ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.7 by root, Fri Dec 30 01:28:31 2005 UTC vs.
Revision 1.228 by root, Wed Jul 8 01:11:12 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, see below) to
		182	watch for events (AnyEvent might or might not keep a reference to this
		183	file 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	=head3 GETTING A FILE HANDLE FROM A FILE DESCRIPTOR
		215
		216	It is not uncommon to only have a file descriptor, while AnyEvent requires
		217	a Perl file handle.
		218
		219	There are basically two methods to convert a file descriptor into a file handle. If you own
		220	the file descriptor, you can open it with C<&=>, as in:
		221
		222	open my $fh, "<&=$fileno" or die "xxx: §!";
		223
		224	This will "own" the file descriptor, meaning that when C<$fh> is
		225	destroyed, it will automatically close the C<$fileno>. Also, note that
		226	the open mode (read, write, read/write) must correspond with how the
		227	underlying file descriptor was opened.
		228
		229	In many cases, taking over the file descriptor is now what you want, in
		230	which case the only alternative is to dup the file descriptor:
		231
		232	open my $fh, "<&$fileno" or die "xxx: $!";
		233
		234	This has the advantage of not closing the file descriptor and the
		235	disadvantage of making a slow copy.
		236
		237	=head2 TIME WATCHERS
		238
		239	You can create a time watcher by calling the C<< AnyEvent->timer >>
		240	method with the following mandatory arguments:
		241
		242	C<after> specifies after how many seconds (fractional values are
		243	supported) the callback should be invoked. C<cb> is the callback to invoke
		244	in that case.
		245
		246	Although the callback might get passed parameters, their value and
		247	presence is undefined and you cannot rely on them. Portable AnyEvent
		248	callbacks cannot use arguments passed to time watcher callbacks.
		249
		250	The callback will normally be invoked once only. If you specify another
		251	parameter, C<interval>, as a strictly positive number (> 0), then the
		252	callback will be invoked regularly at that interval (in fractional
		253	seconds) after the first invocation. If C<interval> is specified with a
		254	false value, then it is treated as if it were missing.
		255
		256	The callback will be rescheduled before invoking the callback, but no
		257	attempt is done to avoid timer drift in most backends, so the interval is
		258	only approximate.
		259
		260	Example: fire an event after 7.7 seconds.
		261
		262	my $w = AnyEvent->timer (after => 7.7, cb => sub {
		263	warn "timeout\n";
		264	});
		265
		266	# to cancel the timer:
		267	undef $w;
		268
		269	Example 2: fire an event after 0.5 seconds, then roughly every second.
		270
		271	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
		272	warn "timeout\n";
		273	};
		274
		275	=head3 TIMING ISSUES
		276
		277	There are two ways to handle timers: based on real time (relative, "fire
		278	in 10 seconds") and based on wallclock time (absolute, "fire at 12
		279	o'clock").
		280
		281	While most event loops expect timers to specified in a relative way, they
		282	use absolute time internally. This makes a difference when your clock
		283	"jumps", for example, when ntp decides to set your clock backwards from
		284	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		285	fire "after" a second might actually take six years to finally fire.
		286
		287	AnyEvent cannot compensate for this. The only event loop that is conscious
		288	about these issues is L<EV>, which offers both relative (ev_timer, based
		289	on true relative time) and absolute (ev_periodic, based on wallclock time)
		290	timers.
		291
		292	AnyEvent always prefers relative timers, if available, matching the
		293	AnyEvent API.
		294
		295	AnyEvent has two additional methods that return the "current time":
62		296
63	=over 4	297	=over 4
64		298
		299	=item AnyEvent->time
		300
		301	This returns the "current wallclock time" as a fractional number of
		302	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		303	return, and the result is guaranteed to be compatible with those).
		304
		305	It progresses independently of any event loop processing, i.e. each call
		306	will check the system clock, which usually gets updated frequently.
		307
		308	=item AnyEvent->now
		309
		310	This also returns the "current wallclock time", but unlike C<time>, above,
		311	this value might change only once per event loop iteration, depending on
		312	the event loop (most return the same time as C<time>, above). This is the
		313	time that AnyEvent's timers get scheduled against.
		314
		315	I<In almost all cases (in all cases if you don't care), this is the
		316	function to call when you want to know the current time.>
		317
		318	This function is also often faster then C<< AnyEvent->time >>, and
		319	thus the preferred method if you want some timestamp (for example,
		320	L<AnyEvent::Handle> uses this to update it's activity timeouts).
		321
		322	The rest of this section is only of relevance if you try to be very exact
		323	with your timing, you can skip it without bad conscience.
		324
		325	For a practical example of when these times differ, consider L<Event::Lib>
		326	and L<EV> and the following set-up:
		327
		328	The event loop is running and has just invoked one of your callback at
		329	time=500 (assume no other callbacks delay processing). In your callback,
		330	you wait a second by executing C<sleep 1> (blocking the process for a
		331	second) and then (at time=501) you create a relative timer that fires
		332	after three seconds.
		333
		334	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		335	both return C<501>, because that is the current time, and the timer will
		336	be scheduled to fire at time=504 (C<501> + C<3>).
		337
		338	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		339	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		340	last event processing phase started. With L<EV>, your timer gets scheduled
		341	to run at time=503 (C<500> + C<3>).
		342
		343	In one sense, L<Event::Lib> is more exact, as it uses the current time
		344	regardless of any delays introduced by event processing. However, most
		345	callbacks do not expect large delays in processing, so this causes a
		346	higher drift (and a lot more system calls to get the current time).
		347
		348	In another sense, L<EV> is more exact, as your timer will be scheduled at
		349	the same time, regardless of how long event processing actually took.
		350
		351	In either case, if you care (and in most cases, you don't), then you
		352	can get whatever behaviour you want with any event loop, by taking the
		353	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		354	account.
		355
		356	=item AnyEvent->now_update
		357
		358	Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
		359	the current time for each loop iteration (see the discussion of L<<
		360	AnyEvent->now >>, above).
		361
		362	When a callback runs for a long time (or when the process sleeps), then
		363	this "current" time will differ substantially from the real time, which
		364	might affect timers and time-outs.
		365
		366	When this is the case, you can call this method, which will update the
		367	event loop's idea of "current time".
		368
		369	Note that updating the time I<might> cause some events to be handled.
		370
		371	=back
		372
		373	=head2 SIGNAL WATCHERS
		374
		375	You can watch for signals using a signal watcher, C<signal> is the signal
		376	I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
		377	callback to be invoked whenever a signal occurs.
		378
		379	Although the callback might get passed parameters, their value and
		380	presence is undefined and you cannot rely on them. Portable AnyEvent
		381	callbacks cannot use arguments passed to signal watcher callbacks.
		382
		383	Multiple signal occurrences can be clumped together into one callback
		384	invocation, and callback invocation will be synchronous. Synchronous means
		385	that it might take a while until the signal gets handled by the process,
		386	but it is guaranteed not to interrupt any other callbacks.
		387
		388	The main advantage of using these watchers is that you can share a signal
		389	between multiple watchers.
		390
		391	This watcher might use C<%SIG>, so programs overwriting those signals
		392	directly will likely not work correctly.
		393
		394	Example: exit on SIGINT
		395
		396	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		397
		398	=head2 CHILD PROCESS WATCHERS
		399
		400	You can also watch on a child process exit and catch its exit status.
		401
		402	The child process is specified by the C<pid> argument (if set to C<0>, it
		403	watches for any child process exit). The watcher will triggered only when
		404	the child process has finished and an exit status is available, not on
		405	any trace events (stopped/continued).
		406
		407	The callback will be called with the pid and exit status (as returned by
		408	waitpid), so unlike other watcher types, you I<can> rely on child watcher
		409	callback arguments.
		410
		411	This watcher type works by installing a signal handler for C<SIGCHLD>,
		412	and since it cannot be shared, nothing else should use SIGCHLD or reap
		413	random child processes (waiting for specific child processes, e.g. inside
		414	C<system>, is just fine).
		415
		416	There is a slight catch to child watchers, however: you usually start them
		417	I<after> the child process was created, and this means the process could
		418	have exited already (and no SIGCHLD will be sent anymore).
		419
		420	Not all event models handle this correctly (neither POE nor IO::Async do,
		421	see their AnyEvent::Impl manpages for details), but even for event models
		422	that I<do> handle this correctly, they usually need to be loaded before
		423	the process exits (i.e. before you fork in the first place). AnyEvent's
		424	pure perl event loop handles all cases correctly regardless of when you
		425	start the watcher.
		426
		427	This means you cannot create a child watcher as the very first
		428	thing in an AnyEvent program, you I<have> to create at least one
		429	watcher before you C<fork> the child (alternatively, you can call
		430	C<AnyEvent::detect>).
		431
		432	Example: fork a process and wait for it
		433
		434	my $done = AnyEvent->condvar;
		435
		436	my $pid = fork or exit 5;
		437
		438	my $w = AnyEvent->child (
		439	pid => $pid,
		440	cb => sub {
		441	my ($pid, $status) = @_;
		442	warn "pid $pid exited with status $status";
		443	$done->send;
		444	},
		445	);
		446
		447	# do something else, then wait for process exit
		448	$done->recv;
		449
		450	=head2 IDLE WATCHERS
		451
		452	Sometimes there is a need to do something, but it is not so important
		453	to do it instantly, but only when there is nothing better to do. This
		454	"nothing better to do" is usually defined to be "no other events need
		455	attention by the event loop".
		456
		457	Idle watchers ideally get invoked when the event loop has nothing
		458	better to do, just before it would block the process to wait for new
		459	events. Instead of blocking, the idle watcher is invoked.
		460
		461	Most event loops unfortunately do not really support idle watchers (only
		462	EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
		463	will simply call the callback "from time to time".
		464
		465	Example: read lines from STDIN, but only process them when the
		466	program is otherwise idle:
		467
		468	my @lines; # read data
		469	my $idle_w;
		470	my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		471	push @lines, scalar <STDIN>;
		472
		473	# start an idle watcher, if not already done
		474	$idle_w ||= AnyEvent->idle (cb => sub {
		475	# handle only one line, when there are lines left
		476	if (my $line = shift @lines) {
		477	print "handled when idle: $line";
		478	} else {
		479	# otherwise disable the idle watcher again
		480	undef $idle_w;
		481	}
		482	});
		483	});
		484
		485	=head2 CONDITION VARIABLES
		486
		487	If you are familiar with some event loops you will know that all of them
		488	require you to run some blocking "loop", "run" or similar function that
		489	will actively watch for new events and call your callbacks.
		490
		491	AnyEvent is different, it expects somebody else to run the event loop and
		492	will only block when necessary (usually when told by the user).
		493
		494	The instrument to do that is called a "condition variable", so called
		495	because they represent a condition that must become true.
		496
		497	Condition variables can be created by calling the C<< AnyEvent->condvar
		498	>> method, usually without arguments. The only argument pair allowed is
		499
		500	C<cb>, which specifies a callback to be called when the condition variable
		501	becomes true, with the condition variable as the first argument (but not
		502	the results).
		503
		504	After creation, the condition variable is "false" until it becomes "true"
		505	by calling the C<send> method (or calling the condition variable as if it
		506	were a callback, read about the caveats in the description for the C<<
		507	->send >> method).
		508
		509	Condition variables are similar to callbacks, except that you can
		510	optionally wait for them. They can also be called merge points - points
		511	in time where multiple outstanding events have been processed. And yet
		512	another way to call them is transactions - each condition variable can be
		513	used to represent a transaction, which finishes at some point and delivers
		514	a result.
		515
		516	Condition variables are very useful to signal that something has finished,
		517	for example, if you write a module that does asynchronous http requests,
		518	then a condition variable would be the ideal candidate to signal the
		519	availability of results. The user can either act when the callback is
		520	called or can synchronously C<< ->recv >> for the results.
		521
		522	You can also use them to simulate traditional event loops - for example,
		523	you can block your main program until an event occurs - for example, you
		524	could C<< ->recv >> in your main program until the user clicks the Quit
		525	button of your app, which would C<< ->send >> the "quit" event.
		526
		527	Note that condition variables recurse into the event loop - if you have
		528	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		529	lose. Therefore, condition variables are good to export to your caller, but
		530	you should avoid making a blocking wait yourself, at least in callbacks,
		531	as this asks for trouble.
		532
		533	Condition variables are represented by hash refs in perl, and the keys
		534	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		535	easy (it is often useful to build your own transaction class on top of
		536	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		537	it's C<new> method in your own C<new> method.
		538
		539	There are two "sides" to a condition variable - the "producer side" which
		540	eventually calls C<< -> send >>, and the "consumer side", which waits
		541	for the send to occur.
		542
		543	Example: wait for a timer.
		544
		545	# wait till the result is ready
		546	my $result_ready = AnyEvent->condvar;
		547
		548	# do something such as adding a timer
		549	# or socket watcher the calls $result_ready->send
		550	# when the "result" is ready.
		551	# in this case, we simply use a timer:
		552	my $w = AnyEvent->timer (
		553	after => 1,
		554	cb => sub { $result_ready->send },
		555	);
		556
		557	# this "blocks" (while handling events) till the callback
		558	# calls send
		559	$result_ready->recv;
		560
		561	Example: wait for a timer, but take advantage of the fact that
		562	condition variables are also code references.
		563
		564	my $done = AnyEvent->condvar;
		565	my $delay = AnyEvent->timer (after => 5, cb => $done);
		566	$done->recv;
		567
		568	Example: Imagine an API that returns a condvar and doesn't support
		569	callbacks. This is how you make a synchronous call, for example from
		570	the main program:
		571
		572	use AnyEvent::CouchDB;
		573
		574	...
		575
		576	my @info = $couchdb->info->recv;
		577
		578	And this is how you would just ste a callback to be called whenever the
		579	results are available:
		580
		581	$couchdb->info->cb (sub {
		582	my @info = $_[0]->recv;
		583	});
		584
		585	=head3 METHODS FOR PRODUCERS
		586
		587	These methods should only be used by the producing side, i.e. the
		588	code/module that eventually sends the signal. Note that it is also
		589	the producer side which creates the condvar in most cases, but it isn't
		590	uncommon for the consumer to create it as well.
		591
		592	=over 4
		593
		594	=item $cv->send (...)
		595
		596	Flag the condition as ready - a running C<< ->recv >> and all further
		597	calls to C<recv> will (eventually) return after this method has been
		598	called. If nobody is waiting the send will be remembered.
		599
		600	If a callback has been set on the condition variable, it is called
		601	immediately from within send.
		602
		603	Any arguments passed to the C<send> call will be returned by all
		604	future C<< ->recv >> calls.
		605
		606	Condition variables are overloaded so one can call them directly
		607	(as a code reference). Calling them directly is the same as calling
		608	C<send>. Note, however, that many C-based event loops do not handle
		609	overloading, so as tempting as it may be, passing a condition variable
		610	instead of a callback does not work. Both the pure perl and EV loops
		611	support overloading, however, as well as all functions that use perl to
		612	invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
		613	example).
		614
		615	=item $cv->croak ($error)
		616
		617	Similar to send, but causes all call's to C<< ->recv >> to invoke
		618	C<Carp::croak> with the given error message/object/scalar.
		619
		620	This can be used to signal any errors to the condition variable
		621	user/consumer.
		622
		623	=item $cv->begin ([group callback])
		624
		625	=item $cv->end
		626
		627	These two methods can be used to combine many transactions/events into
		628	one. For example, a function that pings many hosts in parallel might want
		629	to use a condition variable for the whole process.
		630
		631	Every call to C<< ->begin >> will increment a counter, and every call to
		632	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		633	>>, the (last) callback passed to C<begin> will be executed. That callback
		634	is I<supposed> to call C<< ->send >>, but that is not required. If no
		635	callback was set, C<send> will be called without any arguments.
		636
		637	You can think of C<< $cv->send >> giving you an OR condition (one call
		638	sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
		639	condition (all C<begin> calls must be C<end>'ed before the condvar sends).
		640
		641	Let's start with a simple example: you have two I/O watchers (for example,
		642	STDOUT and STDERR for a program), and you want to wait for both streams to
		643	close before activating a condvar:
		644
		645	my $cv = AnyEvent->condvar;
		646
		647	$cv->begin; # first watcher
		648	my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
		649	defined sysread $fh1, my $buf, 4096
		650	or $cv->end;
		651	});
		652
		653	$cv->begin; # second watcher
		654	my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
		655	defined sysread $fh2, my $buf, 4096
		656	or $cv->end;
		657	});
		658
		659	$cv->recv;
		660
		661	This works because for every event source (EOF on file handle), there is
		662	one call to C<begin>, so the condvar waits for all calls to C<end> before
		663	sending.
		664
		665	The ping example mentioned above is slightly more complicated, as the
		666	there are results to be passwd back, and the number of tasks that are
		667	begung can potentially be zero:
		668
		669	my $cv = AnyEvent->condvar;
		670
		671	my %result;
		672	$cv->begin (sub { $cv->send (\%result) });
		673
		674	for my $host (@list_of_hosts) {
		675	$cv->begin;
		676	ping_host_then_call_callback $host, sub {
		677	$result{$host} = ...;
		678	$cv->end;
		679	};
		680	}
		681
		682	$cv->end;
		683
		684	This code fragment supposedly pings a number of hosts and calls
		685	C<send> after results for all then have have been gathered - in any
		686	order. To achieve this, the code issues a call to C<begin> when it starts
		687	each ping request and calls C<end> when it has received some result for
		688	it. Since C<begin> and C<end> only maintain a counter, the order in which
		689	results arrive is not relevant.
		690
		691	There is an additional bracketing call to C<begin> and C<end> outside the
		692	loop, which serves two important purposes: first, it sets the callback
		693	to be called once the counter reaches C<0>, and second, it ensures that
		694	C<send> is called even when C<no> hosts are being pinged (the loop
		695	doesn't execute once).
		696
		697	This is the general pattern when you "fan out" into multiple (but
		698	potentially none) subrequests: use an outer C<begin>/C<end> pair to set
		699	the callback and ensure C<end> is called at least once, and then, for each
		700	subrequest you start, call C<begin> and for each subrequest you finish,
		701	call C<end>.
		702
		703	=back
		704
		705	=head3 METHODS FOR CONSUMERS
		706
		707	These methods should only be used by the consuming side, i.e. the
		708	code awaits the condition.
		709
		710	=over 4
		711
		712	=item $cv->recv
		713
		714	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		715	>> methods have been called on c<$cv>, while servicing other watchers
		716	normally.
		717
		718	You can only wait once on a condition - additional calls are valid but
		719	will return immediately.
		720
		721	If an error condition has been set by calling C<< ->croak >>, then this
		722	function will call C<croak>.
		723
		724	In list context, all parameters passed to C<send> will be returned,
		725	in scalar context only the first one will be returned.
		726
		727	Not all event models support a blocking wait - some die in that case
		728	(programs might want to do that to stay interactive), so I<if you are
		729	using this from a module, never require a blocking wait>, but let the
		730	caller decide whether the call will block or not (for example, by coupling
		731	condition variables with some kind of request results and supporting
		732	callbacks so the caller knows that getting the result will not block,
		733	while still supporting blocking waits if the caller so desires).
		734
		735	Another reason I<never> to C<< ->recv >> in a module is that you cannot
		736	sensibly have two C<< ->recv >>'s in parallel, as that would require
		737	multiple interpreters or coroutines/threads, none of which C<AnyEvent>
		738	can supply.
		739
		740	The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
		741	fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
		742	versions and also integrates coroutines into AnyEvent, making blocking
		743	C<< ->recv >> calls perfectly safe as long as they are done from another
		744	coroutine (one that doesn't run the event loop).
		745
		746	You can ensure that C<< -recv >> never blocks by setting a callback and
		747	only calling C<< ->recv >> from within that callback (or at a later
		748	time). This will work even when the event loop does not support blocking
		749	waits otherwise.
		750
		751	=item $bool = $cv->ready
		752
		753	Returns true when the condition is "true", i.e. whether C<send> or
		754	C<croak> have been called.
		755
		756	=item $cb = $cv->cb ($cb->($cv))
		757
		758	This is a mutator function that returns the callback set and optionally
		759	replaces it before doing so.
		760
		761	The callback will be called when the condition becomes "true", i.e. when
		762	C<send> or C<croak> are called, with the only argument being the condition
		763	variable itself. Calling C<recv> inside the callback or at any later time
		764	is guaranteed not to block.
		765
		766	=back
		767
		768	=head1 GLOBAL VARIABLES AND FUNCTIONS
		769
		770	=over 4
		771
		772	=item $AnyEvent::MODEL
		773
		774	Contains C<undef> until the first watcher is being created. Then it
		775	contains the event model that is being used, which is the name of the
		776	Perl class implementing the model. This class is usually one of the
		777	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case
		778	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).
		779
		780	The known classes so far are:
		781
		782	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
		783	AnyEvent::Impl::Event based on Event, second best choice.
		784	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		785	AnyEvent::Impl::Glib based on Glib, third-best choice.
		786	AnyEvent::Impl::Tk based on Tk, very bad choice.
		787	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
		788	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		789	AnyEvent::Impl::POE based on POE, not generic enough for full support.
		790
		791	# warning, support for IO::Async is only partial, as it is too broken
		792	# and limited toe ven support the AnyEvent API. See AnyEvent::Impl::Async.
		793	AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed (see its docs).
		794
		795	There is no support for WxWidgets, as WxWidgets has no support for
		796	watching file handles. However, you can use WxWidgets through the
		797	POE Adaptor, as POE has a Wx backend that simply polls 20 times per
		798	second, which was considered to be too horrible to even consider for
		799	AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
		800	it's adaptor.
		801
		802	AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
		803	autodetecting them.
		804
		805	=item AnyEvent::detect
		806
		807	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
		808	if necessary. You should only call this function right before you would
		809	have created an AnyEvent watcher anyway, that is, as late as possible at
		810	runtime.
		811
		812	=item $guard = AnyEvent::post_detect { BLOCK }
		813
		814	Arranges for the code block to be executed as soon as the event model is
		815	autodetected (or immediately if this has already happened).
		816
		817	If called in scalar or list context, then it creates and returns an object
		818	that automatically removes the callback again when it is destroyed. See
		819	L<Coro::BDB> for a case where this is useful.
		820
		821	=item @AnyEvent::post_detect
		822
		823	If there are any code references in this array (you can C<push> to it
		824	before or after loading AnyEvent), then they will called directly after
		825	the event loop has been chosen.
		826
		827	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		828	if it contains a true value then the event loop has already been detected,
		829	and the array will be ignored.
		830
		831	Best use C<AnyEvent::post_detect { BLOCK }> instead.
		832
		833	=back
		834
		835	=head1 WHAT TO DO IN A MODULE
		836
		837	As a module author, you should C<use AnyEvent> and call AnyEvent methods
		838	freely, but you should not load a specific event module or rely on it.
		839
		840	Be careful when you create watchers in the module body - AnyEvent will
		841	decide which event module to use as soon as the first method is called, so
		842	by calling AnyEvent in your module body you force the user of your module
		843	to load the event module first.
		844
		845	Never call C<< ->recv >> on a condition variable unless you I<know> that
		846	the C<< ->send >> method has been called on it already. This is
		847	because it will stall the whole program, and the whole point of using
		848	events is to stay interactive.
		849
		850	It is fine, however, to call C<< ->recv >> when the user of your module
		851	requests it (i.e. if you create a http request object ad have a method
		852	called C<results> that returns the results, it should call C<< ->recv >>
		853	freely, as the user of your module knows what she is doing. always).
		854
		855	=head1 WHAT TO DO IN THE MAIN PROGRAM
		856
		857	There will always be a single main program - the only place that should
		858	dictate which event model to use.
		859
		860	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
		861	do anything special (it does not need to be event-based) and let AnyEvent
		862	decide which implementation to chose if some module relies on it.
		863
		864	If the main program relies on a specific event model - for example, in
		865	Gtk2 programs you have to rely on the Glib module - you should load the
		866	event module before loading AnyEvent or any module that uses it: generally
		867	speaking, you should load it as early as possible. The reason is that
		868	modules might create watchers when they are loaded, and AnyEvent will
		869	decide on the event model to use as soon as it creates watchers, and it
		870	might chose the wrong one unless you load the correct one yourself.
		871
		872	You can chose to use a pure-perl implementation by loading the
		873	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
		874	everywhere, but letting AnyEvent chose the model is generally better.
		875
		876	=head2 MAINLOOP EMULATION
		877
		878	Sometimes (often for short test scripts, or even standalone programs who
		879	only want to use AnyEvent), you do not want to run a specific event loop.
		880
		881	In that case, you can use a condition variable like this:
		882
		883	AnyEvent->condvar->recv;
		884
		885	This has the effect of entering the event loop and looping forever.
		886
		887	Note that usually your program has some exit condition, in which case
		888	it is better to use the "traditional" approach of storing a condition
		889	variable somewhere, waiting for it, and sending it when the program should
		890	exit cleanly.
		891
		892
		893	=head1 OTHER MODULES
		894
		895	The following is a non-exhaustive list of additional modules that use
		896	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		897	in the same program. Some of the modules come with AnyEvent, some are
		898	available via CPAN.
		899
		900	=over 4
		901
		902	=item L<AnyEvent::Util>
		903
		904	Contains various utility functions that replace often-used but blocking
		905	functions such as C<inet_aton> by event-/callback-based versions.
		906
		907	=item L<AnyEvent::Socket>
		908
		909	Provides various utility functions for (internet protocol) sockets,
		910	addresses and name resolution. Also functions to create non-blocking tcp
		911	connections or tcp servers, with IPv6 and SRV record support and more.
		912
		913	=item L<AnyEvent::Handle>
		914
		915	Provide read and write buffers, manages watchers for reads and writes,
		916	supports raw and formatted I/O, I/O queued and fully transparent and
		917	non-blocking SSL/TLS.
		918
		919	=item L<AnyEvent::DNS>
		920
		921	Provides rich asynchronous DNS resolver capabilities.
		922
		923	=item L<AnyEvent::HTTP>
		924
		925	A simple-to-use HTTP library that is capable of making a lot of concurrent
		926	HTTP requests.
		927
		928	=item L<AnyEvent::HTTPD>
		929
		930	Provides a simple web application server framework.
		931
		932	=item L<AnyEvent::FastPing>
		933
		934	The fastest ping in the west.
		935
		936	=item L<AnyEvent::DBI>
		937
		938	Executes L<DBI> requests asynchronously in a proxy process.
		939
		940	=item L<AnyEvent::AIO>
		941
		942	Truly asynchronous I/O, should be in the toolbox of every event
		943	programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
		944	together.
		945
		946	=item L<AnyEvent::BDB>
		947
		948	Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
		949	L<BDB> and AnyEvent together.
		950
		951	=item L<AnyEvent::GPSD>
		952
		953	A non-blocking interface to gpsd, a daemon delivering GPS information.
		954
		955	=item L<AnyEvent::IGS>
		956
		957	A non-blocking interface to the Internet Go Server protocol (used by
		958	L<App::IGS>).
		959
		960	=item L<AnyEvent::IRC>
		961
		962	AnyEvent based IRC client module family (replacing the older Net::IRC3).
		963
		964	=item L<Net::XMPP2>
		965
		966	AnyEvent based XMPP (Jabber protocol) module family.
		967
		968	=item L<Net::FCP>
		969
		970	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		971	of AnyEvent.
		972
		973	=item L<Event::ExecFlow>
		974
		975	High level API for event-based execution flow control.
		976
		977	=item L<Coro>
		978
		979	Has special support for AnyEvent via L<Coro::AnyEvent>.
		980
		981	=item L<IO::Lambda>
		982
		983	The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
		984
		985	=back
		986
65	=cut	987	=cut
66		988
67	package AnyEvent;	989	package AnyEvent;
68		990
69	no warnings;	991	no warnings;
70	use strict 'vars';	992	use strict qw(vars subs);
		993
71	use Carp;	994	use Carp;
72		995
73	our $VERSION = '0.4';	996	our $VERSION = 4.8;
74	our $MODEL;	997	our $MODEL;
75		998
76	our $AUTOLOAD;	999	our $AUTOLOAD;
77	our @ISA;	1000	our @ISA;
78		1001
		1002	our @REGISTRY;
		1003
		1004	our $WIN32;
		1005
		1006	BEGIN {
		1007	eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }";
		1008	eval "sub TAINT(){ " . (${^TAINT}*1) . " }";
		1009
		1010	delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
		1011	if ${^TAINT};
		1012	}
		1013
79	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;	1014	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
80		1015
		1016	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		1017
		1018	{
		1019	my $idx;
		1020	$PROTOCOL{$_} = ++$idx
		1021	for reverse split /\s*,\s*/,
		1022	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		1023	}
		1024
81	my @models = (	1025	my @models = (
82	[Coro => Coro::Event::],	1026	[EV:: => AnyEvent::Impl::EV::],
83	[Event => Event::],	1027	[Event:: => AnyEvent::Impl::Event::],
84	[Glib => Glib::],	1028	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
85	[Tk => Tk::],	1029	# everything below here will not be autoprobed
		1030	# as the pureperl backend should work everywhere
		1031	# and is usually faster
		1032	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		1033	[Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
		1034	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		1035	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		1036	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		1037	[Wx:: => AnyEvent::Impl::POE::],
		1038	[Prima:: => AnyEvent::Impl::POE::],
		1039	# IO::Async is just too broken - we would need workaorunds for its
		1040	# byzantine signal and broken child handling, among others.
		1041	# IO::Async is rather hard to detect, as it doesn't have any
		1042	# obvious default class.
		1043	# [IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1044	# [IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1045	# [IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program
86	);	1046	);
87		1047
88	our %method = map +($_ => 1), qw(io timer condvar broadcast wait cancel DESTROY);	1048	our %method = map +($_ => 1),
		1049	qw(io timer time now now_update signal child idle condvar one_event DESTROY);
89		1050
90	sub AUTOLOAD {	1051	our @post_detect;
91	$AUTOLOAD =~ s/.*://;
92		1052
93	$method{$AUTOLOAD}	1053	sub post_detect(&) {
94	or croak "$AUTOLOAD: not a valid method for AnyEvent objects";	1054	my ($cb) = @_;
95		1055
		1056	if ($MODEL) {
		1057	$cb->();
		1058
		1059	1
		1060	} else {
		1061	push @post_detect, $cb;
		1062
		1063	defined wantarray
		1064	? bless \$cb, "AnyEvent::Util::postdetect"
		1065	: ()
		1066	}
		1067	}
		1068
		1069	sub AnyEvent::Util::postdetect::DESTROY {
		1070	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		1071	}
		1072
		1073	sub detect() {
96	unless ($MODEL) {	1074	unless ($MODEL) {
97	# check for already loaded models	1075	no strict 'refs';
98	for (@models) {	1076	local $SIG{__DIE__};
99	my ($model, $package) = @$_;	1077
100	if (${"$package\::VERSION"} > 0) {	1078	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
101	eval "require AnyEvent::Impl::$model";	1079	my $model = "AnyEvent::Impl::$1";
102	warn "AnyEvent: found model '$model', using it.\n" if $MODEL && $verbose > 1;	1080	if (eval "require $model") {
103	last if $MODEL;	1081	$MODEL = $model;
		1082	warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1;
		1083	} else {
		1084	warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose;
104	}	1085	}
105	}	1086	}
106		1087
		1088	# check for already loaded models
107	unless ($MODEL) {	1089	unless ($MODEL) {
108	# try to load a model
109
110	for (@models) {	1090	for (@REGISTRY, @models) {
111	my ($model, $package) = @$_;	1091	my ($package, $model) = @$_;
112	eval "require AnyEvent::Impl::$model";	1092	if (${"$package\::VERSION"} > 0) {
		1093	if (eval "require $model") {
		1094	$MODEL = $model;
113	warn "AnyEvent: autprobed and loaded model '$model', using it.\n" if $MODEL && $verbose > 1;	1095	warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1;
114	last if $MODEL;	1096	last;
		1097	}
		1098	}
115	}	1099	}
116		1100
		1101	unless ($MODEL) {
		1102	# try to load a model
		1103
		1104	for (@REGISTRY, @models) {
		1105	my ($package, $model) = @$_;
		1106	if (eval "require $package"
		1107	and ${"$package\::VERSION"} > 0
		1108	and eval "require $model") {
		1109	$MODEL = $model;
		1110	warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1;
		1111	last;
		1112	}
		1113	}
		1114
117	$MODEL	1115	$MODEL
118	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: Coro, Event, Glib or Tk.";	1116	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
		1117	}
		1118	}
		1119
		1120	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		1121
		1122	unshift @ISA, $MODEL;
		1123
		1124	require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
		1125
		1126	(shift @post_detect)->() while @post_detect;
		1127	}
		1128
		1129	$MODEL
		1130	}
		1131
		1132	sub AUTOLOAD {
		1133	(my $func = $AUTOLOAD) =~ s/.*://;
		1134
		1135	$method{$func}
		1136	or croak "$func: not a valid method for AnyEvent objects";
		1137
		1138	detect unless $MODEL;
		1139
		1140	my $class = shift;
		1141	$class->$func (@_);
		1142	}
		1143
		1144	# utility function to dup a filehandle. this is used by many backends
		1145	# to support binding more than one watcher per filehandle (they usually
		1146	# allow only one watcher per fd, so we dup it to get a different one).
		1147	sub _dupfh($$;$$) {
		1148	my ($poll, $fh, $r, $w) = @_;
		1149
		1150	# cygwin requires the fh mode to be matching, unix doesn't
		1151	my ($rw, $mode) = $poll eq "r" ? ($r, "<")
		1152	: $poll eq "w" ? ($w, ">")
		1153	: Carp::croak "AnyEvent->io requires poll set to either 'r' or 'w'";
		1154
		1155	open my $fh2, "$mode&" . fileno $fh
		1156	or die "cannot dup() filehandle: $!,";
		1157
		1158	# we assume CLOEXEC is already set by perl in all important cases
		1159
		1160	($fh2, $rw)
		1161	}
		1162
		1163	package AnyEvent::Base;
		1164
		1165	# default implementations for many methods
		1166
		1167	BEGIN {
		1168	if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
		1169	*_time = \&Time::HiRes::time;
		1170	# if (eval "use POSIX (); (POSIX::times())...
		1171	} else {
		1172	*_time = sub { time }; # epic fail
		1173	}
		1174	}
		1175
		1176	sub time { _time }
		1177	sub now { _time }
		1178	sub now_update { }
		1179
		1180	# default implementation for ->condvar
		1181
		1182	sub condvar {
		1183	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
		1184	}
		1185
		1186	# default implementation for ->signal
		1187
		1188	our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
		1189
		1190	sub _signal_exec {
		1191	sysread $SIGPIPE_R, my $dummy, 4;
		1192
		1193	while (%SIG_EV) {
		1194	for (keys %SIG_EV) {
		1195	delete $SIG_EV{$_};
		1196	$_->() for values %{ $SIG_CB{$_} || {} };
119	}	1197	}
120	}	1198	}
		1199	}
121		1200
122	@ISA = $MODEL;	1201	sub signal {
		1202	my (undef, %arg) = @_;
123		1203
		1204	unless ($SIGPIPE_R) {
		1205	require Fcntl;
		1206
		1207	if (AnyEvent::WIN32) {
		1208	require AnyEvent::Util;
		1209
		1210	($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
		1211	AnyEvent::Util::fh_nonblocking ($SIGPIPE_R) if $SIGPIPE_R;
		1212	AnyEvent::Util::fh_nonblocking ($SIGPIPE_W) if $SIGPIPE_W; # just in case
		1213	} else {
		1214	pipe $SIGPIPE_R, $SIGPIPE_W;
		1215	fcntl $SIGPIPE_R, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_R;
		1216	fcntl $SIGPIPE_W, &Fcntl::F_SETFL, &Fcntl::O_NONBLOCK if $SIGPIPE_W; # just in case
		1217
		1218	# not strictly required, as $^F is normally 2, but let's make sure...
		1219	fcntl $SIGPIPE_R, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1220	fcntl $SIGPIPE_W, &Fcntl::F_SETFD, &Fcntl::FD_CLOEXEC;
		1221	}
		1222
		1223	$SIGPIPE_R
		1224	or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
		1225
		1226	$SIG_IO = AnyEvent->io (fh => $SIGPIPE_R, poll => "r", cb => \&_signal_exec);
		1227	}
		1228
		1229	my $signal = uc $arg{signal}
		1230	or Carp::croak "required option 'signal' is missing";
		1231
		1232	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1233	$SIG{$signal} ||= sub {
		1234	local $!;
		1235	syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
		1236	undef $SIG_EV{$signal};
		1237	};
		1238
		1239	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
		1240	}
		1241
		1242	sub AnyEvent::Base::signal::DESTROY {
		1243	my ($signal, $cb) = @{$_[0]};
		1244
		1245	delete $SIG_CB{$signal}{$cb};
		1246
		1247	# delete doesn't work with older perls - they then
		1248	# print weird messages, or just unconditionally exit
		1249	# instead of getting the default action.
		1250	undef $SIG{$signal} unless keys %{ $SIG_CB{$signal} };
		1251	}
		1252
		1253	# default implementation for ->child
		1254
		1255	our %PID_CB;
		1256	our $CHLD_W;
		1257	our $CHLD_DELAY_W;
		1258	our $WNOHANG;
		1259
		1260	sub _sigchld {
		1261	while (0 < (my $pid = waitpid -1, $WNOHANG)) {
		1262	$_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }),
		1263	(values %{ $PID_CB{0} || {} });
		1264	}
		1265	}
		1266
		1267	sub child {
		1268	my (undef, %arg) = @_;
		1269
		1270	defined (my $pid = $arg{pid} + 0)
		1271	or Carp::croak "required option 'pid' is missing";
		1272
		1273	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
		1274
		1275	$WNOHANG ||= eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
		1276
		1277	unless ($CHLD_W) {
		1278	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
		1279	# child could be a zombie already, so make at least one round
		1280	&_sigchld;
		1281	}
		1282
		1283	bless [$pid, $arg{cb}], "AnyEvent::Base::child"
		1284	}
		1285
		1286	sub AnyEvent::Base::child::DESTROY {
		1287	my ($pid, $cb) = @{$_[0]};
		1288
		1289	delete $PID_CB{$pid}{$cb};
		1290	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
		1291
		1292	undef $CHLD_W unless keys %PID_CB;
		1293	}
		1294
		1295	# idle emulation is done by simply using a timer, regardless
		1296	# of whether the process is idle or not, and not letting
		1297	# the callback use more than 50% of the time.
		1298	sub idle {
		1299	my (undef, %arg) = @_;
		1300
		1301	my ($cb, $w, $rcb) = $arg{cb};
		1302
		1303	$rcb = sub {
		1304	if ($cb) {
		1305	$w = _time;
		1306	&$cb;
		1307	$w = _time - $w;
		1308
		1309	# never use more then 50% of the time for the idle watcher,
		1310	# within some limits
		1311	$w = 0.0001 if $w < 0.0001;
		1312	$w = 5 if $w > 5;
		1313
		1314	$w = AnyEvent->timer (after => $w, cb => $rcb);
		1315	} else {
		1316	# clean up...
		1317	undef $w;
		1318	undef $rcb;
		1319	}
		1320	};
		1321
		1322	$w = AnyEvent->timer (after => 0.05, cb => $rcb);
		1323
		1324	bless \\$cb, "AnyEvent::Base::idle"
		1325	}
		1326
		1327	sub AnyEvent::Base::idle::DESTROY {
		1328	undef $${$_[0]};
		1329	}
		1330
		1331	package AnyEvent::CondVar;
		1332
		1333	our @ISA = AnyEvent::CondVar::Base::;
		1334
		1335	package AnyEvent::CondVar::Base;
		1336
		1337	use overload
		1338	'&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1339	fallback => 1;
		1340
		1341	sub _send {
		1342	# nop
		1343	}
		1344
		1345	sub send {
124	my $class = shift;	1346	my $cv = shift;
125	$class->$AUTOLOAD (@_);	1347	$cv->{_ae_sent} = [@_];
		1348	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1349	$cv->_send;
126	}	1350	}
		1351
		1352	sub croak {
		1353	$_[0]{_ae_croak} = $_[1];
		1354	$_[0]->send;
		1355	}
		1356
		1357	sub ready {
		1358	$_[0]{_ae_sent}
		1359	}
		1360
		1361	sub _wait {
		1362	AnyEvent->one_event while !$_[0]{_ae_sent};
		1363	}
		1364
		1365	sub recv {
		1366	$_[0]->_wait;
		1367
		1368	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1369	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1370	}
		1371
		1372	sub cb {
		1373	$_[0]{_ae_cb} = $_[1] if @_ > 1;
		1374	$_[0]{_ae_cb}
		1375	}
		1376
		1377	sub begin {
		1378	++$_[0]{_ae_counter};
		1379	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1380	}
		1381
		1382	sub end {
		1383	return if --$_[0]{_ae_counter};
		1384	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1385	}
		1386
		1387	# undocumented/compatibility with pre-3.4
		1388	*broadcast = \&send;
		1389	*wait = \&_wait;
		1390
		1391	=head1 ERROR AND EXCEPTION HANDLING
		1392
		1393	In general, AnyEvent does not do any error handling - it relies on the
		1394	caller to do that if required. The L<AnyEvent::Strict> module (see also
		1395	the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
		1396	checking of all AnyEvent methods, however, which is highly useful during
		1397	development.
		1398
		1399	As for exception handling (i.e. runtime errors and exceptions thrown while
		1400	executing a callback), this is not only highly event-loop specific, but
		1401	also not in any way wrapped by this module, as this is the job of the main
		1402	program.
		1403
		1404	The pure perl event loop simply re-throws the exception (usually
		1405	within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
		1406	$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
		1407	so on.
		1408
		1409	=head1 ENVIRONMENT VARIABLES
		1410
		1411	The following environment variables are used by this module or its
		1412	submodules.
		1413
		1414	Note that AnyEvent will remove I<all> environment variables starting with
		1415	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		1416	enabled.
		1417
		1418	=over 4
		1419
		1420	=item C<PERL_ANYEVENT_VERBOSE>
		1421
		1422	By default, AnyEvent will be completely silent except in fatal
		1423	conditions. You can set this environment variable to make AnyEvent more
		1424	talkative.
		1425
		1426	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1427	conditions, such as not being able to load the event model specified by
		1428	C<PERL_ANYEVENT_MODEL>.
		1429
		1430	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1431	model it chooses.
		1432
		1433	=item C<PERL_ANYEVENT_STRICT>
		1434
		1435	AnyEvent does not do much argument checking by default, as thorough
		1436	argument checking is very costly. Setting this variable to a true value
		1437	will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
		1438	check the arguments passed to most method calls. If it finds any problems,
		1439	it will croak.
		1440
		1441	In other words, enables "strict" mode.
		1442
		1443	Unlike C<use strict>, it is definitely recommended to keep it off in
		1444	production. Keeping C<PERL_ANYEVENT_STRICT=1> in your environment while
		1445	developing programs can be very useful, however.
		1446
		1447	=item C<PERL_ANYEVENT_MODEL>
		1448
		1449	This can be used to specify the event model to be used by AnyEvent, before
		1450	auto detection and -probing kicks in. It must be a string consisting
		1451	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1452	and the resulting module name is loaded and if the load was successful,
		1453	used as event model. If it fails to load AnyEvent will proceed with
		1454	auto detection and -probing.
		1455
		1456	This functionality might change in future versions.
		1457
		1458	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1459	could start your program like this:
		1460
		1461	PERL_ANYEVENT_MODEL=Perl perl ...
		1462
		1463	=item C<PERL_ANYEVENT_PROTOCOLS>
		1464
		1465	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1466	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1467	of auto probing).
		1468
		1469	Must be set to a comma-separated list of protocols or address families,
		1470	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1471	used, and preference will be given to protocols mentioned earlier in the
		1472	list.
		1473
		1474	This variable can effectively be used for denial-of-service attacks
		1475	against local programs (e.g. when setuid), although the impact is likely
		1476	small, as the program has to handle conenction and other failures anyways.
		1477
		1478	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1479	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1480	- only support IPv4, never try to resolve or contact IPv6
		1481	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1482	IPv6, but prefer IPv6 over IPv4.
		1483
		1484	=item C<PERL_ANYEVENT_EDNS0>
		1485
		1486	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1487	for DNS. This extension is generally useful to reduce DNS traffic, but
		1488	some (broken) firewalls drop such DNS packets, which is why it is off by
		1489	default.
		1490
		1491	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1492	EDNS0 in its DNS requests.
		1493
		1494	=item C<PERL_ANYEVENT_MAX_FORKS>
		1495
		1496	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1497	will create in parallel.
		1498
		1499	=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
		1500
		1501	The default value for the C<max_outstanding> parameter for the default DNS
		1502	resolver - this is the maximum number of parallel DNS requests that are
		1503	sent to the DNS server.
		1504
		1505	=item C<PERL_ANYEVENT_RESOLV_CONF>
		1506
		1507	The file to use instead of F</etc/resolv.conf> (or OS-specific
		1508	configuration) in the default resolver. When set to the empty string, no
		1509	default config will be used.
		1510
		1511	=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
		1512
		1513	When neither C<ca_file> nor C<ca_path> was specified during
		1514	L<AnyEvent::TLS> context creation, and either of these environment
		1515	variables exist, they will be used to specify CA certificate locations
		1516	instead of a system-dependent default.
127		1517
128	=back	1518	=back
129		1519
130	=head1 ENVIRONMENT VARIABLES	1520	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
131		1521
132	The following environment variables are used by this module:	1522	This is an advanced topic that you do not normally need to use AnyEvent in
		1523	a module. This section is only of use to event loop authors who want to
		1524	provide AnyEvent compatibility.
133		1525
134	C<PERL_ANYEVENT_VERBOSE> when set to C<2> or higher, reports which event	1526	If you need to support another event library which isn't directly
135	model gets used.	1527	supported by AnyEvent, you can supply your own interface to it by
		1528	pushing, before the first watcher gets created, the package name of
		1529	the event module and the package name of the interface to use onto
		1530	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
		1531	AnyEvent, so it is reasonably cheap.
136		1532
		1533	Example:
		1534
		1535	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
		1536
		1537	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
		1538	package/class when it finds the C<urxvt> package/module is already loaded.
		1539
		1540	When AnyEvent is loaded and asked to find a suitable event model, it
		1541	will first check for the presence of urxvt by trying to C<use> the
		1542	C<urxvt::anyevent> module.
		1543
		1544	The class should provide implementations for all watcher types. See
		1545	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
		1546	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
		1547	see the sources.
		1548
		1549	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1550	provide suitable (hopefully) replacements.
		1551
		1552	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
		1553	terminal emulator uses the above line as-is. An interface isn't included
		1554	in AnyEvent because it doesn't make sense outside the embedded interpreter
		1555	inside I<rxvt-unicode>, and it is updated and maintained as part of the
		1556	I<rxvt-unicode> distribution.
		1557
		1558	I<rxvt-unicode> also cheats a bit by not providing blocking access to
		1559	condition variables: code blocking while waiting for a condition will
		1560	C<die>. This still works with most modules/usages, and blocking calls must
		1561	not be done in an interactive application, so it makes sense.
		1562
137	=head1 EXAMPLE	1563	=head1 EXAMPLE PROGRAM
138		1564
139	The following program uses an io watcher to read data from stdin, a timer	1565	The following program uses an I/O watcher to read data from STDIN, a timer
140	to display a message once per second, and a condvar to exit the program	1566	to display a message once per second, and a condition variable to quit the
141	when the user enters quit:	1567	program when the user enters quit:
142		1568
143	use AnyEvent;	1569	use AnyEvent;
144		1570
145	my $cv = AnyEvent->condvar;	1571	my $cv = AnyEvent->condvar;
146		1572
147	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	1573	my $io_watcher = AnyEvent->io (
		1574	fh => \*STDIN,
		1575	poll => 'r',
		1576	cb => sub {
148	warn "io event <$_[0]>\n"; # will always output <r>	1577	warn "io event <$_[0]>\n"; # will always output <r>
149	chomp (my $input = <STDIN>); # read a line	1578	chomp (my $input = <STDIN>); # read a line
150	warn "read: $input\n"; # output what has been read	1579	warn "read: $input\n"; # output what has been read
151	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1580	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1581	},
152	});	1582	);
153		1583
154	my $time_watcher; # can only be used once	1584	my $time_watcher; # can only be used once
155		1585
156	sub new_timer {	1586	sub new_timer {
157	$timer = AnyEvent->timer (after => 1, cb => sub {	1587	$timer = AnyEvent->timer (after => 1, cb => sub {
…		…
160	});	1590	});
161	}	1591	}
162		1592
163	new_timer; # create first timer	1593	new_timer; # create first timer
164		1594
165	$cv->wait; # wait until user enters /^q/i	1595	$cv->recv; # wait until user enters /^q/i
166		1596
167	=head1 REAL-WORLD EXAMPLE	1597	=head1 REAL-WORLD EXAMPLE
168		1598
169	Consider the L<Net::FCP> module. It features (among others) the following	1599	Consider the L<Net::FCP> module. It features (among others) the following
170	API calls, which are to freenet what HTTP GET requests are to http:	1600	API calls, which are to freenet what HTTP GET requests are to http:
…		…
220	syswrite $txn->{fh}, $txn->{request}	1650	syswrite $txn->{fh}, $txn->{request}
221	or die "connection or write error";	1651	or die "connection or write error";
222	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	1652	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
223		1653
224	Again, C<fh_ready_r> waits till all data has arrived, and then stores the	1654	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
225	result and signals any possible waiters that the request ahs finished:	1655	result and signals any possible waiters that the request has finished:
226		1656
227	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	1657	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
228		1658
229	if (end-of-file or data complete) {	1659	if (end-of-file or data complete) {
230	$txn->{result} = $txn->{buf};	1660	$txn->{result} = $txn->{buf};
231	$txn->{finished}->broadcast;	1661	$txn->{finished}->send;
232	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	1662	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
233	}	1663	}
234		1664
235	The C<result> method, finally, just waits for the finished signal (if the	1665	The C<result> method, finally, just waits for the finished signal (if the
236	request was already finished, it doesn't wait, of course, and returns the	1666	request was already finished, it doesn't wait, of course, and returns the
237	data:	1667	data:
238		1668
239	$txn->{finished}->wait;	1669	$txn->{finished}->recv;
240	return $txn->{result};	1670	return $txn->{result};
241		1671
242	The actual code goes further and collects all errors (C<die>s, exceptions)	1672	The actual code goes further and collects all errors (C<die>s, exceptions)
243	that occured during request processing. The C<result> method detects	1673	that occurred during request processing. The C<result> method detects
244	wether an exception as thrown (it is stored inside the $txn object)	1674	whether an exception as thrown (it is stored inside the $txn object)
245	and just throws the exception, which means connection errors and other	1675	and just throws the exception, which means connection errors and other
246	problems get reported tot he code that tries to use the result, not in a	1676	problems get reported tot he code that tries to use the result, not in a
247	random callback.	1677	random callback.
248		1678
249	All of this enables the following usage styles:	1679	All of this enables the following usage styles:
250		1680
251	1. Blocking:	1681	1. Blocking:
252		1682
253	my $data = $fcp->client_get ($url);	1683	my $data = $fcp->client_get ($url);
254		1684
255	2. Blocking, but parallelizing:	1685	2. Blocking, but running in parallel:
256		1686
257	my @datas = map $_->result,	1687	my @datas = map $_->result,
258	map $fcp->txn_client_get ($_),	1688	map $fcp->txn_client_get ($_),
259	@urls;	1689	@urls;
260		1690
261	Both blocking examples work without the module user having to know	1691	Both blocking examples work without the module user having to know
262	anything about events.	1692	anything about events.
263		1693
264	3a. Event-based in a main program, using any support Event module:	1694	3a. Event-based in a main program, using any supported event module:
265		1695
266	use Event;	1696	use EV;
267		1697
268	$fcp->txn_client_get ($url)->cb (sub {	1698	$fcp->txn_client_get ($url)->cb (sub {
269	my $txn = shift;	1699	my $txn = shift;
270	my $data = $txn->result;	1700	my $data = $txn->result;
271	...	1701	...
272	});	1702	});
273		1703
274	Event::loop;	1704	EV::loop;
275		1705
276	3b. The module user could use AnyEvent, too:	1706	3b. The module user could use AnyEvent, too:
277		1707
278	use AnyEvent;	1708	use AnyEvent;
279		1709
280	my $quit = AnyEvent->condvar;	1710	my $quit = AnyEvent->condvar;
281		1711
282	$fcp->txn_client_get ($url)->cb (sub {	1712	$fcp->txn_client_get ($url)->cb (sub {
283	...	1713	...
284	$quit->broadcast;	1714	$quit->send;
285	});	1715	});
286		1716
287	$quit->wait;	1717	$quit->recv;
		1718
		1719
		1720	=head1 BENCHMARKS
		1721
		1722	To give you an idea of the performance and overheads that AnyEvent adds
		1723	over the event loops themselves and to give you an impression of the speed
		1724	of various event loops I prepared some benchmarks.
		1725
		1726	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1727
		1728	Here is a benchmark of various supported event models used natively and
		1729	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1730	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1731	which it is), lets them fire exactly once and destroys them again.
		1732
		1733	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		1734	distribution.
		1735
		1736	=head3 Explanation of the columns
		1737
		1738	I<watcher> is the number of event watchers created/destroyed. Since
		1739	different event models feature vastly different performances, each event
		1740	loop was given a number of watchers so that overall runtime is acceptable
		1741	and similar between tested event loop (and keep them from crashing): Glib
		1742	would probably take thousands of years if asked to process the same number
		1743	of watchers as EV in this benchmark.
		1744
		1745	I<bytes> is the number of bytes (as measured by the resident set size,
		1746	RSS) consumed by each watcher. This method of measuring captures both C
		1747	and Perl-based overheads.
		1748
		1749	I<create> is the time, in microseconds (millionths of seconds), that it
		1750	takes to create a single watcher. The callback is a closure shared between
		1751	all watchers, to avoid adding memory overhead. That means closure creation
		1752	and memory usage is not included in the figures.
		1753
		1754	I<invoke> is the time, in microseconds, used to invoke a simple
		1755	callback. The callback simply counts down a Perl variable and after it was
		1756	invoked "watcher" times, it would C<< ->send >> a condvar once to
		1757	signal the end of this phase.
		1758
		1759	I<destroy> is the time, in microseconds, that it takes to destroy a single
		1760	watcher.
		1761
		1762	=head3 Results
		1763
		1764	name watchers bytes create invoke destroy comment
		1765	EV/EV 400000 224 0.47 0.35 0.27 EV native interface
		1766	EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers
		1767	CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal
		1768	Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation
		1769	Event/Event 16000 517 32.20 31.80 0.81 Event native interface
		1770	Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers
		1771	IOAsync/Any 16000 989 38.10 32.77 11.13 via IO::Async::Loop::IO_Poll
		1772	IOAsync/Any 16000 990 37.59 29.50 10.61 via IO::Async::Loop::Epoll
		1773	Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour
		1774	Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers
		1775	POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event
		1776	POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select
		1777
		1778	=head3 Discussion
		1779
		1780	The benchmark does I<not> measure scalability of the event loop very
		1781	well. For example, a select-based event loop (such as the pure perl one)
		1782	can never compete with an event loop that uses epoll when the number of
		1783	file descriptors grows high. In this benchmark, all events become ready at
		1784	the same time, so select/poll-based implementations get an unnatural speed
		1785	boost.
		1786
		1787	Also, note that the number of watchers usually has a nonlinear effect on
		1788	overall speed, that is, creating twice as many watchers doesn't take twice
		1789	the time - usually it takes longer. This puts event loops tested with a
		1790	higher number of watchers at a disadvantage.
		1791
		1792	To put the range of results into perspective, consider that on the
		1793	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1794	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1795	cycles with POE.
		1796
		1797	C<EV> is the sole leader regarding speed and memory use, which are both
		1798	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1799	far less memory than any other event loop and is still faster than Event
		1800	natively.
		1801
		1802	The pure perl implementation is hit in a few sweet spots (both the
		1803	constant timeout and the use of a single fd hit optimisations in the perl
		1804	interpreter and the backend itself). Nevertheless this shows that it
		1805	adds very little overhead in itself. Like any select-based backend its
		1806	performance becomes really bad with lots of file descriptors (and few of
		1807	them active), of course, but this was not subject of this benchmark.
		1808
		1809	The C<Event> module has a relatively high setup and callback invocation
		1810	cost, but overall scores in on the third place.
		1811
		1812	C<IO::Async> performs admirably well, about on par with C<Event>, even
		1813	when using its pure perl backend.
		1814
		1815	C<Glib>'s memory usage is quite a bit higher, but it features a
		1816	faster callback invocation and overall ends up in the same class as
		1817	C<Event>. However, Glib scales extremely badly, doubling the number of
		1818	watchers increases the processing time by more than a factor of four,
		1819	making it completely unusable when using larger numbers of watchers
		1820	(note that only a single file descriptor was used in the benchmark, so
		1821	inefficiencies of C<poll> do not account for this).
		1822
		1823	The C<Tk> adaptor works relatively well. The fact that it crashes with
		1824	more than 2000 watchers is a big setback, however, as correctness takes
		1825	precedence over speed. Nevertheless, its performance is surprising, as the
		1826	file descriptor is dup()ed for each watcher. This shows that the dup()
		1827	employed by some adaptors is not a big performance issue (it does incur a
		1828	hidden memory cost inside the kernel which is not reflected in the figures
		1829	above).
		1830
		1831	C<POE>, regardless of underlying event loop (whether using its pure perl
		1832	select-based backend or the Event module, the POE-EV backend couldn't
		1833	be tested because it wasn't working) shows abysmal performance and
		1834	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		1835	as EV watchers, and 10 times as much memory as Event (the high memory
		1836	requirements are caused by requiring a session for each watcher). Watcher
		1837	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		1838	implementation.
		1839
		1840	The design of the POE adaptor class in AnyEvent can not really account
		1841	for the performance issues, though, as session creation overhead is
		1842	small compared to execution of the state machine, which is coded pretty
		1843	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		1844	using multiple sessions is not a good approach, especially regarding
		1845	memory usage, even the author of POE could not come up with a faster
		1846	design).
		1847
		1848	=head3 Summary
		1849
		1850	=over 4
		1851
		1852	=item * Using EV through AnyEvent is faster than any other event loop
		1853	(even when used without AnyEvent), but most event loops have acceptable
		1854	performance with or without AnyEvent.
		1855
		1856	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		1857	the actual event loop, only with extremely fast event loops such as EV
		1858	adds AnyEvent significant overhead.
		1859
		1860	=item * You should avoid POE like the plague if you want performance or
		1861	reasonable memory usage.
		1862
		1863	=back
		1864
		1865	=head2 BENCHMARKING THE LARGE SERVER CASE
		1866
		1867	This benchmark actually benchmarks the event loop itself. It works by
		1868	creating a number of "servers": each server consists of a socket pair, a
		1869	timeout watcher that gets reset on activity (but never fires), and an I/O
		1870	watcher waiting for input on one side of the socket. Each time the socket
		1871	watcher reads a byte it will write that byte to a random other "server".
		1872
		1873	The effect is that there will be a lot of I/O watchers, only part of which
		1874	are active at any one point (so there is a constant number of active
		1875	fds for each loop iteration, but which fds these are is random). The
		1876	timeout is reset each time something is read because that reflects how
		1877	most timeouts work (and puts extra pressure on the event loops).
		1878
		1879	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		1880	(1%) are active. This mirrors the activity of large servers with many
		1881	connections, most of which are idle at any one point in time.
		1882
		1883	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1884	distribution.
		1885
		1886	=head3 Explanation of the columns
		1887
		1888	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1889	each server has a read and write socket end).
		1890
		1891	I<create> is the time it takes to create a socket pair (which is
		1892	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1893
		1894	I<request>, the most important value, is the time it takes to handle a
		1895	single "request", that is, reading the token from the pipe and forwarding
		1896	it to another server. This includes deleting the old timeout and creating
		1897	a new one that moves the timeout into the future.
		1898
		1899	=head3 Results
		1900
		1901	name sockets create request
		1902	EV 20000 69.01 11.16
		1903	Perl 20000 73.32 35.87
		1904	IOAsync 20000 157.00 98.14 epoll
		1905	IOAsync 20000 159.31 616.06 poll
		1906	Event 20000 212.62 257.32
		1907	Glib 20000 651.16 1896.30
		1908	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1909
		1910	=head3 Discussion
		1911
		1912	This benchmark I<does> measure scalability and overall performance of the
		1913	particular event loop.
		1914
		1915	EV is again fastest. Since it is using epoll on my system, the setup time
		1916	is relatively high, though.
		1917
		1918	Perl surprisingly comes second. It is much faster than the C-based event
		1919	loops Event and Glib.
		1920
		1921	IO::Async performs very well when using its epoll backend, and still quite
		1922	good compared to Glib when using its pure perl backend.
		1923
		1924	Event suffers from high setup time as well (look at its code and you will
		1925	understand why). Callback invocation also has a high overhead compared to
		1926	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1927	uses select or poll in basically all documented configurations.
		1928
		1929	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1930	clearly fails to perform with many filehandles or in busy servers.
		1931
		1932	POE is still completely out of the picture, taking over 1000 times as long
		1933	as EV, and over 100 times as long as the Perl implementation, even though
		1934	it uses a C-based event loop in this case.
		1935
		1936	=head3 Summary
		1937
		1938	=over 4
		1939
		1940	=item * The pure perl implementation performs extremely well.
		1941
		1942	=item * Avoid Glib or POE in large projects where performance matters.
		1943
		1944	=back
		1945
		1946	=head2 BENCHMARKING SMALL SERVERS
		1947
		1948	While event loops should scale (and select-based ones do not...) even to
		1949	large servers, most programs we (or I :) actually write have only a few
		1950	I/O watchers.
		1951
		1952	In this benchmark, I use the same benchmark program as in the large server
		1953	case, but it uses only eight "servers", of which three are active at any
		1954	one time. This should reflect performance for a small server relatively
		1955	well.
		1956
		1957	The columns are identical to the previous table.
		1958
		1959	=head3 Results
		1960
		1961	name sockets create request
		1962	EV 16 20.00 6.54
		1963	Perl 16 25.75 12.62
		1964	Event 16 81.27 35.86
		1965	Glib 16 32.63 15.48
		1966	POE 16 261.87 276.28 uses POE::Loop::Event
		1967
		1968	=head3 Discussion
		1969
		1970	The benchmark tries to test the performance of a typical small
		1971	server. While knowing how various event loops perform is interesting, keep
		1972	in mind that their overhead in this case is usually not as important, due
		1973	to the small absolute number of watchers (that is, you need efficiency and
		1974	speed most when you have lots of watchers, not when you only have a few of
		1975	them).
		1976
		1977	EV is again fastest.
		1978
		1979	Perl again comes second. It is noticeably faster than the C-based event
		1980	loops Event and Glib, although the difference is too small to really
		1981	matter.
		1982
		1983	POE also performs much better in this case, but is is still far behind the
		1984	others.
		1985
		1986	=head3 Summary
		1987
		1988	=over 4
		1989
		1990	=item * C-based event loops perform very well with small number of
		1991	watchers, as the management overhead dominates.
		1992
		1993	=back
		1994
		1995	=head2 THE IO::Lambda BENCHMARK
		1996
		1997	Recently I was told about the benchmark in the IO::Lambda manpage, which
		1998	could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
		1999	simply compares IO::Lambda with POE, and IO::Lambda looks better (which
		2000	shouldn't come as a surprise to anybody). As such, the benchmark is
		2001	fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
		2002	very optimal. But how would AnyEvent compare when used without the extra
		2003	baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
		2004
		2005	The benchmark itself creates an echo-server, and then, for 500 times,
		2006	connects to the echo server, sends a line, waits for the reply, and then
		2007	creates the next connection. This is a rather bad benchmark, as it doesn't
		2008	test the efficiency of the framework or much non-blocking I/O, but it is a
		2009	benchmark nevertheless.
		2010
		2011	name runtime
		2012	Lambda/select 0.330 sec
		2013	+ optimized 0.122 sec
		2014	Lambda/AnyEvent 0.327 sec
		2015	+ optimized 0.138 sec
		2016	Raw sockets/select 0.077 sec
		2017	POE/select, components 0.662 sec
		2018	POE/select, raw sockets 0.226 sec
		2019	POE/select, optimized 0.404 sec
		2020
		2021	AnyEvent/select/nb 0.085 sec
		2022	AnyEvent/EV/nb 0.068 sec
		2023	+state machine 0.134 sec
		2024
		2025	The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
		2026	benchmarks actually make blocking connects and use 100% blocking I/O,
		2027	defeating the purpose of an event-based solution. All of the newly
		2028	written AnyEvent benchmarks use 100% non-blocking connects (using
		2029	AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
		2030	resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
		2031	generally require a lot more bookkeeping and event handling than blocking
		2032	connects (which involve a single syscall only).
		2033
		2034	The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
		2035	offers similar expressive power as POE and IO::Lambda, using conventional
		2036	Perl syntax. This means that both the echo server and the client are 100%
		2037	non-blocking, further placing it at a disadvantage.
		2038
		2039	As you can see, the AnyEvent + EV combination even beats the
		2040	hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
		2041	backend easily beats IO::Lambda and POE.
		2042
		2043	And even the 100% non-blocking version written using the high-level (and
		2044	slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda by a
		2045	large margin, even though it does all of DNS, tcp-connect and socket I/O
		2046	in a non-blocking way.
		2047
		2048	The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
		2049	F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
		2050	part of the IO::lambda distribution and were used without any changes.
		2051
		2052
		2053	=head1 SIGNALS
		2054
		2055	AnyEvent currently installs handlers for these signals:
		2056
		2057	=over 4
		2058
		2059	=item SIGCHLD
		2060
		2061	A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
		2062	emulation for event loops that do not support them natively. Also, some
		2063	event loops install a similar handler.
		2064
		2065	If, when AnyEvent is loaded, SIGCHLD is set to IGNORE, then AnyEvent will
		2066	reset it to default, to avoid losing child exit statuses.
		2067
		2068	=item SIGPIPE
		2069
		2070	A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
		2071	when AnyEvent gets loaded.
		2072
		2073	The rationale for this is that AnyEvent users usually do not really depend
		2074	on SIGPIPE delivery (which is purely an optimisation for shell use, or
		2075	badly-written programs), but C<SIGPIPE> can cause spurious and rare
		2076	program exits as a lot of people do not expect C<SIGPIPE> when writing to
		2077	some random socket.
		2078
		2079	The rationale for installing a no-op handler as opposed to ignoring it is
		2080	that this way, the handler will be restored to defaults on exec.
		2081
		2082	Feel free to install your own handler, or reset it to defaults.
		2083
		2084	=back
		2085
		2086	=cut
		2087
		2088	undef $SIG{CHLD}
		2089	if $SIG{CHLD} eq 'IGNORE';
		2090
		2091	$SIG{PIPE} = sub { }
		2092	unless defined $SIG{PIPE};
		2093
		2094	=head1 FORK
		2095
		2096	Most event libraries are not fork-safe. The ones who are usually are
		2097	because they rely on inefficient but fork-safe C<select> or C<poll>
		2098	calls. Only L<EV> is fully fork-aware.
		2099
		2100	If you have to fork, you must either do so I<before> creating your first
		2101	watcher OR you must not use AnyEvent at all in the child.
		2102
		2103
		2104	=head1 SECURITY CONSIDERATIONS
		2105
		2106	AnyEvent can be forced to load any event model via
		2107	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		2108	execute arbitrary code or directly gain access, it can easily be used to
		2109	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		2110	will not be active when the program uses a different event model than
		2111	specified in the variable.
		2112
		2113	You can make AnyEvent completely ignore this variable by deleting it
		2114	before the first watcher gets created, e.g. with a C<BEGIN> block:
		2115
		2116	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		2117
		2118	use AnyEvent;
		2119
		2120	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		2121	be used to probe what backend is used and gain other information (which is
		2122	probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
		2123	$ENV{PERL_ANYEVENT_STRICT}.
		2124
		2125	Note that AnyEvent will remove I<all> environment variables starting with
		2126	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		2127	enabled.
		2128
		2129
		2130	=head1 BUGS
		2131
		2132	Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
		2133	to work around. If you suffer from memleaks, first upgrade to Perl 5.10
		2134	and check wether the leaks still show up. (Perl 5.10.0 has other annoying
		2135	memleaks, such as leaking on C<map> and C<grep> but it is usually not as
		2136	pronounced).
		2137
288		2138
289	=head1 SEE ALSO	2139	=head1 SEE ALSO
290		2140
291	Event modules: L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>.	2141	Utility functions: L<AnyEvent::Util>.
292		2142
293	Implementations: L<AnyEvent::Impl::Coro>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>.	2143	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
		2144	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
294		2145
295	Nontrivial usage example: L<Net::FCP>.	2146	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		2147	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		2148	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		2149	L<AnyEvent::Impl::POE>.
296		2150
297	=head1	2151	Non-blocking file handles, sockets, TCP clients and
		2152	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>.
		2153
		2154	Asynchronous DNS: L<AnyEvent::DNS>.
		2155
		2156	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>,
		2157
		2158	Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>.
		2159
		2160
		2161	=head1 AUTHOR
		2162
		2163	Marc Lehmann <schmorp@schmorp.de>
		2164	http://home.schmorp.de/
298		2165
299	=cut	2166	=cut
300		2167
301	1	2168	1
302		2169

Diff Legend

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