ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.60 by root, Fri Apr 25 01:05:26 2008 UTC vs.
Revision 1.145 by root, Thu May 29 03:45:37 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	EV, Event, Coro::EV, Coro::Event, Glib, Tk, Perl, Event::Lib, Qt - 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
…		…
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 whether 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		22
23	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)	23	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
24		24
25	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen	25	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
26	nowadays. So what is different about AnyEvent?	26	nowadays. So what is different about AnyEvent?
…		…
48	isn't itself. What's worse, all the potential users of your module are	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.	49	I<also> forced to use the same event loop you use.
50		50
51	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works	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	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	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,	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	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	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	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).	58	event loops to AnyEvent, too, so it is future-proof).
59		59
60	In addition to being free of having to use I<the one and only true event	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	61	model>, AnyEvent also is free of bloat and policy: with POE or similar
62	modules, you get an enourmous amount of code and strict rules you have to	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	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	64	offering the functionality that is necessary, in as thin as a wrapper as
65	technically possible.	65	technically possible.
66		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
67	Of course, if you want lots of policy (this can arguably be somewhat	73	Now, if you I<do want> lots of policy (this can arguably be somewhat
68	useful) and you want to force your users to use the one and only event	74	useful) and you want to force your users to use the one and only event
69	model, you should I<not> use this module.	75	model, you should I<not> use this module.
70
71		76
72	=head1 DESCRIPTION	77	=head1 DESCRIPTION
73		78
74	L<AnyEvent> provides an identical interface to multiple event loops. This	79	L<AnyEvent> provides an identical interface to multiple event loops. This
75	allows module authors to utilise an event loop without forcing module	80	allows module authors to utilise an event loop without forcing module
…		…
78		83
79	The interface itself is vaguely similar, but not identical to the L<Event>	84	The interface itself is vaguely similar, but not identical to the L<Event>
80	module.	85	module.
81		86
82	During the first call of any watcher-creation method, the module tries	87	During the first call of any watcher-creation method, the module tries
83	to detect the currently loaded event loop by probing whether one of	88	to detect the currently loaded event loop by probing whether one of the
84	the following modules is already loaded: L<Coro::EV>, L<Coro::Event>,	89	following modules is already loaded: L<EV>,
85	L<EV>, L<Event>, L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>. The first one	90	L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>,
86	found is used. If none are found, the module tries to load these modules	91	L<POE>. The first one found is used. If none are found, the module tries
87	(excluding Event::Lib and Qt) in the order given. The first one that can	92	to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl
		93	adaptor should always succeed) in the order given. The first one that can
88	be successfully loaded will be used. If, after this, still none could be	94	be successfully loaded will be used. If, after this, still none could be
89	found, AnyEvent will fall back to a pure-perl event loop, which is not	95	found, AnyEvent will fall back to a pure-perl event loop, which is not
90	very efficient, but should work everywhere.	96	very efficient, but should work everywhere.
91		97
92	Because AnyEvent first checks for modules that are already loaded, loading	98	Because AnyEvent first checks for modules that are already loaded, loading
…		…
102	starts using it, all bets are off. Maybe you should tell their authors to	108	starts using it, all bets are off. Maybe you should tell their authors to
103	use AnyEvent so their modules work together with others seamlessly...	109	use AnyEvent so their modules work together with others seamlessly...
104		110
105	The pure-perl implementation of AnyEvent is called	111	The pure-perl implementation of AnyEvent is called
106	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
107	explicitly.	113	explicitly and enjoy the high availability of that event loop :)
108		114
109	=head1 WATCHERS	115	=head1 WATCHERS
110		116
111	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
112	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
113	the callback to call, the filehandle to watch, etc.	119	the callback to call, the file handle to watch, etc.
114		120
115	These watchers are normal Perl objects with normal Perl lifetime. After	121	These watchers are normal Perl objects with normal Perl lifetime. After
116	creating a watcher it will immediately "watch" for events and invoke the	122	creating a watcher it will immediately "watch" for events and invoke the
117	callback when the event occurs (of course, only when the event model	123	callback when the event occurs (of course, only when the event model
118	is in control).	124	is in control).
…		…
135		141
136	Note that C<my $w; $w => combination. This is necessary because in Perl,	142	Note that C<my $w; $w => combination. This is necessary because in Perl,
137	my variables are only visible after the statement in which they are	143	my variables are only visible after the statement in which they are
138	declared.	144	declared.
139		145
140	=head2 IO WATCHERS	146	=head2 I/O WATCHERS
141		147
142	You can create an I/O watcher by calling the C<< AnyEvent->io >> method	148	You can create an I/O watcher by calling the C<< AnyEvent->io >> method
143	with the following mandatory key-value pairs as arguments:	149	with the following mandatory key-value pairs as arguments:
144		150
145	C<fh> the Perl I<file handle> (I<not> file descriptor) to watch for	151	C<fh> the Perl I<file handle> (I<not> file descriptor) to watch
146	events. C<poll> must be a string that is either C<r> or C<w>, which	152	for events. C<poll> must be a string that is either C<r> or C<w>,
147	creates a watcher waiting for "r"eadable or "w"ritable events,	153	which creates a watcher waiting for "r"eadable or "w"ritable events,
148	respectively. C<cb> is the callback to invoke each time the file handle	154	respectively. C<cb> is the callback to invoke each time the file handle
149	becomes ready.	155	becomes ready.
150		156
151	As long as the I/O watcher exists it will keep the file descriptor or a	157	Although the callback might get passed parameters, their value and
152	copy of it alive/open.	158	presence is undefined and you cannot rely on them. Portable AnyEvent
		159	callbacks cannot use arguments passed to I/O watcher callbacks.
153		160
		161	The I/O watcher might use the underlying file descriptor or a copy of it.
154	It is not allowed to close a file handle as long as any watcher is active	162	You must not close a file handle as long as any watcher is active on the
155	on the underlying file descriptor.	163	underlying file descriptor.
156		164
157	Some event loops issue spurious readyness notifications, so you should	165	Some event loops issue spurious readyness notifications, so you should
158	always use non-blocking calls when reading/writing from/to your file	166	always use non-blocking calls when reading/writing from/to your file
159	handles.	167	handles.
160		168
…		…
171		179
172	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 >>
173	method with the following mandatory arguments:	181	method with the following mandatory arguments:
174		182
175	C<after> specifies after how many seconds (fractional values are	183	C<after> specifies after how many seconds (fractional values are
176	supported) should the timer activate. C<cb> the callback to invoke in that	184	supported) the callback should be invoked. C<cb> is the callback to invoke
177	case.	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.
178		190
179	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
180	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
181	and Glib).	193	and Glib).
182		194
…		…
221	timers.	233	timers.
222		234
223	AnyEvent always prefers relative timers, if available, matching the	235	AnyEvent always prefers relative timers, if available, matching the
224	AnyEvent API.	236	AnyEvent API.
225		237
		238	AnyEvent has two additional methods that return the "current time":
		239
		240	=over 4
		241
		242	=item AnyEvent->time
		243
		244	This returns the "current wallclock time" as a fractional number of
		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).
		247
		248	It progresses independently of any event loop processing, i.e. each call
		249	will check the system clock, which usually gets updated frequently.
		250
		251	=item AnyEvent->now
		252
		253	This also returns the "current wallclock time", but unlike C<time>, above,
		254	this value might change only once per event loop iteration, depending on
		255	the event loop (most return the same time as C<time>, above). This is the
		256	time that AnyEvent's timers get scheduled against.
		257
		258	I<In almost all cases (in all cases if you don't care), this is the
		259	function to call when you want to know the current time.>
		260
		261	This function is also often faster then C<< AnyEvent->time >>, and
		262	thus the preferred method if you want some timestamp (for example,
		263	L<AnyEvent::Handle> uses this to update it's activity timeouts).
		264
		265	The rest of this section is only of relevance if you try to be very exact
		266	with your timing, you can skip it without bad conscience.
		267
		268	For a practical example of when these times differ, consider L<Event::Lib>
		269	and L<EV> and the following set-up:
		270
		271	The event loop is running and has just invoked one of your callback at
		272	time=500 (assume no other callbacks delay processing). In your callback,
		273	you wait a second by executing C<sleep 1> (blocking the process for a
		274	second) and then (at time=501) you create a relative timer that fires
		275	after three seconds.
		276
		277	With L<Event::Lib>, C<< AnyEvent->time >> and C<< AnyEvent->now >> will
		278	both return C<501>, because that is the current time, and the timer will
		279	be scheduled to fire at time=504 (C<501> + C<3>).
		280
		281	With L<EV>, C<< AnyEvent->time >> returns C<501> (as that is the current
		282	time), but C<< AnyEvent->now >> returns C<500>, as that is the time the
		283	last event processing phase started. With L<EV>, your timer gets scheduled
		284	to run at time=503 (C<500> + C<3>).
		285
		286	In one sense, L<Event::Lib> is more exact, as it uses the current time
		287	regardless of any delays introduced by event processing. However, most
		288	callbacks do not expect large delays in processing, so this causes a
		289	higher drift (and a lot more system calls to get the current time).
		290
		291	In another sense, L<EV> is more exact, as your timer will be scheduled at
		292	the same time, regardless of how long event processing actually took.
		293
		294	In either case, if you care (and in most cases, you don't), then you
		295	can get whatever behaviour you want with any event loop, by taking the
		296	difference between C<< AnyEvent->time >> and C<< AnyEvent->now >> into
		297	account.
		298
		299	=back
		300
226	=head2 SIGNAL WATCHERS	301	=head2 SIGNAL WATCHERS
227		302
228	You can watch for signals using a signal watcher, C<signal> is the signal	303	You can watch for signals using a signal watcher, C<signal> is the signal
229	I<name> without any C<SIG> prefix, C<cb> is the Perl callback to	304	I<name> without any C<SIG> prefix, C<cb> is the Perl callback to
230	be invoked whenever a signal occurs.	305	be invoked whenever a signal occurs.
231		306
		307	Although the callback might get passed parameters, their value and
		308	presence is undefined and you cannot rely on them. Portable AnyEvent
		309	callbacks cannot use arguments passed to signal watcher callbacks.
		310
232	Multiple signal occurances can be clumped together into one callback	311	Multiple signal occurrences can be clumped together into one callback
233	invocation, and callback invocation will be synchronous. synchronous means	312	invocation, and callback invocation will be synchronous. Synchronous means
234	that it might take a while until the signal gets handled by the process,	313	that it might take a while until the signal gets handled by the process,
235	but it is guarenteed not to interrupt any other callbacks.	314	but it is guaranteed not to interrupt any other callbacks.
236		315
237	The main advantage of using these watchers is that you can share a signal	316	The main advantage of using these watchers is that you can share a signal
238	between multiple watchers.	317	between multiple watchers.
239		318
240	This watcher might use C<%SIG>, so programs overwriting those signals	319	This watcher might use C<%SIG>, so programs overwriting those signals
…		…
250		329
251	The child process is specified by the C<pid> argument (if set to C<0>, it	330	The child process is specified by the C<pid> argument (if set to C<0>, it
252	watches for any child process exit). The watcher will trigger as often	331	watches for any child process exit). The watcher will trigger as often
253	as status change for the child are received. This works by installing a	332	as status change for the child are received. This works by installing a
254	signal handler for C<SIGCHLD>. The callback will be called with the pid	333	signal handler for C<SIGCHLD>. The callback will be called with the pid
255	and exit status (as returned by waitpid).	334	and exit status (as returned by waitpid), so unlike other watcher types,
		335	you I<can> rely on child watcher callback arguments.
256		336
257	Example: wait for pid 1333	337	There is a slight catch to child watchers, however: you usually start them
		338	I<after> the child process was created, and this means the process could
		339	have exited already (and no SIGCHLD will be sent anymore).
		340
		341	Not all event models handle this correctly (POE doesn't), but even for
		342	event models that I<do> handle this correctly, they usually need to be
		343	loaded before the process exits (i.e. before you fork in the first place).
		344
		345	This means you cannot create a child watcher as the very first thing in an
		346	AnyEvent program, you I<have> to create at least one watcher before you
		347	C<fork> the child (alternatively, you can call C<AnyEvent::detect>).
		348
		349	Example: fork a process and wait for it
		350
		351	my $done = AnyEvent->condvar;
		352
		353	my $pid = fork or exit 5;
258		354
259	my $w = AnyEvent->child (	355	my $w = AnyEvent->child (
260	pid => 1333,	356	pid => $pid,
261	cb => sub {	357	cb => sub {
262	my ($pid, $status) = @_;	358	my ($pid, $status) = @_;
263	warn "pid $pid exited with status $status";	359	warn "pid $pid exited with status $status";
		360	$done->send;
264	},	361	},
265	);	362	);
266		363
		364	# do something else, then wait for process exit
		365	$done->recv;
		366
267	=head2 CONDITION VARIABLES	367	=head2 CONDITION VARIABLES
268		368
		369	If you are familiar with some event loops you will know that all of them
		370	require you to run some blocking "loop", "run" or similar function that
		371	will actively watch for new events and call your callbacks.
		372
		373	AnyEvent is different, it expects somebody else to run the event loop and
		374	will only block when necessary (usually when told by the user).
		375
		376	The instrument to do that is called a "condition variable", so called
		377	because they represent a condition that must become true.
		378
269	Condition variables can be created by calling the C<< AnyEvent->condvar >>	379	Condition variables can be created by calling the C<< AnyEvent->condvar
270	method without any arguments.	380	>> method, usually without arguments. The only argument pair allowed is
		381	C<cb>, which specifies a callback to be called when the condition variable
		382	becomes true.
271		383
272	A condition variable waits for a condition - precisely that the C<<	384	After creation, the condition variable is "false" until it becomes "true"
273	->broadcast >> method has been called.	385	by calling the C<send> method (or calling the condition variable as if it
		386	were a callback, read about the caveats in the description for the C<<
		387	->send >> method).
274		388
275	They are very useful to signal that a condition has been fulfilled, for	389	Condition variables are similar to callbacks, except that you can
		390	optionally wait for them. They can also be called merge points - points
		391	in time where multiple outstanding events have been processed. And yet
		392	another way to call them is transactions - each condition variable can be
		393	used to represent a transaction, which finishes at some point and delivers
		394	a result.
		395
		396	Condition variables are very useful to signal that something has finished,
276	example, if you write a module that does asynchronous http requests,	397	for example, if you write a module that does asynchronous http requests,
277	then a condition variable would be the ideal candidate to signal the	398	then a condition variable would be the ideal candidate to signal the
278	availability of results.	399	availability of results. The user can either act when the callback is
		400	called or can synchronously C<< ->recv >> for the results.
279		401
280	You can also use condition variables to block your main program until	402	You can also use them to simulate traditional event loops - for example,
281	an event occurs - for example, you could C<< ->wait >> in your main	403	you can block your main program until an event occurs - for example, you
282	program until the user clicks the Quit button in your app, which would C<<	404	could C<< ->recv >> in your main program until the user clicks the Quit
283	->broadcast >> the "quit" event.	405	button of your app, which would C<< ->send >> the "quit" event.
284		406
285	Note that condition variables recurse into the event loop - if you have	407	Note that condition variables recurse into the event loop - if you have
286	two pirces of code that call C<< ->wait >> in a round-robbin fashion, you	408	two pieces of code that call C<< ->recv >> in a round-robin fashion, you
287	lose. Therefore, condition variables are good to export to your caller, but	409	lose. Therefore, condition variables are good to export to your caller, but
288	you should avoid making a blocking wait yourself, at least in callbacks,	410	you should avoid making a blocking wait yourself, at least in callbacks,
289	as this asks for trouble.	411	as this asks for trouble.
290		412
291	This object has two methods:	413	Condition variables are represented by hash refs in perl, and the keys
		414	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		415	easy (it is often useful to build your own transaction class on top of
		416	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		417	it's C<new> method in your own C<new> method.
		418
		419	There are two "sides" to a condition variable - the "producer side" which
		420	eventually calls C<< -> send >>, and the "consumer side", which waits
		421	for the send to occur.
		422
		423	Example: wait for a timer.
		424
		425	# wait till the result is ready
		426	my $result_ready = AnyEvent->condvar;
		427
		428	# do something such as adding a timer
		429	# or socket watcher the calls $result_ready->send
		430	# when the "result" is ready.
		431	# in this case, we simply use a timer:
		432	my $w = AnyEvent->timer (
		433	after => 1,
		434	cb => sub { $result_ready->send },
		435	);
		436
		437	# this "blocks" (while handling events) till the callback
		438	# calls send
		439	$result_ready->recv;
		440
		441	Example: wait for a timer, but take advantage of the fact that
		442	condition variables are also code references.
		443
		444	my $done = AnyEvent->condvar;
		445	my $delay = AnyEvent->timer (after => 5, cb => $done);
		446	$done->recv;
		447
		448	=head3 METHODS FOR PRODUCERS
		449
		450	These methods should only be used by the producing side, i.e. the
		451	code/module that eventually sends the signal. Note that it is also
		452	the producer side which creates the condvar in most cases, but it isn't
		453	uncommon for the consumer to create it as well.
292		454
293	=over 4	455	=over 4
294		456
		457	=item $cv->send (...)
		458
		459	Flag the condition as ready - a running C<< ->recv >> and all further
		460	calls to C<recv> will (eventually) return after this method has been
		461	called. If nobody is waiting the send will be remembered.
		462
		463	If a callback has been set on the condition variable, it is called
		464	immediately from within send.
		465
		466	Any arguments passed to the C<send> call will be returned by all
		467	future C<< ->recv >> calls.
		468
		469	Condition variables are overloaded so one can call them directly
		470	(as a code reference). Calling them directly is the same as calling
		471	C<send>. Note, however, that many C-based event loops do not handle
		472	overloading, so as tempting as it may be, passing a condition variable
		473	instead of a callback does not work. Both the pure perl and EV loops
		474	support overloading, however, as well as all functions that use perl to
		475	invoke a callback (as in L<AnyEvent::Socket> and L<AnyEvent::DNS> for
		476	example).
		477
		478	=item $cv->croak ($error)
		479
		480	Similar to send, but causes all call's to C<< ->recv >> to invoke
		481	C<Carp::croak> with the given error message/object/scalar.
		482
		483	This can be used to signal any errors to the condition variable
		484	user/consumer.
		485
		486	=item $cv->begin ([group callback])
		487
295	=item $cv->wait	488	=item $cv->end
296		489
297	Wait (blocking if necessary) until the C<< ->broadcast >> method has been	490	These two methods are EXPERIMENTAL and MIGHT CHANGE.
		491
		492	These two methods can be used to combine many transactions/events into
		493	one. For example, a function that pings many hosts in parallel might want
		494	to use a condition variable for the whole process.
		495
		496	Every call to C<< ->begin >> will increment a counter, and every call to
		497	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		498	>>, the (last) callback passed to C<begin> will be executed. That callback
		499	is I<supposed> to call C<< ->send >>, but that is not required. If no
		500	callback was set, C<send> will be called without any arguments.
		501
		502	Let's clarify this with the ping example:
		503
		504	my $cv = AnyEvent->condvar;
		505
		506	my %result;
		507	$cv->begin (sub { $cv->send (\%result) });
		508
		509	for my $host (@list_of_hosts) {
		510	$cv->begin;
		511	ping_host_then_call_callback $host, sub {
		512	$result{$host} = ...;
		513	$cv->end;
		514	};
		515	}
		516
		517	$cv->end;
		518
		519	This code fragment supposedly pings a number of hosts and calls
		520	C<send> after results for all then have have been gathered - in any
		521	order. To achieve this, the code issues a call to C<begin> when it starts
		522	each ping request and calls C<end> when it has received some result for
		523	it. Since C<begin> and C<end> only maintain a counter, the order in which
		524	results arrive is not relevant.
		525
		526	There is an additional bracketing call to C<begin> and C<end> outside the
		527	loop, which serves two important purposes: first, it sets the callback
		528	to be called once the counter reaches C<0>, and second, it ensures that
		529	C<send> is called even when C<no> hosts are being pinged (the loop
		530	doesn't execute once).
		531
		532	This is the general pattern when you "fan out" into multiple subrequests:
		533	use an outer C<begin>/C<end> pair to set the callback and ensure C<end>
		534	is called at least once, and then, for each subrequest you start, call
		535	C<begin> and for each subrequest you finish, call C<end>.
		536
		537	=back
		538
		539	=head3 METHODS FOR CONSUMERS
		540
		541	These methods should only be used by the consuming side, i.e. the
		542	code awaits the condition.
		543
		544	=over 4
		545
		546	=item $cv->recv
		547
		548	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
298	called on c<$cv>, while servicing other watchers normally.	549	>> methods have been called on c<$cv>, while servicing other watchers
		550	normally.
299		551
300	You can only wait once on a condition - additional calls will return	552	You can only wait once on a condition - additional calls are valid but
301	immediately.	553	will return immediately.
		554
		555	If an error condition has been set by calling C<< ->croak >>, then this
		556	function will call C<croak>.
		557
		558	In list context, all parameters passed to C<send> will be returned,
		559	in scalar context only the first one will be returned.
302		560
303	Not all event models support a blocking wait - some die in that case	561	Not all event models support a blocking wait - some die in that case
304	(programs might want to do that to stay interactive), so I<if you are	562	(programs might want to do that to stay interactive), so I<if you are
305	using this from a module, never require a blocking wait>, but let the	563	using this from a module, never require a blocking wait>, but let the
306	caller decide whether the call will block or not (for example, by coupling	564	caller decide whether the call will block or not (for example, by coupling
307	condition variables with some kind of request results and supporting	565	condition variables with some kind of request results and supporting
308	callbacks so the caller knows that getting the result will not block,	566	callbacks so the caller knows that getting the result will not block,
309	while still suppporting blocking waits if the caller so desires).	567	while still supporting blocking waits if the caller so desires).
310		568
311	Another reason I<never> to C<< ->wait >> in a module is that you cannot	569	Another reason I<never> to C<< ->recv >> in a module is that you cannot
312	sensibly have two C<< ->wait >>'s in parallel, as that would require	570	sensibly have two C<< ->recv >>'s in parallel, as that would require
313	multiple interpreters or coroutines/threads, none of which C<AnyEvent>	571	multiple interpreters or coroutines/threads, none of which C<AnyEvent>
314	can supply (the coroutine-aware backends L<AnyEvent::Impl::CoroEV> and	572	can supply.
315	L<AnyEvent::Impl::CoroEvent> explicitly support concurrent C<< ->wait >>'s
316	from different coroutines, however).
317		573
318	=item $cv->broadcast	574	The L<Coro> module, however, I<can> and I<does> supply coroutines and, in
		575	fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe
		576	versions and also integrates coroutines into AnyEvent, making blocking
		577	C<< ->recv >> calls perfectly safe as long as they are done from another
		578	coroutine (one that doesn't run the event loop).
319		579
320	Flag the condition as ready - a running C<< ->wait >> and all further	580	You can ensure that C<< -recv >> never blocks by setting a callback and
321	calls to C<wait> will (eventually) return after this method has been	581	only calling C<< ->recv >> from within that callback (or at a later
322	called. If nobody is waiting the broadcast will be remembered..	582	time). This will work even when the event loop does not support blocking
		583	waits otherwise.
		584
		585	=item $bool = $cv->ready
		586
		587	Returns true when the condition is "true", i.e. whether C<send> or
		588	C<croak> have been called.
		589
		590	=item $cb = $cv->cb ([new callback])
		591
		592	This is a mutator function that returns the callback set and optionally
		593	replaces it before doing so.
		594
		595	The callback will be called when the condition becomes "true", i.e. when
		596	C<send> or C<croak> are called. Calling C<recv> inside the callback
		597	or at any later time is guaranteed not to block.
323		598
324	=back	599	=back
325
326	Example:
327
328	# wait till the result is ready
329	my $result_ready = AnyEvent->condvar;
330
331	# do something such as adding a timer
332	# or socket watcher the calls $result_ready->broadcast
333	# when the "result" is ready.
334	# in this case, we simply use a timer:
335	my $w = AnyEvent->timer (
336	after => 1,
337	cb => sub { $result_ready->broadcast },
338	);
339
340	# this "blocks" (while handling events) till the watcher
341	# calls broadcast
342	$result_ready->wait;
343		600
344	=head1 GLOBAL VARIABLES AND FUNCTIONS	601	=head1 GLOBAL VARIABLES AND FUNCTIONS
345		602
346	=over 4	603	=over 4
347		604
…		…
353	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case	610	C<AnyEvent::Impl:xxx> modules, but can be any other class in the case
354	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).	611	AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>).
355		612
356	The known classes so far are:	613	The known classes so far are:
357		614
358	AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
359	AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
360	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).	615	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
361	AnyEvent::Impl::Event based on Event, second best choice.	616	AnyEvent::Impl::Event based on Event, second best choice.
		617	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
362	AnyEvent::Impl::Glib based on Glib, third-best choice.	618	AnyEvent::Impl::Glib based on Glib, third-best choice.
363	AnyEvent::Impl::Tk based on Tk, very bad choice.	619	AnyEvent::Impl::Tk based on Tk, very bad choice.
364	AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable.
365	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).	620	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
366	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.	621	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		622	AnyEvent::Impl::POE based on POE, not generic enough for full support.
		623
		624	There is no support for WxWidgets, as WxWidgets has no support for
		625	watching file handles. However, you can use WxWidgets through the
		626	POE Adaptor, as POE has a Wx backend that simply polls 20 times per
		627	second, which was considered to be too horrible to even consider for
		628	AnyEvent. Likewise, other POE backends can be used by AnyEvent by using
		629	it's adaptor.
		630
		631	AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when
		632	autodetecting them.
367		633
368	=item AnyEvent::detect	634	=item AnyEvent::detect
369		635
370	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model	636	Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model
371	if necessary. You should only call this function right before you would	637	if necessary. You should only call this function right before you would
372	have created an AnyEvent watcher anyway, that is, as late as possible at	638	have created an AnyEvent watcher anyway, that is, as late as possible at
373	runtime.	639	runtime.
374		640
		641	=item $guard = AnyEvent::post_detect { BLOCK }
		642
		643	Arranges for the code block to be executed as soon as the event model is
		644	autodetected (or immediately if this has already happened).
		645
		646	If called in scalar or list context, then it creates and returns an object
		647	that automatically removes the callback again when it is destroyed. See
		648	L<Coro::BDB> for a case where this is useful.
		649
		650	=item @AnyEvent::post_detect
		651
		652	If there are any code references in this array (you can C<push> to it
		653	before or after loading AnyEvent), then they will called directly after
		654	the event loop has been chosen.
		655
		656	You should check C<$AnyEvent::MODEL> before adding to this array, though:
		657	if it contains a true value then the event loop has already been detected,
		658	and the array will be ignored.
		659
		660	Best use C<AnyEvent::post_detect { BLOCK }> instead.
		661
375	=back	662	=back
376		663
377	=head1 WHAT TO DO IN A MODULE	664	=head1 WHAT TO DO IN A MODULE
378		665
379	As a module author, you should C<use AnyEvent> and call AnyEvent methods	666	As a module author, you should C<use AnyEvent> and call AnyEvent methods
…		…
382	Be careful when you create watchers in the module body - AnyEvent will	669	Be careful when you create watchers in the module body - AnyEvent will
383	decide which event module to use as soon as the first method is called, so	670	decide which event module to use as soon as the first method is called, so
384	by calling AnyEvent in your module body you force the user of your module	671	by calling AnyEvent in your module body you force the user of your module
385	to load the event module first.	672	to load the event module first.
386		673
387	Never call C<< ->wait >> on a condition variable unless you I<know> that	674	Never call C<< ->recv >> on a condition variable unless you I<know> that
388	the C<< ->broadcast >> method has been called on it already. This is	675	the C<< ->send >> method has been called on it already. This is
389	because it will stall the whole program, and the whole point of using	676	because it will stall the whole program, and the whole point of using
390	events is to stay interactive.	677	events is to stay interactive.
391		678
392	It is fine, however, to call C<< ->wait >> when the user of your module	679	It is fine, however, to call C<< ->recv >> when the user of your module
393	requests it (i.e. if you create a http request object ad have a method	680	requests it (i.e. if you create a http request object ad have a method
394	called C<results> that returns the results, it should call C<< ->wait >>	681	called C<results> that returns the results, it should call C<< ->recv >>
395	freely, as the user of your module knows what she is doing. always).	682	freely, as the user of your module knows what she is doing. always).
396		683
397	=head1 WHAT TO DO IN THE MAIN PROGRAM	684	=head1 WHAT TO DO IN THE MAIN PROGRAM
398		685
399	There will always be a single main program - the only place that should	686	There will always be a single main program - the only place that should
…		…
401		688
402	If it doesn't care, it can just "use AnyEvent" and use it itself, or not	689	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
403	do anything special (it does not need to be event-based) and let AnyEvent	690	do anything special (it does not need to be event-based) and let AnyEvent
404	decide which implementation to chose if some module relies on it.	691	decide which implementation to chose if some module relies on it.
405		692
406	If the main program relies on a specific event model. For example, in	693	If the main program relies on a specific event model - for example, in
407	Gtk2 programs you have to rely on the Glib module. You should load the	694	Gtk2 programs you have to rely on the Glib module - you should load the
408	event module before loading AnyEvent or any module that uses it: generally	695	event module before loading AnyEvent or any module that uses it: generally
409	speaking, you should load it as early as possible. The reason is that	696	speaking, you should load it as early as possible. The reason is that
410	modules might create watchers when they are loaded, and AnyEvent will	697	modules might create watchers when they are loaded, and AnyEvent will
411	decide on the event model to use as soon as it creates watchers, and it	698	decide on the event model to use as soon as it creates watchers, and it
412	might chose the wrong one unless you load the correct one yourself.	699	might chose the wrong one unless you load the correct one yourself.
413		700
414	You can chose to use a rather inefficient pure-perl implementation by	701	You can chose to use a pure-perl implementation by loading the
415	loading the C<AnyEvent::Impl::Perl> module, which gives you similar	702	C<AnyEvent::Impl::Perl> module, which gives you similar behaviour
416	behaviour everywhere, but letting AnyEvent chose is generally better.	703	everywhere, but letting AnyEvent chose the model is generally better.
		704
		705	=head2 MAINLOOP EMULATION
		706
		707	Sometimes (often for short test scripts, or even standalone programs who
		708	only want to use AnyEvent), you do not want to run a specific event loop.
		709
		710	In that case, you can use a condition variable like this:
		711
		712	AnyEvent->condvar->recv;
		713
		714	This has the effect of entering the event loop and looping forever.
		715
		716	Note that usually your program has some exit condition, in which case
		717	it is better to use the "traditional" approach of storing a condition
		718	variable somewhere, waiting for it, and sending it when the program should
		719	exit cleanly.
		720
		721
		722	=head1 OTHER MODULES
		723
		724	The following is a non-exhaustive list of additional modules that use
		725	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		726	in the same program. Some of the modules come with AnyEvent, some are
		727	available via CPAN.
		728
		729	=over 4
		730
		731	=item L<AnyEvent::Util>
		732
		733	Contains various utility functions that replace often-used but blocking
		734	functions such as C<inet_aton> by event-/callback-based versions.
		735
		736	=item L<AnyEvent::Handle>
		737
		738	Provide read and write buffers and manages watchers for reads and writes.
		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::DNS>
		747
		748	Provides rich asynchronous DNS resolver capabilities.
		749
		750	=item L<AnyEvent::HTTPD>
		751
		752	Provides a simple web application server framework.
		753
		754	=item L<AnyEvent::FastPing>
		755
		756	The fastest ping in the west.
		757
		758	=item L<Net::IRC3>
		759
		760	AnyEvent based IRC client module family.
		761
		762	=item L<Net::XMPP2>
		763
		764	AnyEvent based XMPP (Jabber protocol) module family.
		765
		766	=item L<Net::FCP>
		767
		768	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		769	of AnyEvent.
		770
		771	=item L<Event::ExecFlow>
		772
		773	High level API for event-based execution flow control.
		774
		775	=item L<Coro>
		776
		777	Has special support for AnyEvent via L<Coro::AnyEvent>.
		778
		779	=item L<AnyEvent::AIO>, L<IO::AIO>
		780
		781	Truly asynchronous I/O, should be in the toolbox of every event
		782	programmer. AnyEvent::AIO transparently fuses IO::AIO and AnyEvent
		783	together.
		784
		785	=item L<AnyEvent::BDB>, L<BDB>
		786
		787	Truly asynchronous Berkeley DB access. AnyEvent::AIO transparently fuses
		788	IO::AIO and AnyEvent together.
		789
		790	=item L<IO::Lambda>
		791
		792	The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
		793
		794	=back
417		795
418	=cut	796	=cut
419		797
420	package AnyEvent;	798	package AnyEvent;
421		799
422	no warnings;	800	no warnings;
423	use strict;	801	use strict;
424		802
425	use Carp;	803	use Carp;
426		804
427	our $VERSION = '3.2';	805	our $VERSION = '4.1';
428	our $MODEL;	806	our $MODEL;
429		807
430	our $AUTOLOAD;	808	our $AUTOLOAD;
431	our @ISA;	809	our @ISA;
432		810
		811	our @REGISTRY;
		812
		813	our $WIN32;
		814
		815	BEGIN {
		816	my $win32 = ! ! ($^O =~ /mswin32/i);
		817	eval "sub WIN32(){ $win32 }";
		818	}
		819
433	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;	820	our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1;
434		821
435	our @REGISTRY;	822	our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred
		823
		824	{
		825	my $idx;
		826	$PROTOCOL{$_} = ++$idx
		827	for reverse split /\s*,\s*/,
		828	$ENV{PERL_ANYEVENT_PROTOCOLS} || "ipv4,ipv6";
		829	}
436		830
437	my @models = (	831	my @models = (
438	[Coro::EV:: => AnyEvent::Impl::CoroEV::],
439	[Coro::Event:: => AnyEvent::Impl::CoroEvent::],
440	[EV:: => AnyEvent::Impl::EV::],	832	[EV:: => AnyEvent::Impl::EV::],
441	[Event:: => AnyEvent::Impl::Event::],	833	[Event:: => AnyEvent::Impl::Event::],
442	[Glib:: => AnyEvent::Impl::Glib::],
443	[Tk:: => AnyEvent::Impl::Tk::],
444	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],	834	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
		835	# everything below here will not be autoprobed
		836	# as the pureperl backend should work everywhere
		837	# and is usually faster
		838	[Tk:: => AnyEvent::Impl::Tk::], # crashes with many handles
		839	[Glib:: => AnyEvent::Impl::Glib::], # becomes extremely slow with many watchers
		840	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
		841	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
		842	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
		843	[Wx:: => AnyEvent::Impl::POE::],
		844	[Prima:: => AnyEvent::Impl::POE::],
445	);	845	);
446	my @models_detect = (
447	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
448	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
449	);
450		846
451	our %method = map +($_ => 1), qw(io timer signal child condvar broadcast wait one_event DESTROY);	847	our %method = map +($_ => 1), qw(io timer time now signal child condvar one_event DESTROY);
		848
		849	our @post_detect;
		850
		851	sub post_detect(&) {
		852	my ($cb) = @_;
		853
		854	if ($MODEL) {
		855	$cb->();
		856
		857	1
		858	} else {
		859	push @post_detect, $cb;
		860
		861	defined wantarray
		862	? bless \$cb, "AnyEvent::Util::PostDetect"
		863	: ()
		864	}
		865	}
		866
		867	sub AnyEvent::Util::PostDetect::DESTROY {
		868	@post_detect = grep $_ != ${$_[0]}, @post_detect;
		869	}
452		870
453	sub detect() {	871	sub detect() {
454	unless ($MODEL) {	872	unless ($MODEL) {
455	no strict 'refs';	873	no strict 'refs';
		874	local $SIG{__DIE__};
456		875
457	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {	876	if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) {
458	my $model = "AnyEvent::Impl::$1";	877	my $model = "AnyEvent::Impl::$1";
459	if (eval "require $model") {	878	if (eval "require $model") {
460	$MODEL = $model;	879	$MODEL = $model;
…		…
464	}	883	}
465	}	884	}
466		885
467	# check for already loaded models	886	# check for already loaded models
468	unless ($MODEL) {	887	unless ($MODEL) {
469	for (@REGISTRY, @models, @models_detect) {	888	for (@REGISTRY, @models) {
470	my ($package, $model) = @$_;	889	my ($package, $model) = @$_;
471	if (${"$package\::VERSION"} > 0) {	890	if (${"$package\::VERSION"} > 0) {
472	if (eval "require $model") {	891	if (eval "require $model") {
473	$MODEL = $model;	892	$MODEL = $model;
474	warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1;	893	warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1;
…		…
490	last;	909	last;
491	}	910	}
492	}	911	}
493		912
494	$MODEL	913	$MODEL
495	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) or Glib.";	914	or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib.";
496	}	915	}
497	}	916	}
498		917
499	unshift @ISA, $MODEL;	918	unshift @ISA, $MODEL;
500	push @{"$MODEL\::ISA"}, "AnyEvent::Base";	919	push @{"$MODEL\::ISA"}, "AnyEvent::Base";
		920
		921	(shift @post_detect)->() while @post_detect;
501	}	922	}
502		923
503	$MODEL	924	$MODEL
504	}	925	}
505		926
…		…
515	$class->$func (@_);	936	$class->$func (@_);
516	}	937	}
517		938
518	package AnyEvent::Base;	939	package AnyEvent::Base;
519		940
		941	# default implementation for now and time
		942
		943	use Time::HiRes ();
		944
		945	sub time { Time::HiRes::time }
		946	sub now { Time::HiRes::time }
		947
520	# default implementation for ->condvar, ->wait, ->broadcast	948	# default implementation for ->condvar
521		949
522	sub condvar {	950	sub condvar {
523	bless \my $flag, "AnyEvent::Base::CondVar"	951	bless { @_ == 3 ? (_ae_cb => $_[2]) : () }, AnyEvent::CondVar::
524	}
525
526	sub AnyEvent::Base::CondVar::broadcast {
527	${$_[0]}++;
528	}
529
530	sub AnyEvent::Base::CondVar::wait {
531	AnyEvent->one_event while !${$_[0]};
532	}	952	}
533		953
534	# default implementation for ->signal	954	# default implementation for ->signal
535		955
536	our %SIG_CB;	956	our %SIG_CB;
…		…
589	or Carp::croak "required option 'pid' is missing";	1009	or Carp::croak "required option 'pid' is missing";
590		1010
591	$PID_CB{$pid}{$arg{cb}} = $arg{cb};	1011	$PID_CB{$pid}{$arg{cb}} = $arg{cb};
592		1012
593	unless ($WNOHANG) {	1013	unless ($WNOHANG) {
594	$WNOHANG = eval { require POSIX; &POSIX::WNOHANG } || 1;	1014	$WNOHANG = eval { local $SIG{__DIE__}; require POSIX; &POSIX::WNOHANG } || 1;
595	}	1015	}
596		1016
597	unless ($CHLD_W) {	1017	unless ($CHLD_W) {
598	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);	1018	$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld);
599	# child could be a zombie already, so make at least one round	1019	# child could be a zombie already, so make at least one round
…		…
609	delete $PID_CB{$pid}{$cb};	1029	delete $PID_CB{$pid}{$cb};
610	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };	1030	delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} };
611		1031
612	undef $CHLD_W unless keys %PID_CB;	1032	undef $CHLD_W unless keys %PID_CB;
613	}	1033	}
		1034
		1035	package AnyEvent::CondVar;
		1036
		1037	our @ISA = AnyEvent::CondVar::Base::;
		1038
		1039	package AnyEvent::CondVar::Base;
		1040
		1041	use overload
		1042	'&{}' => sub { my $self = shift; sub { $self->send (@_) } },
		1043	fallback => 1;
		1044
		1045	sub _send {
		1046	# nop
		1047	}
		1048
		1049	sub send {
		1050	my $cv = shift;
		1051	$cv->{_ae_sent} = [@_];
		1052	(delete $cv->{_ae_cb})->($cv) if $cv->{_ae_cb};
		1053	$cv->_send;
		1054	}
		1055
		1056	sub croak {
		1057	$_[0]{_ae_croak} = $_[1];
		1058	$_[0]->send;
		1059	}
		1060
		1061	sub ready {
		1062	$_[0]{_ae_sent}
		1063	}
		1064
		1065	sub _wait {
		1066	AnyEvent->one_event while !$_[0]{_ae_sent};
		1067	}
		1068
		1069	sub recv {
		1070	$_[0]->_wait;
		1071
		1072	Carp::croak $_[0]{_ae_croak} if $_[0]{_ae_croak};
		1073	wantarray ? @{ $_[0]{_ae_sent} } : $_[0]{_ae_sent}[0]
		1074	}
		1075
		1076	sub cb {
		1077	$_[0]{_ae_cb} = $_[1] if @_ > 1;
		1078	$_[0]{_ae_cb}
		1079	}
		1080
		1081	sub begin {
		1082	++$_[0]{_ae_counter};
		1083	$_[0]{_ae_end_cb} = $_[1] if @_ > 1;
		1084	}
		1085
		1086	sub end {
		1087	return if --$_[0]{_ae_counter};
		1088	&{ $_[0]{_ae_end_cb} || sub { $_[0]->send } };
		1089	}
		1090
		1091	# undocumented/compatibility with pre-3.4
		1092	*broadcast = \&send;
		1093	*wait = \&_wait;
614		1094
615	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE	1095	=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE
616		1096
617	This is an advanced topic that you do not normally need to use AnyEvent in	1097	This is an advanced topic that you do not normally need to use AnyEvent in
618	a module. This section is only of use to event loop authors who want to	1098	a module. This section is only of use to event loop authors who want to
…		…
675	model it chooses.	1155	model it chooses.
676		1156
677	=item C<PERL_ANYEVENT_MODEL>	1157	=item C<PERL_ANYEVENT_MODEL>
678		1158
679	This can be used to specify the event model to be used by AnyEvent, before	1159	This can be used to specify the event model to be used by AnyEvent, before
680	autodetection and -probing kicks in. It must be a string consisting	1160	auto detection and -probing kicks in. It must be a string consisting
681	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended	1161	entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended
682	and the resulting module name is loaded and if the load was successful,	1162	and the resulting module name is loaded and if the load was successful,
683	used as event model. If it fails to load AnyEvent will proceed with	1163	used as event model. If it fails to load AnyEvent will proceed with
684	autodetection and -probing.	1164	auto detection and -probing.
685		1165
686	This functionality might change in future versions.	1166	This functionality might change in future versions.
687		1167
688	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you	1168	For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you
689	could start your program like this:	1169	could start your program like this:
690		1170
691	PERL_ANYEVENT_MODEL=Perl perl ...	1171	PERL_ANYEVENT_MODEL=Perl perl ...
692		1172
		1173	=item C<PERL_ANYEVENT_PROTOCOLS>
		1174
		1175	Used by both L<AnyEvent::DNS> and L<AnyEvent::Socket> to determine preferences
		1176	for IPv4 or IPv6. The default is unspecified (and might change, or be the result
		1177	of auto probing).
		1178
		1179	Must be set to a comma-separated list of protocols or address families,
		1180	current supported: C<ipv4> and C<ipv6>. Only protocols mentioned will be
		1181	used, and preference will be given to protocols mentioned earlier in the
		1182	list.
		1183
		1184	This variable can effectively be used for denial-of-service attacks
		1185	against local programs (e.g. when setuid), although the impact is likely
		1186	small, as the program has to handle connection errors already-
		1187
		1188	Examples: C<PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6> - prefer IPv4 over IPv6,
		1189	but support both and try to use both. C<PERL_ANYEVENT_PROTOCOLS=ipv4>
		1190	- only support IPv4, never try to resolve or contact IPv6
		1191	addresses. C<PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4> support either IPv4 or
		1192	IPv6, but prefer IPv6 over IPv4.
		1193
		1194	=item C<PERL_ANYEVENT_EDNS0>
		1195
		1196	Used by L<AnyEvent::DNS> to decide whether to use the EDNS0 extension
		1197	for DNS. This extension is generally useful to reduce DNS traffic, but
		1198	some (broken) firewalls drop such DNS packets, which is why it is off by
		1199	default.
		1200
		1201	Setting this variable to C<1> will cause L<AnyEvent::DNS> to announce
		1202	EDNS0 in its DNS requests.
		1203
		1204	=item C<PERL_ANYEVENT_MAX_FORKS>
		1205
		1206	The maximum number of child processes that C<AnyEvent::Util::fork_call>
		1207	will create in parallel.
		1208
693	=back	1209	=back
694		1210
695	=head1 EXAMPLE PROGRAM	1211	=head1 EXAMPLE PROGRAM
696		1212
697	The following program uses an IO watcher to read data from STDIN, a timer	1213	The following program uses an I/O watcher to read data from STDIN, a timer
698	to display a message once per second, and a condition variable to quit the	1214	to display a message once per second, and a condition variable to quit the
699	program when the user enters quit:	1215	program when the user enters quit:
700		1216
701	use AnyEvent;	1217	use AnyEvent;
702		1218
…		…
707	poll => 'r',	1223	poll => 'r',
708	cb => sub {	1224	cb => sub {
709	warn "io event <$_[0]>\n"; # will always output <r>	1225	warn "io event <$_[0]>\n"; # will always output <r>
710	chomp (my $input = <STDIN>); # read a line	1226	chomp (my $input = <STDIN>); # read a line
711	warn "read: $input\n"; # output what has been read	1227	warn "read: $input\n"; # output what has been read
712	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	1228	$cv->send if $input =~ /^q/i; # quit program if /^q/i
713	},	1229	},
714	);	1230	);
715		1231
716	my $time_watcher; # can only be used once	1232	my $time_watcher; # can only be used once
717		1233
…		…
722	});	1238	});
723	}	1239	}
724		1240
725	new_timer; # create first timer	1241	new_timer; # create first timer
726		1242
727	$cv->wait; # wait until user enters /^q/i	1243	$cv->recv; # wait until user enters /^q/i
728		1244
729	=head1 REAL-WORLD EXAMPLE	1245	=head1 REAL-WORLD EXAMPLE
730		1246
731	Consider the L<Net::FCP> module. It features (among others) the following	1247	Consider the L<Net::FCP> module. It features (among others) the following
732	API calls, which are to freenet what HTTP GET requests are to http:	1248	API calls, which are to freenet what HTTP GET requests are to http:
…		…
782	syswrite $txn->{fh}, $txn->{request}	1298	syswrite $txn->{fh}, $txn->{request}
783	or die "connection or write error";	1299	or die "connection or write error";
784	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	1300	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
785		1301
786	Again, C<fh_ready_r> waits till all data has arrived, and then stores the	1302	Again, C<fh_ready_r> waits till all data has arrived, and then stores the
787	result and signals any possible waiters that the request ahs finished:	1303	result and signals any possible waiters that the request has finished:
788		1304
789	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	1305	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
790		1306
791	if (end-of-file or data complete) {	1307	if (end-of-file or data complete) {
792	$txn->{result} = $txn->{buf};	1308	$txn->{result} = $txn->{buf};
793	$txn->{finished}->broadcast;	1309	$txn->{finished}->send;
794	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	1310	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
795	}	1311	}
796		1312
797	The C<result> method, finally, just waits for the finished signal (if the	1313	The C<result> method, finally, just waits for the finished signal (if the
798	request was already finished, it doesn't wait, of course, and returns the	1314	request was already finished, it doesn't wait, of course, and returns the
799	data:	1315	data:
800		1316
801	$txn->{finished}->wait;	1317	$txn->{finished}->recv;
802	return $txn->{result};	1318	return $txn->{result};
803		1319
804	The actual code goes further and collects all errors (C<die>s, exceptions)	1320	The actual code goes further and collects all errors (C<die>s, exceptions)
805	that occured during request processing. The C<result> method detects	1321	that occurred during request processing. The C<result> method detects
806	whether an exception as thrown (it is stored inside the $txn object)	1322	whether an exception as thrown (it is stored inside the $txn object)
807	and just throws the exception, which means connection errors and other	1323	and just throws the exception, which means connection errors and other
808	problems get reported tot he code that tries to use the result, not in a	1324	problems get reported tot he code that tries to use the result, not in a
809	random callback.	1325	random callback.
810		1326
…		…
841		1357
842	my $quit = AnyEvent->condvar;	1358	my $quit = AnyEvent->condvar;
843		1359
844	$fcp->txn_client_get ($url)->cb (sub {	1360	$fcp->txn_client_get ($url)->cb (sub {
845	...	1361	...
846	$quit->broadcast;	1362	$quit->send;
847	});	1363	});
848		1364
849	$quit->wait;	1365	$quit->recv;
		1366
		1367
		1368	=head1 BENCHMARKS
		1369
		1370	To give you an idea of the performance and overheads that AnyEvent adds
		1371	over the event loops themselves and to give you an impression of the speed
		1372	of various event loops I prepared some benchmarks.
		1373
		1374	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1375
		1376	Here is a benchmark of various supported event models used natively and
		1377	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1378	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1379	which it is), lets them fire exactly once and destroys them again.
		1380
		1381	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
		1382	distribution.
		1383
		1384	=head3 Explanation of the columns
		1385
		1386	I<watcher> is the number of event watchers created/destroyed. Since
		1387	different event models feature vastly different performances, each event
		1388	loop was given a number of watchers so that overall runtime is acceptable
		1389	and similar between tested event loop (and keep them from crashing): Glib
		1390	would probably take thousands of years if asked to process the same number
		1391	of watchers as EV in this benchmark.
		1392
		1393	I<bytes> is the number of bytes (as measured by the resident set size,
		1394	RSS) consumed by each watcher. This method of measuring captures both C
		1395	and Perl-based overheads.
		1396
		1397	I<create> is the time, in microseconds (millionths of seconds), that it
		1398	takes to create a single watcher. The callback is a closure shared between
		1399	all watchers, to avoid adding memory overhead. That means closure creation
		1400	and memory usage is not included in the figures.
		1401
		1402	I<invoke> is the time, in microseconds, used to invoke a simple
		1403	callback. The callback simply counts down a Perl variable and after it was
		1404	invoked "watcher" times, it would C<< ->send >> a condvar once to
		1405	signal the end of this phase.
		1406
		1407	I<destroy> is the time, in microseconds, that it takes to destroy a single
		1408	watcher.
		1409
		1410	=head3 Results
		1411
		1412	name watchers bytes create invoke destroy comment
		1413	EV/EV 400000 244 0.56 0.46 0.31 EV native interface
		1414	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
		1415	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
		1416	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
		1417	Event/Event 16000 516 31.88 31.30 0.85 Event native interface
		1418	Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
		1419	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
		1420	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
		1421	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
		1422	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
		1423
		1424	=head3 Discussion
		1425
		1426	The benchmark does I<not> measure scalability of the event loop very
		1427	well. For example, a select-based event loop (such as the pure perl one)
		1428	can never compete with an event loop that uses epoll when the number of
		1429	file descriptors grows high. In this benchmark, all events become ready at
		1430	the same time, so select/poll-based implementations get an unnatural speed
		1431	boost.
		1432
		1433	Also, note that the number of watchers usually has a nonlinear effect on
		1434	overall speed, that is, creating twice as many watchers doesn't take twice
		1435	the time - usually it takes longer. This puts event loops tested with a
		1436	higher number of watchers at a disadvantage.
		1437
		1438	To put the range of results into perspective, consider that on the
		1439	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1440	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1441	cycles with POE.
		1442
		1443	C<EV> is the sole leader regarding speed and memory use, which are both
		1444	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1445	far less memory than any other event loop and is still faster than Event
		1446	natively.
		1447
		1448	The pure perl implementation is hit in a few sweet spots (both the
		1449	constant timeout and the use of a single fd hit optimisations in the perl
		1450	interpreter and the backend itself). Nevertheless this shows that it
		1451	adds very little overhead in itself. Like any select-based backend its
		1452	performance becomes really bad with lots of file descriptors (and few of
		1453	them active), of course, but this was not subject of this benchmark.
		1454
		1455	The C<Event> module has a relatively high setup and callback invocation
		1456	cost, but overall scores in on the third place.
		1457
		1458	C<Glib>'s memory usage is quite a bit higher, but it features a
		1459	faster callback invocation and overall ends up in the same class as
		1460	C<Event>. However, Glib scales extremely badly, doubling the number of
		1461	watchers increases the processing time by more than a factor of four,
		1462	making it completely unusable when using larger numbers of watchers
		1463	(note that only a single file descriptor was used in the benchmark, so
		1464	inefficiencies of C<poll> do not account for this).
		1465
		1466	The C<Tk> adaptor works relatively well. The fact that it crashes with
		1467	more than 2000 watchers is a big setback, however, as correctness takes
		1468	precedence over speed. Nevertheless, its performance is surprising, as the
		1469	file descriptor is dup()ed for each watcher. This shows that the dup()
		1470	employed by some adaptors is not a big performance issue (it does incur a
		1471	hidden memory cost inside the kernel which is not reflected in the figures
		1472	above).
		1473
		1474	C<POE>, regardless of underlying event loop (whether using its pure perl
		1475	select-based backend or the Event module, the POE-EV backend couldn't
		1476	be tested because it wasn't working) shows abysmal performance and
		1477	memory usage with AnyEvent: Watchers use almost 30 times as much memory
		1478	as EV watchers, and 10 times as much memory as Event (the high memory
		1479	requirements are caused by requiring a session for each watcher). Watcher
		1480	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		1481	implementation.
		1482
		1483	The design of the POE adaptor class in AnyEvent can not really account
		1484	for the performance issues, though, as session creation overhead is
		1485	small compared to execution of the state machine, which is coded pretty
		1486	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		1487	using multiple sessions is not a good approach, especially regarding
		1488	memory usage, even the author of POE could not come up with a faster
		1489	design).
		1490
		1491	=head3 Summary
		1492
		1493	=over 4
		1494
		1495	=item * Using EV through AnyEvent is faster than any other event loop
		1496	(even when used without AnyEvent), but most event loops have acceptable
		1497	performance with or without AnyEvent.
		1498
		1499	=item * The overhead AnyEvent adds is usually much smaller than the overhead of
		1500	the actual event loop, only with extremely fast event loops such as EV
		1501	adds AnyEvent significant overhead.
		1502
		1503	=item * You should avoid POE like the plague if you want performance or
		1504	reasonable memory usage.
		1505
		1506	=back
		1507
		1508	=head2 BENCHMARKING THE LARGE SERVER CASE
		1509
		1510	This benchmark actually benchmarks the event loop itself. It works by
		1511	creating a number of "servers": each server consists of a socket pair, a
		1512	timeout watcher that gets reset on activity (but never fires), and an I/O
		1513	watcher waiting for input on one side of the socket. Each time the socket
		1514	watcher reads a byte it will write that byte to a random other "server".
		1515
		1516	The effect is that there will be a lot of I/O watchers, only part of which
		1517	are active at any one point (so there is a constant number of active
		1518	fds for each loop iteration, but which fds these are is random). The
		1519	timeout is reset each time something is read because that reflects how
		1520	most timeouts work (and puts extra pressure on the event loops).
		1521
		1522	In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100
		1523	(1%) are active. This mirrors the activity of large servers with many
		1524	connections, most of which are idle at any one point in time.
		1525
		1526	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1527	distribution.
		1528
		1529	=head3 Explanation of the columns
		1530
		1531	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1532	each server has a read and write socket end).
		1533
		1534	I<create> is the time it takes to create a socket pair (which is
		1535	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1536
		1537	I<request>, the most important value, is the time it takes to handle a
		1538	single "request", that is, reading the token from the pipe and forwarding
		1539	it to another server. This includes deleting the old timeout and creating
		1540	a new one that moves the timeout into the future.
		1541
		1542	=head3 Results
		1543
		1544	name sockets create request
		1545	EV 20000 69.01 11.16
		1546	Perl 20000 73.32 35.87
		1547	Event 20000 212.62 257.32
		1548	Glib 20000 651.16 1896.30
		1549	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1550
		1551	=head3 Discussion
		1552
		1553	This benchmark I<does> measure scalability and overall performance of the
		1554	particular event loop.
		1555
		1556	EV is again fastest. Since it is using epoll on my system, the setup time
		1557	is relatively high, though.
		1558
		1559	Perl surprisingly comes second. It is much faster than the C-based event
		1560	loops Event and Glib.
		1561
		1562	Event suffers from high setup time as well (look at its code and you will
		1563	understand why). Callback invocation also has a high overhead compared to
		1564	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1565	uses select or poll in basically all documented configurations.
		1566
		1567	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1568	clearly fails to perform with many filehandles or in busy servers.
		1569
		1570	POE is still completely out of the picture, taking over 1000 times as long
		1571	as EV, and over 100 times as long as the Perl implementation, even though
		1572	it uses a C-based event loop in this case.
		1573
		1574	=head3 Summary
		1575
		1576	=over 4
		1577
		1578	=item * The pure perl implementation performs extremely well.
		1579
		1580	=item * Avoid Glib or POE in large projects where performance matters.
		1581
		1582	=back
		1583
		1584	=head2 BENCHMARKING SMALL SERVERS
		1585
		1586	While event loops should scale (and select-based ones do not...) even to
		1587	large servers, most programs we (or I :) actually write have only a few
		1588	I/O watchers.
		1589
		1590	In this benchmark, I use the same benchmark program as in the large server
		1591	case, but it uses only eight "servers", of which three are active at any
		1592	one time. This should reflect performance for a small server relatively
		1593	well.
		1594
		1595	The columns are identical to the previous table.
		1596
		1597	=head3 Results
		1598
		1599	name sockets create request
		1600	EV 16 20.00 6.54
		1601	Perl 16 25.75 12.62
		1602	Event 16 81.27 35.86
		1603	Glib 16 32.63 15.48
		1604	POE 16 261.87 276.28 uses POE::Loop::Event
		1605
		1606	=head3 Discussion
		1607
		1608	The benchmark tries to test the performance of a typical small
		1609	server. While knowing how various event loops perform is interesting, keep
		1610	in mind that their overhead in this case is usually not as important, due
		1611	to the small absolute number of watchers (that is, you need efficiency and
		1612	speed most when you have lots of watchers, not when you only have a few of
		1613	them).
		1614
		1615	EV is again fastest.
		1616
		1617	Perl again comes second. It is noticeably faster than the C-based event
		1618	loops Event and Glib, although the difference is too small to really
		1619	matter.
		1620
		1621	POE also performs much better in this case, but is is still far behind the
		1622	others.
		1623
		1624	=head3 Summary
		1625
		1626	=over 4
		1627
		1628	=item * C-based event loops perform very well with small number of
		1629	watchers, as the management overhead dominates.
		1630
		1631	=back
		1632
850		1633
851	=head1 FORK	1634	=head1 FORK
852		1635
853	Most event libraries are not fork-safe. The ones who are usually are	1636	Most event libraries are not fork-safe. The ones who are usually are
854	because they are so inefficient. Only L<EV> is fully fork-aware.	1637	because they rely on inefficient but fork-safe C<select> or C<poll>
		1638	calls. Only L<EV> is fully fork-aware.
855		1639
856	If you have to fork, you must either do so I<before> creating your first	1640	If you have to fork, you must either do so I<before> creating your first
857	watcher OR you must not use AnyEvent at all in the child.	1641	watcher OR you must not use AnyEvent at all in the child.
		1642
858		1643
859	=head1 SECURITY CONSIDERATIONS	1644	=head1 SECURITY CONSIDERATIONS
860		1645
861	AnyEvent can be forced to load any event model via	1646	AnyEvent can be forced to load any event model via
862	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to	1647	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to
…		…
870		1655
871	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }	1656	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
872		1657
873	use AnyEvent;	1658	use AnyEvent;
874		1659
		1660	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		1661	be used to probe what backend is used and gain other information (which is
		1662	probably even less useful to an attacker than PERL_ANYEVENT_MODEL).
		1663
		1664
875	=head1 SEE ALSO	1665	=head1 SEE ALSO
876		1666
877	Event modules: L<Coro::EV>, L<EV>, L<EV::Glib>, L<Glib::EV>,	1667	Utility functions: L<AnyEvent::Util>.
878	L<Coro::Event>, L<Event>, L<Glib::Event>, L<Glib>, L<Coro>, L<Tk>,
879	L<Event::Lib>, L<Qt>.
880		1668
881	Implementations: L<AnyEvent::Impl::CoroEV>, L<AnyEvent::Impl::EV>,	1669	Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>,
882	L<AnyEvent::Impl::CoroEvent>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>,	1670	L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>.
883	L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, L<AnyEvent::Impl::EventLib>,	1671
		1672	Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>,
		1673	L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>,
		1674	L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>,
884	L<AnyEvent::Impl::Qt>.	1675	L<AnyEvent::Impl::POE>.
885		1676
		1677	Non-blocking file handles, sockets, TCP clients and
		1678	servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>.
		1679
		1680	Asynchronous DNS: L<AnyEvent::DNS>.
		1681
		1682	Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>,
		1683
886	Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>.	1684	Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>.
		1685
887		1686
888	=head1 AUTHOR	1687	=head1 AUTHOR
889		1688
890	Marc Lehmann <schmorp@schmorp.de>	1689	Marc Lehmann <schmorp@schmorp.de>
891	http://home.schmorp.de/	1690	http://home.schmorp.de/

Diff Legend

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