ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.4 by root, Thu Dec 1 22:04:50 2005 UTC vs.
Revision 1.142 by root, Tue May 27 02:34:30 2008 UTC

1	=head1 NAME	1	=head1 => NAME
2		2
3	AnyEvent - provide framework for multiple event loops	3	AnyEvent - provide framework for multiple event loops
4		4
5	Event, Coro, Glib, Tk - various supported event loops	5	EV, Event, Glib, Tk, Perl, Event::Lib, Qt, POE - various supported event loops
6		6
7	=head1 SYNOPSIS	7	=head1 SYNOPSIS
8		8
9	use AnyEvent;	9	use AnyEvent;
10		10
11	my $w = AnyEvent->timer (fh => ..., poll => "[rw]+", cb => sub {	11	my $w = AnyEvent->io (fh => $fh, poll => "r|w", cb => sub {
12	my ($poll_got) = @_;
13	...	12	...
14	});	13	});
		14
15	my $w = AnyEvent->io (after => $seconds, cb => sub {	15	my $w = AnyEvent->timer (after => $seconds, cb => sub {
16	...	16	...
17	});	17	});
18		18
19	# watchers get canceled whenever $w is destroyed	19	my $w = AnyEvent->condvar; # stores whether a condition was flagged
20	# only one watcher per $fh and $poll type is allowed	20	$w->send; # wake up current and all future recv's
21	# (i.e. on a socket you cna have one r + one w or one rw
22	# watcher, not any more.
23	# timers can only be used once
24
25	my $w = AnyEvent->condvar; # kind of main loop replacement
26	# can only be used once
27	$w->wait; # enters main loop till $condvar gets ->send	21	$w->recv; # enters "main loop" till $condvar gets ->send
28	$w->broadcast; # wake up waiting and future wait's	22
		23	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
		24
		25	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
		26	nowadays. So what is different about AnyEvent?
		27
		28	Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of
		29	policy> and AnyEvent is I<small and efficient>.
		30
		31	First and foremost, I<AnyEvent is not an event model> itself, it only
		32	interfaces to whatever event model the main program happens to use in a
		33	pragmatic way. For event models and certain classes of immortals alike,
		34	the statement "there can only be one" is a bitter reality: In general,
		35	only one event loop can be active at the same time in a process. AnyEvent
		36	helps hiding the differences between those event loops.
		37
		38	The goal of AnyEvent is to offer module authors the ability to do event
		39	programming (waiting for I/O or timer events) without subscribing to a
		40	religion, a way of living, and most importantly: without forcing your
		41	module users into the same thing by forcing them to use the same event
		42	model you use.
		43
		44	For modules like POE or IO::Async (which is a total misnomer as it is
		45	actually doing all I/O I<synchronously>...), using them in your module is
		46	like joining a cult: After you joined, you are dependent on them and you
		47	cannot use anything else, as it is simply incompatible to everything that
		48	isn't itself. What's worse, all the potential users of your module are
		49	I<also> forced to use the same event loop you use.
		50
		51	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
		52	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
		53	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
		54	your module uses one of those, every user of your module has to use it,
		55	too. But if your module uses AnyEvent, it works transparently with all
		56	event models it supports (including stuff like POE and IO::Async, as long
		57	as those use one of the supported event loops. It is trivial to add new
		58	event loops to AnyEvent, too, so it is future-proof).
		59
		60	In addition to being free of having to use I<the one and only true event
		61	model>, AnyEvent also is free of bloat and policy: with POE or similar
		62	modules, you get an enormous amount of code and strict rules you have to
		63	follow. AnyEvent, on the other hand, is lean and up to the point, by only
		64	offering the functionality that is necessary, in as thin as a wrapper as
		65	technically possible.
		66
		67	Of course, AnyEvent comes with a big (and fully optional!) toolbox
		68	of useful functionality, such as an asynchronous DNS resolver, 100%
		69	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		70	such as Windows) and lots of real-world knowledge and workarounds for
		71	platform bugs and differences.
		72
		73	Now, if you I<do want> lots of policy (this can arguably be somewhat
		74	useful) and you want to force your users to use the one and only event
		75	model, you should I<not> use this module.
29		76
30	=head1 DESCRIPTION	77	=head1 DESCRIPTION
31		78
32	L<AnyEvent> provides an identical interface to multiple event loops. This	79	L<AnyEvent> provides an identical interface to multiple event loops. This
33	allows module authors to utilizy an event loop without forcing module	80	allows module authors to utilise an event loop without forcing module
34	users to use the same event loop (as only a single event loop can coexist	81	users to use the same event loop (as only a single event loop can coexist
35	peacefully at any one time).	82	peacefully at any one time).
36		83
37	The interface itself is vaguely similar but not identical to the Event	84	The interface itself is vaguely similar, but not identical to the L<Event>
38	module.	85	module.
39		86
40	On the first call of any method, the module tries to detect the currently	87	During the first call of any watcher-creation method, the module tries
41	loaded event loop by probing wether any of the following modules is	88	to detect the currently loaded event loop by probing whether one of the
42	loaded: L<Coro::Event>, L<Event>, L<Glib>, L<Tk>. The first one found is	89	following modules is already loaded: L<EV>,
43	used. If none is found, the module tries to load these modules in the	90	L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>,
44	order given. The first one that could be successfully loaded will be	91	L<POE>. The first one found is used. If none are found, the module tries
45	used. If still none could be found, it will issue an error.	92	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
		93	adaptor should always succeed) in the order given. The first one that can
		94	be successfully loaded will be used. If, after this, still none could be
		95	found, AnyEvent will fall back to a pure-perl event loop, which is not
		96	very efficient, but should work everywhere.
		97
		98	Because AnyEvent first checks for modules that are already loaded, loading
		99	an event model explicitly before first using AnyEvent will likely make
		100	that model the default. For example:
		101
		102	use Tk;
		103	use AnyEvent;
		104
		105	# .. AnyEvent will likely default to Tk
		106
		107	The I<likely> means that, if any module loads another event model and
		108	starts using it, all bets are off. Maybe you should tell their authors to
		109	use AnyEvent so their modules work together with others seamlessly...
		110
		111	The pure-perl implementation of AnyEvent is called
		112	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
		113	explicitly and enjoy the high availability of that event loop :)
		114
		115	=head1 WATCHERS
		116
		117	AnyEvent has the central concept of a I<watcher>, which is an object that
		118	stores relevant data for each kind of event you are waiting for, such as
		119	the callback to call, the file handle to watch, etc.
		120
		121	These watchers are normal Perl objects with normal Perl lifetime. After
		122	creating a watcher it will immediately "watch" for events and invoke the
		123	callback when the event occurs (of course, only when the event model
		124	is in control).
		125
		126	To disable the watcher you have to destroy it (e.g. by setting the
		127	variable you store it in to C<undef> or otherwise deleting all references
		128	to it).
		129
		130	All watchers are created by calling a method on the C<AnyEvent> class.
		131
		132	Many watchers either are used with "recursion" (repeating timers for
		133	example), or need to refer to their watcher object in other ways.
		134
		135	An any way to achieve that is this pattern:
		136
		137	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
		138	# you can use $w here, for example to undef it
		139	undef $w;
		140	});
		141
		142	Note that C<my $w; $w => combination. This is necessary because in Perl,
		143	my variables are only visible after the statement in which they are
		144	declared.
		145
		146	=head2 I/O WATCHERS
		147
		148	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
		149	with the following mandatory key-value pairs as arguments:
		150
		151	C<fh> the Perl I<file handle> (I<not> file descriptor) to watch
		152	for events. C<poll> must be a string that is either C<r> or C<w>,
		153	which creates a watcher waiting for "r"eadable or "w"ritable events,
		154	respectively. C<cb> is the callback to invoke each time the file handle
		155	becomes ready.
		156
		157	Although the callback might get passed parameters, their value and
		158	presence is undefined and you cannot rely on them. Portable AnyEvent
		159	callbacks cannot use arguments passed to I/O watcher callbacks.
		160
		161	The I/O watcher might use the underlying file descriptor or a copy of it.
		162	You must not close a file handle as long as any watcher is active on the
		163	underlying file descriptor.
		164
		165	Some event loops issue spurious readyness notifications, so you should
		166	always use non-blocking calls when reading/writing from/to your file
		167	handles.
		168
		169	Example:
		170
		171	# wait for readability of STDIN, then read a line and disable the watcher
		172	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
		173	chomp (my $input = <STDIN>);
		174	warn "read: $input\n";
		175	undef $w;
		176	});
		177
		178	=head2 TIME WATCHERS
		179
		180	You can create a time watcher by calling the C<< AnyEvent->timer >>
		181	method with the following mandatory arguments:
		182
		183	C<after> specifies after how many seconds (fractional values are
		184	supported) the callback should be invoked. C<cb> is the callback to invoke
		185	in that case.
		186
		187	Although the callback might get passed parameters, their value and
		188	presence is undefined and you cannot rely on them. Portable AnyEvent
		189	callbacks cannot use arguments passed to time watcher callbacks.
		190
		191	The timer callback will be invoked at most once: if you want a repeating
		192	timer you have to create a new watcher (this is a limitation by both Tk
		193	and Glib).
		194
		195	Example:
		196
		197	# fire an event after 7.7 seconds
		198	my $w = AnyEvent->timer (after => 7.7, cb => sub {
		199	warn "timeout\n";
		200	});
		201
		202	# to cancel the timer:
		203	undef $w;
		204
		205	Example 2:
		206
		207	# fire an event after 0.5 seconds, then roughly every second
		208	my $w;
		209
		210	my $cb = sub {
		211	# cancel the old timer while creating a new one
		212	$w = AnyEvent->timer (after => 1, cb => $cb);
		213	};
		214
		215	# start the "loop" by creating the first watcher
		216	$w = AnyEvent->timer (after => 0.5, cb => $cb);
		217
		218	=head3 TIMING ISSUES
		219
		220	There are two ways to handle timers: based on real time (relative, "fire
		221	in 10 seconds") and based on wallclock time (absolute, "fire at 12
		222	o'clock").
		223
		224	While most event loops expect timers to specified in a relative way, they
		225	use absolute time internally. This makes a difference when your clock
		226	"jumps", for example, when ntp decides to set your clock backwards from
		227	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		228	fire "after" a second might actually take six years to finally fire.
		229
		230	AnyEvent cannot compensate for this. The only event loop that is conscious
		231	about these issues is L<EV>, which offers both relative (ev_timer, based
		232	on true relative time) and absolute (ev_periodic, based on wallclock time)
		233	timers.
		234
		235	AnyEvent always prefers relative timers, if available, matching the
		236	AnyEvent API.
		237
		238	=head2 SIGNAL WATCHERS
		239
		240	You can watch for signals using a signal watcher, C<signal> is the signal
		241	I<name> without any C<SIG> prefix, C<cb> is the Perl callback to
		242	be invoked whenever a signal occurs.
		243
		244	Although the callback might get passed parameters, their value and
		245	presence is undefined and you cannot rely on them. Portable AnyEvent
		246	callbacks cannot use arguments passed to signal watcher callbacks.
		247
		248	Multiple signal occurrences can be clumped together into one callback
		249	invocation, and callback invocation will be synchronous. Synchronous means
		250	that it might take a while until the signal gets handled by the process,
		251	but it is guaranteed not to interrupt any other callbacks.
		252
		253	The main advantage of using these watchers is that you can share a signal
		254	between multiple watchers.
		255
		256	This watcher might use C<%SIG>, so programs overwriting those signals
		257	directly will likely not work correctly.
		258
		259	Example: exit on SIGINT
		260
		261	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		262
		263	=head2 CHILD PROCESS WATCHERS
		264
		265	You can also watch on a child process exit and catch its exit status.
		266
		267	The child process is specified by the C<pid> argument (if set to C<0>, it
		268	watches for any child process exit). The watcher will trigger as often
		269	as status change for the child are received. This works by installing a
		270	signal handler for C<SIGCHLD>. The callback will be called with the pid
		271	and exit status (as returned by waitpid), so unlike other watcher types,
		272	you I<can> rely on child watcher callback arguments.
		273
		274	There is a slight catch to child watchers, however: you usually start them
		275	I<after> the child process was created, and this means the process could
		276	have exited already (and no SIGCHLD will be sent anymore).
		277
		278	Not all event models handle this correctly (POE doesn't), but even for
		279	event models that I<do> handle this correctly, they usually need to be
		280	loaded before the process exits (i.e. before you fork in the first place).
		281
		282	This means you cannot create a child watcher as the very first thing in an
		283	AnyEvent program, you I<have> to create at least one watcher before you
		284	C<fork> the child (alternatively, you can call C<AnyEvent::detect>).
		285
		286	Example: fork a process and wait for it
		287
		288	my $done = AnyEvent->condvar;
		289
		290	my $pid = fork or exit 5;
		291
		292	my $w = AnyEvent->child (
		293	pid => $pid,
		294	cb => sub {
		295	my ($pid, $status) = @_;
		296	warn "pid $pid exited with status $status";
		297	$done->send;
		298	},
		299	);
		300
		301	# do something else, then wait for process exit
		302	$done->recv;
		303
		304	=head2 CONDITION VARIABLES
		305
		306	If you are familiar with some event loops you will know that all of them
		307	require you to run some blocking "loop", "run" or similar function that
		308	will actively watch for new events and call your callbacks.
		309
		310	AnyEvent is different, it expects somebody else to run the event loop and
		311	will only block when necessary (usually when told by the user).
		312
		313	The instrument to do that is called a "condition variable", so called
		314	because they represent a condition that must become true.
		315
		316	Condition variables can be created by calling the C<< AnyEvent->condvar
		317	>> method, usually without arguments. The only argument pair allowed is
		318	C<cb>, which specifies a callback to be called when the condition variable
		319	becomes true.
		320
		321	After creation, the condition variable is "false" until it becomes "true"
		322	by calling the C<send> method (or calling the condition variable as if it
		323	were a callback, read about the caveats in the description for the C<<
		324	->send >> method).
		325
		326	Condition variables are similar to callbacks, except that you can
		327	optionally wait for them. They can also be called merge points - points
		328	in time where multiple outstanding events have been processed. And yet
		329	another way to call them is transactions - each condition variable can be
		330	used to represent a transaction, which finishes at some point and delivers
		331	a result.
		332
		333	Condition variables are very useful to signal that something has finished,
		334	for example, if you write a module that does asynchronous http requests,
		335	then a condition variable would be the ideal candidate to signal the
		336	availability of results. The user can either act when the callback is
		337	called or can synchronously C<< ->recv >> for the results.
		338
		339	You can also use them to simulate traditional event loops - for example,
		340	you can block your main program until an event occurs - for example, you
		341	could C<< ->recv >> in your main program until the user clicks the Quit
		342	button of your app, which would C<< ->send >> the "quit" event.
		343
		344	Note that condition variables recurse into the event loop - if you have
		345	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		346	lose. Therefore, condition variables are good to export to your caller, but
		347	you should avoid making a blocking wait yourself, at least in callbacks,
		348	as this asks for trouble.
		349
		350	Condition variables are represented by hash refs in perl, and the keys
		351	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		352	easy (it is often useful to build your own transaction class on top of
		353	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		354	it's C<new> method in your own C<new> method.
		355
		356	There are two "sides" to a condition variable - the "producer side" which
		357	eventually calls C<< -> send >>, and the "consumer side", which waits
		358	for the send to occur.
		359
		360	Example: wait for a timer.
		361
		362	# wait till the result is ready
		363	my $result_ready = AnyEvent->condvar;
		364
		365	# do something such as adding a timer
		366	# or socket watcher the calls $result_ready->send
		367	# when the "result" is ready.
		368	# in this case, we simply use a timer:
		369	my $w = AnyEvent->timer (
		370	after => 1,
		371	cb => sub { $result_ready->send },
		372	);
		373
		374	# this "blocks" (while handling events) till the callback
		375	# calls send
		376	$result_ready->recv;
		377
		378	Example: wait for a timer, but take advantage of the fact that
		379	condition variables are also code references.
		380
		381	my $done = AnyEvent->condvar;
		382	my $delay = AnyEvent->timer (after => 5, cb => $done);
		383	$done->recv;
		384
		385	=head3 METHODS FOR PRODUCERS
		386
		387	These methods should only be used by the producing side, i.e. the
		388	code/module that eventually sends the signal. Note that it is also
		389	the producer side which creates the condvar in most cases, but it isn't
		390	uncommon for the consumer to create it as well.
46		391
47	=over 4	392	=over 4
48		393
		394	=item $cv->send (...)
		395
		396	Flag the condition as ready - a running C<< ->recv >> and all further
		397	calls to C<recv> will (eventually) return after this method has been
		398	called. If nobody is waiting the send will be remembered.
		399
		400	If a callback has been set on the condition variable, it is called
		401	immediately from within send.
		402
		403	Any arguments passed to the C<send> call will be returned by all
		404	future C<< ->recv >> calls.
		405
		406	Condition variables are overloaded so one can call them directly
		407	(as a code reference). Calling them directly is the same as calling
		408	C<send>. Note, however, that many C-based event loops do not handle
		409	overloading, so as tempting as it may be, passing a condition variable
		410	instead of a callback does not work. Both the pure perl and EV loops
		411	support overloading, however, as well as all functions that use perl to
		412	invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
		413	example).
		414
		415	=item $cv->croak ($error)
		416
		417	Similar to send, but causes all call's to C<< ->recv >> to invoke
		418	C<Carp::croak> with the given error message/object/scalar.
		419
		420	This can be used to signal any errors to the condition variable
		421	user/consumer.
		422
		423	=item $cv->begin ([group callback])
		424
		425	=item $cv->end
		426
		427	These two methods are EXPERIMENTAL and MIGHT CHANGE.
		428
		429	These two methods can be used to combine many transactions/events into
		430	one. For example, a function that pings many hosts in parallel might want
		431	to use a condition variable for the whole process.
		432
		433	Every call to C<< ->begin >> will increment a counter, and every call to
		434	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		435	>>, the (last) callback passed to C<begin> will be executed. That callback
		436	is I<supposed> to call C<< ->send >>, but that is not required. If no
		437	callback was set, C<send> will be called without any arguments.
		438
		439	Let's clarify this with the ping example:
		440
		441	my $cv = AnyEvent->condvar;
		442
		443	my %result;
		444	$cv->begin (sub { $cv->send (\%result) });
		445
		446	for my $host (@list_of_hosts) {
		447	$cv->begin;
		448	ping_host_then_call_callback $host, sub {
		449	$result{$host} = ...;
		450	$cv->end;
		451	};
		452	}
		453
		454	$cv->end;
		455
		456	This code fragment supposedly pings a number of hosts and calls
		457	C<send> after results for all then have have been gathered - in any
		458	order. To achieve this, the code issues a call to C<begin> when it starts
		459	each ping request and calls C<end> when it has received some result for
		460	it. Since C<begin> and C<end> only maintain a counter, the order in which
		461	results arrive is not relevant.
		462
		463	There is an additional bracketing call to C<begin> and C<end> outside the
		464	loop, which serves two important purposes: first, it sets the callback
		465	to be called once the counter reaches C<0>, and second, it ensures that
		466	C<send> is called even when C<no> hosts are being pinged (the loop
		467	doesn't execute once).
		468
		469	This is the general pattern when you "fan out" into multiple subrequests:
		470	use an outer C<begin>/C<end> pair to set the callback and ensure C<end>
		471	is called at least once, and then, for each subrequest you start, call
		472	C<begin> and for each subrequest you finish, call C<end>.
		473
		474	=back
		475
		476	=head3 METHODS FOR CONSUMERS
		477
		478	These methods should only be used by the consuming side, i.e. the
		479	code awaits the condition.
		480
		481	=over 4
		482
		483	=item $cv->recv
		484
		485	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		486	>> methods have been called on c<$cv>, while servicing other watchers
		487	normally.
		488
		489	You can only wait once on a condition - additional calls are valid but
		490	will return immediately.
		491
		492	If an error condition has been set by calling C<< ->croak >>, then this
		493	function will call C<croak>.
		494
		495	In list context, all parameters passed to C<send> will be returned,
		496	in scalar context only the first one will be returned.
		497
		498	Not all event models support a blocking wait - some die in that case
		499	(programs might want to do that to stay interactive), so I<if you are
		500	using this from a module, never require a blocking wait>, but let the
		501	caller decide whether the call will block or not (for example, by coupling
		502	condition variables with some kind of request results and supporting
		503	callbacks so the caller knows that getting the result will not block,
		504	while still supporting blocking waits if the caller so desires).
		505
		506	Another reason I<never> to C<< ->recv >> in a module is that you cannot
		507	sensibly have two C<< ->recv >>'s in parallel, as that would require
		508	multiple interpreters or coroutines/threads, none of which C<AnyEvent>
		509	can supply.
		510
		511	The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
		512	fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
		513	versions and also integrates coroutines into AnyEvent, making blocking
		514	C<< ->recv >> calls perfectly safe as long as they are done from another
		515	coroutine (one that doesn't run the event loop).
		516
		517	You can ensure that C<< -recv >> never blocks by setting a callback and
		518	only calling C<< ->recv >> from within that callback (or at a later
		519	time). This will work even when the event loop does not support blocking
		520	waits otherwise.
		521
		522	=item $bool = $cv->ready
		523
		524	Returns true when the condition is "true", i.e. whether C<send> or
		525	C<croak> have been called.
		526
		527	=item $cb = $cv->cb ([new callback])
		528
		529	This is a mutator function that returns the callback set and optionally
		530	replaces it before doing so.
		531
		532	The callback will be called when the condition becomes "true", i.e. when
		533	C<send> or C<croak> are called. Calling C<recv> inside the callback
		534	or at any later time is guaranteed not to block.
		535
		536	=back
		537
		538	=head1 GLOBAL VARIABLES AND FUNCTIONS
		539
		540	=over 4
		541
		542	=item $AnyEvent::MODEL
		543
		544	Contains C<undef> until the first watcher is being created. Then it
		545	contains the event model that is being used, which is the name of the
		546	Perl class implementing the model. This class is usually one of the
		547	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case
		548	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).
		549
		550	The known classes so far are:
		551
		552	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
		553	AnyEvent::Impl::Event based on Event, second best choice.
		554	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
		555	AnyEvent::Impl::Glib based on Glib, third-best choice.
		556	AnyEvent::Impl::Tk based on Tk, very bad choice.
		557	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
		558	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		559	AnyEvent::Impl::POE based on POE, not generic enough for full support.
		560
		561	There is no support for WxWidgets, as WxWidgets has no support for
		562	watching file handles. However, you can use WxWidgets through the
		563	POE Adaptor, as POE has a Wx backend that simply polls 20 times per
		564	second, which was considered to be too horrible to even consider for
		565	AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
		566	it's adaptor.
		567
		568	AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
		569	autodetecting them.
		570
		571	=item AnyEvent::detect
		572
		573	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
		574	if necessary. You should only call this function right before you would
		575	have created an AnyEvent watcher anyway, that is, as late as possible at
		576	runtime.
		577
		578	=item $guard = AnyEvent::post_detect { BLOCK }
		579
		580	Arranges for the code block to be executed as soon as the event model is
		581	autodetected (or immediately if this has already happened).
		582
		583	If called in scalar or list context, then it creates and returns an object
		584	that automatically removes the callback again when it is destroyed. See
		585	L<Coro::BDB> for a case where this is useful.
		586
		587	=item @AnyEvent::post_detect
		588
		589	If there are any code references in this array (you can C<push> to it
		590	before or after loading AnyEvent), then they will called directly after
		591	the event loop has been chosen.
		592
		593	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		594	if it contains a true value then the event loop has already been detected,
		595	and the array will be ignored.
		596
		597	Best use C<AnyEvent::post_detect { BLOCK }> instead.
		598
		599	=back
		600
		601	=head1 WHAT TO DO IN A MODULE
		602
		603	As a module author, you should C<use AnyEvent> and call AnyEvent methods
		604	freely, but you should not load a specific event module or rely on it.
		605
		606	Be careful when you create watchers in the module body - AnyEvent will
		607	decide which event module to use as soon as the first method is called, so
		608	by calling AnyEvent in your module body you force the user of your module
		609	to load the event module first.
		610
		611	Never call C<< ->recv >> on a condition variable unless you I<know> that
		612	the C<< ->send >> method has been called on it already. This is
		613	because it will stall the whole program, and the whole point of using
		614	events is to stay interactive.
		615
		616	It is fine, however, to call C<< ->recv >> when the user of your module
		617	requests it (i.e. if you create a http request object ad have a method
		618	called C<results> that returns the results, it should call C<< ->recv >>
		619	freely, as the user of your module knows what she is doing. always).
		620
		621	=head1 WHAT TO DO IN THE MAIN PROGRAM
		622
		623	There will always be a single main program - the only place that should
		624	dictate which event model to use.
		625
		626	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
		627	do anything special (it does not need to be event-based) and let AnyEvent
		628	decide which implementation to chose if some module relies on it.
		629
		630	If the main program relies on a specific event model - for example, in
		631	Gtk2 programs you have to rely on the Glib module - you should load the
		632	event module before loading AnyEvent or any module that uses it: generally
		633	speaking, you should load it as early as possible. The reason is that
		634	modules might create watchers when they are loaded, and AnyEvent will
		635	decide on the event model to use as soon as it creates watchers, and it
		636	might chose the wrong one unless you load the correct one yourself.
		637
		638	You can chose to use a pure-perl implementation by loading the
		639	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
		640	everywhere, but letting AnyEvent chose the model is generally better.
		641
		642	=head2 MAINLOOP EMULATION
		643
		644	Sometimes (often for short test scripts, or even standalone programs who
		645	only want to use AnyEvent), you do not want to run a specific event loop.
		646
		647	In that case, you can use a condition variable like this:
		648
		649	AnyEvent->condvar->recv;
		650
		651	This has the effect of entering the event loop and looping forever.
		652
		653	Note that usually your program has some exit condition, in which case
		654	it is better to use the "traditional" approach of storing a condition
		655	variable somewhere, waiting for it, and sending it when the program should
		656	exit cleanly.
		657
		658
		659	=head1 OTHER MODULES
		660
		661	The following is a non-exhaustive list of additional modules that use
		662	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		663	in the same program. Some of the modules come with AnyEvent, some are
		664	available via CPAN.
		665
		666	=over 4
		667
		668	=item L<AnyEvent::Util>
		669
		670	Contains various utility functions that replace often-used but blocking
		671	functions such as C<inet_aton> by event-/callback-based versions.
		672
		673	=item L<AnyEvent::Handle>
		674
		675	Provide read and write buffers and manages watchers for reads and writes.
		676
		677	=item L<AnyEvent::Socket>
		678
		679	Provides various utility functions for (internet protocol) sockets,
		680	addresses and name resolution. Also functions to create non-blocking tcp
		681	connections or tcp servers, with IPv6 and SRV record support and more.
		682
		683	=item L<AnyEvent::DNS>
		684
		685	Provides rich asynchronous DNS resolver capabilities.
		686
		687	=item L<AnyEvent::HTTPD>
		688
		689	Provides a simple web application server framework.
		690
		691	=item L<AnyEvent::FastPing>
		692
		693	The fastest ping in the west.
		694
		695	=item L<Net::IRC3>
		696
		697	AnyEvent based IRC client module family.
		698
		699	=item L<Net::XMPP2>
		700
		701	AnyEvent based XMPP (Jabber protocol) module family.
		702
		703	=item L<Net::FCP>
		704
		705	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		706	of AnyEvent.
		707
		708	=item L<Event::ExecFlow>
		709
		710	High level API for event-based execution flow control.
		711
		712	=item L<Coro>
		713
		714	Has special support for AnyEvent via L<Coro::AnyEvent>.
		715
		716	=item L<AnyEvent::AIO>, L<IO::AIO>
		717
		718	Truly asynchronous I/O, should be in the toolbox of every event
		719	programmer. AnyEvent::AIO transparently fuses IO::AIO and AnyEvent
		720	together.
		721
		722	=item L<AnyEvent::BDB>, L<BDB>
		723
		724	Truly asynchronous Berkeley DB access. AnyEvent::AIO transparently fuses
		725	IO::AIO and AnyEvent together.
		726
		727	=item L<IO::Lambda>
		728
		729	The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
		730
		731	=back
		732
49	=cut	733	=cut
50		734
51	package AnyEvent;	735	package AnyEvent;
52		736
53	no warnings;	737	no warnings;
54	use strict 'vars';	738	use strict;
		739
55	use Carp;	740	use Carp;
56		741
57	our $VERSION = 0.2;	742	our $VERSION = '4.05';
58	our $MODEL;	743	our $MODEL;
59		744
60	our $AUTOLOAD;	745	our $AUTOLOAD;
61	our @ISA;	746	our @ISA;
62		747
		748	our @REGISTRY;
		749
		750	our $WIN32;
		751
		752	BEGIN {
		753	my $win32 = ! ! ($^O =~ /mswin32/i);
		754	eval "sub WIN32(){ $win32 }";
		755	}
		756
		757	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
		758
		759	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		760
		761	{
		762	my $idx;
		763	$PROTOCOL{$_} = ++$idx
		764	for reverse split /\s*,\s*/,
		765	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		766	}
		767
63	my @models = (	768	my @models = (
64	[Coro => Coro::Event::],	769	[EV:: => AnyEvent::Impl::EV::],
65	[Event => Event::],	770	[Event:: => AnyEvent::Impl::Event::],
66	[Glib => Glib::],	771	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
67	[Tk => Tk::],	772	# everything below here will not be autoprobed
		773	# as the pureperl backend should work everywhere
		774	# and is usually faster
		775	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		776	[Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
		777	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		778	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		779	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		780	[Wx:: => AnyEvent::Impl::POE::],
		781	[Prima:: => AnyEvent::Impl::POE::],
68	);	782	);
69		783
70	our %method = map +($_ => 1), qw(io timer condvar broadcast wait cancel DESTROY);	784	our %method = map +($_ => 1), qw(io timer signal child condvar one_event DESTROY);
71		785
72	sub AUTOLOAD {	786	our @post_detect;
73	$AUTOLOAD =~ s/.*://;
74		787
75	$method{$AUTOLOAD}	788	sub post_detect(&) {
76	or croak "$AUTOLOAD: not a valid method for AnyEvent objects";	789	my ($cb) = @_;
77		790
		791	if ($MODEL) {
		792	$cb->();
		793
		794	1
		795	} else {
		796	push @post_detect, $cb;
		797
		798	defined wantarray
		799	? bless \$cb, "AnyEvent::Util::PostDetect"
		800	: ()
		801	}
		802	}
		803
		804	sub AnyEvent::Util::PostDetect::DESTROY {
		805	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		806	}
		807
		808	sub detect() {
78	unless ($MODEL) {	809	unless ($MODEL) {
79	# check for already loaded models	810	no strict 'refs';
80	for (@models) {	811	local $SIG{__DIE__};
81	my ($model, $package) = @$_;	812
82	if (scalar keys %{ *{"$package\::"} }) {	813	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
83	eval "require AnyEvent::Impl::$model";	814	my $model = "AnyEvent::Impl::$1";
84	last if $MODEL;	815	if (eval "require $model") {
		816	$MODEL = $model;
		817	warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1;
		818	} else {
		819	warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose;
85	}	820	}
86	}	821	}
87		822
		823	# check for already loaded models
88	unless ($MODEL) {	824	unless ($MODEL) {
89	# try to load a model
90
91	for (@models) {	825	for (@REGISTRY, @models) {
92	my ($model, $package) = @$_;	826	my ($package, $model) = @$_;
93	eval "require AnyEvent::Impl::$model";	827	if (${"$package\::VERSION"} > 0) {
94	last if $MODEL;	828	if (eval "require $model") {
		829	$MODEL = $model;
		830	warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1;
		831	last;
		832	}
		833	}
95	}	834	}
96		835
		836	unless ($MODEL) {
		837	# try to load a model
		838
		839	for (@REGISTRY, @models) {
		840	my ($package, $model) = @$_;
		841	if (eval "require $package"
		842	and ${"$package\::VERSION"} > 0
		843	and eval "require $model") {
		844	$MODEL = $model;
		845	warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1;
		846	last;
		847	}
		848	}
		849
97	$MODEL	850	$MODEL
98	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: Coro, Event, Glib or Tk.";	851	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.";
		852	}
99	}	853	}
		854
		855	unshift @ISA, $MODEL;
		856	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		857
		858	(shift @post_detect)->() while @post_detect;
100	}	859	}
101		860
102	@ISA = $MODEL;	861	$MODEL
		862	}
		863
		864	sub AUTOLOAD {
		865	(my $func = $AUTOLOAD) =~ s/.*://;
		866
		867	$method{$func}
		868	or croak "$func: not a valid method for AnyEvent objects";
		869
		870	detect unless $MODEL;
103		871
104	my $class = shift;	872	my $class = shift;
105	$class->$AUTOLOAD (@_);	873	$class->$func (@_);
106	}	874	}
		875
		876	package AnyEvent::Base;
		877
		878	# default implementation for ->condvar
		879
		880	sub condvar {
		881	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, AnyEvent::CondVar::
		882	}
		883
		884	# default implementation for ->signal
		885
		886	our %SIG_CB;
		887
		888	sub signal {
		889	my (undef, %arg) = @_;
		890
		891	my $signal = uc $arg{signal}
		892	or Carp::croak "required option 'signal' is missing";
		893
		894	$SIG_CB{$signal}{$arg{cb}} = $arg{cb};
		895	$SIG{$signal} ||= sub {
		896	$_->() for values %{ $SIG_CB{$signal} || {} };
		897	};
		898
		899	bless [$signal, $arg{cb}], "AnyEvent::Base::Signal"
		900	}
		901
		902	sub AnyEvent::Base::Signal::DESTROY {
		903	my ($signal, $cb) = @{$_[0]};
		904
		905	delete $SIG_CB{$signal}{$cb};
		906
		907	$SIG{$signal} = 'DEFAULT' unless keys %{ $SIG_CB{$signal} };
		908	}
		909
		910	# default implementation for ->child
		911
		912	our %PID_CB;
		913	our $CHLD_W;
		914	our $CHLD_DELAY_W;
		915	our $PID_IDLE;
		916	our $WNOHANG;
		917
		918	sub _child_wait {
		919	while (0 < (my $pid = waitpid -1, $WNOHANG)) {
		920	$_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }),
		921	(values %{ $PID_CB{0} || {} });
		922	}
		923
		924	undef $PID_IDLE;
		925	}
		926
		927	sub _sigchld {
		928	# make sure we deliver these changes "synchronous" with the event loop.
		929	$CHLD_DELAY_W ||= AnyEvent->timer (after => 0, cb => sub {
		930	undef $CHLD_DELAY_W;
		931	&_child_wait;
		932	});
		933	}
		934
		935	sub child {
		936	my (undef, %arg) = @_;
		937
		938	defined (my $pid = $arg{pid} + 0)
		939	or Carp::croak "required option 'pid' is missing";
		940
		941	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
		942
		943	unless ($WNOHANG) {
		944	$WNOHANG = eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
		945	}
		946
		947	unless ($CHLD_W) {
		948	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
		949	# child could be a zombie already, so make at least one round
		950	&_sigchld;
		951	}
		952
		953	bless [$pid, $arg{cb}], "AnyEvent::Base::Child"
		954	}
		955
		956	sub AnyEvent::Base::Child::DESTROY {
		957	my ($pid, $cb) = @{$_[0]};
		958
		959	delete $PID_CB{$pid}{$cb};
		960	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
		961
		962	undef $CHLD_W unless keys %PID_CB;
		963	}
		964
		965	package AnyEvent::CondVar;
		966
		967	our @ISA = AnyEvent::CondVar::Base::;
		968
		969	package AnyEvent::CondVar::Base;
		970
		971	use overload
		972	'&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		973	fallback => 1;
		974
		975	sub _send {
		976	# nop
		977	}
		978
		979	sub send {
		980	my $cv = shift;
		981	$cv->{_ae_sent} = [@_];
		982	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		983	$cv->_send;
		984	}
		985
		986	sub croak {
		987	$_[0]{_ae_croak} = $_[1];
		988	$_[0]->send;
		989	}
		990
		991	sub ready {
		992	$_[0]{_ae_sent}
		993	}
		994
		995	sub _wait {
		996	AnyEvent->one_event while !$_[0]{_ae_sent};
		997	}
		998
		999	sub recv {
		1000	$_[0]->_wait;
		1001
		1002	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1003	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1004	}
		1005
		1006	sub cb {
		1007	$_[0]{_ae_cb} = $_[1] if @_ > 1;
		1008	$_[0]{_ae_cb}
		1009	}
		1010
		1011	sub begin {
		1012	++$_[0]{_ae_counter};
		1013	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1014	}
		1015
		1016	sub end {
		1017	return if --$_[0]{_ae_counter};
		1018	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1019	}
		1020
		1021	# undocumented/compatibility with pre-3.4
		1022	*broadcast = \&send;
		1023	*wait = \&_wait;
		1024
		1025	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1026
		1027	This is an advanced topic that you do not normally need to use AnyEvent in
		1028	a module. This section is only of use to event loop authors who want to
		1029	provide AnyEvent compatibility.
		1030
		1031	If you need to support another event library which isn't directly
		1032	supported by AnyEvent, you can supply your own interface to it by
		1033	pushing, before the first watcher gets created, the package name of
		1034	the event module and the package name of the interface to use onto
		1035	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
		1036	AnyEvent, so it is reasonably cheap.
		1037
		1038	Example:
		1039
		1040	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
		1041
		1042	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
		1043	package/class when it finds the C<urxvt> package/module is already loaded.
		1044
		1045	When AnyEvent is loaded and asked to find a suitable event model, it
		1046	will first check for the presence of urxvt by trying to C<use> the
		1047	C<urxvt::anyevent> module.
		1048
		1049	The class should provide implementations for all watcher types. See
		1050	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
		1051	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
		1052	see the sources.
		1053
		1054	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1055	provide suitable (hopefully) replacements.
		1056
		1057	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
		1058	terminal emulator uses the above line as-is. An interface isn't included
		1059	in AnyEvent because it doesn't make sense outside the embedded interpreter
		1060	inside I<rxvt-unicode>, and it is updated and maintained as part of the
		1061	I<rxvt-unicode> distribution.
		1062
		1063	I<rxvt-unicode> also cheats a bit by not providing blocking access to
		1064	condition variables: code blocking while waiting for a condition will
		1065	C<die>. This still works with most modules/usages, and blocking calls must
		1066	not be done in an interactive application, so it makes sense.
		1067
		1068	=head1 ENVIRONMENT VARIABLES
		1069
		1070	The following environment variables are used by this module:
		1071
		1072	=over 4
		1073
		1074	=item C<PERL_ANYEVENT_VERBOSE>
		1075
		1076	By default, AnyEvent will be completely silent except in fatal
		1077	conditions. You can set this environment variable to make AnyEvent more
		1078	talkative.
		1079
		1080	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1081	conditions, such as not being able to load the event model specified by
		1082	C<PERL_ANYEVENT_MODEL>.
		1083
		1084	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1085	model it chooses.
		1086
		1087	=item C<PERL_ANYEVENT_MODEL>
		1088
		1089	This can be used to specify the event model to be used by AnyEvent, before
		1090	auto detection and -probing kicks in. It must be a string consisting
		1091	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1092	and the resulting module name is loaded and if the load was successful,
		1093	used as event model. If it fails to load AnyEvent will proceed with
		1094	auto detection and -probing.
		1095
		1096	This functionality might change in future versions.
		1097
		1098	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1099	could start your program like this:
		1100
		1101	PERL_ANYEVENT_MODEL=Perl perl ...
		1102
		1103	=item C<PERL_ANYEVENT_PROTOCOLS>
		1104
		1105	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1106	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1107	of auto probing).
		1108
		1109	Must be set to a comma-separated list of protocols or address families,
		1110	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1111	used, and preference will be given to protocols mentioned earlier in the
		1112	list.
		1113
		1114	This variable can effectively be used for denial-of-service attacks
		1115	against local programs (e.g. when setuid), although the impact is likely
		1116	small, as the program has to handle connection errors already-
		1117
		1118	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1119	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1120	- only support IPv4, never try to resolve or contact IPv6
		1121	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1122	IPv6, but prefer IPv6 over IPv4.
		1123
		1124	=item C<PERL_ANYEVENT_EDNS0>
		1125
		1126	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1127	for DNS. This extension is generally useful to reduce DNS traffic, but
		1128	some (broken) firewalls drop such DNS packets, which is why it is off by
		1129	default.
		1130
		1131	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1132	EDNS0 in its DNS requests.
		1133
		1134	=item C<PERL_ANYEVENT_MAX_FORKS>
		1135
		1136	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1137	will create in parallel.
107		1138
108	=back	1139	=back
109		1140
110	=head1 EXAMPLE	1141	=head1 EXAMPLE PROGRAM
111		1142
112	The following program uses an io watcher to read data from stdin, a timer	1143	The following program uses an I/O watcher to read data from STDIN, a timer
113	to display a message once per second, and a condvar to exit the program	1144	to display a message once per second, and a condition variable to quit the
114	when the user enters quit:	1145	program when the user enters quit:
115		1146
116	use AnyEvent;	1147	use AnyEvent;
117		1148
118	my $cv = AnyEvent->condvar;	1149	my $cv = AnyEvent->condvar;
119		1150
120	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	1151	my $io_watcher = AnyEvent->io (
		1152	fh => \*STDIN,
		1153	poll => 'r',
		1154	cb => sub {
121	warn "io event <$_[0]>\n"; # will always output <r>	1155	warn "io event <$_[0]>\n"; # will always output <r>
122	chomp (my $input = <STDIN>); # read a line	1156	chomp (my $input = <STDIN>); # read a line
123	warn "read: $input\n"; # output what has been read	1157	warn "read: $input\n"; # output what has been read
124	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1158	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1159	},
125	});	1160	);
126		1161
127	my $time_watcher; # can only be used once	1162	my $time_watcher; # can only be used once
128		1163
129	sub new_timer {	1164	sub new_timer {
130	$timer = AnyEvent->timer (after => 1, cb => sub {	1165	$timer = AnyEvent->timer (after => 1, cb => sub {
…		…
133	});	1168	});
134	}	1169	}
135		1170
136	new_timer; # create first timer	1171	new_timer; # create first timer
137		1172
138	$cv->wait; # wait until user enters /^q/i	1173	$cv->recv; # wait until user enters /^q/i
		1174
		1175	=head1 REAL-WORLD EXAMPLE
		1176
		1177	Consider the L<Net::FCP> module. It features (among others) the following
		1178	API calls, which are to freenet what HTTP GET requests are to http:
		1179
		1180	my $data = $fcp->client_get ($url); # blocks
		1181
		1182	my $transaction = $fcp->txn_client_get ($url); # does not block
		1183	$transaction->cb ( sub { ... } ); # set optional result callback
		1184	my $data = $transaction->result; # possibly blocks
		1185
		1186	The C<client_get> method works like C<LWP::Simple::get>: it requests the
		1187	given URL and waits till the data has arrived. It is defined to be:
		1188
		1189	sub client_get { $_[0]->txn_client_get ($_[1])->result }
		1190
		1191	And in fact is automatically generated. This is the blocking API of
		1192	L<Net::FCP>, and it works as simple as in any other, similar, module.
		1193
		1194	More complicated is C<txn_client_get>: It only creates a transaction
		1195	(completion, result, ...) object and initiates the transaction.
		1196
		1197	my $txn = bless { }, Net::FCP::Txn::;
		1198
		1199	It also creates a condition variable that is used to signal the completion
		1200	of the request:
		1201
		1202	$txn->{finished} = AnyAvent->condvar;
		1203
		1204	It then creates a socket in non-blocking mode.
		1205
		1206	socket $txn->{fh}, ...;
		1207	fcntl $txn->{fh}, F_SETFL, O_NONBLOCK;
		1208	connect $txn->{fh}, ...
		1209	and !$!{EWOULDBLOCK}
		1210	and !$!{EINPROGRESS}
		1211	and Carp::croak "unable to connect: $!\n";
		1212
		1213	Then it creates a write-watcher which gets called whenever an error occurs
		1214	or the connection succeeds:
		1215
		1216	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'w', cb => sub { $txn->fh_ready_w });
		1217
		1218	And returns this transaction object. The C<fh_ready_w> callback gets
		1219	called as soon as the event loop detects that the socket is ready for
		1220	writing.
		1221
		1222	The C<fh_ready_w> method makes the socket blocking again, writes the
		1223	request data and replaces the watcher by a read watcher (waiting for reply
		1224	data). The actual code is more complicated, but that doesn't matter for
		1225	this example:
		1226
		1227	fcntl $txn->{fh}, F_SETFL, 0;
		1228	syswrite $txn->{fh}, $txn->{request}
		1229	or die "connection or write error";
		1230	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
		1231
		1232	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
		1233	result and signals any possible waiters that the request has finished:
		1234
		1235	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
		1236
		1237	if (end-of-file or data complete) {
		1238	$txn->{result} = $txn->{buf};
		1239	$txn->{finished}->send;
		1240	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
		1241	}
		1242
		1243	The C<result> method, finally, just waits for the finished signal (if the
		1244	request was already finished, it doesn't wait, of course, and returns the
		1245	data:
		1246
		1247	$txn->{finished}->recv;
		1248	return $txn->{result};
		1249
		1250	The actual code goes further and collects all errors (C<die>s, exceptions)
		1251	that occurred during request processing. The C<result> method detects
		1252	whether an exception as thrown (it is stored inside the $txn object)
		1253	and just throws the exception, which means connection errors and other
		1254	problems get reported tot he code that tries to use the result, not in a
		1255	random callback.
		1256
		1257	All of this enables the following usage styles:
		1258
		1259	1. Blocking:
		1260
		1261	my $data = $fcp->client_get ($url);
		1262
		1263	2. Blocking, but running in parallel:
		1264
		1265	my @datas = map $_->result,
		1266	map $fcp->txn_client_get ($_),
		1267	@urls;
		1268
		1269	Both blocking examples work without the module user having to know
		1270	anything about events.
		1271
		1272	3a. Event-based in a main program, using any supported event module:
		1273
		1274	use EV;
		1275
		1276	$fcp->txn_client_get ($url)->cb (sub {
		1277	my $txn = shift;
		1278	my $data = $txn->result;
		1279	...
		1280	});
		1281
		1282	EV::loop;
		1283
		1284	3b. The module user could use AnyEvent, too:
		1285
		1286	use AnyEvent;
		1287
		1288	my $quit = AnyEvent->condvar;
		1289
		1290	$fcp->txn_client_get ($url)->cb (sub {
		1291	...
		1292	$quit->send;
		1293	});
		1294
		1295	$quit->recv;
		1296
		1297
		1298	=head1 BENCHMARKS
		1299
		1300	To give you an idea of the performance and overheads that AnyEvent adds
		1301	over the event loops themselves and to give you an impression of the speed
		1302	of various event loops I prepared some benchmarks.
		1303
		1304	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1305
		1306	Here is a benchmark of various supported event models used natively and
		1307	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1308	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1309	which it is), lets them fire exactly once and destroys them again.
		1310
		1311	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		1312	distribution.
		1313
		1314	=head3 Explanation of the columns
		1315
		1316	I<watcher> is the number of event watchers created/destroyed. Since
		1317	different event models feature vastly different performances, each event
		1318	loop was given a number of watchers so that overall runtime is acceptable
		1319	and similar between tested event loop (and keep them from crashing): Glib
		1320	would probably take thousands of years if asked to process the same number
		1321	of watchers as EV in this benchmark.
		1322
		1323	I<bytes> is the number of bytes (as measured by the resident set size,
		1324	RSS) consumed by each watcher. This method of measuring captures both C
		1325	and Perl-based overheads.
		1326
		1327	I<create> is the time, in microseconds (millionths of seconds), that it
		1328	takes to create a single watcher. The callback is a closure shared between
		1329	all watchers, to avoid adding memory overhead. That means closure creation
		1330	and memory usage is not included in the figures.
		1331
		1332	I<invoke> is the time, in microseconds, used to invoke a simple
		1333	callback. The callback simply counts down a Perl variable and after it was
		1334	invoked "watcher" times, it would C<< ->send >> a condvar once to
		1335	signal the end of this phase.
		1336
		1337	I<destroy> is the time, in microseconds, that it takes to destroy a single
		1338	watcher.
		1339
		1340	=head3 Results
		1341
		1342	name watchers bytes create invoke destroy comment
		1343	EV/EV 400000 244 0.56 0.46 0.31 EV native interface
		1344	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
		1345	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
		1346	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
		1347	Event/Event 16000 516 31.88 31.30 0.85 Event native interface
		1348	Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
		1349	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
		1350	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
		1351	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
		1352	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
		1353
		1354	=head3 Discussion
		1355
		1356	The benchmark does I<not> measure scalability of the event loop very
		1357	well. For example, a select-based event loop (such as the pure perl one)
		1358	can never compete with an event loop that uses epoll when the number of
		1359	file descriptors grows high. In this benchmark, all events become ready at
		1360	the same time, so select/poll-based implementations get an unnatural speed
		1361	boost.
		1362
		1363	Also, note that the number of watchers usually has a nonlinear effect on
		1364	overall speed, that is, creating twice as many watchers doesn't take twice
		1365	the time - usually it takes longer. This puts event loops tested with a
		1366	higher number of watchers at a disadvantage.
		1367
		1368	To put the range of results into perspective, consider that on the
		1369	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1370	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1371	cycles with POE.
		1372
		1373	C<EV> is the sole leader regarding speed and memory use, which are both
		1374	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1375	far less memory than any other event loop and is still faster than Event
		1376	natively.
		1377
		1378	The pure perl implementation is hit in a few sweet spots (both the
		1379	constant timeout and the use of a single fd hit optimisations in the perl
		1380	interpreter and the backend itself). Nevertheless this shows that it
		1381	adds very little overhead in itself. Like any select-based backend its
		1382	performance becomes really bad with lots of file descriptors (and few of
		1383	them active), of course, but this was not subject of this benchmark.
		1384
		1385	The C<Event> module has a relatively high setup and callback invocation
		1386	cost, but overall scores in on the third place.
		1387
		1388	C<Glib>'s memory usage is quite a bit higher, but it features a
		1389	faster callback invocation and overall ends up in the same class as
		1390	C<Event>. However, Glib scales extremely badly, doubling the number of
		1391	watchers increases the processing time by more than a factor of four,
		1392	making it completely unusable when using larger numbers of watchers
		1393	(note that only a single file descriptor was used in the benchmark, so
		1394	inefficiencies of C<poll> do not account for this).
		1395
		1396	The C<Tk> adaptor works relatively well. The fact that it crashes with
		1397	more than 2000 watchers is a big setback, however, as correctness takes
		1398	precedence over speed. Nevertheless, its performance is surprising, as the
		1399	file descriptor is dup()ed for each watcher. This shows that the dup()
		1400	employed by some adaptors is not a big performance issue (it does incur a
		1401	hidden memory cost inside the kernel which is not reflected in the figures
		1402	above).
		1403
		1404	C<POE>, regardless of underlying event loop (whether using its pure perl
		1405	select-based backend or the Event module, the POE-EV backend couldn't
		1406	be tested because it wasn't working) shows abysmal performance and
		1407	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		1408	as EV watchers, and 10 times as much memory as Event (the high memory
		1409	requirements are caused by requiring a session for each watcher). Watcher
		1410	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		1411	implementation.
		1412
		1413	The design of the POE adaptor class in AnyEvent can not really account
		1414	for the performance issues, though, as session creation overhead is
		1415	small compared to execution of the state machine, which is coded pretty
		1416	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		1417	using multiple sessions is not a good approach, especially regarding
		1418	memory usage, even the author of POE could not come up with a faster
		1419	design).
		1420
		1421	=head3 Summary
		1422
		1423	=over 4
		1424
		1425	=item * Using EV through AnyEvent is faster than any other event loop
		1426	(even when used without AnyEvent), but most event loops have acceptable
		1427	performance with or without AnyEvent.
		1428
		1429	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		1430	the actual event loop, only with extremely fast event loops such as EV
		1431	adds AnyEvent significant overhead.
		1432
		1433	=item * You should avoid POE like the plague if you want performance or
		1434	reasonable memory usage.
		1435
		1436	=back
		1437
		1438	=head2 BENCHMARKING THE LARGE SERVER CASE
		1439
		1440	This benchmark actually benchmarks the event loop itself. It works by
		1441	creating a number of "servers": each server consists of a socket pair, a
		1442	timeout watcher that gets reset on activity (but never fires), and an I/O
		1443	watcher waiting for input on one side of the socket. Each time the socket
		1444	watcher reads a byte it will write that byte to a random other "server".
		1445
		1446	The effect is that there will be a lot of I/O watchers, only part of which
		1447	are active at any one point (so there is a constant number of active
		1448	fds for each loop iteration, but which fds these are is random). The
		1449	timeout is reset each time something is read because that reflects how
		1450	most timeouts work (and puts extra pressure on the event loops).
		1451
		1452	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		1453	(1%) are active. This mirrors the activity of large servers with many
		1454	connections, most of which are idle at any one point in time.
		1455
		1456	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1457	distribution.
		1458
		1459	=head3 Explanation of the columns
		1460
		1461	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1462	each server has a read and write socket end).
		1463
		1464	I<create> is the time it takes to create a socket pair (which is
		1465	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1466
		1467	I<request>, the most important value, is the time it takes to handle a
		1468	single "request", that is, reading the token from the pipe and forwarding
		1469	it to another server. This includes deleting the old timeout and creating
		1470	a new one that moves the timeout into the future.
		1471
		1472	=head3 Results
		1473
		1474	name sockets create request
		1475	EV 20000 69.01 11.16
		1476	Perl 20000 73.32 35.87
		1477	Event 20000 212.62 257.32
		1478	Glib 20000 651.16 1896.30
		1479	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1480
		1481	=head3 Discussion
		1482
		1483	This benchmark I<does> measure scalability and overall performance of the
		1484	particular event loop.
		1485
		1486	EV is again fastest. Since it is using epoll on my system, the setup time
		1487	is relatively high, though.
		1488
		1489	Perl surprisingly comes second. It is much faster than the C-based event
		1490	loops Event and Glib.
		1491
		1492	Event suffers from high setup time as well (look at its code and you will
		1493	understand why). Callback invocation also has a high overhead compared to
		1494	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1495	uses select or poll in basically all documented configurations.
		1496
		1497	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1498	clearly fails to perform with many filehandles or in busy servers.
		1499
		1500	POE is still completely out of the picture, taking over 1000 times as long
		1501	as EV, and over 100 times as long as the Perl implementation, even though
		1502	it uses a C-based event loop in this case.
		1503
		1504	=head3 Summary
		1505
		1506	=over 4
		1507
		1508	=item * The pure perl implementation performs extremely well.
		1509
		1510	=item * Avoid Glib or POE in large projects where performance matters.
		1511
		1512	=back
		1513
		1514	=head2 BENCHMARKING SMALL SERVERS
		1515
		1516	While event loops should scale (and select-based ones do not...) even to
		1517	large servers, most programs we (or I :) actually write have only a few
		1518	I/O watchers.
		1519
		1520	In this benchmark, I use the same benchmark program as in the large server
		1521	case, but it uses only eight "servers", of which three are active at any
		1522	one time. This should reflect performance for a small server relatively
		1523	well.
		1524
		1525	The columns are identical to the previous table.
		1526
		1527	=head3 Results
		1528
		1529	name sockets create request
		1530	EV 16 20.00 6.54
		1531	Perl 16 25.75 12.62
		1532	Event 16 81.27 35.86
		1533	Glib 16 32.63 15.48
		1534	POE 16 261.87 276.28 uses POE::Loop::Event
		1535
		1536	=head3 Discussion
		1537
		1538	The benchmark tries to test the performance of a typical small
		1539	server. While knowing how various event loops perform is interesting, keep
		1540	in mind that their overhead in this case is usually not as important, due
		1541	to the small absolute number of watchers (that is, you need efficiency and
		1542	speed most when you have lots of watchers, not when you only have a few of
		1543	them).
		1544
		1545	EV is again fastest.
		1546
		1547	Perl again comes second. It is noticeably faster than the C-based event
		1548	loops Event and Glib, although the difference is too small to really
		1549	matter.
		1550
		1551	POE also performs much better in this case, but is is still far behind the
		1552	others.
		1553
		1554	=head3 Summary
		1555
		1556	=over 4
		1557
		1558	=item * C-based event loops perform very well with small number of
		1559	watchers, as the management overhead dominates.
		1560
		1561	=back
		1562
		1563
		1564	=head1 FORK
		1565
		1566	Most event libraries are not fork-safe. The ones who are usually are
		1567	because they rely on inefficient but fork-safe C<select> or C<poll>
		1568	calls. Only L<EV> is fully fork-aware.
		1569
		1570	If you have to fork, you must either do so I<before> creating your first
		1571	watcher OR you must not use AnyEvent at all in the child.
		1572
		1573
		1574	=head1 SECURITY CONSIDERATIONS
		1575
		1576	AnyEvent can be forced to load any event model via
		1577	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		1578	execute arbitrary code or directly gain access, it can easily be used to
		1579	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		1580	will not be active when the program uses a different event model than
		1581	specified in the variable.
		1582
		1583	You can make AnyEvent completely ignore this variable by deleting it
		1584	before the first watcher gets created, e.g. with a C<BEGIN> block:
		1585
		1586	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		1587
		1588	use AnyEvent;
		1589
		1590	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		1591	be used to probe what backend is used and gain other information (which is
		1592	probably even less useful to an attacker than PERL_ANYEVENT_MODEL).
		1593
139		1594
140	=head1 SEE ALSO	1595	=head1 SEE ALSO
141		1596
142	L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>,	1597	Utility functions: L<AnyEvent::Util>.
143	L<AnyEvent::Impl::Coro>,	1598
144	L<AnyEvent::Impl::Event>,	1599	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
145	L<AnyEvent::Impl::Glib>,	1600	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
		1601
		1602	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		1603	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		1604	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
146	L<AnyEvent::Impl::Tk>.	1605	L<AnyEvent::Impl::POE>.
147		1606
148	=head1	1607	Non-blocking file handles, sockets, TCP clients and
		1608	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>.
		1609
		1610	Asynchronous DNS: L<AnyEvent::DNS>.
		1611
		1612	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>,
		1613
		1614	Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>.
		1615
		1616
		1617	=head1 AUTHOR
		1618
		1619	Marc Lehmann <schmorp@schmorp.de>
		1620	http://home.schmorp.de/
149		1621
150	=cut	1622	=cut
151		1623
152	1	1624	1
153		1625

Diff Legend

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