ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.13 by root, Thu Jul 20 08:26:00 2006 UTC vs.
Revision 1.324 by root, Thu May 20 23:55:51 2010 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent - provide framework for multiple event loops	3	AnyEvent - the DBI of event loop programming
4		4
5	Event, Coro, Glib, Tk - various supported event loops	5	EV, Event, Glib, Tk, Perl, Event::Lib, Irssi, rxvt-unicode, IO::Async, Qt
		6	and POE are various supported event loops/environments.
6		7
7	=head1 SYNOPSIS	8	=head1 SYNOPSIS
8		9
9	use AnyEvent;	10	use AnyEvent;
10		11
		12	# if you prefer function calls, look at the AE manpage for
		13	# an alternative API.
		14
		15	# file handle or descriptor readable
11	my $w = AnyEvent->io (fh => ..., poll => "[rw]+", cb => sub {	16	my $w = AnyEvent->io (fh => $fh, poll => "r", cb => sub { ... });
12	my ($poll_got) = @_;	17
		18	# one-shot or repeating timers
		19	my $w = AnyEvent->timer (after => $seconds, cb => sub { ... });
		20	my $w = AnyEvent->timer (after => $seconds, interval => $seconds, cb => ...
		21
		22	print AnyEvent->now; # prints current event loop time
		23	print AnyEvent->time; # think Time::HiRes::time or simply CORE::time.
		24
		25	# POSIX signal
		26	my $w = AnyEvent->signal (signal => "TERM", cb => sub { ... });
		27
		28	# child process exit
		29	my $w = AnyEvent->child (pid => $pid, cb => sub {
		30	my ($pid, $status) = @_;
13	...	31	...
14	});	32	});
15		33
16	* only one io watcher per $fh and $poll type is allowed (i.e. on a socket	34	# called when event loop idle (if applicable)
17	you can have one r + one w or one rw watcher, not any more (limitation by	35	my $w = AnyEvent->idle (cb => sub { ... });
18	Tk).
19		36
20	* the C<$poll_got> passed to the handler needs to be checked by looking	37	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	38	$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	39	$w->recv; # enters "main loop" till $condvar gets ->send
39	$w->broadcast; # wake up current and all future wait's	40	# use a condvar in callback mode:
		41	$w->cb (sub { $_[0]->recv });
40		42
41	* condvars are used to give blocking behaviour when neccessary. Create	43	=head1 INTRODUCTION/TUTORIAL
42	a condvar for any "request" or "event" your module might create, C<<	44
43	->broadcast >> it when the event happens and provide a function that calls	45	This manpage is mainly a reference manual. If you are interested
44	C<< ->wait >> for it. See the examples below.	46	in a tutorial or some gentle introduction, have a look at the
		47	L<AnyEvent::Intro> manpage.
		48
		49	=head1 SUPPORT
		50
		51	There is a mailinglist for discussing all things AnyEvent, and an IRC
		52	channel, too.
		53
		54	See the AnyEvent project page at the B<Schmorpforge Ta-Sa Software
		55	Repository>, at L<http://anyevent.schmorp.de>, for more info.
		56
		57	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
		58
		59	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
		60	nowadays. So what is different about AnyEvent?
		61
		62	Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
		63	policy> and AnyEvent is I<small and efficient>.
		64
		65	First and foremost, I<AnyEvent is not an event model> itself, it only
		66	interfaces to whatever event model the main program happens to use, in a
		67	pragmatic way. For event models and certain classes of immortals alike,
		68	the statement "there can only be one" is a bitter reality: In general,
		69	only one event loop can be active at the same time in a process. AnyEvent
		70	cannot change this, but it can hide the differences between those event
		71	loops.
		72
		73	The goal of AnyEvent is to offer module authors the ability to do event
		74	programming (waiting for I/O or timer events) without subscribing to a
		75	religion, a way of living, and most importantly: without forcing your
		76	module users into the same thing by forcing them to use the same event
		77	model you use.
		78
		79	For modules like POE or IO::Async (which is a total misnomer as it is
		80	actually doing all I/O I<synchronously>...), using them in your module is
		81	like joining a cult: After you joined, you are dependent on them and you
		82	cannot use anything else, as they are simply incompatible to everything
		83	that isn't them. What's worse, all the potential users of your
		84	module are I<also> forced to use the same event loop you use.
		85
		86	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
		87	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
		88	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
		89	your module uses one of those, every user of your module has to use it,
		90	too. But if your module uses AnyEvent, it works transparently with all
		91	event models it supports (including stuff like IO::Async, as long as those
		92	use one of the supported event loops. It is trivial to add new event loops
		93	to AnyEvent, too, so it is future-proof).
		94
		95	In addition to being free of having to use I<the one and only true event
		96	model>, AnyEvent also is free of bloat and policy: with POE or similar
		97	modules, you get an enormous amount of code and strict rules you have to
		98	follow. AnyEvent, on the other hand, is lean and up to the point, by only
		99	offering the functionality that is necessary, in as thin as a wrapper as
		100	technically possible.
		101
		102	Of course, AnyEvent comes with a big (and fully optional!) toolbox
		103	of useful functionality, such as an asynchronous DNS resolver, 100%
		104	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		105	such as Windows) and lots of real-world knowledge and workarounds for
		106	platform bugs and differences.
		107
		108	Now, if you I<do want> lots of policy (this can arguably be somewhat
		109	useful) and you want to force your users to use the one and only event
		110	model, you should I<not> use this module.
45		111
46	=head1 DESCRIPTION	112	=head1 DESCRIPTION
47		113
48	L<AnyEvent> provides an identical interface to multiple event loops. This	114	L<AnyEvent> provides an identical interface to multiple event loops. This
49	allows module authors to utilise an event loop without forcing module	115	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	116	users to use the same event loop (as only a single event loop can coexist
51	peacefully at any one time).	117	peacefully at any one time).
52		118
53	The interface itself is vaguely similar but not identical to the Event	119	The interface itself is vaguely similar, but not identical to the L<Event>
54	module.	120	module.
55		121
56	On the first call of any method, the module tries to detect the currently	122	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	123	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	124	following modules is already loaded: L<EV>,
59	used. If none is found, the module tries to load these modules in the	125	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	126	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.	127	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
		128	adaptor should always succeed) in the order given. The first one that can
		129	be successfully loaded will be used. If, after this, still none could be
		130	found, AnyEvent will fall back to a pure-perl event loop, which is not
		131	very efficient, but should work everywhere.
		132
		133	Because AnyEvent first checks for modules that are already loaded, loading
		134	an event model explicitly before first using AnyEvent will likely make
		135	that model the default. For example:
		136
		137	use Tk;
		138	use AnyEvent;
		139
		140	# .. AnyEvent will likely default to Tk
		141
		142	The I<likely> means that, if any module loads another event model and
		143	starts using it, all bets are off. Maybe you should tell their authors to
		144	use AnyEvent so their modules work together with others seamlessly...
		145
		146	The pure-perl implementation of AnyEvent is called
		147	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
		148	explicitly and enjoy the high availability of that event loop :)
		149
		150	=head1 WATCHERS
		151
		152	AnyEvent has the central concept of a I<watcher>, which is an object that
		153	stores relevant data for each kind of event you are waiting for, such as
		154	the callback to call, the file handle to watch, etc.
		155
		156	These watchers are normal Perl objects with normal Perl lifetime. After
		157	creating a watcher it will immediately "watch" for events and invoke the
		158	callback when the event occurs (of course, only when the event model
		159	is in control).
		160
		161	Note that B<callbacks must not permanently change global variables>
		162	potentially in use by the event loop (such as C<$_> or C<$[>) and that B<<
		163	callbacks must not C<die> >>. The former is good programming practise in
		164	Perl and the latter stems from the fact that exception handling differs
		165	widely between event loops.
		166
		167	To disable the watcher you have to destroy it (e.g. by setting the
		168	variable you store it in to C<undef> or otherwise deleting all references
		169	to it).
		170
		171	All watchers are created by calling a method on the C<AnyEvent> class.
		172
		173	Many watchers either are used with "recursion" (repeating timers for
		174	example), or need to refer to their watcher object in other ways.
		175
		176	An any way to achieve that is this pattern:
		177
		178	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
		179	# you can use $w here, for example to undef it
		180	undef $w;
		181	});
		182
		183	Note that C<my $w; $w => combination. This is necessary because in Perl,
		184	my variables are only visible after the statement in which they are
		185	declared.
		186
		187	=head2 I/O WATCHERS
		188
		189	$w = AnyEvent->io (
		190	fh => <filehandle_or_fileno>,
		191	poll => <"r" or "w">,
		192	cb => <callback>,
		193	);
		194
		195	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
		196	with the following mandatory key-value pairs as arguments:
		197
		198	C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch
		199	for events (AnyEvent might or might not keep a reference to this file
		200	handle). Note that only file handles pointing to things for which
		201	non-blocking operation makes sense are allowed. This includes sockets,
		202	most character devices, pipes, fifos and so on, but not for example files
		203	or block devices.
		204
		205	C<poll> must be a string that is either C<r> or C<w>, which creates a
		206	watcher waiting for "r"eadable or "w"ritable events, respectively.
		207
		208	C<cb> is the callback to invoke each time the file handle becomes ready.
		209
		210	Although the callback might get passed parameters, their value and
		211	presence is undefined and you cannot rely on them. Portable AnyEvent
		212	callbacks cannot use arguments passed to I/O watcher callbacks.
		213
		214	The I/O watcher might use the underlying file descriptor or a copy of it.
		215	You must not close a file handle as long as any watcher is active on the
		216	underlying file descriptor.
		217
		218	Some event loops issue spurious readyness notifications, so you should
		219	always use non-blocking calls when reading/writing from/to your file
		220	handles.
		221
		222	Example: wait for readability of STDIN, then read a line and disable the
		223	watcher.
		224
		225	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		226	chomp (my $input = <STDIN>);
		227	warn "read: $input\n";
		228	undef $w;
		229	});
		230
		231	=head2 TIME WATCHERS
		232
		233	$w = AnyEvent->timer (after => <seconds>, cb => <callback>);
		234
		235	$w = AnyEvent->timer (
		236	after => <fractional_seconds>,
		237	interval => <fractional_seconds>,
		238	cb => <callback>,
		239	);
		240
		241	You can create a time watcher by calling the C<< AnyEvent->timer >>
		242	method with the following mandatory arguments:
		243
		244	C<after> specifies after how many seconds (fractional values are
		245	supported) the callback should be invoked. C<cb> is the callback to invoke
		246	in that case.
		247
		248	Although the callback might get passed parameters, their value and
		249	presence is undefined and you cannot rely on them. Portable AnyEvent
		250	callbacks cannot use arguments passed to time watcher callbacks.
		251
		252	The callback will normally be invoked once only. If you specify another
		253	parameter, C<interval>, as a strictly positive number (> 0), then the
		254	callback will be invoked regularly at that interval (in fractional
		255	seconds) after the first invocation. If C<interval> is specified with a
		256	false value, then it is treated as if it were missing.
		257
		258	The callback will be rescheduled before invoking the callback, but no
		259	attempt is done to avoid timer drift in most backends, so the interval is
		260	only approximate.
		261
		262	Example: fire an event after 7.7 seconds.
		263
		264	my $w = AnyEvent->timer (after => 7.7, cb => sub {
		265	warn "timeout\n";
		266	});
		267
		268	# to cancel the timer:
		269	undef $w;
		270
		271	Example 2: fire an event after 0.5 seconds, then roughly every second.
		272
		273	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
		274	warn "timeout\n";
		275	};
		276
		277	=head3 TIMING ISSUES
		278
		279	There are two ways to handle timers: based on real time (relative, "fire
		280	in 10 seconds") and based on wallclock time (absolute, "fire at 12
		281	o'clock").
		282
		283	While most event loops expect timers to specified in a relative way, they
		284	use absolute time internally. This makes a difference when your clock
		285	"jumps", for example, when ntp decides to set your clock backwards from
		286	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		287	fire "after" a second might actually take six years to finally fire.
		288
		289	AnyEvent cannot compensate for this. The only event loop that is conscious
		290	about these issues is L<EV>, which offers both relative (ev_timer, based
		291	on true relative time) and absolute (ev_periodic, based on wallclock time)
		292	timers.
		293
		294	AnyEvent always prefers relative timers, if available, matching the
		295	AnyEvent API.
		296
		297	AnyEvent has two additional methods that return the "current time":
62		298
63	=over 4	299	=over 4
64		300
		301	=item AnyEvent->time
		302
		303	This returns the "current wallclock time" as a fractional number of
		304	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		305	return, and the result is guaranteed to be compatible with those).
		306
		307	It progresses independently of any event loop processing, i.e. each call
		308	will check the system clock, which usually gets updated frequently.
		309
		310	=item AnyEvent->now
		311
		312	This also returns the "current wallclock time", but unlike C<time>, above,
		313	this value might change only once per event loop iteration, depending on
		314	the event loop (most return the same time as C<time>, above). This is the
		315	time that AnyEvent's timers get scheduled against.
		316
		317	I<In almost all cases (in all cases if you don't care), this is the
		318	function to call when you want to know the current time.>
		319
		320	This function is also often faster then C<< AnyEvent->time >>, and
		321	thus the preferred method if you want some timestamp (for example,
		322	L<AnyEvent::Handle> uses this to update it's activity timeouts).
		323
		324	The rest of this section is only of relevance if you try to be very exact
		325	with your timing, you can skip it without bad conscience.
		326
		327	For a practical example of when these times differ, consider L<Event::Lib>
		328	and L<EV> and the following set-up:
		329
		330	The event loop is running and has just invoked one of your callback at
		331	time=500 (assume no other callbacks delay processing). In your callback,
		332	you wait a second by executing C<sleep 1> (blocking the process for a
		333	second) and then (at time=501) you create a relative timer that fires
		334	after three seconds.
		335
		336	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		337	both return C<501>, because that is the current time, and the timer will
		338	be scheduled to fire at time=504 (C<501> + C<3>).
		339
		340	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		341	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		342	last event processing phase started. With L<EV>, your timer gets scheduled
		343	to run at time=503 (C<500> + C<3>).
		344
		345	In one sense, L<Event::Lib> is more exact, as it uses the current time
		346	regardless of any delays introduced by event processing. However, most
		347	callbacks do not expect large delays in processing, so this causes a
		348	higher drift (and a lot more system calls to get the current time).
		349
		350	In another sense, L<EV> is more exact, as your timer will be scheduled at
		351	the same time, regardless of how long event processing actually took.
		352
		353	In either case, if you care (and in most cases, you don't), then you
		354	can get whatever behaviour you want with any event loop, by taking the
		355	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		356	account.
		357
		358	=item AnyEvent->now_update
		359
		360	Some event loops (such as L<EV> or L<AnyEvent::Impl::Perl>) cache
		361	the current time for each loop iteration (see the discussion of L<<
		362	AnyEvent->now >>, above).
		363
		364	When a callback runs for a long time (or when the process sleeps), then
		365	this "current" time will differ substantially from the real time, which
		366	might affect timers and time-outs.
		367
		368	When this is the case, you can call this method, which will update the
		369	event loop's idea of "current time".
		370
		371	A typical example would be a script in a web server (e.g. C<mod_perl>) -
		372	when mod_perl executes the script, then the event loop will have the wrong
		373	idea about the "current time" (being potentially far in the past, when the
		374	script ran the last time). In that case you should arrange a call to C<<
		375	AnyEvent->now_update >> each time the web server process wakes up again
		376	(e.g. at the start of your script, or in a handler).
		377
		378	Note that updating the time I<might> cause some events to be handled.
		379
		380	=back
		381
		382	=head2 SIGNAL WATCHERS
		383
		384	$w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>);
		385
		386	You can watch for signals using a signal watcher, C<signal> is the signal
		387	I<name> in uppercase and without any C<SIG> prefix, C<cb> is the Perl
		388	callback to be invoked whenever a signal occurs.
		389
		390	Although the callback might get passed parameters, their value and
		391	presence is undefined and you cannot rely on them. Portable AnyEvent
		392	callbacks cannot use arguments passed to signal watcher callbacks.
		393
		394	Multiple signal occurrences can be clumped together into one callback
		395	invocation, and callback invocation will be synchronous. Synchronous means
		396	that it might take a while until the signal gets handled by the process,
		397	but it is guaranteed not to interrupt any other callbacks.
		398
		399	The main advantage of using these watchers is that you can share a signal
		400	between multiple watchers, and AnyEvent will ensure that signals will not
		401	interrupt your program at bad times.
		402
		403	This watcher might use C<%SIG> (depending on the event loop used),
		404	so programs overwriting those signals directly will likely not work
		405	correctly.
		406
		407	Example: exit on SIGINT
		408
		409	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		410
		411	=head3 Restart Behaviour
		412
		413	While restart behaviour is up to the event loop implementation, most will
		414	not restart syscalls (that includes L<Async::Interrupt> and AnyEvent's
		415	pure perl implementation).
		416
		417	=head3 Safe/Unsafe Signals
		418
		419	Perl signals can be either "safe" (synchronous to opcode handling) or
		420	"unsafe" (asynchronous) - the former might get delayed indefinitely, the
		421	latter might corrupt your memory.
		422
		423	AnyEvent signal handlers are, in addition, synchronous to the event loop,
		424	i.e. they will not interrupt your running perl program but will only be
		425	called as part of the normal event handling (just like timer, I/O etc.
		426	callbacks, too).
		427
		428	=head3 Signal Races, Delays and Workarounds
		429
		430	Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching
		431	callbacks to signals in a generic way, which is a pity, as you cannot
		432	do race-free signal handling in perl, requiring C libraries for
		433	this. AnyEvent will try to do it's best, which means in some cases,
		434	signals will be delayed. The maximum time a signal might be delayed is
		435	specified in C<$AnyEvent::MAX_SIGNAL_LATENCY> (default: 10 seconds). This
		436	variable can be changed only before the first signal watcher is created,
		437	and should be left alone otherwise. This variable determines how often
		438	AnyEvent polls for signals (in case a wake-up was missed). Higher values
		439	will cause fewer spurious wake-ups, which is better for power and CPU
		440	saving.
		441
		442	All these problems can be avoided by installing the optional
		443	L<Async::Interrupt> module, which works with most event loops. It will not
		444	work with inherently broken event loops such as L<Event> or L<Event::Lib>
		445	(and not with L<POE> currently, as POE does it's own workaround with
		446	one-second latency). For those, you just have to suffer the delays.
		447
		448	=head2 CHILD PROCESS WATCHERS
		449
		450	$w = AnyEvent->child (pid => <process id>, cb => <callback>);
		451
		452	You can also watch on a child process exit and catch its exit status.
		453
		454	The child process is specified by the C<pid> argument (one some backends,
		455	using C<0> watches for any child process exit, on others this will
		456	croak). The watcher will be triggered only when the child process has
		457	finished and an exit status is available, not on any trace events
		458	(stopped/continued).
		459
		460	The callback will be called with the pid and exit status (as returned by
		461	waitpid), so unlike other watcher types, you I<can> rely on child watcher
		462	callback arguments.
		463
		464	This watcher type works by installing a signal handler for C<SIGCHLD>,
		465	and since it cannot be shared, nothing else should use SIGCHLD or reap
		466	random child processes (waiting for specific child processes, e.g. inside
		467	C<system>, is just fine).
		468
		469	There is a slight catch to child watchers, however: you usually start them
		470	I<after> the child process was created, and this means the process could
		471	have exited already (and no SIGCHLD will be sent anymore).
		472
		473	Not all event models handle this correctly (neither POE nor IO::Async do,
		474	see their AnyEvent::Impl manpages for details), but even for event models
		475	that I<do> handle this correctly, they usually need to be loaded before
		476	the process exits (i.e. before you fork in the first place). AnyEvent's
		477	pure perl event loop handles all cases correctly regardless of when you
		478	start the watcher.
		479
		480	This means you cannot create a child watcher as the very first
		481	thing in an AnyEvent program, you I<have> to create at least one
		482	watcher before you C<fork> the child (alternatively, you can call
		483	C<AnyEvent::detect>).
		484
		485	As most event loops do not support waiting for child events, they will be
		486	emulated by AnyEvent in most cases, in which the latency and race problems
		487	mentioned in the description of signal watchers apply.
		488
		489	Example: fork a process and wait for it
		490
		491	my $done = AnyEvent->condvar;
		492
		493	my $pid = fork or exit 5;
		494
		495	my $w = AnyEvent->child (
		496	pid => $pid,
		497	cb => sub {
		498	my ($pid, $status) = @_;
		499	warn "pid $pid exited with status $status";
		500	$done->send;
		501	},
		502	);
		503
		504	# do something else, then wait for process exit
		505	$done->recv;
		506
		507	=head2 IDLE WATCHERS
		508
		509	$w = AnyEvent->idle (cb => <callback>);
		510
		511	Repeatedly invoke the callback after the process becomes idle, until
		512	either the watcher is destroyed or new events have been detected.
		513
		514	Idle watchers are useful when there is a need to do something, but it
		515	is not so important (or wise) to do it instantly. The callback will be
		516	invoked only when there is "nothing better to do", which is usually
		517	defined as "all outstanding events have been handled and no new events
		518	have been detected". That means that idle watchers ideally get invoked
		519	when the event loop has just polled for new events but none have been
		520	detected. Instead of blocking to wait for more events, the idle watchers
		521	will be invoked.
		522
		523	Unfortunately, most event loops do not really support idle watchers (only
		524	EV, Event and Glib do it in a usable fashion) - for the rest, AnyEvent
		525	will simply call the callback "from time to time".
		526
		527	Example: read lines from STDIN, but only process them when the
		528	program is otherwise idle:
		529
		530	my @lines; # read data
		531	my $idle_w;
		532	my $io_w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		533	push @lines, scalar <STDIN>;
		534
		535	# start an idle watcher, if not already done
		536	$idle_w ||= AnyEvent->idle (cb => sub {
		537	# handle only one line, when there are lines left
		538	if (my $line = shift @lines) {
		539	print "handled when idle: $line";
		540	} else {
		541	# otherwise disable the idle watcher again
		542	undef $idle_w;
		543	}
		544	});
		545	});
		546
		547	=head2 CONDITION VARIABLES
		548
		549	$cv = AnyEvent->condvar;
		550
		551	$cv->send (<list>);
		552	my @res = $cv->recv;
		553
		554	If you are familiar with some event loops you will know that all of them
		555	require you to run some blocking "loop", "run" or similar function that
		556	will actively watch for new events and call your callbacks.
		557
		558	AnyEvent is slightly different: it expects somebody else to run the event
		559	loop and will only block when necessary (usually when told by the user).
		560
		561	The instrument to do that is called a "condition variable", so called
		562	because they represent a condition that must become true.
		563
		564	Now is probably a good time to look at the examples further below.
		565
		566	Condition variables can be created by calling the C<< AnyEvent->condvar
		567	>> method, usually without arguments. The only argument pair allowed is
		568	C<cb>, which specifies a callback to be called when the condition variable
		569	becomes true, with the condition variable as the first argument (but not
		570	the results).
		571
		572	After creation, the condition variable is "false" until it becomes "true"
		573	by calling the C<send> method (or calling the condition variable as if it
		574	were a callback, read about the caveats in the description for the C<<
		575	->send >> method).
		576
		577	Condition variables are similar to callbacks, except that you can
		578	optionally wait for them. They can also be called merge points - points
		579	in time where multiple outstanding events have been processed. And yet
		580	another way to call them is transactions - each condition variable can be
		581	used to represent a transaction, which finishes at some point and delivers
		582	a result. And yet some people know them as "futures" - a promise to
		583	compute/deliver something that you can wait for.
		584
		585	Condition variables are very useful to signal that something has finished,
		586	for example, if you write a module that does asynchronous http requests,
		587	then a condition variable would be the ideal candidate to signal the
		588	availability of results. The user can either act when the callback is
		589	called or can synchronously C<< ->recv >> for the results.
		590
		591	You can also use them to simulate traditional event loops - for example,
		592	you can block your main program until an event occurs - for example, you
		593	could C<< ->recv >> in your main program until the user clicks the Quit
		594	button of your app, which would C<< ->send >> the "quit" event.
		595
		596	Note that condition variables recurse into the event loop - if you have
		597	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		598	lose. Therefore, condition variables are good to export to your caller, but
		599	you should avoid making a blocking wait yourself, at least in callbacks,
		600	as this asks for trouble.
		601
		602	Condition variables are represented by hash refs in perl, and the keys
		603	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		604	easy (it is often useful to build your own transaction class on top of
		605	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		606	it's C<new> method in your own C<new> method.
		607
		608	There are two "sides" to a condition variable - the "producer side" which
		609	eventually calls C<< -> send >>, and the "consumer side", which waits
		610	for the send to occur.
		611
		612	Example: wait for a timer.
		613
		614	# condition: "wait till the timer is fired"
		615	my $timer_fired = AnyEvent->condvar;
		616
		617	# create the timer - we could wait for, say
		618	# a handle becomign ready, or even an
		619	# AnyEvent::HTTP request to finish, but
		620	# in this case, we simply use a timer:
		621	my $w = AnyEvent->timer (
		622	after => 1,
		623	cb => sub { $timer_fired->send },
		624	);
		625
		626	# this "blocks" (while handling events) till the callback
		627	# calls ->send
		628	$timer_fired->recv;
		629
		630	Example: wait for a timer, but take advantage of the fact that condition
		631	variables are also callable directly.
		632
		633	my $done = AnyEvent->condvar;
		634	my $delay = AnyEvent->timer (after => 5, cb => $done);
		635	$done->recv;
		636
		637	Example: Imagine an API that returns a condvar and doesn't support
		638	callbacks. This is how you make a synchronous call, for example from
		639	the main program:
		640
		641	use AnyEvent::CouchDB;
		642
		643	...
		644
		645	my @info = $couchdb->info->recv;
		646
		647	And this is how you would just set a callback to be called whenever the
		648	results are available:
		649
		650	$couchdb->info->cb (sub {
		651	my @info = $_[0]->recv;
		652	});
		653
		654	=head3 METHODS FOR PRODUCERS
		655
		656	These methods should only be used by the producing side, i.e. the
		657	code/module that eventually sends the signal. Note that it is also
		658	the producer side which creates the condvar in most cases, but it isn't
		659	uncommon for the consumer to create it as well.
		660
		661	=over 4
		662
		663	=item $cv->send (...)
		664
		665	Flag the condition as ready - a running C<< ->recv >> and all further
		666	calls to C<recv> will (eventually) return after this method has been
		667	called. If nobody is waiting the send will be remembered.
		668
		669	If a callback has been set on the condition variable, it is called
		670	immediately from within send.
		671
		672	Any arguments passed to the C<send> call will be returned by all
		673	future C<< ->recv >> calls.
		674
		675	Condition variables are overloaded so one can call them directly (as if
		676	they were a code reference). Calling them directly is the same as calling
		677	C<send>.
		678
		679	=item $cv->croak ($error)
		680
		681	Similar to send, but causes all call's to C<< ->recv >> to invoke
		682	C<Carp::croak> with the given error message/object/scalar.
		683
		684	This can be used to signal any errors to the condition variable
		685	user/consumer. Doing it this way instead of calling C<croak> directly
		686	delays the error detetcion, but has the overwhelmign advantage that it
		687	diagnoses the error at the place where the result is expected, and not
		688	deep in some event clalback without connection to the actual code causing
		689	the problem.
		690
		691	=item $cv->begin ([group callback])
		692
		693	=item $cv->end
		694
		695	These two methods can be used to combine many transactions/events into
		696	one. For example, a function that pings many hosts in parallel might want
		697	to use a condition variable for the whole process.
		698
		699	Every call to C<< ->begin >> will increment a counter, and every call to
		700	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		701	>>, the (last) callback passed to C<begin> will be executed, passing the
		702	condvar as first argument. That callback is I<supposed> to call C<< ->send
		703	>>, but that is not required. If no group callback was set, C<send> will
		704	be called without any arguments.
		705
		706	You can think of C<< $cv->send >> giving you an OR condition (one call
		707	sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND
		708	condition (all C<begin> calls must be C<end>'ed before the condvar sends).
		709
		710	Let's start with a simple example: you have two I/O watchers (for example,
		711	STDOUT and STDERR for a program), and you want to wait for both streams to
		712	close before activating a condvar:
		713
		714	my $cv = AnyEvent->condvar;
		715
		716	$cv->begin; # first watcher
		717	my $w1 = AnyEvent->io (fh => $fh1, cb => sub {
		718	defined sysread $fh1, my $buf, 4096
		719	or $cv->end;
		720	});
		721
		722	$cv->begin; # second watcher
		723	my $w2 = AnyEvent->io (fh => $fh2, cb => sub {
		724	defined sysread $fh2, my $buf, 4096
		725	or $cv->end;
		726	});
		727
		728	$cv->recv;
		729
		730	This works because for every event source (EOF on file handle), there is
		731	one call to C<begin>, so the condvar waits for all calls to C<end> before
		732	sending.
		733
		734	The ping example mentioned above is slightly more complicated, as the
		735	there are results to be passwd back, and the number of tasks that are
		736	begung can potentially be zero:
		737
		738	my $cv = AnyEvent->condvar;
		739
		740	my %result;
		741	$cv->begin (sub { shift->send (\%result) });
		742
		743	for my $host (@list_of_hosts) {
		744	$cv->begin;
		745	ping_host_then_call_callback $host, sub {
		746	$result{$host} = ...;
		747	$cv->end;
		748	};
		749	}
		750
		751	$cv->end;
		752
		753	This code fragment supposedly pings a number of hosts and calls
		754	C<send> after results for all then have have been gathered - in any
		755	order. To achieve this, the code issues a call to C<begin> when it starts
		756	each ping request and calls C<end> when it has received some result for
		757	it. Since C<begin> and C<end> only maintain a counter, the order in which
		758	results arrive is not relevant.
		759
		760	There is an additional bracketing call to C<begin> and C<end> outside the
		761	loop, which serves two important purposes: first, it sets the callback
		762	to be called once the counter reaches C<0>, and second, it ensures that
		763	C<send> is called even when C<no> hosts are being pinged (the loop
		764	doesn't execute once).
		765
		766	This is the general pattern when you "fan out" into multiple (but
		767	potentially none) subrequests: use an outer C<begin>/C<end> pair to set
		768	the callback and ensure C<end> is called at least once, and then, for each
		769	subrequest you start, call C<begin> and for each subrequest you finish,
		770	call C<end>.
		771
		772	=back
		773
		774	=head3 METHODS FOR CONSUMERS
		775
		776	These methods should only be used by the consuming side, i.e. the
		777	code awaits the condition.
		778
		779	=over 4
		780
		781	=item $cv->recv
		782
		783	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		784	>> methods have been called on c<$cv>, while servicing other watchers
		785	normally.
		786
		787	You can only wait once on a condition - additional calls are valid but
		788	will return immediately.
		789
		790	If an error condition has been set by calling C<< ->croak >>, then this
		791	function will call C<croak>.
		792
		793	In list context, all parameters passed to C<send> will be returned,
		794	in scalar context only the first one will be returned.
		795
		796	Note that doing a blocking wait in a callback is not supported by any
		797	event loop, that is, recursive invocation of a blocking C<< ->recv
		798	>> is not allowed, and the C<recv> call will C<croak> if such a
		799	condition is detected. This condition can be slightly loosened by using
		800	L<Coro::AnyEvent>, which allows you to do a blocking C<< ->recv >> from
		801	any thread that doesn't run the event loop itself.
		802
		803	Not all event models support a blocking wait - some die in that case
		804	(programs might want to do that to stay interactive), so I<if you are
		805	using this from a module, never require a blocking wait>. Instead, let the
		806	caller decide whether the call will block or not (for example, by coupling
		807	condition variables with some kind of request results and supporting
		808	callbacks so the caller knows that getting the result will not block,
		809	while still supporting blocking waits if the caller so desires).
		810
		811	You can ensure that C<< -recv >> never blocks by setting a callback and
		812	only calling C<< ->recv >> from within that callback (or at a later
		813	time). This will work even when the event loop does not support blocking
		814	waits otherwise.
		815
		816	=item $bool = $cv->ready
		817
		818	Returns true when the condition is "true", i.e. whether C<send> or
		819	C<croak> have been called.
		820
		821	=item $cb = $cv->cb ($cb->($cv))
		822
		823	This is a mutator function that returns the callback set and optionally
		824	replaces it before doing so.
		825
		826	The callback will be called when the condition becomes (or already was)
		827	"true", i.e. when C<send> or C<croak> are called (or were called), with
		828	the only argument being the condition variable itself. Calling C<recv>
		829	inside the callback or at any later time is guaranteed not to block.
		830
		831	=back
		832
		833	=head1 SUPPORTED EVENT LOOPS/BACKENDS
		834
		835	The available backend classes are (every class has its own manpage):
		836
		837	=over 4
		838
		839	=item Backends that are autoprobed when no other event loop can be found.
		840
		841	EV is the preferred backend when no other event loop seems to be in
		842	use. If EV is not installed, then AnyEvent will fall back to its own
		843	pure-perl implementation, which is available everywhere as it comes with
		844	AnyEvent itself.
		845
		846	AnyEvent::Impl::EV based on EV (interface to libev, best choice).
		847	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		848
		849	=item Backends that are transparently being picked up when they are used.
		850
		851	These will be used when they are currently loaded when the first watcher
		852	is created, in which case it is assumed that the application is using
		853	them. This means that AnyEvent will automatically pick the right backend
		854	when the main program loads an event module before anything starts to
		855	create watchers. Nothing special needs to be done by the main program.
		856
		857	AnyEvent::Impl::Event based on Event, very stable, few glitches.
		858	AnyEvent::Impl::Glib based on Glib, slow but very stable.
		859	AnyEvent::Impl::Tk based on Tk, very broken.
		860	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		861	AnyEvent::Impl::POE based on POE, very slow, some limitations.
		862	AnyEvent::Impl::Irssi used when running within irssi.
		863
		864	=item Backends with special needs.
		865
		866	Qt requires the Qt::Application to be instantiated first, but will
		867	otherwise be picked up automatically. As long as the main program
		868	instantiates the application before any AnyEvent watchers are created,
		869	everything should just work.
		870
		871	AnyEvent::Impl::Qt based on Qt.
		872
		873	Support for IO::Async can only be partial, as it is too broken and
		874	architecturally limited to even support the AnyEvent API. It also
		875	is the only event loop that needs the loop to be set explicitly, so
		876	it can only be used by a main program knowing about AnyEvent. See
		877	L<AnyEvent::Impl::Async> for the gory details.
		878
		879	AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed.
		880
		881	=item Event loops that are indirectly supported via other backends.
		882
		883	Some event loops can be supported via other modules:
		884
		885	There is no direct support for WxWidgets (L<Wx>) or L<Prima>.
		886
		887	B<WxWidgets> has no support for watching file handles. However, you can
		888	use WxWidgets through the POE adaptor, as POE has a Wx backend that simply
		889	polls 20 times per second, which was considered to be too horrible to even
		890	consider for AnyEvent.
		891
		892	B<Prima> is not supported as nobody seems to be using it, but it has a POE
		893	backend, so it can be supported through POE.
		894
		895	AnyEvent knows about both L<Prima> and L<Wx>, however, and will try to
		896	load L<POE> when detecting them, in the hope that POE will pick them up,
		897	in which case everything will be automatic.
		898
		899	=back
		900
		901	=head1 GLOBAL VARIABLES AND FUNCTIONS
		902
		903	These are not normally required to use AnyEvent, but can be useful to
		904	write AnyEvent extension modules.
		905
		906	=over 4
		907
		908	=item $AnyEvent::MODEL
		909
		910	Contains C<undef> until the first watcher is being created, before the
		911	backend has been autodetected.
		912
		913	Afterwards it contains the event model that is being used, which is the
		914	name of the Perl class implementing the model. This class is usually one
		915	of the C<AnyEvent::Impl:xxx> modules, but can be any other class in the
		916	case AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode> it
		917	will be C<urxvt::anyevent>).
		918
		919	=item AnyEvent::detect
		920
		921	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
		922	if necessary. You should only call this function right before you would
		923	have created an AnyEvent watcher anyway, that is, as late as possible at
		924	runtime, and not e.g. while initialising of your module.
		925
		926	If you need to do some initialisation before AnyEvent watchers are
		927	created, use C<post_detect>.
		928
		929	=item $guard = AnyEvent::post_detect { BLOCK }
		930
		931	Arranges for the code block to be executed as soon as the event model is
		932	autodetected (or immediately if this has already happened).
		933
		934	The block will be executed I<after> the actual backend has been detected
		935	(C<$AnyEvent::MODEL> is set), but I<before> any watchers have been
		936	created, so it is possible to e.g. patch C<@AnyEvent::ISA> or do
		937	other initialisations - see the sources of L<AnyEvent::Strict> or
		938	L<AnyEvent::AIO> to see how this is used.
		939
		940	The most common usage is to create some global watchers, without forcing
		941	event module detection too early, for example, L<AnyEvent::AIO> creates
		942	and installs the global L<IO::AIO> watcher in a C<post_detect> block to
		943	avoid autodetecting the event module at load time.
		944
		945	If called in scalar or list context, then it creates and returns an object
		946	that automatically removes the callback again when it is destroyed (or
		947	C<undef> when the hook was immediately executed). See L<AnyEvent::AIO> for
		948	a case where this is useful.
		949
		950	Example: Create a watcher for the IO::AIO module and store it in
		951	C<$WATCHER>. Only do so after the event loop is initialised, though.
		952
		953	our WATCHER;
		954
		955	my $guard = AnyEvent::post_detect {
		956	$WATCHER = AnyEvent->io (fh => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb);
		957	};
		958
		959	# the ||= is important in case post_detect immediately runs the block,
		960	# as to not clobber the newly-created watcher. assigning both watcher and
		961	# post_detect guard to the same variable has the advantage of users being
		962	# able to just C<undef $WATCHER> if the watcher causes them grief.
		963
		964	$WATCHER ||= $guard;
		965
		966	=item @AnyEvent::post_detect
		967
		968	If there are any code references in this array (you can C<push> to it
		969	before or after loading AnyEvent), then they will called directly after
		970	the event loop has been chosen.
		971
		972	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		973	if it is defined then the event loop has already been detected, and the
		974	array will be ignored.
		975
		976	Best use C<AnyEvent::post_detect { BLOCK }> when your application allows
		977	it, as it takes care of these details.
		978
		979	This variable is mainly useful for modules that can do something useful
		980	when AnyEvent is used and thus want to know when it is initialised, but do
		981	not need to even load it by default. This array provides the means to hook
		982	into AnyEvent passively, without loading it.
		983
		984	Example: To load Coro::AnyEvent whenever Coro and AnyEvent are used
		985	together, you could put this into Coro (this is the actual code used by
		986	Coro to accomplish this):
		987
		988	if (defined $AnyEvent::MODEL) {
		989	# AnyEvent already initialised, so load Coro::AnyEvent
		990	require Coro::AnyEvent;
		991	} else {
		992	# AnyEvent not yet initialised, so make sure to load Coro::AnyEvent
		993	# as soon as it is
		994	push @AnyEvent::post_detect, sub { require Coro::AnyEvent };
		995	}
		996
		997	=back
		998
		999	=head1 WHAT TO DO IN A MODULE
		1000
		1001	As a module author, you should C<use AnyEvent> and call AnyEvent methods
		1002	freely, but you should not load a specific event module or rely on it.
		1003
		1004	Be careful when you create watchers in the module body - AnyEvent will
		1005	decide which event module to use as soon as the first method is called, so
		1006	by calling AnyEvent in your module body you force the user of your module
		1007	to load the event module first.
		1008
		1009	Never call C<< ->recv >> on a condition variable unless you I<know> that
		1010	the C<< ->send >> method has been called on it already. This is
		1011	because it will stall the whole program, and the whole point of using
		1012	events is to stay interactive.
		1013
		1014	It is fine, however, to call C<< ->recv >> when the user of your module
		1015	requests it (i.e. if you create a http request object ad have a method
		1016	called C<results> that returns the results, it should call C<< ->recv >>
		1017	freely, as the user of your module knows what she is doing. always).
		1018
		1019	=head1 WHAT TO DO IN THE MAIN PROGRAM
		1020
		1021	There will always be a single main program - the only place that should
		1022	dictate which event model to use.
		1023
		1024	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
		1025	do anything special (it does not need to be event-based) and let AnyEvent
		1026	decide which implementation to chose if some module relies on it.
		1027
		1028	If the main program relies on a specific event model - for example, in
		1029	Gtk2 programs you have to rely on the Glib module - you should load the
		1030	event module before loading AnyEvent or any module that uses it: generally
		1031	speaking, you should load it as early as possible. The reason is that
		1032	modules might create watchers when they are loaded, and AnyEvent will
		1033	decide on the event model to use as soon as it creates watchers, and it
		1034	might chose the wrong one unless you load the correct one yourself.
		1035
		1036	You can chose to use a pure-perl implementation by loading the
		1037	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
		1038	everywhere, but letting AnyEvent chose the model is generally better.
		1039
		1040	=head2 MAINLOOP EMULATION
		1041
		1042	Sometimes (often for short test scripts, or even standalone programs who
		1043	only want to use AnyEvent), you do not want to run a specific event loop.
		1044
		1045	In that case, you can use a condition variable like this:
		1046
		1047	AnyEvent->condvar->recv;
		1048
		1049	This has the effect of entering the event loop and looping forever.
		1050
		1051	Note that usually your program has some exit condition, in which case
		1052	it is better to use the "traditional" approach of storing a condition
		1053	variable somewhere, waiting for it, and sending it when the program should
		1054	exit cleanly.
		1055
		1056
		1057	=head1 OTHER MODULES
		1058
		1059	The following is a non-exhaustive list of additional modules that use
		1060	AnyEvent as a client and can therefore be mixed easily with other AnyEvent
		1061	modules and other event loops in the same program. Some of the modules
		1062	come with AnyEvent, the others are available via CPAN.
		1063
		1064	=over 4
		1065
		1066	=item L<AnyEvent::Util>
		1067
		1068	Contains various utility functions that replace often-used but blocking
		1069	functions such as C<inet_aton> by event-/callback-based versions.
		1070
		1071	=item L<AnyEvent::Socket>
		1072
		1073	Provides various utility functions for (internet protocol) sockets,
		1074	addresses and name resolution. Also functions to create non-blocking tcp
		1075	connections or tcp servers, with IPv6 and SRV record support and more.
		1076
		1077	=item L<AnyEvent::Handle>
		1078
		1079	Provide read and write buffers, manages watchers for reads and writes,
		1080	supports raw and formatted I/O, I/O queued and fully transparent and
		1081	non-blocking SSL/TLS (via L<AnyEvent::TLS>.
		1082
		1083	=item L<AnyEvent::DNS>
		1084
		1085	Provides rich asynchronous DNS resolver capabilities.
		1086
		1087	=item L<AnyEvent::HTTP>, L<AnyEvent::IRC>, L<AnyEvent::XMPP>, L<AnyEvent::GPSD>, L<AnyEvent::IGS>, L<AnyEvent::FCP>
		1088
		1089	Implement event-based interfaces to the protocols of the same name (for
		1090	the curious, IGS is the International Go Server and FCP is the Freenet
		1091	Client Protocol).
		1092
		1093	=item L<AnyEvent::Handle::UDP>
		1094
		1095	Here be danger!
		1096
		1097	As Pauli would put it, "Not only is it not right, it's not even wrong!" -
		1098	there are so many things wrong with AnyEvent::Handle::UDP, most notably
		1099	it's use of a stream-based API with a protocol that isn't streamable, that
		1100	the only way to improve it is to delete it.
		1101
		1102	It features data corruption (but typically only under load) and general
		1103	confusion. On top, the author is not only clueless about UDP but also
		1104	fact-resistant - some gems of his understanding: "connect doesn't work
		1105	with UDP", "UDP packets are not IP packets", "UDP only has datagrams, not
		1106	packets", "I don't need to implement proper error checking as UDP doesn't
		1107	support error checking" and so on - he doesn't even understand what's
		1108	wrong with his module when it is explained to him.
		1109
		1110	=item L<AnyEvent::DBI>
		1111
		1112	Executes L<DBI> requests asynchronously in a proxy process for you,
		1113	notifying you in an event-bnased way when the operation is finished.
		1114
		1115	=item L<AnyEvent::AIO>
		1116
		1117	Truly asynchronous (as opposed to non-blocking) I/O, should be in the
		1118	toolbox of every event programmer. AnyEvent::AIO transparently fuses
		1119	L<IO::AIO> and AnyEvent together, giving AnyEvent access to event-based
		1120	file I/O, and much more.
		1121
		1122	=item L<AnyEvent::HTTPD>
		1123
		1124	A simple embedded webserver.
		1125
		1126	=item L<AnyEvent::FastPing>
		1127
		1128	The fastest ping in the west.
		1129
		1130	=item L<Coro>
		1131
		1132	Has special support for AnyEvent via L<Coro::AnyEvent>.
		1133
		1134	=back
		1135
65	=cut	1136	=cut
66		1137
67	package AnyEvent;	1138	package AnyEvent;
68		1139
69	no warnings;	1140	# basically a tuned-down version of common::sense
70	use strict 'vars';	1141	sub common_sense {
		1142	# from common:.sense 1.0
		1143	${^WARNING_BITS} = "\xfc\x3f\x33\x00\x0f\xf3\xcf\xc0\xf3\xfc\x33\x00";
		1144	# use strict vars subs - NO UTF-8, as Util.pm doesn't like this atm. (uts46data.pl)
		1145	$^H |= 0x00000600;
		1146	}
		1147
		1148	BEGIN { AnyEvent::common_sense }
		1149
71	use Carp;	1150	use Carp ();
72		1151
73	our $VERSION = '1.02';	1152	our $VERSION = '5.261';
74	our $MODEL;	1153	our $MODEL;
75		1154
76	our $AUTOLOAD;	1155	our $AUTOLOAD;
77	our @ISA;	1156	our @ISA;
78		1157
79	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
80
81	our @REGISTRY;	1158	our @REGISTRY;
82		1159
		1160	our $VERBOSE;
		1161
		1162	BEGIN {
		1163	require "AnyEvent/constants.pl";
		1164
		1165	eval "sub TAINT (){" . (${^TAINT}*1) . "}";
		1166
		1167	delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV}
		1168	if ${^TAINT};
		1169
		1170	$VERBOSE = $ENV{PERL_ANYEVENT_VERBOSE}*1;
		1171
		1172	}
		1173
		1174	our $MAX_SIGNAL_LATENCY = 10;
		1175
		1176	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		1177
		1178	{
		1179	my $idx;
		1180	$PROTOCOL{$_} = ++$idx
		1181	for reverse split /\s*,\s*/,
		1182	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		1183	}
		1184
83	my @models = (	1185	my @models = (
84	[Coro::Event:: => AnyEvent::Impl::Coro::],	1186	[EV:: => AnyEvent::Impl::EV:: , 1],
		1187	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl:: , 1],
		1188	# everything below here will not (normally) be autoprobed
		1189	# as the pureperl backend should work everywhere
		1190	# and is usually faster
85	[Event:: => AnyEvent::Impl::Event::],	1191	[Event:: => AnyEvent::Impl::Event::, 1],
86	[Glib:: => AnyEvent::Impl::Glib::],	1192	[Glib:: => AnyEvent::Impl::Glib:: , 1], # becomes extremely slow with many watchers
		1193	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		1194	[Irssi:: => AnyEvent::Impl::Irssi::], # Irssi has a bogus "Event" package
		1195	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		1196	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		1197	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
87	[Tk:: => AnyEvent::Impl::Tk::],	1198	[Wx:: => AnyEvent::Impl::POE::],
		1199	[Prima:: => AnyEvent::Impl::POE::],
		1200	# IO::Async is just too broken - we would need workarounds for its
		1201	# byzantine signal and broken child handling, among others.
		1202	# IO::Async is rather hard to detect, as it doesn't have any
		1203	# obvious default class.
		1204	[IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1205	[IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1206	[IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program
		1207	[AnyEvent::Impl::IOAsync:: => AnyEvent::Impl::IOAsync::], # requires special main program
88	);	1208	);
89		1209
90	our %method = map +($_ => 1), qw(io timer condvar broadcast wait cancel DESTROY);	1210	our %method = map +($_ => 1),
		1211	qw(io timer time now now_update signal child idle condvar one_event DESTROY);
91		1212
92	sub AUTOLOAD {	1213	our @post_detect;
93	$AUTOLOAD =~ s/.*://;
94		1214
95	$method{$AUTOLOAD}	1215	sub post_detect(&) {
96	or croak "$AUTOLOAD: not a valid method for AnyEvent objects";	1216	my ($cb) = @_;
97		1217
		1218	push @post_detect, $cb;
		1219
		1220	defined wantarray
		1221	? bless \$cb, "AnyEvent::Util::postdetect"
		1222	: ()
		1223	}
		1224
		1225	sub AnyEvent::Util::postdetect::DESTROY {
		1226	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		1227	}
		1228
		1229	sub detect() {
		1230	# free some memory
		1231	*detect = sub () { $MODEL };
		1232
		1233	local $!; # for good measure
		1234	local $SIG{__DIE__};
		1235
		1236	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
		1237	my $model = "AnyEvent::Impl::$1";
		1238	if (eval "require $model") {
		1239	$MODEL = $model;
		1240	warn "AnyEvent: loaded model '$model' (forced by \$ENV{PERL_ANYEVENT_MODEL}), using it.\n" if $VERBOSE >= 2;
		1241	} else {
		1242	warn "AnyEvent: unable to load model '$model' (from \$ENV{PERL_ANYEVENT_MODEL}):\n$@" if $VERBOSE;
		1243	}
		1244	}
		1245
		1246	# check for already loaded models
98	unless ($MODEL) {	1247	unless ($MODEL) {
99	# check for already loaded models
100	for (@REGISTRY, @models) {	1248	for (@REGISTRY, @models) {
101	my ($package, $model) = @$_;	1249	my ($package, $model) = @$_;
102	if (${"$package\::VERSION"} > 0) {	1250	if (${"$package\::VERSION"} > 0) {
103	if (eval "require $model") {	1251	if (eval "require $model") {
104	$MODEL = $model;	1252	$MODEL = $model;
105	warn "AnyEvent: found model '$model', using it.\n" if $verbose > 1;	1253	warn "AnyEvent: autodetected model '$model', using it.\n" if $VERBOSE >= 2;
106	last;	1254	last;
107	}	1255	}
108	}	1256	}
109	}	1257	}
110		1258
111	unless ($MODEL) {	1259	unless ($MODEL) {
112	# try to load a model	1260	# try to autoload a model
113
114	for (@REGISTRY, @models) {	1261	for (@REGISTRY, @models) {
115	my ($package, $model) = @$_;	1262	my ($package, $model, $autoload) = @$_;
		1263	if (
		1264	$autoload
		1265	and eval "require $package"
		1266	and ${"$package\::VERSION"} > 0
116	if (eval "require $model") {	1267	and eval "require $model"
		1268	) {
117	$MODEL = $model;	1269	$MODEL = $model;
118	warn "AnyEvent: autoprobed and loaded model '$model', using it.\n" if $verbose > 1;	1270	warn "AnyEvent: autoloaded model '$model', using it.\n" if $VERBOSE >= 2;
119	last;	1271	last;
120	}	1272	}
121	}	1273	}
122		1274
123	$MODEL	1275	$MODEL
124	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: Coro, Event, Glib or Tk.";	1276	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.\n";
125	}	1277	}
126	}	1278	}
127		1279
128	@ISA = $MODEL;	1280	@models = (); # free probe data
		1281
		1282	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		1283	unshift @ISA, $MODEL;
		1284
		1285	# now nuke some methods that are overriden by the backend.
		1286	# SUPER is not allowed.
		1287	for (qw(time signal child idle)) {
		1288	undef &{"AnyEvent::Base::$_"}
		1289	if defined &{"$MODEL\::$_"};
		1290	}
		1291
		1292	require AnyEvent::Strict if $ENV{PERL_ANYEVENT_STRICT};
		1293
		1294	(shift @post_detect)->() while @post_detect;
		1295
		1296	*post_detect = sub(&) {
		1297	shift->();
		1298
		1299	undef
		1300	};
		1301
		1302	$MODEL
		1303	}
		1304
		1305	sub AUTOLOAD {
		1306	(my $func = $AUTOLOAD) =~ s/.*://;
		1307
		1308	$method{$func}
		1309	or Carp::croak "$func: not a valid AnyEvent class method";
		1310
		1311	detect;
129		1312
130	my $class = shift;	1313	my $class = shift;
131	$class->$AUTOLOAD (@_);	1314	$class->$func (@_);
132	}	1315	}
		1316
		1317	# utility function to dup a filehandle. this is used by many backends
		1318	# to support binding more than one watcher per filehandle (they usually
		1319	# allow only one watcher per fd, so we dup it to get a different one).
		1320	sub _dupfh($$;$$) {
		1321	my ($poll, $fh, $r, $w) = @_;
		1322
		1323	# cygwin requires the fh mode to be matching, unix doesn't
		1324	my ($rw, $mode) = $poll eq "r" ? ($r, "<&") : ($w, ">&");
		1325
		1326	open my $fh2, $mode, $fh
		1327	or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,";
		1328
		1329	# we assume CLOEXEC is already set by perl in all important cases
		1330
		1331	($fh2, $rw)
		1332	}
		1333
		1334	=head1 SIMPLIFIED AE API
		1335
		1336	Starting with version 5.0, AnyEvent officially supports a second, much
		1337	simpler, API that is designed to reduce the calling, typing and memory
		1338	overhead by using function call syntax and a fixed number of parameters.
		1339
		1340	See the L<AE> manpage for details.
		1341
		1342	=cut
		1343
		1344	package AE;
		1345
		1346	our $VERSION = $AnyEvent::VERSION;
		1347
		1348	# fall back to the main API by default - backends and AnyEvent::Base
		1349	# implementations can overwrite these.
		1350
		1351	sub io($$$) {
		1352	AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2])
		1353	}
		1354
		1355	sub timer($$$) {
		1356	AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2])
		1357	}
		1358
		1359	sub signal($$) {
		1360	AnyEvent->signal (signal => $_[0], cb => $_[1])
		1361	}
		1362
		1363	sub child($$) {
		1364	AnyEvent->child (pid => $_[0], cb => $_[1])
		1365	}
		1366
		1367	sub idle($) {
		1368	AnyEvent->idle (cb => $_[0])
		1369	}
		1370
		1371	sub cv(;&) {
		1372	AnyEvent->condvar (@_ ? (cb => $_[0]) : ())
		1373	}
		1374
		1375	sub now() {
		1376	AnyEvent->now
		1377	}
		1378
		1379	sub now_update() {
		1380	AnyEvent->now_update
		1381	}
		1382
		1383	sub time() {
		1384	AnyEvent->time
		1385	}
		1386
		1387	package AnyEvent::Base;
		1388
		1389	# default implementations for many methods
		1390
		1391	sub time {
		1392	eval q{ # poor man's autoloading {}
		1393	# probe for availability of Time::HiRes
		1394	if (eval "use Time::HiRes (); Time::HiRes::time (); 1") {
		1395	warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8;
		1396	*AE::time = \&Time::HiRes::time;
		1397	# if (eval "use POSIX (); (POSIX::times())...
		1398	} else {
		1399	warn "AnyEvent: using built-in time(), WARNING, no sub-second resolution!\n" if $VERBOSE;
		1400	*AE::time = sub (){ time }; # epic fail
		1401	}
		1402
		1403	*time = sub { AE::time }; # different prototypes
		1404	};
		1405	die if $@;
		1406
		1407	&time
		1408	}
		1409
		1410	*now = \&time;
		1411
		1412	sub now_update { }
		1413
		1414	# default implementation for ->condvar
		1415
		1416	sub condvar {
		1417	eval q{ # poor man's autoloading {}
		1418	*condvar = sub {
		1419	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, "AnyEvent::CondVar"
		1420	};
		1421
		1422	*AE::cv = sub (;&) {
		1423	bless { @_ ? (_ae_cb => shift) : () }, "AnyEvent::CondVar"
		1424	};
		1425	};
		1426	die if $@;
		1427
		1428	&condvar
		1429	}
		1430
		1431	# default implementation for ->signal
		1432
		1433	our $HAVE_ASYNC_INTERRUPT;
		1434
		1435	sub _have_async_interrupt() {
		1436	$HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT}
		1437	&& eval "use Async::Interrupt 1.02 (); 1")
		1438	unless defined $HAVE_ASYNC_INTERRUPT;
		1439
		1440	$HAVE_ASYNC_INTERRUPT
		1441	}
		1442
		1443	our ($SIGPIPE_R, $SIGPIPE_W, %SIG_CB, %SIG_EV, $SIG_IO);
		1444	our (%SIG_ASY, %SIG_ASY_W);
		1445	our ($SIG_COUNT, $SIG_TW);
		1446
		1447	# install a dummy wakeup watcher to reduce signal catching latency
		1448	# used by Impls
		1449	sub _sig_add() {
		1450	unless ($SIG_COUNT++) {
		1451	# try to align timer on a full-second boundary, if possible
		1452	my $NOW = AE::now;
		1453
		1454	$SIG_TW = AE::timer
		1455	$MAX_SIGNAL_LATENCY - ($NOW - int $NOW),
		1456	$MAX_SIGNAL_LATENCY,
		1457	sub { } # just for the PERL_ASYNC_CHECK
		1458	;
		1459	}
		1460	}
		1461
		1462	sub _sig_del {
		1463	undef $SIG_TW
		1464	unless --$SIG_COUNT;
		1465	}
		1466
		1467	our $_sig_name_init; $_sig_name_init = sub {
		1468	eval q{ # poor man's autoloading {}
		1469	undef $_sig_name_init;
		1470
		1471	if (_have_async_interrupt) {
		1472	*sig2num = \&Async::Interrupt::sig2num;
		1473	*sig2name = \&Async::Interrupt::sig2name;
		1474	} else {
		1475	require Config;
		1476
		1477	my %signame2num;
		1478	@signame2num{ split ' ', $Config::Config{sig_name} }
		1479	= split ' ', $Config::Config{sig_num};
		1480
		1481	my @signum2name;
		1482	@signum2name[values %signame2num] = keys %signame2num;
		1483
		1484	*sig2num = sub($) {
		1485	$_[0] > 0 ? shift : $signame2num{+shift}
		1486	};
		1487	*sig2name = sub ($) {
		1488	$_[0] > 0 ? $signum2name[+shift] : shift
		1489	};
		1490	}
		1491	};
		1492	die if $@;
		1493	};
		1494
		1495	sub sig2num ($) { &$_sig_name_init; &sig2num }
		1496	sub sig2name($) { &$_sig_name_init; &sig2name }
		1497
		1498	sub signal {
		1499	eval q{ # poor man's autoloading {}
		1500	# probe for availability of Async::Interrupt
		1501	if (_have_async_interrupt) {
		1502	warn "AnyEvent: using Async::Interrupt for race-free signal handling.\n" if $VERBOSE >= 8;
		1503
		1504	$SIGPIPE_R = new Async::Interrupt::EventPipe;
		1505	$SIG_IO = AE::io $SIGPIPE_R->fileno, 0, \&_signal_exec;
		1506
		1507	} else {
		1508	warn "AnyEvent: using emulated perl signal handling with latency timer.\n" if $VERBOSE >= 8;
		1509
		1510	if (AnyEvent::WIN32) {
		1511	require AnyEvent::Util;
		1512
		1513	($SIGPIPE_R, $SIGPIPE_W) = AnyEvent::Util::portable_pipe ();
		1514	AnyEvent::Util::fh_nonblocking ($SIGPIPE_R, 1) if $SIGPIPE_R;
		1515	AnyEvent::Util::fh_nonblocking ($SIGPIPE_W, 1) if $SIGPIPE_W; # just in case
		1516	} else {
		1517	pipe $SIGPIPE_R, $SIGPIPE_W;
		1518	fcntl $SIGPIPE_R, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_R;
		1519	fcntl $SIGPIPE_W, AnyEvent::F_SETFL, AnyEvent::O_NONBLOCK if $SIGPIPE_W; # just in case
		1520
		1521	# not strictly required, as $^F is normally 2, but let's make sure...
		1522	fcntl $SIGPIPE_R, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
		1523	fcntl $SIGPIPE_W, AnyEvent::F_SETFD, AnyEvent::FD_CLOEXEC;
		1524	}
		1525
		1526	$SIGPIPE_R
		1527	or Carp::croak "AnyEvent: unable to create a signal reporting pipe: $!\n";
		1528
		1529	$SIG_IO = AE::io $SIGPIPE_R, 0, \&_signal_exec;
		1530	}
		1531
		1532	*signal = $HAVE_ASYNC_INTERRUPT
		1533	? sub {
		1534	my (undef, %arg) = @_;
		1535
		1536	# async::interrupt
		1537	my $signal = sig2num $arg{signal};
		1538	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1539
		1540	$SIG_ASY{$signal} ||= new Async::Interrupt
		1541	cb => sub { undef $SIG_EV{$signal} },
		1542	signal => $signal,
		1543	pipe => [$SIGPIPE_R->filenos],
		1544	pipe_autodrain => 0,
		1545	;
		1546
		1547	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
		1548	}
		1549	: sub {
		1550	my (undef, %arg) = @_;
		1551
		1552	# pure perl
		1553	my $signal = sig2name $arg{signal};
		1554	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		1555
		1556	$SIG{$signal} ||= sub {
		1557	local $!;
		1558	syswrite $SIGPIPE_W, "\x00", 1 unless %SIG_EV;
		1559	undef $SIG_EV{$signal};
		1560	};
		1561
		1562	# can't do signal processing without introducing races in pure perl,
		1563	# so limit the signal latency.
		1564	_sig_add;
		1565
		1566	bless [$signal, $arg{cb}], "AnyEvent::Base::signal"
		1567	}
		1568	;
		1569
		1570	*AnyEvent::Base::signal::DESTROY = sub {
		1571	my ($signal, $cb) = @{$_[0]};
		1572
		1573	_sig_del;
		1574
		1575	delete $SIG_CB{$signal}{$cb};
		1576
		1577	$HAVE_ASYNC_INTERRUPT
		1578	? delete $SIG_ASY{$signal}
		1579	: # delete doesn't work with older perls - they then
		1580	# print weird messages, or just unconditionally exit
		1581	# instead of getting the default action.
		1582	undef $SIG{$signal}
		1583	unless keys %{ $SIG_CB{$signal} };
		1584	};
		1585
		1586	*_signal_exec = sub {
		1587	$HAVE_ASYNC_INTERRUPT
		1588	? $SIGPIPE_R->drain
		1589	: sysread $SIGPIPE_R, (my $dummy), 9;
		1590
		1591	while (%SIG_EV) {
		1592	for (keys %SIG_EV) {
		1593	delete $SIG_EV{$_};
		1594	$_->() for values %{ $SIG_CB{$_} || {} };
		1595	}
		1596	}
		1597	};
		1598	};
		1599	die if $@;
		1600
		1601	&signal
		1602	}
		1603
		1604	# default implementation for ->child
		1605
		1606	our %PID_CB;
		1607	our $CHLD_W;
		1608	our $CHLD_DELAY_W;
		1609	our $WNOHANG;
		1610
		1611	# used by many Impl's
		1612	sub _emit_childstatus($$) {
		1613	my (undef, $rpid, $rstatus) = @_;
		1614
		1615	$_->($rpid, $rstatus)
		1616	for values %{ $PID_CB{$rpid} || {} },
		1617	values %{ $PID_CB{0} || {} };
		1618	}
		1619
		1620	sub child {
		1621	eval q{ # poor man's autoloading {}
		1622	*_sigchld = sub {
		1623	my $pid;
		1624
		1625	AnyEvent->_emit_childstatus ($pid, $?)
		1626	while ($pid = waitpid -1, $WNOHANG) > 0;
		1627	};
		1628
		1629	*child = sub {
		1630	my (undef, %arg) = @_;
		1631
		1632	defined (my $pid = $arg{pid} + 0)
		1633	or Carp::croak "required option 'pid' is missing";
		1634
		1635	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
		1636
		1637	# WNOHANG is almost cetrainly 1 everywhere
		1638	$WNOHANG ||= $^O =~ /^(?:openbsd|netbsd|linux|freebsd|cygwin|MSWin32)$/
		1639	? 1
		1640	: eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
		1641
		1642	unless ($CHLD_W) {
		1643	$CHLD_W = AE::signal CHLD => \&_sigchld;
		1644	# child could be a zombie already, so make at least one round
		1645	&_sigchld;
		1646	}
		1647
		1648	bless [$pid, $arg{cb}], "AnyEvent::Base::child"
		1649	};
		1650
		1651	*AnyEvent::Base::child::DESTROY = sub {
		1652	my ($pid, $cb) = @{$_[0]};
		1653
		1654	delete $PID_CB{$pid}{$cb};
		1655	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
		1656
		1657	undef $CHLD_W unless keys %PID_CB;
		1658	};
		1659	};
		1660	die if $@;
		1661
		1662	&child
		1663	}
		1664
		1665	# idle emulation is done by simply using a timer, regardless
		1666	# of whether the process is idle or not, and not letting
		1667	# the callback use more than 50% of the time.
		1668	sub idle {
		1669	eval q{ # poor man's autoloading {}
		1670	*idle = sub {
		1671	my (undef, %arg) = @_;
		1672
		1673	my ($cb, $w, $rcb) = $arg{cb};
		1674
		1675	$rcb = sub {
		1676	if ($cb) {
		1677	$w = _time;
		1678	&$cb;
		1679	$w = _time - $w;
		1680
		1681	# never use more then 50% of the time for the idle watcher,
		1682	# within some limits
		1683	$w = 0.0001 if $w < 0.0001;
		1684	$w = 5 if $w > 5;
		1685
		1686	$w = AE::timer $w, 0, $rcb;
		1687	} else {
		1688	# clean up...
		1689	undef $w;
		1690	undef $rcb;
		1691	}
		1692	};
		1693
		1694	$w = AE::timer 0.05, 0, $rcb;
		1695
		1696	bless \\$cb, "AnyEvent::Base::idle"
		1697	};
		1698
		1699	*AnyEvent::Base::idle::DESTROY = sub {
		1700	undef $${$_[0]};
		1701	};
		1702	};
		1703	die if $@;
		1704
		1705	&idle
		1706	}
		1707
		1708	package AnyEvent::CondVar;
		1709
		1710	our @ISA = AnyEvent::CondVar::Base::;
		1711
		1712	package AnyEvent::CondVar::Base;
		1713
		1714	#use overload
		1715	# '&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1716	# fallback => 1;
		1717
		1718	# save 300+ kilobytes by dirtily hardcoding overloading
		1719	${"AnyEvent::CondVar::Base::OVERLOAD"}{dummy}++; # Register with magic by touching.
		1720	*{'AnyEvent::CondVar::Base::()'} = sub { }; # "Make it findable via fetchmethod."
		1721	*{'AnyEvent::CondVar::Base::(&{}'} = sub { my $self = shift; sub { $self->send (@_) } }; # &{}
		1722	${'AnyEvent::CondVar::Base::()'} = 1; # fallback
		1723
		1724	our $WAITING;
		1725
		1726	sub _send {
		1727	# nop
		1728	}
		1729
		1730	sub send {
		1731	my $cv = shift;
		1732	$cv->{_ae_sent} = [@_];
		1733	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1734	$cv->_send;
		1735	}
		1736
		1737	sub croak {
		1738	$_[0]{_ae_croak} = $_[1];
		1739	$_[0]->send;
		1740	}
		1741
		1742	sub ready {
		1743	$_[0]{_ae_sent}
		1744	}
		1745
		1746	sub _wait {
		1747	$WAITING
		1748	and !$_[0]{_ae_sent}
		1749	and Carp::croak "AnyEvent::CondVar: recursive blocking wait detected";
		1750
		1751	local $WAITING = 1;
		1752	AnyEvent->one_event while !$_[0]{_ae_sent};
		1753	}
		1754
		1755	sub recv {
		1756	$_[0]->_wait;
		1757
		1758	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1759	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1760	}
		1761
		1762	sub cb {
		1763	my $cv = shift;
		1764
		1765	@_
		1766	and $cv->{_ae_cb} = shift
		1767	and $cv->{_ae_sent}
		1768	and (delete $cv->{_ae_cb})->($cv);
		1769
		1770	$cv->{_ae_cb}
		1771	}
		1772
		1773	sub begin {
		1774	++$_[0]{_ae_counter};
		1775	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1776	}
		1777
		1778	sub end {
		1779	return if --$_[0]{_ae_counter};
		1780	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1781	}
		1782
		1783	# undocumented/compatibility with pre-3.4
		1784	*broadcast = \&send;
		1785	*wait = \&_wait;
		1786
		1787	=head1 ERROR AND EXCEPTION HANDLING
		1788
		1789	In general, AnyEvent does not do any error handling - it relies on the
		1790	caller to do that if required. The L<AnyEvent::Strict> module (see also
		1791	the C<PERL_ANYEVENT_STRICT> environment variable, below) provides strict
		1792	checking of all AnyEvent methods, however, which is highly useful during
		1793	development.
		1794
		1795	As for exception handling (i.e. runtime errors and exceptions thrown while
		1796	executing a callback), this is not only highly event-loop specific, but
		1797	also not in any way wrapped by this module, as this is the job of the main
		1798	program.
		1799
		1800	The pure perl event loop simply re-throws the exception (usually
		1801	within C<< condvar->recv >>), the L<Event> and L<EV> modules call C<<
		1802	$Event/EV::DIED->() >>, L<Glib> uses C<< install_exception_handler >> and
		1803	so on.
		1804
		1805	=head1 ENVIRONMENT VARIABLES
		1806
		1807	The following environment variables are used by this module or its
		1808	submodules.
		1809
		1810	Note that AnyEvent will remove I<all> environment variables starting with
		1811	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		1812	enabled.
		1813
		1814	=over 4
		1815
		1816	=item C<PERL_ANYEVENT_VERBOSE>
		1817
		1818	By default, AnyEvent will be completely silent except in fatal
		1819	conditions. You can set this environment variable to make AnyEvent more
		1820	talkative.
		1821
		1822	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1823	conditions, such as not being able to load the event model specified by
		1824	C<PERL_ANYEVENT_MODEL>.
		1825
		1826	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1827	model it chooses.
		1828
		1829	When set to C<8> or higher, then AnyEvent will report extra information on
		1830	which optional modules it loads and how it implements certain features.
		1831
		1832	=item C<PERL_ANYEVENT_STRICT>
		1833
		1834	AnyEvent does not do much argument checking by default, as thorough
		1835	argument checking is very costly. Setting this variable to a true value
		1836	will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly
		1837	check the arguments passed to most method calls. If it finds any problems,
		1838	it will croak.
		1839
		1840	In other words, enables "strict" mode.
		1841
		1842	Unlike C<use strict> (or it's modern cousin, C<< use L<common::sense>
		1843	>>, it is definitely recommended to keep it off in production. Keeping
		1844	C<PERL_ANYEVENT_STRICT=1> in your environment while developing programs
		1845	can be very useful, however.
		1846
		1847	=item C<PERL_ANYEVENT_MODEL>
		1848
		1849	This can be used to specify the event model to be used by AnyEvent, before
		1850	auto detection and -probing kicks in. It must be a string consisting
		1851	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1852	and the resulting module name is loaded and if the load was successful,
		1853	used as event model. If it fails to load AnyEvent will proceed with
		1854	auto detection and -probing.
		1855
		1856	This functionality might change in future versions.
		1857
		1858	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1859	could start your program like this:
		1860
		1861	PERL_ANYEVENT_MODEL=Perl perl ...
		1862
		1863	=item C<PERL_ANYEVENT_PROTOCOLS>
		1864
		1865	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1866	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1867	of auto probing).
		1868
		1869	Must be set to a comma-separated list of protocols or address families,
		1870	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1871	used, and preference will be given to protocols mentioned earlier in the
		1872	list.
		1873
		1874	This variable can effectively be used for denial-of-service attacks
		1875	against local programs (e.g. when setuid), although the impact is likely
		1876	small, as the program has to handle conenction and other failures anyways.
		1877
		1878	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1879	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1880	- only support IPv4, never try to resolve or contact IPv6
		1881	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1882	IPv6, but prefer IPv6 over IPv4.
		1883
		1884	=item C<PERL_ANYEVENT_EDNS0>
		1885
		1886	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1887	for DNS. This extension is generally useful to reduce DNS traffic, but
		1888	some (broken) firewalls drop such DNS packets, which is why it is off by
		1889	default.
		1890
		1891	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1892	EDNS0 in its DNS requests.
		1893
		1894	=item C<PERL_ANYEVENT_MAX_FORKS>
		1895
		1896	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1897	will create in parallel.
		1898
		1899	=item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS>
		1900
		1901	The default value for the C<max_outstanding> parameter for the default DNS
		1902	resolver - this is the maximum number of parallel DNS requests that are
		1903	sent to the DNS server.
		1904
		1905	=item C<PERL_ANYEVENT_RESOLV_CONF>
		1906
		1907	The file to use instead of F</etc/resolv.conf> (or OS-specific
		1908	configuration) in the default resolver. When set to the empty string, no
		1909	default config will be used.
		1910
		1911	=item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>.
		1912
		1913	When neither C<ca_file> nor C<ca_path> was specified during
		1914	L<AnyEvent::TLS> context creation, and either of these environment
		1915	variables exist, they will be used to specify CA certificate locations
		1916	instead of a system-dependent default.
		1917
		1918	=item C<PERL_ANYEVENT_AVOID_GUARD> and C<PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT>
		1919
		1920	When these are set to C<1>, then the respective modules are not
		1921	loaded. Mostly good for testing AnyEvent itself.
133		1922
134	=back	1923	=back
135		1924
136	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE	1925	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1926
		1927	This is an advanced topic that you do not normally need to use AnyEvent in
		1928	a module. This section is only of use to event loop authors who want to
		1929	provide AnyEvent compatibility.
137		1930
138	If you need to support another event library which isn't directly	1931	If you need to support another event library which isn't directly
139	supported by AnyEvent, you can supply your own interface to it by	1932	supported by AnyEvent, you can supply your own interface to it by
140	pushing, before the first watcher gets created, the package name of	1933	pushing, before the first watcher gets created, the package name of
141	the event module and the package name of the interface to use onto	1934	the event module and the package name of the interface to use onto
142	C<@AnyEvent::REGISTRY>. You can do that before and even without loading	1935	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
143	AnyEvent.	1936	AnyEvent, so it is reasonably cheap.
144		1937
145	Example:	1938	Example:
146		1939
147	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];	1940	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
148		1941
149	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>	1942	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
150	package/class when it finds the C<urxvt> package/module is loaded. When	1943	package/class when it finds the C<urxvt> package/module is already loaded.
		1944
151	AnyEvent is loaded and asked to find a suitable event model, it will	1945	When AnyEvent is loaded and asked to find a suitable event model, it
152	first check for the presence of urxvt.	1946	will first check for the presence of urxvt by trying to C<use> the
		1947	C<urxvt::anyevent> module.
153		1948
154	The class should prove implementations for all watcher types (see	1949	The class should provide implementations for all watcher types. See
155	L<AnyEvent::Impl::Event> (source code), L<AnyEvent::Impl::Glib>	1950	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
156	(Source code) and so on for actual examples, use C<perldoc -m	1951	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
157	AnyEvent::Impl::Glib> to see the sources).	1952	see the sources.
158		1953
		1954	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1955	provide suitable (hopefully) replacements.
		1956
159	The above isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)	1957	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
160	uses the above line as-is. An interface isn't included in AnyEvent	1958	terminal emulator uses the above line as-is. An interface isn't included
161	because it doesn't make sense outside the embedded interpreter inside	1959	in AnyEvent because it doesn't make sense outside the embedded interpreter
162	I<rxvt-unicode>, and it is updated and maintained as part of the	1960	inside I<rxvt-unicode>, and it is updated and maintained as part of the
163	I<rxvt-unicode> distribution.	1961	I<rxvt-unicode> distribution.
164		1962
165	I<rxvt-unicode> also cheats a bit by not providing blocking access to	1963	I<rxvt-unicode> also cheats a bit by not providing blocking access to
166	condition variables: code blocking while waiting for a condition will	1964	condition variables: code blocking while waiting for a condition will
167	C<die>. This still works with most modules/usages, and blocking calls must	1965	C<die>. This still works with most modules/usages, and blocking calls must
168	not be in an interactive appliation, so it makes sense.	1966	not be done in an interactive application, so it makes sense.
169		1967
170	=head1 ENVIRONMENT VARIABLES
171
172	The following environment variables are used by this module:
173
174	C<PERL_ANYEVENT_VERBOSE> when set to C<2> or higher, reports which event
175	model gets used.
176
177	=head1 EXAMPLE	1968	=head1 EXAMPLE PROGRAM
178		1969
179	The following program uses an io watcher to read data from stdin, a timer	1970	The following program uses an I/O watcher to read data from STDIN, a timer
180	to display a message once per second, and a condvar to exit the program	1971	to display a message once per second, and a condition variable to quit the
181	when the user enters quit:	1972	program when the user enters quit:
182		1973
183	use AnyEvent;	1974	use AnyEvent;
184		1975
185	my $cv = AnyEvent->condvar;	1976	my $cv = AnyEvent->condvar;
186		1977
187	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	1978	my $io_watcher = AnyEvent->io (
		1979	fh => \*STDIN,
		1980	poll => 'r',
		1981	cb => sub {
188	warn "io event <$_[0]>\n"; # will always output <r>	1982	warn "io event <$_[0]>\n"; # will always output <r>
189	chomp (my $input = <STDIN>); # read a line	1983	chomp (my $input = <STDIN>); # read a line
190	warn "read: $input\n"; # output what has been read	1984	warn "read: $input\n"; # output what has been read
191	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1985	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1986	},
		1987	);
		1988
		1989	my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub {
		1990	warn "timeout\n"; # print 'timeout' at most every second
192	});	1991	});
193		1992
194	my $time_watcher; # can only be used once
195
196	sub new_timer {
197	$timer = AnyEvent->timer (after => 1, cb => sub {
198	warn "timeout\n"; # print 'timeout' about every second
199	&new_timer; # and restart the time
200	});
201	}
202
203	new_timer; # create first timer
204
205	$cv->wait; # wait until user enters /^q/i	1993	$cv->recv; # wait until user enters /^q/i
206		1994
207	=head1 REAL-WORLD EXAMPLE	1995	=head1 REAL-WORLD EXAMPLE
208		1996
209	Consider the L<Net::FCP> module. It features (among others) the following	1997	Consider the L<Net::FCP> module. It features (among others) the following
210	API calls, which are to freenet what HTTP GET requests are to http:	1998	API calls, which are to freenet what HTTP GET requests are to http:
…		…
260	syswrite $txn->{fh}, $txn->{request}	2048	syswrite $txn->{fh}, $txn->{request}
261	or die "connection or write error";	2049	or die "connection or write error";
262	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	2050	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
263		2051
264	Again, C<fh_ready_r> waits till all data has arrived, and then stores the	2052	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
265	result and signals any possible waiters that the request ahs finished:	2053	result and signals any possible waiters that the request has finished:
266		2054
267	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	2055	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
268		2056
269	if (end-of-file or data complete) {	2057	if (end-of-file or data complete) {
270	$txn->{result} = $txn->{buf};	2058	$txn->{result} = $txn->{buf};
271	$txn->{finished}->broadcast;	2059	$txn->{finished}->send;
272	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	2060	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
273	}	2061	}
274		2062
275	The C<result> method, finally, just waits for the finished signal (if the	2063	The C<result> method, finally, just waits for the finished signal (if the
276	request was already finished, it doesn't wait, of course, and returns the	2064	request was already finished, it doesn't wait, of course, and returns the
277	data:	2065	data:
278		2066
279	$txn->{finished}->wait;	2067	$txn->{finished}->recv;
280	return $txn->{result};	2068	return $txn->{result};
281		2069
282	The actual code goes further and collects all errors (C<die>s, exceptions)	2070	The actual code goes further and collects all errors (C<die>s, exceptions)
283	that occured during request processing. The C<result> method detects	2071	that occurred during request processing. The C<result> method detects
284	wether an exception as thrown (it is stored inside the $txn object)	2072	whether an exception as thrown (it is stored inside the $txn object)
285	and just throws the exception, which means connection errors and other	2073	and just throws the exception, which means connection errors and other
286	problems get reported tot he code that tries to use the result, not in a	2074	problems get reported to the code that tries to use the result, not in a
287	random callback.	2075	random callback.
288		2076
289	All of this enables the following usage styles:	2077	All of this enables the following usage styles:
290		2078
291	1. Blocking:	2079	1. Blocking:
292		2080
293	my $data = $fcp->client_get ($url);	2081	my $data = $fcp->client_get ($url);
294		2082
295	2. Blocking, but parallelizing:	2083	2. Blocking, but running in parallel:
296		2084
297	my @datas = map $_->result,	2085	my @datas = map $_->result,
298	map $fcp->txn_client_get ($_),	2086	map $fcp->txn_client_get ($_),
299	@urls;	2087	@urls;
300		2088
301	Both blocking examples work without the module user having to know	2089	Both blocking examples work without the module user having to know
302	anything about events.	2090	anything about events.
303		2091
304	3a. Event-based in a main program, using any support Event module:	2092	3a. Event-based in a main program, using any supported event module:
305		2093
306	use Event;	2094	use EV;
307		2095
308	$fcp->txn_client_get ($url)->cb (sub {	2096	$fcp->txn_client_get ($url)->cb (sub {
309	my $txn = shift;	2097	my $txn = shift;
310	my $data = $txn->result;	2098	my $data = $txn->result;
311	...	2099	...
312	});	2100	});
313		2101
314	Event::loop;	2102	EV::loop;
315		2103
316	3b. The module user could use AnyEvent, too:	2104	3b. The module user could use AnyEvent, too:
317		2105
318	use AnyEvent;	2106	use AnyEvent;
319		2107
320	my $quit = AnyEvent->condvar;	2108	my $quit = AnyEvent->condvar;
321		2109
322	$fcp->txn_client_get ($url)->cb (sub {	2110	$fcp->txn_client_get ($url)->cb (sub {
323	...	2111	...
324	$quit->broadcast;	2112	$quit->send;
325	});	2113	});
326		2114
327	$quit->wait;	2115	$quit->recv;
		2116
		2117
		2118	=head1 BENCHMARKS
		2119
		2120	To give you an idea of the performance and overheads that AnyEvent adds
		2121	over the event loops themselves and to give you an impression of the speed
		2122	of various event loops I prepared some benchmarks.
		2123
		2124	=head2 BENCHMARKING ANYEVENT OVERHEAD
		2125
		2126	Here is a benchmark of various supported event models used natively and
		2127	through AnyEvent. The benchmark creates a lot of timers (with a zero
		2128	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		2129	which it is), lets them fire exactly once and destroys them again.
		2130
		2131	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		2132	distribution. It uses the L<AE> interface, which makes a real difference
		2133	for the EV and Perl backends only.
		2134
		2135	=head3 Explanation of the columns
		2136
		2137	I<watcher> is the number of event watchers created/destroyed. Since
		2138	different event models feature vastly different performances, each event
		2139	loop was given a number of watchers so that overall runtime is acceptable
		2140	and similar between tested event loop (and keep them from crashing): Glib
		2141	would probably take thousands of years if asked to process the same number
		2142	of watchers as EV in this benchmark.
		2143
		2144	I<bytes> is the number of bytes (as measured by the resident set size,
		2145	RSS) consumed by each watcher. This method of measuring captures both C
		2146	and Perl-based overheads.
		2147
		2148	I<create> is the time, in microseconds (millionths of seconds), that it
		2149	takes to create a single watcher. The callback is a closure shared between
		2150	all watchers, to avoid adding memory overhead. That means closure creation
		2151	and memory usage is not included in the figures.
		2152
		2153	I<invoke> is the time, in microseconds, used to invoke a simple
		2154	callback. The callback simply counts down a Perl variable and after it was
		2155	invoked "watcher" times, it would C<< ->send >> a condvar once to
		2156	signal the end of this phase.
		2157
		2158	I<destroy> is the time, in microseconds, that it takes to destroy a single
		2159	watcher.
		2160
		2161	=head3 Results
		2162
		2163	name watchers bytes create invoke destroy comment
		2164	EV/EV 100000 223 0.47 0.43 0.27 EV native interface
		2165	EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers
		2166	Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal
		2167	Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation
		2168	Event/Event 16000 516 31.16 31.84 0.82 Event native interface
		2169	Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers
		2170	IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll
		2171	IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll
		2172	Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour
		2173	Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers
		2174	POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event
		2175	POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select
		2176
		2177	=head3 Discussion
		2178
		2179	The benchmark does I<not> measure scalability of the event loop very
		2180	well. For example, a select-based event loop (such as the pure perl one)
		2181	can never compete with an event loop that uses epoll when the number of
		2182	file descriptors grows high. In this benchmark, all events become ready at
		2183	the same time, so select/poll-based implementations get an unnatural speed
		2184	boost.
		2185
		2186	Also, note that the number of watchers usually has a nonlinear effect on
		2187	overall speed, that is, creating twice as many watchers doesn't take twice
		2188	the time - usually it takes longer. This puts event loops tested with a
		2189	higher number of watchers at a disadvantage.
		2190
		2191	To put the range of results into perspective, consider that on the
		2192	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		2193	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		2194	cycles with POE.
		2195
		2196	C<EV> is the sole leader regarding speed and memory use, which are both
		2197	maximal/minimal, respectively. When using the L<AE> API there is zero
		2198	overhead (when going through the AnyEvent API create is about 5-6 times
		2199	slower, with other times being equal, so still uses far less memory than
		2200	any other event loop and is still faster than Event natively).
		2201
		2202	The pure perl implementation is hit in a few sweet spots (both the
		2203	constant timeout and the use of a single fd hit optimisations in the perl
		2204	interpreter and the backend itself). Nevertheless this shows that it
		2205	adds very little overhead in itself. Like any select-based backend its
		2206	performance becomes really bad with lots of file descriptors (and few of
		2207	them active), of course, but this was not subject of this benchmark.
		2208
		2209	The C<Event> module has a relatively high setup and callback invocation
		2210	cost, but overall scores in on the third place.
		2211
		2212	C<IO::Async> performs admirably well, about on par with C<Event>, even
		2213	when using its pure perl backend.
		2214
		2215	C<Glib>'s memory usage is quite a bit higher, but it features a
		2216	faster callback invocation and overall ends up in the same class as
		2217	C<Event>. However, Glib scales extremely badly, doubling the number of
		2218	watchers increases the processing time by more than a factor of four,
		2219	making it completely unusable when using larger numbers of watchers
		2220	(note that only a single file descriptor was used in the benchmark, so
		2221	inefficiencies of C<poll> do not account for this).
		2222
		2223	The C<Tk> adaptor works relatively well. The fact that it crashes with
		2224	more than 2000 watchers is a big setback, however, as correctness takes
		2225	precedence over speed. Nevertheless, its performance is surprising, as the
		2226	file descriptor is dup()ed for each watcher. This shows that the dup()
		2227	employed by some adaptors is not a big performance issue (it does incur a
		2228	hidden memory cost inside the kernel which is not reflected in the figures
		2229	above).
		2230
		2231	C<POE>, regardless of underlying event loop (whether using its pure perl
		2232	select-based backend or the Event module, the POE-EV backend couldn't
		2233	be tested because it wasn't working) shows abysmal performance and
		2234	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		2235	as EV watchers, and 10 times as much memory as Event (the high memory
		2236	requirements are caused by requiring a session for each watcher). Watcher
		2237	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		2238	implementation.
		2239
		2240	The design of the POE adaptor class in AnyEvent can not really account
		2241	for the performance issues, though, as session creation overhead is
		2242	small compared to execution of the state machine, which is coded pretty
		2243	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		2244	using multiple sessions is not a good approach, especially regarding
		2245	memory usage, even the author of POE could not come up with a faster
		2246	design).
		2247
		2248	=head3 Summary
		2249
		2250	=over 4
		2251
		2252	=item * Using EV through AnyEvent is faster than any other event loop
		2253	(even when used without AnyEvent), but most event loops have acceptable
		2254	performance with or without AnyEvent.
		2255
		2256	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		2257	the actual event loop, only with extremely fast event loops such as EV
		2258	adds AnyEvent significant overhead.
		2259
		2260	=item * You should avoid POE like the plague if you want performance or
		2261	reasonable memory usage.
		2262
		2263	=back
		2264
		2265	=head2 BENCHMARKING THE LARGE SERVER CASE
		2266
		2267	This benchmark actually benchmarks the event loop itself. It works by
		2268	creating a number of "servers": each server consists of a socket pair, a
		2269	timeout watcher that gets reset on activity (but never fires), and an I/O
		2270	watcher waiting for input on one side of the socket. Each time the socket
		2271	watcher reads a byte it will write that byte to a random other "server".
		2272
		2273	The effect is that there will be a lot of I/O watchers, only part of which
		2274	are active at any one point (so there is a constant number of active
		2275	fds for each loop iteration, but which fds these are is random). The
		2276	timeout is reset each time something is read because that reflects how
		2277	most timeouts work (and puts extra pressure on the event loops).
		2278
		2279	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		2280	(1%) are active. This mirrors the activity of large servers with many
		2281	connections, most of which are idle at any one point in time.
		2282
		2283	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		2284	distribution. It uses the L<AE> interface, which makes a real difference
		2285	for the EV and Perl backends only.
		2286
		2287	=head3 Explanation of the columns
		2288
		2289	I<sockets> is the number of sockets, and twice the number of "servers" (as
		2290	each server has a read and write socket end).
		2291
		2292	I<create> is the time it takes to create a socket pair (which is
		2293	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		2294
		2295	I<request>, the most important value, is the time it takes to handle a
		2296	single "request", that is, reading the token from the pipe and forwarding
		2297	it to another server. This includes deleting the old timeout and creating
		2298	a new one that moves the timeout into the future.
		2299
		2300	=head3 Results
		2301
		2302	name sockets create request
		2303	EV 20000 62.66 7.99
		2304	Perl 20000 68.32 32.64
		2305	IOAsync 20000 174.06 101.15 epoll
		2306	IOAsync 20000 174.67 610.84 poll
		2307	Event 20000 202.69 242.91
		2308	Glib 20000 557.01 1689.52
		2309	POE 20000 341.54 12086.32 uses POE::Loop::Event
		2310
		2311	=head3 Discussion
		2312
		2313	This benchmark I<does> measure scalability and overall performance of the
		2314	particular event loop.
		2315
		2316	EV is again fastest. Since it is using epoll on my system, the setup time
		2317	is relatively high, though.
		2318
		2319	Perl surprisingly comes second. It is much faster than the C-based event
		2320	loops Event and Glib.
		2321
		2322	IO::Async performs very well when using its epoll backend, and still quite
		2323	good compared to Glib when using its pure perl backend.
		2324
		2325	Event suffers from high setup time as well (look at its code and you will
		2326	understand why). Callback invocation also has a high overhead compared to
		2327	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		2328	uses select or poll in basically all documented configurations.
		2329
		2330	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		2331	clearly fails to perform with many filehandles or in busy servers.
		2332
		2333	POE is still completely out of the picture, taking over 1000 times as long
		2334	as EV, and over 100 times as long as the Perl implementation, even though
		2335	it uses a C-based event loop in this case.
		2336
		2337	=head3 Summary
		2338
		2339	=over 4
		2340
		2341	=item * The pure perl implementation performs extremely well.
		2342
		2343	=item * Avoid Glib or POE in large projects where performance matters.
		2344
		2345	=back
		2346
		2347	=head2 BENCHMARKING SMALL SERVERS
		2348
		2349	While event loops should scale (and select-based ones do not...) even to
		2350	large servers, most programs we (or I :) actually write have only a few
		2351	I/O watchers.
		2352
		2353	In this benchmark, I use the same benchmark program as in the large server
		2354	case, but it uses only eight "servers", of which three are active at any
		2355	one time. This should reflect performance for a small server relatively
		2356	well.
		2357
		2358	The columns are identical to the previous table.
		2359
		2360	=head3 Results
		2361
		2362	name sockets create request
		2363	EV 16 20.00 6.54
		2364	Perl 16 25.75 12.62
		2365	Event 16 81.27 35.86
		2366	Glib 16 32.63 15.48
		2367	POE 16 261.87 276.28 uses POE::Loop::Event
		2368
		2369	=head3 Discussion
		2370
		2371	The benchmark tries to test the performance of a typical small
		2372	server. While knowing how various event loops perform is interesting, keep
		2373	in mind that their overhead in this case is usually not as important, due
		2374	to the small absolute number of watchers (that is, you need efficiency and
		2375	speed most when you have lots of watchers, not when you only have a few of
		2376	them).
		2377
		2378	EV is again fastest.
		2379
		2380	Perl again comes second. It is noticeably faster than the C-based event
		2381	loops Event and Glib, although the difference is too small to really
		2382	matter.
		2383
		2384	POE also performs much better in this case, but is is still far behind the
		2385	others.
		2386
		2387	=head3 Summary
		2388
		2389	=over 4
		2390
		2391	=item * C-based event loops perform very well with small number of
		2392	watchers, as the management overhead dominates.
		2393
		2394	=back
		2395
		2396	=head2 THE IO::Lambda BENCHMARK
		2397
		2398	Recently I was told about the benchmark in the IO::Lambda manpage, which
		2399	could be misinterpreted to make AnyEvent look bad. In fact, the benchmark
		2400	simply compares IO::Lambda with POE, and IO::Lambda looks better (which
		2401	shouldn't come as a surprise to anybody). As such, the benchmark is
		2402	fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't
		2403	very optimal. But how would AnyEvent compare when used without the extra
		2404	baggage? To explore this, I wrote the equivalent benchmark for AnyEvent.
		2405
		2406	The benchmark itself creates an echo-server, and then, for 500 times,
		2407	connects to the echo server, sends a line, waits for the reply, and then
		2408	creates the next connection. This is a rather bad benchmark, as it doesn't
		2409	test the efficiency of the framework or much non-blocking I/O, but it is a
		2410	benchmark nevertheless.
		2411
		2412	name runtime
		2413	Lambda/select 0.330 sec
		2414	+ optimized 0.122 sec
		2415	Lambda/AnyEvent 0.327 sec
		2416	+ optimized 0.138 sec
		2417	Raw sockets/select 0.077 sec
		2418	POE/select, components 0.662 sec
		2419	POE/select, raw sockets 0.226 sec
		2420	POE/select, optimized 0.404 sec
		2421
		2422	AnyEvent/select/nb 0.085 sec
		2423	AnyEvent/EV/nb 0.068 sec
		2424	+state machine 0.134 sec
		2425
		2426	The benchmark is also a bit unfair (my fault): the IO::Lambda/POE
		2427	benchmarks actually make blocking connects and use 100% blocking I/O,
		2428	defeating the purpose of an event-based solution. All of the newly
		2429	written AnyEvent benchmarks use 100% non-blocking connects (using
		2430	AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS
		2431	resolver), so AnyEvent is at a disadvantage here, as non-blocking connects
		2432	generally require a lot more bookkeeping and event handling than blocking
		2433	connects (which involve a single syscall only).
		2434
		2435	The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which
		2436	offers similar expressive power as POE and IO::Lambda, using conventional
		2437	Perl syntax. This means that both the echo server and the client are 100%
		2438	non-blocking, further placing it at a disadvantage.
		2439
		2440	As you can see, the AnyEvent + EV combination even beats the
		2441	hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl
		2442	backend easily beats IO::Lambda and POE.
		2443
		2444	And even the 100% non-blocking version written using the high-level (and
		2445	slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda
		2446	higher level ("unoptimised") abstractions by a large margin, even though
		2447	it does all of DNS, tcp-connect and socket I/O in a non-blocking way.
		2448
		2449	The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and
		2450	F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are
		2451	part of the IO::Lambda distribution and were used without any changes.
		2452
		2453
		2454	=head1 SIGNALS
		2455
		2456	AnyEvent currently installs handlers for these signals:
		2457
		2458	=over 4
		2459
		2460	=item SIGCHLD
		2461
		2462	A handler for C<SIGCHLD> is installed by AnyEvent's child watcher
		2463	emulation for event loops that do not support them natively. Also, some
		2464	event loops install a similar handler.
		2465
		2466	Additionally, when AnyEvent is loaded and SIGCHLD is set to IGNORE, then
		2467	AnyEvent will reset it to default, to avoid losing child exit statuses.
		2468
		2469	=item SIGPIPE
		2470
		2471	A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef>
		2472	when AnyEvent gets loaded.
		2473
		2474	The rationale for this is that AnyEvent users usually do not really depend
		2475	on SIGPIPE delivery (which is purely an optimisation for shell use, or
		2476	badly-written programs), but C<SIGPIPE> can cause spurious and rare
		2477	program exits as a lot of people do not expect C<SIGPIPE> when writing to
		2478	some random socket.
		2479
		2480	The rationale for installing a no-op handler as opposed to ignoring it is
		2481	that this way, the handler will be restored to defaults on exec.
		2482
		2483	Feel free to install your own handler, or reset it to defaults.
		2484
		2485	=back
		2486
		2487	=cut
		2488
		2489	undef $SIG{CHLD}
		2490	if $SIG{CHLD} eq 'IGNORE';
		2491
		2492	$SIG{PIPE} = sub { }
		2493	unless defined $SIG{PIPE};
		2494
		2495	=head1 RECOMMENDED/OPTIONAL MODULES
		2496
		2497	One of AnyEvent's main goals is to be 100% Pure-Perl(tm): only perl (and
		2498	it's built-in modules) are required to use it.
		2499
		2500	That does not mean that AnyEvent won't take advantage of some additional
		2501	modules if they are installed.
		2502
		2503	This section explains which additional modules will be used, and how they
		2504	affect AnyEvent's operation.
		2505
		2506	=over 4
		2507
		2508	=item L<Async::Interrupt>
		2509
		2510	This slightly arcane module is used to implement fast signal handling: To
		2511	my knowledge, there is no way to do completely race-free and quick
		2512	signal handling in pure perl. To ensure that signals still get
		2513	delivered, AnyEvent will start an interval timer to wake up perl (and
		2514	catch the signals) with some delay (default is 10 seconds, look for
		2515	C<$AnyEvent::MAX_SIGNAL_LATENCY>).
		2516
		2517	If this module is available, then it will be used to implement signal
		2518	catching, which means that signals will not be delayed, and the event loop
		2519	will not be interrupted regularly, which is more efficient (and good for
		2520	battery life on laptops).
		2521
		2522	This affects not just the pure-perl event loop, but also other event loops
		2523	that have no signal handling on their own (e.g. Glib, Tk, Qt).
		2524
		2525	Some event loops (POE, Event, Event::Lib) offer signal watchers natively,
		2526	and either employ their own workarounds (POE) or use AnyEvent's workaround
		2527	(using C<$AnyEvent::MAX_SIGNAL_LATENCY>). Installing L<Async::Interrupt>
		2528	does nothing for those backends.
		2529
		2530	=item L<EV>
		2531
		2532	This module isn't really "optional", as it is simply one of the backend
		2533	event loops that AnyEvent can use. However, it is simply the best event
		2534	loop available in terms of features, speed and stability: It supports
		2535	the AnyEvent API optimally, implements all the watcher types in XS, does
		2536	automatic timer adjustments even when no monotonic clock is available,
		2537	can take avdantage of advanced kernel interfaces such as C<epoll> and
		2538	C<kqueue>, and is the fastest backend I<by far>. You can even embed
		2539	L<Glib>/L<Gtk2> in it (or vice versa, see L<EV::Glib> and L<Glib::EV>).
		2540
		2541	If you only use backends that rely on another event loop (e.g. C<Tk>),
		2542	then this module will do nothing for you.
		2543
		2544	=item L<Guard>
		2545
		2546	The guard module, when used, will be used to implement
		2547	C<AnyEvent::Util::guard>. This speeds up guards considerably (and uses a
		2548	lot less memory), but otherwise doesn't affect guard operation much. It is
		2549	purely used for performance.
		2550
		2551	=item L<JSON> and L<JSON::XS>
		2552
		2553	One of these modules is required when you want to read or write JSON data
		2554	via L<AnyEvent::Handle>. L<JSON> is also written in pure-perl, but can take
		2555	advantage of the ultra-high-speed L<JSON::XS> module when it is installed.
		2556
		2557	=item L<Net::SSLeay>
		2558
		2559	Implementing TLS/SSL in Perl is certainly interesting, but not very
		2560	worthwhile: If this module is installed, then L<AnyEvent::Handle> (with
		2561	the help of L<AnyEvent::TLS>), gains the ability to do TLS/SSL.
		2562
		2563	=item L<Time::HiRes>
		2564
		2565	This module is part of perl since release 5.008. It will be used when the
		2566	chosen event library does not come with a timing source on it's own. The
		2567	pure-perl event loop (L<AnyEvent::Impl::Perl>) will additionally use it to
		2568	try to use a monotonic clock for timing stability.
		2569
		2570	=back
		2571
		2572
		2573	=head1 FORK
		2574
		2575	Most event libraries are not fork-safe. The ones who are usually are
		2576	because they rely on inefficient but fork-safe C<select> or C<poll> calls
		2577	- higher performance APIs such as BSD's kqueue or the dreaded Linux epoll
		2578	are usually badly thought-out hacks that are incompatible with fork in
		2579	one way or another. Only L<EV> is fully fork-aware and ensures that you
		2580	continue event-processing in both parent and child (or both, if you know
		2581	what you are doing).
		2582
		2583	This means that, in general, you cannot fork and do event processing in
		2584	the child if the event library was initialised before the fork (which
		2585	usually happens when the first AnyEvent watcher is created, or the library
		2586	is loaded).
		2587
		2588	If you have to fork, you must either do so I<before> creating your first
		2589	watcher OR you must not use AnyEvent at all in the child OR you must do
		2590	something completely out of the scope of AnyEvent.
		2591
		2592	The problem of doing event processing in the parent I<and> the child
		2593	is much more complicated: even for backends that I<are> fork-aware or
		2594	fork-safe, their behaviour is not usually what you want: fork clones all
		2595	watchers, that means all timers, I/O watchers etc. are active in both
		2596	parent and child, which is almost never what you want. USing C<exec>
		2597	to start worker children from some kind of manage rprocess is usually
		2598	preferred, because it is much easier and cleaner, at the expense of having
		2599	to have another binary.
		2600
		2601
		2602	=head1 SECURITY CONSIDERATIONS
		2603
		2604	AnyEvent can be forced to load any event model via
		2605	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		2606	execute arbitrary code or directly gain access, it can easily be used to
		2607	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		2608	will not be active when the program uses a different event model than
		2609	specified in the variable.
		2610
		2611	You can make AnyEvent completely ignore this variable by deleting it
		2612	before the first watcher gets created, e.g. with a C<BEGIN> block:
		2613
		2614	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		2615
		2616	use AnyEvent;
		2617
		2618	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		2619	be used to probe what backend is used and gain other information (which is
		2620	probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and
		2621	$ENV{PERL_ANYEVENT_STRICT}.
		2622
		2623	Note that AnyEvent will remove I<all> environment variables starting with
		2624	C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is
		2625	enabled.
		2626
		2627
		2628	=head1 BUGS
		2629
		2630	Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
		2631	to work around. If you suffer from memleaks, first upgrade to Perl 5.10
		2632	and check wether the leaks still show up. (Perl 5.10.0 has other annoying
		2633	memleaks, such as leaking on C<map> and C<grep> but it is usually not as
		2634	pronounced).
		2635
328		2636
329	=head1 SEE ALSO	2637	=head1 SEE ALSO
330		2638
331	Event modules: L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>.	2639	Utility functions: L<AnyEvent::Util>.
332		2640
333	Implementations: L<AnyEvent::Impl::Coro>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>.	2641	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
		2642	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
334		2643
335	Nontrivial usage example: L<Net::FCP>.	2644	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		2645	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		2646	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		2647	L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>, L<Anyevent::Impl::Irssi>.
336		2648
337	=head1	2649	Non-blocking file handles, sockets, TCP clients and
		2650	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>.
		2651
		2652	Asynchronous DNS: L<AnyEvent::DNS>.
		2653
		2654	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>,
		2655	L<Coro::Event>,
		2656
		2657	Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::XMPP>,
		2658	L<AnyEvent::HTTP>.
		2659
		2660
		2661	=head1 AUTHOR
		2662
		2663	Marc Lehmann <schmorp@schmorp.de>
		2664	http://home.schmorp.de/
338		2665
339	=cut	2666	=cut
340		2667
341	1	2668	1
342		2669

Diff Legend

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