specified in C<$AnyEvent::MAX_SIGNAL_LATENCY> (default: 10 seconds). This
		435	variable can be changed only before the first signal watcher is created,
		436	and should be left alone otherwise. This variable determines how often
		437	AnyEvent polls for signals (in case a wake-up was missed). Higher values
		438	will cause fewer spurious wake-ups, which is better for power and CPU
		439	saving.
		440
		441	All these problems can be avoided by installing the optional
		442	L<Async::Interrupt> module, which works with most event loops. It will not
		443	work with inherently broken event loops such as L<Event> or L<Event::Lib>
		444	(and not with L<POE> currently, as POE does its own workaround with
		445	one-second latency). For those, you just have to suffer the delays.
		446
		447	=head2 CHILD PROCESS WATCHERS
		448
		449	$w = AnyEvent->child (pid => <process id>, cb => <callback>);
		450
		451	You can also watch for a child process exit and catch its exit status.
		452
		453	The child process is specified by the C<pid> argument (on some backends,
		454	using C<0> watches for any child process exit, on others this will
		455	croak). The watcher will be triggered only when the child process has
		456	finished and an exit status is available, not on any trace events
		457	(stopped/continued).
		458
		459	The callback will be called with the pid and exit status (as returned by
		460	waitpid), so unlike other watcher types, you I<can> rely on child watcher
		461	callback arguments.
		462
		463	This watcher type works by installing a signal handler for C<SIGCHLD>,
		464	and since it cannot be shared, nothing else should use SIGCHLD or reap
		465	random child processes (waiting for specific child processes, e.g. inside
		466	C<system>, is just fine).
		467
		468	There is a slight catch to child watchers, however: you usually start them
		469	I<after> the child process was created, and this means the process could
		470	have exited already (and no SIGCHLD will be sent anymore).
		471
		472	Not all event models handle this correctly (neither POE nor IO::Async do,
		473	see their AnyEvent::Impl manpages for details), but even for event models
		474	that I<do> handle this correctly, they usually need to be loaded before
		475	the process exits (i.e. before you fork in the first place). AnyEvent's
		476	pure perl event loop handles all cases correctly regardless of when you
		477	start the watcher.
		478
		479	This means you cannot create a child watcher as the very first
		480	thing in an AnyEvent program, you I<have> to create at least one
		481	watcher before you C<fork> the child (alternatively, you can call
		482	C<AnyEvent::detect>).
		483
		484	As most event loops do not support waiting for child events, they will be
		485	emulated by AnyEvent in most cases, in which the latency and race problems
		486	mentioned in the description of signal watchers apply.
		487
		488	Example: fork a process and wait for it
		489
		490	my $done = AnyEvent->condvar;
		491
		492	my $pid = fork or exit 5;
		493
		494	my $w = AnyEvent->child (
		495	pid => $pid,
		496	cb => sub {
		497	my ($pid, $status) = @_;
		498	warn "pid $pid exited with status $status";
		499	$done->send;
		500	},
		501	);
		502
		503	# do something else, then wait for process exit
		504	$done->recv;
		505
		506	=head2 IDLE WATCHERS
		507
		508	$w = AnyEvent->idle (cb => <callback>);
		509
		510	This will repeatedly invoke the callback after the process becomes idle,
		511	until either the watcher is destroyed or new events have been detected.
		512
		513	Idle watchers are useful when there is a need to do something, but it
		514	is not so important (or wise) to do it instantly. The callback will be
		515	invoked only when there is "nothing better to do", which is usually
		516	defined as "all outstanding events have been handled and no new events
		517	have been detected". That means that idle watchers ideally get invoked
		518	when the event loop has just polled for new events but none have been
		519	detected. Instead of blocking to wait for more events, the idle watchers
		520	will be invoked.
		521
		522	Unfortunately, most event loops do not really support idle watchers (only
		523	EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
		524	will simply call the callback "from time to time".
		525
		526	Example: read lines from STDIN, but only process them when the
		527	program is otherwise idle:
		528
		529	my @lines; # read data
		530	my $idle_w;
		531	my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		532	push @lines, scalar <STDIN>;
		533
		534	# start an idle watcher, if not already done
		535	$idle_w ||= AnyEvent->idle (cb => sub {
		536	# handle only one line, when there are lines left
		537	if (my $line = shift @lines) {
		538	print "handled when idle: $line";
		539	} else {
		540	# otherwise disable the idle watcher again
		541	undef $idle_w;
		542	}
		543	});
		544	});
		545
		546	=head2 CONDITION VARIABLES
		547
		548	$cv = AnyEvent->condvar;
		549
		550	$cv->send (<list>);
		551	my @res = $cv->recv;
		552
		553	If you are familiar with some event loops you will know that all of them
		554	require you to run some blocking "loop", "run" or similar function that
		555	will actively watch for new events and call your callbacks.
		556
		557	AnyEvent is slightly different: it expects somebody else to run the event
		558	loop and will only block when necessary (usually when told by the user).
		559
		560	The tool to do that is called a "condition variable", so called because
		561	they represent a condition that must become true.
		562
		563	Now is probably a good time to look at the examples further below.
		564
		565	Condition variables can be created by calling the C<< AnyEvent->condvar
		566	>> method, usually without arguments. The only argument pair allowed is
		567	C<cb>, which specifies a callback to be called when the condition variable
		568	becomes true, with the condition variable as the first argument (but not
		569	the results).
		570
		571	After creation, the condition variable is "false" until it becomes "true"
		572	by calling the C<send> method (or calling the condition variable as if it
		573	were a callback, read about the caveats in the description for the C<<
		574	->send >> method).
		575
		576	Since condition variables are the most complex part of the AnyEvent API, here are
		577	some different mental models of what they are - pick the ones you can connect to:
		578
		579	=over 4
		580
		581	=item * Condition variables are like callbacks - you can call them (and pass them instead
		582	of callbacks). Unlike callbacks however, you can also wait for them to be called.
		583
		584	=item * Condition variables are signals - one side can emit or send them,
		585	the other side can wait for them, or install a handler that is called when
		586	the signal fires.
		587
		588	=item * Condition variables are like "Merge Points" - points in your program
		589	where you merge multiple independent results/control flows into one.
		590
		591	=item * Condition variables represent a transaction - functions that start
		592	some kind of transaction can return them, leaving the caller the choice
		593	between waiting in a blocking fashion, or setting a callback.
		594
		595	=item * Condition variables represent future values, or promises to deliver
		596	some result, long before the result is available.
		597
		598	=back
		599
		600	Condition variables are very useful to signal that something has finished,
		601	for example, if you write a module that does asynchronous http requests,
		602	then a condition variable would be the ideal candidate to signal the
		603	availability of results. The user can either act when the callback is
		604	called or can synchronously C<< ->recv >> for the results.
		605
		606	You can also use them to simulate traditional event loops - for example,
		607	you can block your main program until an event occurs - for example, you
		608	could C<< ->recv >> in your main program until the user clicks the Quit
		609	button of your app, which would C<< ->send >> the "quit" event.
		610
		611	Note that condition variables recurse into the event loop - if you have
		612	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		613	lose. Therefore, condition variables are good to export to your caller, but
		614	you should avoid making a blocking wait yourself, at least in callbacks,
		615	as this asks for trouble.
		616
		617	Condition variables are represented by hash refs in perl, and the keys
		618	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		619	easy (it is often useful to build your own transaction class on top of
		620	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		621	its C<new> method in your own C<new> method.
		622
		623	There are two "sides" to a condition variable - the "producer side" which
		624	eventually calls C<< -> send >>, and the "consumer side", which waits
		625	for the send to occur.
		626
		627	Example: wait for a timer.
		628
		629	# condition: "wait till the timer is fired"
		630	my $timer_fired = AnyEvent->condvar;
		631
		632	# create the timer - we could wait for, say
		633	# a handle becomign ready, or even an
		634	# AnyEvent::HTTP request to finish, but
		635	# in this case, we simply use a timer:
		636	my $w = AnyEvent->timer (
		637	after => 1,
		638	cb => sub { $timer_fired->send },
		639	);
		640
		641	# this "blocks" (while handling events) till the callback
		642	# calls ->send
		643	$timer_fired->recv;
		644
		645	Example: wait for a timer, but take advantage of the fact that condition
		646	variables are also callable directly.
		647
		648	my $done = AnyEvent->condvar;
		649	my $delay = AnyEvent->timer (after => 5, cb => $done);
		650	$done->recv;
		651
		652	Example: Imagine an API that returns a condvar and doesn't support
		653	callbacks. This is how you make a synchronous call, for example from
		654	the main program:
		655
		656	use AnyEvent::CouchDB;
		657
		658	...
		659
		660	my @info = $couchdb->info->recv;
		661
		662	And this is how you would just set a callback to be called whenever the
		663	results are available:
		664
		665	$couchdb->info->cb (sub {
		666	my @info = $_[0]->recv;
		667	});
		668
		669	=head3 METHODS FOR PRODUCERS
		670
		671	These methods should only be used by the producing side, i.e. the
		672	code/module that eventually sends the signal. Note that it is also
		673	the producer side which creates the condvar in most cases, but it isn't
		674	uncommon for the consumer to create it as well.
		675
		676	=over 4
		677
		678	=item $cv->send (...)
		679
		680	Flag the condition as ready - a running C<< ->recv >> and all further
		681	calls to C<recv> will (eventually) return after this method has been
		682	called. If nobody is waiting the send will be remembered.
		683
		684	If a callback has been set on the condition variable, it is called
		685	immediately from within send.
		686
		687	Any arguments passed to the C<send> call will be returned by all
		688	future C<< ->recv >> calls.
		689
		690	Condition variables are overloaded so one can call them directly (as if
		691	they were a code reference). Calling them directly is the same as calling
		692	C<send>.
		693
		694	=item $cv->croak ($error)
		695
		696	Similar to send, but causes all calls to C<< ->recv >> to invoke
		697	C<Carp::croak> with the given error message/object/scalar.
		698
		699	This can be used to signal any errors to the condition variable
		700	user/consumer. Doing it this way instead of calling C<croak> directly
		701	delays the error detection, but has the overwhelming advantage that it
		702	diagnoses the error at the place where the result is expected, and not
		703	deep in some event callback with no connection to the actual code causing
		704	the problem.
		705
		706	=item $cv->begin ([group callback])
		707
		708	=item $cv->end
		709
		710	These two methods can be used to combine many transactions/events into
		711	one. For example, a function that pings many hosts in parallel might want
		712	to use a condition variable for the whole process.
		713
		714	Every call to C<< ->begin >> will increment a counter, and every call to
		715	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		716	>>, the (last) callback passed to C<begin> will be executed, passing the
		717	condvar as first argument. That callback is I<supposed> to call C<< ->send
		718	>>, but that is not required. If no group callback was set, C<send> will
		719	be called without any arguments.
		720
		721	You can think of C<< $cv->send >> giving you an OR condition (one call
		722	sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
		723	condition (all C<begin> calls must be C<end>'ed before the condvar sends).
		724
		725	Let's start with a simple example: you have two I/O watchers (for example,
		726	STDOUT and STDERR for a program), and you want to wait for both streams to
		727	close before activating a condvar:
		728
		729	my $cv = AnyEvent->condvar;
		730
		731	$cv->begin; # first watcher
		732	my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
		733	defined sysread $fh1, my $buf, 4096
		734	or $cv->end;
		735	});
		736
		737	$cv->begin; # second watcher
		738	my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
		739	defined sysread $fh2, my $buf, 4096
		740	or $cv->end;
		741	});
		742
		743	$cv->recv;
		744
		745	This works because for every event source (EOF on file handle), there is
		746	one call to C<begin>, so the condvar waits for all calls to C<end> before
		747	sending.
		748
		749	The ping example mentioned above is slightly more complicated, as the
		750	there are results to be passwd back, and the number of tasks that are
		751	begun can potentially be zero:
		752
		753	my $cv = AnyEvent->condvar;
		754
		755	my %result;
		756	$cv->begin (sub { shift->send (\%result) });
		757
		758	for my $host (@list_of_hosts) {
		759	$cv->begin;
		760	ping_host_then_call_callback $host, sub {
		761	$result{$host} = ...;
		762	$cv->end;
		763	};
		764	}
		765
		766	$cv->end;
		767
		768	This code fragment supposedly pings a number of hosts and calls
		769	C<send> after results for all then have have been gathered - in any
		770	order. To achieve this, the code issues a call to C<begin> when it starts
		771	each ping request and calls C<end> when it has received some result for
		772	it. Since C<begin> and C<end> only maintain a counter, the order in which
		773	results arrive is not relevant.
		774
		775	There is an additional bracketing call to C<begin> and C<end> outside the
		776	loop, which serves two important purposes: first, it sets the callback
		777	to be called once the counter reaches C<0>, and second, it ensures that
		778	C<send> is called even when C<no> hosts are being pinged (the loop
		779	doesn't execute once).
		780
		781	This is the general pattern when you "fan out" into multiple (but
		782	potentially zero) subrequests: use an outer C<begin>/C<end> pair to set
		783	the callback and ensure C<end> is called at least once, and then, for each
		784	subrequest you start, call C<begin> and for each subrequest you finish,
		785	call C<end>.
		786
		787	=back
		788
		789	=head3 METHODS FOR CONSUMERS
		790
		791	These methods should only be used by the consuming side, i.e. the
		792	code awaits the condition.
		793
		794	=over 4
		795
		796	=item $cv->recv
		797
		798	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		799	>> methods have been called on C<$cv>, while servicing other watchers
		800	normally.
		801
		802	You can only wait once on a condition - additional calls are valid but
		803	will return immediately.
		804
		805	If an error condition has been set by calling C<< ->croak >>, then this
		806	function will call C<croak>.
		807
		808	In list context, all parameters passed to C<send> will be returned,
		809	in scalar context only the first one will be returned.
		810
		811	Note that doing a blocking wait in a callback is not supported by any
		812	event loop, that is, recursive invocation of a blocking C<< ->recv
		813	>> is not allowed, and the C<recv> call will C<croak> if such a
		814	condition is detected. This condition can be slightly loosened by using
		815	L<Coro::AnyEvent>, which allows you to do a blocking C<< ->recv >> from
		816	any thread that doesn't run the event loop itself.
		817
		818	Not all event models support a blocking wait - some die in that case
		819	(programs might want to do that to stay interactive), so I<if you are
		820	using this from a module, never require a blocking wait>. Instead, let the
		821	caller decide whether the call will block or not (for example, by coupling
		822	condition variables with some kind of request results and supporting
		823	callbacks so the caller knows that getting the result will not block,
		824	while still supporting blocking waits if the caller so desires).
		825
		826	You can ensure that C<< ->recv >> never blocks by setting a callback and
		827	only calling C<< ->recv >> from within that callback (or at a later
		828	time). This will work even when the event loop does not support blocking
		829	waits otherwise.
		830
		831	=item $bool = $cv->ready
		832
		833	Returns true when the condition is "true", i.e. whether C<send> or
		834	C<croak> have been called.
		835
		836	=item $cb = $cv->cb ($cb->($cv))
		837
		838	This is a mutator function that returns the callback set and optionally
		839	replaces it before doing so.
		840
		841	The callback will be called when the condition becomes "true", i.e. when
		842	C<send> or C<croak> are called, with the only argument being the
		843	condition variable itself. If the condition is already true, the
		844	callback is called immediately when it is set. Calling C<recv> inside
		845	the callback or at any later time is guaranteed not to block.
		846
		847	=back
		848
		849	=head1 SUPPORTED EVENT LOOPS/BACKENDS
		850
		851	The available backend classes are (every class has its own manpage):
		852
		853	=over 4
		854
		855	=item Backends that are autoprobed when no other event loop can be found.
		856
		857	EV is the preferred backend when no other event loop seems to be in
		858	use. If EV is not installed, then AnyEvent will fall back to its own
		859	pure-perl implementation, which is available everywhere as it comes with
		860	AnyEvent itself.
		861
		862	AnyEvent::Impl::EV based on EV (interface to libev, best choice).
		863	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		864
		865	=item Backends that are transparently being picked up when they are used.
		866
		867	These will be used if they are already loaded when the first watcher
		868	is created, in which case it is assumed that the application is using
		869	them. This means that AnyEvent will automatically pick the right backend
		870	when the main program loads an event module before anything starts to
		871	create watchers. Nothing special needs to be done by the main program.
		872
		873	AnyEvent::Impl::Event based on Event, very stable, few glitches.
		874	AnyEvent::Impl::Glib based on Glib, slow but very stable.
		875	AnyEvent::Impl::Tk based on Tk, very broken.
		876	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		877	AnyEvent::Impl::POE based on POE, very slow, some limitations.
		878	AnyEvent::Impl::Irssi used when running within irssi.
		879
		880	=item Backends with special needs.
		881
		882	Qt requires the Qt::Application to be instantiated first, but will
		883	otherwise be picked up automatically. As long as the main program
		884	instantiates the application before any AnyEvent watchers are created,
		885	everything should just work.
		886
		887	AnyEvent::Impl::Qt based on Qt.
		888
		889	Support for IO::Async can only be partial, as it is too broken and
		890	architecturally limited to even support the AnyEvent API. It also
		891	is the only event loop that needs the loop to be set explicitly, so
		892	it can only be used by a main program knowing about AnyEvent. See
		893	L<AnyEvent::Impl::IOAsync> for the gory details.
		894
		895	AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed.
		896
		897	=item Event loops that are indirectly supported via other backends.
		898
		899	Some event loops can be supported via other modules:
		900
		901	There is no direct support for WxWidgets (L<Wx>) or L<Prima>.
		902
		903	B<WxWidgets> has no support for watching file handles. However, you can
		904	use WxWidgets through the POE adaptor, as POE has a Wx backend that simply
		905	polls 20 times per second, which was considered to be too horrible to even
		906	consider for AnyEvent.
		907
		908	B<Prima> is not supported as nobody seems to be using it, but it has a POE
		909	backend, so it can be supported through POE.
		910
		911	AnyEvent knows about both L<Prima> and L<Wx>, however, and will try to
		912	load L<POE> when detecting them, in the hope that POE will pick them up,
		913	in which case everything will be automatic.
		914
		915	=back
		916
		917	=head1 GLOBAL VARIABLES AND FUNCTIONS
		918
		919	These are not normally required to use AnyEvent, but can be useful to
		920	write AnyEvent extension modules.
		921
		922	=over 4
		923
		924	=item $AnyEvent::MODEL
		925
		926	Contains C<undef> until the first watcher is being created, before the
		927	backend has been autodetected.
		928
		929	Afterwards it contains the event model that is being used, which is the
		930	name of the Perl class implementing the model. This class is usually one
		931	of the C<AnyEvent::Impl::xxx> modules, but can be any other class in the
		932	case AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode> it
		933	will be C<urxvt::anyevent>).
		934
		935	=item AnyEvent::detect
		936
		937	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
		938	if necessary. You should only call this function right before you would
		939	have created an AnyEvent watcher anyway, that is, as late as possible at
		940	runtime, and not e.g. during initialisation of your module.
		941
		942	If you need to do some initialisation before AnyEvent watchers are
		943	created, use C<post_detect>.
		944
		945	=item $guard = AnyEvent::post_detect { BLOCK }
		946
		947	Arranges for the code block to be executed as soon as the event model is
		948	autodetected (or immediately if that has already happened).
		949
		950	The block will be executed I<after> the actual backend has been detected
		951	(C<$AnyEvent::MODEL> is set), but I<before> any watchers have been
		952	created, so it is possible to e.g. patch C<@AnyEvent::ISA> or do
		953	other initialisations - see the sources of L<AnyEvent::Strict> or
		954	L<AnyEvent::AIO> to see how this is used.
		955
		956	The most common usage is to create some global watchers, without forcing
		957	event module detection too early, for example, L<AnyEvent::AIO> creates
		958	and installs the global L<IO::AIO> watcher in a C<post_detect> block to
		959	avoid autodetecting the event module at load time.
		960
		961	If called in scalar or list context, then it creates and returns an object
		962	that automatically removes the callback again when it is destroyed (or
		963	C<undef> when the hook was immediately executed). See L<AnyEvent::AIO> for
		964	a case where this is useful.
		965
		966	Example: Create a watcher for the IO::AIO module and store it in
		967	C<$WATCHER>, but do so only do so after the event loop is initialised.
		968
		969	our WATCHER;
		970
		971	my $guard = AnyEvent::post_detect {
		972	$WATCHER = AnyEvent->io (fh => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb);
		973	};
		974
		975	# the ||= is important in case post_detect immediately runs the block,
		976	# as to not clobber the newly-created watcher. assigning both watcher and
		977	# post_detect guard to the same variable has the advantage of users being
		978	# able to just C<undef $WATCHER> if the watcher causes them grief.
		979
		980	$WATCHER ||= $guard;
		981
		982	=item @AnyEvent::post_detect
		983
		984	If there are any code references in this array (you can C<push> to it
		985	before or after loading AnyEvent), then they will be called directly
		986	after the event loop has been chosen.
		987
		988	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		989	if it is defined then the event loop has already been detected, and the
		990	array will be ignored.
		991
		992	Best use C<AnyEvent::post_detect { BLOCK }> when your application allows
		993	it, as it takes care of these details.
		994
		995	This variable is mainly useful for modules that can do something useful
		996	when AnyEvent is used and thus want to know when it is initialised, but do
		997	not need to even load it by default. This array provides the means to hook
		998	into AnyEvent passively, without loading it.
		999
		1000	Example: To load Coro::AnyEvent whenever Coro and AnyEvent are used
		1001	together, you could put this into Coro (this is the actual code used by
		1002	Coro to accomplish this):
		1003
		1004	if (defined $AnyEvent::MODEL) {
		1005	# AnyEvent already initialised, so load Coro::AnyEvent
		1006	require Coro::AnyEvent;
		1007	} else {
		1008	# AnyEvent not yet initialised, so make sure to load Coro::AnyEvent
		1009	# as soon as it is
		1010	push @AnyEvent::post_detect, sub { require Coro::AnyEvent };
		1011	}
		1012
		1013	=back
		1014
		1015	=head1 WHAT TO DO IN A MODULE
		1016
		1017	As a module author, you should C<use AnyEvent> and call AnyEvent methods
		1018	freely, but you should not load a specific event module or rely on it.
		1019
		1020	Be careful when you create watchers in the module body - AnyEvent will
		1021	decide which event module to use as soon as the first method is called, so
		1022	by calling AnyEvent in your module body you force the user of your module
		1023	to load the event module first.
		1024
		1025	Never call C<< ->recv >> on a condition variable unless you I<know> that
		1026	the C<< ->send >> method has been called on it already. This is
		1027	because it will stall the whole program, and the whole point of using
		1028	events is to stay interactive.
		1029
		1030	It is fine, however, to call C<< ->recv >> when the user of your module
		1031	requests it (i.e. if you create a http request object ad have a method
		1032	called C<results> that returns the results, it may call C<< ->recv >>
		1033	freely, as the user of your module knows what she is doing. Always).
		1034
		1035	=head1 WHAT TO DO IN THE MAIN PROGRAM
		1036
		1037	There will always be a single main program - the only place that should
		1038	dictate which event model to use.
		1039
		1040	If the program is not event-based, it need not do anything special, even
		1041	when it depends on a module that uses an AnyEvent. If the program itself
		1042	uses AnyEvent, but does not care which event loop is used, all it needs
		1043	to do is C<use AnyEvent>. In either case, AnyEvent will choose the best
		1044	available loop implementation.
		1045
		1046	If the main program relies on a specific event model - for example, in
		1047	Gtk2 programs you have to rely on the Glib module - you should load the
		1048	event module before loading AnyEvent or any module that uses it: generally
		1049	speaking, you should load it as early as possible. The reason is that
		1050	modules might create watchers when they are loaded, and AnyEvent will
		1051	decide on the event model to use as soon as it creates watchers, and it
		1052	might choose the wrong one unless you load the correct one yourself.
		1053
		1054	You can chose to use a pure-perl implementation by loading the
		1055	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
		1056	everywhere, but letting AnyEvent chose the model is generally better.
		1057
		1058	=head2 MAINLOOP EMULATION
		1059
		1060	Sometimes (often for short test scripts, or even standalone programs who
		1061	only want to use AnyEvent), you do not want to run a specific event loop.
		1062
		1063	In that case, you can use a condition variable like this:
		1064
		1065	AnyEvent->condvar->recv;
		1066
		1067	This has the effect of entering the event loop and looping forever.
		1068
		1069	Note that usually your program has some exit condition, in which case
		1070	it is better to use the "traditional" approach of storing a condition
		1071	variable somewhere, waiting for it, and sending it when the program should
		1072	exit cleanly.
		1073
		1074
		1075	=head1 OTHER MODULES
		1076
		1077	The following is a non-exhaustive list of additional modules that use
		1078	AnyEvent as a client and can therefore be mixed easily with other AnyEvent
		1079	modules and other event loops in the same program. Some of the modules
		1080	come as part of AnyEvent, the others are available via CPAN.
		1081
		1082	=over 4
		1083
		1084	=item L<AnyEvent::Util>
		1085
		1086	Contains various utility functions that replace often-used blocking
		1087	functions such as C<inet_aton> with event/callback-based versions.
		1088
		1089	=item L<AnyEvent::Socket>
		1090
		1091	Provides various utility functions for (internet protocol) sockets,
		1092	addresses and name resolution. Also functions to create non-blocking tcp
		1093	connections or tcp servers, with IPv6 and SRV record support and more.
		1094
		1095	=item L<AnyEvent::Handle>
		1096
		1097	Provide read and write buffers, manages watchers for reads and writes,
		1098	supports raw and formatted I/O, I/O queued and fully transparent and
		1099	non-blocking SSL/TLS (via L<AnyEvent::TLS>).
		1100
		1101	=item L<AnyEvent::DNS>
		1102
		1103	Provides rich asynchronous DNS resolver capabilities.
		1104
		1105	=item L<AnyEvent::HTTP>, L<AnyEvent::IRC>, L<AnyEvent::XMPP>, L<AnyEvent::GPSD>, L<AnyEvent::IGS>, L<AnyEvent::FCP>
		1106
		1107	Implement event-based interfaces to the protocols of the same name (for
		1108	the curious, IGS is the International Go Server and FCP is the Freenet
		1109	Client Protocol).
		1110
		1111	=item L<AnyEvent::Handle::UDP>
		1112
		1113	Here be danger!
		1114
		1115	As Pauli would put it, "Not only is it not right, it's not even wrong!" -
		1116	there are so many things wrong with AnyEvent::Handle::UDP, most notably
		1117	its use of a stream-based API with a protocol that isn't streamable, that
		1118	the only way to improve it is to delete it.
		1119
		1120	It features data corruption (but typically only under load) and general
		1121	confusion. On top, the author is not only clueless about UDP but also
		1122	fact-resistant - some gems of his understanding: "connect doesn't work
		1123	with UDP", "UDP packets are not IP packets", "UDP only has datagrams, not
		1124	packets", "I don't need to implement proper error checking as UDP doesn't
		1125	support error checking" and so on - he doesn't even understand what's
		1126	wrong with his module when it is explained to him.
		1127
		1128	=item L<AnyEvent::DBI>
		1129
		1130	Executes L<DBI> requests asynchronously in a proxy process for you,
		1131	notifying you in an event-based way when the operation is finished.
		1132
		1133	=item L<AnyEvent::AIO>
		1134
		1135	Truly asynchronous (as opposed to non-blocking) I/O, should be in the
		1136	toolbox of every event programmer. AnyEvent::AIO transparently fuses
		1137	L<IO::AIO> and AnyEvent together, giving AnyEvent access to event-based
		1138	file I/O, and much more.
		1139
		1140	=item L<AnyEvent::HTTPD>
		1141
		1142	A simple embedded webserver.
		1143
		1144	=item L<AnyEvent::FastPing>
		1145
		1146	The fastest ping in the west.
		1147
		1148	=item L<Coro>
		1149
		1150	Has special support for AnyEvent via L<Coro::AnyEvent>.
		1151
		1152	=back
		1153
		1154	=cut
		1155
		1156	package AnyEvent;
		1157
		1158	# basically a tuned-down version of common::sense
		1159	sub common_sense {
		1160	# from common:.sense 3.3
		1161	${^WARNING_BITS} ^= ${^WARNING_BITS} ^ "\x3c\x3f\x33\x00\x0f\xf3\x0f\xc0\xf0\xfc\x33\x00";
		1162	# use strict vars subs - NO UTF-8, as Util.pm doesn't like this atm. (uts46data.pl)
		1163	$^H |= 0x00000600;
		1164	}
		1165
		1166	BEGIN { AnyEvent::common_sense }
		1167
		1168	use Carp ();
		1169
		1170	our $VERSION = '5.271';
		1171	our $MODEL;
		1172
		1173	our $AUTOLOAD;
		1174	our @ISA;
		1175
		1176	our @REGISTRY;
		1177
		1178	our $VERBOSE;
		1179
		1180	BEGIN {
		1181	require "AnyEvent/constants.pl";
		1182
		1183	eval "sub TAINT (){" . (${^TAINT}*1) . "}";
		1184
		1185	delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
		1186	if ${^TAINT};
		1187
		1188	$VERBOSE = $ENV{PERL_ANYEVENT_VERBOSE}*1;
		1189
		1190	}
		1191
		1192	our $MAX_SIGNAL_LATENCY = 10;
		1193
		1194	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		1195
		1196	{
		1197	my $idx;
		1198	$PROTOCOL{$_} = ++$idx
		1199	for reverse split /\s*,\s*/,
		1200	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		1201	}
		1202
		1203	my @models = (
		1204	[EV:: => AnyEvent::Impl::EV:: , 1],
		1205	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl:: , 1],
		1206	# everything below here will not (normally) be autoprobed
		1207	# as the pureperl backend should work everywhere
		1208	# and is usually faster
		1209	[Event:: => AnyEvent::Impl::Event::, 1],
		1210	[Glib:: => AnyEvent::Impl::Glib:: , 1], # becomes extremely slow with many watchers
		1211	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		1212	[Irssi:: => AnyEvent::Impl::Irssi::], # Irssi has a bogus "Event" package
		1213	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		1214	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		1215	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		1216	[Wx:: => AnyEvent::Impl::POE::],
		1217	[Prima:: => AnyEvent::Impl::POE::],
		1218	# IO::Async is just too broken - we would need workarounds for its
		1219	# byzantine signal and broken child handling, among others.
		1220	# IO::Async is rather hard to detect, as it doesn't have any
		1221	# obvious default class.
		1222	[IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1223	[IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1224	[IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1225	[AnyEvent::Impl::IOAsync:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1226	);
		1227
		1228	our %method = map +($_ => 1),
		1229	qw(io timer time now now_update signal child idle condvar one_event DESTROY);
		1230
		1231	our @post_detect;
		1232
		1233	sub post_detect(&) {
		1234	my ($cb) = @_;
		1235
		1236	push @post_detect, $cb;
		1237
		1238	defined wantarray
		1239	? bless \$cb, "AnyEvent::Util::postdetect"
		1240	: ()
		1241	}
		1242
		1243	sub AnyEvent::Util::postdetect::DESTROY {
		1244	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		1245	}
		1246
		1247	sub detect() {
		1248	# free some memory
		1249	*detect = sub () { $MODEL };
		1250
		1251	local $!; # for good measure
		1252	local $SIG{__DIE__};
		1253
		1254	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
		1255	my $model = "AnyEvent::Impl::$1";
		1256	if (eval "require $model") {
		1257	$MODEL = $model;
		1258	warn "AnyEvent: loaded model '$model' (forced by \$ENV{PERL_ANYEVENT_MODEL}), using it.\n" if $VERBOSE >= 2;
		1259	} else {
		1260	warn "AnyEvent: unable to load model '$model' (from \$ENV{PERL_ANYEVENT_MODEL}):\n$@" if $VERBOSE;
		1261	}
		1262	}
		1263
		1264	# check for already loaded models
		1265	unless ($MODEL) {
		1266	for (@REGISTRY, @models) {
		1267	my ($package, $model) = @$_;
		1268	if (${"$package\::VERSION"} > 0) {
		1269	if (eval "require $model") {
		1270	$MODEL = $model;
		1271	warn "AnyEvent: autodetected model '$model', using it.\n" if $VERBOSE >= 2;
		1272	last;
		1273	}
		1274	}
		1275	}
		1276
		1277	unless ($MODEL) {
		1278	# try to autoload a model
		1279	for (@REGISTRY, @models) {
		1280	my ($package, $model, $autoload) = @$_;
		1281	if (
		1282	$autoload
		1283	and eval "require $package"
		1284	and ${"$package\::VERSION"} > 0
		1285	and eval "require $model"
		1286	) {
		1287	$MODEL = $model;
		1288	warn "AnyEvent: autoloaded model '$model', using it.\n" if $VERBOSE >= 2;
		1289	last;
		1290	}
		1291	}
		1292
		1293	$MODEL
		1294	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
		1295	}
		1296	}
		1297
		1298	@models = (); # free probe data
		1299
		1300	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		1301	unshift @ISA, $MODEL;
		1302
		1303	# now nuke some methods that are overriden by the backend.
		1304	# SUPER is not allowed.
		1305	for (qw(time signal child idle)) {
		1306	undef &{"AnyEvent::Base::$_"}
		1307	if defined &{"$MODEL\::$_"};
		1308	}
		1309
		1310	require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
		1311
		1312	(shift @post_detect)->() while @post_detect;
		1313
		1314	*post_detect = sub(&) {
		1315	shift->();
		1316
		1317	undef
		1318	};
		1319
		1320	$MODEL
		1321	}
		1322
		1323	sub AUTOLOAD {
		1324	(my $func = $AUTOLOAD) =~ s/.*://;
		1325
		1326	$method{$func}
		1327	or Carp::croak "$func: not a valid AnyEvent class method";
		1328
		1329	detect;
		1330
		1331	my $class = shift;
		1332	$class->$func (@_);
		1333	}
		1334
		1335	# utility function to dup a filehandle. this is used by many backends
		1336	# to support binding more than one watcher per filehandle (they usually
		1337	# allow only one watcher per fd, so we dup it to get a different one).
		1338	sub _dupfh($$;$$) {
		1339	my ($poll, $fh, $r, $w) = @_;
		1340
		1341	# cygwin requires the fh mode to be matching, unix doesn't
		1342	my ($rw, $mode) = $poll eq "r" ? ($r, "<&") : ($w, ">&");
		1343
		1344	open my $fh2, $mode, $fh
		1345	or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,";
		1346
		1347	# we assume CLOEXEC is already set by perl in all important cases
		1348
		1349	($fh2, $rw)
		1350	}
		1351
		1352	=head1 SIMPLIFIED AE API
		1353
		1354	Starting with version 5.0, AnyEvent officially supports a second, much
		1355	simpler, API that is designed to reduce the calling, typing and memory
		1356	overhead by using function call syntax and a fixed number of parameters.
		1357
		1358	See the L<AE> manpage for details.
		1359
		1360	=cut
		1361
		1362	package AE;
		1363
		1364	our $VERSION = $AnyEvent::VERSION;
		1365
		1366	# fall back to the main API by default - backends and AnyEvent::Base
		1367	# implementations can overwrite these.
		1368
		1369	sub io($$$) {
		1370	AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2])
		1371	}
		1372
		1373	sub timer($$$) {
		1374	AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2])
		1375	}
		1376
		1377	sub signal($$) {
		1378	AnyEvent->signal (signal => $_[0], cb => $_[1])
		1379	}
		1380
		1381	sub child($$) {
		1382	AnyEvent->child (pid => $_[0], cb => $_[1])
		1383	}
		1384
		1385	sub idle($) {
		1386	AnyEvent->idle (cb => $_[0])
		1387	}
		1388
		1389	sub cv(;&) {
		1390	AnyEvent->condvar (@_ ? (cb => $_[0]) : ())
		1391	}
		1392
		1393	sub now() {
		1394	AnyEvent->now
		1395	}
		1396
		1397	sub now_update() {
		1398	AnyEvent->now_update
		1399	}
		1400
		1401	sub time() {
		1402	AnyEvent->time
		1403	}
		1404
		1405	package AnyEvent::Base;
		1406
		1407	# default implementations for many methods
		1408
		1409	sub time {
		1410	eval q{ # poor man's autoloading {}
		1411	# probe for availability of Time::HiRes
		1412	if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
		1413	warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8;
		1414	*AE::time = \&Time::HiRes::time;
		1415	# if (eval "use POSIX (); (POSIX::times())...
		1416	} else {
		1417	warn "AnyEvent: using built-in time(), WARNING, no sub-second resolution!\n" if $VERBOSE;
		1418	*AE::time = sub (){ time }; # epic fail
		1419	}
		1420
		1421	*time = sub { AE::time }; # different prototypes
		1422	};
		1423	die if $@;
		1424
		1425	&time
		1426	}
		1427
		1428	*now = \&time;
		1429
		1430	sub now_update { }
		1431
		1432	# default implementation for ->condvar
		1433
		1434	sub condvar {
		1435	eval q{ # poor man's autoloading {}
		1436	*condvar = sub {
		1437	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
		1438	};
		1439
		1440	*AE::cv = sub (;&) {
		1441	bless { @_ ? (_ae_cb => shift) : () }, "AnyEvent::CondVar"
		1442	};
		1443	};
		1444	die if $@;
		1445
		1446	&condvar
		1447	}
		1448
		1449	# default implementation for ->signal
		1450
		1451	our $HAVE_ASYNC_INTERRUPT;
		1452
		1453	sub _have_async_interrupt() {
		1454	$HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT}
		1455	&& eval "use Async::Interrupt 1.02 (); 1")
		1456	unless defined $HAVE_ASYNC_INTERRUPT;
		1457
		1458	$HAVE_ASYNC_INTERRUPT
		1459	}
		1460
		1461	our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
		1462	our (%SIG_ASY, %SIG_ASY_W);
		1463	our ($SIG_COUNT, $SIG_TW);
		1464
		1465	# install a dummy wakeup watcher to reduce signal catching latency
		1466	# used by Impls
		1467	sub _sig_add() {
		1468	unless ($SIG_COUNT++) {
		1469	# try to align timer on a full-second boundary, if possible
		1470	my $NOW = AE::now;
		1471
		1472	$SIG_TW = AE::timer
		1473	$MAX_SIGNAL_LATENCY - ($NOW - int $NOW),
		1474	$MAX_SIGNAL_LATENCY,
		1475	sub { } # just for the PERL_ASYNC_CHECK
		1476	;
		1477	}
		1478	}
		1479
		1480	sub _sig_del {
		1481	undef $SIG_TW
		1482	unless --$SIG_COUNT;
		1483	}
		1484
		1485	our $_sig_name_init; $_sig_name_init = sub {
		1486	eval q{ # poor man's autoloading {}
		1487	undef $_sig_name_init;
		1488
		1489	if (_have_async_interrupt) {
		1490	*sig2num = \&Async::Interrupt::sig2num;
		1491	*sig2name = \&Async::Interrupt::sig2name;
		1492	} else {
		1493	require Config;
		1494
		1495	my %signame2num;
		1496	@signame2num{ split ' ', $Config::Config{sig_name} }
		1497	= split ' ', $Config::Config{sig_num};
		1498
		1499	my @signum2name;
		1500	@signum2name[values %signame2num] = keys %signame2num;
		1501
		1502	*sig2num = sub($) {
		1503	$_[0] > 0 ? shift : $signame2num{+shift}
		1504	};
		1505	*sig2name = sub ($) {
		1506	$_[0] > 0 ? $signum2name[+shift] : shift
		1507	};
		1508	}
		1509	};
		1510	die if $@;
		1511	};
		1512
		1513	sub sig2num ($) { &$_sig_name_init; &sig2num }
		1514	sub sig2name($) { &$_sig_name_init; &sig2name }
		1515
		1516	sub signal {
		1517	eval q{ # poor man's autoloading {}
		1518	# probe for availability of Async::Interrupt
		1519	if (_have_async_interrupt) {
		1520	warn "AnyEvent: using Async::Interrupt for race-free signal handling.\n" if $VERBOSE >= 8;
		1521
		1522	$SIGPIPE_R = new Async::Interrupt::EventPipe;
		1523	$SIG_IO = AE::io $SIGPIPE_R->fileno, 0, \&_signal_exec;
		1524
		1525	} else {
		1526	warn "AnyEvent: using emulated perl signal handling with latency timer.\n" if $VERBOSE >= 8;
		1527
		1528	if (AnyEvent::WIN32) {
		1529	require AnyEvent::Util;
		1530
		1531	($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
		1532	AnyEvent::Util::fh_nonblocking ($SIGPIPE_R, 1) if $SIGPIPE_R;
		1533	AnyEvent::Util::fh_nonblocking ($SIGPIPE_W, 1) if $SIGPIPE_W; # just in case
		1534	} else {
		1535	pipe $SIGPIPE_R, $SIGPIPE_W;
		1536	fcntl $SIGPIPE_R, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_R;
		1537	fcntl $SIGPIPE_W, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_W; # just in case
		1538
		1539	# not strictly required, as $^F is normally 2, but let's make sure...
		1540	fcntl $SIGPIPE_R, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
		1541	fcntl $SIGPIPE_W, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
		1542	}
		1543
		1544	$SIGPIPE_R
		1545	or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
		1546
		1547	$SIG_IO = AE::io $SIGPIPE_R, 0, \&_signal_exec;
		1548	}
		1549
		1550	*signal = $HAVE_ASYNC_INTERRUPT
		1551	? sub {
		1552	my (undef, %arg) = @_;
		1553
		1554	# async::interrupt
		1555	my $signal = sig2num $arg{signal};
		1556	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1557
		1558	$SIG_ASY{$signal} ||= new Async::Interrupt
		1559	cb => sub { undef $SIG_EV{$signal} },
		1560	signal => $signal,
		1561	pipe => [$SIGPIPE_R->filenos],
		1562	pipe_autodrain => 0,
		1563	;
		1564
		1565	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
		1566	}
		1567	: sub {
		1568	my (undef, %arg) = @_;
		1569
		1570	# pure perl
		1571	my $signal = sig2name $arg{signal};
		1572	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1573
		1574	$SIG{$signal} ||= sub {
		1575	local $!;
		1576	syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
		1577	undef $SIG_EV{$signal};
		1578	};
		1579
		1580	# can't do signal processing without introducing races in pure perl,
		1581	# so limit the signal latency.
		1582	_sig_add;
		1583
		1584	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
		1585	}
		1586	;
		1587
		1588	*AnyEvent::Base::signal::DESTROY = sub {
		1589	my ($signal, $cb) = @{$_[0]};
		1590
		1591	_sig_del;
		1592
		1593	delete $SIG_CB{$signal}{$cb};
		1594
		1595	$HAVE_ASYNC_INTERRUPT
		1596	? delete $SIG_ASY{$signal}
		1597	: # delete doesn't work with older perls - they then
		1598	# print weird messages, or just unconditionally exit
		1599	# instead of getting the default action.
		1600	undef $SIG{$signal}
		1601	unless keys %{ $SIG_CB{$signal} };
		1602	};
		1603
		1604	*_signal_exec = sub {
		1605	$HAVE_ASYNC_INTERRUPT
		1606	? $SIGPIPE_R->drain
		1607	: sysread $SIGPIPE_R, (my $dummy), 9;
		1608
		1609	while (%SIG_EV) {
		1610	for (keys %SIG_EV) {
		1611	delete $SIG_EV{$_};
		1612	$_->() for values %{ $SIG_CB{$_} || {} };
		1613	}
		1614	}
		1615	};
		1616	};
		1617	die if $@;
		1618
		1619	&signal
		1620	}
		1621
		1622	# default implementation for ->child
		1623
		1624	our %PID_CB;
		1625	our $CHLD_W;
		1626	our $CHLD_DELAY_W;
		1627	our $WNOHANG;
		1628
		1629	# used by many Impl's
		1630	sub _emit_childstatus($$) {
		1631	my (undef, $rpid, $rstatus) = @_;
		1632
		1633	$_->($rpid, $rstatus)
		1634	for values %{ $PID_CB{$rpid} || {} },
		1635	values %{ $PID_CB{0} || {} };
		1636	}
		1637
		1638	sub child {
		1639	eval q{ # poor man's autoloading {}
		1640	*_sigchld = sub {
		1641	my $pid;
		1642
		1643	AnyEvent->_emit_childstatus ($pid, $?)
		1644	while ($pid = waitpid -1, $WNOHANG) > 0;
		1645	};
		1646
		1647	*child = sub {
		1648	my (undef, %arg) = @_;
		1649
		1650	defined (my $pid = $arg{pid} + 0)
		1651	or Carp::croak "required option 'pid' is missing";
		1652
		1653	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
		1654
		1655	# WNOHANG is almost cetrainly 1 everywhere
		1656	$WNOHANG ||= $^O =~ /^(?:openbsd|netbsd|linux|freebsd|cygwin|MSWin32)$/
		1657	? 1
		1658	: eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
		1659
		1660	unless ($CHLD_W) {
		1661	$CHLD_W = AE::signal CHLD => \&_sigchld;
		1662	# child could be a zombie already, so make at least one round
		1663	&_sigchld;
		1664	}
		1665
		1666	bless [$pid, $arg{cb}], "AnyEvent::Base::child"
		1667	};
		1668
		1669	*AnyEvent::Base::child::DESTROY = sub {
		1670	my ($pid, $cb) = @{$_[0]};
		1671
		1672	delete $PID_CB{$pid}{$cb};
		1673	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
		1674
		1675	undef $CHLD_W unless keys %PID_CB;
		1676	};
		1677	};
		1678	die if $@;
		1679
		1680	&child
		1681	}
		1682
		1683	# idle emulation is done by simply using a timer, regardless
		1684	# of whether the process is idle or not, and not letting
		1685	# the callback use more than 50% of the time.
		1686	sub idle {
		1687	eval q{ # poor man's autoloading {}
		1688	*idle = sub {
		1689	my (undef, %arg) = @_;
		1690
		1691	my ($cb, $w, $rcb) = $arg{cb};
		1692
		1693	$rcb = sub {
		1694	if ($cb) {
		1695	$w = _time;
		1696	&$cb;
		1697	$w = _time - $w;
		1698
		1699	# never use more then 50% of the time for the idle watcher,
		1700	# within some limits
		1701	$w = 0.0001 if $w < 0.0001;
		1702	$w = 5 if $w > 5;
		1703
		1704	$w = AE::timer $w, 0, $rcb;
		1705	} else {
		1706	# clean up...
		1707	undef $w;
		1708	undef $rcb;
		1709	}
		1710	};
		1711
		1712	$w = AE::timer 0.05, 0, $rcb;
		1713
		1714	bless \\$cb, "AnyEvent::Base::idle"
		1715	};
		1716
		1717	*AnyEvent::Base::idle::DESTROY = sub {
		1718	undef $${$_[0]};
		1719	};
		1720	};
		1721	die if $@;
		1722
		1723	&idle
		1724	}
		1725
		1726	package AnyEvent::CondVar;
		1727
		1728	our @ISA = AnyEvent::CondVar::Base::;
		1729
		1730	package AnyEvent::CondVar::Base;
		1731
		1732	#use overload
		1733	# '&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1734	# fallback => 1;
		1735
		1736	# save 300+ kilobytes by dirtily hardcoding overloading
		1737	${"AnyEvent::CondVar::Base::OVERLOAD"}{dummy}++; # Register with magic by touching.
		1738	*{'AnyEvent::CondVar::Base::()'} = sub { }; # "Make it findable via fetchmethod."
		1739	*{'AnyEvent::CondVar::Base::(&{}'} = sub { my $self = shift; sub { $self->send (@_) } }; # &{}
		1740	${'AnyEvent::CondVar::Base::()'} = 1; # fallback
		1741
		1742	our $WAITING;
		1743
		1744	sub _send {
		1745	# nop
		1746	}
		1747
		1748	sub send {
		1749	my $cv = shift;
		1750	$cv->{_ae_sent} = [@_];
		1751	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1752	$cv->_send;
		1753	}
		1754
		1755	sub croak {
		1756	$_[0]{_ae_croak} = $_[1];
		1757	$_[0]->send;
		1758	}
		1759
		1760	sub ready {
		1761	$_[0]{_ae_sent}
		1762	}
		1763
		1764	sub _wait {
		1765	$WAITING
		1766	and !$_[0]{_ae_sent}
		1767	and Carp::croak "AnyEvent::CondVar: recursive blocking wait detected";
		1768
		1769	local $WAITING = 1;
		1770	AnyEvent->one_event while !$_[0]{_ae_sent};
		1771	}
		1772
		1773	sub recv {
		1774	$_[0]->_wait;
		1775
		1776	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1777	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1778	}
		1779
		1780	sub cb {
		1781	my $cv = shift;
		1782
		1783	@_
		1784	and $cv->{_ae_cb} = shift
		1785	and $cv->{_ae_sent}
		1786	and (delete $cv->{_ae_cb})->($cv);
		1787
		1788	$cv->{_ae_cb}
		1789	}
		1790
		1791	sub begin {
		1792	++$_[0]{_ae_counter};
		1793	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1794	}
		1795
		1796	sub end {
		1797	return if --$_[0]{_ae_counter};
		1798	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1799	}
		1800
		1801	# undocumented/compatibility with pre-3.4
		1802	*broadcast = \&send;
		1803	*wait = \&_wait;
		1804
		1805	=head1 ERROR AND EXCEPTION HANDLING
		1806
		1807	In general, AnyEvent does not do any error handling - it relies on the
		1808	caller to do that if required. The L<AnyEvent::Strict> module (see also
		1809	the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
		1810	checking of all AnyEvent methods, however, which is highly useful during
		1811	development.
		1812
		1813	As for exception handling (i.e. runtime errors and exceptions thrown while
		1814	executing a callback), this is not only highly event-loop specific, but
		1815	also not in any way wrapped by this module, as this is the job of the main
		1816	program.
		1817
		1818	The pure perl event loop simply re-throws the exception (usually
		1819	within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
		1820	$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
		1821	so on.
		1822
		1823	=head1 ENVIRONMENT VARIABLES
		1824
		1825	The following environment variables are used by this module or its
		1826	submodules.
		1827
		1828	Note that AnyEvent will remove I<all> environment variables starting with
		1829	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		1830	enabled.
		1831
		1832	=over 4
		1833
		1834	=item C<PERL_ANYEVENT_VERBOSE>
		1835
		1836	By default, AnyEvent will be completely silent except in fatal
		1837	conditions. You can set this environment variable to make AnyEvent more
		1838	talkative.
		1839
		1840	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1841	conditions, such as not being able to load the event model specified by
		1842	C<PERL_ANYEVENT_MODEL>.
		1843
		1844	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1845	model it chooses.
		1846
		1847	When set to C<8> or higher, then AnyEvent will report extra information on
		1848	which optional modules it loads and how it implements certain features.
		1849
		1850	=item C<PERL_ANYEVENT_STRICT>
		1851
		1852	AnyEvent does not do much argument checking by default, as thorough
		1853	argument checking is very costly. Setting this variable to a true value
		1854	will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
		1855	check the arguments passed to most method calls. If it finds any problems,
		1856	it will croak.
		1857
		1858	In other words, enables "strict" mode.
		1859
		1860	Unlike C<use strict> (or its modern cousin, C<< use L<common::sense>
		1861	>>, it is definitely recommended to keep it off in production. Keeping
		1862	C<PERL_ANYEVENT_STRICT=1> in your environment while developing programs
		1863	can be very useful, however.
		1864
		1865	=item C<PERL_ANYEVENT_MODEL>
		1866
		1867	This can be used to specify the event model to be used by AnyEvent, before
		1868	auto detection and -probing kicks in. It must be a string consisting
		1869	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1870	and the resulting module name is loaded and if the load was successful,
		1871	used as event model. If it fails to load AnyEvent will proceed with
		1872	auto detection and -probing.
		1873
		1874	This functionality might change in future versions.
		1875
		1876	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1877	could start your program like this:
		1878
		1879	PERL_ANYEVENT_MODEL=Perl perl ...
		1880
		1881	=item C<PERL_ANYEVENT_PROTOCOLS>
		1882
		1883	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1884	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1885	of auto probing).
		1886
		1887	Must be set to a comma-separated list of protocols or address families,
		1888	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1889	used, and preference will be given to protocols mentioned earlier in the
		1890	list.
		1891
		1892	This variable can effectively be used for denial-of-service attacks
		1893	against local programs (e.g. when setuid), although the impact is likely
		1894	small, as the program has to handle conenction and other failures anyways.
		1895
		1896	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1897	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1898	- only support IPv4, never try to resolve or contact IPv6
		1899	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1900	IPv6, but prefer IPv6 over IPv4.
		1901
		1902	=item C<PERL_ANYEVENT_EDNS0>
		1903
		1904	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1905	for DNS. This extension is generally useful to reduce DNS traffic, but
		1906	some (broken) firewalls drop such DNS packets, which is why it is off by
		1907	default.
		1908
		1909	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1910	EDNS0 in its DNS requests.
		1911
		1912	=item C<PERL_ANYEVENT_MAX_FORKS>
		1913
		1914	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1915	will create in parallel.
		1916
		1917	=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
		1918
		1919	The default value for the C<max_outstanding> parameter for the default DNS
		1920	resolver - this is the maximum number of parallel DNS requests that are
		1921	sent to the DNS server.
		1922
		1923	=item C<PERL_ANYEVENT_RESOLV_CONF>
		1924
		1925	The file to use instead of F</etc/resolv.conf> (or OS-specific
		1926	configuration) in the default resolver. When set to the empty string, no
		1927	default config will be used.
		1928
		1929	=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
		1930
		1931	When neither C<ca_file> nor C<ca_path> was specified during
		1932	L<AnyEvent::TLS> context creation, and either of these environment
		1933	variables exist, they will be used to specify CA certificate locations
		1934	instead of a system-dependent default.
		1935
		1936	=item C<PERL_ANYEVENT_AVOID_GUARD> and C<PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT>
		1937
		1938	When these are set to C<1>, then the respective modules are not
		1939	loaded. Mostly good for testing AnyEvent itself.
		1940
		1941	=back
		1942
		1943	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1944
		1945	This is an advanced topic that you do not normally need to use AnyEvent in
		1946	a module. This section is only of use to event loop authors who want to
		1947	provide AnyEvent compatibility.
		1948
		1949	If you need to support another event library which isn't directly
		1950	supported by AnyEvent, you can supply your own interface to it by
		1951	pushing, before the first watcher gets created, the package name of
		1952	the event module and the package name of the interface to use onto
		1953	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
		1954	AnyEvent, so it is reasonably cheap.
		1955
		1956	Example:
		1957
		1958	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
		1959
		1960	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
		1961	package/class when it finds the C<urxvt> package/module is already loaded.
		1962
		1963	When AnyEvent is loaded and asked to find a suitable event model, it
		1964	will first check for the presence of urxvt by trying to C<use> the
		1965	C<urxvt::anyevent> module.
		1966
		1967	The class should provide implementations for all watcher types. See
		1968	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
		1969	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
		1970	see the sources.
		1971
		1972	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1973	provide suitable (hopefully) replacements.
		1974
		1975	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
		1976	terminal emulator uses the above line as-is. An interface isn't included
		1977	in AnyEvent because it doesn't make sense outside the embedded interpreter
		1978	inside I<rxvt-unicode>, and it is updated and maintained as part of the
		1979	I<rxvt-unicode> distribution.
		1980
		1981	I<rxvt-unicode> also cheats a bit by not providing blocking access to
		1982	condition variables: code blocking while waiting for a condition will
		1983	C<die>. This still works with most modules/usages, and blocking calls must
		1984	not be done in an interactive application, so it makes sense.
		1985
		1986	=head1 EXAMPLE PROGRAM
		1987
		1988	The following program uses an I/O watcher to read data from STDIN, a timer
		1989	to display a message once per second, and a condition variable to quit the
		1990	program when the user enters quit:
		1991
		1992	use AnyEvent;
		1993
		1994	my $cv = AnyEvent->condvar;
		1995
		1996	my $io_watcher = AnyEvent->io (
		1997	fh => \*STDIN,
		1998	poll => 'r',
		1999	cb => sub {
		2000	warn "io event <$_[0]>\n"; # will always output <r>
		2001	chomp (my $input = <STDIN>); # read a line
		2002	warn "read: $input\n"; # output what has been read
		2003	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		2004	},
		2005	);
		2006
		2007	my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub {
		2008	warn "timeout\n"; # print 'timeout' at most every second
		2009	});
		2010
		2011	$cv->recv; # wait until user enters /^q/i
		2012
		2013	=head1 REAL-WORLD EXAMPLE
		2014
		2015	Consider the L<Net::FCP> module. It features (among others) the following
		2016	API calls, which are to freenet what HTTP GET requests are to http:
		2017
		2018	my $data = $fcp->client_get ($url); # blocks
		2019
		2020	my $transaction = $fcp->txn_client_get ($url); # does not block
		2021	$transaction->cb ( sub { ... } ); # set optional result callback
		2022	my $data = $transaction->result; # possibly blocks
		2023
		2024	The C<client_get> method works like C<LWP::Simple::get>: it requests the
		2025	given URL and waits till the data has arrived. It is defined to be:
		2026
		2027	sub client_get { $_[0]->txn_client_get ($_[1])->result }
		2028
		2029	And in fact is automatically generated. This is the blocking API of
		2030	L<Net::FCP>, and it works as simple as in any other, similar, module.
		2031
		2032	More complicated is C<txn_client_get>: It only creates a transaction
		2033	(completion, result, ...) object and initiates the transaction.
		2034
		2035	my $txn = bless { }, Net::FCP::Txn::;
		2036
		2037	It also creates a condition variable that is used to signal the completion
		2038	of the request:
		2039
		2040	$txn->{finished} = AnyAvent->condvar;
		2041
		2042	It then creates a socket in non-blocking mode.
		2043
		2044	socket $txn->{fh}, ...;
		2045	fcntl $txn->{fh}, F_SETFL, O_NONBLOCK;
		2046	connect $txn->{fh}, ...
		2047	and !$!{EWOULDBLOCK}
		2048	and !$!{EINPROGRESS}
		2049	and Carp::croak "unable to connect: $!\n";
		2050
		2051	Then it creates a write-watcher which gets called whenever an error occurs
		2052	or the connection succeeds:
		2053
		2054	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'w', cb => sub { $txn->fh_ready_w });
		2055
		2056	And returns this transaction object. The C<fh_ready_w> callback gets
		2057	called as soon as the event loop detects that the socket is ready for
		2058	writing.
		2059
		2060	The C<fh_ready_w> method makes the socket blocking again, writes the
		2061	request data and replaces the watcher by a read watcher (waiting for reply
		2062	data). The actual code is more complicated, but that doesn't matter for
		2063	this example:
		2064
		2065	fcntl $txn->{fh}, F_SETFL, 0;
		2066	syswrite $txn->{fh}, $txn->{request}
		2067	or die "connection or write error";
		2068	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
		2069
		2070	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
		2071	result and signals any possible waiters that the request has finished:
		2072
		2073	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
		2074
		2075	if (end-of-file or data complete) {
		2076	$txn->{result} = $txn->{buf};
		2077	$txn->{finished}->send;
		2078	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
		2079	}
		2080
		2081	The C<result> method, finally, just waits for the finished signal (if the
		2082	request was already finished, it doesn't wait, of course, and returns the
		2083	data:
		2084
		2085	$txn->{finished}->recv;
		2086	return $txn->{result};
		2087
		2088	The actual code goes further and collects all errors (C<die>s, exceptions)
		2089	that occurred during request processing. The C<result> method detects
		2090	whether an exception as thrown (it is stored inside the $txn object)
		2091	and just throws the exception, which means connection errors and other
		2092	problems get reported to the code that tries to use the result, not in a
		2093	random callback.
		2094
		2095	All of this enables the following usage styles:
		2096
		2097	1. Blocking:
		2098
		2099	my $data = $fcp->client_get ($url);
		2100
		2101	2. Blocking, but running in parallel:
		2102
		2103	my @datas = map $_->result,
		2104	map $fcp->txn_client_get ($_),
		2105	@urls;
		2106
		2107	Both blocking examples work without the module user having to know
		2108	anything about events.
		2109
		2110	3a. Event-based in a main program, using any supported event module:
		2111
		2112	use EV;
		2113
		2114	$fcp->txn_client_get ($url)->cb (sub {
		2115	my $txn = shift;
		2116	my $data = $txn->result;
16	...	2117	...
17	});	2118	});
18		2119
19	# watchers get canceled whenever $w is destroyed	2120	EV::loop;
20	# only one watcher per $fh and $poll type is allowed
21	# (i.e. on a socket you cna have one r + one w or one rw
22	# watcher, not any more.
23	# timers can only be used once
24		2121
25	my $w = AnyEvent->condvar; # kind of main loop replacement	2122	3b. The module user could use AnyEvent, too:
26	# can only be used once
27	$w->wait; # enters main loop till $condvar gets ->send
28	$w->broadcast; # wake up waiting and future wait's
29		2123
30	=head1 DESCRIPTION	2124	use AnyEvent;
31		2125
32	L<AnyEvent> provides an identical interface to multiple event loops. This	2126	my $quit = AnyEvent->condvar;
33	allows module authors to utilizy an event loop without forcing module
34	users to use the same event loop (as only a single event loop can coexist
35	peacefully at any one time).
36		2127
37	The interface itself is vaguely similar but not identical to the Event	2128	$fcp->txn_client_get ($url)->cb (sub {
38	module.	2129	...
		2130	$quit->send;
		2131	});
39		2132
40	On the first call of any method, the module tries to detect the currently	2133	$quit->recv;
41	loaded event loop by probing wether any of the following modules is	2134
42	loaded: L<Coro::Event>, L<Event>, L<Glib>, L<Tk>. The first one found is	2135
43	used. If none is found, the module tries to load these modules in the	2136	=head1 BENCHMARKS
44	order given. The first one that could be successfully loaded will be	2137
45	used. If still none could be found, it will issue an error.	2138	To give you an idea of the performance and overheads that AnyEvent adds
		2139	over the event loops themselves and to give you an impression of the speed
		2140	of various event loops I prepared some benchmarks.
		2141
		2142	=head2 BENCHMARKING ANYEVENT OVERHEAD
		2143
		2144	Here is a benchmark of various supported event models used natively and
		2145	through AnyEvent. The benchmark creates a lot of timers (with a zero
		2146	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		2147	which it is), lets them fire exactly once and destroys them again.
		2148
		2149	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		2150	distribution. It uses the L<AE> interface, which makes a real difference
		2151	for the EV and Perl backends only.
		2152
		2153	=head3 Explanation of the columns
		2154
		2155	I<watcher> is the number of event watchers created/destroyed. Since
		2156	different event models feature vastly different performances, each event
		2157	loop was given a number of watchers so that overall runtime is acceptable
		2158	and similar between tested event loop (and keep them from crashing): Glib
		2159	would probably take thousands of years if asked to process the same number
		2160	of watchers as EV in this benchmark.
		2161
		2162	I<bytes> is the number of bytes (as measured by the resident set size,
		2163	RSS) consumed by each watcher. This method of measuring captures both C
		2164	and Perl-based overheads.
		2165
		2166	I<create> is the time, in microseconds (millionths of seconds), that it
		2167	takes to create a single watcher. The callback is a closure shared between
		2168	all watchers, to avoid adding memory overhead. That means closure creation
		2169	and memory usage is not included in the figures.
		2170
		2171	I<invoke> is the time, in microseconds, used to invoke a simple
		2172	callback. The callback simply counts down a Perl variable and after it was
		2173	invoked "watcher" times, it would C<< ->send >> a condvar once to
		2174	signal the end of this phase.
		2175
		2176	I<destroy> is the time, in microseconds, that it takes to destroy a single
		2177	watcher.
		2178
		2179	=head3 Results
		2180
		2181	name watchers bytes create invoke destroy comment
		2182	EV/EV 100000 223 0.47 0.43 0.27 EV native interface
		2183	EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers
		2184	Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal
		2185	Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation
		2186	Event/Event 16000 516 31.16 31.84 0.82 Event native interface
		2187	Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers
		2188	IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll
		2189	IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll
		2190	Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour
		2191	Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers
		2192	POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event
		2193	POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select
		2194
		2195	=head3 Discussion
		2196
		2197	The benchmark does I<not> measure scalability of the event loop very
		2198	well. For example, a select-based event loop (such as the pure perl one)
		2199	can never compete with an event loop that uses epoll when the number of
		2200	file descriptors grows high. In this benchmark, all events become ready at
		2201	the same time, so select/poll-based implementations get an unnatural speed
		2202	boost.
		2203
		2204	Also, note that the number of watchers usually has a nonlinear effect on
		2205	overall speed, that is, creating twice as many watchers doesn't take twice
		2206	the time - usually it takes longer. This puts event loops tested with a
		2207	higher number of watchers at a disadvantage.
		2208
		2209	To put the range of results into perspective, consider that on the
		2210	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		2211	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		2212	cycles with POE.
		2213
		2214	C<EV> is the sole leader regarding speed and memory use, which are both
		2215	maximal/minimal, respectively. When using the L<AE> API there is zero
		2216	overhead (when going through the AnyEvent API create is about 5-6 times
		2217	slower, with other times being equal, so still uses far less memory than
		2218	any other event loop and is still faster than Event natively).
		2219
		2220	The pure perl implementation is hit in a few sweet spots (both the
		2221	constant timeout and the use of a single fd hit optimisations in the perl
		2222	interpreter and the backend itself). Nevertheless this shows that it
		2223	adds very little overhead in itself. Like any select-based backend its
		2224	performance becomes really bad with lots of file descriptors (and few of
		2225	them active), of course, but this was not subject of this benchmark.
		2226
		2227	The C<Event> module has a relatively high setup and callback invocation
		2228	cost, but overall scores in on the third place.
		2229
		2230	C<IO::Async> performs admirably well, about on par with C<Event>, even
		2231	when using its pure perl backend.
		2232
		2233	C<Glib>'s memory usage is quite a bit higher, but it features a
		2234	faster callback invocation and overall ends up in the same class as
		2235	C<Event>. However, Glib scales extremely badly, doubling the number of
		2236	watchers increases the processing time by more than a factor of four,
		2237	making it completely unusable when using larger numbers of watchers
		2238	(note that only a single file descriptor was used in the benchmark, so
		2239	inefficiencies of C<poll> do not account for this).
		2240
		2241	The C<Tk> adaptor works relatively well. The fact that it crashes with
		2242	more than 2000 watchers is a big setback, however, as correctness takes
		2243	precedence over speed. Nevertheless, its performance is surprising, as the
		2244	file descriptor is dup()ed for each watcher. This shows that the dup()
		2245	employed by some adaptors is not a big performance issue (it does incur a
		2246	hidden memory cost inside the kernel which is not reflected in the figures
		2247	above).
		2248
		2249	C<POE>, regardless of underlying event loop (whether using its pure perl
		2250	select-based backend or the Event module, the POE-EV backend couldn't
		2251	be tested because it wasn't working) shows abysmal performance and
		2252	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		2253	as EV watchers, and 10 times as much memory as Event (the high memory
		2254	requirements are caused by requiring a session for each watcher). Watcher
		2255	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		2256	implementation.
		2257
		2258	The design of the POE adaptor class in AnyEvent can not really account
		2259	for the performance issues, though, as session creation overhead is
		2260	small compared to execution of the state machine, which is coded pretty
		2261	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		2262	using multiple sessions is not a good approach, especially regarding
		2263	memory usage, even the author of POE could not come up with a faster
		2264	design).
		2265
		2266	=head3 Summary
46		2267
47	=over 4	2268	=over 4
48		2269
		2270	=item * Using EV through AnyEvent is faster than any other event loop
		2271	(even when used without AnyEvent), but most event loops have acceptable
		2272	performance with or without AnyEvent.
		2273
		2274	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		2275	the actual event loop, only with extremely fast event loops such as EV
		2276	adds AnyEvent significant overhead.
		2277
		2278	=item * You should avoid POE like the plague if you want performance or
		2279	reasonable memory usage.
		2280
		2281	=back
		2282
		2283	=head2 BENCHMARKING THE LARGE SERVER CASE
		2284
		2285	This benchmark actually benchmarks the event loop itself. It works by
		2286	creating a number of "servers": each server consists of a socket pair, a
		2287	timeout watcher that gets reset on activity (but never fires), and an I/O
		2288	watcher waiting for input on one side of the socket. Each time the socket
		2289	watcher reads a byte it will write that byte to a random other "server".
		2290
		2291	The effect is that there will be a lot of I/O watchers, only part of which
		2292	are active at any one point (so there is a constant number of active
		2293	fds for each loop iteration, but which fds these are is random). The
		2294	timeout is reset each time something is read because that reflects how
		2295	most timeouts work (and puts extra pressure on the event loops).
		2296
		2297	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		2298	(1%) are active. This mirrors the activity of large servers with many
		2299	connections, most of which are idle at any one point in time.
		2300
		2301	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		2302	distribution. It uses the L<AE> interface, which makes a real difference
		2303	for the EV and Perl backends only.
		2304
		2305	=head3 Explanation of the columns
		2306
		2307	I<sockets> is the number of sockets, and twice the number of "servers" (as
		2308	each server has a read and write socket end).
		2309
		2310	I<create> is the time it takes to create a socket pair (which is
		2311	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		2312
		2313	I<request>, the most important value, is the time it takes to handle a
		2314	single "request", that is, reading the token from the pipe and forwarding
		2315	it to another server. This includes deleting the old timeout and creating
		2316	a new one that moves the timeout into the future.
		2317
		2318	=head3 Results
		2319
		2320	name sockets create request
		2321	EV 20000 62.66 7.99
		2322	Perl 20000 68.32 32.64
		2323	IOAsync 20000 174.06 101.15 epoll
		2324	IOAsync 20000 174.67 610.84 poll
		2325	Event 20000 202.69 242.91
		2326	Glib 20000 557.01 1689.52
		2327	POE 20000 341.54 12086.32 uses POE::Loop::Event
		2328
		2329	=head3 Discussion
		2330
		2331	This benchmark I<does> measure scalability and overall performance of the
		2332	particular event loop.
		2333
		2334	EV is again fastest. Since it is using epoll on my system, the setup time
		2335	is relatively high, though.
		2336
		2337	Perl surprisingly comes second. It is much faster than the C-based event
		2338	loops Event and Glib.
		2339
		2340	IO::Async performs very well when using its epoll backend, and still quite
		2341	good compared to Glib when using its pure perl backend.
		2342
		2343	Event suffers from high setup time as well (look at its code and you will
		2344	understand why). Callback invocation also has a high overhead compared to
		2345	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		2346	uses select or poll in basically all documented configurations.
		2347
		2348	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		2349	clearly fails to perform with many filehandles or in busy servers.
		2350
		2351	POE is still completely out of the picture, taking over 1000 times as long
		2352	as EV, and over 100 times as long as the Perl implementation, even though
		2353	it uses a C-based event loop in this case.
		2354
		2355	=head3 Summary
		2356
		2357	=over 4
		2358
		2359	=item * The pure perl implementation performs extremely well.
		2360
		2361	=item * Avoid Glib or POE in large projects where performance matters.
		2362
		2363	=back
		2364
		2365	=head2 BENCHMARKING SMALL SERVERS
		2366
		2367	While event loops should scale (and select-based ones do not...) even to
		2368	large servers, most programs we (or I :) actually write have only a few
		2369	I/O watchers.
		2370
		2371	In this benchmark, I use the same benchmark program as in the large server
		2372	case, but it uses only eight "servers", of which three are active at any
		2373	one time. This should reflect performance for a small server relatively
		2374	well.
		2375
		2376	The columns are identical to the previous table.
		2377
		2378	=head3 Results
		2379
		2380	name sockets create request
		2381	EV 16 20.00 6.54
		2382	Perl 16 25.75 12.62
		2383	Event 16 81.27 35.86
		2384	Glib 16 32.63 15.48
		2385	POE 16 261.87 276.28 uses POE::Loop::Event
		2386
		2387	=head3 Discussion
		2388
		2389	The benchmark tries to test the performance of a typical small
		2390	server. While knowing how various event loops perform is interesting, keep
		2391	in mind that their overhead in this case is usually not as important, due
		2392	to the small absolute number of watchers (that is, you need efficiency and
		2393	speed most when you have lots of watchers, not when you only have a few of
		2394	them).
		2395
		2396	EV is again fastest.
		2397
		2398	Perl again comes second. It is noticeably faster than the C-based event
		2399	loops Event and Glib, although the difference is too small to really
		2400	matter.
		2401
		2402	POE also performs much better in this case, but is is still far behind the
		2403	others.
		2404
		2405	=head3 Summary
		2406
		2407	=over 4
		2408
		2409	=item * C-based event loops perform very well with small number of
		2410	watchers, as the management overhead dominates.
		2411
		2412	=back
		2413
		2414	=head2 THE IO::Lambda BENCHMARK
		2415
		2416	Recently I was told about the benchmark in the IO::Lambda manpage, which
		2417	could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
		2418	simply compares IO::Lambda with POE, and IO::Lambda looks better (which
		2419	shouldn't come as a surprise to anybody). As such, the benchmark is
		2420	fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
		2421	very optimal. But how would AnyEvent compare when used without the extra
		2422	baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
		2423
		2424	The benchmark itself creates an echo-server, and then, for 500 times,
		2425	connects to the echo server, sends a line, waits for the reply, and then
		2426	creates the next connection. This is a rather bad benchmark, as it doesn't
		2427	test the efficiency of the framework or much non-blocking I/O, but it is a
		2428	benchmark nevertheless.
		2429
		2430	name runtime
		2431	Lambda/select 0.330 sec
		2432	+ optimized 0.122 sec
		2433	Lambda/AnyEvent 0.327 sec
		2434	+ optimized 0.138 sec
		2435	Raw sockets/select 0.077 sec
		2436	POE/select, components 0.662 sec
		2437	POE/select, raw sockets 0.226 sec
		2438	POE/select, optimized 0.404 sec
		2439
		2440	AnyEvent/select/nb 0.085 sec
		2441	AnyEvent/EV/nb 0.068 sec
		2442	+state machine 0.134 sec
		2443
		2444	The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
		2445	benchmarks actually make blocking connects and use 100% blocking I/O,
		2446	defeating the purpose of an event-based solution. All of the newly
		2447	written AnyEvent benchmarks use 100% non-blocking connects (using
		2448	AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
		2449	resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
		2450	generally require a lot more bookkeeping and event handling than blocking
		2451	connects (which involve a single syscall only).
		2452
		2453	The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
		2454	offers similar expressive power as POE and IO::Lambda, using conventional
		2455	Perl syntax. This means that both the echo server and the client are 100%
		2456	non-blocking, further placing it at a disadvantage.
		2457
		2458	As you can see, the AnyEvent + EV combination even beats the
		2459	hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
		2460	backend easily beats IO::Lambda and POE.
		2461
		2462	And even the 100% non-blocking version written using the high-level (and
		2463	slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda
		2464	higher level ("unoptimised") abstractions by a large margin, even though
		2465	it does all of DNS, tcp-connect and socket I/O in a non-blocking way.
		2466
		2467	The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
		2468	F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
		2469	part of the IO::Lambda distribution and were used without any changes.
		2470
		2471
		2472	=head1 SIGNALS
		2473
		2474	AnyEvent currently installs handlers for these signals:
		2475
		2476	=over 4
		2477
		2478	=item SIGCHLD
		2479
		2480	A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
		2481	emulation for event loops that do not support them natively. Also, some
		2482	event loops install a similar handler.
		2483
		2484	Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE, then
		2485	AnyEvent will reset it to default, to avoid losing child exit statuses.
		2486
		2487	=item SIGPIPE
		2488
		2489	A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
		2490	when AnyEvent gets loaded.
		2491
		2492	The rationale for this is that AnyEvent users usually do not really depend
		2493	on SIGPIPE delivery (which is purely an optimisation for shell use, or
		2494	badly-written programs), but C<SIGPIPE> can cause spurious and rare
		2495	program exits as a lot of people do not expect C<SIGPIPE> when writing to
		2496	some random socket.
		2497
		2498	The rationale for installing a no-op handler as opposed to ignoring it is
		2499	that this way, the handler will be restored to defaults on exec.
		2500
		2501	Feel free to install your own handler, or reset it to defaults.
		2502
		2503	=back
		2504
49	=cut	2505	=cut
50		2506
51	package AnyEvent;	2507	undef $SIG{CHLD}
		2508	if $SIG{CHLD} eq 'IGNORE';
52		2509
53	no warnings;	2510	$SIG{PIPE} = sub { }
54	use strict 'vars';	2511	unless defined $SIG{PIPE};
55	use Carp;
56		2512
57	our $VERSION = 0.2;	2513	=head1 RECOMMENDED/OPTIONAL MODULES
58	our $MODEL;
59		2514
60	our $AUTOLOAD;	2515	One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
61	our @ISA;	2516	its built-in modules) are required to use it.
62		2517
63	my @models = (	2518	That does not mean that AnyEvent won't take advantage of some additional
64	[Coro => Coro::Event::],	2519	modules if they are installed.
65	[Event => Event::],
66	[Glib => Glib::],
67	[Tk => Tk::],
68	);
69		2520
70	our %method = map +($_ => 1), qw(io timer condvar broadcast wait cancel DESTROY);	2521	This section explains which additional modules will be used, and how they
		2522	affect AnyEvent's operation.
71		2523
72	sub AUTOLOAD {	2524	=over 4
73	$AUTOLOAD =~ s/.*://;
74		2525
75	$method{$AUTOLOAD}	2526	=item L<Async::Interrupt>
76	or croak "$AUTOLOAD: not a valid method for AnyEvent objects";
77		2527
78	unless ($MODEL) {	2528	This slightly arcane module is used to implement fast signal handling: To
79	# check for already loaded models	2529	my knowledge, there is no way to do completely race-free and quick
80	for (@models) {	2530	signal handling in pure perl. To ensure that signals still get
81	my ($model, $package) = @$_;	2531	delivered, AnyEvent will start an interval timer to wake up perl (and
82	if (scalar keys %{ *{"$package\::"} }) {	2532	catch the signals) with some delay (default is 10 seconds, look for
83	eval "require AnyEvent::Impl::$model";	2533	C<$AnyEvent::MAX_SIGNAL_LATENCY>).
84	last if $MODEL;
85	}
86	}
87		2534
88	unless ($MODEL) {	2535	If this module is available, then it will be used to implement signal
89	# try to load a model	2536	catching, which means that signals will not be delayed, and the event loop
		2537	will not be interrupted regularly, which is more efficient (and good for
		2538	battery life on laptops).
90		2539
91	for (@models) {	2540	This affects not just the pure-perl event loop, but also other event loops
92	my ($model, $package) = @$_;	2541	that have no signal handling on their own (e.g. Glib, Tk, Qt).
93	eval "require AnyEvent::Impl::$model";
94	last if $MODEL;
95	}
96		2542
97	$MODEL	2543	Some event loops (POE, Event, Event::Lib) offer signal watchers natively,
98	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: Coro, Event, Glib or Tk.";	2544	and either employ their own workarounds (POE) or use AnyEvent's workaround
99	}	2545	(using C<$AnyEvent::MAX_SIGNAL_LATENCY>). Installing L<Async::Interrupt>
100	}	2546	does nothing for those backends.
101		2547
102	@ISA = $MODEL;	2548	=item L<EV>
103		2549
104	my $class = shift;	2550	This module isn't really "optional", as it is simply one of the backend
105	$class->$AUTOLOAD (@_);	2551	event loops that AnyEvent can use. However, it is simply the best event
106	}	2552	loop available in terms of features, speed and stability: It supports
		2553	the AnyEvent API optimally, implements all the watcher types in XS, does
		2554	automatic timer adjustments even when no monotonic clock is available,
		2555	can take avdantage of advanced kernel interfaces such as C<epoll> and
		2556	C<kqueue>, and is the fastest backend I<by far>. You can even embed
		2557	L<Glib>/L<Gtk2> in it (or vice versa, see L<EV::Glib> and L<Glib::EV>).
		2558
		2559	If you only use backends that rely on another event loop (e.g. C<Tk>),
		2560	then this module will do nothing for you.
		2561
		2562	=item L<Guard>
		2563
		2564	The guard module, when used, will be used to implement
		2565	C<AnyEvent::Util::guard>. This speeds up guards considerably (and uses a
		2566	lot less memory), but otherwise doesn't affect guard operation much. It is
		2567	purely used for performance.
		2568
		2569	=item L<JSON> and L<JSON::XS>
		2570
		2571	One of these modules is required when you want to read or write JSON data
		2572	via L<AnyEvent::Handle>. L<JSON> is also written in pure-perl, but can take
		2573	advantage of the ultra-high-speed L<JSON::XS> module when it is installed.
		2574
		2575	=item L<Net::SSLeay>
		2576
		2577	Implementing TLS/SSL in Perl is certainly interesting, but not very
		2578	worthwhile: If this module is installed, then L<AnyEvent::Handle> (with
		2579	the help of L<AnyEvent::TLS>), gains the ability to do TLS/SSL.
		2580
		2581	=item L<Time::HiRes>
		2582
		2583	This module is part of perl since release 5.008. It will be used when the
		2584	chosen event library does not come with a timing source of its own. The
		2585	pure-perl event loop (L<AnyEvent::Impl::Perl>) will additionally use it to
		2586	try to use a monotonic clock for timing stability.
107		2587
108	=back	2588	=back
109		2589
110	=head1 EXAMPLE
111		2590
112	The following program uses an io watcher to read data from stdin, a timer	2591	=head1 FORK
113	to display a message once per second, and a condvar to exit the program
114	when the user enters quit:
115		2592
		2593	Most event libraries are not fork-safe. The ones who are usually are
		2594	because they rely on inefficient but fork-safe C<select> or C<poll> calls
		2595	- higher performance APIs such as BSD's kqueue or the dreaded Linux epoll
		2596	are usually badly thought-out hacks that are incompatible with fork in
		2597	one way or another. Only L<EV> is fully fork-aware and ensures that you
		2598	continue event-processing in both parent and child (or both, if you know
		2599	what you are doing).
		2600
		2601	This means that, in general, you cannot fork and do event processing in
		2602	the child if the event library was initialised before the fork (which
		2603	usually happens when the first AnyEvent watcher is created, or the library
		2604	is loaded).
		2605
		2606	If you have to fork, you must either do so I<before> creating your first
		2607	watcher OR you must not use AnyEvent at all in the child OR you must do
		2608	something completely out of the scope of AnyEvent.
		2609
		2610	The problem of doing event processing in the parent I<and> the child
		2611	is much more complicated: even for backends that I<are> fork-aware or
		2612	fork-safe, their behaviour is not usually what you want: fork clones all
		2613	watchers, that means all timers, I/O watchers etc. are active in both
		2614	parent and child, which is almost never what you want. USing C<exec>
		2615	to start worker children from some kind of manage rprocess is usually
		2616	preferred, because it is much easier and cleaner, at the expense of having
		2617	to have another binary.
		2618
		2619
		2620	=head1 SECURITY CONSIDERATIONS
		2621
		2622	AnyEvent can be forced to load any event model via
		2623	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		2624	execute arbitrary code or directly gain access, it can easily be used to
		2625	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		2626	will not be active when the program uses a different event model than
		2627	specified in the variable.
		2628
		2629	You can make AnyEvent completely ignore this variable by deleting it
		2630	before the first watcher gets created, e.g. with a C<BEGIN> block:
		2631
		2632	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		2633
116	use AnyEvent;	2634	use AnyEvent;
117		2635
118	my $cv = AnyEvent->condvar;	2636	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		2637	be used to probe what backend is used and gain other information (which is
		2638	probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
		2639	$ENV{PERL_ANYEVENT_STRICT}.
119		2640
120	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	2641	Note that AnyEvent will remove I<all> environment variables starting with
121	warn "io event <$_[0]>\n"; # will always output <r>	2642	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
122	chomp (my $input = <STDIN>); # read a line	2643	enabled.
123	warn "read: $input\n"; # output what has been read
124	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i
125	});
126		2644
127	my $time_watcher; # can only be used once
128		2645
129	sub new_timer {	2646	=head1 BUGS
130	$timer = AnyEvent->timer (after => 1, cb => sub {
131	warn "timeout\n"; # print 'timeout' about every second
132	&new_timer; # and restart the time
133	});
134	}
135		2647
136	new_timer; # create first timer	2648	Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
		2649	to work around. If you suffer from memleaks, first upgrade to Perl 5.10
		2650	and check wether the leaks still show up. (Perl 5.10.0 has other annoying
		2651	memleaks, such as leaking on C<map> and C<grep> but it is usually not as
		2652	pronounced).
137		2653
138	$cv->wait; # wait until user enters /^q/i
139		2654
140	=head1 SEE ALSO	2655	=head1 SEE ALSO
141		2656
142	L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>,	2657	Utility functions: L<AnyEvent::Util>.
143	L<AnyEvent::Impl::Coro>,	2658
144	L<AnyEvent::Impl::Event>,	2659	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
145	L<AnyEvent::Impl::Glib>,	2660	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
		2661
		2662	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		2663	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		2664	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		2665	L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>, L<Anyevent::Impl::Irssi>.
		2666
		2667	Non-blocking file handles, sockets, TCP clients and
		2668	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>.
		2669
		2670	Asynchronous DNS: L<AnyEvent::DNS>.
		2671
		2672	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>,
		2673	L<Coro::Event>,
		2674
		2675	Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::XMPP>,
146	L<AnyEvent::Impl::Tk>.	2676	L<AnyEvent::HTTP>.
147		2677
148	=head1	2678
		2679	=head1 AUTHOR
		2680
		2681	Marc Lehmann <schmorp@schmorp.de>
		2682	http://home.schmorp.de/
149		2683
150	=cut	2684	=cut
151		2685
152	1	2686	1
153		2687

Diff Legend

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