ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.10 by root, Fri Jan 13 13:16:15 2006 UTC vs.
Revision 1.216 by root, Tue Jun 23 12:21:34 2009 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent - provide framework for multiple event loops	3	AnyEvent - provide framework for multiple event loops
4		4
5	Event, Coro, Glib, Tk - various supported event loops	5	EV, Event, Glib, Tk, Perl, Event::Lib, Qt and POE are various supported
		6	event loops.
6		7
7	=head1 SYNOPSIS	8	=head1 SYNOPSIS
8		9
9	use AnyEvent;	10	use AnyEvent;
10		11
		12	# file descriptor readable
11	my $w = AnyEvent->io (fh => ..., 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	});
15		30
16	* only one io watcher per $fh and $poll type is allowed (i.e. on a socket	31	# called when event loop idle (if applicable)
17	you can have one r + one w or one rw watcher, not any more (limitation by	32	my $w = AnyEvent->idle (cb => sub { ... });
18	Tk).
19		33
20	* the C<$poll_got> passed to the handler needs to be checked by looking	34	my $w = AnyEvent->condvar; # stores whether a condition was flagged
21	for single characters (e.g. with a regex), as it can contain more event	35	$w->send; # wake up current and all future recv's
22	types than were requested (e.g. a 'w' watcher might generate 'rw' events,
23	limitation by Glib).
24
25	* AnyEvent will keep filehandles alive, so as long as the watcher exists,
26	the filehandle exists.
27
28	my $w = AnyEvent->timer (after => $seconds, cb => sub {
29	...
30	});
31
32	* io and time watchers get canceled whenever $w is destroyed, so keep a copy
33
34	* timers can only be used once and must be recreated for repeated
35	operation (limitation by Glib and Tk).
36
37	my $w = AnyEvent->condvar; # kind of main loop replacement
38	$w->wait; # enters main loop till $condvar gets ->broadcast	36	$w->recv; # enters "main loop" till $condvar gets ->send
39	$w->broadcast; # wake up current and all future wait's	37	# use a condvar in callback mode:
		38	$w->cb (sub { $_[0]->recv });
40		39
41	* condvars are used to give blocking behaviour when neccessary. Create	40	=head1 INTRODUCTION/TUTORIAL
42	a condvar for any "request" or "event" your module might create, C<<	41
43	->broadcast >> it when the event happens and provide a function that calls	42	This manpage is mainly a reference manual. If you are interested
44	C<< ->wait >> for it. See the examples below.	43	in a tutorial or some gentle introduction, have a look at the
		44	L<AnyEvent::Intro> manpage.
		45
		46	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
		47
		48	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
		49	nowadays. So what is different about AnyEvent?
		50
		51	Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
		52	policy> and AnyEvent is I<small and efficient>.
		53
		54	First and foremost, I<AnyEvent is not an event model> itself, it only
		55	interfaces to whatever event model the main program happens to use, in a
		56	pragmatic way. For event models and certain classes of immortals alike,
		57	the statement "there can only be one" is a bitter reality: In general,
		58	only one event loop can be active at the same time in a process. AnyEvent
		59	cannot change this, but it can hide the differences between those event
		60	loops.
		61
		62	The goal of AnyEvent is to offer module authors the ability to do event
		63	programming (waiting for I/O or timer events) without subscribing to a
		64	religion, a way of living, and most importantly: without forcing your
		65	module users into the same thing by forcing them to use the same event
		66	model you use.
		67
		68	For modules like POE or IO::Async (which is a total misnomer as it is
		69	actually doing all I/O I<synchronously>...), using them in your module is
		70	like joining a cult: After you joined, you are dependent on them and you
		71	cannot use anything else, as they are simply incompatible to everything
		72	that isn't them. What's worse, all the potential users of your
		73	module are I<also> forced to use the same event loop you use.
		74
		75	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
		76	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
		77	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
		78	your module uses one of those, every user of your module has to use it,
		79	too. But if your module uses AnyEvent, it works transparently with all
		80	event models it supports (including stuff like IO::Async, as long as those
		81	use one of the supported event loops. It is trivial to add new event loops
		82	to AnyEvent, too, so it is future-proof).
		83
		84	In addition to being free of having to use I<the one and only true event
		85	model>, AnyEvent also is free of bloat and policy: with POE or similar
		86	modules, you get an enormous amount of code and strict rules you have to
		87	follow. AnyEvent, on the other hand, is lean and up to the point, by only
		88	offering the functionality that is necessary, in as thin as a wrapper as
		89	technically possible.
		90
		91	Of course, AnyEvent comes with a big (and fully optional!) toolbox
		92	of useful functionality, such as an asynchronous DNS resolver, 100%
		93	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		94	such as Windows) and lots of real-world knowledge and workarounds for
		95	platform bugs and differences.
		96
		97	Now, if you I<do want> lots of policy (this can arguably be somewhat
		98	useful) and you want to force your users to use the one and only event
		99	model, you should I<not> use this module.
45		100
46	=head1 DESCRIPTION	101	=head1 DESCRIPTION
47		102
48	L<AnyEvent> provides an identical interface to multiple event loops. This	103	L<AnyEvent> provides an identical interface to multiple event loops. This
49	allows module authors to utilizy an event loop without forcing module	104	allows module authors to utilise an event loop without forcing module
50	users to use the same event loop (as only a single event loop can coexist	105	users to use the same event loop (as only a single event loop can coexist
51	peacefully at any one time).	106	peacefully at any one time).
52		107
53	The interface itself is vaguely similar but not identical to the Event	108	The interface itself is vaguely similar, but not identical to the L<Event>
54	module.	109	module.
55		110
56	On the first call of any method, the module tries to detect the currently	111	During the first call of any watcher-creation method, the module tries
57	loaded event loop by probing wether any of the following modules is	112	to detect the currently loaded event loop by probing whether one of the
58	loaded: L<Coro::Event>, L<Event>, L<Glib>, L<Tk>. The first one found is	113	following modules is already loaded: L<EV>,
59	used. If none is found, the module tries to load these modules in the	114	L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>,
60	order given. The first one that could be successfully loaded will be	115	L<POE>. The first one found is used. If none are found, the module tries
61	used. If still none could be found, it will issue an error.	116	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
		117	adaptor should always succeed) in the order given. The first one that can
		118	be successfully loaded will be used. If, after this, still none could be
		119	found, AnyEvent will fall back to a pure-perl event loop, which is not
		120	very efficient, but should work everywhere.
		121
		122	Because AnyEvent first checks for modules that are already loaded, loading
		123	an event model explicitly before first using AnyEvent will likely make
		124	that model the default. For example:
		125
		126	use Tk;
		127	use AnyEvent;
		128
		129	# .. AnyEvent will likely default to Tk
		130
		131	The I<likely> means that, if any module loads another event model and
		132	starts using it, all bets are off. Maybe you should tell their authors to
		133	use AnyEvent so their modules work together with others seamlessly...
		134
		135	The pure-perl implementation of AnyEvent is called
		136	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
		137	explicitly and enjoy the high availability of that event loop :)
		138
		139	=head1 WATCHERS
		140
		141	AnyEvent has the central concept of a I<watcher>, which is an object that
		142	stores relevant data for each kind of event you are waiting for, such as
		143	the callback to call, the file handle to watch, etc.
		144
		145	These watchers are normal Perl objects with normal Perl lifetime. After
		146	creating a watcher it will immediately "watch" for events and invoke the
		147	callback when the event occurs (of course, only when the event model
		148	is in control).
		149
		150	Note that B<callbacks must not permanently change global variables>
		151	potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
		152	callbacks must not C<die> >>. The former is good programming practise in
		153	Perl and the latter stems from the fact that exception handling differs
		154	widely between event loops.
		155
		156	To disable the watcher you have to destroy it (e.g. by setting the
		157	variable you store it in to C<undef> or otherwise deleting all references
		158	to it).
		159
		160	All watchers are created by calling a method on the C<AnyEvent> class.
		161
		162	Many watchers either are used with "recursion" (repeating timers for
		163	example), or need to refer to their watcher object in other ways.
		164
		165	An any way to achieve that is this pattern:
		166
		167	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
		168	# you can use $w here, for example to undef it
		169	undef $w;
		170	});
		171
		172	Note that C<my $w; $w => combination. This is necessary because in Perl,
		173	my variables are only visible after the statement in which they are
		174	declared.
		175
		176	=head2 I/O WATCHERS
		177
		178	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
		179	with the following mandatory key-value pairs as arguments:
		180
		181	C<fh> is the Perl I<file handle> (I<not> file descriptor) to watch
		182	for events (AnyEvent might or might not keep a reference to this file
		183	handle). Note that only file handles pointing to things for which
		184	non-blocking operation makes sense are allowed. This includes sockets,
		185	most character devices, pipes, fifos and so on, but not for example files
		186	or block devices.
		187
		188	C<poll> must be a string that is either C<r> or C<w>, which creates a
		189	watcher waiting for "r"eadable or "w"ritable events, respectively.
		190
		191	C<cb> is the callback to invoke each time the file handle becomes ready.
		192
		193	Although the callback might get passed parameters, their value and
		194	presence is undefined and you cannot rely on them. Portable AnyEvent
		195	callbacks cannot use arguments passed to I/O watcher callbacks.
		196
		197	The I/O watcher might use the underlying file descriptor or a copy of it.
		198	You must not close a file handle as long as any watcher is active on the
		199	underlying file descriptor.
		200
		201	Some event loops issue spurious readyness notifications, so you should
		202	always use non-blocking calls when reading/writing from/to your file
		203	handles.
		204
		205	Example: wait for readability of STDIN, then read a line and disable the
		206	watcher.
		207
		208	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		209	chomp (my $input = <STDIN>);
		210	warn "read: $input\n";
		211	undef $w;
		212	});
		213
		214	=head2 TIME WATCHERS
		215
		216	You can create a time watcher by calling the C<< AnyEvent->timer >>
		217	method with the following mandatory arguments:
		218
		219	C<after> specifies after how many seconds (fractional values are
		220	supported) the callback should be invoked. C<cb> is the callback to invoke
		221	in that case.
		222
		223	Although the callback might get passed parameters, their value and
		224	presence is undefined and you cannot rely on them. Portable AnyEvent
		225	callbacks cannot use arguments passed to time watcher callbacks.
		226
		227	The callback will normally be invoked once only. If you specify another
		228	parameter, C<interval>, as a strictly positive number (> 0), then the
		229	callback will be invoked regularly at that interval (in fractional
		230	seconds) after the first invocation. If C<interval> is specified with a
		231	false value, then it is treated as if it were missing.
		232
		233	The callback will be rescheduled before invoking the callback, but no
		234	attempt is done to avoid timer drift in most backends, so the interval is
		235	only approximate.
		236
		237	Example: fire an event after 7.7 seconds.
		238
		239	my $w = AnyEvent->timer (after => 7.7, cb => sub {
		240	warn "timeout\n";
		241	});
		242
		243	# to cancel the timer:
		244	undef $w;
		245
		246	Example 2: fire an event after 0.5 seconds, then roughly every second.
		247
		248	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
		249	warn "timeout\n";
		250	};
		251
		252	=head3 TIMING ISSUES
		253
		254	There are two ways to handle timers: based on real time (relative, "fire
		255	in 10 seconds") and based on wallclock time (absolute, "fire at 12
		256	o'clock").
		257
		258	While most event loops expect timers to specified in a relative way, they
		259	use absolute time internally. This makes a difference when your clock
		260	"jumps", for example, when ntp decides to set your clock backwards from
		261	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		262	fire "after" a second might actually take six years to finally fire.
		263
		264	AnyEvent cannot compensate for this. The only event loop that is conscious
		265	about these issues is L<EV>, which offers both relative (ev_timer, based
		266	on true relative time) and absolute (ev_periodic, based on wallclock time)
		267	timers.
		268
		269	AnyEvent always prefers relative timers, if available, matching the
		270	AnyEvent API.
		271
		272	AnyEvent has two additional methods that return the "current time":
62		273
63	=over 4	274	=over 4
64		275
		276	=item AnyEvent->time
		277
		278	This returns the "current wallclock time" as a fractional number of
		279	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		280	return, and the result is guaranteed to be compatible with those).
		281
		282	It progresses independently of any event loop processing, i.e. each call
		283	will check the system clock, which usually gets updated frequently.
		284
		285	=item AnyEvent->now
		286
		287	This also returns the "current wallclock time", but unlike C<time>, above,
		288	this value might change only once per event loop iteration, depending on
		289	the event loop (most return the same time as C<time>, above). This is the
		290	time that AnyEvent's timers get scheduled against.
		291
		292	I<In almost all cases (in all cases if you don't care), this is the
		293	function to call when you want to know the current time.>
		294
		295	This function is also often faster then C<< AnyEvent->time >>, and
		296	thus the preferred method if you want some timestamp (for example,
		297	L<AnyEvent::Handle> uses this to update it's activity timeouts).
		298
		299	The rest of this section is only of relevance if you try to be very exact
		300	with your timing, you can skip it without bad conscience.
		301
		302	For a practical example of when these times differ, consider L<Event::Lib>
		303	and L<EV> and the following set-up:
		304
		305	The event loop is running and has just invoked one of your callback at
		306	time=500 (assume no other callbacks delay processing). In your callback,
		307	you wait a second by executing C<sleep 1> (blocking the process for a
		308	second) and then (at time=501) you create a relative timer that fires
		309	after three seconds.
		310
		311	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		312	both return C<501>, because that is the current time, and the timer will
		313	be scheduled to fire at time=504 (C<501> + C<3>).
		314
		315	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		316	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		317	last event processing phase started. With L<EV>, your timer gets scheduled
		318	to run at time=503 (C<500> + C<3>).
		319
		320	In one sense, L<Event::Lib> is more exact, as it uses the current time
		321	regardless of any delays introduced by event processing. However, most
		322	callbacks do not expect large delays in processing, so this causes a
		323	higher drift (and a lot more system calls to get the current time).
		324
		325	In another sense, L<EV> is more exact, as your timer will be scheduled at
		326	the same time, regardless of how long event processing actually took.
		327
		328	In either case, if you care (and in most cases, you don't), then you
		329	can get whatever behaviour you want with any event loop, by taking the
		330	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		331	account.
		332
		333	=item AnyEvent->now_update
		334
		335	Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
		336	the current time for each loop iteration (see the discussion of L<<
		337	AnyEvent->now >>, above).
		338
		339	When a callback runs for a long time (or when the process sleeps), then
		340	this "current" time will differ substantially from the real time, which
		341	might affect timers and time-outs.
		342
		343	When this is the case, you can call this method, which will update the
		344	event loop's idea of "current time".
		345
		346	Note that updating the time I<might> cause some events to be handled.
		347
		348	=back
		349
		350	=head2 SIGNAL WATCHERS
		351
		352	You can watch for signals using a signal watcher, C<signal> is the signal
		353	I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
		354	callback to be invoked whenever a signal occurs.
		355
		356	Although the callback might get passed parameters, their value and
		357	presence is undefined and you cannot rely on them. Portable AnyEvent
		358	callbacks cannot use arguments passed to signal watcher callbacks.
		359
		360	Multiple signal occurrences can be clumped together into one callback
		361	invocation, and callback invocation will be synchronous. Synchronous means
		362	that it might take a while until the signal gets handled by the process,
		363	but it is guaranteed not to interrupt any other callbacks.
		364
		365	The main advantage of using these watchers is that you can share a signal
		366	between multiple watchers.
		367
		368	This watcher might use C<%SIG>, so programs overwriting those signals
		369	directly will likely not work correctly.
		370
		371	Example: exit on SIGINT
		372
		373	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		374
		375	=head2 CHILD PROCESS WATCHERS
		376
		377	You can also watch on a child process exit and catch its exit status.
		378
		379	The child process is specified by the C<pid> argument (if set to C<0>, it
		380	watches for any child process exit). The watcher will triggered only when
		381	the child process has finished and an exit status is available, not on
		382	any trace events (stopped/continued).
		383
		384	The callback will be called with the pid and exit status (as returned by
		385	waitpid), so unlike other watcher types, you I<can> rely on child watcher
		386	callback arguments.
		387
		388	This watcher type works by installing a signal handler for C<SIGCHLD>,
		389	and since it cannot be shared, nothing else should use SIGCHLD or reap
		390	random child processes (waiting for specific child processes, e.g. inside
		391	C<system>, is just fine).
		392
		393	There is a slight catch to child watchers, however: you usually start them
		394	I<after> the child process was created, and this means the process could
		395	have exited already (and no SIGCHLD will be sent anymore).
		396
		397	Not all event models handle this correctly (POE doesn't), but even for
		398	event models that I<do> handle this correctly, they usually need to be
		399	loaded before the process exits (i.e. before you fork in the first place).
		400
		401	This means you cannot create a child watcher as the very first thing in an
		402	AnyEvent program, you I<have> to create at least one watcher before you
		403	C<fork> the child (alternatively, you can call C<AnyEvent::detect>).
		404
		405	Example: fork a process and wait for it
		406
		407	my $done = AnyEvent->condvar;
		408
		409	my $pid = fork or exit 5;
		410
		411	my $w = AnyEvent->child (
		412	pid => $pid,
		413	cb => sub {
		414	my ($pid, $status) = @_;
		415	warn "pid $pid exited with status $status";
		416	$done->send;
		417	},
		418	);
		419
		420	# do something else, then wait for process exit
		421	$done->recv;
		422
		423	=head2 IDLE WATCHERS
		424
		425	Sometimes there is a need to do something, but it is not so important
		426	to do it instantly, but only when there is nothing better to do. This
		427	"nothing better to do" is usually defined to be "no other events need
		428	attention by the event loop".
		429
		430	Idle watchers ideally get invoked when the event loop has nothing
		431	better to do, just before it would block the process to wait for new
		432	events. Instead of blocking, the idle watcher is invoked.
		433
		434	Most event loops unfortunately do not really support idle watchers (only
		435	EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
		436	will simply call the callback "from time to time".
		437
		438	Example: read lines from STDIN, but only process them when the
		439	program is otherwise idle:
		440
		441	my @lines; # read data
		442	my $idle_w;
		443	my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		444	push @lines, scalar <STDIN>;
		445
		446	# start an idle watcher, if not already done
		447	$idle_w ||= AnyEvent->idle (cb => sub {
		448	# handle only one line, when there are lines left
		449	if (my $line = shift @lines) {
		450	print "handled when idle: $line";
		451	} else {
		452	# otherwise disable the idle watcher again
		453	undef $idle_w;
		454	}
		455	});
		456	});
		457
		458	=head2 CONDITION VARIABLES
		459
		460	If you are familiar with some event loops you will know that all of them
		461	require you to run some blocking "loop", "run" or similar function that
		462	will actively watch for new events and call your callbacks.
		463
		464	AnyEvent is different, it expects somebody else to run the event loop and
		465	will only block when necessary (usually when told by the user).
		466
		467	The instrument to do that is called a "condition variable", so called
		468	because they represent a condition that must become true.
		469
		470	Condition variables can be created by calling the C<< AnyEvent->condvar
		471	>> method, usually without arguments. The only argument pair allowed is
		472
		473	C<cb>, which specifies a callback to be called when the condition variable
		474	becomes true, with the condition variable as the first argument (but not
		475	the results).
		476
		477	After creation, the condition variable is "false" until it becomes "true"
		478	by calling the C<send> method (or calling the condition variable as if it
		479	were a callback, read about the caveats in the description for the C<<
		480	->send >> method).
		481
		482	Condition variables are similar to callbacks, except that you can
		483	optionally wait for them. They can also be called merge points - points
		484	in time where multiple outstanding events have been processed. And yet
		485	another way to call them is transactions - each condition variable can be
		486	used to represent a transaction, which finishes at some point and delivers
		487	a result.
		488
		489	Condition variables are very useful to signal that something has finished,
		490	for example, if you write a module that does asynchronous http requests,
		491	then a condition variable would be the ideal candidate to signal the
		492	availability of results. The user can either act when the callback is
		493	called or can synchronously C<< ->recv >> for the results.
		494
		495	You can also use them to simulate traditional event loops - for example,
		496	you can block your main program until an event occurs - for example, you
		497	could C<< ->recv >> in your main program until the user clicks the Quit
		498	button of your app, which would C<< ->send >> the "quit" event.
		499
		500	Note that condition variables recurse into the event loop - if you have
		501	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		502	lose. Therefore, condition variables are good to export to your caller, but
		503	you should avoid making a blocking wait yourself, at least in callbacks,
		504	as this asks for trouble.
		505
		506	Condition variables are represented by hash refs in perl, and the keys
		507	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		508	easy (it is often useful to build your own transaction class on top of
		509	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		510	it's C<new> method in your own C<new> method.
		511
		512	There are two "sides" to a condition variable - the "producer side" which
		513	eventually calls C<< -> send >>, and the "consumer side", which waits
		514	for the send to occur.
		515
		516	Example: wait for a timer.
		517
		518	# wait till the result is ready
		519	my $result_ready = AnyEvent->condvar;
		520
		521	# do something such as adding a timer
		522	# or socket watcher the calls $result_ready->send
		523	# when the "result" is ready.
		524	# in this case, we simply use a timer:
		525	my $w = AnyEvent->timer (
		526	after => 1,
		527	cb => sub { $result_ready->send },
		528	);
		529
		530	# this "blocks" (while handling events) till the callback
		531	# calls send
		532	$result_ready->recv;
		533
		534	Example: wait for a timer, but take advantage of the fact that
		535	condition variables are also code references.
		536
		537	my $done = AnyEvent->condvar;
		538	my $delay = AnyEvent->timer (after => 5, cb => $done);
		539	$done->recv;
		540
		541	Example: Imagine an API that returns a condvar and doesn't support
		542	callbacks. This is how you make a synchronous call, for example from
		543	the main program:
		544
		545	use AnyEvent::CouchDB;
		546
		547	...
		548
		549	my @info = $couchdb->info->recv;
		550
		551	And this is how you would just ste a callback to be called whenever the
		552	results are available:
		553
		554	$couchdb->info->cb (sub {
		555	my @info = $_[0]->recv;
		556	});
		557
		558	=head3 METHODS FOR PRODUCERS
		559
		560	These methods should only be used by the producing side, i.e. the
		561	code/module that eventually sends the signal. Note that it is also
		562	the producer side which creates the condvar in most cases, but it isn't
		563	uncommon for the consumer to create it as well.
		564
		565	=over 4
		566
		567	=item $cv->send (...)
		568
		569	Flag the condition as ready - a running C<< ->recv >> and all further
		570	calls to C<recv> will (eventually) return after this method has been
		571	called. If nobody is waiting the send will be remembered.
		572
		573	If a callback has been set on the condition variable, it is called
		574	immediately from within send.
		575
		576	Any arguments passed to the C<send> call will be returned by all
		577	future C<< ->recv >> calls.
		578
		579	Condition variables are overloaded so one can call them directly
		580	(as a code reference). Calling them directly is the same as calling
		581	C<send>. Note, however, that many C-based event loops do not handle
		582	overloading, so as tempting as it may be, passing a condition variable
		583	instead of a callback does not work. Both the pure perl and EV loops
		584	support overloading, however, as well as all functions that use perl to
		585	invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
		586	example).
		587
		588	=item $cv->croak ($error)
		589
		590	Similar to send, but causes all call's to C<< ->recv >> to invoke
		591	C<Carp::croak> with the given error message/object/scalar.
		592
		593	This can be used to signal any errors to the condition variable
		594	user/consumer.
		595
		596	=item $cv->begin ([group callback])
		597
		598	=item $cv->end
		599
		600	These two methods are EXPERIMENTAL and MIGHT CHANGE.
		601
		602	These two methods can be used to combine many transactions/events into
		603	one. For example, a function that pings many hosts in parallel might want
		604	to use a condition variable for the whole process.
		605
		606	Every call to C<< ->begin >> will increment a counter, and every call to
		607	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		608	>>, the (last) callback passed to C<begin> will be executed. That callback
		609	is I<supposed> to call C<< ->send >>, but that is not required. If no
		610	callback was set, C<send> will be called without any arguments.
		611
		612	Let's clarify this with the ping example:
		613
		614	my $cv = AnyEvent->condvar;
		615
		616	my %result;
		617	$cv->begin (sub { $cv->send (\%result) });
		618
		619	for my $host (@list_of_hosts) {
		620	$cv->begin;
		621	ping_host_then_call_callback $host, sub {
		622	$result{$host} = ...;
		623	$cv->end;
		624	};
		625	}
		626
		627	$cv->end;
		628
		629	This code fragment supposedly pings a number of hosts and calls
		630	C<send> after results for all then have have been gathered - in any
		631	order. To achieve this, the code issues a call to C<begin> when it starts
		632	each ping request and calls C<end> when it has received some result for
		633	it. Since C<begin> and C<end> only maintain a counter, the order in which
		634	results arrive is not relevant.
		635
		636	There is an additional bracketing call to C<begin> and C<end> outside the
		637	loop, which serves two important purposes: first, it sets the callback
		638	to be called once the counter reaches C<0>, and second, it ensures that
		639	C<send> is called even when C<no> hosts are being pinged (the loop
		640	doesn't execute once).
		641
		642	This is the general pattern when you "fan out" into multiple subrequests:
		643	use an outer C<begin>/C<end> pair to set the callback and ensure C<end>
		644	is called at least once, and then, for each subrequest you start, call
		645	C<begin> and for each subrequest you finish, call C<end>.
		646
		647	=back
		648
		649	=head3 METHODS FOR CONSUMERS
		650
		651	These methods should only be used by the consuming side, i.e. the
		652	code awaits the condition.
		653
		654	=over 4
		655
		656	=item $cv->recv
		657
		658	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		659	>> methods have been called on c<$cv>, while servicing other watchers
		660	normally.
		661
		662	You can only wait once on a condition - additional calls are valid but
		663	will return immediately.
		664
		665	If an error condition has been set by calling C<< ->croak >>, then this
		666	function will call C<croak>.
		667
		668	In list context, all parameters passed to C<send> will be returned,
		669	in scalar context only the first one will be returned.
		670
		671	Not all event models support a blocking wait - some die in that case
		672	(programs might want to do that to stay interactive), so I<if you are
		673	using this from a module, never require a blocking wait>, but let the
		674	caller decide whether the call will block or not (for example, by coupling
		675	condition variables with some kind of request results and supporting
		676	callbacks so the caller knows that getting the result will not block,
		677	while still supporting blocking waits if the caller so desires).
		678
		679	Another reason I<never> to C<< ->recv >> in a module is that you cannot
		680	sensibly have two C<< ->recv >>'s in parallel, as that would require
		681	multiple interpreters or coroutines/threads, none of which C<AnyEvent>
		682	can supply.
		683
		684	The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
		685	fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
		686	versions and also integrates coroutines into AnyEvent, making blocking
		687	C<< ->recv >> calls perfectly safe as long as they are done from another
		688	coroutine (one that doesn't run the event loop).
		689
		690	You can ensure that C<< -recv >> never blocks by setting a callback and
		691	only calling C<< ->recv >> from within that callback (or at a later
		692	time). This will work even when the event loop does not support blocking
		693	waits otherwise.
		694
		695	=item $bool = $cv->ready
		696
		697	Returns true when the condition is "true", i.e. whether C<send> or
		698	C<croak> have been called.
		699
		700	=item $cb = $cv->cb ($cb->($cv))
		701
		702	This is a mutator function that returns the callback set and optionally
		703	replaces it before doing so.
		704
		705	The callback will be called when the condition becomes "true", i.e. when
		706	C<send> or C<croak> are called, with the only argument being the condition
		707	variable itself. Calling C<recv> inside the callback or at any later time
		708	is guaranteed not to block.
		709
		710	=back
		711
		712	=head1 GLOBAL VARIABLES AND FUNCTIONS
		713
		714	=over 4
		715
		716	=item $AnyEvent::MODEL
		717
		718	Contains C<undef> until the first watcher is being created. Then it
		719	contains the event model that is being used, which is the name of the
		720	Perl class implementing the model. This class is usually one of the
		721	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case
		722	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).
		723
		724	The known classes so far are:
		725
		726	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
		727	AnyEvent::Impl::Event based on Event, second best choice.
		728	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		729	AnyEvent::Impl::Glib based on Glib, third-best choice.
		730	AnyEvent::Impl::Tk based on Tk, very bad choice.
		731	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
		732	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		733	AnyEvent::Impl::POE based on POE, not generic enough for full support.
		734
		735	There is no support for WxWidgets, as WxWidgets has no support for
		736	watching file handles. However, you can use WxWidgets through the
		737	POE Adaptor, as POE has a Wx backend that simply polls 20 times per
		738	second, which was considered to be too horrible to even consider for
		739	AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
		740	it's adaptor.
		741
		742	AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
		743	autodetecting them.
		744
		745	=item AnyEvent::detect
		746
		747	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
		748	if necessary. You should only call this function right before you would
		749	have created an AnyEvent watcher anyway, that is, as late as possible at
		750	runtime.
		751
		752	=item $guard = AnyEvent::post_detect { BLOCK }
		753
		754	Arranges for the code block to be executed as soon as the event model is
		755	autodetected (or immediately if this has already happened).
		756
		757	If called in scalar or list context, then it creates and returns an object
		758	that automatically removes the callback again when it is destroyed. See
		759	L<Coro::BDB> for a case where this is useful.
		760
		761	=item @AnyEvent::post_detect
		762
		763	If there are any code references in this array (you can C<push> to it
		764	before or after loading AnyEvent), then they will called directly after
		765	the event loop has been chosen.
		766
		767	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		768	if it contains a true value then the event loop has already been detected,
		769	and the array will be ignored.
		770
		771	Best use C<AnyEvent::post_detect { BLOCK }> instead.
		772
		773	=back
		774
		775	=head1 WHAT TO DO IN A MODULE
		776
		777	As a module author, you should C<use AnyEvent> and call AnyEvent methods
		778	freely, but you should not load a specific event module or rely on it.
		779
		780	Be careful when you create watchers in the module body - AnyEvent will
		781	decide which event module to use as soon as the first method is called, so
		782	by calling AnyEvent in your module body you force the user of your module
		783	to load the event module first.
		784
		785	Never call C<< ->recv >> on a condition variable unless you I<know> that
		786	the C<< ->send >> method has been called on it already. This is
		787	because it will stall the whole program, and the whole point of using
		788	events is to stay interactive.
		789
		790	It is fine, however, to call C<< ->recv >> when the user of your module
		791	requests it (i.e. if you create a http request object ad have a method
		792	called C<results> that returns the results, it should call C<< ->recv >>
		793	freely, as the user of your module knows what she is doing. always).
		794
		795	=head1 WHAT TO DO IN THE MAIN PROGRAM
		796
		797	There will always be a single main program - the only place that should
		798	dictate which event model to use.
		799
		800	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
		801	do anything special (it does not need to be event-based) and let AnyEvent
		802	decide which implementation to chose if some module relies on it.
		803
		804	If the main program relies on a specific event model - for example, in
		805	Gtk2 programs you have to rely on the Glib module - you should load the
		806	event module before loading AnyEvent or any module that uses it: generally
		807	speaking, you should load it as early as possible. The reason is that
		808	modules might create watchers when they are loaded, and AnyEvent will
		809	decide on the event model to use as soon as it creates watchers, and it
		810	might chose the wrong one unless you load the correct one yourself.
		811
		812	You can chose to use a pure-perl implementation by loading the
		813	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
		814	everywhere, but letting AnyEvent chose the model is generally better.
		815
		816	=head2 MAINLOOP EMULATION
		817
		818	Sometimes (often for short test scripts, or even standalone programs who
		819	only want to use AnyEvent), you do not want to run a specific event loop.
		820
		821	In that case, you can use a condition variable like this:
		822
		823	AnyEvent->condvar->recv;
		824
		825	This has the effect of entering the event loop and looping forever.
		826
		827	Note that usually your program has some exit condition, in which case
		828	it is better to use the "traditional" approach of storing a condition
		829	variable somewhere, waiting for it, and sending it when the program should
		830	exit cleanly.
		831
		832
		833	=head1 OTHER MODULES
		834
		835	The following is a non-exhaustive list of additional modules that use
		836	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		837	in the same program. Some of the modules come with AnyEvent, some are
		838	available via CPAN.
		839
		840	=over 4
		841
		842	=item L<AnyEvent::Util>
		843
		844	Contains various utility functions that replace often-used but blocking
		845	functions such as C<inet_aton> by event-/callback-based versions.
		846
		847	=item L<AnyEvent::Socket>
		848
		849	Provides various utility functions for (internet protocol) sockets,
		850	addresses and name resolution. Also functions to create non-blocking tcp
		851	connections or tcp servers, with IPv6 and SRV record support and more.
		852
		853	=item L<AnyEvent::Handle>
		854
		855	Provide read and write buffers, manages watchers for reads and writes,
		856	supports raw and formatted I/O, I/O queued and fully transparent and
		857	non-blocking SSL/TLS.
		858
		859	=item L<AnyEvent::DNS>
		860
		861	Provides rich asynchronous DNS resolver capabilities.
		862
		863	=item L<AnyEvent::HTTP>
		864
		865	A simple-to-use HTTP library that is capable of making a lot of concurrent
		866	HTTP requests.
		867
		868	=item L<AnyEvent::HTTPD>
		869
		870	Provides a simple web application server framework.
		871
		872	=item L<AnyEvent::FastPing>
		873
		874	The fastest ping in the west.
		875
		876	=item L<AnyEvent::DBI>
		877
		878	Executes L<DBI> requests asynchronously in a proxy process.
		879
		880	=item L<AnyEvent::AIO>
		881
		882	Truly asynchronous I/O, should be in the toolbox of every event
		883	programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
		884	together.
		885
		886	=item L<AnyEvent::BDB>
		887
		888	Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
		889	L<BDB> and AnyEvent together.
		890
		891	=item L<AnyEvent::GPSD>
		892
		893	A non-blocking interface to gpsd, a daemon delivering GPS information.
		894
		895	=item L<AnyEvent::IGS>
		896
		897	A non-blocking interface to the Internet Go Server protocol (used by
		898	L<App::IGS>).
		899
		900	=item L<AnyEvent::IRC>
		901
		902	AnyEvent based IRC client module family (replacing the older Net::IRC3).
		903
		904	=item L<Net::XMPP2>
		905
		906	AnyEvent based XMPP (Jabber protocol) module family.
		907
		908	=item L<Net::FCP>
		909
		910	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		911	of AnyEvent.
		912
		913	=item L<Event::ExecFlow>
		914
		915	High level API for event-based execution flow control.
		916
		917	=item L<Coro>
		918
		919	Has special support for AnyEvent via L<Coro::AnyEvent>.
		920
		921	=item L<IO::Lambda>
		922
		923	The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
		924
		925	=back
		926
65	=cut	927	=cut
66		928
67	package AnyEvent;	929	package AnyEvent;
68		930
69	no warnings;	931	no warnings;
70	use strict 'vars';	932	use strict qw(vars subs);
		933
71	use Carp;	934	use Carp;
72		935
73	our $VERSION = '1.02';	936	our $VERSION = 4.411;
74	our $MODEL;	937	our $MODEL;
75		938
76	our $AUTOLOAD;	939	our $AUTOLOAD;
77	our @ISA;	940	our @ISA;
78		941
		942	our @REGISTRY;
		943
		944	our $WIN32;
		945
		946	BEGIN {
		947	eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }";
		948	eval "sub TAINT(){ " . (${^TAINT}*1) . " }";
		949
		950	delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
		951	if ${^TAINT};
		952	}
		953
79	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;	954	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
80		955
81	our @REGISTRY;	956	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		957
		958	{
		959	my $idx;
		960	$PROTOCOL{$_} = ++$idx
		961	for reverse split /\s*,\s*/,
		962	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		963	}
82		964
83	my @models = (	965	my @models = (
84	[Coro::Event:: => AnyEvent::Impl::Coro::],	966	[EV:: => AnyEvent::Impl::EV::],
85	[Event:: => AnyEvent::Impl::Event::],	967	[Event:: => AnyEvent::Impl::Event::],
86	[Glib:: => AnyEvent::Impl::Glib::],	968	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
		969	# everything below here will not be autoprobed
		970	# as the pureperl backend should work everywhere
		971	# and is usually faster
		972	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		973	[Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
		974	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		975	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		976	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
87	[Tk:: => AnyEvent::Impl::Tk::],	977	[Wx:: => AnyEvent::Impl::POE::],
		978	[Prima:: => AnyEvent::Impl::POE::],
88	);	979	);
89		980
90	our %method = map +($_ => 1), qw(io timer condvar broadcast wait cancel DESTROY);	981	our %method = map +($_ => 1),
		982	qw(io timer time now now_update signal child idle condvar one_event DESTROY);
91		983
92	sub AUTOLOAD {	984	our @post_detect;
93	$AUTOLOAD =~ s/.*://;
94		985
95	$method{$AUTOLOAD}	986	sub post_detect(&) {
96	or croak "$AUTOLOAD: not a valid method for AnyEvent objects";	987	my ($cb) = @_;
97		988
		989	if ($MODEL) {
		990	$cb->();
		991
		992	1
		993	} else {
		994	push @post_detect, $cb;
		995
		996	defined wantarray
		997	? bless \$cb, "AnyEvent::Util::postdetect"
		998	: ()
		999	}
		1000	}
		1001
		1002	sub AnyEvent::Util::postdetect::DESTROY {
		1003	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		1004	}
		1005
		1006	sub detect() {
98	unless ($MODEL) {	1007	unless ($MODEL) {
		1008	no strict 'refs';
		1009	local $SIG{__DIE__};
		1010
		1011	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
		1012	my $model = "AnyEvent::Impl::$1";
		1013	if (eval "require $model") {
		1014	$MODEL = $model;
		1015	warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1;
		1016	} else {
		1017	warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose;
		1018	}
		1019	}
		1020
99	# check for already loaded models	1021	# check for already loaded models
		1022	unless ($MODEL) {
100	for (@REGISTRY, @models) {	1023	for (@REGISTRY, @models) {
101	my ($package, $model) = @$_;	1024	my ($package, $model) = @$_;
102	if (${"$package\::VERSION"} > 0) {	1025	if (${"$package\::VERSION"} > 0) {
103	if (eval "require $model") {	1026	if (eval "require $model") {
104	$MODEL = $model;	1027	$MODEL = $model;
105	warn "AnyEvent: found model '$model', using it.\n" if $verbose > 1;	1028	warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1;
106	last;	1029	last;
		1030	}
107	}	1031	}
108	}	1032	}
		1033
		1034	unless ($MODEL) {
		1035	# try to load a model
		1036
		1037	for (@REGISTRY, @models) {
		1038	my ($package, $model) = @$_;
		1039	if (eval "require $package"
		1040	and ${"$package\::VERSION"} > 0
		1041	and eval "require $model") {
		1042	$MODEL = $model;
		1043	warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1;
		1044	last;
		1045	}
		1046	}
		1047
		1048	$MODEL
		1049	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
		1050	}
109	}	1051	}
110		1052
111	unless ($MODEL) {	1053	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
112	# try to load a model
113		1054
114	for (@REGISTRY, @models) {	1055	unshift @ISA, $MODEL;
115	my ($package, $model) = @$_;
116	if (eval "require $model") {
117	$MODEL = $model;
118	warn "AnyEvent: autoprobed and loaded model '$model', using it.\n" if $verbose > 1;
119	last;
120	}
121	}
122		1056
		1057	require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
		1058
		1059	(shift @post_detect)->() while @post_detect;
		1060	}
		1061
123	$MODEL	1062	$MODEL
124	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: Coro, Event, Glib or Tk.";	1063	}
		1064
		1065	sub AUTOLOAD {
		1066	(my $func = $AUTOLOAD) =~ s/.*://;
		1067
		1068	$method{$func}
		1069	or croak "$func: not a valid method for AnyEvent objects";
		1070
		1071	detect unless $MODEL;
		1072
		1073	my $class = shift;
		1074	$class->$func (@_);
		1075	}
		1076
		1077	# utility function to dup a filehandle. this is used by many backends
		1078	# to support binding more than one watcher per filehandle (they usually
		1079	# allow only one watcher per fd, so we dup it to get a different one).
		1080	sub _dupfh($$$$) {
		1081	my ($poll, $fh, $r, $w) = @_;
		1082
		1083	# cygwin requires the fh mode to be matching, unix doesn't
		1084	my ($rw, $mode) = $poll eq "r" ? ($r, "<")
		1085	: $poll eq "w" ? ($w, ">")
		1086	: Carp::croak "AnyEvent->io requires poll set to either 'r' or 'w'";
		1087
		1088	open my $fh2, "$mode&" . fileno $fh
		1089	or die "cannot dup() filehandle: $!,";
		1090
		1091	# we assume CLOEXEC is already set by perl in all important cases
		1092
		1093	($fh2, $rw)
		1094	}
		1095
		1096	package AnyEvent::Base;
		1097
		1098	# default implementations for many methods
		1099
		1100	BEGIN {
		1101	if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
		1102	*_time = \&Time::HiRes::time;
		1103	# if (eval "use POSIX (); (POSIX::times())...
		1104	} else {
		1105	*_time = sub { time }; # epic fail
		1106	}
		1107	}
		1108
		1109	sub time { _time }
		1110	sub now { _time }
		1111	sub now_update { }
		1112
		1113	# default implementation for ->condvar
		1114
		1115	sub condvar {
		1116	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
		1117	}
		1118
		1119	# default implementation for ->signal
		1120
		1121	our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
		1122
		1123	sub _signal_exec {
		1124	sysread $SIGPIPE_R, my $dummy, 4;
		1125
		1126	while (%SIG_EV) {
		1127	for (keys %SIG_EV) {
		1128	delete $SIG_EV{$_};
		1129	$_->() for values %{ $SIG_CB{$_} || {} };
125	}	1130	}
126	}	1131	}
		1132	}
127		1133
128	@ISA = $MODEL;	1134	sub signal {
		1135	my (undef, %arg) = @_;
129		1136
		1137	unless ($SIGPIPE_R) {
		1138	require Fcntl;
		1139
		1140	if (AnyEvent::WIN32) {
		1141	require AnyEvent::Util;
		1142
		1143	($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
		1144	AnyEvent::Util::fh_nonblocking ($SIGPIPE_R) if $SIGPIPE_R;
		1145	AnyEvent::Util::fh_nonblocking ($SIGPIPE_W) if $SIGPIPE_W; # just in case
		1146	} else {
		1147	pipe $SIGPIPE_R, $SIGPIPE_W;
		1148	fcntl $SIGPIPE_R, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_R;
		1149	fcntl $SIGPIPE_W, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_W; # just in case
		1150
		1151	# not strictly required, as $^F is normally 2, but let's make sure...
		1152	fcntl $SIGPIPE_R, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1153	fcntl $SIGPIPE_W, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1154	}
		1155
		1156	$SIGPIPE_R
		1157	or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
		1158
		1159	$SIG_IO = AnyEvent->io (fh => $SIGPIPE_R, poll => "r", cb => \&_signal_exec);
		1160	}
		1161
		1162	my $signal = uc $arg{signal}
		1163	or Carp::croak "required option 'signal' is missing";
		1164
		1165	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1166	$SIG{$signal} ||= sub {
		1167	local $!;
		1168	syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
		1169	undef $SIG_EV{$signal};
		1170	};
		1171
		1172	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
		1173	}
		1174
		1175	sub AnyEvent::Base::signal::DESTROY {
		1176	my ($signal, $cb) = @{$_[0]};
		1177
		1178	delete $SIG_CB{$signal}{$cb};
		1179
		1180	# delete doesn't work with older perls - they then
		1181	# print weird messages, or just unconditionally exit
		1182	# instead of getting the default action.
		1183	undef $SIG{$signal} unless keys %{ $SIG_CB{$signal} };
		1184	}
		1185
		1186	# default implementation for ->child
		1187
		1188	our %PID_CB;
		1189	our $CHLD_W;
		1190	our $CHLD_DELAY_W;
		1191	our $WNOHANG;
		1192
		1193	sub _sigchld {
		1194	while (0 < (my $pid = waitpid -1, $WNOHANG)) {
		1195	$_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }),
		1196	(values %{ $PID_CB{0} || {} });
		1197	}
		1198	}
		1199
		1200	sub child {
		1201	my (undef, %arg) = @_;
		1202
		1203	defined (my $pid = $arg{pid} + 0)
		1204	or Carp::croak "required option 'pid' is missing";
		1205
		1206	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
		1207
		1208	$WNOHANG ||= eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
		1209
		1210	unless ($CHLD_W) {
		1211	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
		1212	# child could be a zombie already, so make at least one round
		1213	&_sigchld;
		1214	}
		1215
		1216	bless [$pid, $arg{cb}], "AnyEvent::Base::child"
		1217	}
		1218
		1219	sub AnyEvent::Base::child::DESTROY {
		1220	my ($pid, $cb) = @{$_[0]};
		1221
		1222	delete $PID_CB{$pid}{$cb};
		1223	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
		1224
		1225	undef $CHLD_W unless keys %PID_CB;
		1226	}
		1227
		1228	# idle emulation is done by simply using a timer, regardless
		1229	# of whether the process is idle or not, and not letting
		1230	# the callback use more than 50% of the time.
		1231	sub idle {
		1232	my (undef, %arg) = @_;
		1233
		1234	my ($cb, $w, $rcb) = $arg{cb};
		1235
		1236	$rcb = sub {
		1237	if ($cb) {
		1238	$w = _time;
		1239	&$cb;
		1240	$w = _time - $w;
		1241
		1242	# never use more then 50% of the time for the idle watcher,
		1243	# within some limits
		1244	$w = 0.0001 if $w < 0.0001;
		1245	$w = 5 if $w > 5;
		1246
		1247	$w = AnyEvent->timer (after => $w, cb => $rcb);
		1248	} else {
		1249	# clean up...
		1250	undef $w;
		1251	undef $rcb;
		1252	}
		1253	};
		1254
		1255	$w = AnyEvent->timer (after => 0.05, cb => $rcb);
		1256
		1257	bless \\$cb, "AnyEvent::Base::idle"
		1258	}
		1259
		1260	sub AnyEvent::Base::idle::DESTROY {
		1261	undef $${$_[0]};
		1262	}
		1263
		1264	package AnyEvent::CondVar;
		1265
		1266	our @ISA = AnyEvent::CondVar::Base::;
		1267
		1268	package AnyEvent::CondVar::Base;
		1269
		1270	use overload
		1271	'&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1272	fallback => 1;
		1273
		1274	sub _send {
		1275	# nop
		1276	}
		1277
		1278	sub send {
130	my $class = shift;	1279	my $cv = shift;
131	$class->$AUTOLOAD (@_);	1280	$cv->{_ae_sent} = [@_];
		1281	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1282	$cv->_send;
132	}	1283	}
		1284
		1285	sub croak {
		1286	$_[0]{_ae_croak} = $_[1];
		1287	$_[0]->send;
		1288	}
		1289
		1290	sub ready {
		1291	$_[0]{_ae_sent}
		1292	}
		1293
		1294	sub _wait {
		1295	AnyEvent->one_event while !$_[0]{_ae_sent};
		1296	}
		1297
		1298	sub recv {
		1299	$_[0]->_wait;
		1300
		1301	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1302	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1303	}
		1304
		1305	sub cb {
		1306	$_[0]{_ae_cb} = $_[1] if @_ > 1;
		1307	$_[0]{_ae_cb}
		1308	}
		1309
		1310	sub begin {
		1311	++$_[0]{_ae_counter};
		1312	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1313	}
		1314
		1315	sub end {
		1316	return if --$_[0]{_ae_counter};
		1317	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1318	}
		1319
		1320	# undocumented/compatibility with pre-3.4
		1321	*broadcast = \&send;
		1322	*wait = \&_wait;
		1323
		1324	=head1 ERROR AND EXCEPTION HANDLING
		1325
		1326	In general, AnyEvent does not do any error handling - it relies on the
		1327	caller to do that if required. The L<AnyEvent::Strict> module (see also
		1328	the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
		1329	checking of all AnyEvent methods, however, which is highly useful during
		1330	development.
		1331
		1332	As for exception handling (i.e. runtime errors and exceptions thrown while
		1333	executing a callback), this is not only highly event-loop specific, but
		1334	also not in any way wrapped by this module, as this is the job of the main
		1335	program.
		1336
		1337	The pure perl event loop simply re-throws the exception (usually
		1338	within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
		1339	$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
		1340	so on.
		1341
		1342	=head1 ENVIRONMENT VARIABLES
		1343
		1344	The following environment variables are used by this module or its
		1345	submodules.
		1346
		1347	Note that AnyEvent will remove I<all> environment variables starting with
		1348	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		1349	enabled.
		1350
		1351	=over 4
		1352
		1353	=item C<PERL_ANYEVENT_VERBOSE>
		1354
		1355	By default, AnyEvent will be completely silent except in fatal
		1356	conditions. You can set this environment variable to make AnyEvent more
		1357	talkative.
		1358
		1359	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1360	conditions, such as not being able to load the event model specified by
		1361	C<PERL_ANYEVENT_MODEL>.
		1362
		1363	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1364	model it chooses.
		1365
		1366	=item C<PERL_ANYEVENT_STRICT>
		1367
		1368	AnyEvent does not do much argument checking by default, as thorough
		1369	argument checking is very costly. Setting this variable to a true value
		1370	will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
		1371	check the arguments passed to most method calls. If it finds any problems
		1372	it will croak.
		1373
		1374	In other words, enables "strict" mode.
		1375
		1376	Unlike C<use strict>, it is definitely recommended ot keep it off in
		1377	production. Keeping C<PERL_ANYEVENT_STRICT=1> in your environment while
		1378	developing programs can be very useful, however.
		1379
		1380	=item C<PERL_ANYEVENT_MODEL>
		1381
		1382	This can be used to specify the event model to be used by AnyEvent, before
		1383	auto detection and -probing kicks in. It must be a string consisting
		1384	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1385	and the resulting module name is loaded and if the load was successful,
		1386	used as event model. If it fails to load AnyEvent will proceed with
		1387	auto detection and -probing.
		1388
		1389	This functionality might change in future versions.
		1390
		1391	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1392	could start your program like this:
		1393
		1394	PERL_ANYEVENT_MODEL=Perl perl ...
		1395
		1396	=item C<PERL_ANYEVENT_PROTOCOLS>
		1397
		1398	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1399	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1400	of auto probing).
		1401
		1402	Must be set to a comma-separated list of protocols or address families,
		1403	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1404	used, and preference will be given to protocols mentioned earlier in the
		1405	list.
		1406
		1407	This variable can effectively be used for denial-of-service attacks
		1408	against local programs (e.g. when setuid), although the impact is likely
		1409	small, as the program has to handle conenction and other failures anyways.
		1410
		1411	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1412	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1413	- only support IPv4, never try to resolve or contact IPv6
		1414	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1415	IPv6, but prefer IPv6 over IPv4.
		1416
		1417	=item C<PERL_ANYEVENT_EDNS0>
		1418
		1419	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1420	for DNS. This extension is generally useful to reduce DNS traffic, but
		1421	some (broken) firewalls drop such DNS packets, which is why it is off by
		1422	default.
		1423
		1424	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1425	EDNS0 in its DNS requests.
		1426
		1427	=item C<PERL_ANYEVENT_MAX_FORKS>
		1428
		1429	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1430	will create in parallel.
133		1431
134	=back	1432	=back
135		1433
136	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE	1434	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1435
		1436	This is an advanced topic that you do not normally need to use AnyEvent in
		1437	a module. This section is only of use to event loop authors who want to
		1438	provide AnyEvent compatibility.
137		1439
138	If you need to support another event library which isn't directly	1440	If you need to support another event library which isn't directly
139	supported by AnyEvent, you can supply your own interface to it by	1441	supported by AnyEvent, you can supply your own interface to it by
140	pushing, before the first watch gets created, the package name of	1442	pushing, before the first watcher gets created, the package name of
141	the event module and the package name of the interface to use onto	1443	the event module and the package name of the interface to use onto
142	C<@AnyEvent::REGISTRY>. You can do that before and even without loading	1444	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
143	AnyEvent.	1445	AnyEvent, so it is reasonably cheap.
144		1446
145	Example:	1447	Example:
146		1448
147	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];	1449	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
148		1450
149	This tells AnyEvent to (literally) use the C<urxvt::anyevent::> module	1451	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
150	when it finds the C<urxvt> module is loaded. When AnyEvent is loaded and	1452	package/class when it finds the C<urxvt> package/module is already loaded.
151	requested to find a suitable event model, it will first check for the
152	urxvt module.
153		1453
		1454	When AnyEvent is loaded and asked to find a suitable event model, it
		1455	will first check for the presence of urxvt by trying to C<use> the
		1456	C<urxvt::anyevent> module.
		1457
		1458	The class should provide implementations for all watcher types. See
		1459	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
		1460	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
		1461	see the sources.
		1462
		1463	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1464	provide suitable (hopefully) replacements.
		1465
154	The above isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt) uses	1466	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
155	the above line exactly. An interface isn't included in AnyEvent	1467	terminal emulator uses the above line as-is. An interface isn't included
156	because it doesn't make sense outside the embedded interpreter inside	1468	in AnyEvent because it doesn't make sense outside the embedded interpreter
157	I<rxvt-unicode>, and it is updated and maintained as part of the	1469	inside I<rxvt-unicode>, and it is updated and maintained as part of the
158	I<rxvt-unicode> distribution.	1470	I<rxvt-unicode> distribution.
159		1471
160	=head1 ENVIRONMENT VARIABLES	1472	I<rxvt-unicode> also cheats a bit by not providing blocking access to
		1473	condition variables: code blocking while waiting for a condition will
		1474	C<die>. This still works with most modules/usages, and blocking calls must
		1475	not be done in an interactive application, so it makes sense.
161		1476
162	The following environment variables are used by this module:
163
164	C<PERL_ANYEVENT_VERBOSE> when set to C<2> or higher, reports which event
165	model gets used.
166
167	=head1 EXAMPLE	1477	=head1 EXAMPLE PROGRAM
168		1478
169	The following program uses an io watcher to read data from stdin, a timer	1479	The following program uses an I/O watcher to read data from STDIN, a timer
170	to display a message once per second, and a condvar to exit the program	1480	to display a message once per second, and a condition variable to quit the
171	when the user enters quit:	1481	program when the user enters quit:
172		1482
173	use AnyEvent;	1483	use AnyEvent;
174		1484
175	my $cv = AnyEvent->condvar;	1485	my $cv = AnyEvent->condvar;
176		1486
177	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	1487	my $io_watcher = AnyEvent->io (
		1488	fh => \*STDIN,
		1489	poll => 'r',
		1490	cb => sub {
178	warn "io event <$_[0]>\n"; # will always output <r>	1491	warn "io event <$_[0]>\n"; # will always output <r>
179	chomp (my $input = <STDIN>); # read a line	1492	chomp (my $input = <STDIN>); # read a line
180	warn "read: $input\n"; # output what has been read	1493	warn "read: $input\n"; # output what has been read
181	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1494	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1495	},
182	});	1496	);
183		1497
184	my $time_watcher; # can only be used once	1498	my $time_watcher; # can only be used once
185		1499
186	sub new_timer {	1500	sub new_timer {
187	$timer = AnyEvent->timer (after => 1, cb => sub {	1501	$timer = AnyEvent->timer (after => 1, cb => sub {
…		…
190	});	1504	});
191	}	1505	}
192		1506
193	new_timer; # create first timer	1507	new_timer; # create first timer
194		1508
195	$cv->wait; # wait until user enters /^q/i	1509	$cv->recv; # wait until user enters /^q/i
196		1510
197	=head1 REAL-WORLD EXAMPLE	1511	=head1 REAL-WORLD EXAMPLE
198		1512
199	Consider the L<Net::FCP> module. It features (among others) the following	1513	Consider the L<Net::FCP> module. It features (among others) the following
200	API calls, which are to freenet what HTTP GET requests are to http:	1514	API calls, which are to freenet what HTTP GET requests are to http:
…		…
250	syswrite $txn->{fh}, $txn->{request}	1564	syswrite $txn->{fh}, $txn->{request}
251	or die "connection or write error";	1565	or die "connection or write error";
252	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	1566	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
253		1567
254	Again, C<fh_ready_r> waits till all data has arrived, and then stores the	1568	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
255	result and signals any possible waiters that the request ahs finished:	1569	result and signals any possible waiters that the request has finished:
256		1570
257	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	1571	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
258		1572
259	if (end-of-file or data complete) {	1573	if (end-of-file or data complete) {
260	$txn->{result} = $txn->{buf};	1574	$txn->{result} = $txn->{buf};
261	$txn->{finished}->broadcast;	1575	$txn->{finished}->send;
262	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	1576	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
263	}	1577	}
264		1578
265	The C<result> method, finally, just waits for the finished signal (if the	1579	The C<result> method, finally, just waits for the finished signal (if the
266	request was already finished, it doesn't wait, of course, and returns the	1580	request was already finished, it doesn't wait, of course, and returns the
267	data:	1581	data:
268		1582
269	$txn->{finished}->wait;	1583	$txn->{finished}->recv;
270	return $txn->{result};	1584	return $txn->{result};
271		1585
272	The actual code goes further and collects all errors (C<die>s, exceptions)	1586	The actual code goes further and collects all errors (C<die>s, exceptions)
273	that occured during request processing. The C<result> method detects	1587	that occurred during request processing. The C<result> method detects
274	wether an exception as thrown (it is stored inside the $txn object)	1588	whether an exception as thrown (it is stored inside the $txn object)
275	and just throws the exception, which means connection errors and other	1589	and just throws the exception, which means connection errors and other
276	problems get reported tot he code that tries to use the result, not in a	1590	problems get reported tot he code that tries to use the result, not in a
277	random callback.	1591	random callback.
278		1592
279	All of this enables the following usage styles:	1593	All of this enables the following usage styles:
280		1594
281	1. Blocking:	1595	1. Blocking:
282		1596
283	my $data = $fcp->client_get ($url);	1597	my $data = $fcp->client_get ($url);
284		1598
285	2. Blocking, but parallelizing:	1599	2. Blocking, but running in parallel:
286		1600
287	my @datas = map $_->result,	1601	my @datas = map $_->result,
288	map $fcp->txn_client_get ($_),	1602	map $fcp->txn_client_get ($_),
289	@urls;	1603	@urls;
290		1604
291	Both blocking examples work without the module user having to know	1605	Both blocking examples work without the module user having to know
292	anything about events.	1606	anything about events.
293		1607
294	3a. Event-based in a main program, using any support Event module:	1608	3a. Event-based in a main program, using any supported event module:
295		1609
296	use Event;	1610	use EV;
297		1611
298	$fcp->txn_client_get ($url)->cb (sub {	1612	$fcp->txn_client_get ($url)->cb (sub {
299	my $txn = shift;	1613	my $txn = shift;
300	my $data = $txn->result;	1614	my $data = $txn->result;
301	...	1615	...
302	});	1616	});
303		1617
304	Event::loop;	1618	EV::loop;
305		1619
306	3b. The module user could use AnyEvent, too:	1620	3b. The module user could use AnyEvent, too:
307		1621
308	use AnyEvent;	1622	use AnyEvent;
309		1623
310	my $quit = AnyEvent->condvar;	1624	my $quit = AnyEvent->condvar;
311		1625
312	$fcp->txn_client_get ($url)->cb (sub {	1626	$fcp->txn_client_get ($url)->cb (sub {
313	...	1627	...
314	$quit->broadcast;	1628	$quit->send;
315	});	1629	});
316		1630
317	$quit->wait;	1631	$quit->recv;
		1632
		1633
		1634	=head1 BENCHMARKS
		1635
		1636	To give you an idea of the performance and overheads that AnyEvent adds
		1637	over the event loops themselves and to give you an impression of the speed
		1638	of various event loops I prepared some benchmarks.
		1639
		1640	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1641
		1642	Here is a benchmark of various supported event models used natively and
		1643	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1644	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1645	which it is), lets them fire exactly once and destroys them again.
		1646
		1647	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		1648	distribution.
		1649
		1650	=head3 Explanation of the columns
		1651
		1652	I<watcher> is the number of event watchers created/destroyed. Since
		1653	different event models feature vastly different performances, each event
		1654	loop was given a number of watchers so that overall runtime is acceptable
		1655	and similar between tested event loop (and keep them from crashing): Glib
		1656	would probably take thousands of years if asked to process the same number
		1657	of watchers as EV in this benchmark.
		1658
		1659	I<bytes> is the number of bytes (as measured by the resident set size,
		1660	RSS) consumed by each watcher. This method of measuring captures both C
		1661	and Perl-based overheads.
		1662
		1663	I<create> is the time, in microseconds (millionths of seconds), that it
		1664	takes to create a single watcher. The callback is a closure shared between
		1665	all watchers, to avoid adding memory overhead. That means closure creation
		1666	and memory usage is not included in the figures.
		1667
		1668	I<invoke> is the time, in microseconds, used to invoke a simple
		1669	callback. The callback simply counts down a Perl variable and after it was
		1670	invoked "watcher" times, it would C<< ->send >> a condvar once to
		1671	signal the end of this phase.
		1672
		1673	I<destroy> is the time, in microseconds, that it takes to destroy a single
		1674	watcher.
		1675
		1676	=head3 Results
		1677
		1678	name watchers bytes create invoke destroy comment
		1679	EV/EV 400000 224 0.47 0.35 0.27 EV native interface
		1680	EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers
		1681	CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal
		1682	Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation
		1683	Event/Event 16000 517 32.20 31.80 0.81 Event native interface
		1684	Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers
		1685	Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour
		1686	Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers
		1687	POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event
		1688	POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select
		1689
		1690	=head3 Discussion
		1691
		1692	The benchmark does I<not> measure scalability of the event loop very
		1693	well. For example, a select-based event loop (such as the pure perl one)
		1694	can never compete with an event loop that uses epoll when the number of
		1695	file descriptors grows high. In this benchmark, all events become ready at
		1696	the same time, so select/poll-based implementations get an unnatural speed
		1697	boost.
		1698
		1699	Also, note that the number of watchers usually has a nonlinear effect on
		1700	overall speed, that is, creating twice as many watchers doesn't take twice
		1701	the time - usually it takes longer. This puts event loops tested with a
		1702	higher number of watchers at a disadvantage.
		1703
		1704	To put the range of results into perspective, consider that on the
		1705	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1706	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1707	cycles with POE.
		1708
		1709	C<EV> is the sole leader regarding speed and memory use, which are both
		1710	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1711	far less memory than any other event loop and is still faster than Event
		1712	natively.
		1713
		1714	The pure perl implementation is hit in a few sweet spots (both the
		1715	constant timeout and the use of a single fd hit optimisations in the perl
		1716	interpreter and the backend itself). Nevertheless this shows that it
		1717	adds very little overhead in itself. Like any select-based backend its
		1718	performance becomes really bad with lots of file descriptors (and few of
		1719	them active), of course, but this was not subject of this benchmark.
		1720
		1721	The C<Event> module has a relatively high setup and callback invocation
		1722	cost, but overall scores in on the third place.
		1723
		1724	C<Glib>'s memory usage is quite a bit higher, but it features a
		1725	faster callback invocation and overall ends up in the same class as
		1726	C<Event>. However, Glib scales extremely badly, doubling the number of
		1727	watchers increases the processing time by more than a factor of four,
		1728	making it completely unusable when using larger numbers of watchers
		1729	(note that only a single file descriptor was used in the benchmark, so
		1730	inefficiencies of C<poll> do not account for this).
		1731
		1732	The C<Tk> adaptor works relatively well. The fact that it crashes with
		1733	more than 2000 watchers is a big setback, however, as correctness takes
		1734	precedence over speed. Nevertheless, its performance is surprising, as the
		1735	file descriptor is dup()ed for each watcher. This shows that the dup()
		1736	employed by some adaptors is not a big performance issue (it does incur a
		1737	hidden memory cost inside the kernel which is not reflected in the figures
		1738	above).
		1739
		1740	C<POE>, regardless of underlying event loop (whether using its pure perl
		1741	select-based backend or the Event module, the POE-EV backend couldn't
		1742	be tested because it wasn't working) shows abysmal performance and
		1743	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		1744	as EV watchers, and 10 times as much memory as Event (the high memory
		1745	requirements are caused by requiring a session for each watcher). Watcher
		1746	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		1747	implementation.
		1748
		1749	The design of the POE adaptor class in AnyEvent can not really account
		1750	for the performance issues, though, as session creation overhead is
		1751	small compared to execution of the state machine, which is coded pretty
		1752	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		1753	using multiple sessions is not a good approach, especially regarding
		1754	memory usage, even the author of POE could not come up with a faster
		1755	design).
		1756
		1757	=head3 Summary
		1758
		1759	=over 4
		1760
		1761	=item * Using EV through AnyEvent is faster than any other event loop
		1762	(even when used without AnyEvent), but most event loops have acceptable
		1763	performance with or without AnyEvent.
		1764
		1765	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		1766	the actual event loop, only with extremely fast event loops such as EV
		1767	adds AnyEvent significant overhead.
		1768
		1769	=item * You should avoid POE like the plague if you want performance or
		1770	reasonable memory usage.
		1771
		1772	=back
		1773
		1774	=head2 BENCHMARKING THE LARGE SERVER CASE
		1775
		1776	This benchmark actually benchmarks the event loop itself. It works by
		1777	creating a number of "servers": each server consists of a socket pair, a
		1778	timeout watcher that gets reset on activity (but never fires), and an I/O
		1779	watcher waiting for input on one side of the socket. Each time the socket
		1780	watcher reads a byte it will write that byte to a random other "server".
		1781
		1782	The effect is that there will be a lot of I/O watchers, only part of which
		1783	are active at any one point (so there is a constant number of active
		1784	fds for each loop iteration, but which fds these are is random). The
		1785	timeout is reset each time something is read because that reflects how
		1786	most timeouts work (and puts extra pressure on the event loops).
		1787
		1788	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		1789	(1%) are active. This mirrors the activity of large servers with many
		1790	connections, most of which are idle at any one point in time.
		1791
		1792	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1793	distribution.
		1794
		1795	=head3 Explanation of the columns
		1796
		1797	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1798	each server has a read and write socket end).
		1799
		1800	I<create> is the time it takes to create a socket pair (which is
		1801	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1802
		1803	I<request>, the most important value, is the time it takes to handle a
		1804	single "request", that is, reading the token from the pipe and forwarding
		1805	it to another server. This includes deleting the old timeout and creating
		1806	a new one that moves the timeout into the future.
		1807
		1808	=head3 Results
		1809
		1810	name sockets create request
		1811	EV 20000 69.01 11.16
		1812	Perl 20000 73.32 35.87
		1813	Event 20000 212.62 257.32
		1814	Glib 20000 651.16 1896.30
		1815	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1816
		1817	=head3 Discussion
		1818
		1819	This benchmark I<does> measure scalability and overall performance of the
		1820	particular event loop.
		1821
		1822	EV is again fastest. Since it is using epoll on my system, the setup time
		1823	is relatively high, though.
		1824
		1825	Perl surprisingly comes second. It is much faster than the C-based event
		1826	loops Event and Glib.
		1827
		1828	Event suffers from high setup time as well (look at its code and you will
		1829	understand why). Callback invocation also has a high overhead compared to
		1830	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1831	uses select or poll in basically all documented configurations.
		1832
		1833	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1834	clearly fails to perform with many filehandles or in busy servers.
		1835
		1836	POE is still completely out of the picture, taking over 1000 times as long
		1837	as EV, and over 100 times as long as the Perl implementation, even though
		1838	it uses a C-based event loop in this case.
		1839
		1840	=head3 Summary
		1841
		1842	=over 4
		1843
		1844	=item * The pure perl implementation performs extremely well.
		1845
		1846	=item * Avoid Glib or POE in large projects where performance matters.
		1847
		1848	=back
		1849
		1850	=head2 BENCHMARKING SMALL SERVERS
		1851
		1852	While event loops should scale (and select-based ones do not...) even to
		1853	large servers, most programs we (or I :) actually write have only a few
		1854	I/O watchers.
		1855
		1856	In this benchmark, I use the same benchmark program as in the large server
		1857	case, but it uses only eight "servers", of which three are active at any
		1858	one time. This should reflect performance for a small server relatively
		1859	well.
		1860
		1861	The columns are identical to the previous table.
		1862
		1863	=head3 Results
		1864
		1865	name sockets create request
		1866	EV 16 20.00 6.54
		1867	Perl 16 25.75 12.62
		1868	Event 16 81.27 35.86
		1869	Glib 16 32.63 15.48
		1870	POE 16 261.87 276.28 uses POE::Loop::Event
		1871
		1872	=head3 Discussion
		1873
		1874	The benchmark tries to test the performance of a typical small
		1875	server. While knowing how various event loops perform is interesting, keep
		1876	in mind that their overhead in this case is usually not as important, due
		1877	to the small absolute number of watchers (that is, you need efficiency and
		1878	speed most when you have lots of watchers, not when you only have a few of
		1879	them).
		1880
		1881	EV is again fastest.
		1882
		1883	Perl again comes second. It is noticeably faster than the C-based event
		1884	loops Event and Glib, although the difference is too small to really
		1885	matter.
		1886
		1887	POE also performs much better in this case, but is is still far behind the
		1888	others.
		1889
		1890	=head3 Summary
		1891
		1892	=over 4
		1893
		1894	=item * C-based event loops perform very well with small number of
		1895	watchers, as the management overhead dominates.
		1896
		1897	=back
		1898
		1899	=head2 THE IO::Lambda BENCHMARK
		1900
		1901	Recently I was told about the benchmark in the IO::Lambda manpage, which
		1902	could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
		1903	simply compares IO::Lambda with POE, and IO::Lambda looks better (which
		1904	shouldn't come as a surprise to anybody). As such, the benchmark is
		1905	fine, and shows that the AnyEvent backend from IO::Lambda isn't very
		1906	optimal. But how would AnyEvent compare when used without the extra
		1907	baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
		1908
		1909	The benchmark itself creates an echo-server, and then, for 500 times,
		1910	connects to the echo server, sends a line, waits for the reply, and then
		1911	creates the next connection. This is a rather bad benchmark, as it doesn't
		1912	test the efficiency of the framework, but it is a benchmark nevertheless.
		1913
		1914	name runtime
		1915	Lambda/select 0.330 sec
		1916	+ optimized 0.122 sec
		1917	Lambda/AnyEvent 0.327 sec
		1918	+ optimized 0.138 sec
		1919	Raw sockets/select 0.077 sec
		1920	POE/select, components 0.662 sec
		1921	POE/select, raw sockets 0.226 sec
		1922	POE/select, optimized 0.404 sec
		1923
		1924	AnyEvent/select/nb 0.085 sec
		1925	AnyEvent/EV/nb 0.068 sec
		1926	+state machine 0.134 sec
		1927
		1928	The benchmark is also a bit unfair (my fault) - the IO::Lambda
		1929	benchmarks actually make blocking connects and use 100% blocking I/O,
		1930	defeating the purpose of an event-based solution. All of the newly
		1931	written AnyEvent benchmarks use 100% non-blocking connects (using
		1932	AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
		1933	resolver), so AnyEvent is at a disadvantage here as non-blocking connects
		1934	generally require a lot more bookkeeping and event handling than blocking
		1935	connects (which involve a single syscall only).
		1936
		1937	The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
		1938	offers similar expressive power as POE and IO::Lambda (using conventional
		1939	Perl syntax), which means both the echo server and the client are 100%
		1940	non-blocking w.r.t. I/O, further placing it at a disadvantage.
		1941
		1942	As you can see, AnyEvent + EV even beats the hand-optimised "raw sockets
		1943	benchmark", while AnyEvent + its pure perl backend easily beats
		1944	IO::Lambda and POE.
		1945
		1946	And even the 100% non-blocking version written using the high-level (and
		1947	slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda,
		1948	even thought it does all of DNS, tcp-connect and socket I/O in a
		1949	non-blocking way.
		1950
		1951	The two AnyEvent benchmarks can be found as F<eg/ae0.pl> and F<eg/ae2.pl>
		1952	in the AnyEvent distribution, the remaining benchmarks are part of the
		1953	IO::lambda distribution and were used without any changes.
		1954
		1955
		1956	=head1 SIGNALS
		1957
		1958	AnyEvent currently installs handlers for these signals:
		1959
		1960	=over 4
		1961
		1962	=item SIGCHLD
		1963
		1964	A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
		1965	emulation for event loops that do not support them natively. Also, some
		1966	event loops install a similar handler.
		1967
		1968	=item SIGPIPE
		1969
		1970	A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
		1971	when AnyEvent gets loaded.
		1972
		1973	The rationale for this is that AnyEvent users usually do not really depend
		1974	on SIGPIPE delivery (which is purely an optimisation for shell use, or
		1975	badly-written programs), but C<SIGPIPE> can cause spurious and rare
		1976	program exits as a lot of people do not expect C<SIGPIPE> when writing to
		1977	some random socket.
		1978
		1979	The rationale for installing a no-op handler as opposed to ignoring it is
		1980	that this way, the handler will be restored to defaults on exec.
		1981
		1982	Feel free to install your own handler, or reset it to defaults.
		1983
		1984	=back
		1985
		1986	=cut
		1987
		1988	$SIG{PIPE} = sub { }
		1989	unless defined $SIG{PIPE};
		1990
		1991
		1992	=head1 FORK
		1993
		1994	Most event libraries are not fork-safe. The ones who are usually are
		1995	because they rely on inefficient but fork-safe C<select> or C<poll>
		1996	calls. Only L<EV> is fully fork-aware.
		1997
		1998	If you have to fork, you must either do so I<before> creating your first
		1999	watcher OR you must not use AnyEvent at all in the child.
		2000
		2001
		2002	=head1 SECURITY CONSIDERATIONS
		2003
		2004	AnyEvent can be forced to load any event model via
		2005	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		2006	execute arbitrary code or directly gain access, it can easily be used to
		2007	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		2008	will not be active when the program uses a different event model than
		2009	specified in the variable.
		2010
		2011	You can make AnyEvent completely ignore this variable by deleting it
		2012	before the first watcher gets created, e.g. with a C<BEGIN> block:
		2013
		2014	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		2015
		2016	use AnyEvent;
		2017
		2018	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		2019	be used to probe what backend is used and gain other information (which is
		2020	probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
		2021	$ENV{PERL_ANYEVENT_STRICT}.
		2022
		2023
		2024	=head1 BUGS
		2025
		2026	Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
		2027	to work around. If you suffer from memleaks, first upgrade to Perl 5.10
		2028	and check wether the leaks still show up. (Perl 5.10.0 has other annoying
		2029	memleaks, such as leaking on C<map> and C<grep> but it is usually not as
		2030	pronounced).
		2031
318		2032
319	=head1 SEE ALSO	2033	=head1 SEE ALSO
320		2034
321	Event modules: L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>.	2035	Utility functions: L<AnyEvent::Util>.
322		2036
323	Implementations: L<AnyEvent::Impl::Coro>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>.	2037	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
		2038	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
324		2039
325	Nontrivial usage example: L<Net::FCP>.	2040	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		2041	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		2042	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		2043	L<AnyEvent::Impl::POE>.
326		2044
327	=head1	2045	Non-blocking file handles, sockets, TCP clients and
		2046	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>.
		2047
		2048	Asynchronous DNS: L<AnyEvent::DNS>.
		2049
		2050	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>,
		2051
		2052	Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>.
		2053
		2054
		2055	=head1 AUTHOR
		2056
		2057	Marc Lehmann <schmorp@schmorp.de>
		2058	http://home.schmorp.de/
328		2059
329	=cut	2060	=cut
330		2061
331	1	2062	1
332		2063

Diff Legend

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