ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.44 by root, Mon Apr 7 19:42:18 2008 UTC vs.
Revision 1.164 by root, Tue Jul 8 19:50:25 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 INTRODUCTION/TUTORIAL
		24
		25	This manpage is mainly a reference manual. If you are interested
		26	in a tutorial or some gentle introduction, have a look at the
		27	L<AnyEvent::Intro> manpage.
22		28
23	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)	29	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
24		30
25	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen	31	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
26	nowadays. So what is different about AnyEvent?	32	nowadays. So what is different about AnyEvent?
…		…
29	policy> and AnyEvent is I<small and efficient>.	35	policy> and AnyEvent is I<small and efficient>.
30		36
31	First and foremost, I<AnyEvent is not an event model> itself, it only	37	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	38	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,	39	pragmatic way. For event models and certain classes of immortals alike,
34	the statement "there can only be one" is a bitter reality, and AnyEvent	40	the statement "there can only be one" is a bitter reality: In general,
35	helps hiding the differences.	41	only one event loop can be active at the same time in a process. AnyEvent
		42	helps hiding the differences between those event loops.
36		43
37	The goal of AnyEvent is to offer module authors the ability to do event	44	The goal of AnyEvent is to offer module authors the ability to do event
38	programming (waiting for I/O or timer events) without subscribing to a	45	programming (waiting for I/O or timer events) without subscribing to a
39	religion, a way of living, and most importantly: without forcing your	46	religion, a way of living, and most importantly: without forcing your
40	module users into the same thing by forcing them to use the same event	47	module users into the same thing by forcing them to use the same event
41	model you use.	48	model you use.
42		49
43	For modules like POE or IO::Async (which is actually doing all I/O	50	For modules like POE or IO::Async (which is a total misnomer as it is
44	I<synchronously>...), using them in your module is like joining a	51	actually doing all I/O I<synchronously>...), using them in your module is
45	cult: After you joined, you are dependent on them and you cannot use	52	like joining a cult: After you joined, you are dependent on them and you
46	anything else, as it is simply incompatible to everything that isn't	53	cannot use anything else, as it is simply incompatible to everything that
47	itself.	54	isn't itself. What's worse, all the potential users of your module are
		55	I<also> forced to use the same event loop you use.
48		56
49	AnyEvent + POE works fine. AnyEvent + Glib works fine. AnyEvent + Tk	57	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
50	works fine etc. etc. but none of these work together with the rest: POE	58	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
51	+ IO::Async? no go. Tk + Event? no go. If your module uses one of	59	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if
52	those, every user of your module has to use it, too. If your module	60	your module uses one of those, every user of your module has to use it,
53	uses AnyEvent, it works transparently with all event models it supports	61	too. But if your module uses AnyEvent, it works transparently with all
54	(including stuff like POE and IO::Async).	62	event models it supports (including stuff like POE and IO::Async, as long
		63	as those use one of the supported event loops. It is trivial to add new
		64	event loops to AnyEvent, too, so it is future-proof).
55		65
56	In addition of being free of having to use I<the one and only true event	66	In addition to being free of having to use I<the one and only true event
57	model>, AnyEvent also is free of bloat and policy: with POE or similar	67	model>, AnyEvent also is free of bloat and policy: with POE or similar
58	modules, you get an enourmous amount of code and strict rules you have	68	modules, you get an enormous amount of code and strict rules you have to
59	to follow. AnyEvent, on the other hand, is lean and to the point by only	69	follow. AnyEvent, on the other hand, is lean and up to the point, by only
60	offering the functionality that is useful, in as thin as a wrapper as	70	offering the functionality that is necessary, in as thin as a wrapper as
61	technically possible.	71	technically possible.
62		72
		73	Of course, AnyEvent comes with a big (and fully optional!) toolbox
		74	of useful functionality, such as an asynchronous DNS resolver, 100%
		75	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		76	such as Windows) and lots of real-world knowledge and workarounds for
		77	platform bugs and differences.
		78
63	Of course, if you want lots of policy (this is arguably somewhat useful	79	Now, if you I<do want> lots of policy (this can arguably be somewhat
64	in many cases) and you want to force your users to the one and only event	80	useful) and you want to force your users to use the one and only event
65	model your module forces on them, you should I<not> use this module.	81	model, you should I<not> use this module.
66
67		82
68	=head1 DESCRIPTION	83	=head1 DESCRIPTION
69		84
70	L<AnyEvent> provides an identical interface to multiple event loops. This	85	L<AnyEvent> provides an identical interface to multiple event loops. This
71	allows module authors to utilise an event loop without forcing module	86	allows module authors to utilise an event loop without forcing module
72	users to use the same event loop (as only a single event loop can coexist	87	users to use the same event loop (as only a single event loop can coexist
73	peacefully at any one time).	88	peacefully at any one time).
74		89
75	The interface itself is vaguely similar but not identical to the Event	90	The interface itself is vaguely similar, but not identical to the L<Event>
76	module.	91	module.
77		92
78	On the first call of any method, the module tries to detect the currently	93	During the first call of any watcher-creation method, the module tries
79	loaded event loop by probing wether any of the following modules is	94	to detect the currently loaded event loop by probing whether one of the
80	loaded: L<Coro::Event>, L<Event>, L<Glib>, L<Tk>. The first one found is	95	following modules is already loaded: L<EV>,
81	used. If none is found, the module tries to load these modules in the	96	L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>,
82	order given. The first one that could be successfully loaded will be	97	L<POE>. The first one found is used. If none are found, the module tries
83	used. If still none could be found, AnyEvent will fall back to a pure-perl	98	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
84	event loop, which is also not very efficient.	99	adaptor should always succeed) in the order given. The first one that can
		100	be successfully loaded will be used. If, after this, still none could be
		101	found, AnyEvent will fall back to a pure-perl event loop, which is not
		102	very efficient, but should work everywhere.
85		103
86	Because AnyEvent first checks for modules that are already loaded, loading	104	Because AnyEvent first checks for modules that are already loaded, loading
87	an Event model explicitly before first using AnyEvent will likely make	105	an event model explicitly before first using AnyEvent will likely make
88	that model the default. For example:	106	that model the default. For example:
89		107
90	use Tk;	108	use Tk;
91	use AnyEvent;	109	use AnyEvent;
92		110
93	# .. AnyEvent will likely default to Tk	111	# .. AnyEvent will likely default to Tk
94		112
		113	The I<likely> means that, if any module loads another event model and
		114	starts using it, all bets are off. Maybe you should tell their authors to
		115	use AnyEvent so their modules work together with others seamlessly...
		116
95	The pure-perl implementation of AnyEvent is called	117	The pure-perl implementation of AnyEvent is called
96	C<AnyEvent::Impl::Perl>. Like other event modules you can load it	118	C<AnyEvent::Impl::Perl>. Like other event modules you can load it
97	explicitly.	119	explicitly and enjoy the high availability of that event loop :)
98		120
99	=head1 WATCHERS	121	=head1 WATCHERS
100		122
101	AnyEvent has the central concept of a I<watcher>, which is an object that	123	AnyEvent has the central concept of a I<watcher>, which is an object that
102	stores relevant data for each kind of event you are waiting for, such as	124	stores relevant data for each kind of event you are waiting for, such as
103	the callback to call, the filehandle to watch, etc.	125	the callback to call, the file handle to watch, etc.
104		126
105	These watchers are normal Perl objects with normal Perl lifetime. After	127	These watchers are normal Perl objects with normal Perl lifetime. After
106	creating a watcher it will immediately "watch" for events and invoke	128	creating a watcher it will immediately "watch" for events and invoke the
		129	callback when the event occurs (of course, only when the event model
		130	is in control).
		131
107	the callback. To disable the watcher you have to destroy it (e.g. by	132	To disable the watcher you have to destroy it (e.g. by setting the
108	setting the variable that stores it to C<undef> or otherwise deleting all	133	variable you store it in to C<undef> or otherwise deleting all references
109	references to it).	134	to it).
110		135
111	All watchers are created by calling a method on the C<AnyEvent> class.	136	All watchers are created by calling a method on the C<AnyEvent> class.
112		137
		138	Many watchers either are used with "recursion" (repeating timers for
		139	example), or need to refer to their watcher object in other ways.
		140
		141	An any way to achieve that is this pattern:
		142
		143	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
		144	# you can use $w here, for example to undef it
		145	undef $w;
		146	});
		147
		148	Note that C<my $w; $w => combination. This is necessary because in Perl,
		149	my variables are only visible after the statement in which they are
		150	declared.
		151
113	=head2 IO WATCHERS	152	=head2 I/O WATCHERS
114		153
115	You can create I/O watcher by calling the C<< AnyEvent->io >> method with	154	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
116	the following mandatory arguments:	155	with the following mandatory key-value pairs as arguments:
117		156
118	C<fh> the Perl I<filehandle> (not filedescriptor) to watch for	157	C<fh> the Perl I<file handle> (I<not> file descriptor) to watch
119	events. C<poll> must be a string that is either C<r> or C<w>, that creates	158	for events. C<poll> must be a string that is either C<r> or C<w>,
120	a watcher waiting for "r"eadable or "w"ritable events. C<cb> the callback	159	which creates a watcher waiting for "r"eadable or "w"ritable events,
121	to invoke everytime the filehandle becomes ready.	160	respectively. C<cb> is the callback to invoke each time the file handle
		161	becomes ready.
122		162
123	Only one io watcher per C<fh> and C<poll> combination is allowed (i.e. on	163	Although the callback might get passed parameters, their value and
124	a socket you can have one r + one w, not any more (limitation comes from	164	presence is undefined and you cannot rely on them. Portable AnyEvent
125	Tk - if you are sure you are not using Tk this limitation is gone).	165	callbacks cannot use arguments passed to I/O watcher callbacks.
126		166
127	Filehandles will be kept alive, so as long as the watcher exists, the	167	The I/O watcher might use the underlying file descriptor or a copy of it.
128	filehandle exists, too.	168	You must not close a file handle as long as any watcher is active on the
		169	underlying file descriptor.
129		170
130	Example:	171	Some event loops issue spurious readyness notifications, so you should
		172	always use non-blocking calls when reading/writing from/to your file
		173	handles.
131		174
132	# wait for readability of STDIN, then read a line and disable the watcher	175	Example: wait for readability of STDIN, then read a line and disable the
		176	watcher.
		177
133	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	178	my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
134	chomp (my $input = <STDIN>);	179	chomp (my $input = <STDIN>);
135	warn "read: $input\n";	180	warn "read: $input\n";
136	undef $w;	181	undef $w;
137	});	182	});
…		…
139	=head2 TIME WATCHERS	184	=head2 TIME WATCHERS
140		185
141	You can create a time watcher by calling the C<< AnyEvent->timer >>	186	You can create a time watcher by calling the C<< AnyEvent->timer >>
142	method with the following mandatory arguments:	187	method with the following mandatory arguments:
143		188
144	C<after> after how many seconds (fractions are supported) should the timer	189	C<after> specifies after how many seconds (fractional values are
145	activate. C<cb> the callback to invoke.	190	supported) the callback should be invoked. C<cb> is the callback to invoke
		191	in that case.
146		192
147	The timer callback will be invoked at most once: if you want a repeating	193	Although the callback might get passed parameters, their value and
148	timer you have to create a new watcher (this is a limitation by both Tk	194	presence is undefined and you cannot rely on them. Portable AnyEvent
149	and Glib).	195	callbacks cannot use arguments passed to time watcher callbacks.
150		196
151	Example:	197	The callback will normally be invoked once only. If you specify another
		198	parameter, C<interval>, as a positive number, then the callback will be
		199	invoked regularly at that interval (in fractional seconds) after the first
		200	invocation.
152		201
		202	The callback will be rescheduled before invoking the callback, but no
		203	attempt is done to avoid timer drift in most backends, so the interval is
		204	only approximate.
		205
153	# fire an event after 7.7 seconds	206	Example: fire an event after 7.7 seconds.
		207
154	my $w = AnyEvent->timer (after => 7.7, cb => sub {	208	my $w = AnyEvent->timer (after => 7.7, cb => sub {
155	warn "timeout\n";	209	warn "timeout\n";
156	});	210	});
157		211
158	# to cancel the timer:	212	# to cancel the timer:
159	undef $w;	213	undef $w;
160		214
161	=head2 CONDITION WATCHERS	215	Example 2: fire an event after 0.5 seconds, then roughly every second.
162		216
163	Condition watchers can be created by calling the C<< AnyEvent->condvar >>	217	my $w = AnyEvent->timer (after => 0.5, interval => 1, cb => sub {
164	method without any arguments.	218	warn "timeout\n";
		219	};
165		220
166	A condition watcher watches for a condition - precisely that the C<<	221	=head3 TIMING ISSUES
167	->broadcast >> method has been called.
168		222
169	Note that condition watchers recurse into the event loop - if you have	223	There are two ways to handle timers: based on real time (relative, "fire
170	two watchers that call C<< ->wait >> in a round-robbin fashion, you	224	in 10 seconds") and based on wallclock time (absolute, "fire at 12
171	lose. Therefore, condition watchers are good to export to your caller, but	225	o'clock").
172	you should avoid making a blocking wait, at least in callbacks, as this
173	usually asks for trouble.
174		226
175	The watcher has only two methods:	227	While most event loops expect timers to specified in a relative way, they
		228	use absolute time internally. This makes a difference when your clock
		229	"jumps", for example, when ntp decides to set your clock backwards from
		230	the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to
		231	fire "after" a second might actually take six years to finally fire.
		232
		233	AnyEvent cannot compensate for this. The only event loop that is conscious
		234	about these issues is L<EV>, which offers both relative (ev_timer, based
		235	on true relative time) and absolute (ev_periodic, based on wallclock time)
		236	timers.
		237
		238	AnyEvent always prefers relative timers, if available, matching the
		239	AnyEvent API.
		240
		241	AnyEvent has two additional methods that return the "current time":
176		242
177	=over 4	243	=over 4
178		244
179	=item $cv->wait	245	=item AnyEvent->time
180		246
181	Wait (blocking if necessary) until the C<< ->broadcast >> method has been	247	This returns the "current wallclock time" as a fractional number of
182	called on c<$cv>, while servicing other watchers normally.	248	seconds since the Epoch (the same thing as C<time> or C<Time::HiRes::time>
		249	return, and the result is guaranteed to be compatible with those).
183		250
184	Not all event models support a blocking wait - some die in that case, so	251	It progresses independently of any event loop processing, i.e. each call
185	if you are using this from a module, never require a blocking wait, but	252	will check the system clock, which usually gets updated frequently.
186	let the caller decide wether the call will block or not (for example,
187	by coupling condition variables with some kind of request results and
188	supporting callbacks so the caller knows that getting the result will not
189	block, while still suppporting blockign waits if the caller so desires).
190		253
191	You can only wait once on a condition - additional calls will return	254	=item AnyEvent->now
192	immediately.
193		255
194	=item $cv->broadcast	256	This also returns the "current wallclock time", but unlike C<time>, above,
		257	this value might change only once per event loop iteration, depending on
		258	the event loop (most return the same time as C<time>, above). This is the
		259	time that AnyEvent's timers get scheduled against.
195		260
196	Flag the condition as ready - a running C<< ->wait >> and all further	261	I<In almost all cases (in all cases if you don't care), this is the
197	calls to C<wait> will return after this method has been called. If nobody	262	function to call when you want to know the current time.>
198	is waiting the broadcast will be remembered..
199		263
200	Example:	264	This function is also often faster then C<< AnyEvent->time >>, and
		265	thus the preferred method if you want some timestamp (for example,
		266	L<AnyEvent::Handle> uses this to update it's activity timeouts).
		267
		268	The rest of this section is only of relevance if you try to be very exact
		269	with your timing, you can skip it without bad conscience.
		270
		271	For a practical example of when these times differ, consider L<Event::Lib>
		272	and L<EV> and the following set-up:
		273
		274	The event loop is running and has just invoked one of your callback at
		275	time=500 (assume no other callbacks delay processing). In your callback,
		276	you wait a second by executing C<sleep 1> (blocking the process for a
		277	second) and then (at time=501) you create a relative timer that fires
		278	after three seconds.
		279
		280	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		281	both return C<501>, because that is the current time, and the timer will
		282	be scheduled to fire at time=504 (C<501> + C<3>).
		283
		284	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		285	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		286	last event processing phase started. With L<EV>, your timer gets scheduled
		287	to run at time=503 (C<500> + C<3>).
		288
		289	In one sense, L<Event::Lib> is more exact, as it uses the current time
		290	regardless of any delays introduced by event processing. However, most
		291	callbacks do not expect large delays in processing, so this causes a
		292	higher drift (and a lot more system calls to get the current time).
		293
		294	In another sense, L<EV> is more exact, as your timer will be scheduled at
		295	the same time, regardless of how long event processing actually took.
		296
		297	In either case, if you care (and in most cases, you don't), then you
		298	can get whatever behaviour you want with any event loop, by taking the
		299	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		300	account.
		301
		302	=back
		303
		304	=head2 SIGNAL WATCHERS
		305
		306	You can watch for signals using a signal watcher, C<signal> is the signal
		307	I<name> without any C<SIG> prefix, C<cb> is the Perl callback to
		308	be invoked whenever a signal occurs.
		309
		310	Although the callback might get passed parameters, their value and
		311	presence is undefined and you cannot rely on them. Portable AnyEvent
		312	callbacks cannot use arguments passed to signal watcher callbacks.
		313
		314	Multiple signal occurrences can be clumped together into one callback
		315	invocation, and callback invocation will be synchronous. Synchronous means
		316	that it might take a while until the signal gets handled by the process,
		317	but it is guaranteed not to interrupt any other callbacks.
		318
		319	The main advantage of using these watchers is that you can share a signal
		320	between multiple watchers.
		321
		322	This watcher might use C<%SIG>, so programs overwriting those signals
		323	directly will likely not work correctly.
		324
		325	Example: exit on SIGINT
		326
		327	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
		328
		329	=head2 CHILD PROCESS WATCHERS
		330
		331	You can also watch on a child process exit and catch its exit status.
		332
		333	The child process is specified by the C<pid> argument (if set to C<0>, it
		334	watches for any child process exit). The watcher will trigger as often
		335	as status change for the child are received. This works by installing a
		336	signal handler for C<SIGCHLD>. The callback will be called with the pid
		337	and exit status (as returned by waitpid), so unlike other watcher types,
		338	you I<can> rely on child watcher callback arguments.
		339
		340	There is a slight catch to child watchers, however: you usually start them
		341	I<after> the child process was created, and this means the process could
		342	have exited already (and no SIGCHLD will be sent anymore).
		343
		344	Not all event models handle this correctly (POE doesn't), but even for
		345	event models that I<do> handle this correctly, they usually need to be
		346	loaded before the process exits (i.e. before you fork in the first place).
		347
		348	This means you cannot create a child watcher as the very first thing in an
		349	AnyEvent program, you I<have> to create at least one watcher before you
		350	C<fork> the child (alternatively, you can call C<AnyEvent::detect>).
		351
		352	Example: fork a process and wait for it
		353
		354	my $done = AnyEvent->condvar;
		355
		356	my $pid = fork or exit 5;
		357
		358	my $w = AnyEvent->child (
		359	pid => $pid,
		360	cb => sub {
		361	my ($pid, $status) = @_;
		362	warn "pid $pid exited with status $status";
		363	$done->send;
		364	},
		365	);
		366
		367	# do something else, then wait for process exit
		368	$done->recv;
		369
		370	=head2 CONDITION VARIABLES
		371
		372	If you are familiar with some event loops you will know that all of them
		373	require you to run some blocking "loop", "run" or similar function that
		374	will actively watch for new events and call your callbacks.
		375
		376	AnyEvent is different, it expects somebody else to run the event loop and
		377	will only block when necessary (usually when told by the user).
		378
		379	The instrument to do that is called a "condition variable", so called
		380	because they represent a condition that must become true.
		381
		382	Condition variables can be created by calling the C<< AnyEvent->condvar
		383	>> method, usually without arguments. The only argument pair allowed is
		384	C<cb>, which specifies a callback to be called when the condition variable
		385	becomes true.
		386
		387	After creation, the condition variable is "false" until it becomes "true"
		388	by calling the C<send> method (or calling the condition variable as if it
		389	were a callback, read about the caveats in the description for the C<<
		390	->send >> method).
		391
		392	Condition variables are similar to callbacks, except that you can
		393	optionally wait for them. They can also be called merge points - points
		394	in time where multiple outstanding events have been processed. And yet
		395	another way to call them is transactions - each condition variable can be
		396	used to represent a transaction, which finishes at some point and delivers
		397	a result.
		398
		399	Condition variables are very useful to signal that something has finished,
		400	for example, if you write a module that does asynchronous http requests,
		401	then a condition variable would be the ideal candidate to signal the
		402	availability of results. The user can either act when the callback is
		403	called or can synchronously C<< ->recv >> for the results.
		404
		405	You can also use them to simulate traditional event loops - for example,
		406	you can block your main program until an event occurs - for example, you
		407	could C<< ->recv >> in your main program until the user clicks the Quit
		408	button of your app, which would C<< ->send >> the "quit" event.
		409
		410	Note that condition variables recurse into the event loop - if you have
		411	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
		412	lose. Therefore, condition variables are good to export to your caller, but
		413	you should avoid making a blocking wait yourself, at least in callbacks,
		414	as this asks for trouble.
		415
		416	Condition variables are represented by hash refs in perl, and the keys
		417	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		418	easy (it is often useful to build your own transaction class on top of
		419	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		420	it's C<new> method in your own C<new> method.
		421
		422	There are two "sides" to a condition variable - the "producer side" which
		423	eventually calls C<< -> send >>, and the "consumer side", which waits
		424	for the send to occur.
		425
		426	Example: wait for a timer.
201		427
202	# wait till the result is ready	428	# wait till the result is ready
203	my $result_ready = AnyEvent->condvar;	429	my $result_ready = AnyEvent->condvar;
204		430
205	# do something such as adding a timer	431	# do something such as adding a timer
206	# or socket watcher the calls $result_ready->broadcast	432	# or socket watcher the calls $result_ready->send
207	# when the "result" is ready.	433	# when the "result" is ready.
		434	# in this case, we simply use a timer:
		435	my $w = AnyEvent->timer (
		436	after => 1,
		437	cb => sub { $result_ready->send },
		438	);
208		439
		440	# this "blocks" (while handling events) till the callback
		441	# calls send
209	$result_ready->wait;	442	$result_ready->recv;
		443
		444	Example: wait for a timer, but take advantage of the fact that
		445	condition variables are also code references.
		446
		447	my $done = AnyEvent->condvar;
		448	my $delay = AnyEvent->timer (after => 5, cb => $done);
		449	$done->recv;
		450
		451	=head3 METHODS FOR PRODUCERS
		452
		453	These methods should only be used by the producing side, i.e. the
		454	code/module that eventually sends the signal. Note that it is also
		455	the producer side which creates the condvar in most cases, but it isn't
		456	uncommon for the consumer to create it as well.
		457
		458	=over 4
		459
		460	=item $cv->send (...)
		461
		462	Flag the condition as ready - a running C<< ->recv >> and all further
		463	calls to C<recv> will (eventually) return after this method has been
		464	called. If nobody is waiting the send will be remembered.
		465
		466	If a callback has been set on the condition variable, it is called
		467	immediately from within send.
		468
		469	Any arguments passed to the C<send> call will be returned by all
		470	future C<< ->recv >> calls.
		471
		472	Condition variables are overloaded so one can call them directly
		473	(as a code reference). Calling them directly is the same as calling
		474	C<send>. Note, however, that many C-based event loops do not handle
		475	overloading, so as tempting as it may be, passing a condition variable
		476	instead of a callback does not work. Both the pure perl and EV loops
		477	support overloading, however, as well as all functions that use perl to
		478	invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
		479	example).
		480
		481	=item $cv->croak ($error)
		482
		483	Similar to send, but causes all call's to C<< ->recv >> to invoke
		484	C<Carp::croak> with the given error message/object/scalar.
		485
		486	This can be used to signal any errors to the condition variable
		487	user/consumer.
		488
		489	=item $cv->begin ([group callback])
		490
		491	=item $cv->end
		492
		493	These two methods are EXPERIMENTAL and MIGHT CHANGE.
		494
		495	These two methods can be used to combine many transactions/events into
		496	one. For example, a function that pings many hosts in parallel might want
		497	to use a condition variable for the whole process.
		498
		499	Every call to C<< ->begin >> will increment a counter, and every call to
		500	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		501	>>, the (last) callback passed to C<begin> will be executed. That callback
		502	is I<supposed> to call C<< ->send >>, but that is not required. If no
		503	callback was set, C<send> will be called without any arguments.
		504
		505	Let's clarify this with the ping example:
		506
		507	my $cv = AnyEvent->condvar;
		508
		509	my %result;
		510	$cv->begin (sub { $cv->send (\%result) });
		511
		512	for my $host (@list_of_hosts) {
		513	$cv->begin;
		514	ping_host_then_call_callback $host, sub {
		515	$result{$host} = ...;
		516	$cv->end;
		517	};
		518	}
		519
		520	$cv->end;
		521
		522	This code fragment supposedly pings a number of hosts and calls
		523	C<send> after results for all then have have been gathered - in any
		524	order. To achieve this, the code issues a call to C<begin> when it starts
		525	each ping request and calls C<end> when it has received some result for
		526	it. Since C<begin> and C<end> only maintain a counter, the order in which
		527	results arrive is not relevant.
		528
		529	There is an additional bracketing call to C<begin> and C<end> outside the
		530	loop, which serves two important purposes: first, it sets the callback
		531	to be called once the counter reaches C<0>, and second, it ensures that
		532	C<send> is called even when C<no> hosts are being pinged (the loop
		533	doesn't execute once).
		534
		535	This is the general pattern when you "fan out" into multiple subrequests:
		536	use an outer C<begin>/C<end> pair to set the callback and ensure C<end>
		537	is called at least once, and then, for each subrequest you start, call
		538	C<begin> and for each subrequest you finish, call C<end>.
210		539
211	=back	540	=back
212		541
213	=head2 SIGNAL WATCHERS	542	=head3 METHODS FOR CONSUMERS
214		543
215	You can listen for signals using a signal watcher, C<signal> is the signal	544	These methods should only be used by the consuming side, i.e. the
216	I<name> without any C<SIG> prefix. Multiple signals events can be clumped	545	code awaits the condition.
217	together into one callback invocation, and callback invocation might or
218	might not be asynchronous.
219		546
220	These watchers might use C<%SIG>, so programs overwriting those signals	547	=over 4
221	directly will likely not work correctly.
222		548
223	Example: exit on SIGINT	549	=item $cv->recv
224		550
225	my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });	551	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
		552	>> methods have been called on c<$cv>, while servicing other watchers
		553	normally.
226		554
227	=head2 CHILD PROCESS WATCHERS	555	You can only wait once on a condition - additional calls are valid but
		556	will return immediately.
228		557
229	You can also listen for the status of a child process specified by the	558	If an error condition has been set by calling C<< ->croak >>, then this
230	C<pid> argument (or any child if the pid argument is 0). The watcher will	559	function will call C<croak>.
231	trigger as often as status change for the child are received. This works
232	by installing a signal handler for C<SIGCHLD>. The callback will be called with
233	the pid and exit status (as returned by waitpid).
234		560
235	Example: wait for pid 1333	561	In list context, all parameters passed to C<send> will be returned,
		562	in scalar context only the first one will be returned.
236		563
237	my $w = AnyEvent->child (pid => 1333, cb => sub { warn "exit status $?" });	564	Not all event models support a blocking wait - some die in that case
		565	(programs might want to do that to stay interactive), so I<if you are
		566	using this from a module, never require a blocking wait>, but let the
		567	caller decide whether the call will block or not (for example, by coupling
		568	condition variables with some kind of request results and supporting
		569	callbacks so the caller knows that getting the result will not block,
		570	while still supporting blocking waits if the caller so desires).
238		571
239	=head1 GLOBALS	572	Another reason I<never> to C<< ->recv >> in a module is that you cannot
		573	sensibly have two C<< ->recv >>'s in parallel, as that would require
		574	multiple interpreters or coroutines/threads, none of which C<AnyEvent>
		575	can supply.
		576
		577	The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
		578	fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
		579	versions and also integrates coroutines into AnyEvent, making blocking
		580	C<< ->recv >> calls perfectly safe as long as they are done from another
		581	coroutine (one that doesn't run the event loop).
		582
		583	You can ensure that C<< -recv >> never blocks by setting a callback and
		584	only calling C<< ->recv >> from within that callback (or at a later
		585	time). This will work even when the event loop does not support blocking
		586	waits otherwise.
		587
		588	=item $bool = $cv->ready
		589
		590	Returns true when the condition is "true", i.e. whether C<send> or
		591	C<croak> have been called.
		592
		593	=item $cb = $cv->cb ([new callback])
		594
		595	This is a mutator function that returns the callback set and optionally
		596	replaces it before doing so.
		597
		598	The callback will be called when the condition becomes "true", i.e. when
		599	C<send> or C<croak> are called, with the only argument being the condition
		600	variable itself. Calling C<recv> inside the callback or at any later time
		601	is guaranteed not to block.
		602
		603	=back
		604
		605	=head1 GLOBAL VARIABLES AND FUNCTIONS
240		606
241	=over 4	607	=over 4
242		608
243	=item $AnyEvent::MODEL	609	=item $AnyEvent::MODEL
244		610
…		…
248	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case	614	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case
249	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).	615	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).
250		616
251	The known classes so far are:	617	The known classes so far are:
252		618
253	AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
254	AnyEvent::Impl::EV based on EV (an interface to libev, also best choice).	619	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
255	AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
256	AnyEvent::Impl::Event based on Event, also second best choice :)	620	AnyEvent::Impl::Event based on Event, second best choice.
		621	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
257	AnyEvent::Impl::Glib based on Glib, second-best choice.	622	AnyEvent::Impl::Glib based on Glib, third-best choice.
258	AnyEvent::Impl::Tk based on Tk, very bad choice.	623	AnyEvent::Impl::Tk based on Tk, very bad choice.
259	AnyEvent::Impl::Perl pure-perl implementation, inefficient.	624	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
		625	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		626	AnyEvent::Impl::POE based on POE, not generic enough for full support.
		627
		628	There is no support for WxWidgets, as WxWidgets has no support for
		629	watching file handles. However, you can use WxWidgets through the
		630	POE Adaptor, as POE has a Wx backend that simply polls 20 times per
		631	second, which was considered to be too horrible to even consider for
		632	AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
		633	it's adaptor.
		634
		635	AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
		636	autodetecting them.
260		637
261	=item AnyEvent::detect	638	=item AnyEvent::detect
262		639
263	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model if	640	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
264	necessary. You should only call this function right before you would have	641	if necessary. You should only call this function right before you would
265	created an AnyEvent watcher anyway, that is, very late at runtime.	642	have created an AnyEvent watcher anyway, that is, as late as possible at
		643	runtime.
		644
		645	=item $guard = AnyEvent::post_detect { BLOCK }
		646
		647	Arranges for the code block to be executed as soon as the event model is
		648	autodetected (or immediately if this has already happened).
		649
		650	If called in scalar or list context, then it creates and returns an object
		651	that automatically removes the callback again when it is destroyed. See
		652	L<Coro::BDB> for a case where this is useful.
		653
		654	=item @AnyEvent::post_detect
		655
		656	If there are any code references in this array (you can C<push> to it
		657	before or after loading AnyEvent), then they will called directly after
		658	the event loop has been chosen.
		659
		660	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		661	if it contains a true value then the event loop has already been detected,
		662	and the array will be ignored.
		663
		664	Best use C<AnyEvent::post_detect { BLOCK }> instead.
266		665
267	=back	666	=back
268		667
269	=head1 WHAT TO DO IN A MODULE	668	=head1 WHAT TO DO IN A MODULE
270		669
271	As a module author, you should "use AnyEvent" and call AnyEvent methods	670	As a module author, you should C<use AnyEvent> and call AnyEvent methods
272	freely, but you should not load a specific event module or rely on it.	671	freely, but you should not load a specific event module or rely on it.
273		672
274	Be careful when you create watchers in the module body - Anyevent will	673	Be careful when you create watchers in the module body - AnyEvent will
275	decide which event module to use as soon as the first method is called, so	674	decide which event module to use as soon as the first method is called, so
276	by calling AnyEvent in your module body you force the user of your module	675	by calling AnyEvent in your module body you force the user of your module
277	to load the event module first.	676	to load the event module first.
278		677
		678	Never call C<< ->recv >> on a condition variable unless you I<know> that
		679	the C<< ->send >> method has been called on it already. This is
		680	because it will stall the whole program, and the whole point of using
		681	events is to stay interactive.
		682
		683	It is fine, however, to call C<< ->recv >> when the user of your module
		684	requests it (i.e. if you create a http request object ad have a method
		685	called C<results> that returns the results, it should call C<< ->recv >>
		686	freely, as the user of your module knows what she is doing. always).
		687
279	=head1 WHAT TO DO IN THE MAIN PROGRAM	688	=head1 WHAT TO DO IN THE MAIN PROGRAM
280		689
281	There will always be a single main program - the only place that should	690	There will always be a single main program - the only place that should
282	dictate which event model to use.	691	dictate which event model to use.
283		692
284	If it doesn't care, it can just "use AnyEvent" and use it itself, or not	693	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
285	do anything special and let AnyEvent decide which implementation to chose.	694	do anything special (it does not need to be event-based) and let AnyEvent
		695	decide which implementation to chose if some module relies on it.
286		696
287	If the main program relies on a specific event model (for example, in Gtk2	697	If the main program relies on a specific event model - for example, in
288	programs you have to rely on either Glib or Glib::Event), you should load	698	Gtk2 programs you have to rely on the Glib module - you should load the
289	it before loading AnyEvent or any module that uses it, generally, as early	699	event module before loading AnyEvent or any module that uses it: generally
290	as possible. The reason is that modules might create watchers when they	700	speaking, you should load it as early as possible. The reason is that
291	are loaded, and AnyEvent will decide on the event model to use as soon as	701	modules might create watchers when they are loaded, and AnyEvent will
292	it creates watchers, and it might chose the wrong one unless you load the	702	decide on the event model to use as soon as it creates watchers, and it
293	correct one yourself.	703	might chose the wrong one unless you load the correct one yourself.
294		704
295	You can chose to use a rather inefficient pure-perl implementation by	705	You can chose to use a pure-perl implementation by loading the
296	loading the C<AnyEvent::Impl::Perl> module, but letting AnyEvent chose is	706	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
297	generally better.	707	everywhere, but letting AnyEvent chose the model is generally better.
		708
		709	=head2 MAINLOOP EMULATION
		710
		711	Sometimes (often for short test scripts, or even standalone programs who
		712	only want to use AnyEvent), you do not want to run a specific event loop.
		713
		714	In that case, you can use a condition variable like this:
		715
		716	AnyEvent->condvar->recv;
		717
		718	This has the effect of entering the event loop and looping forever.
		719
		720	Note that usually your program has some exit condition, in which case
		721	it is better to use the "traditional" approach of storing a condition
		722	variable somewhere, waiting for it, and sending it when the program should
		723	exit cleanly.
		724
		725
		726	=head1 OTHER MODULES
		727
		728	The following is a non-exhaustive list of additional modules that use
		729	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		730	in the same program. Some of the modules come with AnyEvent, some are
		731	available via CPAN.
		732
		733	=over 4
		734
		735	=item L<AnyEvent::Util>
		736
		737	Contains various utility functions that replace often-used but blocking
		738	functions such as C<inet_aton> by event-/callback-based versions.
		739
		740	=item L<AnyEvent::Socket>
		741
		742	Provides various utility functions for (internet protocol) sockets,
		743	addresses and name resolution. Also functions to create non-blocking tcp
		744	connections or tcp servers, with IPv6 and SRV record support and more.
		745
		746	=item L<AnyEvent::Handle>
		747
		748	Provide read and write buffers, manages watchers for reads and writes,
		749	supports raw and formatted I/O, I/O queued and fully transparent and
		750	non-blocking SSL/TLS.
		751
		752	=item L<AnyEvent::DNS>
		753
		754	Provides rich asynchronous DNS resolver capabilities.
		755
		756	=item L<AnyEvent::HTTP>
		757
		758	A simple-to-use HTTP library that is capable of making a lot of concurrent
		759	HTTP requests.
		760
		761	=item L<AnyEvent::HTTPD>
		762
		763	Provides a simple web application server framework.
		764
		765	=item L<AnyEvent::FastPing>
		766
		767	The fastest ping in the west.
		768
		769	=item L<AnyEvent::DBI>
		770
		771	Executes L<DBI> requests asynchronously in a proxy process.
		772
		773	=item L<AnyEvent::AIO>
		774
		775	Truly asynchronous I/O, should be in the toolbox of every event
		776	programmer. AnyEvent::AIO transparently fuses L<IO::AIO> and AnyEvent
		777	together.
		778
		779	=item L<AnyEvent::BDB>
		780
		781	Truly asynchronous Berkeley DB access. AnyEvent::BDB transparently fuses
		782	L<BDB> and AnyEvent together.
		783
		784	=item L<AnyEvent::GPSD>
		785
		786	A non-blocking interface to gpsd, a daemon delivering GPS information.
		787
		788	=item L<AnyEvent::IGS>
		789
		790	A non-blocking interface to the Internet Go Server protocol (used by
		791	L<App::IGS>).
		792
		793	=item L<Net::IRC3>
		794
		795	AnyEvent based IRC client module family.
		796
		797	=item L<Net::XMPP2>
		798
		799	AnyEvent based XMPP (Jabber protocol) module family.
		800
		801	=item L<Net::FCP>
		802
		803	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		804	of AnyEvent.
		805
		806	=item L<Event::ExecFlow>
		807
		808	High level API for event-based execution flow control.
		809
		810	=item L<Coro>
		811
		812	Has special support for AnyEvent via L<Coro::AnyEvent>.
		813
		814	=item L<IO::Lambda>
		815
		816	The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
		817
		818	=back
298		819
299	=cut	820	=cut
300		821
301	package AnyEvent;	822	package AnyEvent;
302		823
303	no warnings;	824	no warnings;
304	use strict;	825	use strict;
305		826
306	use Carp;	827	use Carp;
307		828
308	our $VERSION = '3.0';	829	our $VERSION = 4.2;
309	our $MODEL;	830	our $MODEL;
310		831
311	our $AUTOLOAD;	832	our $AUTOLOAD;
312	our @ISA;	833	our @ISA;
313		834
		835	our @REGISTRY;
		836
		837	our $WIN32;
		838
		839	BEGIN {
		840	my $win32 = ! ! ($^O =~ /mswin32/i);
		841	eval "sub WIN32(){ $win32 }";
		842	}
		843
314	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;	844	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
315		845
316	our @REGISTRY;	846	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		847
		848	{
		849	my $idx;
		850	$PROTOCOL{$_} = ++$idx
		851	for reverse split /\s*,\s*/,
		852	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		853	}
317		854
318	my @models = (	855	my @models = (
319	[Coro::EV:: => AnyEvent::Impl::CoroEV::],
320	[EV:: => AnyEvent::Impl::EV::],	856	[EV:: => AnyEvent::Impl::EV::],
321	[Coro::Event:: => AnyEvent::Impl::CoroEvent::],
322	[Event:: => AnyEvent::Impl::Event::],	857	[Event:: => AnyEvent::Impl::Event::],
323	[Glib:: => AnyEvent::Impl::Glib::],
324	[Tk:: => AnyEvent::Impl::Tk::],
325	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],	858	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
		859	# everything below here will not be autoprobed
		860	# as the pureperl backend should work everywhere
		861	# and is usually faster
		862	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		863	[Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
		864	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		865	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		866	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		867	[Wx:: => AnyEvent::Impl::POE::],
		868	[Prima:: => AnyEvent::Impl::POE::],
326	);	869	);
327		870
328	our %method = map +($_ => 1), qw(io timer condvar broadcast wait signal one_event DESTROY);	871	our %method = map +($_ => 1), qw(io timer time now signal child condvar one_event DESTROY);
		872
		873	our @post_detect;
		874
		875	sub post_detect(&) {
		876	my ($cb) = @_;
		877
		878	if ($MODEL) {
		879	$cb->();
		880
		881	1
		882	} else {
		883	push @post_detect, $cb;
		884
		885	defined wantarray
		886	? bless \$cb, "AnyEvent::Util::PostDetect"
		887	: ()
		888	}
		889	}
		890
		891	sub AnyEvent::Util::PostDetect::DESTROY {
		892	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		893	}
329		894
330	sub detect() {	895	sub detect() {
331	unless ($MODEL) {	896	unless ($MODEL) {
332	no strict 'refs';	897	no strict 'refs';
		898	local $SIG{__DIE__};
		899
		900	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
		901	my $model = "AnyEvent::Impl::$1";
		902	if (eval "require $model") {
		903	$MODEL = $model;
		904	warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1;
		905	} else {
		906	warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose;
		907	}
		908	}
333		909
334	# check for already loaded models	910	# check for already loaded models
		911	unless ($MODEL) {
335	for (@REGISTRY, @models) {	912	for (@REGISTRY, @models) {
336	my ($package, $model) = @$_;	913	my ($package, $model) = @$_;
337	if (${"$package\::VERSION"} > 0) {	914	if (${"$package\::VERSION"} > 0) {
338	if (eval "require $model") {	915	if (eval "require $model") {
339	$MODEL = $model;	916	$MODEL = $model;
340	warn "AnyEvent: found model '$model', using it.\n" if $verbose > 1;	917	warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1;
341	last;	918	last;
		919	}
342	}	920	}
343	}	921	}
344	}
345		922
346	unless ($MODEL) {	923	unless ($MODEL) {
347	# try to load a model	924	# try to load a model
348		925
349	for (@REGISTRY, @models) {	926	for (@REGISTRY, @models) {
350	my ($package, $model) = @$_;	927	my ($package, $model) = @$_;
351	if (eval "require $package"	928	if (eval "require $package"
352	and ${"$package\::VERSION"} > 0	929	and ${"$package\::VERSION"} > 0
353	and eval "require $model") {	930	and eval "require $model") {
354	$MODEL = $model;	931	$MODEL = $model;
355	warn "AnyEvent: autoprobed and loaded model '$model', using it.\n" if $verbose > 1;	932	warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1;
356	last;	933	last;
		934	}
357	}	935	}
		936
		937	$MODEL
		938	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.";
358	}	939	}
359
360	$MODEL
361	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV (or Coro+EV), Event (or Coro+Event), Glib or Tk.";
362	}	940	}
363		941
364	unshift @ISA, $MODEL;	942	unshift @ISA, $MODEL;
365	push @{"$MODEL\::ISA"}, "AnyEvent::Base";	943	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		944
		945	(shift @post_detect)->() while @post_detect;
366	}	946	}
367		947
368	$MODEL	948	$MODEL
369	}	949	}
370		950
…		…
380	$class->$func (@_);	960	$class->$func (@_);
381	}	961	}
382		962
383	package AnyEvent::Base;	963	package AnyEvent::Base;
384		964
		965	# default implementation for now and time
		966
		967	use Time::HiRes ();
		968
		969	sub time { Time::HiRes::time }
		970	sub now { Time::HiRes::time }
		971
385	# default implementation for ->condvar, ->wait, ->broadcast	972	# default implementation for ->condvar
386		973
387	sub condvar {	974	sub condvar {
388	bless \my $flag, "AnyEvent::Base::CondVar"	975	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, AnyEvent::CondVar::
389	}
390
391	sub AnyEvent::Base::CondVar::broadcast {
392	${$_[0]}++;
393	}
394
395	sub AnyEvent::Base::CondVar::wait {
396	AnyEvent->one_event while !${$_[0]};
397	}	976	}
398		977
399	# default implementation for ->signal	978	# default implementation for ->signal
400		979
401	our %SIG_CB;	980	our %SIG_CB;
…		…
417	sub AnyEvent::Base::Signal::DESTROY {	996	sub AnyEvent::Base::Signal::DESTROY {
418	my ($signal, $cb) = @{$_[0]};	997	my ($signal, $cb) = @{$_[0]};
419		998
420	delete $SIG_CB{$signal}{$cb};	999	delete $SIG_CB{$signal}{$cb};
421		1000
422	$SIG{$signal} = 'DEFAULT' unless keys %{ $SIG_CB{$signal} };	1001	delete $SIG{$signal} unless keys %{ $SIG_CB{$signal} };
423	}	1002	}
424		1003
425	# default implementation for ->child	1004	# default implementation for ->child
426		1005
427	our %PID_CB;	1006	our %PID_CB;
…		…
454	or Carp::croak "required option 'pid' is missing";	1033	or Carp::croak "required option 'pid' is missing";
455		1034
456	$PID_CB{$pid}{$arg{cb}} = $arg{cb};	1035	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
457		1036
458	unless ($WNOHANG) {	1037	unless ($WNOHANG) {
459	$WNOHANG = eval { require POSIX; &POSIX::WNOHANG } || 1;	1038	$WNOHANG = eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
460	}	1039	}
461		1040
462	unless ($CHLD_W) {	1041	unless ($CHLD_W) {
463	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);	1042	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
464	# child could be a zombie already, so make at least one round	1043	# child could be a zombie already, so make at least one round
…		…
475	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };	1054	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
476		1055
477	undef $CHLD_W unless keys %PID_CB;	1056	undef $CHLD_W unless keys %PID_CB;
478	}	1057	}
479		1058
		1059	package AnyEvent::CondVar;
		1060
		1061	our @ISA = AnyEvent::CondVar::Base::;
		1062
		1063	package AnyEvent::CondVar::Base;
		1064
		1065	use overload
		1066	'&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1067	fallback => 1;
		1068
		1069	sub _send {
		1070	# nop
		1071	}
		1072
		1073	sub send {
		1074	my $cv = shift;
		1075	$cv->{_ae_sent} = [@_];
		1076	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1077	$cv->_send;
		1078	}
		1079
		1080	sub croak {
		1081	$_[0]{_ae_croak} = $_[1];
		1082	$_[0]->send;
		1083	}
		1084
		1085	sub ready {
		1086	$_[0]{_ae_sent}
		1087	}
		1088
		1089	sub _wait {
		1090	AnyEvent->one_event while !$_[0]{_ae_sent};
		1091	}
		1092
		1093	sub recv {
		1094	$_[0]->_wait;
		1095
		1096	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1097	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1098	}
		1099
		1100	sub cb {
		1101	$_[0]{_ae_cb} = $_[1] if @_ > 1;
		1102	$_[0]{_ae_cb}
		1103	}
		1104
		1105	sub begin {
		1106	++$_[0]{_ae_counter};
		1107	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1108	}
		1109
		1110	sub end {
		1111	return if --$_[0]{_ae_counter};
		1112	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1113	}
		1114
		1115	# undocumented/compatibility with pre-3.4
		1116	*broadcast = \&send;
		1117	*wait = \&_wait;
		1118
480	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE	1119	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
		1120
		1121	This is an advanced topic that you do not normally need to use AnyEvent in
		1122	a module. This section is only of use to event loop authors who want to
		1123	provide AnyEvent compatibility.
481		1124
482	If you need to support another event library which isn't directly	1125	If you need to support another event library which isn't directly
483	supported by AnyEvent, you can supply your own interface to it by	1126	supported by AnyEvent, you can supply your own interface to it by
484	pushing, before the first watcher gets created, the package name of	1127	pushing, before the first watcher gets created, the package name of
485	the event module and the package name of the interface to use onto	1128	the event module and the package name of the interface to use onto
486	C<@AnyEvent::REGISTRY>. You can do that before and even without loading	1129	C<@AnyEvent::REGISTRY>. You can do that before and even without loading
487	AnyEvent.	1130	AnyEvent, so it is reasonably cheap.
488		1131
489	Example:	1132	Example:
490		1133
491	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];	1134	push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
492		1135
493	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>	1136	This tells AnyEvent to (literally) use the C<urxvt::anyevent::>
494	package/class when it finds the C<urxvt> package/module is loaded. When	1137	package/class when it finds the C<urxvt> package/module is already loaded.
		1138
495	AnyEvent is loaded and asked to find a suitable event model, it will	1139	When AnyEvent is loaded and asked to find a suitable event model, it
496	first check for the presence of urxvt.	1140	will first check for the presence of urxvt by trying to C<use> the
		1141	C<urxvt::anyevent> module.
497		1142
498	The class should provide implementations for all watcher types (see	1143	The class should provide implementations for all watcher types. See
499	L<AnyEvent::Impl::Event> (source code), L<AnyEvent::Impl::Glib>	1144	L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code)
500	(Source code) and so on for actual examples, use C<perldoc -m	1145	and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to
501	AnyEvent::Impl::Glib> to see the sources).	1146	see the sources.
502		1147
		1148	If you don't provide C<signal> and C<child> watchers than AnyEvent will
		1149	provide suitable (hopefully) replacements.
		1150
503	The above isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)	1151	The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt)
504	uses the above line as-is. An interface isn't included in AnyEvent	1152	terminal emulator uses the above line as-is. An interface isn't included
505	because it doesn't make sense outside the embedded interpreter inside	1153	in AnyEvent because it doesn't make sense outside the embedded interpreter
506	I<rxvt-unicode>, and it is updated and maintained as part of the	1154	inside I<rxvt-unicode>, and it is updated and maintained as part of the
507	I<rxvt-unicode> distribution.	1155	I<rxvt-unicode> distribution.
508		1156
509	I<rxvt-unicode> also cheats a bit by not providing blocking access to	1157	I<rxvt-unicode> also cheats a bit by not providing blocking access to
510	condition variables: code blocking while waiting for a condition will	1158	condition variables: code blocking while waiting for a condition will
511	C<die>. This still works with most modules/usages, and blocking calls must	1159	C<die>. This still works with most modules/usages, and blocking calls must
512	not be in an interactive application, so it makes sense.	1160	not be done in an interactive application, so it makes sense.
513		1161
514	=head1 ENVIRONMENT VARIABLES	1162	=head1 ENVIRONMENT VARIABLES
515		1163
516	The following environment variables are used by this module:	1164	The following environment variables are used by this module:
517		1165
518	C<PERL_ANYEVENT_VERBOSE> when set to C<2> or higher, reports which event	1166	=over 4
519	model gets used.
520		1167
		1168	=item C<PERL_ANYEVENT_VERBOSE>
		1169
		1170	By default, AnyEvent will be completely silent except in fatal
		1171	conditions. You can set this environment variable to make AnyEvent more
		1172	talkative.
		1173
		1174	When set to C<1> or higher, causes AnyEvent to warn about unexpected
		1175	conditions, such as not being able to load the event model specified by
		1176	C<PERL_ANYEVENT_MODEL>.
		1177
		1178	When set to C<2> or higher, cause AnyEvent to report to STDERR which event
		1179	model it chooses.
		1180
		1181	=item C<PERL_ANYEVENT_MODEL>
		1182
		1183	This can be used to specify the event model to be used by AnyEvent, before
		1184	auto detection and -probing kicks in. It must be a string consisting
		1185	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
		1186	and the resulting module name is loaded and if the load was successful,
		1187	used as event model. If it fails to load AnyEvent will proceed with
		1188	auto detection and -probing.
		1189
		1190	This functionality might change in future versions.
		1191
		1192	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
		1193	could start your program like this:
		1194
		1195	PERL_ANYEVENT_MODEL=Perl perl ...
		1196
		1197	=item C<PERL_ANYEVENT_PROTOCOLS>
		1198
		1199	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1200	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1201	of auto probing).
		1202
		1203	Must be set to a comma-separated list of protocols or address families,
		1204	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1205	used, and preference will be given to protocols mentioned earlier in the
		1206	list.
		1207
		1208	This variable can effectively be used for denial-of-service attacks
		1209	against local programs (e.g. when setuid), although the impact is likely
		1210	small, as the program has to handle connection errors already-
		1211
		1212	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1213	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1214	- only support IPv4, never try to resolve or contact IPv6
		1215	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1216	IPv6, but prefer IPv6 over IPv4.
		1217
		1218	=item C<PERL_ANYEVENT_EDNS0>
		1219
		1220	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1221	for DNS. This extension is generally useful to reduce DNS traffic, but
		1222	some (broken) firewalls drop such DNS packets, which is why it is off by
		1223	default.
		1224
		1225	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1226	EDNS0 in its DNS requests.
		1227
		1228	=item C<PERL_ANYEVENT_MAX_FORKS>
		1229
		1230	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1231	will create in parallel.
		1232
		1233	=back
		1234
521	=head1 EXAMPLE	1235	=head1 EXAMPLE PROGRAM
522		1236
523	The following program uses an io watcher to read data from stdin, a timer	1237	The following program uses an I/O watcher to read data from STDIN, a timer
524	to display a message once per second, and a condvar to exit the program	1238	to display a message once per second, and a condition variable to quit the
525	when the user enters quit:	1239	program when the user enters quit:
526		1240
527	use AnyEvent;	1241	use AnyEvent;
528		1242
529	my $cv = AnyEvent->condvar;	1243	my $cv = AnyEvent->condvar;
530		1244
531	my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {	1245	my $io_watcher = AnyEvent->io (
		1246	fh => \*STDIN,
		1247	poll => 'r',
		1248	cb => sub {
532	warn "io event <$_[0]>\n"; # will always output <r>	1249	warn "io event <$_[0]>\n"; # will always output <r>
533	chomp (my $input = <STDIN>); # read a line	1250	chomp (my $input = <STDIN>); # read a line
534	warn "read: $input\n"; # output what has been read	1251	warn "read: $input\n"; # output what has been read
535	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1252	$cv->send if $input =~ /^q/i; # quit program if /^q/i
		1253	},
536	});	1254	);
537		1255
538	my $time_watcher; # can only be used once	1256	my $time_watcher; # can only be used once
539		1257
540	sub new_timer {	1258	sub new_timer {
541	$timer = AnyEvent->timer (after => 1, cb => sub {	1259	$timer = AnyEvent->timer (after => 1, cb => sub {
…		…
544	});	1262	});
545	}	1263	}
546		1264
547	new_timer; # create first timer	1265	new_timer; # create first timer
548		1266
549	$cv->wait; # wait until user enters /^q/i	1267	$cv->recv; # wait until user enters /^q/i
550		1268
551	=head1 REAL-WORLD EXAMPLE	1269	=head1 REAL-WORLD EXAMPLE
552		1270
553	Consider the L<Net::FCP> module. It features (among others) the following	1271	Consider the L<Net::FCP> module. It features (among others) the following
554	API calls, which are to freenet what HTTP GET requests are to http:	1272	API calls, which are to freenet what HTTP GET requests are to http:
…		…
604	syswrite $txn->{fh}, $txn->{request}	1322	syswrite $txn->{fh}, $txn->{request}
605	or die "connection or write error";	1323	or die "connection or write error";
606	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	1324	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
607		1325
608	Again, C<fh_ready_r> waits till all data has arrived, and then stores the	1326	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
609	result and signals any possible waiters that the request ahs finished:	1327	result and signals any possible waiters that the request has finished:
610		1328
611	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	1329	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
612		1330
613	if (end-of-file or data complete) {	1331	if (end-of-file or data complete) {
614	$txn->{result} = $txn->{buf};	1332	$txn->{result} = $txn->{buf};
615	$txn->{finished}->broadcast;	1333	$txn->{finished}->send;
616	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	1334	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
617	}	1335	}
618		1336
619	The C<result> method, finally, just waits for the finished signal (if the	1337	The C<result> method, finally, just waits for the finished signal (if the
620	request was already finished, it doesn't wait, of course, and returns the	1338	request was already finished, it doesn't wait, of course, and returns the
621	data:	1339	data:
622		1340
623	$txn->{finished}->wait;	1341	$txn->{finished}->recv;
624	return $txn->{result};	1342	return $txn->{result};
625		1343
626	The actual code goes further and collects all errors (C<die>s, exceptions)	1344	The actual code goes further and collects all errors (C<die>s, exceptions)
627	that occured during request processing. The C<result> method detects	1345	that occurred during request processing. The C<result> method detects
628	wether an exception as thrown (it is stored inside the $txn object)	1346	whether an exception as thrown (it is stored inside the $txn object)
629	and just throws the exception, which means connection errors and other	1347	and just throws the exception, which means connection errors and other
630	problems get reported tot he code that tries to use the result, not in a	1348	problems get reported tot he code that tries to use the result, not in a
631	random callback.	1349	random callback.
632		1350
633	All of this enables the following usage styles:	1351	All of this enables the following usage styles:
634		1352
635	1. Blocking:	1353	1. Blocking:
636		1354
637	my $data = $fcp->client_get ($url);	1355	my $data = $fcp->client_get ($url);
638		1356
639	2. Blocking, but parallelizing:	1357	2. Blocking, but running in parallel:
640		1358
641	my @datas = map $_->result,	1359	my @datas = map $_->result,
642	map $fcp->txn_client_get ($_),	1360	map $fcp->txn_client_get ($_),
643	@urls;	1361	@urls;
644		1362
645	Both blocking examples work without the module user having to know	1363	Both blocking examples work without the module user having to know
646	anything about events.	1364	anything about events.
647		1365
648	3a. Event-based in a main program, using any support Event module:	1366	3a. Event-based in a main program, using any supported event module:
649		1367
650	use Event;	1368	use EV;
651		1369
652	$fcp->txn_client_get ($url)->cb (sub {	1370	$fcp->txn_client_get ($url)->cb (sub {
653	my $txn = shift;	1371	my $txn = shift;
654	my $data = $txn->result;	1372	my $data = $txn->result;
655	...	1373	...
656	});	1374	});
657		1375
658	Event::loop;	1376	EV::loop;
659		1377
660	3b. The module user could use AnyEvent, too:	1378	3b. The module user could use AnyEvent, too:
661		1379
662	use AnyEvent;	1380	use AnyEvent;
663		1381
664	my $quit = AnyEvent->condvar;	1382	my $quit = AnyEvent->condvar;
665		1383
666	$fcp->txn_client_get ($url)->cb (sub {	1384	$fcp->txn_client_get ($url)->cb (sub {
667	...	1385	...
668	$quit->broadcast;	1386	$quit->send;
669	});	1387	});
670		1388
671	$quit->wait;	1389	$quit->recv;
		1390
		1391
		1392	=head1 BENCHMARKS
		1393
		1394	To give you an idea of the performance and overheads that AnyEvent adds
		1395	over the event loops themselves and to give you an impression of the speed
		1396	of various event loops I prepared some benchmarks.
		1397
		1398	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1399
		1400	Here is a benchmark of various supported event models used natively and
		1401	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1402	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1403	which it is), lets them fire exactly once and destroys them again.
		1404
		1405	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		1406	distribution.
		1407
		1408	=head3 Explanation of the columns
		1409
		1410	I<watcher> is the number of event watchers created/destroyed. Since
		1411	different event models feature vastly different performances, each event
		1412	loop was given a number of watchers so that overall runtime is acceptable
		1413	and similar between tested event loop (and keep them from crashing): Glib
		1414	would probably take thousands of years if asked to process the same number
		1415	of watchers as EV in this benchmark.
		1416
		1417	I<bytes> is the number of bytes (as measured by the resident set size,
		1418	RSS) consumed by each watcher. This method of measuring captures both C
		1419	and Perl-based overheads.
		1420
		1421	I<create> is the time, in microseconds (millionths of seconds), that it
		1422	takes to create a single watcher. The callback is a closure shared between
		1423	all watchers, to avoid adding memory overhead. That means closure creation
		1424	and memory usage is not included in the figures.
		1425
		1426	I<invoke> is the time, in microseconds, used to invoke a simple
		1427	callback. The callback simply counts down a Perl variable and after it was
		1428	invoked "watcher" times, it would C<< ->send >> a condvar once to
		1429	signal the end of this phase.
		1430
		1431	I<destroy> is the time, in microseconds, that it takes to destroy a single
		1432	watcher.
		1433
		1434	=head3 Results
		1435
		1436	name watchers bytes create invoke destroy comment
		1437	EV/EV 400000 244 0.56 0.46 0.31 EV native interface
		1438	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
		1439	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
		1440	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
		1441	Event/Event 16000 516 31.88 31.30 0.85 Event native interface
		1442	Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
		1443	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
		1444	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
		1445	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
		1446	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
		1447
		1448	=head3 Discussion
		1449
		1450	The benchmark does I<not> measure scalability of the event loop very
		1451	well. For example, a select-based event loop (such as the pure perl one)
		1452	can never compete with an event loop that uses epoll when the number of
		1453	file descriptors grows high. In this benchmark, all events become ready at
		1454	the same time, so select/poll-based implementations get an unnatural speed
		1455	boost.
		1456
		1457	Also, note that the number of watchers usually has a nonlinear effect on
		1458	overall speed, that is, creating twice as many watchers doesn't take twice
		1459	the time - usually it takes longer. This puts event loops tested with a
		1460	higher number of watchers at a disadvantage.
		1461
		1462	To put the range of results into perspective, consider that on the
		1463	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1464	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1465	cycles with POE.
		1466
		1467	C<EV> is the sole leader regarding speed and memory use, which are both
		1468	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1469	far less memory than any other event loop and is still faster than Event
		1470	natively.
		1471
		1472	The pure perl implementation is hit in a few sweet spots (both the
		1473	constant timeout and the use of a single fd hit optimisations in the perl
		1474	interpreter and the backend itself). Nevertheless this shows that it
		1475	adds very little overhead in itself. Like any select-based backend its
		1476	performance becomes really bad with lots of file descriptors (and few of
		1477	them active), of course, but this was not subject of this benchmark.
		1478
		1479	The C<Event> module has a relatively high setup and callback invocation
		1480	cost, but overall scores in on the third place.
		1481
		1482	C<Glib>'s memory usage is quite a bit higher, but it features a
		1483	faster callback invocation and overall ends up in the same class as
		1484	C<Event>. However, Glib scales extremely badly, doubling the number of
		1485	watchers increases the processing time by more than a factor of four,
		1486	making it completely unusable when using larger numbers of watchers
		1487	(note that only a single file descriptor was used in the benchmark, so
		1488	inefficiencies of C<poll> do not account for this).
		1489
		1490	The C<Tk> adaptor works relatively well. The fact that it crashes with
		1491	more than 2000 watchers is a big setback, however, as correctness takes
		1492	precedence over speed. Nevertheless, its performance is surprising, as the
		1493	file descriptor is dup()ed for each watcher. This shows that the dup()
		1494	employed by some adaptors is not a big performance issue (it does incur a
		1495	hidden memory cost inside the kernel which is not reflected in the figures
		1496	above).
		1497
		1498	C<POE>, regardless of underlying event loop (whether using its pure perl
		1499	select-based backend or the Event module, the POE-EV backend couldn't
		1500	be tested because it wasn't working) shows abysmal performance and
		1501	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		1502	as EV watchers, and 10 times as much memory as Event (the high memory
		1503	requirements are caused by requiring a session for each watcher). Watcher
		1504	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		1505	implementation.
		1506
		1507	The design of the POE adaptor class in AnyEvent can not really account
		1508	for the performance issues, though, as session creation overhead is
		1509	small compared to execution of the state machine, which is coded pretty
		1510	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		1511	using multiple sessions is not a good approach, especially regarding
		1512	memory usage, even the author of POE could not come up with a faster
		1513	design).
		1514
		1515	=head3 Summary
		1516
		1517	=over 4
		1518
		1519	=item * Using EV through AnyEvent is faster than any other event loop
		1520	(even when used without AnyEvent), but most event loops have acceptable
		1521	performance with or without AnyEvent.
		1522
		1523	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		1524	the actual event loop, only with extremely fast event loops such as EV
		1525	adds AnyEvent significant overhead.
		1526
		1527	=item * You should avoid POE like the plague if you want performance or
		1528	reasonable memory usage.
		1529
		1530	=back
		1531
		1532	=head2 BENCHMARKING THE LARGE SERVER CASE
		1533
		1534	This benchmark actually benchmarks the event loop itself. It works by
		1535	creating a number of "servers": each server consists of a socket pair, a
		1536	timeout watcher that gets reset on activity (but never fires), and an I/O
		1537	watcher waiting for input on one side of the socket. Each time the socket
		1538	watcher reads a byte it will write that byte to a random other "server".
		1539
		1540	The effect is that there will be a lot of I/O watchers, only part of which
		1541	are active at any one point (so there is a constant number of active
		1542	fds for each loop iteration, but which fds these are is random). The
		1543	timeout is reset each time something is read because that reflects how
		1544	most timeouts work (and puts extra pressure on the event loops).
		1545
		1546	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		1547	(1%) are active. This mirrors the activity of large servers with many
		1548	connections, most of which are idle at any one point in time.
		1549
		1550	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1551	distribution.
		1552
		1553	=head3 Explanation of the columns
		1554
		1555	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1556	each server has a read and write socket end).
		1557
		1558	I<create> is the time it takes to create a socket pair (which is
		1559	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1560
		1561	I<request>, the most important value, is the time it takes to handle a
		1562	single "request", that is, reading the token from the pipe and forwarding
		1563	it to another server. This includes deleting the old timeout and creating
		1564	a new one that moves the timeout into the future.
		1565
		1566	=head3 Results
		1567
		1568	name sockets create request
		1569	EV 20000 69.01 11.16
		1570	Perl 20000 73.32 35.87
		1571	Event 20000 212.62 257.32
		1572	Glib 20000 651.16 1896.30
		1573	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1574
		1575	=head3 Discussion
		1576
		1577	This benchmark I<does> measure scalability and overall performance of the
		1578	particular event loop.
		1579
		1580	EV is again fastest. Since it is using epoll on my system, the setup time
		1581	is relatively high, though.
		1582
		1583	Perl surprisingly comes second. It is much faster than the C-based event
		1584	loops Event and Glib.
		1585
		1586	Event suffers from high setup time as well (look at its code and you will
		1587	understand why). Callback invocation also has a high overhead compared to
		1588	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1589	uses select or poll in basically all documented configurations.
		1590
		1591	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1592	clearly fails to perform with many filehandles or in busy servers.
		1593
		1594	POE is still completely out of the picture, taking over 1000 times as long
		1595	as EV, and over 100 times as long as the Perl implementation, even though
		1596	it uses a C-based event loop in this case.
		1597
		1598	=head3 Summary
		1599
		1600	=over 4
		1601
		1602	=item * The pure perl implementation performs extremely well.
		1603
		1604	=item * Avoid Glib or POE in large projects where performance matters.
		1605
		1606	=back
		1607
		1608	=head2 BENCHMARKING SMALL SERVERS
		1609
		1610	While event loops should scale (and select-based ones do not...) even to
		1611	large servers, most programs we (or I :) actually write have only a few
		1612	I/O watchers.
		1613
		1614	In this benchmark, I use the same benchmark program as in the large server
		1615	case, but it uses only eight "servers", of which three are active at any
		1616	one time. This should reflect performance for a small server relatively
		1617	well.
		1618
		1619	The columns are identical to the previous table.
		1620
		1621	=head3 Results
		1622
		1623	name sockets create request
		1624	EV 16 20.00 6.54
		1625	Perl 16 25.75 12.62
		1626	Event 16 81.27 35.86
		1627	Glib 16 32.63 15.48
		1628	POE 16 261.87 276.28 uses POE::Loop::Event
		1629
		1630	=head3 Discussion
		1631
		1632	The benchmark tries to test the performance of a typical small
		1633	server. While knowing how various event loops perform is interesting, keep
		1634	in mind that their overhead in this case is usually not as important, due
		1635	to the small absolute number of watchers (that is, you need efficiency and
		1636	speed most when you have lots of watchers, not when you only have a few of
		1637	them).
		1638
		1639	EV is again fastest.
		1640
		1641	Perl again comes second. It is noticeably faster than the C-based event
		1642	loops Event and Glib, although the difference is too small to really
		1643	matter.
		1644
		1645	POE also performs much better in this case, but is is still far behind the
		1646	others.
		1647
		1648	=head3 Summary
		1649
		1650	=over 4
		1651
		1652	=item * C-based event loops perform very well with small number of
		1653	watchers, as the management overhead dominates.
		1654
		1655	=back
		1656
		1657
		1658	=head1 FORK
		1659
		1660	Most event libraries are not fork-safe. The ones who are usually are
		1661	because they rely on inefficient but fork-safe C<select> or C<poll>
		1662	calls. Only L<EV> is fully fork-aware.
		1663
		1664	If you have to fork, you must either do so I<before> creating your first
		1665	watcher OR you must not use AnyEvent at all in the child.
		1666
		1667
		1668	=head1 SECURITY CONSIDERATIONS
		1669
		1670	AnyEvent can be forced to load any event model via
		1671	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
		1672	execute arbitrary code or directly gain access, it can easily be used to
		1673	make the program hang or malfunction in subtle ways, as AnyEvent watchers
		1674	will not be active when the program uses a different event model than
		1675	specified in the variable.
		1676
		1677	You can make AnyEvent completely ignore this variable by deleting it
		1678	before the first watcher gets created, e.g. with a C<BEGIN> block:
		1679
		1680	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		1681
		1682	use AnyEvent;
		1683
		1684	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		1685	be used to probe what backend is used and gain other information (which is
		1686	probably even less useful to an attacker than PERL_ANYEVENT_MODEL).
		1687
		1688
		1689	=head1 BUGS
		1690
		1691	Perl 5.8 has numerous memleaks that sometimes hit this module and are hard
		1692	to work around. If you suffer from memleaks, first upgrade to Perl 5.10
		1693	and check wether the leaks still show up. (Perl 5.10.0 has other annoying
		1694	mamleaks, such as leaking on C<map> and C<grep> but it is usually not as
		1695	pronounced).
		1696
672		1697
673	=head1 SEE ALSO	1698	=head1 SEE ALSO
674		1699
675	Event modules: L<Coro::Event>, L<Coro>, L<Event>, L<Glib::Event>, L<Glib>.	1700	Utility functions: L<AnyEvent::Util>.
676		1701
677	Implementations: L<AnyEvent::Impl::Coro>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>.	1702	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
		1703	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
678		1704
679	Nontrivial usage example: L<Net::FCP>.	1705	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		1706	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		1707	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
		1708	L<AnyEvent::Impl::POE>.
680		1709
681	=head1	1710	Non-blocking file handles, sockets, TCP clients and
		1711	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>.
		1712
		1713	Asynchronous DNS: L<AnyEvent::DNS>.
		1714
		1715	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>,
		1716
		1717	Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>.
		1718
		1719
		1720	=head1 AUTHOR
		1721
		1722	Marc Lehmann <schmorp@schmorp.de>
		1723	http://home.schmorp.de/
682		1724
683	=cut	1725	=cut
684		1726
685	1	1727	1
686		1728

Diff Legend

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