ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.28 by root, Sat Oct 27 15:10:09 2007 UTC vs.
Revision 1.143 by root, Wed May 28 23:57:38 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, Perl - 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
…		…
14		14
15	my $w = AnyEvent->timer (after => $seconds, cb => sub {	15	my $w = AnyEvent->timer (after => $seconds, cb => sub {
16	...	16	...
17	});	17	});
18		18
19	my $w = AnyEvent->condvar; # stores wether a condition was flagged	19	my $w = AnyEvent->condvar; # stores whether a condition was flagged
		20	$w->send; # wake up current and all future recv's
20	$w->wait; # enters "main loop" till $condvar gets ->broadcast	21	$w->recv; # enters "main loop" till $condvar gets ->send
21	$w->broadcast; # wake up current and all 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.
22		76
23	=head1 DESCRIPTION	77	=head1 DESCRIPTION
24		78
25	L<AnyEvent> provides an identical interface to multiple event loops. This	79	L<AnyEvent> provides an identical interface to multiple event loops. This
26	allows module authors to utilise an event loop without forcing module	80	allows module authors to utilise an event loop without forcing module
27	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
28	peacefully at any one time).	82	peacefully at any one time).
29		83
30	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>
31	module.	85	module.
32		86
33	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
34	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
35	loaded: L<Coro::Event>, L<Event>, L<Glib>, L<Tk>. The first one found is	89	following modules is already loaded: L<EV>,
36	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>,
37	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
38	used. If still none could be found, AnyEvent will fall back to a pure-perl	92	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
39	event loop, which is also not very efficient.	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.
40		97
41	Because AnyEvent first checks for modules that are already loaded, loading	98	Because AnyEvent first checks for modules that are already loaded, loading
42	an Event model explicitly before first using AnyEvent will likely make	99	an event model explicitly before first using AnyEvent will likely make
43	that model the default. For example:	100	that model the default. For example:
44		101
45	use Tk;	102	use Tk;
46	use AnyEvent;	103	use AnyEvent;
47		104
48	# .. AnyEvent will likely default to Tk	105	# .. AnyEvent will likely default to Tk
49		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
50	The pure-perl implementation of AnyEvent is called	111	The pure-perl implementation of AnyEvent is called
51	C<AnyEvent::Impl::Perl>. Like other event modules you can load it	112	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
52	explicitly.	113	explicitly and enjoy the high availability of that event loop :)
53		114
54	=head1 WATCHERS	115	=head1 WATCHERS
55		116
56	AnyEvent has the central concept of a I<watcher>, which is an object that	117	AnyEvent has the central concept of a I<watcher>, which is an object that
57	stores relevant data for each kind of event you are waiting for, such as	118	stores relevant data for each kind of event you are waiting for, such as
58	the callback to call, the filehandle to watch, etc.	119	the callback to call, the file handle to watch, etc.
59		120
60	These watchers are normal Perl objects with normal Perl lifetime. After	121	These watchers are normal Perl objects with normal Perl lifetime. After
61	creating a watcher it will immediately "watch" for events and invoke	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
62	the callback. To disable the watcher you have to destroy it (e.g. by	126	To disable the watcher you have to destroy it (e.g. by setting the
63	setting the variable that stores it to C<undef> or otherwise deleting all	127	variable you store it in to C<undef> or otherwise deleting all references
64	references to it).	128	to it).
65		129
66	All watchers are created by calling a method on the C<AnyEvent> class.	130	All watchers are created by calling a method on the C<AnyEvent> class.
67		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
68	=head2 IO WATCHERS	146	=head2 I/O WATCHERS
69		147
70	You can create I/O watcher by calling the C<< AnyEvent->io >> method with	148	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
71	the following mandatory arguments:	149	with the following mandatory key-value pairs as arguments:
72		150
73	C<fh> the Perl I<filehandle> (not filedescriptor) to watch for	151	C<fh> the Perl I<file handle> (I<not> file descriptor) to watch
74	events. C<poll> must be a string that is either C<r> or C<w>, that creates	152	for events. C<poll> must be a string that is either C<r> or C<w>,
75	a watcher waiting for "r"eadable or "w"ritable events. C<cb> teh callback	153	which creates a watcher waiting for "r"eadable or "w"ritable events,
76	to invoke everytime the filehandle becomes ready.	154	respectively. C<cb> is the callback to invoke each time the file handle
		155	becomes ready.
77		156
78	Only one io watcher per C<fh> and C<poll> combination is allowed (i.e. on	157	Although the callback might get passed parameters, their value and
79	a socket you can have one r + one w, not any more (limitation comes from	158	presence is undefined and you cannot rely on them. Portable AnyEvent
80	Tk - if you are sure you are not using Tk this limitation is gone).	159	callbacks cannot use arguments passed to I/O watcher callbacks.
81		160
82	Filehandles will be kept alive, so as long as the watcher exists, the	161	The I/O watcher might use the underlying file descriptor or a copy of it.
83	filehandle exists, too.	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.
84		168
85	Example:	169	Example:
86		170
87	# wait for readability of STDIN, then read a line and disable the watcher	171	# wait for readability of STDIN, then read a line and disable the watcher
88	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	172	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
…		…
94	=head2 TIME WATCHERS	178	=head2 TIME WATCHERS
95		179
96	You can create a time watcher by calling the C<< AnyEvent->timer >>	180	You can create a time watcher by calling the C<< AnyEvent->timer >>
97	method with the following mandatory arguments:	181	method with the following mandatory arguments:
98		182
99	C<after> after how many seconds (fractions are supported) should the timer	183	C<after> specifies after how many seconds (fractional values are
100	activate. C<cb> the callback to invoke.	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.
101		190
102	The timer callback will be invoked at most once: if you want a repeating	191	The timer callback will be invoked at most once: if you want a repeating
103	timer you have to create a new watcher (this is a limitation by both Tk	192	timer you have to create a new watcher (this is a limitation by both Tk
104	and Glib).	193	and Glib).
105		194
…		…
109	my $w = AnyEvent->timer (after => 7.7, cb => sub {	198	my $w = AnyEvent->timer (after => 7.7, cb => sub {
110	warn "timeout\n";	199	warn "timeout\n";
111	});	200	});
112		201
113	# to cancel the timer:	202	# to cancel the timer:
114	undef $w	203	undef $w;
115		204
116	=head2 CONDITION WATCHERS	205	Example 2:
117		206
118	Condition watchers can be created by calling the C<< AnyEvent->condvar >>	207	# fire an event after 0.5 seconds, then roughly every second
119	method without any arguments.	208	my $w;
120		209
121	A condition watcher watches for a condition - precisely that the C<<	210	my $cb = sub {
122	->broadcast >> method has been called.	211	# cancel the old timer while creating a new one
		212	$w = AnyEvent->timer (after => 1, cb => $cb);
		213	};
123		214
124	The watcher has only two methods:	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	AnyEvent has two additional methods that return the "current time":
125		239
126	=over 4	240	=over 4
127		241
128	=item $cv->wait	242	=item AnyEvent->time
129		243
130	Wait (blocking if necessary) until the C<< ->broadcast >> method has been	244	This returns the "current wallclock time" as a fractional number of
131	called on c<$cv>, while servicing other watchers normally.	245	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		246	return, and the result is guaranteed to be compatible with those).
132		247
133	Not all event models support a blocking wait - some die in that case, so	248	It progresses independently of any event loop processing.
134	if you are using this from a module, never require a blocking wait, but
135	let the caller decide wether the call will block or not (for example,
136	by coupling condition variables with some kind of request results and
137	supporting callbacks so the caller knows that getting the result will not
138	block, while still suppporting blockign waits if the caller so desires).
139		249
140	You can only wait once on a condition - additional calls will return	250	In almost all cases (in all cases if you don't care), this is the function
141	immediately.	251	to call when you want to know the current time.
142		252
143	=item $cv->broadcast	253	=item AnyEvent->now
144		254
145	Flag the condition as ready - a running C<< ->wait >> and all further	255	This also returns the "current wallclock time", but unlike C<time>, above,
146	calls to C<wait> will return after this method has been called. If nobody	256	this value might change only once per event loop iteration, depending on
147	is waiting the broadcast will be remembered..	257	the event loop (most return the same time as C<time>, above). This is the
		258	time that AnyEvent timers get scheduled against.
148		259
149	Example:	260	For a practical example of when these times differ, consider L<Event::Lib>
		261	and L<EV> and the following set-up:
		262
		263	The event loop is running and has just invoked one of your callback at
		264	time=500 (assume no other callbacks delay processing). In your callback,
		265	you wait a second by executing C<sleep 1> (blocking the process for a
		266	second) and then (at time=501) you create a relative timer that fires
		267	after three seconds.
		268
		269	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		270	both return C<501>, because that is the current time, and the timer will
		271	be scheduled to fire at time=504 (C<501> + C<3>).
		272
		273	With L<EV>m C<< AnyEvent->time >> returns C<501> (as that is the current
		274	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		275	last event processing phase started. With L<EV>, your timer gets scheduled
		276	to run at time=503 (C<500> + C<3>).
		277
		278	In one sense, L<Event::Lib> is more exact, as it uses the current time
		279	regardless of any delays introduced by event processing. However, most
		280	callbacks do not expect large delays in processing, so this causes a
		281	higher drift (and a lot more syscalls to get the current time).
		282
		283	In another sense, L<EV> is more exact, as your timer will be scheduled at
		284	the same time, regardless of how long event processing actually took.
		285
		286	In either case, if you care (and in most cases, you don't), then you
		287	can get whatever behaviour you want with any event loop, by taking the
		288	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		289	account.
		290
		291	=back
		292
		293	=head2 SIGNAL WATCHERS
		294
		295	You can watch for signals using a signal watcher, C<signal> is the signal
		296	I<name> without any C<SIG> prefix, C<cb> is the Perl callback to
		297	be invoked whenever a signal occurs.
		298
		299	Although the callback might get passed parameters, their value and
		300	presence is undefined and you cannot rely on them. Portable AnyEvent
		301	callbacks cannot use arguments passed to signal watcher callbacks.
		302
		303	Multiple signal occurrences can be clumped together into one callback
		304	invocation, and callback invocation will be synchronous. Synchronous means
		305	that it might take a while until the signal gets handled by the process,
		306	but it is guaranteed not to interrupt any other callbacks.
		307
		308	The main advantage of using these watchers is that you can share a signal
		309	between multiple watchers.
		310
		311	This watcher might use C<%SIG>, so programs overwriting those signals
		312	directly will likely not work correctly.
		313
		314	Example: exit on SIGINT
		315
		316	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		317
		318	=head2 CHILD PROCESS WATCHERS
		319
		320	You can also watch on a child process exit and catch its exit status.
		321
		322	The child process is specified by the C<pid> argument (if set to C<0>, it
		323	watches for any child process exit). The watcher will trigger as often
		324	as status change for the child are received. This works by installing a
		325	signal handler for C<SIGCHLD>. The callback will be called with the pid
		326	and exit status (as returned by waitpid), so unlike other watcher types,
		327	you I<can> rely on child watcher callback arguments.
		328
		329	There is a slight catch to child watchers, however: you usually start them
		330	I<after> the child process was created, and this means the process could
		331	have exited already (and no SIGCHLD will be sent anymore).
		332
		333	Not all event models handle this correctly (POE doesn't), but even for
		334	event models that I<do> handle this correctly, they usually need to be
		335	loaded before the process exits (i.e. before you fork in the first place).
		336
		337	This means you cannot create a child watcher as the very first thing in an
		338	AnyEvent program, you I<have> to create at least one watcher before you
		339	C<fork> the child (alternatively, you can call C<AnyEvent::detect>).
		340
		341	Example: fork a process and wait for it
		342
		343	my $done = AnyEvent->condvar;
		344
		345	my $pid = fork or exit 5;
		346
		347	my $w = AnyEvent->child (
		348	pid => $pid,
		349	cb => sub {
		350	my ($pid, $status) = @_;
		351	warn "pid $pid exited with status $status";
		352	$done->send;
		353	},
		354	);
		355
		356	# do something else, then wait for process exit
		357	$done->recv;
		358
		359	=head2 CONDITION VARIABLES
		360
		361	If you are familiar with some event loops you will know that all of them
		362	require you to run some blocking "loop", "run" or similar function that
		363	will actively watch for new events and call your callbacks.
		364
		365	AnyEvent is different, it expects somebody else to run the event loop and
		366	will only block when necessary (usually when told by the user).
		367
		368	The instrument to do that is called a "condition variable", so called
		369	because they represent a condition that must become true.
		370
		371	Condition variables can be created by calling the C<< AnyEvent->condvar
		372	>> method, usually without arguments. The only argument pair allowed is
		373	C<cb>, which specifies a callback to be called when the condition variable
		374	becomes true.
		375
		376	After creation, the condition variable is "false" until it becomes "true"
		377	by calling the C<send> method (or calling the condition variable as if it
		378	were a callback, read about the caveats in the description for the C<<
		379	->send >> method).
		380
		381	Condition variables are similar to callbacks, except that you can
		382	optionally wait for them. They can also be called merge points - points
		383	in time where multiple outstanding events have been processed. And yet
		384	another way to call them is transactions - each condition variable can be
		385	used to represent a transaction, which finishes at some point and delivers
		386	a result.
		387
		388	Condition variables are very useful to signal that something has finished,
		389	for example, if you write a module that does asynchronous http requests,
		390	then a condition variable would be the ideal candidate to signal the
		391	availability of results. The user can either act when the callback is
		392	called or can synchronously C<< ->recv >> for the results.
		393
		394	You can also use them to simulate traditional event loops - for example,
		395	you can block your main program until an event occurs - for example, you
		396	could C<< ->recv >> in your main program until the user clicks the Quit
		397	button of your app, which would C<< ->send >> the "quit" event.
		398
		399	Note that condition variables recurse into the event loop - if you have
		400	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		401	lose. Therefore, condition variables are good to export to your caller, but
		402	you should avoid making a blocking wait yourself, at least in callbacks,
		403	as this asks for trouble.
		404
		405	Condition variables are represented by hash refs in perl, and the keys
		406	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		407	easy (it is often useful to build your own transaction class on top of
		408	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		409	it's C<new> method in your own C<new> method.
		410
		411	There are two "sides" to a condition variable - the "producer side" which
		412	eventually calls C<< -> send >>, and the "consumer side", which waits
		413	for the send to occur.
		414
		415	Example: wait for a timer.
150		416
151	# wait till the result is ready	417	# wait till the result is ready
152	my $result_ready = AnyEvent->condvar;	418	my $result_ready = AnyEvent->condvar;
153		419
154	# do something such as adding a timer	420	# do something such as adding a timer
155	# or socket watcher the calls $result_ready->broadcast	421	# or socket watcher the calls $result_ready->send
156	# when the "result" is ready.	422	# when the "result" is ready.
		423	# in this case, we simply use a timer:
		424	my $w = AnyEvent->timer (
		425	after => 1,
		426	cb => sub { $result_ready->send },
		427	);
157		428
		429	# this "blocks" (while handling events) till the callback
		430	# calls send
158	$result_ready->wait;	431	$result_ready->recv;
		432
		433	Example: wait for a timer, but take advantage of the fact that
		434	condition variables are also code references.
		435
		436	my $done = AnyEvent->condvar;
		437	my $delay = AnyEvent->timer (after => 5, cb => $done);
		438	$done->recv;
		439
		440	=head3 METHODS FOR PRODUCERS
		441
		442	These methods should only be used by the producing side, i.e. the
		443	code/module that eventually sends the signal. Note that it is also
		444	the producer side which creates the condvar in most cases, but it isn't
		445	uncommon for the consumer to create it as well.
		446
		447	=over 4
		448
		449	=item $cv->send (...)
		450
		451	Flag the condition as ready - a running C<< ->recv >> and all further
		452	calls to C<recv> will (eventually) return after this method has been
		453	called. If nobody is waiting the send will be remembered.
		454
		455	If a callback has been set on the condition variable, it is called
		456	immediately from within send.
		457
		458	Any arguments passed to the C<send> call will be returned by all
		459	future C<< ->recv >> calls.
		460
		461	Condition variables are overloaded so one can call them directly
		462	(as a code reference). Calling them directly is the same as calling
		463	C<send>. Note, however, that many C-based event loops do not handle
		464	overloading, so as tempting as it may be, passing a condition variable
		465	instead of a callback does not work. Both the pure perl and EV loops
		466	support overloading, however, as well as all functions that use perl to
		467	invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
		468	example).
		469
		470	=item $cv->croak ($error)
		471
		472	Similar to send, but causes all call's to C<< ->recv >> to invoke
		473	C<Carp::croak> with the given error message/object/scalar.
		474
		475	This can be used to signal any errors to the condition variable
		476	user/consumer.
		477
		478	=item $cv->begin ([group callback])
		479
		480	=item $cv->end
		481
		482	These two methods are EXPERIMENTAL and MIGHT CHANGE.
		483
		484	These two methods can be used to combine many transactions/events into
		485	one. For example, a function that pings many hosts in parallel might want
		486	to use a condition variable for the whole process.
		487
		488	Every call to C<< ->begin >> will increment a counter, and every call to
		489	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		490	>>, the (last) callback passed to C<begin> will be executed. That callback
		491	is I<supposed> to call C<< ->send >>, but that is not required. If no
		492	callback was set, C<send> will be called without any arguments.
		493
		494	Let's clarify this with the ping example:
		495
		496	my $cv = AnyEvent->condvar;
		497
		498	my %result;
		499	$cv->begin (sub { $cv->send (\%result) });
		500
		501	for my $host (@list_of_hosts) {
		502	$cv->begin;
		503	ping_host_then_call_callback $host, sub {
		504	$result{$host} = ...;
		505	$cv->end;
		506	};
		507	}
		508
		509	$cv->end;
		510
		511	This code fragment supposedly pings a number of hosts and calls
		512	C<send> after results for all then have have been gathered - in any
		513	order. To achieve this, the code issues a call to C<begin> when it starts
		514	each ping request and calls C<end> when it has received some result for
		515	it. Since C<begin> and C<end> only maintain a counter, the order in which
		516	results arrive is not relevant.
		517
		518	There is an additional bracketing call to C<begin> and C<end> outside the
		519	loop, which serves two important purposes: first, it sets the callback
		520	to be called once the counter reaches C<0>, and second, it ensures that
		521	C<send> is called even when C<no> hosts are being pinged (the loop
		522	doesn't execute once).
		523
		524	This is the general pattern when you "fan out" into multiple subrequests:
		525	use an outer C<begin>/C<end> pair to set the callback and ensure C<end>
		526	is called at least once, and then, for each subrequest you start, call
		527	C<begin> and for each subrequest you finish, call C<end>.
159		528
160	=back	529	=back
161		530
162	=head2 SIGNAL WATCHERS	531	=head3 METHODS FOR CONSUMERS
163		532
164	You can listen for signals using a signal watcher, C<signal> is the signal	533	These methods should only be used by the consuming side, i.e. the
165	I<name> without any C<SIG> prefix. Multiple signals events can be clumped	534	code awaits the condition.
166	together into one callback invocation, and callback invocation might or
167	might not be asynchronous.
168		535
169	These watchers might use C<%SIG>, so programs overwriting those signals	536	=over 4
170	directly will likely not work correctly.
171		537
172	Example: exit on SIGINT	538	=item $cv->recv
173		539
174	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });	540	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		541	>> methods have been called on c<$cv>, while servicing other watchers
		542	normally.
175		543
176	=head2 CHILD PROCESS WATCHERS	544	You can only wait once on a condition - additional calls are valid but
		545	will return immediately.
177		546
178	You can also listen for the status of a child process specified by the	547	If an error condition has been set by calling C<< ->croak >>, then this
179	C<pid> argument. The watcher will only trigger once. This works by	548	function will call C<croak>.
180	installing a signal handler for C<SIGCHLD>.
181		549
182	Example: wait for pid 1333	550	In list context, all parameters passed to C<send> will be returned,
		551	in scalar context only the first one will be returned.
183		552
184	my $w = AnyEvent->child (pid => 1333, cb => sub { warn "exit status $?" });	553	Not all event models support a blocking wait - some die in that case
		554	(programs might want to do that to stay interactive), so I<if you are
		555	using this from a module, never require a blocking wait>, but let the
		556	caller decide whether the call will block or not (for example, by coupling
		557	condition variables with some kind of request results and supporting
		558	callbacks so the caller knows that getting the result will not block,
		559	while still supporting blocking waits if the caller so desires).
185		560
186	=head1 GLOBALS	561	Another reason I<never> to C<< ->recv >> in a module is that you cannot
		562	sensibly have two C<< ->recv >>'s in parallel, as that would require
		563	multiple interpreters or coroutines/threads, none of which C<AnyEvent>
		564	can supply.
		565
		566	The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
		567	fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
		568	versions and also integrates coroutines into AnyEvent, making blocking
		569	C<< ->recv >> calls perfectly safe as long as they are done from another
		570	coroutine (one that doesn't run the event loop).
		571
		572	You can ensure that C<< -recv >> never blocks by setting a callback and
		573	only calling C<< ->recv >> from within that callback (or at a later
		574	time). This will work even when the event loop does not support blocking
		575	waits otherwise.
		576
		577	=item $bool = $cv->ready
		578
		579	Returns true when the condition is "true", i.e. whether C<send> or
		580	C<croak> have been called.
		581
		582	=item $cb = $cv->cb ([new callback])
		583
		584	This is a mutator function that returns the callback set and optionally
		585	replaces it before doing so.
		586
		587	The callback will be called when the condition becomes "true", i.e. when
		588	C<send> or C<croak> are called. Calling C<recv> inside the callback
		589	or at any later time is guaranteed not to block.
		590
		591	=back
		592
		593	=head1 GLOBAL VARIABLES AND FUNCTIONS
187		594
188	=over 4	595	=over 4
189		596
190	=item $AnyEvent::MODEL	597	=item $AnyEvent::MODEL
191		598
…		…
195	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case	602	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case
196	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).	603	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).
197		604
198	The known classes so far are:	605	The known classes so far are:
199		606
200	AnyEvent::Impl::Coro based on Coro::Event, best choice.
201	EV::AnyEvent based on EV (an interface to libevent)	607	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
202	AnyEvent::Impl::Event based on Event, also best choice :)	608	AnyEvent::Impl::Event based on Event, second best choice.
		609	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
203	AnyEvent::Impl::Glib based on Glib, second-best choice.	610	AnyEvent::Impl::Glib based on Glib, third-best choice.
204	AnyEvent::Impl::Tk based on Tk, very bad choice.	611	AnyEvent::Impl::Tk based on Tk, very bad choice.
205	AnyEvent::Impl::Perl pure-perl implementation, inefficient.	612	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
		613	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		614	AnyEvent::Impl::POE based on POE, not generic enough for full support.
		615
		616	There is no support for WxWidgets, as WxWidgets has no support for
		617	watching file handles. However, you can use WxWidgets through the
		618	POE Adaptor, as POE has a Wx backend that simply polls 20 times per
		619	second, which was considered to be too horrible to even consider for
		620	AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
		621	it's adaptor.
		622
		623	AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
		624	autodetecting them.
206		625
207	=item AnyEvent::detect	626	=item AnyEvent::detect
208		627
209	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model if	628	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
210	necessary. You should only call this function right before you would have	629	if necessary. You should only call this function right before you would
211	created an AnyEvent watcher anyway, that is, very late at runtime.	630	have created an AnyEvent watcher anyway, that is, as late as possible at
		631	runtime.
		632
		633	=item $guard = AnyEvent::post_detect { BLOCK }
		634
		635	Arranges for the code block to be executed as soon as the event model is
		636	autodetected (or immediately if this has already happened).
		637
		638	If called in scalar or list context, then it creates and returns an object
		639	that automatically removes the callback again when it is destroyed. See
		640	L<Coro::BDB> for a case where this is useful.
		641
		642	=item @AnyEvent::post_detect
		643
		644	If there are any code references in this array (you can C<push> to it
		645	before or after loading AnyEvent), then they will called directly after
		646	the event loop has been chosen.
		647
		648	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		649	if it contains a true value then the event loop has already been detected,
		650	and the array will be ignored.
		651
		652	Best use C<AnyEvent::post_detect { BLOCK }> instead.
212		653
213	=back	654	=back
214		655
215	=head1 WHAT TO DO IN A MODULE	656	=head1 WHAT TO DO IN A MODULE
216		657
217	As a module author, you should "use AnyEvent" and call AnyEvent methods	658	As a module author, you should C<use AnyEvent> and call AnyEvent methods
218	freely, but you should not load a specific event module or rely on it.	659	freely, but you should not load a specific event module or rely on it.
219		660
220	Be careful when you create watchers in the module body - Anyevent will	661	Be careful when you create watchers in the module body - AnyEvent will
221	decide which event module to use as soon as the first method is called, so	662	decide which event module to use as soon as the first method is called, so
222	by calling AnyEvent in your module body you force the user of your module	663	by calling AnyEvent in your module body you force the user of your module
223	to load the event module first.	664	to load the event module first.
224		665
		666	Never call C<< ->recv >> on a condition variable unless you I<know> that
		667	the C<< ->send >> method has been called on it already. This is
		668	because it will stall the whole program, and the whole point of using
		669	events is to stay interactive.
		670
		671	It is fine, however, to call C<< ->recv >> when the user of your module
		672	requests it (i.e. if you create a http request object ad have a method
		673	called C<results> that returns the results, it should call C<< ->recv >>
		674	freely, as the user of your module knows what she is doing. always).
		675
225	=head1 WHAT TO DO IN THE MAIN PROGRAM	676	=head1 WHAT TO DO IN THE MAIN PROGRAM
226		677
227	There will always be a single main program - the only place that should	678	There will always be a single main program - the only place that should
228	dictate which event model to use.	679	dictate which event model to use.
229		680
230	If it doesn't care, it can just "use AnyEvent" and use it itself, or not	681	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
231	do anything special and let AnyEvent decide which implementation to chose.	682	do anything special (it does not need to be event-based) and let AnyEvent
		683	decide which implementation to chose if some module relies on it.
232		684
233	If the main program relies on a specific event model (for example, in Gtk2	685	If the main program relies on a specific event model - for example, in
234	programs you have to rely on either Glib or Glib::Event), you should load	686	Gtk2 programs you have to rely on the Glib module - you should load the
235	it before loading AnyEvent or any module that uses it, generally, as early	687	event module before loading AnyEvent or any module that uses it: generally
236	as possible. The reason is that modules might create watchers when they	688	speaking, you should load it as early as possible. The reason is that
237	are loaded, and AnyEvent will decide on the event model to use as soon as	689	modules might create watchers when they are loaded, and AnyEvent will
238	it creates watchers, and it might chose the wrong one unless you load the	690	decide on the event model to use as soon as it creates watchers, and it
239	correct one yourself.	691	might chose the wrong one unless you load the correct one yourself.
240		692
241	You can chose to use a rather inefficient pure-perl implementation by	693	You can chose to use a pure-perl implementation by loading the
242	loading the C<AnyEvent::Impl::Perl> module, but letting AnyEvent chose is	694	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
243	generally better.	695	everywhere, but letting AnyEvent chose the model is generally better.
		696
		697	=head2 MAINLOOP EMULATION
		698
		699	Sometimes (often for short test scripts, or even standalone programs who
		700	only want to use AnyEvent), you do not want to run a specific event loop.
		701
		702	In that case, you can use a condition variable like this:
		703
		704	AnyEvent->condvar->recv;
		705
		706	This has the effect of entering the event loop and looping forever.
		707
		708	Note that usually your program has some exit condition, in which case
		709	it is better to use the "traditional" approach of storing a condition
		710	variable somewhere, waiting for it, and sending it when the program should
		711	exit cleanly.
		712
		713
		714	=head1 OTHER MODULES
		715
		716	The following is a non-exhaustive list of additional modules that use
		717	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		718	in the same program. Some of the modules come with AnyEvent, some are
		719	available via CPAN.
		720
		721	=over 4
		722
		723	=item L<AnyEvent::Util>
		724
		725	Contains various utility functions that replace often-used but blocking
		726	functions such as C<inet_aton> by event-/callback-based versions.
		727
		728	=item L<AnyEvent::Handle>
		729
		730	Provide read and write buffers and manages watchers for reads and writes.
		731
		732	=item L<AnyEvent::Socket>
		733
		734	Provides various utility functions for (internet protocol) sockets,
		735	addresses and name resolution. Also functions to create non-blocking tcp
		736	connections or tcp servers, with IPv6 and SRV record support and more.
		737
		738	=item L<AnyEvent::DNS>
		739
		740	Provides rich asynchronous DNS resolver capabilities.
		741
		742	=item L<AnyEvent::HTTPD>
		743
		744	Provides a simple web application server framework.
		745
		746	=item L<AnyEvent::FastPing>
		747
		748	The fastest ping in the west.
		749
		750	=item L<Net::IRC3>
		751
		752	AnyEvent based IRC client module family.
		753
		754	=item L<Net::XMPP2>
		755
		756	AnyEvent based XMPP (Jabber protocol) module family.
		757
		758	=item L<Net::FCP>
		759
		760	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		761	of AnyEvent.
		762
		763	=item L<Event::ExecFlow>
		764
		765	High level API for event-based execution flow control.
		766
		767	=item L<Coro>
		768
		769	Has special support for AnyEvent via L<Coro::AnyEvent>.
		770
		771	=item L<AnyEvent::AIO>, L<IO::AIO>
		772
		773	Truly asynchronous I/O, should be in the toolbox of every event
		774	programmer. AnyEvent::AIO transparently fuses IO::AIO and AnyEvent
		775	together.
		776
		777	=item L<AnyEvent::BDB>, L<BDB>
		778
		779	Truly asynchronous Berkeley DB access. AnyEvent::AIO transparently fuses
		780	IO::AIO and AnyEvent together.
		781
		782	=item L<IO::Lambda>
		783
		784	The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
		785
		786	=back
244		787
245	=cut	788	=cut
246		789
247	package AnyEvent;	790	package AnyEvent;
248		791
249	no warnings;	792	no warnings;
250	use strict;	793	use strict;
251		794
252	use Carp;	795	use Carp;
253		796
254	our $VERSION = '2.55';	797	our $VERSION = '4.05';
255	our $MODEL;	798	our $MODEL;
256		799
257	our $AUTOLOAD;	800	our $AUTOLOAD;
258	our @ISA;	801	our @ISA;
259		802
		803	our @REGISTRY;
		804
		805	our $WIN32;
		806
		807	BEGIN {
		808	my $win32 = ! ! ($^O =~ /mswin32/i);
		809	eval "sub WIN32(){ $win32 }";
		810	}
		811
260	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;	812	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
261		813
262	our @REGISTRY;	814	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		815
		816	{
		817	my $idx;
		818	$PROTOCOL{$_} = ++$idx
		819	for reverse split /\s*,\s*/,
		820	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		821	}
263		822
264	my @models = (	823	my @models = (
265	[Coro::Event:: => AnyEvent::Impl::Coro::],
266	[EV:: => EV::AnyEvent::],	824	[EV:: => AnyEvent::Impl::EV::],
267	[Event:: => AnyEvent::Impl::Event::],	825	[Event:: => AnyEvent::Impl::Event::],
268	[Glib:: => AnyEvent::Impl::Glib::],
269	[Tk:: => AnyEvent::Impl::Tk::],
270	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],	826	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
		827	# everything below here will not be autoprobed
		828	# as the pureperl backend should work everywhere
		829	# and is usually faster
		830	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		831	[Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
		832	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		833	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		834	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		835	[Wx:: => AnyEvent::Impl::POE::],
		836	[Prima:: => AnyEvent::Impl::POE::],
271	);	837	);
272		838
273	our %method = map +($_ => 1), qw(io timer condvar broadcast wait signal one_event DESTROY);	839	our %method = map +($_ => 1), qw(io timer time now signal child condvar one_event DESTROY);
		840
		841	our @post_detect;
		842
		843	sub post_detect(&) {
		844	my ($cb) = @_;
		845
		846	if ($MODEL) {
		847	$cb->();
		848
		849	1
		850	} else {
		851	push @post_detect, $cb;
		852
		853	defined wantarray
		854	? bless \$cb, "AnyEvent::Util::PostDetect"
		855	: ()
		856	}
		857	}
		858
		859	sub AnyEvent::Util::PostDetect::DESTROY {
		860	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		861	}
274		862
275	sub detect() {	863	sub detect() {
276	unless ($MODEL) {	864	unless ($MODEL) {
277	no strict 'refs';	865	no strict 'refs';
		866	local $SIG{__DIE__};
		867
		868	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
		869	my $model = "AnyEvent::Impl::$1";
		870	if (eval "require $model") {
		871	$MODEL = $model;
		872	warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1;
		873	} else {
		874	warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose;
		875	}
		876	}
278		877
279	# check for already loaded models	878	# check for already loaded models
		879	unless ($MODEL) {
280	for (@REGISTRY, @models) {	880	for (@REGISTRY, @models) {
281	my ($package, $model) = @$_;	881	my ($package, $model) = @$_;
282	if (${"$package\::VERSION"} > 0) {	882	if (${"$package\::VERSION"} > 0) {
283	if (eval "require $model") {	883	if (eval "require $model") {
284	$MODEL = $model;	884	$MODEL = $model;
285	warn "AnyEvent: found model '$model', using it.\n" if $verbose > 1;	885	warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1;
286	last;	886	last;
		887	}
287	}	888	}
288	}	889	}
289	}
290		890
291	unless ($MODEL) {	891	unless ($MODEL) {
292	# try to load a model	892	# try to load a model
293		893
294	for (@REGISTRY, @models) {	894	for (@REGISTRY, @models) {
295	my ($package, $model) = @$_;	895	my ($package, $model) = @$_;
296	if (eval "require $package"	896	if (eval "require $package"
297	and ${"$package\::VERSION"} > 0	897	and ${"$package\::VERSION"} > 0
298	and eval "require $model") {	898	and eval "require $model") {
299	$MODEL = $model;	899	$MODEL = $model;
300	warn "AnyEvent: autoprobed and loaded model '$model', using it.\n" if $verbose > 1;	900	warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1;
301	last;	901	last;
		902	}
302	}	903	}
		904
		905	$MODEL
		906	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.";
303	}	907	}
304
305	$MODEL
306	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: Event (or Coro+Event), Glib or Tk.";
307	}	908	}
308		909
309	unshift @ISA, $MODEL;	910	unshift @ISA, $MODEL;
310	push @{"$MODEL\::ISA"}, "AnyEvent::Base";	911	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		912
		913	(shift @post_detect)->() while @post_detect;
311	}	914	}
312		915
313	$MODEL	916	$MODEL
314	}	917	}
315		918
…		…
325	$class->$func (@_);	928	$class->$func (@_);
326	}	929	}
327		930
328	package AnyEvent::Base;	931	package AnyEvent::Base;
329		932
		933	# default implementation for now and time
		934
		935	use Time::HiRes ();
		936
		937	sub time { Time::HiRes::time }
		938	sub now { Time::HiRes::time }
		939
330	# default implementation for ->condvar, ->wait, ->broadcast	940	# default implementation for ->condvar
331		941
332	sub condvar {	942	sub condvar {
333	bless \my $flag, "AnyEvent::Base::CondVar"	943	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, AnyEvent::CondVar::
334	}
335
336	sub AnyEvent::Base::CondVar::broadcast {
337	${$_[0]}++;
338	}
339
340	sub AnyEvent::Base::CondVar::wait {
341	AnyEvent->one_event while !${$_[0]};
342	}	944	}
343		945
344	# default implementation for ->signal	946	# default implementation for ->signal
345		947
346	our %SIG_CB;	948	our %SIG_CB;
…		…
369		971
370	# default implementation for ->child	972	# default implementation for ->child
371		973
372	our %PID_CB;	974	our %PID_CB;
373	our $CHLD_W;	975	our $CHLD_W;
		976	our $CHLD_DELAY_W;
374	our $PID_IDLE;	977	our $PID_IDLE;
375	our $WNOHANG;	978	our $WNOHANG;
376		979
377	sub _child_wait {	980	sub _child_wait {
378	while (0 < (my $pid = waitpid -1, $WNOHANG)) {	981	while (0 < (my $pid = waitpid -1, $WNOHANG)) {
379	$_->() for values %{ (delete $PID_CB{$pid}) || {} };	982	$_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }),
		983	(values %{ $PID_CB{0} || {} });
380	}	984	}
381		985
382	undef $PID_IDLE;	986	undef $PID_IDLE;
		987	}
		988
		989	sub _sigchld {
		990	# make sure we deliver these changes "synchronous" with the event loop.
		991	$CHLD_DELAY_W ||= AnyEvent->timer (after => 0, cb => sub {
		992	undef $CHLD_DELAY_W;
		993	&_child_wait;
		994	});
383	}	995	}
384		996
385	sub child {	997	sub child {
386	my (undef, %arg) = @_;	998	my (undef, %arg) = @_;
387		999
388	my $pid = uc $arg{pid}	1000	defined (my $pid = $arg{pid} + 0)
389	or Carp::croak "required option 'pid' is missing";	1001	or Carp::croak "required option 'pid' is missing";
390		1002
391	$PID_CB{$pid}{$arg{cb}} = $arg{cb};	1003	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
392		1004
393	unless ($WNOHANG) {	1005	unless ($WNOHANG) {
394	$WNOHANG = eval { require POSIX; &POSIX::WNOHANG } || 1;	1006	$WNOHANG = eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
395	}	1007	}
396		1008
397	unless ($CHLD_W) {	1009	unless ($CHLD_W) {
398	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_child_wait);	1010	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
399	# child could be a zombie already	1011	# child could be a zombie already, so make at least one round
400	$PID_IDLE ||= AnyEvent->timer (after => 0, cb => \&_child_wait);	1012	&_sigchld;
401	}	1013	}
402		1014
403	bless [$pid, $arg{cb}], "AnyEvent::Base::Child"	1015	bless [$pid, $arg{cb}], "AnyEvent::Base::Child"
404	}	1016	}
405		1017
…		…
410	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };	1022	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
411		1023
412	undef $CHLD_W unless keys %PID_CB;	1024	undef $CHLD_W unless keys %PID_CB;
413	}	1025	}
414		1026
		1027	package AnyEvent::CondVar;
		1028
		1029	our @ISA = AnyEvent::CondVar::Base::;
		1030
		1031	package AnyEvent::CondVar::Base;
		1032
		1033	use overload
		1034	'&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1035	fallback => 1;
		1036
		1037	sub _send {
		1038	# nop
		1039	}
		1040
		1041	sub send {
		1042	my $cv = shift;
		1043	$cv->{_ae_sent} = [@_];
		1044	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1045	$cv->_send;
		1046	}
		1047
		1048	sub croak {
		1049	$_[0]{_ae_croak} = $_[1];
		1050	$_[0]->send;
		1051	}
		1052
		1053	sub ready {
		1054	$_[0]{_ae_sent}
		1055	}
		1056
		1057	sub _wait {
		1058	AnyEvent->one_event while !$_[0]{_ae_sent};
		1059	}
		1060
		1061	sub recv {
		1062	$_[0]->_wait;
		1063
		1064	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1065	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1066	}
		1067
		1068	sub cb {
		1069	$_[0]{_ae_cb} = $_[1] if @_ > 1;
		1070	$_[0]{_ae_cb}
		1071	}
		1072
		1073	sub begin {
		1074	++$_[0]{_ae_counter};
		1075	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1076	}
		1077
		1078	sub end {
		1079	return if --$_[0]{_ae_counter};
		1080	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1081	}
		1082
		1083	# undocumented/compatibility with pre-3.4
		1084	*broadcast = \&send;
		1085	*wait = \&_wait;
		1086
415	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE	1087	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1088
		1089	This is an advanced topic that you do not normally need to use AnyEvent in
		1090	a module. This section is only of use to event loop authors who want to
		1091	provide AnyEvent compatibility.
416		1092
417	If you need to support another event library which isn't directly	1093	If you need to support another event library which isn't directly
418	supported by AnyEvent, you can supply your own interface to it by	1094	supported by AnyEvent, you can supply your own interface to it by
419	pushing, before the first watcher gets created, the package name of	1095	pushing, before the first watcher gets created, the package name of
420	the event module and the package name of the interface to use onto	1096	the event module and the package name of the interface to use onto
421	C<@AnyEvent::REGISTRY>. You can do that before and even without loading	1097	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
422	AnyEvent.	1098	AnyEvent, so it is reasonably cheap.
423		1099
424	Example:	1100	Example:
425		1101
426	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];	1102	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
427		1103
428	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>	1104	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
429	package/class when it finds the C<urxvt> package/module is loaded. When	1105	package/class when it finds the C<urxvt> package/module is already loaded.
		1106
430	AnyEvent is loaded and asked to find a suitable event model, it will	1107	When AnyEvent is loaded and asked to find a suitable event model, it
431	first check for the presence of urxvt.	1108	will first check for the presence of urxvt by trying to C<use> the
		1109	C<urxvt::anyevent> module.
432		1110
433	The class should provide implementations for all watcher types (see	1111	The class should provide implementations for all watcher types. See
434	L<AnyEvent::Impl::Event> (source code), L<AnyEvent::Impl::Glib>	1112	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
435	(Source code) and so on for actual examples, use C<perldoc -m	1113	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
436	AnyEvent::Impl::Glib> to see the sources).	1114	see the sources.
437		1115
		1116	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1117	provide suitable (hopefully) replacements.
		1118
438	The above isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)	1119	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
439	uses the above line as-is. An interface isn't included in AnyEvent	1120	terminal emulator uses the above line as-is. An interface isn't included
440	because it doesn't make sense outside the embedded interpreter inside	1121	in AnyEvent because it doesn't make sense outside the embedded interpreter
441	I<rxvt-unicode>, and it is updated and maintained as part of the	1122	inside I<rxvt-unicode>, and it is updated and maintained as part of the
442	I<rxvt-unicode> distribution.	1123	I<rxvt-unicode> distribution.
443		1124
444	I<rxvt-unicode> also cheats a bit by not providing blocking access to	1125	I<rxvt-unicode> also cheats a bit by not providing blocking access to
445	condition variables: code blocking while waiting for a condition will	1126	condition variables: code blocking while waiting for a condition will
446	C<die>. This still works with most modules/usages, and blocking calls must	1127	C<die>. This still works with most modules/usages, and blocking calls must
447	not be in an interactive application, so it makes sense.	1128	not be done in an interactive application, so it makes sense.
448		1129
449	=head1 ENVIRONMENT VARIABLES	1130	=head1 ENVIRONMENT VARIABLES
450		1131
451	The following environment variables are used by this module:	1132	The following environment variables are used by this module:
452		1133
453	C<PERL_ANYEVENT_VERBOSE> when set to C<2> or higher, reports which event	1134	=over 4
454	model gets used.
455		1135
		1136	=item C<PERL_ANYEVENT_VERBOSE>
		1137
		1138	By default, AnyEvent will be completely silent except in fatal
		1139	conditions. You can set this environment variable to make AnyEvent more
		1140	talkative.
		1141
		1142	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1143	conditions, such as not being able to load the event model specified by
		1144	C<PERL_ANYEVENT_MODEL>.
		1145
		1146	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1147	model it chooses.
		1148
		1149	=item C<PERL_ANYEVENT_MODEL>
		1150
		1151	This can be used to specify the event model to be used by AnyEvent, before
		1152	auto detection and -probing kicks in. It must be a string consisting
		1153	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1154	and the resulting module name is loaded and if the load was successful,
		1155	used as event model. If it fails to load AnyEvent will proceed with
		1156	auto detection and -probing.
		1157
		1158	This functionality might change in future versions.
		1159
		1160	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1161	could start your program like this:
		1162
		1163	PERL_ANYEVENT_MODEL=Perl perl ...
		1164
		1165	=item C<PERL_ANYEVENT_PROTOCOLS>
		1166
		1167	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1168	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1169	of auto probing).
		1170
		1171	Must be set to a comma-separated list of protocols or address families,
		1172	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1173	used, and preference will be given to protocols mentioned earlier in the
		1174	list.
		1175
		1176	This variable can effectively be used for denial-of-service attacks
		1177	against local programs (e.g. when setuid), although the impact is likely
		1178	small, as the program has to handle connection errors already-
		1179
		1180	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1181	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1182	- only support IPv4, never try to resolve or contact IPv6
		1183	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1184	IPv6, but prefer IPv6 over IPv4.
		1185
		1186	=item C<PERL_ANYEVENT_EDNS0>
		1187
		1188	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1189	for DNS. This extension is generally useful to reduce DNS traffic, but
		1190	some (broken) firewalls drop such DNS packets, which is why it is off by
		1191	default.
		1192
		1193	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1194	EDNS0 in its DNS requests.
		1195
		1196	=item C<PERL_ANYEVENT_MAX_FORKS>
		1197
		1198	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1199	will create in parallel.
		1200
		1201	=back
		1202
456	=head1 EXAMPLE	1203	=head1 EXAMPLE PROGRAM
457		1204
458	The following program uses an io watcher to read data from stdin, a timer	1205	The following program uses an I/O watcher to read data from STDIN, a timer
459	to display a message once per second, and a condvar to exit the program	1206	to display a message once per second, and a condition variable to quit the
460	when the user enters quit:	1207	program when the user enters quit:
461		1208
462	use AnyEvent;	1209	use AnyEvent;
463		1210
464	my $cv = AnyEvent->condvar;	1211	my $cv = AnyEvent->condvar;
465		1212
466	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	1213	my $io_watcher = AnyEvent->io (
		1214	fh => \*STDIN,
		1215	poll => 'r',
		1216	cb => sub {
467	warn "io event <$_[0]>\n"; # will always output <r>	1217	warn "io event <$_[0]>\n"; # will always output <r>
468	chomp (my $input = <STDIN>); # read a line	1218	chomp (my $input = <STDIN>); # read a line
469	warn "read: $input\n"; # output what has been read	1219	warn "read: $input\n"; # output what has been read
470	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1220	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1221	},
471	});	1222	);
472		1223
473	my $time_watcher; # can only be used once	1224	my $time_watcher; # can only be used once
474		1225
475	sub new_timer {	1226	sub new_timer {
476	$timer = AnyEvent->timer (after => 1, cb => sub {	1227	$timer = AnyEvent->timer (after => 1, cb => sub {
…		…
479	});	1230	});
480	}	1231	}
481		1232
482	new_timer; # create first timer	1233	new_timer; # create first timer
483		1234
484	$cv->wait; # wait until user enters /^q/i	1235	$cv->recv; # wait until user enters /^q/i
485		1236
486	=head1 REAL-WORLD EXAMPLE	1237	=head1 REAL-WORLD EXAMPLE
487		1238
488	Consider the L<Net::FCP> module. It features (among others) the following	1239	Consider the L<Net::FCP> module. It features (among others) the following
489	API calls, which are to freenet what HTTP GET requests are to http:	1240	API calls, which are to freenet what HTTP GET requests are to http:
…		…
539	syswrite $txn->{fh}, $txn->{request}	1290	syswrite $txn->{fh}, $txn->{request}
540	or die "connection or write error";	1291	or die "connection or write error";
541	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	1292	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
542		1293
543	Again, C<fh_ready_r> waits till all data has arrived, and then stores the	1294	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
544	result and signals any possible waiters that the request ahs finished:	1295	result and signals any possible waiters that the request has finished:
545		1296
546	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	1297	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
547		1298
548	if (end-of-file or data complete) {	1299	if (end-of-file or data complete) {
549	$txn->{result} = $txn->{buf};	1300	$txn->{result} = $txn->{buf};
550	$txn->{finished}->broadcast;	1301	$txn->{finished}->send;
551	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	1302	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
552	}	1303	}
553		1304
554	The C<result> method, finally, just waits for the finished signal (if the	1305	The C<result> method, finally, just waits for the finished signal (if the
555	request was already finished, it doesn't wait, of course, and returns the	1306	request was already finished, it doesn't wait, of course, and returns the
556	data:	1307	data:
557		1308
558	$txn->{finished}->wait;	1309	$txn->{finished}->recv;
559	return $txn->{result};	1310	return $txn->{result};
560		1311
561	The actual code goes further and collects all errors (C<die>s, exceptions)	1312	The actual code goes further and collects all errors (C<die>s, exceptions)
562	that occured during request processing. The C<result> method detects	1313	that occurred during request processing. The C<result> method detects
563	wether an exception as thrown (it is stored inside the $txn object)	1314	whether an exception as thrown (it is stored inside the $txn object)
564	and just throws the exception, which means connection errors and other	1315	and just throws the exception, which means connection errors and other
565	problems get reported tot he code that tries to use the result, not in a	1316	problems get reported tot he code that tries to use the result, not in a
566	random callback.	1317	random callback.
567		1318
568	All of this enables the following usage styles:	1319	All of this enables the following usage styles:
569		1320
570	1. Blocking:	1321	1. Blocking:
571		1322
572	my $data = $fcp->client_get ($url);	1323	my $data = $fcp->client_get ($url);
573		1324
574	2. Blocking, but parallelizing:	1325	2. Blocking, but running in parallel:
575		1326
576	my @datas = map $_->result,	1327	my @datas = map $_->result,
577	map $fcp->txn_client_get ($_),	1328	map $fcp->txn_client_get ($_),
578	@urls;	1329	@urls;
579		1330
580	Both blocking examples work without the module user having to know	1331	Both blocking examples work without the module user having to know
581	anything about events.	1332	anything about events.
582		1333
583	3a. Event-based in a main program, using any support Event module:	1334	3a. Event-based in a main program, using any supported event module:
584		1335
585	use Event;	1336	use EV;
586		1337
587	$fcp->txn_client_get ($url)->cb (sub {	1338	$fcp->txn_client_get ($url)->cb (sub {
588	my $txn = shift;	1339	my $txn = shift;
589	my $data = $txn->result;	1340	my $data = $txn->result;
590	...	1341	...
591	});	1342	});
592		1343
593	Event::loop;	1344	EV::loop;
594		1345
595	3b. The module user could use AnyEvent, too:	1346	3b. The module user could use AnyEvent, too:
596		1347
597	use AnyEvent;	1348	use AnyEvent;
598		1349
599	my $quit = AnyEvent->condvar;	1350	my $quit = AnyEvent->condvar;
600		1351
601	$fcp->txn_client_get ($url)->cb (sub {	1352	$fcp->txn_client_get ($url)->cb (sub {
602	...	1353	...
603	$quit->broadcast;	1354	$quit->send;
604	});	1355	});
605		1356
606	$quit->wait;	1357	$quit->recv;
		1358
		1359
		1360	=head1 BENCHMARKS
		1361
		1362	To give you an idea of the performance and overheads that AnyEvent adds
		1363	over the event loops themselves and to give you an impression of the speed
		1364	of various event loops I prepared some benchmarks.
		1365
		1366	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1367
		1368	Here is a benchmark of various supported event models used natively and
		1369	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1370	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1371	which it is), lets them fire exactly once and destroys them again.
		1372
		1373	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		1374	distribution.
		1375
		1376	=head3 Explanation of the columns
		1377
		1378	I<watcher> is the number of event watchers created/destroyed. Since
		1379	different event models feature vastly different performances, each event
		1380	loop was given a number of watchers so that overall runtime is acceptable
		1381	and similar between tested event loop (and keep them from crashing): Glib
		1382	would probably take thousands of years if asked to process the same number
		1383	of watchers as EV in this benchmark.
		1384
		1385	I<bytes> is the number of bytes (as measured by the resident set size,
		1386	RSS) consumed by each watcher. This method of measuring captures both C
		1387	and Perl-based overheads.
		1388
		1389	I<create> is the time, in microseconds (millionths of seconds), that it
		1390	takes to create a single watcher. The callback is a closure shared between
		1391	all watchers, to avoid adding memory overhead. That means closure creation
		1392	and memory usage is not included in the figures.
		1393
		1394	I<invoke> is the time, in microseconds, used to invoke a simple
		1395	callback. The callback simply counts down a Perl variable and after it was
		1396	invoked "watcher" times, it would C<< ->send >> a condvar once to
		1397	signal the end of this phase.
		1398
		1399	I<destroy> is the time, in microseconds, that it takes to destroy a single
		1400	watcher.
		1401
		1402	=head3 Results
		1403
		1404	name watchers bytes create invoke destroy comment
		1405	EV/EV 400000 244 0.56 0.46 0.31 EV native interface
		1406	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
		1407	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
		1408	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
		1409	Event/Event 16000 516 31.88 31.30 0.85 Event native interface
		1410	Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
		1411	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
		1412	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
		1413	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
		1414	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
		1415
		1416	=head3 Discussion
		1417
		1418	The benchmark does I<not> measure scalability of the event loop very
		1419	well. For example, a select-based event loop (such as the pure perl one)
		1420	can never compete with an event loop that uses epoll when the number of
		1421	file descriptors grows high. In this benchmark, all events become ready at
		1422	the same time, so select/poll-based implementations get an unnatural speed
		1423	boost.
		1424
		1425	Also, note that the number of watchers usually has a nonlinear effect on
		1426	overall speed, that is, creating twice as many watchers doesn't take twice
		1427	the time - usually it takes longer. This puts event loops tested with a
		1428	higher number of watchers at a disadvantage.
		1429
		1430	To put the range of results into perspective, consider that on the
		1431	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1432	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1433	cycles with POE.
		1434
		1435	C<EV> is the sole leader regarding speed and memory use, which are both
		1436	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1437	far less memory than any other event loop and is still faster than Event
		1438	natively.
		1439
		1440	The pure perl implementation is hit in a few sweet spots (both the
		1441	constant timeout and the use of a single fd hit optimisations in the perl
		1442	interpreter and the backend itself). Nevertheless this shows that it
		1443	adds very little overhead in itself. Like any select-based backend its
		1444	performance becomes really bad with lots of file descriptors (and few of
		1445	them active), of course, but this was not subject of this benchmark.
		1446
		1447	The C<Event> module has a relatively high setup and callback invocation
		1448	cost, but overall scores in on the third place.
		1449
		1450	C<Glib>'s memory usage is quite a bit higher, but it features a
		1451	faster callback invocation and overall ends up in the same class as
		1452	C<Event>. However, Glib scales extremely badly, doubling the number of
		1453	watchers increases the processing time by more than a factor of four,
		1454	making it completely unusable when using larger numbers of watchers
		1455	(note that only a single file descriptor was used in the benchmark, so
		1456	inefficiencies of C<poll> do not account for this).
		1457
		1458	The C<Tk> adaptor works relatively well. The fact that it crashes with
		1459	more than 2000 watchers is a big setback, however, as correctness takes
		1460	precedence over speed. Nevertheless, its performance is surprising, as the
		1461	file descriptor is dup()ed for each watcher. This shows that the dup()
		1462	employed by some adaptors is not a big performance issue (it does incur a
		1463	hidden memory cost inside the kernel which is not reflected in the figures
		1464	above).
		1465
		1466	C<POE>, regardless of underlying event loop (whether using its pure perl
		1467	select-based backend or the Event module, the POE-EV backend couldn't
		1468	be tested because it wasn't working) shows abysmal performance and
		1469	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		1470	as EV watchers, and 10 times as much memory as Event (the high memory
		1471	requirements are caused by requiring a session for each watcher). Watcher
		1472	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		1473	implementation.
		1474
		1475	The design of the POE adaptor class in AnyEvent can not really account
		1476	for the performance issues, though, as session creation overhead is
		1477	small compared to execution of the state machine, which is coded pretty
		1478	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		1479	using multiple sessions is not a good approach, especially regarding
		1480	memory usage, even the author of POE could not come up with a faster
		1481	design).
		1482
		1483	=head3 Summary
		1484
		1485	=over 4
		1486
		1487	=item * Using EV through AnyEvent is faster than any other event loop
		1488	(even when used without AnyEvent), but most event loops have acceptable
		1489	performance with or without AnyEvent.
		1490
		1491	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		1492	the actual event loop, only with extremely fast event loops such as EV
		1493	adds AnyEvent significant overhead.
		1494
		1495	=item * You should avoid POE like the plague if you want performance or
		1496	reasonable memory usage.
		1497
		1498	=back
		1499
		1500	=head2 BENCHMARKING THE LARGE SERVER CASE
		1501
		1502	This benchmark actually benchmarks the event loop itself. It works by
		1503	creating a number of "servers": each server consists of a socket pair, a
		1504	timeout watcher that gets reset on activity (but never fires), and an I/O
		1505	watcher waiting for input on one side of the socket. Each time the socket
		1506	watcher reads a byte it will write that byte to a random other "server".
		1507
		1508	The effect is that there will be a lot of I/O watchers, only part of which
		1509	are active at any one point (so there is a constant number of active
		1510	fds for each loop iteration, but which fds these are is random). The
		1511	timeout is reset each time something is read because that reflects how
		1512	most timeouts work (and puts extra pressure on the event loops).
		1513
		1514	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		1515	(1%) are active. This mirrors the activity of large servers with many
		1516	connections, most of which are idle at any one point in time.
		1517
		1518	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1519	distribution.
		1520
		1521	=head3 Explanation of the columns
		1522
		1523	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1524	each server has a read and write socket end).
		1525
		1526	I<create> is the time it takes to create a socket pair (which is
		1527	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1528
		1529	I<request>, the most important value, is the time it takes to handle a
		1530	single "request", that is, reading the token from the pipe and forwarding
		1531	it to another server. This includes deleting the old timeout and creating
		1532	a new one that moves the timeout into the future.
		1533
		1534	=head3 Results
		1535
		1536	name sockets create request
		1537	EV 20000 69.01 11.16
		1538	Perl 20000 73.32 35.87
		1539	Event 20000 212.62 257.32
		1540	Glib 20000 651.16 1896.30
		1541	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1542
		1543	=head3 Discussion
		1544
		1545	This benchmark I<does> measure scalability and overall performance of the
		1546	particular event loop.
		1547
		1548	EV is again fastest. Since it is using epoll on my system, the setup time
		1549	is relatively high, though.
		1550
		1551	Perl surprisingly comes second. It is much faster than the C-based event
		1552	loops Event and Glib.
		1553
		1554	Event suffers from high setup time as well (look at its code and you will
		1555	understand why). Callback invocation also has a high overhead compared to
		1556	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1557	uses select or poll in basically all documented configurations.
		1558
		1559	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1560	clearly fails to perform with many filehandles or in busy servers.
		1561
		1562	POE is still completely out of the picture, taking over 1000 times as long
		1563	as EV, and over 100 times as long as the Perl implementation, even though
		1564	it uses a C-based event loop in this case.
		1565
		1566	=head3 Summary
		1567
		1568	=over 4
		1569
		1570	=item * The pure perl implementation performs extremely well.
		1571
		1572	=item * Avoid Glib or POE in large projects where performance matters.
		1573
		1574	=back
		1575
		1576	=head2 BENCHMARKING SMALL SERVERS
		1577
		1578	While event loops should scale (and select-based ones do not...) even to
		1579	large servers, most programs we (or I :) actually write have only a few
		1580	I/O watchers.
		1581
		1582	In this benchmark, I use the same benchmark program as in the large server
		1583	case, but it uses only eight "servers", of which three are active at any
		1584	one time. This should reflect performance for a small server relatively
		1585	well.
		1586
		1587	The columns are identical to the previous table.
		1588
		1589	=head3 Results
		1590
		1591	name sockets create request
		1592	EV 16 20.00 6.54
		1593	Perl 16 25.75 12.62
		1594	Event 16 81.27 35.86
		1595	Glib 16 32.63 15.48
		1596	POE 16 261.87 276.28 uses POE::Loop::Event
		1597
		1598	=head3 Discussion
		1599
		1600	The benchmark tries to test the performance of a typical small
		1601	server. While knowing how various event loops perform is interesting, keep
		1602	in mind that their overhead in this case is usually not as important, due
		1603	to the small absolute number of watchers (that is, you need efficiency and
		1604	speed most when you have lots of watchers, not when you only have a few of
		1605	them).
		1606
		1607	EV is again fastest.
		1608
		1609	Perl again comes second. It is noticeably faster than the C-based event
		1610	loops Event and Glib, although the difference is too small to really
		1611	matter.
		1612
		1613	POE also performs much better in this case, but is is still far behind the
		1614	others.
		1615
		1616	=head3 Summary
		1617
		1618	=over 4
		1619
		1620	=item * C-based event loops perform very well with small number of
		1621	watchers, as the management overhead dominates.
		1622
		1623	=back
		1624
		1625
		1626	=head1 FORK
		1627
		1628	Most event libraries are not fork-safe. The ones who are usually are
		1629	because they rely on inefficient but fork-safe C<select> or C<poll>
		1630	calls. Only L<EV> is fully fork-aware.
		1631
		1632	If you have to fork, you must either do so I<before> creating your first
		1633	watcher OR you must not use AnyEvent at all in the child.
		1634
		1635
		1636	=head1 SECURITY CONSIDERATIONS
		1637
		1638	AnyEvent can be forced to load any event model via
		1639	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		1640	execute arbitrary code or directly gain access, it can easily be used to
		1641	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		1642	will not be active when the program uses a different event model than
		1643	specified in the variable.
		1644
		1645	You can make AnyEvent completely ignore this variable by deleting it
		1646	before the first watcher gets created, e.g. with a C<BEGIN> block:
		1647
		1648	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		1649
		1650	use AnyEvent;
		1651
		1652	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		1653	be used to probe what backend is used and gain other information (which is
		1654	probably even less useful to an attacker than PERL_ANYEVENT_MODEL).
		1655
607		1656
608	=head1 SEE ALSO	1657	=head1 SEE ALSO
609		1658
610	Event modules: L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>.	1659	Utility functions: L<AnyEvent::Util>.
611		1660
612	Implementations: L<AnyEvent::Impl::Coro>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>.	1661	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
		1662	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
613		1663
614	Nontrivial usage example: L<Net::FCP>.	1664	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		1665	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		1666	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		1667	L<AnyEvent::Impl::POE>.
615		1668
616	=head1	1669	Non-blocking file handles, sockets, TCP clients and
		1670	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>.
		1671
		1672	Asynchronous DNS: L<AnyEvent::DNS>.
		1673
		1674	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>,
		1675
		1676	Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>.
		1677
		1678
		1679	=head1 AUTHOR
		1680
		1681	Marc Lehmann <schmorp@schmorp.de>
		1682	http://home.schmorp.de/
617		1683
618	=cut	1684	=cut
619		1685
620	1	1686	1
621		1687

Diff Legend

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