ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.2 by root, Thu Dec 1 18:56:18 2005 UTC vs.
Revision 1.309 by root, Sat Dec 26 08:59:35 2009 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	# file descriptor readable
11	my $w = AnyEvent->timer (fh => ..., poll => "[rw]+", cb => sub {	13	my $w = AnyEvent->io (fh => $fh, poll => "r", cb => sub { ... });
12	my ($poll_got) = @_;	14
		15	# one-shot or repeating timers
		16	my $w = AnyEvent->timer (after => $seconds, cb => sub { ... });
		17	my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ...
		18
		19	print AnyEvent->now; # prints current event loop time
		20	print AnyEvent->time; # think Time::HiRes::time or simply CORE::time.
		21
		22	# POSIX signal
		23	my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... });
		24
		25	# child process exit
		26	my $w = AnyEvent->child (pid => $pid, cb => sub {
		27	my ($pid, $status) = @_;
13	...	28	...
14	});	29	});
		30
		31	# called when event loop idle (if applicable)
		32	my $w = AnyEvent->idle (cb => sub { ... });
		33
		34	my $w = AnyEvent->condvar; # stores whether a condition was flagged
		35	$w->send; # wake up current and all future recv's
		36	$w->recv; # enters "main loop" till $condvar gets ->send
		37	# use a condvar in callback mode:
		38	$w->cb (sub { $_[0]->recv });
		39
		40	=head1 INTRODUCTION/TUTORIAL
		41
		42	This manpage is mainly a reference manual. If you are interested
		43	in a tutorial or some gentle introduction, have a look at the
		44	L<AnyEvent::Intro> manpage.
		45
		46	=head1 SUPPORT
		47
		48	There is a mailinglist for discussing all things AnyEvent, and an IRC
		49	channel, too.
		50
		51	See the AnyEvent project page at the B<Schmorpforge Ta-Sa Software
		52	Repository>, at L<http://anyevent.schmorp.de>, for more info.
		53
		54	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
		55
		56	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
		57	nowadays. So what is different about AnyEvent?
		58
		59	Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
		60	policy> and AnyEvent is I<small and efficient>.
		61
		62	First and foremost, I<AnyEvent is not an event model> itself, it only
		63	interfaces to whatever event model the main program happens to use, in a
		64	pragmatic way. For event models and certain classes of immortals alike,
		65	the statement "there can only be one" is a bitter reality: In general,
		66	only one event loop can be active at the same time in a process. AnyEvent
		67	cannot change this, but it can hide the differences between those event
		68	loops.
		69
		70	The goal of AnyEvent is to offer module authors the ability to do event
		71	programming (waiting for I/O or timer events) without subscribing to a
		72	religion, a way of living, and most importantly: without forcing your
		73	module users into the same thing by forcing them to use the same event
		74	model you use.
		75
		76	For modules like POE or IO::Async (which is a total misnomer as it is
		77	actually doing all I/O I<synchronously>...), using them in your module is
		78	like joining a cult: After you joined, you are dependent on them and you
		79	cannot use anything else, as they are simply incompatible to everything
		80	that isn't them. What's worse, all the potential users of your
		81	module are I<also> forced to use the same event loop you use.
		82
		83	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
		84	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
		85	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
		86	your module uses one of those, every user of your module has to use it,
		87	too. But if your module uses AnyEvent, it works transparently with all
		88	event models it supports (including stuff like IO::Async, as long as those
		89	use one of the supported event loops. It is trivial to add new event loops
		90	to AnyEvent, too, so it is future-proof).
		91
		92	In addition to being free of having to use I<the one and only true event
		93	model>, AnyEvent also is free of bloat and policy: with POE or similar
		94	modules, you get an enormous amount of code and strict rules you have to
		95	follow. AnyEvent, on the other hand, is lean and up to the point, by only
		96	offering the functionality that is necessary, in as thin as a wrapper as
		97	technically possible.
		98
		99	Of course, AnyEvent comes with a big (and fully optional!) toolbox
		100	of useful functionality, such as an asynchronous DNS resolver, 100%
		101	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		102	such as Windows) and lots of real-world knowledge and workarounds for
		103	platform bugs and differences.
		104
		105	Now, if you I<do want> lots of policy (this can arguably be somewhat
		106	useful) and you want to force your users to use the one and only event
		107	model, you should I<not> use this module.
		108
		109	=head1 DESCRIPTION
		110
		111	L<AnyEvent> provides an identical interface to multiple event loops. This
		112	allows module authors to utilise an event loop without forcing module
		113	users to use the same event loop (as only a single event loop can coexist
		114	peacefully at any one time).
		115
		116	The interface itself is vaguely similar, but not identical to the L<Event>
		117	module.
		118
		119	During the first call of any watcher-creation method, the module tries
		120	to detect the currently loaded event loop by probing whether one of the
		121	following modules is already loaded: L<EV>,
		122	L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>,
		123	L<POE>. The first one found is used. If none are found, the module tries
		124	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
		125	adaptor should always succeed) in the order given. The first one that can
		126	be successfully loaded will be used. If, after this, still none could be
		127	found, AnyEvent will fall back to a pure-perl event loop, which is not
		128	very efficient, but should work everywhere.
		129
		130	Because AnyEvent first checks for modules that are already loaded, loading
		131	an event model explicitly before first using AnyEvent will likely make
		132	that model the default. For example:
		133
		134	use Tk;
		135	use AnyEvent;
		136
		137	# .. AnyEvent will likely default to Tk
		138
		139	The I<likely> means that, if any module loads another event model and
		140	starts using it, all bets are off. Maybe you should tell their authors to
		141	use AnyEvent so their modules work together with others seamlessly...
		142
		143	The pure-perl implementation of AnyEvent is called
		144	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
		145	explicitly and enjoy the high availability of that event loop :)
		146
		147	=head1 WATCHERS
		148
		149	AnyEvent has the central concept of a I<watcher>, which is an object that
		150	stores relevant data for each kind of event you are waiting for, such as
		151	the callback to call, the file handle to watch, etc.
		152
		153	These watchers are normal Perl objects with normal Perl lifetime. After
		154	creating a watcher it will immediately "watch" for events and invoke the
		155	callback when the event occurs (of course, only when the event model
		156	is in control).
		157
		158	Note that B<callbacks must not permanently change global variables>
		159	potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
		160	callbacks must not C<die> >>. The former is good programming practise in
		161	Perl and the latter stems from the fact that exception handling differs
		162	widely between event loops.
		163
		164	To disable the watcher you have to destroy it (e.g. by setting the
		165	variable you store it in to C<undef> or otherwise deleting all references
		166	to it).
		167
		168	All watchers are created by calling a method on the C<AnyEvent> class.
		169
		170	Many watchers either are used with "recursion" (repeating timers for
		171	example), or need to refer to their watcher object in other ways.
		172
		173	An any way to achieve that is this pattern:
		174
		175	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
		176	# you can use $w here, for example to undef it
		177	undef $w;
		178	});
		179
		180	Note that C<my $w; $w => combination. This is necessary because in Perl,
		181	my variables are only visible after the statement in which they are
		182	declared.
		183
		184	=head2 I/O WATCHERS
		185
		186	$w = AnyEvent->io (
		187	fh => <filehandle_or_fileno>,
		188	poll => <"r" or "w">,
		189	cb => <callback>,
		190	);
		191
		192	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
		193	with the following mandatory key-value pairs as arguments:
		194
		195	C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch
		196	for events (AnyEvent might or might not keep a reference to this file
		197	handle). Note that only file handles pointing to things for which
		198	non-blocking operation makes sense are allowed. This includes sockets,
		199	most character devices, pipes, fifos and so on, but not for example files
		200	or block devices.
		201
		202	C<poll> must be a string that is either C<r> or C<w>, which creates a
		203	watcher waiting for "r"eadable or "w"ritable events, respectively.
		204
		205	C<cb> is the callback to invoke each time the file handle becomes ready.
		206
		207	Although the callback might get passed parameters, their value and
		208	presence is undefined and you cannot rely on them. Portable AnyEvent
		209	callbacks cannot use arguments passed to I/O watcher callbacks.
		210
		211	The I/O watcher might use the underlying file descriptor or a copy of it.
		212	You must not close a file handle as long as any watcher is active on the
		213	underlying file descriptor.
		214
		215	Some event loops issue spurious readyness notifications, so you should
		216	always use non-blocking calls when reading/writing from/to your file
		217	handles.
		218
		219	Example: wait for readability of STDIN, then read a line and disable the
		220	watcher.
		221
		222	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		223	chomp (my $input = <STDIN>);
		224	warn "read: $input\n";
		225	undef $w;
		226	});
		227
		228	=head2 TIME WATCHERS
		229
		230	$w = AnyEvent->timer (after => <seconds>, cb => <callback>);
		231
		232	$w = AnyEvent->timer (
		233	after => <fractional_seconds>,
		234	interval => <fractional_seconds>,
		235	cb => <callback>,
		236	);
		237
		238	You can create a time watcher by calling the C<< AnyEvent->timer >>
		239	method with the following mandatory arguments:
		240
		241	C<after> specifies after how many seconds (fractional values are
		242	supported) the callback should be invoked. C<cb> is the callback to invoke
		243	in that case.
		244
		245	Although the callback might get passed parameters, their value and
		246	presence is undefined and you cannot rely on them. Portable AnyEvent
		247	callbacks cannot use arguments passed to time watcher callbacks.
		248
		249	The callback will normally be invoked once only. If you specify another
		250	parameter, C<interval>, as a strictly positive number (> 0), then the
		251	callback will be invoked regularly at that interval (in fractional
		252	seconds) after the first invocation. If C<interval> is specified with a
		253	false value, then it is treated as if it were missing.
		254
		255	The callback will be rescheduled before invoking the callback, but no
		256	attempt is done to avoid timer drift in most backends, so the interval is
		257	only approximate.
		258
		259	Example: fire an event after 7.7 seconds.
		260
15	my $w = AnyEvent->io (after => $seconds, cb => sub {	261	my $w = AnyEvent->timer (after => 7.7, cb => sub {
		262	warn "timeout\n";
		263	});
		264
		265	# to cancel the timer:
		266	undef $w;
		267
		268	Example 2: fire an event after 0.5 seconds, then roughly every second.
		269
		270	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
		271	warn "timeout\n";
		272	};
		273
		274	=head3 TIMING ISSUES
		275
		276	There are two ways to handle timers: based on real time (relative, "fire
		277	in 10 seconds") and based on wallclock time (absolute, "fire at 12
		278	o'clock").
		279
		280	While most event loops expect timers to specified in a relative way, they
		281	use absolute time internally. This makes a difference when your clock
		282	"jumps", for example, when ntp decides to set your clock backwards from
		283	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		284	fire "after" a second might actually take six years to finally fire.
		285
		286	AnyEvent cannot compensate for this. The only event loop that is conscious
		287	about these issues is L<EV>, which offers both relative (ev_timer, based
		288	on true relative time) and absolute (ev_periodic, based on wallclock time)
		289	timers.
		290
		291	AnyEvent always prefers relative timers, if available, matching the
		292	AnyEvent API.
		293
		294	AnyEvent has two additional methods that return the "current time":
		295
		296	=over 4
		297
		298	=item AnyEvent->time
		299
		300	This returns the "current wallclock time" as a fractional number of
		301	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		302	return, and the result is guaranteed to be compatible with those).
		303
		304	It progresses independently of any event loop processing, i.e. each call
		305	will check the system clock, which usually gets updated frequently.
		306
		307	=item AnyEvent->now
		308
		309	This also returns the "current wallclock time", but unlike C<time>, above,
		310	this value might change only once per event loop iteration, depending on
		311	the event loop (most return the same time as C<time>, above). This is the
		312	time that AnyEvent's timers get scheduled against.
		313
		314	I<In almost all cases (in all cases if you don't care), this is the
		315	function to call when you want to know the current time.>
		316
		317	This function is also often faster then C<< AnyEvent->time >>, and
		318	thus the preferred method if you want some timestamp (for example,
		319	L<AnyEvent::Handle> uses this to update it's activity timeouts).
		320
		321	The rest of this section is only of relevance if you try to be very exact
		322	with your timing, you can skip it without bad conscience.
		323
		324	For a practical example of when these times differ, consider L<Event::Lib>
		325	and L<EV> and the following set-up:
		326
		327	The event loop is running and has just invoked one of your callback at
		328	time=500 (assume no other callbacks delay processing). In your callback,
		329	you wait a second by executing C<sleep 1> (blocking the process for a
		330	second) and then (at time=501) you create a relative timer that fires
		331	after three seconds.
		332
		333	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		334	both return C<501>, because that is the current time, and the timer will
		335	be scheduled to fire at time=504 (C<501> + C<3>).
		336
		337	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		338	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		339	last event processing phase started. With L<EV>, your timer gets scheduled
		340	to run at time=503 (C<500> + C<3>).
		341
		342	In one sense, L<Event::Lib> is more exact, as it uses the current time
		343	regardless of any delays introduced by event processing. However, most
		344	callbacks do not expect large delays in processing, so this causes a
		345	higher drift (and a lot more system calls to get the current time).
		346
		347	In another sense, L<EV> is more exact, as your timer will be scheduled at
		348	the same time, regardless of how long event processing actually took.
		349
		350	In either case, if you care (and in most cases, you don't), then you
		351	can get whatever behaviour you want with any event loop, by taking the
		352	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		353	account.
		354
		355	=item AnyEvent->now_update
		356
		357	Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
		358	the current time for each loop iteration (see the discussion of L<<
		359	AnyEvent->now >>, above).
		360
		361	When a callback runs for a long time (or when the process sleeps), then
		362	this "current" time will differ substantially from the real time, which
		363	might affect timers and time-outs.
		364
		365	When this is the case, you can call this method, which will update the
		366	event loop's idea of "current time".
		367
		368	A typical example would be a script in a web server (e.g. C<mod_perl>) -
		369	when mod_perl executes the script, then the event loop will have the wrong
		370	idea about the "current time" (being potentially far in the past, when the
		371	script ran the last time). In that case you should arrange a call to C<<
		372	AnyEvent->now_update >> each time the web server process wakes up again
		373	(e.g. at the start of your script, or in a handler).
		374
		375	Note that updating the time I<might> cause some events to be handled.
		376
		377	=back
		378
		379	=head2 SIGNAL WATCHERS
		380
		381	$w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>);
		382
		383	You can watch for signals using a signal watcher, C<signal> is the signal
		384	I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
		385	callback to be invoked whenever a signal occurs.
		386
		387	Although the callback might get passed parameters, their value and
		388	presence is undefined and you cannot rely on them. Portable AnyEvent
		389	callbacks cannot use arguments passed to signal watcher callbacks.
		390
		391	Multiple signal occurrences can be clumped together into one callback
		392	invocation, and callback invocation will be synchronous. Synchronous means
		393	that it might take a while until the signal gets handled by the process,
		394	but it is guaranteed not to interrupt any other callbacks.
		395
		396	The main advantage of using these watchers is that you can share a signal
		397	between multiple watchers, and AnyEvent will ensure that signals will not
		398	interrupt your program at bad times.
		399
		400	This watcher might use C<%SIG> (depending on the event loop used),
		401	so programs overwriting those signals directly will likely not work
		402	correctly.
		403
		404	Example: exit on SIGINT
		405
		406	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		407
		408	=head3 Restart Behaviour
		409
		410	While restart behaviour is up to the event loop implementation, most will
		411	not restart syscalls (that includes L<Async::Interrupt> and AnyEvent's
		412	pure perl implementation).
		413
		414	=head3 Safe/Unsafe Signals
		415
		416	Perl signals can be either "safe" (synchronous to opcode handling) or
		417	"unsafe" (asynchronous) - the former might get delayed indefinitely, the
		418	latter might corrupt your memory.
		419
		420	AnyEvent signal handlers are, in addition, synchronous to the event loop,
		421	i.e. they will not interrupt your running perl program but will only be
		422	called as part of the normal event handling (just like timer, I/O etc.
		423	callbacks, too).
		424
		425	=head3 Signal Races, Delays and Workarounds
		426
		427	Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
		428	callbacks to signals in a generic way, which is a pity, as you cannot
		429	do race-free signal handling in perl, requiring C libraries for
		430	this. AnyEvent will try to do it's best, which means in some cases,
		431	signals will be delayed. The maximum time a signal might be delayed is
		432	specified in C<$AnyEvent::MAX_SIGNAL_LATENCY> (default: 10 seconds). This
		433	variable can be changed only before the first signal watcher is created,
		434	and should be left alone otherwise. This variable determines how often
		435	AnyEvent polls for signals (in case a wake-up was missed). Higher values
		436	will cause fewer spurious wake-ups, which is better for power and CPU
		437	saving.
		438
		439	All these problems can be avoided by installing the optional
		440	L<Async::Interrupt> module, which works with most event loops. It will not
		441	work with inherently broken event loops such as L<Event> or L<Event::Lib>
		442	(and not with L<POE> currently, as POE does it's own workaround with
		443	one-second latency). For those, you just have to suffer the delays.
		444
		445	=head2 CHILD PROCESS WATCHERS
		446
		447	$w = AnyEvent->child (pid => <process id>, cb => <callback>);
		448
		449	You can also watch on a child process exit and catch its exit status.
		450
		451	The child process is specified by the C<pid> argument (one some backends,
		452	using C<0> watches for any child process exit, on others this will
		453	croak). The watcher will be triggered only when the child process has
		454	finished and an exit status is available, not on any trace events
		455	(stopped/continued).
		456
		457	The callback will be called with the pid and exit status (as returned by
		458	waitpid), so unlike other watcher types, you I<can> rely on child watcher
		459	callback arguments.
		460
		461	This watcher type works by installing a signal handler for C<SIGCHLD>,
		462	and since it cannot be shared, nothing else should use SIGCHLD or reap
		463	random child processes (waiting for specific child processes, e.g. inside
		464	C<system>, is just fine).
		465
		466	There is a slight catch to child watchers, however: you usually start them
		467	I<after> the child process was created, and this means the process could
		468	have exited already (and no SIGCHLD will be sent anymore).
		469
		470	Not all event models handle this correctly (neither POE nor IO::Async do,
		471	see their AnyEvent::Impl manpages for details), but even for event models
		472	that I<do> handle this correctly, they usually need to be loaded before
		473	the process exits (i.e. before you fork in the first place). AnyEvent's
		474	pure perl event loop handles all cases correctly regardless of when you
		475	start the watcher.
		476
		477	This means you cannot create a child watcher as the very first
		478	thing in an AnyEvent program, you I<have> to create at least one
		479	watcher before you C<fork> the child (alternatively, you can call
		480	C<AnyEvent::detect>).
		481
		482	As most event loops do not support waiting for child events, they will be
		483	emulated by AnyEvent in most cases, in which the latency and race problems
		484	mentioned in the description of signal watchers apply.
		485
		486	Example: fork a process and wait for it
		487
		488	my $done = AnyEvent->condvar;
		489
		490	my $pid = fork or exit 5;
		491
		492	my $w = AnyEvent->child (
		493	pid => $pid,
		494	cb => sub {
		495	my ($pid, $status) = @_;
		496	warn "pid $pid exited with status $status";
		497	$done->send;
		498	},
		499	);
		500
		501	# do something else, then wait for process exit
		502	$done->recv;
		503
		504	=head2 IDLE WATCHERS
		505
		506	$w = AnyEvent->idle (cb => <callback>);
		507
		508	Repeatedly invoke the callback after the process becomes idle, until
		509	either the watcher is destroyed or new events have been detected.
		510
		511	Idle watchers are useful when there is a need to do something, but it
		512	is not so important (or wise) to do it instantly. The callback will be
		513	invoked only when there is "nothing better to do", which is usually
		514	defined as "all outstanding events have been handled and no new events
		515	have been detected". That means that idle watchers ideally get invoked
		516	when the event loop has just polled for new events but none have been
		517	detected. Instead of blocking to wait for more events, the idle watchers
		518	will be invoked.
		519
		520	Unfortunately, most event loops do not really support idle watchers (only
		521	EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
		522	will simply call the callback "from time to time".
		523
		524	Example: read lines from STDIN, but only process them when the
		525	program is otherwise idle:
		526
		527	my @lines; # read data
		528	my $idle_w;
		529	my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		530	push @lines, scalar <STDIN>;
		531
		532	# start an idle watcher, if not already done
		533	$idle_w ||= AnyEvent->idle (cb => sub {
		534	# handle only one line, when there are lines left
		535	if (my $line = shift @lines) {
		536	print "handled when idle: $line";
		537	} else {
		538	# otherwise disable the idle watcher again
		539	undef $idle_w;
		540	}
		541	});
		542	});
		543
		544	=head2 CONDITION VARIABLES
		545
		546	$cv = AnyEvent->condvar;
		547
		548	$cv->send (<list>);
		549	my @res = $cv->recv;
		550
		551	If you are familiar with some event loops you will know that all of them
		552	require you to run some blocking "loop", "run" or similar function that
		553	will actively watch for new events and call your callbacks.
		554
		555	AnyEvent is slightly different: it expects somebody else to run the event
		556	loop and will only block when necessary (usually when told by the user).
		557
		558	The instrument to do that is called a "condition variable", so called
		559	because they represent a condition that must become true.
		560
		561	Now is probably a good time to look at the examples further below.
		562
		563	Condition variables can be created by calling the C<< AnyEvent->condvar
		564	>> method, usually without arguments. The only argument pair allowed is
		565	C<cb>, which specifies a callback to be called when the condition variable
		566	becomes true, with the condition variable as the first argument (but not
		567	the results).
		568
		569	After creation, the condition variable is "false" until it becomes "true"
		570	by calling the C<send> method (or calling the condition variable as if it
		571	were a callback, read about the caveats in the description for the C<<
		572	->send >> method).
		573
		574	Condition variables are similar to callbacks, except that you can
		575	optionally wait for them. They can also be called merge points - points
		576	in time where multiple outstanding events have been processed. And yet
		577	another way to call them is transactions - each condition variable can be
		578	used to represent a transaction, which finishes at some point and delivers
		579	a result. And yet some people know them as "futures" - a promise to
		580	compute/deliver something that you can wait for.
		581
		582	Condition variables are very useful to signal that something has finished,
		583	for example, if you write a module that does asynchronous http requests,
		584	then a condition variable would be the ideal candidate to signal the
		585	availability of results. The user can either act when the callback is
		586	called or can synchronously C<< ->recv >> for the results.
		587
		588	You can also use them to simulate traditional event loops - for example,
		589	you can block your main program until an event occurs - for example, you
		590	could C<< ->recv >> in your main program until the user clicks the Quit
		591	button of your app, which would C<< ->send >> the "quit" event.
		592
		593	Note that condition variables recurse into the event loop - if you have
		594	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		595	lose. Therefore, condition variables are good to export to your caller, but
		596	you should avoid making a blocking wait yourself, at least in callbacks,
		597	as this asks for trouble.
		598
		599	Condition variables are represented by hash refs in perl, and the keys
		600	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		601	easy (it is often useful to build your own transaction class on top of
		602	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		603	it's C<new> method in your own C<new> method.
		604
		605	There are two "sides" to a condition variable - the "producer side" which
		606	eventually calls C<< -> send >>, and the "consumer side", which waits
		607	for the send to occur.
		608
		609	Example: wait for a timer.
		610
		611	# wait till the result is ready
		612	my $result_ready = AnyEvent->condvar;
		613
		614	# do something such as adding a timer
		615	# or socket watcher the calls $result_ready->send
		616	# when the "result" is ready.
		617	# in this case, we simply use a timer:
		618	my $w = AnyEvent->timer (
		619	after => 1,
		620	cb => sub { $result_ready->send },
		621	);
		622
		623	# this "blocks" (while handling events) till the callback
		624	# calls ->send
		625	$result_ready->recv;
		626
		627	Example: wait for a timer, but take advantage of the fact that condition
		628	variables are also callable directly.
		629
		630	my $done = AnyEvent->condvar;
		631	my $delay = AnyEvent->timer (after => 5, cb => $done);
		632	$done->recv;
		633
		634	Example: Imagine an API that returns a condvar and doesn't support
		635	callbacks. This is how you make a synchronous call, for example from
		636	the main program:
		637
		638	use AnyEvent::CouchDB;
		639
		640	...
		641
		642	my @info = $couchdb->info->recv;
		643
		644	And this is how you would just set a callback to be called whenever the
		645	results are available:
		646
		647	$couchdb->info->cb (sub {
		648	my @info = $_[0]->recv;
		649	});
		650
		651	=head3 METHODS FOR PRODUCERS
		652
		653	These methods should only be used by the producing side, i.e. the
		654	code/module that eventually sends the signal. Note that it is also
		655	the producer side which creates the condvar in most cases, but it isn't
		656	uncommon for the consumer to create it as well.
		657
		658	=over 4
		659
		660	=item $cv->send (...)
		661
		662	Flag the condition as ready - a running C<< ->recv >> and all further
		663	calls to C<recv> will (eventually) return after this method has been
		664	called. If nobody is waiting the send will be remembered.
		665
		666	If a callback has been set on the condition variable, it is called
		667	immediately from within send.
		668
		669	Any arguments passed to the C<send> call will be returned by all
		670	future C<< ->recv >> calls.
		671
		672	Condition variables are overloaded so one can call them directly (as if
		673	they were a code reference). Calling them directly is the same as calling
		674	C<send>.
		675
		676	=item $cv->croak ($error)
		677
		678	Similar to send, but causes all call's to C<< ->recv >> to invoke
		679	C<Carp::croak> with the given error message/object/scalar.
		680
		681	This can be used to signal any errors to the condition variable
		682	user/consumer. Doing it this way instead of calling C<croak> directly
		683	delays the error detetcion, but has the overwhelmign advantage that it
		684	diagnoses the error at the place where the result is expected, and not
		685	deep in some event clalback without connection to the actual code causing
		686	the problem.
		687
		688	=item $cv->begin ([group callback])
		689
		690	=item $cv->end
		691
		692	These two methods can be used to combine many transactions/events into
		693	one. For example, a function that pings many hosts in parallel might want
		694	to use a condition variable for the whole process.
		695
		696	Every call to C<< ->begin >> will increment a counter, and every call to
		697	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		698	>>, the (last) callback passed to C<begin> will be executed, passing the
		699	condvar as first argument. That callback is I<supposed> to call C<< ->send
		700	>>, but that is not required. If no group callback was set, C<send> will
		701	be called without any arguments.
		702
		703	You can think of C<< $cv->send >> giving you an OR condition (one call
		704	sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
		705	condition (all C<begin> calls must be C<end>'ed before the condvar sends).
		706
		707	Let's start with a simple example: you have two I/O watchers (for example,
		708	STDOUT and STDERR for a program), and you want to wait for both streams to
		709	close before activating a condvar:
		710
		711	my $cv = AnyEvent->condvar;
		712
		713	$cv->begin; # first watcher
		714	my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
		715	defined sysread $fh1, my $buf, 4096
		716	or $cv->end;
		717	});
		718
		719	$cv->begin; # second watcher
		720	my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
		721	defined sysread $fh2, my $buf, 4096
		722	or $cv->end;
		723	});
		724
		725	$cv->recv;
		726
		727	This works because for every event source (EOF on file handle), there is
		728	one call to C<begin>, so the condvar waits for all calls to C<end> before
		729	sending.
		730
		731	The ping example mentioned above is slightly more complicated, as the
		732	there are results to be passwd back, and the number of tasks that are
		733	begung can potentially be zero:
		734
		735	my $cv = AnyEvent->condvar;
		736
		737	my %result;
		738	$cv->begin (sub { shift->send (\%result) });
		739
		740	for my $host (@list_of_hosts) {
		741	$cv->begin;
		742	ping_host_then_call_callback $host, sub {
		743	$result{$host} = ...;
		744	$cv->end;
		745	};
		746	}
		747
		748	$cv->end;
		749
		750	This code fragment supposedly pings a number of hosts and calls
		751	C<send> after results for all then have have been gathered - in any
		752	order. To achieve this, the code issues a call to C<begin> when it starts
		753	each ping request and calls C<end> when it has received some result for
		754	it. Since C<begin> and C<end> only maintain a counter, the order in which
		755	results arrive is not relevant.
		756
		757	There is an additional bracketing call to C<begin> and C<end> outside the
		758	loop, which serves two important purposes: first, it sets the callback
		759	to be called once the counter reaches C<0>, and second, it ensures that
		760	C<send> is called even when C<no> hosts are being pinged (the loop
		761	doesn't execute once).
		762
		763	This is the general pattern when you "fan out" into multiple (but
		764	potentially none) subrequests: use an outer C<begin>/C<end> pair to set
		765	the callback and ensure C<end> is called at least once, and then, for each
		766	subrequest you start, call C<begin> and for each subrequest you finish,
		767	call C<end>.
		768
		769	=back
		770
		771	=head3 METHODS FOR CONSUMERS
		772
		773	These methods should only be used by the consuming side, i.e. the
		774	code awaits the condition.
		775
		776	=over 4
		777
		778	=item $cv->recv
		779
		780	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		781	>> methods have been called on c<$cv>, while servicing other watchers
		782	normally.
		783
		784	You can only wait once on a condition - additional calls are valid but
		785	will return immediately.
		786
		787	If an error condition has been set by calling C<< ->croak >>, then this
		788	function will call C<croak>.
		789
		790	In list context, all parameters passed to C<send> will be returned,
		791	in scalar context only the first one will be returned.
		792
		793	Note that doing a blocking wait in a callback is not supported by any
		794	event loop, that is, recursive invocation of a blocking C<< ->recv
		795	>> is not allowed, and the C<recv> call will C<croak> if such a
		796	condition is detected. This condition can be slightly loosened by using
		797	L<Coro::AnyEvent>, which allows you to do a blocking C<< ->recv >> from
		798	any thread that doesn't run the event loop itself.
		799
		800	Not all event models support a blocking wait - some die in that case
		801	(programs might want to do that to stay interactive), so I<if you are
		802	using this from a module, never require a blocking wait>. Instead, let the
		803	caller decide whether the call will block or not (for example, by coupling
		804	condition variables with some kind of request results and supporting
		805	callbacks so the caller knows that getting the result will not block,
		806	while still supporting blocking waits if the caller so desires).
		807
		808	You can ensure that C<< -recv >> never blocks by setting a callback and
		809	only calling C<< ->recv >> from within that callback (or at a later
		810	time). This will work even when the event loop does not support blocking
		811	waits otherwise.
		812
		813	=item $bool = $cv->ready
		814
		815	Returns true when the condition is "true", i.e. whether C<send> or
		816	C<croak> have been called.
		817
		818	=item $cb = $cv->cb ($cb->($cv))
		819
		820	This is a mutator function that returns the callback set and optionally
		821	replaces it before doing so.
		822
		823	The callback will be called when the condition becomes (or already was)
		824	"true", i.e. when C<send> or C<croak> are called (or were called), with
		825	the only argument being the condition variable itself. Calling C<recv>
		826	inside the callback or at any later time is guaranteed not to block.
		827
		828	=back
		829
		830	=head1 SUPPORTED EVENT LOOPS/BACKENDS
		831
		832	The available backend classes are (every class has its own manpage):
		833
		834	=over 4
		835
		836	=item Backends that are autoprobed when no other event loop can be found.
		837
		838	EV is the preferred backend when no other event loop seems to be in
		839	use. If EV is not installed, then AnyEvent will fall back to its own
		840	pure-perl implementation, which is available everywhere as it comes with
		841	AnyEvent itself.
		842
		843	AnyEvent::Impl::EV based on EV (interface to libev, best choice).
		844	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		845
		846	=item Backends that are transparently being picked up when they are used.
		847
		848	These will be used when they are currently loaded when the first watcher
		849	is created, in which case it is assumed that the application is using
		850	them. This means that AnyEvent will automatically pick the right backend
		851	when the main program loads an event module before anything starts to
		852	create watchers. Nothing special needs to be done by the main program.
		853
		854	AnyEvent::Impl::Event based on Event, very stable, few glitches.
		855	AnyEvent::Impl::Glib based on Glib, slow but very stable.
		856	AnyEvent::Impl::Tk based on Tk, very broken.
		857	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		858	AnyEvent::Impl::POE based on POE, very slow, some limitations.
		859	AnyEvent::Impl::Irssi used when running within irssi.
		860
		861	=item Backends with special needs.
		862
		863	Qt requires the Qt::Application to be instantiated first, but will
		864	otherwise be picked up automatically. As long as the main program
		865	instantiates the application before any AnyEvent watchers are created,
		866	everything should just work.
		867
		868	AnyEvent::Impl::Qt based on Qt.
		869
		870	Support for IO::Async can only be partial, as it is too broken and
		871	architecturally limited to even support the AnyEvent API. It also
		872	is the only event loop that needs the loop to be set explicitly, so
		873	it can only be used by a main program knowing about AnyEvent. See
		874	L<AnyEvent::Impl::Async> for the gory details.
		875
		876	AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed.
		877
		878	=item Event loops that are indirectly supported via other backends.
		879
		880	Some event loops can be supported via other modules:
		881
		882	There is no direct support for WxWidgets (L<Wx>) or L<Prima>.
		883
		884	B<WxWidgets> has no support for watching file handles. However, you can
		885	use WxWidgets through the POE adaptor, as POE has a Wx backend that simply
		886	polls 20 times per second, which was considered to be too horrible to even
		887	consider for AnyEvent.
		888
		889	B<Prima> is not supported as nobody seems to be using it, but it has a POE
		890	backend, so it can be supported through POE.
		891
		892	AnyEvent knows about both L<Prima> and L<Wx>, however, and will try to
		893	load L<POE> when detecting them, in the hope that POE will pick them up,
		894	in which case everything will be automatic.
		895
		896	=back
		897
		898	=head1 GLOBAL VARIABLES AND FUNCTIONS
		899
		900	These are not normally required to use AnyEvent, but can be useful to
		901	write AnyEvent extension modules.
		902
		903	=over 4
		904
		905	=item $AnyEvent::MODEL
		906
		907	Contains C<undef> until the first watcher is being created, before the
		908	backend has been autodetected.
		909
		910	Afterwards it contains the event model that is being used, which is the
		911	name of the Perl class implementing the model. This class is usually one
		912	of the C<AnyEvent::Impl:xxx> modules, but can be any other class in the
		913	case AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode> it
		914	will be C<urxvt::anyevent>).
		915
		916	=item AnyEvent::detect
		917
		918	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
		919	if necessary. You should only call this function right before you would
		920	have created an AnyEvent watcher anyway, that is, as late as possible at
		921	runtime, and not e.g. while initialising of your module.
		922
		923	If you need to do some initialisation before AnyEvent watchers are
		924	created, use C<post_detect>.
		925
		926	=item $guard = AnyEvent::post_detect { BLOCK }
		927
		928	Arranges for the code block to be executed as soon as the event model is
		929	autodetected (or immediately if this has already happened).
		930
		931	The block will be executed I<after> the actual backend has been detected
		932	(C<$AnyEvent::MODEL> is set), but I<before> any watchers have been
		933	created, so it is possible to e.g. patch C<@AnyEvent::ISA> or do
		934	other initialisations - see the sources of L<AnyEvent::Strict> or
		935	L<AnyEvent::AIO> to see how this is used.
		936
		937	The most common usage is to create some global watchers, without forcing
		938	event module detection too early, for example, L<AnyEvent::AIO> creates
		939	and installs the global L<IO::AIO> watcher in a C<post_detect> block to
		940	avoid autodetecting the event module at load time.
		941
		942	If called in scalar or list context, then it creates and returns an object
		943	that automatically removes the callback again when it is destroyed (or
		944	C<undef> when the hook was immediately executed). See L<AnyEvent::AIO> for
		945	a case where this is useful.
		946
		947	Example: Create a watcher for the IO::AIO module and store it in
		948	C<$WATCHER>. Only do so after the event loop is initialised, though.
		949
		950	our WATCHER;
		951
		952	my $guard = AnyEvent::post_detect {
		953	$WATCHER = AnyEvent->io (fh => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb);
		954	};
		955
		956	# the ||= is important in case post_detect immediately runs the block,
		957	# as to not clobber the newly-created watcher. assigning both watcher and
		958	# post_detect guard to the same variable has the advantage of users being
		959	# able to just C<undef $WATCHER> if the watcher causes them grief.
		960
		961	$WATCHER ||= $guard;
		962
		963	=item @AnyEvent::post_detect
		964
		965	If there are any code references in this array (you can C<push> to it
		966	before or after loading AnyEvent), then they will called directly after
		967	the event loop has been chosen.
		968
		969	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		970	if it is defined then the event loop has already been detected, and the
		971	array will be ignored.
		972
		973	Best use C<AnyEvent::post_detect { BLOCK }> when your application allows
		974	it, as it takes care of these details.
		975
		976	This variable is mainly useful for modules that can do something useful
		977	when AnyEvent is used and thus want to know when it is initialised, but do
		978	not need to even load it by default. This array provides the means to hook
		979	into AnyEvent passively, without loading it.
		980
		981	Example: To load Coro::AnyEvent whenever Coro and AnyEvent are used
		982	together, you could put this into Coro (this is the actual code used by
		983	Coro to accomplish this):
		984
		985	if (defined $AnyEvent::MODEL) {
		986	# AnyEvent already initialised, so load Coro::AnyEvent
		987	require Coro::AnyEvent;
		988	} else {
		989	# AnyEvent not yet initialised, so make sure to load Coro::AnyEvent
		990	# as soon as it is
		991	push @AnyEvent::post_detect, sub { require Coro::AnyEvent };
		992	}
		993
		994	=back
		995
		996	=head1 WHAT TO DO IN A MODULE
		997
		998	As a module author, you should C<use AnyEvent> and call AnyEvent methods
		999	freely, but you should not load a specific event module or rely on it.
		1000
		1001	Be careful when you create watchers in the module body - AnyEvent will
		1002	decide which event module to use as soon as the first method is called, so
		1003	by calling AnyEvent in your module body you force the user of your module
		1004	to load the event module first.
		1005
		1006	Never call C<< ->recv >> on a condition variable unless you I<know> that
		1007	the C<< ->send >> method has been called on it already. This is
		1008	because it will stall the whole program, and the whole point of using
		1009	events is to stay interactive.
		1010
		1011	It is fine, however, to call C<< ->recv >> when the user of your module
		1012	requests it (i.e. if you create a http request object ad have a method
		1013	called C<results> that returns the results, it should call C<< ->recv >>
		1014	freely, as the user of your module knows what she is doing. always).
		1015
		1016	=head1 WHAT TO DO IN THE MAIN PROGRAM
		1017
		1018	There will always be a single main program - the only place that should
		1019	dictate which event model to use.
		1020
		1021	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
		1022	do anything special (it does not need to be event-based) and let AnyEvent
		1023	decide which implementation to chose if some module relies on it.
		1024
		1025	If the main program relies on a specific event model - for example, in
		1026	Gtk2 programs you have to rely on the Glib module - you should load the
		1027	event module before loading AnyEvent or any module that uses it: generally
		1028	speaking, you should load it as early as possible. The reason is that
		1029	modules might create watchers when they are loaded, and AnyEvent will
		1030	decide on the event model to use as soon as it creates watchers, and it
		1031	might chose the wrong one unless you load the correct one yourself.
		1032
		1033	You can chose to use a pure-perl implementation by loading the
		1034	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
		1035	everywhere, but letting AnyEvent chose the model is generally better.
		1036
		1037	=head2 MAINLOOP EMULATION
		1038
		1039	Sometimes (often for short test scripts, or even standalone programs who
		1040	only want to use AnyEvent), you do not want to run a specific event loop.
		1041
		1042	In that case, you can use a condition variable like this:
		1043
		1044	AnyEvent->condvar->recv;
		1045
		1046	This has the effect of entering the event loop and looping forever.
		1047
		1048	Note that usually your program has some exit condition, in which case
		1049	it is better to use the "traditional" approach of storing a condition
		1050	variable somewhere, waiting for it, and sending it when the program should
		1051	exit cleanly.
		1052
		1053
		1054	=head1 OTHER MODULES
		1055
		1056	The following is a non-exhaustive list of additional modules that use
		1057	AnyEvent as a client and can therefore be mixed easily with other AnyEvent
		1058	modules and other event loops in the same program. Some of the modules
		1059	come with AnyEvent, most are available via CPAN.
		1060
		1061	=over 4
		1062
		1063	=item L<AnyEvent::Util>
		1064
		1065	Contains various utility functions that replace often-used but blocking
		1066	functions such as C<inet_aton> by event-/callback-based versions.
		1067
		1068	=item L<AnyEvent::Socket>
		1069
		1070	Provides various utility functions for (internet protocol) sockets,
		1071	addresses and name resolution. Also functions to create non-blocking tcp
		1072	connections or tcp servers, with IPv6 and SRV record support and more.
		1073
		1074	=item L<AnyEvent::Handle>
		1075
		1076	Provide read and write buffers, manages watchers for reads and writes,
		1077	supports raw and formatted I/O, I/O queued and fully transparent and
		1078	non-blocking SSL/TLS (via L<AnyEvent::TLS>.
		1079
		1080	=item L<AnyEvent::DNS>
		1081
		1082	Provides rich asynchronous DNS resolver capabilities.
		1083
		1084	=item L<AnyEvent::HTTP>
		1085
		1086	A simple-to-use HTTP library that is capable of making a lot of concurrent
		1087	HTTP requests.
		1088
		1089	=item L<AnyEvent::HTTPD>
		1090
		1091	Provides a simple web application server framework.
		1092
		1093	=item L<AnyEvent::FastPing>
		1094
		1095	The fastest ping in the west.
		1096
		1097	=item L<AnyEvent::DBI>
		1098
		1099	Executes L<DBI> requests asynchronously in a proxy process.
		1100
		1101	=item L<AnyEvent::AIO>
		1102
		1103	Truly asynchronous I/O, should be in the toolbox of every event
		1104	programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
		1105	together.
		1106
		1107	=item L<AnyEvent::BDB>
		1108
		1109	Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
		1110	L<BDB> and AnyEvent together.
		1111
		1112	=item L<AnyEvent::GPSD>
		1113
		1114	A non-blocking interface to gpsd, a daemon delivering GPS information.
		1115
		1116	=item L<AnyEvent::IRC>
		1117
		1118	AnyEvent based IRC client module family (replacing the older Net::IRC3).
		1119
		1120	=item L<AnyEvent::XMPP>
		1121
		1122	AnyEvent based XMPP (Jabber protocol) module family (replacing the older
		1123	Net::XMPP2>.
		1124
		1125	=item L<AnyEvent::IGS>
		1126
		1127	A non-blocking interface to the Internet Go Server protocol (used by
		1128	L<App::IGS>).
		1129
		1130	=item L<Net::FCP>
		1131
		1132	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		1133	of AnyEvent.
		1134
		1135	=item L<Event::ExecFlow>
		1136
		1137	High level API for event-based execution flow control.
		1138
		1139	=item L<Coro>
		1140
		1141	Has special support for AnyEvent via L<Coro::AnyEvent>.
		1142
		1143	=back
		1144
		1145	=cut
		1146
		1147	package AnyEvent;
		1148
		1149	# basically a tuned-down version of common::sense
		1150	sub common_sense {
		1151	# from common:.sense 1.0
		1152	${^WARNING_BITS} = "\xfc\x3f\x33\x00\x0f\xf3\xcf\xc0\xf3\xfc\x33\x00";
		1153	# use strict vars subs - NO UTF-8, as Util.pm doesn't like this atm. (uts46data.pl)
		1154	$^H |= 0x00000600;
		1155	}
		1156
		1157	BEGIN { AnyEvent::common_sense }
		1158
		1159	use Carp ();
		1160
		1161	our $VERSION = '5.23';
		1162	our $MODEL;
		1163
		1164	our $AUTOLOAD;
		1165	our @ISA;
		1166
		1167	our @REGISTRY;
		1168
		1169	our $VERBOSE;
		1170
		1171	BEGIN {
		1172	eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }";
		1173	eval "sub TAINT(){ " . (${^TAINT}*1) . " }";
		1174
		1175	delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
		1176	if ${^TAINT};
		1177
		1178	$VERBOSE = $ENV{PERL_ANYEVENT_VERBOSE}*1;
		1179
		1180	}
		1181
		1182	our $MAX_SIGNAL_LATENCY = 10;
		1183
		1184	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		1185
		1186	{
		1187	my $idx;
		1188	$PROTOCOL{$_} = ++$idx
		1189	for reverse split /\s*,\s*/,
		1190	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		1191	}
		1192
		1193	my @models = (
		1194	[EV:: => AnyEvent::Impl::EV:: , 1],
		1195	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl:: , 1],
		1196	# everything below here will not (normally) be autoprobed
		1197	# as the pureperl backend should work everywhere
		1198	# and is usually faster
		1199	[Event:: => AnyEvent::Impl::Event::, 1],
		1200	[Glib:: => AnyEvent::Impl::Glib:: , 1], # becomes extremely slow with many watchers
		1201	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		1202	[Irssi:: => AnyEvent::Impl::Irssi::], # Irssi has a bogus "Event" package
		1203	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		1204	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		1205	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		1206	[Wx:: => AnyEvent::Impl::POE::],
		1207	[Prima:: => AnyEvent::Impl::POE::],
		1208	# IO::Async is just too broken - we would need workarounds for its
		1209	# byzantine signal and broken child handling, among others.
		1210	# IO::Async is rather hard to detect, as it doesn't have any
		1211	# obvious default class.
		1212	[IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1213	[IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1214	[IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1215	[AnyEvent::Impl::IOAsync:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1216	);
		1217
		1218	our %method = map +($_ => 1),
		1219	qw(io timer time now now_update signal child idle condvar one_event DESTROY);
		1220
		1221	our @post_detect;
		1222
		1223	sub post_detect(&) {
		1224	my ($cb) = @_;
		1225
		1226	if ($MODEL) {
		1227	$cb->();
		1228
		1229	undef
		1230	} else {
		1231	push @post_detect, $cb;
		1232
		1233	defined wantarray
		1234	? bless \$cb, "AnyEvent::Util::postdetect"
		1235	: ()
		1236	}
		1237	}
		1238
		1239	sub AnyEvent::Util::postdetect::DESTROY {
		1240	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		1241	}
		1242
		1243	sub detect() {
		1244	unless ($MODEL) {
		1245	local $SIG{__DIE__};
		1246
		1247	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
		1248	my $model = "AnyEvent::Impl::$1";
		1249	if (eval "require $model") {
		1250	$MODEL = $model;
		1251	warn "AnyEvent: loaded model '$model' (forced by \$ENV{PERL_ANYEVENT_MODEL}), using it.\n" if $VERBOSE >= 2;
		1252	} else {
		1253	warn "AnyEvent: unable to load model '$model' (from \$ENV{PERL_ANYEVENT_MODEL}):\n$@" if $VERBOSE;
		1254	}
		1255	}
		1256
		1257	# check for already loaded models
		1258	unless ($MODEL) {
		1259	for (@REGISTRY, @models) {
		1260	my ($package, $model) = @$_;
		1261	if (${"$package\::VERSION"} > 0) {
		1262	if (eval "require $model") {
		1263	$MODEL = $model;
		1264	warn "AnyEvent: autodetected model '$model', using it.\n" if $VERBOSE >= 2;
		1265	last;
		1266	}
		1267	}
		1268	}
		1269
		1270	unless ($MODEL) {
		1271	# try to autoload a model
		1272	for (@REGISTRY, @models) {
		1273	my ($package, $model, $autoload) = @$_;
		1274	if (
		1275	$autoload
		1276	and eval "require $package"
		1277	and ${"$package\::VERSION"} > 0
		1278	and eval "require $model"
		1279	) {
		1280	$MODEL = $model;
		1281	warn "AnyEvent: autoloaded model '$model', using it.\n" if $VERBOSE >= 2;
		1282	last;
		1283	}
		1284	}
		1285
		1286	$MODEL
		1287	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
		1288	}
		1289	}
		1290
		1291	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		1292
		1293	unshift @ISA, $MODEL;
		1294
		1295	require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
		1296
		1297	(shift @post_detect)->() while @post_detect;
		1298	}
		1299
		1300	$MODEL
		1301	}
		1302
		1303	sub AUTOLOAD {
		1304	(my $func = $AUTOLOAD) =~ s/.*://;
		1305
		1306	$method{$func}
		1307	or Carp::croak "$func: not a valid method for AnyEvent objects";
		1308
		1309	detect unless $MODEL;
		1310
		1311	my $class = shift;
		1312	$class->$func (@_);
		1313	}
		1314
		1315	# utility function to dup a filehandle. this is used by many backends
		1316	# to support binding more than one watcher per filehandle (they usually
		1317	# allow only one watcher per fd, so we dup it to get a different one).
		1318	sub _dupfh($$;$$) {
		1319	my ($poll, $fh, $r, $w) = @_;
		1320
		1321	# cygwin requires the fh mode to be matching, unix doesn't
		1322	my ($rw, $mode) = $poll eq "r" ? ($r, "<&") : ($w, ">&");
		1323
		1324	open my $fh2, $mode, $fh
		1325	or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,";
		1326
		1327	# we assume CLOEXEC is already set by perl in all important cases
		1328
		1329	($fh2, $rw)
		1330	}
		1331
		1332	=head1 SIMPLIFIED AE API
		1333
		1334	Starting with version 5.0, AnyEvent officially supports a second, much
		1335	simpler, API that is designed to reduce the calling, typing and memory
		1336	overhead.
		1337
		1338	See the L<AE> manpage for details.
		1339
		1340	=cut
		1341
		1342	package AE;
		1343
		1344	our $VERSION = $AnyEvent::VERSION;
		1345
		1346	sub io($$$) {
		1347	AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2])
		1348	}
		1349
		1350	sub timer($$$) {
		1351	AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2])
		1352	}
		1353
		1354	sub signal($$) {
		1355	AnyEvent->signal (signal => $_[0], cb => $_[1])
		1356	}
		1357
		1358	sub child($$) {
		1359	AnyEvent->child (pid => $_[0], cb => $_[1])
		1360	}
		1361
		1362	sub idle($) {
		1363	AnyEvent->idle (cb => $_[0])
		1364	}
		1365
		1366	sub cv(;&) {
		1367	AnyEvent->condvar (@_ ? (cb => $_[0]) : ())
		1368	}
		1369
		1370	sub now() {
		1371	AnyEvent->now
		1372	}
		1373
		1374	sub now_update() {
		1375	AnyEvent->now_update
		1376	}
		1377
		1378	sub time() {
		1379	AnyEvent->time
		1380	}
		1381
		1382	package AnyEvent::Base;
		1383
		1384	# default implementations for many methods
		1385
		1386	sub _time() {
		1387	# probe for availability of Time::HiRes
		1388	if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
		1389	warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8;
		1390	*_time = \&Time::HiRes::time;
		1391	# if (eval "use POSIX (); (POSIX::times())...
		1392	} else {
		1393	warn "AnyEvent: using built-in time(), WARNING, no sub-second resolution!\n" if $VERBOSE;
		1394	*_time = sub { time }; # epic fail
		1395	}
		1396
		1397	&_time
		1398	}
		1399
		1400	sub time { _time }
		1401	sub now { _time }
		1402	sub now_update { }
		1403
		1404	# default implementation for ->condvar
		1405
		1406	sub condvar {
		1407	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
		1408	}
		1409
		1410	# default implementation for ->signal
		1411
		1412	our $HAVE_ASYNC_INTERRUPT;
		1413
		1414	sub _have_async_interrupt() {
		1415	$HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT}
		1416	&& eval "use Async::Interrupt 1.02 (); 1")
		1417	unless defined $HAVE_ASYNC_INTERRUPT;
		1418
		1419	$HAVE_ASYNC_INTERRUPT
		1420	}
		1421
		1422	our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
		1423	our (%SIG_ASY, %SIG_ASY_W);
		1424	our ($SIG_COUNT, $SIG_TW);
		1425
		1426	sub _signal_exec {
		1427	$HAVE_ASYNC_INTERRUPT
		1428	? $SIGPIPE_R->drain
		1429	: sysread $SIGPIPE_R, (my $dummy), 9;
		1430
		1431	while (%SIG_EV) {
		1432	for (keys %SIG_EV) {
		1433	delete $SIG_EV{$_};
		1434	$_->() for values %{ $SIG_CB{$_} || {} };
		1435	}
		1436	}
		1437	}
		1438
		1439	# install a dummy wakeup watcher to reduce signal catching latency
		1440	sub _sig_add() {
		1441	unless ($SIG_COUNT++) {
		1442	# try to align timer on a full-second boundary, if possible
		1443	my $NOW = AE::now;
		1444
		1445	$SIG_TW = AE::timer
		1446	$MAX_SIGNAL_LATENCY - ($NOW - int $NOW),
		1447	$MAX_SIGNAL_LATENCY,
		1448	sub { } # just for the PERL_ASYNC_CHECK
		1449	;
		1450	}
		1451	}
		1452
		1453	sub _sig_del {
		1454	undef $SIG_TW
		1455	unless --$SIG_COUNT;
		1456	}
		1457
		1458	our $_sig_name_init; $_sig_name_init = sub {
		1459	eval q{ # poor man's autoloading
		1460	undef $_sig_name_init;
		1461
		1462	if (_have_async_interrupt) {
		1463	*sig2num = \&Async::Interrupt::sig2num;
		1464	*sig2name = \&Async::Interrupt::sig2name;
		1465	} else {
		1466	require Config;
		1467
		1468	my %signame2num;
		1469	@signame2num{ split ' ', $Config::Config{sig_name} }
		1470	= split ' ', $Config::Config{sig_num};
		1471
		1472	my @signum2name;
		1473	@signum2name[values %signame2num] = keys %signame2num;
		1474
		1475	*sig2num = sub($) {
		1476	$_[0] > 0 ? shift : $signame2num{+shift}
		1477	};
		1478	*sig2name = sub ($) {
		1479	$_[0] > 0 ? $signum2name[+shift] : shift
		1480	};
		1481	}
		1482	};
		1483	die if $@;
		1484	};
		1485
		1486	sub sig2num ($) { &$_sig_name_init; &sig2num }
		1487	sub sig2name($) { &$_sig_name_init; &sig2name }
		1488
		1489	sub signal {
		1490	eval q{ # poor man's autoloading {}
		1491	# probe for availability of Async::Interrupt
		1492	if (_have_async_interrupt) {
		1493	warn "AnyEvent: using Async::Interrupt for race-free signal handling.\n" if $VERBOSE >= 8;
		1494
		1495	$SIGPIPE_R = new Async::Interrupt::EventPipe;
		1496	$SIG_IO = AE::io $SIGPIPE_R->fileno, 0, \&_signal_exec;
		1497
		1498	} else {
		1499	warn "AnyEvent: using emulated perl signal handling with latency timer.\n" if $VERBOSE >= 8;
		1500
		1501	require Fcntl;
		1502
		1503	if (AnyEvent::WIN32) {
		1504	require AnyEvent::Util;
		1505
		1506	($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
		1507	AnyEvent::Util::fh_nonblocking ($SIGPIPE_R, 1) if $SIGPIPE_R;
		1508	AnyEvent::Util::fh_nonblocking ($SIGPIPE_W, 1) if $SIGPIPE_W; # just in case
		1509	} else {
		1510	pipe $SIGPIPE_R, $SIGPIPE_W;
		1511	fcntl $SIGPIPE_R, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_R;
		1512	fcntl $SIGPIPE_W, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_W; # just in case
		1513
		1514	# not strictly required, as $^F is normally 2, but let's make sure...
		1515	fcntl $SIGPIPE_R, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1516	fcntl $SIGPIPE_W, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1517	}
		1518
		1519	$SIGPIPE_R
		1520	or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
		1521
		1522	$SIG_IO = AE::io $SIGPIPE_R, 0, \&_signal_exec;
		1523	}
		1524
		1525	*signal = sub {
		1526	my (undef, %arg) = @_;
		1527
		1528	my $signal = uc $arg{signal}
		1529	or Carp::croak "required option 'signal' is missing";
		1530
		1531	if ($HAVE_ASYNC_INTERRUPT) {
		1532	# async::interrupt
		1533
		1534	$signal = sig2num $signal;
		1535	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1536
		1537	$SIG_ASY{$signal} ||= new Async::Interrupt
		1538	cb => sub { undef $SIG_EV{$signal} },
		1539	signal => $signal,
		1540	pipe => [$SIGPIPE_R->filenos],
		1541	pipe_autodrain => 0,
		1542	;
		1543
		1544	} else {
		1545	# pure perl
		1546
		1547	# AE::Util has been loaded in signal
		1548	$signal = sig2name $signal;
		1549	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1550
		1551	$SIG{$signal} ||= sub {
		1552	local $!;
		1553	syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
		1554	undef $SIG_EV{$signal};
		1555	};
		1556
		1557	# can't do signal processing without introducing races in pure perl,
		1558	# so limit the signal latency.
		1559	_sig_add;
		1560	}
		1561
		1562	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
		1563	};
		1564
		1565	*AnyEvent::Base::signal::DESTROY = sub {
		1566	my ($signal, $cb) = @{$_[0]};
		1567
		1568	_sig_del;
		1569
		1570	delete $SIG_CB{$signal}{$cb};
		1571
		1572	$HAVE_ASYNC_INTERRUPT
		1573	? delete $SIG_ASY{$signal}
		1574	: # delete doesn't work with older perls - they then
		1575	# print weird messages, or just unconditionally exit
		1576	# instead of getting the default action.
		1577	undef $SIG{$signal}
		1578	unless keys %{ $SIG_CB{$signal} };
		1579	};
		1580	};
		1581	die if $@;
		1582	&signal
		1583	}
		1584
		1585	# default implementation for ->child
		1586
		1587	our %PID_CB;
		1588	our $CHLD_W;
		1589	our $CHLD_DELAY_W;
		1590	our $WNOHANG;
		1591
		1592	sub _emit_childstatus($$) {
		1593	my (undef, $rpid, $rstatus) = @_;
		1594
		1595	$_->($rpid, $rstatus)
		1596	for values %{ $PID_CB{$rpid} || {} },
		1597	values %{ $PID_CB{0} || {} };
		1598	}
		1599
		1600	sub _sigchld {
		1601	my $pid;
		1602
		1603	AnyEvent->_emit_childstatus ($pid, $?)
		1604	while ($pid = waitpid -1, $WNOHANG) > 0;
		1605	}
		1606
		1607	sub child {
		1608	my (undef, %arg) = @_;
		1609
		1610	defined (my $pid = $arg{pid} + 0)
		1611	or Carp::croak "required option 'pid' is missing";
		1612
		1613	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
		1614
		1615	# WNOHANG is almost cetrainly 1 everywhere
		1616	$WNOHANG ||= $^O =~ /^(?:openbsd|netbsd|linux|freebsd|cygwin|MSWin32)$/
		1617	? 1
		1618	: eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
		1619
		1620	unless ($CHLD_W) {
		1621	$CHLD_W = AE::signal CHLD => \&_sigchld;
		1622	# child could be a zombie already, so make at least one round
		1623	&_sigchld;
		1624	}
		1625
		1626	bless [$pid, $arg{cb}], "AnyEvent::Base::child"
		1627	}
		1628
		1629	sub AnyEvent::Base::child::DESTROY {
		1630	my ($pid, $cb) = @{$_[0]};
		1631
		1632	delete $PID_CB{$pid}{$cb};
		1633	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
		1634
		1635	undef $CHLD_W unless keys %PID_CB;
		1636	}
		1637
		1638	# idle emulation is done by simply using a timer, regardless
		1639	# of whether the process is idle or not, and not letting
		1640	# the callback use more than 50% of the time.
		1641	sub idle {
		1642	my (undef, %arg) = @_;
		1643
		1644	my ($cb, $w, $rcb) = $arg{cb};
		1645
		1646	$rcb = sub {
		1647	if ($cb) {
		1648	$w = _time;
		1649	&$cb;
		1650	$w = _time - $w;
		1651
		1652	# never use more then 50% of the time for the idle watcher,
		1653	# within some limits
		1654	$w = 0.0001 if $w < 0.0001;
		1655	$w = 5 if $w > 5;
		1656
		1657	$w = AE::timer $w, 0, $rcb;
		1658	} else {
		1659	# clean up...
		1660	undef $w;
		1661	undef $rcb;
		1662	}
		1663	};
		1664
		1665	$w = AE::timer 0.05, 0, $rcb;
		1666
		1667	bless \\$cb, "AnyEvent::Base::idle"
		1668	}
		1669
		1670	sub AnyEvent::Base::idle::DESTROY {
		1671	undef $${$_[0]};
		1672	}
		1673
		1674	package AnyEvent::CondVar;
		1675
		1676	our @ISA = AnyEvent::CondVar::Base::;
		1677
		1678	package AnyEvent::CondVar::Base;
		1679
		1680	#use overload
		1681	# '&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1682	# fallback => 1;
		1683
		1684	# save 300+ kilobytes by dirtily hardcoding overloading
		1685	${"AnyEvent::CondVar::Base::OVERLOAD"}{dummy}++; # Register with magic by touching.
		1686	*{'AnyEvent::CondVar::Base::()'} = sub { }; # "Make it findable via fetchmethod."
		1687	*{'AnyEvent::CondVar::Base::(&{}'} = sub { my $self = shift; sub { $self->send (@_) } }; # &{}
		1688	${'AnyEvent::CondVar::Base::()'} = 1; # fallback
		1689
		1690	our $WAITING;
		1691
		1692	sub _send {
		1693	# nop
		1694	}
		1695
		1696	sub send {
		1697	my $cv = shift;
		1698	$cv->{_ae_sent} = [@_];
		1699	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1700	$cv->_send;
		1701	}
		1702
		1703	sub croak {
		1704	$_[0]{_ae_croak} = $_[1];
		1705	$_[0]->send;
		1706	}
		1707
		1708	sub ready {
		1709	$_[0]{_ae_sent}
		1710	}
		1711
		1712	sub _wait {
		1713	$WAITING
		1714	and !$_[0]{_ae_sent}
		1715	and Carp::croak "AnyEvent::CondVar: recursive blocking wait detected";
		1716
		1717	local $WAITING = 1;
		1718	AnyEvent->one_event while !$_[0]{_ae_sent};
		1719	}
		1720
		1721	sub recv {
		1722	$_[0]->_wait;
		1723
		1724	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1725	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1726	}
		1727
		1728	sub cb {
		1729	my $cv = shift;
		1730
		1731	@_
		1732	and $cv->{_ae_cb} = shift
		1733	and $cv->{_ae_sent}
		1734	and (delete $cv->{_ae_cb})->($cv);
		1735
		1736	$cv->{_ae_cb}
		1737	}
		1738
		1739	sub begin {
		1740	++$_[0]{_ae_counter};
		1741	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1742	}
		1743
		1744	sub end {
		1745	return if --$_[0]{_ae_counter};
		1746	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1747	}
		1748
		1749	# undocumented/compatibility with pre-3.4
		1750	*broadcast = \&send;
		1751	*wait = \&_wait;
		1752
		1753	=head1 ERROR AND EXCEPTION HANDLING
		1754
		1755	In general, AnyEvent does not do any error handling - it relies on the
		1756	caller to do that if required. The L<AnyEvent::Strict> module (see also
		1757	the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
		1758	checking of all AnyEvent methods, however, which is highly useful during
		1759	development.
		1760
		1761	As for exception handling (i.e. runtime errors and exceptions thrown while
		1762	executing a callback), this is not only highly event-loop specific, but
		1763	also not in any way wrapped by this module, as this is the job of the main
		1764	program.
		1765
		1766	The pure perl event loop simply re-throws the exception (usually
		1767	within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
		1768	$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
		1769	so on.
		1770
		1771	=head1 ENVIRONMENT VARIABLES
		1772
		1773	The following environment variables are used by this module or its
		1774	submodules.
		1775
		1776	Note that AnyEvent will remove I<all> environment variables starting with
		1777	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		1778	enabled.
		1779
		1780	=over 4
		1781
		1782	=item C<PERL_ANYEVENT_VERBOSE>
		1783
		1784	By default, AnyEvent will be completely silent except in fatal
		1785	conditions. You can set this environment variable to make AnyEvent more
		1786	talkative.
		1787
		1788	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1789	conditions, such as not being able to load the event model specified by
		1790	C<PERL_ANYEVENT_MODEL>.
		1791
		1792	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1793	model it chooses.
		1794
		1795	When set to C<8> or higher, then AnyEvent will report extra information on
		1796	which optional modules it loads and how it implements certain features.
		1797
		1798	=item C<PERL_ANYEVENT_STRICT>
		1799
		1800	AnyEvent does not do much argument checking by default, as thorough
		1801	argument checking is very costly. Setting this variable to a true value
		1802	will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
		1803	check the arguments passed to most method calls. If it finds any problems,
		1804	it will croak.
		1805
		1806	In other words, enables "strict" mode.
		1807
		1808	Unlike C<use strict> (or it's modern cousin, C<< use L<common::sense>
		1809	>>, it is definitely recommended to keep it off in production. Keeping
		1810	C<PERL_ANYEVENT_STRICT=1> in your environment while developing programs
		1811	can be very useful, however.
		1812
		1813	=item C<PERL_ANYEVENT_MODEL>
		1814
		1815	This can be used to specify the event model to be used by AnyEvent, before
		1816	auto detection and -probing kicks in. It must be a string consisting
		1817	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1818	and the resulting module name is loaded and if the load was successful,
		1819	used as event model. If it fails to load AnyEvent will proceed with
		1820	auto detection and -probing.
		1821
		1822	This functionality might change in future versions.
		1823
		1824	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1825	could start your program like this:
		1826
		1827	PERL_ANYEVENT_MODEL=Perl perl ...
		1828
		1829	=item C<PERL_ANYEVENT_PROTOCOLS>
		1830
		1831	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1832	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1833	of auto probing).
		1834
		1835	Must be set to a comma-separated list of protocols or address families,
		1836	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1837	used, and preference will be given to protocols mentioned earlier in the
		1838	list.
		1839
		1840	This variable can effectively be used for denial-of-service attacks
		1841	against local programs (e.g. when setuid), although the impact is likely
		1842	small, as the program has to handle conenction and other failures anyways.
		1843
		1844	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1845	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1846	- only support IPv4, never try to resolve or contact IPv6
		1847	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1848	IPv6, but prefer IPv6 over IPv4.
		1849
		1850	=item C<PERL_ANYEVENT_EDNS0>
		1851
		1852	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1853	for DNS. This extension is generally useful to reduce DNS traffic, but
		1854	some (broken) firewalls drop such DNS packets, which is why it is off by
		1855	default.
		1856
		1857	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1858	EDNS0 in its DNS requests.
		1859
		1860	=item C<PERL_ANYEVENT_MAX_FORKS>
		1861
		1862	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1863	will create in parallel.
		1864
		1865	=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
		1866
		1867	The default value for the C<max_outstanding> parameter for the default DNS
		1868	resolver - this is the maximum number of parallel DNS requests that are
		1869	sent to the DNS server.
		1870
		1871	=item C<PERL_ANYEVENT_RESOLV_CONF>
		1872
		1873	The file to use instead of F</etc/resolv.conf> (or OS-specific
		1874	configuration) in the default resolver. When set to the empty string, no
		1875	default config will be used.
		1876
		1877	=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
		1878
		1879	When neither C<ca_file> nor C<ca_path> was specified during
		1880	L<AnyEvent::TLS> context creation, and either of these environment
		1881	variables exist, they will be used to specify CA certificate locations
		1882	instead of a system-dependent default.
		1883
		1884	=item C<PERL_ANYEVENT_AVOID_GUARD> and C<PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT>
		1885
		1886	When these are set to C<1>, then the respective modules are not
		1887	loaded. Mostly good for testing AnyEvent itself.
		1888
		1889	=back
		1890
		1891	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1892
		1893	This is an advanced topic that you do not normally need to use AnyEvent in
		1894	a module. This section is only of use to event loop authors who want to
		1895	provide AnyEvent compatibility.
		1896
		1897	If you need to support another event library which isn't directly
		1898	supported by AnyEvent, you can supply your own interface to it by
		1899	pushing, before the first watcher gets created, the package name of
		1900	the event module and the package name of the interface to use onto
		1901	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
		1902	AnyEvent, so it is reasonably cheap.
		1903
		1904	Example:
		1905
		1906	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
		1907
		1908	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
		1909	package/class when it finds the C<urxvt> package/module is already loaded.
		1910
		1911	When AnyEvent is loaded and asked to find a suitable event model, it
		1912	will first check for the presence of urxvt by trying to C<use> the
		1913	C<urxvt::anyevent> module.
		1914
		1915	The class should provide implementations for all watcher types. See
		1916	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
		1917	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
		1918	see the sources.
		1919
		1920	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1921	provide suitable (hopefully) replacements.
		1922
		1923	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
		1924	terminal emulator uses the above line as-is. An interface isn't included
		1925	in AnyEvent because it doesn't make sense outside the embedded interpreter
		1926	inside I<rxvt-unicode>, and it is updated and maintained as part of the
		1927	I<rxvt-unicode> distribution.
		1928
		1929	I<rxvt-unicode> also cheats a bit by not providing blocking access to
		1930	condition variables: code blocking while waiting for a condition will
		1931	C<die>. This still works with most modules/usages, and blocking calls must
		1932	not be done in an interactive application, so it makes sense.
		1933
		1934	=head1 EXAMPLE PROGRAM
		1935
		1936	The following program uses an I/O watcher to read data from STDIN, a timer
		1937	to display a message once per second, and a condition variable to quit the
		1938	program when the user enters quit:
		1939
		1940	use AnyEvent;
		1941
		1942	my $cv = AnyEvent->condvar;
		1943
		1944	my $io_watcher = AnyEvent->io (
		1945	fh => \*STDIN,
		1946	poll => 'r',
		1947	cb => sub {
		1948	warn "io event <$_[0]>\n"; # will always output <r>
		1949	chomp (my $input = <STDIN>); # read a line
		1950	warn "read: $input\n"; # output what has been read
		1951	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1952	},
		1953	);
		1954
		1955	my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub {
		1956	warn "timeout\n"; # print 'timeout' at most every second
		1957	});
		1958
		1959	$cv->recv; # wait until user enters /^q/i
		1960
		1961	=head1 REAL-WORLD EXAMPLE
		1962
		1963	Consider the L<Net::FCP> module. It features (among others) the following
		1964	API calls, which are to freenet what HTTP GET requests are to http:
		1965
		1966	my $data = $fcp->client_get ($url); # blocks
		1967
		1968	my $transaction = $fcp->txn_client_get ($url); # does not block
		1969	$transaction->cb ( sub { ... } ); # set optional result callback
		1970	my $data = $transaction->result; # possibly blocks
		1971
		1972	The C<client_get> method works like C<LWP::Simple::get>: it requests the
		1973	given URL and waits till the data has arrived. It is defined to be:
		1974
		1975	sub client_get { $_[0]->txn_client_get ($_[1])->result }
		1976
		1977	And in fact is automatically generated. This is the blocking API of
		1978	L<Net::FCP>, and it works as simple as in any other, similar, module.
		1979
		1980	More complicated is C<txn_client_get>: It only creates a transaction
		1981	(completion, result, ...) object and initiates the transaction.
		1982
		1983	my $txn = bless { }, Net::FCP::Txn::;
		1984
		1985	It also creates a condition variable that is used to signal the completion
		1986	of the request:
		1987
		1988	$txn->{finished} = AnyAvent->condvar;
		1989
		1990	It then creates a socket in non-blocking mode.
		1991
		1992	socket $txn->{fh}, ...;
		1993	fcntl $txn->{fh}, F_SETFL, O_NONBLOCK;
		1994	connect $txn->{fh}, ...
		1995	and !$!{EWOULDBLOCK}
		1996	and !$!{EINPROGRESS}
		1997	and Carp::croak "unable to connect: $!\n";
		1998
		1999	Then it creates a write-watcher which gets called whenever an error occurs
		2000	or the connection succeeds:
		2001
		2002	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'w', cb => sub { $txn->fh_ready_w });
		2003
		2004	And returns this transaction object. The C<fh_ready_w> callback gets
		2005	called as soon as the event loop detects that the socket is ready for
		2006	writing.
		2007
		2008	The C<fh_ready_w> method makes the socket blocking again, writes the
		2009	request data and replaces the watcher by a read watcher (waiting for reply
		2010	data). The actual code is more complicated, but that doesn't matter for
		2011	this example:
		2012
		2013	fcntl $txn->{fh}, F_SETFL, 0;
		2014	syswrite $txn->{fh}, $txn->{request}
		2015	or die "connection or write error";
		2016	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
		2017
		2018	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
		2019	result and signals any possible waiters that the request has finished:
		2020
		2021	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
		2022
		2023	if (end-of-file or data complete) {
		2024	$txn->{result} = $txn->{buf};
		2025	$txn->{finished}->send;
		2026	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
		2027	}
		2028
		2029	The C<result> method, finally, just waits for the finished signal (if the
		2030	request was already finished, it doesn't wait, of course, and returns the
		2031	data:
		2032
		2033	$txn->{finished}->recv;
		2034	return $txn->{result};
		2035
		2036	The actual code goes further and collects all errors (C<die>s, exceptions)
		2037	that occurred during request processing. The C<result> method detects
		2038	whether an exception as thrown (it is stored inside the $txn object)
		2039	and just throws the exception, which means connection errors and other
		2040	problems get reported tot he code that tries to use the result, not in a
		2041	random callback.
		2042
		2043	All of this enables the following usage styles:
		2044
		2045	1. Blocking:
		2046
		2047	my $data = $fcp->client_get ($url);
		2048
		2049	2. Blocking, but running in parallel:
		2050
		2051	my @datas = map $_->result,
		2052	map $fcp->txn_client_get ($_),
		2053	@urls;
		2054
		2055	Both blocking examples work without the module user having to know
		2056	anything about events.
		2057
		2058	3a. Event-based in a main program, using any supported event module:
		2059
		2060	use EV;
		2061
		2062	$fcp->txn_client_get ($url)->cb (sub {
		2063	my $txn = shift;
		2064	my $data = $txn->result;
16	...	2065	...
17	});	2066	});
18		2067
19	# watchers get canceled whenever $w is destroyed	2068	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		2069
25	my $w = AnyEvent->condvar; # kind of main loop replacement	2070	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		2071
30	=head1 DESCRIPTION	2072	use AnyEvent;
31		2073
32	L<AnyEvent> provides an identical interface to multiple event loops. This	2074	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		2075
37	The interface itself is vaguely similar but not identical to the Event	2076	$fcp->txn_client_get ($url)->cb (sub {
38	module.	2077	...
		2078	$quit->send;
		2079	});
39		2080
40	On the first call of any method, the module tries to detect the currently	2081	$quit->recv;
41	loaded event loop by probing wether any of the following modules is	2082
42	loaded: L<Coro::Event>, L<Event>, L<Glib>, L<Tk>. The first one found is	2083
43	used. If none is found, the module tries to load these modules in the	2084	=head1 BENCHMARKS
44	order given. The first one that could be successfully loaded will be	2085
45	used. If still none could be found, it will issue an error.	2086	To give you an idea of the performance and overheads that AnyEvent adds
		2087	over the event loops themselves and to give you an impression of the speed
		2088	of various event loops I prepared some benchmarks.
		2089
		2090	=head2 BENCHMARKING ANYEVENT OVERHEAD
		2091
		2092	Here is a benchmark of various supported event models used natively and
		2093	through AnyEvent. The benchmark creates a lot of timers (with a zero
		2094	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		2095	which it is), lets them fire exactly once and destroys them again.
		2096
		2097	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		2098	distribution. It uses the L<AE> interface, which makes a real difference
		2099	for the EV and Perl backends only.
		2100
		2101	=head3 Explanation of the columns
		2102
		2103	I<watcher> is the number of event watchers created/destroyed. Since
		2104	different event models feature vastly different performances, each event
		2105	loop was given a number of watchers so that overall runtime is acceptable
		2106	and similar between tested event loop (and keep them from crashing): Glib
		2107	would probably take thousands of years if asked to process the same number
		2108	of watchers as EV in this benchmark.
		2109
		2110	I<bytes> is the number of bytes (as measured by the resident set size,
		2111	RSS) consumed by each watcher. This method of measuring captures both C
		2112	and Perl-based overheads.
		2113
		2114	I<create> is the time, in microseconds (millionths of seconds), that it
		2115	takes to create a single watcher. The callback is a closure shared between
		2116	all watchers, to avoid adding memory overhead. That means closure creation
		2117	and memory usage is not included in the figures.
		2118
		2119	I<invoke> is the time, in microseconds, used to invoke a simple
		2120	callback. The callback simply counts down a Perl variable and after it was
		2121	invoked "watcher" times, it would C<< ->send >> a condvar once to
		2122	signal the end of this phase.
		2123
		2124	I<destroy> is the time, in microseconds, that it takes to destroy a single
		2125	watcher.
		2126
		2127	=head3 Results
		2128
		2129	name watchers bytes create invoke destroy comment
		2130	EV/EV 100000 223 0.47 0.43 0.27 EV native interface
		2131	EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers
		2132	Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal
		2133	Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation
		2134	Event/Event 16000 516 31.16 31.84 0.82 Event native interface
		2135	Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers
		2136	IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll
		2137	IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll
		2138	Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour
		2139	Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers
		2140	POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event
		2141	POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select
		2142
		2143	=head3 Discussion
		2144
		2145	The benchmark does I<not> measure scalability of the event loop very
		2146	well. For example, a select-based event loop (such as the pure perl one)
		2147	can never compete with an event loop that uses epoll when the number of
		2148	file descriptors grows high. In this benchmark, all events become ready at
		2149	the same time, so select/poll-based implementations get an unnatural speed
		2150	boost.
		2151
		2152	Also, note that the number of watchers usually has a nonlinear effect on
		2153	overall speed, that is, creating twice as many watchers doesn't take twice
		2154	the time - usually it takes longer. This puts event loops tested with a
		2155	higher number of watchers at a disadvantage.
		2156
		2157	To put the range of results into perspective, consider that on the
		2158	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		2159	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		2160	cycles with POE.
		2161
		2162	C<EV> is the sole leader regarding speed and memory use, which are both
		2163	maximal/minimal, respectively. When using the L<AE> API there is zero
		2164	overhead (when going through the AnyEvent API create is about 5-6 times
		2165	slower, with other times being equal, so still uses far less memory than
		2166	any other event loop and is still faster than Event natively).
		2167
		2168	The pure perl implementation is hit in a few sweet spots (both the
		2169	constant timeout and the use of a single fd hit optimisations in the perl
		2170	interpreter and the backend itself). Nevertheless this shows that it
		2171	adds very little overhead in itself. Like any select-based backend its
		2172	performance becomes really bad with lots of file descriptors (and few of
		2173	them active), of course, but this was not subject of this benchmark.
		2174
		2175	The C<Event> module has a relatively high setup and callback invocation
		2176	cost, but overall scores in on the third place.
		2177
		2178	C<IO::Async> performs admirably well, about on par with C<Event>, even
		2179	when using its pure perl backend.
		2180
		2181	C<Glib>'s memory usage is quite a bit higher, but it features a
		2182	faster callback invocation and overall ends up in the same class as
		2183	C<Event>. However, Glib scales extremely badly, doubling the number of
		2184	watchers increases the processing time by more than a factor of four,
		2185	making it completely unusable when using larger numbers of watchers
		2186	(note that only a single file descriptor was used in the benchmark, so
		2187	inefficiencies of C<poll> do not account for this).
		2188
		2189	The C<Tk> adaptor works relatively well. The fact that it crashes with
		2190	more than 2000 watchers is a big setback, however, as correctness takes
		2191	precedence over speed. Nevertheless, its performance is surprising, as the
		2192	file descriptor is dup()ed for each watcher. This shows that the dup()
		2193	employed by some adaptors is not a big performance issue (it does incur a
		2194	hidden memory cost inside the kernel which is not reflected in the figures
		2195	above).
		2196
		2197	C<POE>, regardless of underlying event loop (whether using its pure perl
		2198	select-based backend or the Event module, the POE-EV backend couldn't
		2199	be tested because it wasn't working) shows abysmal performance and
		2200	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		2201	as EV watchers, and 10 times as much memory as Event (the high memory
		2202	requirements are caused by requiring a session for each watcher). Watcher
		2203	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		2204	implementation.
		2205
		2206	The design of the POE adaptor class in AnyEvent can not really account
		2207	for the performance issues, though, as session creation overhead is
		2208	small compared to execution of the state machine, which is coded pretty
		2209	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		2210	using multiple sessions is not a good approach, especially regarding
		2211	memory usage, even the author of POE could not come up with a faster
		2212	design).
		2213
		2214	=head3 Summary
46		2215
47	=over 4	2216	=over 4
48		2217
		2218	=item * Using EV through AnyEvent is faster than any other event loop
		2219	(even when used without AnyEvent), but most event loops have acceptable
		2220	performance with or without AnyEvent.
		2221
		2222	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		2223	the actual event loop, only with extremely fast event loops such as EV
		2224	adds AnyEvent significant overhead.
		2225
		2226	=item * You should avoid POE like the plague if you want performance or
		2227	reasonable memory usage.
		2228
		2229	=back
		2230
		2231	=head2 BENCHMARKING THE LARGE SERVER CASE
		2232
		2233	This benchmark actually benchmarks the event loop itself. It works by
		2234	creating a number of "servers": each server consists of a socket pair, a
		2235	timeout watcher that gets reset on activity (but never fires), and an I/O
		2236	watcher waiting for input on one side of the socket. Each time the socket
		2237	watcher reads a byte it will write that byte to a random other "server".
		2238
		2239	The effect is that there will be a lot of I/O watchers, only part of which
		2240	are active at any one point (so there is a constant number of active
		2241	fds for each loop iteration, but which fds these are is random). The
		2242	timeout is reset each time something is read because that reflects how
		2243	most timeouts work (and puts extra pressure on the event loops).
		2244
		2245	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		2246	(1%) are active. This mirrors the activity of large servers with many
		2247	connections, most of which are idle at any one point in time.
		2248
		2249	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		2250	distribution. It uses the L<AE> interface, which makes a real difference
		2251	for the EV and Perl backends only.
		2252
		2253	=head3 Explanation of the columns
		2254
		2255	I<sockets> is the number of sockets, and twice the number of "servers" (as
		2256	each server has a read and write socket end).
		2257
		2258	I<create> is the time it takes to create a socket pair (which is
		2259	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		2260
		2261	I<request>, the most important value, is the time it takes to handle a
		2262	single "request", that is, reading the token from the pipe and forwarding
		2263	it to another server. This includes deleting the old timeout and creating
		2264	a new one that moves the timeout into the future.
		2265
		2266	=head3 Results
		2267
		2268	name sockets create request
		2269	EV 20000 62.66 7.99
		2270	Perl 20000 68.32 32.64
		2271	IOAsync 20000 174.06 101.15 epoll
		2272	IOAsync 20000 174.67 610.84 poll
		2273	Event 20000 202.69 242.91
		2274	Glib 20000 557.01 1689.52
		2275	POE 20000 341.54 12086.32 uses POE::Loop::Event
		2276
		2277	=head3 Discussion
		2278
		2279	This benchmark I<does> measure scalability and overall performance of the
		2280	particular event loop.
		2281
		2282	EV is again fastest. Since it is using epoll on my system, the setup time
		2283	is relatively high, though.
		2284
		2285	Perl surprisingly comes second. It is much faster than the C-based event
		2286	loops Event and Glib.
		2287
		2288	IO::Async performs very well when using its epoll backend, and still quite
		2289	good compared to Glib when using its pure perl backend.
		2290
		2291	Event suffers from high setup time as well (look at its code and you will
		2292	understand why). Callback invocation also has a high overhead compared to
		2293	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		2294	uses select or poll in basically all documented configurations.
		2295
		2296	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		2297	clearly fails to perform with many filehandles or in busy servers.
		2298
		2299	POE is still completely out of the picture, taking over 1000 times as long
		2300	as EV, and over 100 times as long as the Perl implementation, even though
		2301	it uses a C-based event loop in this case.
		2302
		2303	=head3 Summary
		2304
		2305	=over 4
		2306
		2307	=item * The pure perl implementation performs extremely well.
		2308
		2309	=item * Avoid Glib or POE in large projects where performance matters.
		2310
		2311	=back
		2312
		2313	=head2 BENCHMARKING SMALL SERVERS
		2314
		2315	While event loops should scale (and select-based ones do not...) even to
		2316	large servers, most programs we (or I :) actually write have only a few
		2317	I/O watchers.
		2318
		2319	In this benchmark, I use the same benchmark program as in the large server
		2320	case, but it uses only eight "servers", of which three are active at any
		2321	one time. This should reflect performance for a small server relatively
		2322	well.
		2323
		2324	The columns are identical to the previous table.
		2325
		2326	=head3 Results
		2327
		2328	name sockets create request
		2329	EV 16 20.00 6.54
		2330	Perl 16 25.75 12.62
		2331	Event 16 81.27 35.86
		2332	Glib 16 32.63 15.48
		2333	POE 16 261.87 276.28 uses POE::Loop::Event
		2334
		2335	=head3 Discussion
		2336
		2337	The benchmark tries to test the performance of a typical small
		2338	server. While knowing how various event loops perform is interesting, keep
		2339	in mind that their overhead in this case is usually not as important, due
		2340	to the small absolute number of watchers (that is, you need efficiency and
		2341	speed most when you have lots of watchers, not when you only have a few of
		2342	them).
		2343
		2344	EV is again fastest.
		2345
		2346	Perl again comes second. It is noticeably faster than the C-based event
		2347	loops Event and Glib, although the difference is too small to really
		2348	matter.
		2349
		2350	POE also performs much better in this case, but is is still far behind the
		2351	others.
		2352
		2353	=head3 Summary
		2354
		2355	=over 4
		2356
		2357	=item * C-based event loops perform very well with small number of
		2358	watchers, as the management overhead dominates.
		2359
		2360	=back
		2361
		2362	=head2 THE IO::Lambda BENCHMARK
		2363
		2364	Recently I was told about the benchmark in the IO::Lambda manpage, which
		2365	could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
		2366	simply compares IO::Lambda with POE, and IO::Lambda looks better (which
		2367	shouldn't come as a surprise to anybody). As such, the benchmark is
		2368	fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
		2369	very optimal. But how would AnyEvent compare when used without the extra
		2370	baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
		2371
		2372	The benchmark itself creates an echo-server, and then, for 500 times,
		2373	connects to the echo server, sends a line, waits for the reply, and then
		2374	creates the next connection. This is a rather bad benchmark, as it doesn't
		2375	test the efficiency of the framework or much non-blocking I/O, but it is a
		2376	benchmark nevertheless.
		2377
		2378	name runtime
		2379	Lambda/select 0.330 sec
		2380	+ optimized 0.122 sec
		2381	Lambda/AnyEvent 0.327 sec
		2382	+ optimized 0.138 sec
		2383	Raw sockets/select 0.077 sec
		2384	POE/select, components 0.662 sec
		2385	POE/select, raw sockets 0.226 sec
		2386	POE/select, optimized 0.404 sec
		2387
		2388	AnyEvent/select/nb 0.085 sec
		2389	AnyEvent/EV/nb 0.068 sec
		2390	+state machine 0.134 sec
		2391
		2392	The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
		2393	benchmarks actually make blocking connects and use 100% blocking I/O,
		2394	defeating the purpose of an event-based solution. All of the newly
		2395	written AnyEvent benchmarks use 100% non-blocking connects (using
		2396	AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
		2397	resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
		2398	generally require a lot more bookkeeping and event handling than blocking
		2399	connects (which involve a single syscall only).
		2400
		2401	The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
		2402	offers similar expressive power as POE and IO::Lambda, using conventional
		2403	Perl syntax. This means that both the echo server and the client are 100%
		2404	non-blocking, further placing it at a disadvantage.
		2405
		2406	As you can see, the AnyEvent + EV combination even beats the
		2407	hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
		2408	backend easily beats IO::Lambda and POE.
		2409
		2410	And even the 100% non-blocking version written using the high-level (and
		2411	slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda
		2412	higher level ("unoptimised") abstractions by a large margin, even though
		2413	it does all of DNS, tcp-connect and socket I/O in a non-blocking way.
		2414
		2415	The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
		2416	F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
		2417	part of the IO::Lambda distribution and were used without any changes.
		2418
		2419
		2420	=head1 SIGNALS
		2421
		2422	AnyEvent currently installs handlers for these signals:
		2423
		2424	=over 4
		2425
		2426	=item SIGCHLD
		2427
		2428	A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
		2429	emulation for event loops that do not support them natively. Also, some
		2430	event loops install a similar handler.
		2431
		2432	Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE, then
		2433	AnyEvent will reset it to default, to avoid losing child exit statuses.
		2434
		2435	=item SIGPIPE
		2436
		2437	A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
		2438	when AnyEvent gets loaded.
		2439
		2440	The rationale for this is that AnyEvent users usually do not really depend
		2441	on SIGPIPE delivery (which is purely an optimisation for shell use, or
		2442	badly-written programs), but C<SIGPIPE> can cause spurious and rare
		2443	program exits as a lot of people do not expect C<SIGPIPE> when writing to
		2444	some random socket.
		2445
		2446	The rationale for installing a no-op handler as opposed to ignoring it is
		2447	that this way, the handler will be restored to defaults on exec.
		2448
		2449	Feel free to install your own handler, or reset it to defaults.
		2450
		2451	=back
		2452
49	=cut	2453	=cut
50		2454
51	package AnyEvent;	2455	undef $SIG{CHLD}
		2456	if $SIG{CHLD} eq 'IGNORE';
52		2457
53	no warnings;	2458	$SIG{PIPE} = sub { }
54	use strict 'vars';	2459	unless defined $SIG{PIPE};
55	use Carp;
56		2460
57	our $VERSION = 0.1;	2461	=head1 RECOMMENDED/OPTIONAL MODULES
58	our $MODEL;
59		2462
60	our $AUTOLOAD;	2463	One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
61	our @ISA;	2464	it's built-in modules) are required to use it.
62		2465
63	my @models = (	2466	That does not mean that AnyEvent won't take advantage of some additional
64	[Coro => Coro::Event::],	2467	modules if they are installed.
65	[Event => Event::],
66	[Glib => Glib::],
67	[Tk => Tk::],
68	);
69		2468
70	sub AUTOLOAD {	2469	This section explains which additional modules will be used, and how they
71	$AUTOLOAD =~ s/.*://;	2470	affect AnyEvent's operation.
72		2471
73	unless ($MODEL) {	2472	=over 4
74	# check for already loaded models
75	for (@models) {
76	my ($model, $package) = @$_;
77	if (scalar keys %{ *{"$package\::"} }) {
78	eval "require AnyEvent::Impl::$model"
79	or die;
80		2473
81	last if $MODEL;	2474	=item L<Async::Interrupt>
82	}
83	}
84		2475
85	unless ($MODEL) {	2476	This slightly arcane module is used to implement fast signal handling: To
86	# try to load a model	2477	my knowledge, there is no way to do completely race-free and quick
		2478	signal handling in pure perl. To ensure that signals still get
		2479	delivered, AnyEvent will start an interval timer to wake up perl (and
		2480	catch the signals) with some delay (default is 10 seconds, look for
		2481	C<$AnyEvent::MAX_SIGNAL_LATENCY>).
87		2482
88	for (@models) {	2483	If this module is available, then it will be used to implement signal
89	my ($model, $package) = @$_;	2484	catching, which means that signals will not be delayed, and the event loop
90	eval "require AnyEvent::Impl::$model"	2485	will not be interrupted regularly, which is more efficient (and good for
91	or die;	2486	battery life on laptops).
92		2487
93	last if $MODEL;	2488	This affects not just the pure-perl event loop, but also other event loops
94	}	2489	that have no signal handling on their own (e.g. Glib, Tk, Qt).
95		2490
96	$MODEL	2491	Some event loops (POE, Event, Event::Lib) offer signal watchers natively,
97	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: Coro, Event, Glib or Tk.";	2492	and either employ their own workarounds (POE) or use AnyEvent's workaround
98	}	2493	(using C<$AnyEvent::MAX_SIGNAL_LATENCY>). Installing L<Async::Interrupt>
99	}	2494	does nothing for those backends.
100		2495
101	@ISA = $MODEL;	2496	=item L<EV>
102		2497
103	my $class = shift;	2498	This module isn't really "optional", as it is simply one of the backend
104	$class->$AUTOLOAD (@_);	2499	event loops that AnyEvent can use. However, it is simply the best event
105	}	2500	loop available in terms of features, speed and stability: It supports
		2501	the AnyEvent API optimally, implements all the watcher types in XS, does
		2502	automatic timer adjustments even when no monotonic clock is available,
		2503	can take avdantage of advanced kernel interfaces such as C<epoll> and
		2504	C<kqueue>, and is the fastest backend I<by far>. You can even embed
		2505	L<Glib>/L<Gtk2> in it (or vice versa, see L<EV::Glib> and L<Glib::EV>).
		2506
		2507	=item L<Guard>
		2508
		2509	The guard module, when used, will be used to implement
		2510	C<AnyEvent::Util::guard>. This speeds up guards considerably (and uses a
		2511	lot less memory), but otherwise doesn't affect guard operation much. It is
		2512	purely used for performance.
		2513
		2514	=item L<JSON> and L<JSON::XS>
		2515
		2516	One of these modules is required when you want to read or write JSON data
		2517	via L<AnyEvent::Handle>. It is also written in pure-perl, but can take
		2518	advantage of the ultra-high-speed L<JSON::XS> module when it is installed.
		2519
		2520	In fact, L<AnyEvent::Handle> will use L<JSON::XS> by default if it is
		2521	installed.
		2522
		2523	=item L<Net::SSLeay>
		2524
		2525	Implementing TLS/SSL in Perl is certainly interesting, but not very
		2526	worthwhile: If this module is installed, then L<AnyEvent::Handle> (with
		2527	the help of L<AnyEvent::TLS>), gains the ability to do TLS/SSL.
		2528
		2529	=item L<Time::HiRes>
		2530
		2531	This module is part of perl since release 5.008. It will be used when the
		2532	chosen event library does not come with a timing source on it's own. The
		2533	pure-perl event loop (L<AnyEvent::Impl::Perl>) will additionally use it to
		2534	try to use a monotonic clock for timing stability.
106		2535
107	=back	2536	=back
108		2537
109	=head1 EXAMPLE
110		2538
111	The following program uses an io watcher to read data from stdin, a timer	2539	=head1 FORK
112	to display a message once per second, and a condvar to exit the program
113	when the user enters quit:
114		2540
		2541	Most event libraries are not fork-safe. The ones who are usually are
		2542	because they rely on inefficient but fork-safe C<select> or C<poll> calls
		2543	- higher performance APIs such as BSD's kqueue or the dreaded Linux epoll
		2544	are usually badly thought-out hacks that are incompatible with fork in
		2545	one way or another. Only L<EV> is fully fork-aware and ensures that you
		2546	continue event-processing in both parent and child (or both, if you know
		2547	what you are doing).
		2548
		2549	This means that, in general, you cannot fork and do event processing in
		2550	the child if the event library was initialised before the fork (which
		2551	usually happens when the first AnyEvent watcher is created, or the library
		2552	is loaded).
		2553
		2554	If you have to fork, you must either do so I<before> creating your first
		2555	watcher OR you must not use AnyEvent at all in the child OR you must do
		2556	something completely out of the scope of AnyEvent.
		2557
		2558	The problem of doing event processing in the parent I<and> the child
		2559	is much more complicated: even for backends that I<are> fork-aware or
		2560	fork-safe, their behaviour is not usually what you want: fork clones all
		2561	watchers, that means all timers, I/O watchers etc. are active in both
		2562	parent and child, which is almost never what you want. USing C<exec>
		2563	to start worker children from some kind of manage rprocess is usually
		2564	preferred, because it is much easier and cleaner, at the expense of having
		2565	to have another binary.
		2566
		2567
		2568	=head1 SECURITY CONSIDERATIONS
		2569
		2570	AnyEvent can be forced to load any event model via
		2571	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		2572	execute arbitrary code or directly gain access, it can easily be used to
		2573	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		2574	will not be active when the program uses a different event model than
		2575	specified in the variable.
		2576
		2577	You can make AnyEvent completely ignore this variable by deleting it
		2578	before the first watcher gets created, e.g. with a C<BEGIN> block:
		2579
		2580	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		2581
115	use AnyEvent;	2582	use AnyEvent;
116		2583
117	my $cv = AnyEvent->condvar;	2584	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		2585	be used to probe what backend is used and gain other information (which is
		2586	probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
		2587	$ENV{PERL_ANYEVENT_STRICT}.
118		2588
119	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	2589	Note that AnyEvent will remove I<all> environment variables starting with
120	warn "io event <$_[0]>\n"; # will always output <r>	2590	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
121	chomp (my $input = <STDIN>); # read a line	2591	enabled.
122	warn "read: $input\n"; # output what has been read
123	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i
124	});
125		2592
126	my $time_watcher; # can only be used once
127		2593
128	sub new_timer {	2594	=head1 BUGS
129	$timer = AnyEvent->timer (after => 1, cb => sub {
130	warn "timeout\n"; # print 'timeout' about every second
131	&new_timer; # and restart the time
132	});
133	}
134		2595
135	new_timer; # create first timer	2596	Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
		2597	to work around. If you suffer from memleaks, first upgrade to Perl 5.10
		2598	and check wether the leaks still show up. (Perl 5.10.0 has other annoying
		2599	memleaks, such as leaking on C<map> and C<grep> but it is usually not as
		2600	pronounced).
136		2601
137	$cv->wait; # wait until user enters /^q/i
138		2602
139	=head1 SEE ALSO	2603	=head1 SEE ALSO
140		2604
141	L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>,	2605	Utility functions: L<AnyEvent::Util>.
142	L<AnyEvent::Impl::Coro>,	2606
143	L<AnyEvent::Impl::Event>,	2607	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
144	L<AnyEvent::Impl::Glib>,	2608	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
		2609
		2610	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		2611	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		2612	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		2613	L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>, L<Anyevent::Impl::Irssi>.
		2614
		2615	Non-blocking file handles, sockets, TCP clients and
		2616	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>.
		2617
		2618	Asynchronous DNS: L<AnyEvent::DNS>.
		2619
		2620	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>,
		2621	L<Coro::Event>,
		2622
		2623	Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::XMPP>,
145	L<AnyEvent::Impl::Tk>.	2624	L<AnyEvent::HTTP>.
146		2625
147	=head1	2626
		2627	=head1 AUTHOR
		2628
		2629	Marc Lehmann <schmorp@schmorp.de>
		2630	http://home.schmorp.de/
148		2631
149	=cut	2632	=cut
150		2633
151	1	2634	1
152		2635

Diff Legend

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