[ViewVC] Diff of: cvs/AnyEvent/README

Comparing AnyEvent/README (file contents):
Revision 1.16 by root, Sat Apr 19 04:58:14 2008 UTC vs.
Revision 1.25 by root, Tue Jun 3 09:02:46 2008 UTC

…		…
1	NAME	1	NAME
2	AnyEvent - provide framework for multiple event loops	2	AnyEvent - provide framework for multiple event loops
3		3
4	EV, Event, Coro::EV, Coro::Event, Glib, Tk, Perl - various supported	4	EV, Event, Glib, Tk, Perl, Event::Lib, Qt, POE - various supported event
5	event loops	5	loops
6		6
7	SYNOPSIS	7	SYNOPSIS
8	use AnyEvent;	8	use AnyEvent;
9		9
10	my $w = AnyEvent->io (fh => $fh, poll => "r\|w", cb => sub {	10	my $w = AnyEvent->io (fh => $fh, poll => "r\|w", cb => sub {
14	my $w = AnyEvent->timer (after => $seconds, cb => sub {	14	my $w = AnyEvent->timer (after => $seconds, cb => sub {
15	...	15	...
16	});	16	});
17		17
18	my $w = AnyEvent->condvar; # stores whether a condition was flagged	18	my $w = AnyEvent->condvar; # stores whether a condition was flagged
		19	$w->send; # wake up current and all future recv's
19	$w->wait; # enters "main loop" till $condvar gets ->broadcast	20	$w->recv; # enters "main loop" till $condvar gets ->send
20	$w->broadcast; # wake up current and all future wait's	21
		22	INTRODUCTION/TUTORIAL
		23	This manpage is mainly a reference manual. If you are interested in a
		24	tutorial or some gentle introduction, have a look at the AnyEvent::Intro
		25	manpage.
21		26
22	WHY YOU SHOULD USE THIS MODULE (OR NOT)	27	WHY YOU SHOULD USE THIS MODULE (OR NOT)
23	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen	28	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
24	nowadays. So what is different about AnyEvent?	29	nowadays. So what is different about AnyEvent?
25		30
…		…
46	that isn't itself. What's worse, all the potential users of your module	51	that isn't itself. What's worse, all the potential users of your module
47	are also forced to use the same event loop you use.	52	are also forced to use the same event loop you use.
48		53
49	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works	54	AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
50	fine. AnyEvent + Tk works fine etc. etc. but none of these work together	55	fine. AnyEvent + Tk works fine etc. etc. but none of these work together
51	with the rest: POE + IO::Async? no go. Tk + Event? no go. Again: if your	56	with the rest: POE + IO::Async? No go. Tk + Event? No go. Again: if your
52	module uses one of those, every user of your module has to use it, too.	57	module uses one of those, every user of your module has to use it, too.
53	But if your module uses AnyEvent, it works transparently with all event	58	But if your module uses AnyEvent, it works transparently with all event
54	models it supports (including stuff like POE and IO::Async, as long as	59	models it supports (including stuff like POE and IO::Async, as long as
55	those use one of the supported event loops. It is trivial to add new	60	those use one of the supported event loops. It is trivial to add new
56	event loops to AnyEvent, too, so it is future-proof).	61	event loops to AnyEvent, too, so it is future-proof).
57		62
58	In addition to being free of having to use *the one and only true event	63	In addition to being free of having to use *the one and only true event
59	model*, AnyEvent also is free of bloat and policy: with POE or similar	64	model*, AnyEvent also is free of bloat and policy: with POE or similar
60	modules, you get an enourmous amount of code and strict rules you have	65	modules, you get an enormous amount of code and strict rules you have to
61	to follow. AnyEvent, on the other hand, is lean and up to the point, by	66	follow. AnyEvent, on the other hand, is lean and up to the point, by
62	only offering the functionality that is necessary, in as thin as a	67	only offering the functionality that is necessary, in as thin as a
63	wrapper as technically possible.	68	wrapper as technically possible.
64		69
		70	Of course, AnyEvent comes with a big (and fully optional!) toolbox of
		71	useful functionality, such as an asynchronous DNS resolver, 100%
		72	non-blocking connects (even with TLS/SSL, IPv6 and on broken platforms
		73	such as Windows) and lots of real-world knowledge and workarounds for
		74	platform bugs and differences.
		75
65	Of course, if you want lots of policy (this can arguably be somewhat	76	Now, if you do want lots of policy (this can arguably be somewhat
66	useful) and you want to force your users to use the one and only event	77	useful) and you want to force your users to use the one and only event
67	model, you should not use this module.	78	model, you should not use this module.
68		79
69	DESCRIPTION	80	DESCRIPTION
70	AnyEvent provides an identical interface to multiple event loops. This	81	AnyEvent provides an identical interface to multiple event loops. This
…		…
75	The interface itself is vaguely similar, but not identical to the Event	86	The interface itself is vaguely similar, but not identical to the Event
76	module.	87	module.
77		88
78	During the first call of any watcher-creation method, the module tries	89	During the first call of any watcher-creation method, the module tries
79	to detect the currently loaded event loop by probing whether one of the	90	to detect the currently loaded event loop by probing whether one of the
80	following modules is already loaded: Coro::EV, Coro::Event, EV, Event,	91	following modules is already loaded: EV, Event, Glib,
81	Glib, Tk. The first one found is used. If none are found, the module	92	AnyEvent::Impl::Perl, Tk, Event::Lib, Qt, POE. The first one found is
82	tries to load these modules in the stated order. The first one that can	93	used. If none are found, the module tries to load these modules
		94	(excluding Tk, Event::Lib, Qt and POE as the pure perl adaptor should
		95	always succeed) in the order given. The first one that can be
83	be successfully loaded will be used. If, after this, still none could be	96	successfully loaded will be used. If, after this, still none could be
84	found, AnyEvent will fall back to a pure-perl event loop, which is not	97	found, AnyEvent will fall back to a pure-perl event loop, which is not
85	very efficient, but should work everywhere.	98	very efficient, but should work everywhere.
86		99
87	Because AnyEvent first checks for modules that are already loaded,	100	Because AnyEvent first checks for modules that are already loaded,
88	loading an event model explicitly before first using AnyEvent will	101	loading an event model explicitly before first using AnyEvent will
…		…
97	starts using it, all bets are off. Maybe you should tell their authors	110	starts using it, all bets are off. Maybe you should tell their authors
98	to use AnyEvent so their modules work together with others seamlessly...	111	to use AnyEvent so their modules work together with others seamlessly...
99		112
100	The pure-perl implementation of AnyEvent is called	113	The pure-perl implementation of AnyEvent is called
101	"AnyEvent::Impl::Perl". Like other event modules you can load it	114	"AnyEvent::Impl::Perl". Like other event modules you can load it
102	explicitly.	115	explicitly and enjoy the high availability of that event loop :)
103		116
104	WATCHERS	117	WATCHERS
105	AnyEvent has the central concept of a watcher, which is an object that	118	AnyEvent has the central concept of a watcher, which is an object that
106	stores relevant data for each kind of event you are waiting for, such as	119	stores relevant data for each kind of event you are waiting for, such as
107	the callback to call, the filehandle to watch, etc.	120	the callback to call, the file handle to watch, etc.
108		121
109	These watchers are normal Perl objects with normal Perl lifetime. After	122	These watchers are normal Perl objects with normal Perl lifetime. After
110	creating a watcher it will immediately "watch" for events and invoke the	123	creating a watcher it will immediately "watch" for events and invoke the
111	callback when the event occurs (of course, only when the event model is	124	callback when the event occurs (of course, only when the event model is
112	in control).	125	in control).
…		…
120	Many watchers either are used with "recursion" (repeating timers for	133	Many watchers either are used with "recursion" (repeating timers for
121	example), or need to refer to their watcher object in other ways.	134	example), or need to refer to their watcher object in other ways.
122		135
123	An any way to achieve that is this pattern:	136	An any way to achieve that is this pattern:
124		137
125	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {	138	my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
126	# you can use $w here, for example to undef it	139	# you can use $w here, for example to undef it
127	undef $w;	140	undef $w;
128	});	141	});
129		142
130	Note that "my $w; $w =" combination. This is necessary because in Perl,	143	Note that "my $w; $w =" combination. This is necessary because in Perl,
131	my variables are only visible after the statement in which they are	144	my variables are only visible after the statement in which they are
132	declared.	145	declared.
133		146
134	IO WATCHERS	147	I/O WATCHERS
135	You can create an I/O watcher by calling the "AnyEvent->io" method with	148	You can create an I/O watcher by calling the "AnyEvent->io" method with
136	the following mandatory key-value pairs as arguments:	149	the following mandatory key-value pairs as arguments:
137		150
138	"fh" the Perl file handle (not file descriptor) to watch for events.	151	"fh" the Perl file handle (not file descriptor) to watch for events.
139	"poll" must be a string that is either "r" or "w", which creates a	152	"poll" must be a string that is either "r" or "w", which creates a
140	watcher waiting for "r"eadable or "w"ritable events, respectively. "cb"	153	watcher waiting for "r"eadable or "w"ritable events, respectively. "cb"
141	is the callback to invoke each time the file handle becomes ready.	154	is the callback to invoke each time the file handle becomes ready.
142		155
143	File handles will be kept alive, so as long as the watcher exists, the	156	Although the callback might get passed parameters, their value and
144	file handle exists, too.	157	presence is undefined and you cannot rely on them. Portable AnyEvent
		158	callbacks cannot use arguments passed to I/O watcher callbacks.
145		159
		160	The I/O watcher might use the underlying file descriptor or a copy of
146	It is not allowed to close a file handle as long as any watcher is	161	it. You must not close a file handle as long as any watcher is active on
147	active on the underlying file descriptor.	162	the underlying file descriptor.
148		163
149	Some event loops issue spurious readyness notifications, so you should	164	Some event loops issue spurious readyness notifications, so you should
150	always use non-blocking calls when reading/writing from/to your file	165	always use non-blocking calls when reading/writing from/to your file
151	handles.	166	handles.
152		167
…		…
162	TIME WATCHERS	177	TIME WATCHERS
163	You can create a time watcher by calling the "AnyEvent->timer" method	178	You can create a time watcher by calling the "AnyEvent->timer" method
164	with the following mandatory arguments:	179	with the following mandatory arguments:
165		180
166	"after" specifies after how many seconds (fractional values are	181	"after" specifies after how many seconds (fractional values are
167	supported) should the timer activate. "cb" the callback to invoke in	182	supported) the callback should be invoked. "cb" is the callback to
168	that case.	183	invoke in that case.
		184
		185	Although the callback might get passed parameters, their value and
		186	presence is undefined and you cannot rely on them. Portable AnyEvent
		187	callbacks cannot use arguments passed to time watcher callbacks.
169		188
170	The timer callback will be invoked at most once: if you want a repeating	189	The timer callback will be invoked at most once: if you want a repeating
171	timer you have to create a new watcher (this is a limitation by both Tk	190	timer you have to create a new watcher (this is a limitation by both Tk
172	and Glib).	191	and Glib).
173		192
…		…
200	o'clock").	219	o'clock").
201		220
202	While most event loops expect timers to specified in a relative way,	221	While most event loops expect timers to specified in a relative way,
203	they use absolute time internally. This makes a difference when your	222	they use absolute time internally. This makes a difference when your
204	clock "jumps", for example, when ntp decides to set your clock backwards	223	clock "jumps", for example, when ntp decides to set your clock backwards
205	from the wrong 2014-01-01 to 2008-01-01, a watcher that you created to	224	from the wrong date of 2014-01-01 to 2008-01-01, a watcher that is
206	fire "after" a second might actually take six years to finally fire.	225	supposed to fire "after" a second might actually take six years to
		226	finally fire.
207		227
208	AnyEvent cannot compensate for this. The only event loop that is	228	AnyEvent cannot compensate for this. The only event loop that is
209	conscious about these issues is EV, which offers both relative	229	conscious about these issues is EV, which offers both relative
210	(ev_timer) and absolute (ev_periodic) timers.	230	(ev_timer, based on true relative time) and absolute (ev_periodic, based
		231	on wallclock time) timers.
211		232
212	AnyEvent always prefers relative timers, if available, matching the	233	AnyEvent always prefers relative timers, if available, matching the
213	AnyEvent API.	234	AnyEvent API.
		235
		236	AnyEvent has two additional methods that return the "current time":
		237
		238	AnyEvent->time
		239	This returns the "current wallclock time" as a fractional number of
		240	seconds since the Epoch (the same thing as "time" or
		241	"Time::HiRes::time" return, and the result is guaranteed to be
		242	compatible with those).
		243
		244	It progresses independently of any event loop processing, i.e. each
		245	call will check the system clock, which usually gets updated
		246	frequently.
		247
		248	AnyEvent->now
		249	This also returns the "current wallclock time", but unlike "time",
		250	above, this value might change only once per event loop iteration,
		251	depending on the event loop (most return the same time as "time",
		252	above). This is the time that AnyEvent's timers get scheduled
		253	against.
		254
		255	*In almost all cases (in all cases if you don't care), this is the
		256	function to call when you want to know the current time.*
		257
		258	This function is also often faster then "AnyEvent->time", and thus
		259	the preferred method if you want some timestamp (for example,
		260	AnyEvent::Handle uses this to update it's activity timeouts).
		261
		262	The rest of this section is only of relevance if you try to be very
		263	exact with your timing, you can skip it without bad conscience.
		264
		265	For a practical example of when these times differ, consider
		266	Event::Lib and EV and the following set-up:
		267
		268	The event loop is running and has just invoked one of your callback
		269	at time=500 (assume no other callbacks delay processing). In your
		270	callback, you wait a second by executing "sleep 1" (blocking the
		271	process for a second) and then (at time=501) you create a relative
		272	timer that fires after three seconds.
		273
		274	With Event::Lib, "AnyEvent->time" and "AnyEvent->now" will both
		275	return 501, because that is the current time, and the timer will be
		276	scheduled to fire at time=504 (501 + 3).
		277
		278	With EV, "AnyEvent->time" returns 501 (as that is the current time),
		279	but "AnyEvent->now" returns 500, as that is the time the last event
		280	processing phase started. With EV, your timer gets scheduled to run
		281	at time=503 (500 + 3).
		282
		283	In one sense, Event::Lib is more exact, as it uses the current time
		284	regardless of any delays introduced by event processing. However,
		285	most callbacks do not expect large delays in processing, so this
		286	causes a higher drift (and a lot more system calls to get the
		287	current time).
		288
		289	In another sense, EV is more exact, as your timer will be scheduled
		290	at the same time, regardless of how long event processing actually
		291	took.
		292
		293	In either case, if you care (and in most cases, you don't), then you
		294	can get whatever behaviour you want with any event loop, by taking
		295	the difference between "AnyEvent->time" and "AnyEvent->now" into
		296	account.
214		297
215	SIGNAL WATCHERS	298	SIGNAL WATCHERS
216	You can watch for signals using a signal watcher, "signal" is the signal	299	You can watch for signals using a signal watcher, "signal" is the signal
217	name without any "SIG" prefix, "cb" is the Perl callback to be invoked	300	name without any "SIG" prefix, "cb" is the Perl callback to be invoked
218	whenever a signal occurs.	301	whenever a signal occurs.
219		302
		303	Although the callback might get passed parameters, their value and
		304	presence is undefined and you cannot rely on them. Portable AnyEvent
		305	callbacks cannot use arguments passed to signal watcher callbacks.
		306
220	Multiple signals occurances can be clumped together into one callback	307	Multiple signal occurrences can be clumped together into one callback
221	invocation, and callback invocation will be synchronous. synchronous	308	invocation, and callback invocation will be synchronous. Synchronous
222	means that it might take a while until the signal gets handled by the	309	means that it might take a while until the signal gets handled by the
223	process, but it is guarenteed not to interrupt any other callbacks.	310	process, but it is guaranteed not to interrupt any other callbacks.
224		311
225	The main advantage of using these watchers is that you can share a	312	The main advantage of using these watchers is that you can share a
226	signal between multiple watchers.	313	signal between multiple watchers.
227		314
228	This watcher might use %SIG, so programs overwriting those signals	315	This watcher might use %SIG, so programs overwriting those signals
…		…
237		324
238	The child process is specified by the "pid" argument (if set to 0, it	325	The child process is specified by the "pid" argument (if set to 0, it
239	watches for any child process exit). The watcher will trigger as often	326	watches for any child process exit). The watcher will trigger as often
240	as status change for the child are received. This works by installing a	327	as status change for the child are received. This works by installing a
241	signal handler for "SIGCHLD". The callback will be called with the pid	328	signal handler for "SIGCHLD". The callback will be called with the pid
242	and exit status (as returned by waitpid).	329	and exit status (as returned by waitpid), so unlike other watcher types,
		330	you can rely on child watcher callback arguments.
243		331
244	Example: wait for pid 1333	332	There is a slight catch to child watchers, however: you usually start
		333	them after the child process was created, and this means the process
		334	could have exited already (and no SIGCHLD will be sent anymore).
245		335
		336	Not all event models handle this correctly (POE doesn't), but even for
		337	event models that do handle this correctly, they usually need to be
		338	loaded before the process exits (i.e. before you fork in the first
		339	place).
		340
		341	This means you cannot create a child watcher as the very first thing in
		342	an AnyEvent program, you have to create at least one watcher before
		343	you "fork" the child (alternatively, you can call "AnyEvent::detect").
		344
		345	Example: fork a process and wait for it
		346
		347	my $done = AnyEvent->condvar;
		348
		349	my $pid = fork or exit 5;
		350
246	my $w = AnyEvent->child (	351	my $w = AnyEvent->child (
247	pid => 1333,	352	pid => $pid,
248	cb => sub {	353	cb => sub {
249	my ($pid, $status) = @_;	354	my ($pid, $status) = @_;
250	warn "pid $pid exited with status $status";	355	warn "pid $pid exited with status $status";
		356	$done->send;
251	},	357	},
252	);	358	);
		359
		360	# do something else, then wait for process exit
		361	$done->recv;
253		362
254	CONDITION VARIABLES	363	CONDITION VARIABLES
		364	If you are familiar with some event loops you will know that all of them
		365	require you to run some blocking "loop", "run" or similar function that
		366	will actively watch for new events and call your callbacks.
		367
		368	AnyEvent is different, it expects somebody else to run the event loop
		369	and will only block when necessary (usually when told by the user).
		370
		371	The instrument to do that is called a "condition variable", so called
		372	because they represent a condition that must become true.
		373
255	Condition variables can be created by calling the "AnyEvent->condvar"	374	Condition variables can be created by calling the "AnyEvent->condvar"
256	method without any arguments.	375	method, usually without arguments. The only argument pair allowed is
		376	"cb", which specifies a callback to be called when the condition
		377	variable becomes true.
257		378
258	A condition variable waits for a condition - precisely that the	379	After creation, the condition variable is "false" until it becomes
259	"->broadcast" method has been called.	380	"true" by calling the "send" method (or calling the condition variable
		381	as if it were a callback, read about the caveats in the description for
		382	the "->send" method).
260		383
261	They are very useful to signal that a condition has been fulfilled, for	384	Condition variables are similar to callbacks, except that you can
		385	optionally wait for them. They can also be called merge points - points
		386	in time where multiple outstanding events have been processed. And yet
		387	another way to call them is transactions - each condition variable can
		388	be used to represent a transaction, which finishes at some point and
		389	delivers a result.
		390
		391	Condition variables are very useful to signal that something has
262	example, if you write a module that does asynchronous http requests,	392	finished, for example, if you write a module that does asynchronous http
263	then a condition variable would be the ideal candidate to signal the	393	requests, then a condition variable would be the ideal candidate to
264	availability of results.	394	signal the availability of results. The user can either act when the
		395	callback is called or can synchronously "->recv" for the results.
265		396
266	You can also use condition variables to block your main program until an	397	You can also use them to simulate traditional event loops - for example,
267	event occurs - for example, you could "->wait" in your main program	398	you can block your main program until an event occurs - for example, you
268	until the user clicks the Quit button in your app, which would	399	could "->recv" in your main program until the user clicks the Quit
269	"->broadcast" the "quit" event.	400	button of your app, which would "->send" the "quit" event.
270		401
271	Note that condition variables recurse into the event loop - if you have	402	Note that condition variables recurse into the event loop - if you have
272	two pirces of code that call "->wait" in a round-robbin fashion, you	403	two pieces of code that call "->recv" in a round-robin fashion, you
273	lose. Therefore, condition variables are good to export to your caller,	404	lose. Therefore, condition variables are good to export to your caller,
274	but you should avoid making a blocking wait yourself, at least in	405	but you should avoid making a blocking wait yourself, at least in
275	callbacks, as this asks for trouble.	406	callbacks, as this asks for trouble.
276		407
277	This object has two methods:	408	Condition variables are represented by hash refs in perl, and the keys
		409	used by AnyEvent itself are all named "_ae_XXX" to make subclassing easy
		410	(it is often useful to build your own transaction class on top of
		411	AnyEvent). To subclass, use "AnyEvent::CondVar" as base class and call
		412	it's "new" method in your own "new" method.
278		413
279	$cv->wait	414	There are two "sides" to a condition variable - the "producer side"
		415	which eventually calls "-> send", and the "consumer side", which waits
		416	for the send to occur.
		417
		418	Example: wait for a timer.
		419
		420	# wait till the result is ready
		421	my $result_ready = AnyEvent->condvar;
		422
		423	# do something such as adding a timer
		424	# or socket watcher the calls $result_ready->send
		425	# when the "result" is ready.
		426	# in this case, we simply use a timer:
		427	my $w = AnyEvent->timer (
		428	after => 1,
		429	cb => sub { $result_ready->send },
		430	);
		431
		432	# this "blocks" (while handling events) till the callback
		433	# calls send
		434	$result_ready->recv;
		435
		436	Example: wait for a timer, but take advantage of the fact that condition
		437	variables are also code references.
		438
		439	my $done = AnyEvent->condvar;
		440	my $delay = AnyEvent->timer (after => 5, cb => $done);
		441	$done->recv;
		442
		443	METHODS FOR PRODUCERS
		444	These methods should only be used by the producing side, i.e. the
		445	code/module that eventually sends the signal. Note that it is also the
		446	producer side which creates the condvar in most cases, but it isn't
		447	uncommon for the consumer to create it as well.
		448
		449	$cv->send (...)
		450	Flag the condition as ready - a running "->recv" and all further
		451	calls to "recv" will (eventually) return after this method has been
		452	called. If nobody is waiting the send will be remembered.
		453
		454	If a callback has been set on the condition variable, it is called
		455	immediately from within send.
		456
		457	Any arguments passed to the "send" call will be returned by all
		458	future "->recv" calls.
		459
		460	Condition variables are overloaded so one can call them directly (as
		461	a code reference). Calling them directly is the same as calling
		462	"send". Note, however, that many C-based event loops do not handle
		463	overloading, so as tempting as it may be, passing a condition
		464	variable instead of a callback does not work. Both the pure perl and
		465	EV loops support overloading, however, as well as all functions that
		466	use perl to invoke a callback (as in AnyEvent::Socket and
		467	AnyEvent::DNS for example).
		468
		469	$cv->croak ($error)
		470	Similar to send, but causes all call's to "->recv" to invoke
		471	"Carp::croak" with the given error message/object/scalar.
		472
		473	This can be used to signal any errors to the condition variable
		474	user/consumer.
		475
		476	$cv->begin ([group callback])
		477	$cv->end
		478	These two methods are EXPERIMENTAL and MIGHT CHANGE.
		479
		480	These two methods can be used to combine many transactions/events
		481	into one. For example, a function that pings many hosts in parallel
		482	might want to use a condition variable for the whole process.
		483
		484	Every call to "->begin" will increment a counter, and every call to
		485	"->end" will decrement it. If the counter reaches 0 in "->end", the
		486	(last) callback passed to "begin" will be executed. That callback is
		487	supposed to call "->send", but that is not required. If no
		488	callback was set, "send" will be called without any arguments.
		489
		490	Let's clarify this with the ping example:
		491
		492	my $cv = AnyEvent->condvar;
		493
		494	my %result;
		495	$cv->begin (sub { $cv->send (\%result) });
		496
		497	for my $host (@list_of_hosts) {
		498	$cv->begin;
		499	ping_host_then_call_callback $host, sub {
		500	$result{$host} = ...;
		501	$cv->end;
		502	};
		503	}
		504
		505	$cv->end;
		506
		507	This code fragment supposedly pings a number of hosts and calls
		508	"send" after results for all then have have been gathered - in any
		509	order. To achieve this, the code issues a call to "begin" when it
		510	starts each ping request and calls "end" when it has received some
		511	result for it. Since "begin" and "end" only maintain a counter, the
		512	order in which results arrive is not relevant.
		513
		514	There is an additional bracketing call to "begin" and "end" outside
		515	the loop, which serves two important purposes: first, it sets the
		516	callback to be called once the counter reaches 0, and second, it
		517	ensures that "send" is called even when "no" hosts are being pinged
		518	(the loop doesn't execute once).
		519
		520	This is the general pattern when you "fan out" into multiple
		521	subrequests: use an outer "begin"/"end" pair to set the callback and
		522	ensure "end" is called at least once, and then, for each subrequest
		523	you start, call "begin" and for each subrequest you finish, call
		524	"end".
		525
		526	METHODS FOR CONSUMERS
		527	These methods should only be used by the consuming side, i.e. the code
		528	awaits the condition.
		529
		530	$cv->recv
280	Wait (blocking if necessary) until the "->broadcast" method has been	531	Wait (blocking if necessary) until the "->send" or "->croak" methods
281	called on c<$cv>, while servicing other watchers normally.	532	have been called on c<$cv>, while servicing other watchers normally.
282		533
283	You can only wait once on a condition - additional calls will return	534	You can only wait once on a condition - additional calls are valid
284	immediately.	535	but will return immediately.
		536
		537	If an error condition has been set by calling "->croak", then this
		538	function will call "croak".
		539
		540	In list context, all parameters passed to "send" will be returned,
		541	in scalar context only the first one will be returned.
285		542
286	Not all event models support a blocking wait - some die in that case	543	Not all event models support a blocking wait - some die in that case
287	(programs might want to do that to stay interactive), so *if you are	544	(programs might want to do that to stay interactive), so *if you are
288	using this from a module, never require a blocking wait*, but let	545	using this from a module, never require a blocking wait*, but let
289	the caller decide whether the call will block or not (for example,	546	the caller decide whether the call will block or not (for example,
290	by coupling condition variables with some kind of request results	547	by coupling condition variables with some kind of request results
291	and supporting callbacks so the caller knows that getting the result	548	and supporting callbacks so the caller knows that getting the result
292	will not block, while still suppporting blocking waits if the caller	549	will not block, while still supporting blocking waits if the caller
293	so desires).	550	so desires).
294		551
295	Another reason never to "->wait" in a module is that you cannot	552	Another reason never to "->recv" in a module is that you cannot
296	sensibly have two "->wait"'s in parallel, as that would require	553	sensibly have two "->recv"'s in parallel, as that would require
297	multiple interpreters or coroutines/threads, none of which	554	multiple interpreters or coroutines/threads, none of which
298	"AnyEvent" can supply (the coroutine-aware backends	555	"AnyEvent" can supply.
299	AnyEvent::Impl::CoroEV and AnyEvent::Impl::CoroEvent explicitly
300	support concurrent "->wait"'s from different coroutines, however).
301		556
302	$cv->broadcast	557	The Coro module, however, can and does supply coroutines and, in
303	Flag the condition as ready - a running "->wait" and all further	558	fact, Coro::AnyEvent replaces AnyEvent's condvars by coroutine-safe
304	calls to "wait" will (eventually) return after this method has been	559	versions and also integrates coroutines into AnyEvent, making
305	called. If nobody is waiting the broadcast will be remembered..	560	blocking "->recv" calls perfectly safe as long as they are done from
		561	another coroutine (one that doesn't run the event loop).
306		562
307	Example:	563	You can ensure that "-recv" never blocks by setting a callback and
		564	only calling "->recv" from within that callback (or at a later
		565	time). This will work even when the event loop does not support
		566	blocking waits otherwise.
308		567
309	# wait till the result is ready	568	$bool = $cv->ready
310	my $result_ready = AnyEvent->condvar;	569	Returns true when the condition is "true", i.e. whether "send" or
		570	"croak" have been called.
311		571
312	# do something such as adding a timer	572	$cb = $cv->cb ([new callback])
313	# or socket watcher the calls $result_ready->broadcast	573	This is a mutator function that returns the callback set and
314	# when the "result" is ready.	574	optionally replaces it before doing so.
315	# in this case, we simply use a timer:
316	my $w = AnyEvent->timer (
317	after => 1,
318	cb => sub { $result_ready->broadcast },
319	);
320		575
321	# this "blocks" (while handling events) till the watcher	576	The callback will be called when the condition becomes "true", i.e.
322	# calls broadcast	577	when "send" or "croak" are called, with the only argument being the
323	$result_ready->wait;	578	condition variable itself. Calling "recv" inside the callback or at
		579	any later time is guaranteed not to block.
324		580
325	GLOBAL VARIABLES AND FUNCTIONS	581	GLOBAL VARIABLES AND FUNCTIONS
326	$AnyEvent::MODEL	582	$AnyEvent::MODEL
327	Contains "undef" until the first watcher is being created. Then it	583	Contains "undef" until the first watcher is being created. Then it
328	contains the event model that is being used, which is the name of	584	contains the event model that is being used, which is the name of
…		…
330	the "AnyEvent::Impl:xxx" modules, but can be any other class in the	586	the "AnyEvent::Impl:xxx" modules, but can be any other class in the
331	case AnyEvent has been extended at runtime (e.g. in rxvt-unicode).	587	case AnyEvent has been extended at runtime (e.g. in rxvt-unicode).
332		588
333	The known classes so far are:	589	The known classes so far are:
334		590
335	AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
336	AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
337	AnyEvent::Impl::EV based on EV (an interface to libev, also best choice).	591	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
338	AnyEvent::Impl::Event based on Event, also second best choice :)	592	AnyEvent::Impl::Event based on Event, second best choice.
		593	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
339	AnyEvent::Impl::Glib based on Glib, third-best choice.	594	AnyEvent::Impl::Glib based on Glib, third-best choice.
340	AnyEvent::Impl::Tk based on Tk, very bad choice.	595	AnyEvent::Impl::Tk based on Tk, very bad choice.
341	AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable.	596	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
		597	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
		598	AnyEvent::Impl::POE based on POE, not generic enough for full support.
		599
		600	There is no support for WxWidgets, as WxWidgets has no support for
		601	watching file handles. However, you can use WxWidgets through the
		602	POE Adaptor, as POE has a Wx backend that simply polls 20 times per
		603	second, which was considered to be too horrible to even consider for
		604	AnyEvent. Likewise, other POE backends can be used by AnyEvent by
		605	using it's adaptor.
		606
		607	AnyEvent knows about Prima and Wx and will try to use POE when
		608	autodetecting them.
342		609
343	AnyEvent::detect	610	AnyEvent::detect
344	Returns $AnyEvent::MODEL, forcing autodetection of the event model	611	Returns $AnyEvent::MODEL, forcing autodetection of the event model
345	if necessary. You should only call this function right before you	612	if necessary. You should only call this function right before you
346	would have created an AnyEvent watcher anyway, that is, as late as	613	would have created an AnyEvent watcher anyway, that is, as late as
347	possible at runtime.	614	possible at runtime.
348		615
		616	$guard = AnyEvent::post_detect { BLOCK }
		617	Arranges for the code block to be executed as soon as the event
		618	model is autodetected (or immediately if this has already happened).
		619
		620	If called in scalar or list context, then it creates and returns an
		621	object that automatically removes the callback again when it is
		622	destroyed. See Coro::BDB for a case where this is useful.
		623
		624	@AnyEvent::post_detect
		625	If there are any code references in this array (you can "push" to it
		626	before or after loading AnyEvent), then they will called directly
		627	after the event loop has been chosen.
		628
		629	You should check $AnyEvent::MODEL before adding to this array,
		630	though: if it contains a true value then the event loop has already
		631	been detected, and the array will be ignored.
		632
		633	Best use "AnyEvent::post_detect { BLOCK }" instead.
		634
349	WHAT TO DO IN A MODULE	635	WHAT TO DO IN A MODULE
350	As a module author, you should "use AnyEvent" and call AnyEvent methods	636	As a module author, you should "use AnyEvent" and call AnyEvent methods
351	freely, but you should not load a specific event module or rely on it.	637	freely, but you should not load a specific event module or rely on it.
352		638
353	Be careful when you create watchers in the module body - AnyEvent will	639	Be careful when you create watchers in the module body - AnyEvent will
354	decide which event module to use as soon as the first method is called,	640	decide which event module to use as soon as the first method is called,
355	so by calling AnyEvent in your module body you force the user of your	641	so by calling AnyEvent in your module body you force the user of your
356	module to load the event module first.	642	module to load the event module first.
357		643
358	Never call "->wait" on a condition variable unless you know that the	644	Never call "->recv" on a condition variable unless you know that the
359	"->broadcast" method has been called on it already. This is because it	645	"->send" method has been called on it already. This is because it will
360	will stall the whole program, and the whole point of using events is to	646	stall the whole program, and the whole point of using events is to stay
361	stay interactive.	647	interactive.
362		648
363	It is fine, however, to call "->wait" when the user of your module	649	It is fine, however, to call "->recv" when the user of your module
364	requests it (i.e. if you create a http request object ad have a method	650	requests it (i.e. if you create a http request object ad have a method
365	called "results" that returns the results, it should call "->wait"	651	called "results" that returns the results, it should call "->recv"
366	freely, as the user of your module knows what she is doing. always).	652	freely, as the user of your module knows what she is doing. always).
367		653
368	WHAT TO DO IN THE MAIN PROGRAM	654	WHAT TO DO IN THE MAIN PROGRAM
369	There will always be a single main program - the only place that should	655	There will always be a single main program - the only place that should
370	dictate which event model to use.	656	dictate which event model to use.
…		…
372	If it doesn't care, it can just "use AnyEvent" and use it itself, or not	658	If it doesn't care, it can just "use AnyEvent" and use it itself, or not
373	do anything special (it does not need to be event-based) and let	659	do anything special (it does not need to be event-based) and let
374	AnyEvent decide which implementation to chose if some module relies on	660	AnyEvent decide which implementation to chose if some module relies on
375	it.	661	it.
376		662
377	If the main program relies on a specific event model. For example, in	663	If the main program relies on a specific event model - for example, in
378	Gtk2 programs you have to rely on the Glib module. You should load the	664	Gtk2 programs you have to rely on the Glib module - you should load the
379	event module before loading AnyEvent or any module that uses it:	665	event module before loading AnyEvent or any module that uses it:
380	generally speaking, you should load it as early as possible. The reason	666	generally speaking, you should load it as early as possible. The reason
381	is that modules might create watchers when they are loaded, and AnyEvent	667	is that modules might create watchers when they are loaded, and AnyEvent
382	will decide on the event model to use as soon as it creates watchers,	668	will decide on the event model to use as soon as it creates watchers,
383	and it might chose the wrong one unless you load the correct one	669	and it might chose the wrong one unless you load the correct one
384	yourself.	670	yourself.
385		671
386	You can chose to use a rather inefficient pure-perl implementation by	672	You can chose to use a pure-perl implementation by loading the
387	loading the "AnyEvent::Impl::Perl" module, which gives you similar	673	"AnyEvent::Impl::Perl" module, which gives you similar behaviour
388	behaviour everywhere, but letting AnyEvent chose is generally better.	674	everywhere, but letting AnyEvent chose the model is generally better.
		675
		676	MAINLOOP EMULATION
		677	Sometimes (often for short test scripts, or even standalone programs who
		678	only want to use AnyEvent), you do not want to run a specific event
		679	loop.
		680
		681	In that case, you can use a condition variable like this:
		682
		683	AnyEvent->condvar->recv;
		684
		685	This has the effect of entering the event loop and looping forever.
		686
		687	Note that usually your program has some exit condition, in which case it
		688	is better to use the "traditional" approach of storing a condition
		689	variable somewhere, waiting for it, and sending it when the program
		690	should exit cleanly.
		691
		692	OTHER MODULES
		693	The following is a non-exhaustive list of additional modules that use
		694	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		695	in the same program. Some of the modules come with AnyEvent, some are
		696	available via CPAN.
		697
		698	AnyEvent::Util
		699	Contains various utility functions that replace often-used but
		700	blocking functions such as "inet_aton" by event-/callback-based
		701	versions.
		702
		703	AnyEvent::Handle
		704	Provide read and write buffers and manages watchers for reads and
		705	writes.
		706
		707	AnyEvent::Socket
		708	Provides various utility functions for (internet protocol) sockets,
		709	addresses and name resolution. Also functions to create non-blocking
		710	tcp connections or tcp servers, with IPv6 and SRV record support and
		711	more.
		712
		713	AnyEvent::DNS
		714	Provides rich asynchronous DNS resolver capabilities.
		715
		716	AnyEvent::HTTPD
		717	Provides a simple web application server framework.
		718
		719	AnyEvent::FastPing
		720	The fastest ping in the west.
		721
		722	Net::IRC3
		723	AnyEvent based IRC client module family.
		724
		725	Net::XMPP2
		726	AnyEvent based XMPP (Jabber protocol) module family.
		727
		728	Net::FCP
		729	AnyEvent-based implementation of the Freenet Client Protocol,
		730	birthplace of AnyEvent.
		731
		732	Event::ExecFlow
		733	High level API for event-based execution flow control.
		734
		735	Coro
		736	Has special support for AnyEvent via Coro::AnyEvent.
		737
		738	AnyEvent::AIO, IO::AIO
		739	Truly asynchronous I/O, should be in the toolbox of every event
		740	programmer. AnyEvent::AIO transparently fuses IO::AIO and AnyEvent
		741	together.
		742
		743	AnyEvent::BDB, BDB
		744	Truly asynchronous Berkeley DB access. AnyEvent::AIO transparently
		745	fuses IO::AIO and AnyEvent together.
		746
		747	IO::Lambda
		748	The lambda approach to I/O - don't ask, look there. Can use
		749	AnyEvent.
389		750
390	SUPPLYING YOUR OWN EVENT MODEL INTERFACE	751	SUPPLYING YOUR OWN EVENT MODEL INTERFACE
391	This is an advanced topic that you do not normally need to use AnyEvent	752	This is an advanced topic that you do not normally need to use AnyEvent
392	in a module. This section is only of use to event loop authors who want	753	in a module. This section is only of use to event loop authors who want
393	to provide AnyEvent compatibility.	754	to provide AnyEvent compatibility.
…		…
431	must not be done in an interactive application, so it makes sense.	792	must not be done in an interactive application, so it makes sense.
432		793
433	ENVIRONMENT VARIABLES	794	ENVIRONMENT VARIABLES
434	The following environment variables are used by this module:	795	The following environment variables are used by this module:
435		796
436	"PERL_ANYEVENT_VERBOSE" when set to 2 or higher, cause AnyEvent to	797	"PERL_ANYEVENT_VERBOSE"
437	report to STDERR which event model it chooses.	798	By default, AnyEvent will be completely silent except in fatal
		799	conditions. You can set this environment variable to make AnyEvent
		800	more talkative.
		801
		802	When set to 1 or higher, causes AnyEvent to warn about unexpected
		803	conditions, such as not being able to load the event model specified
		804	by "PERL_ANYEVENT_MODEL".
		805
		806	When set to 2 or higher, cause AnyEvent to report to STDERR which
		807	event model it chooses.
		808
		809	"PERL_ANYEVENT_MODEL"
		810	This can be used to specify the event model to be used by AnyEvent,
		811	before auto detection and -probing kicks in. It must be a string
		812	consisting entirely of ASCII letters. The string "AnyEvent::Impl::"
		813	gets prepended and the resulting module name is loaded and if the
		814	load was successful, used as event model. If it fails to load
		815	AnyEvent will proceed with auto detection and -probing.
		816
		817	This functionality might change in future versions.
		818
		819	For example, to force the pure perl model (AnyEvent::Impl::Perl) you
		820	could start your program like this:
		821
		822	PERL_ANYEVENT_MODEL=Perl perl ...
		823
		824	"PERL_ANYEVENT_PROTOCOLS"
		825	Used by both AnyEvent::DNS and AnyEvent::Socket to determine
		826	preferences for IPv4 or IPv6. The default is unspecified (and might
		827	change, or be the result of auto probing).
		828
		829	Must be set to a comma-separated list of protocols or address
		830	families, current supported: "ipv4" and "ipv6". Only protocols
		831	mentioned will be used, and preference will be given to protocols
		832	mentioned earlier in the list.
		833
		834	This variable can effectively be used for denial-of-service attacks
		835	against local programs (e.g. when setuid), although the impact is
		836	likely small, as the program has to handle connection errors
		837	already-
		838
		839	Examples: "PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6" - prefer IPv4 over
		840	IPv6, but support both and try to use both.
		841	"PERL_ANYEVENT_PROTOCOLS=ipv4" - only support IPv4, never try to
		842	resolve or contact IPv6 addresses.
		843	"PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4" support either IPv4 or IPv6, but
		844	prefer IPv6 over IPv4.
		845
		846	"PERL_ANYEVENT_EDNS0"
		847	Used by AnyEvent::DNS to decide whether to use the EDNS0 extension
		848	for DNS. This extension is generally useful to reduce DNS traffic,
		849	but some (broken) firewalls drop such DNS packets, which is why it
		850	is off by default.
		851
		852	Setting this variable to 1 will cause AnyEvent::DNS to announce
		853	EDNS0 in its DNS requests.
		854
		855	"PERL_ANYEVENT_MAX_FORKS"
		856	The maximum number of child processes that
		857	"AnyEvent::Util::fork_call" will create in parallel.
438		858
439	EXAMPLE PROGRAM	859	EXAMPLE PROGRAM
440	The following program uses an IO watcher to read data from STDIN, a	860	The following program uses an I/O watcher to read data from STDIN, a
441	timer to display a message once per second, and a condition variable to	861	timer to display a message once per second, and a condition variable to
442	quit the program when the user enters quit:	862	quit the program when the user enters quit:
443		863
444	use AnyEvent;	864	use AnyEvent;
445		865
…		…
450	poll => 'r',	870	poll => 'r',
451	cb => sub {	871	cb => sub {
452	warn "io event <$_[0]>\n"; # will always output <r>	872	warn "io event <$_[0]>\n"; # will always output <r>
453	chomp (my $input = <STDIN>); # read a line	873	chomp (my $input = <STDIN>); # read a line
454	warn "read: $input\n"; # output what has been read	874	warn "read: $input\n"; # output what has been read
455	$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i	875	$cv->send if $input =~ /^q/i; # quit program if /^q/i
456	},	876	},
457	);	877	);
458		878
459	my $time_watcher; # can only be used once	879	my $time_watcher; # can only be used once
460		880
…		…
465	});	885	});
466	}	886	}
467		887
468	new_timer; # create first timer	888	new_timer; # create first timer
469		889
470	$cv->wait; # wait until user enters /^q/i	890	$cv->recv; # wait until user enters /^q/i
471		891
472	REAL-WORLD EXAMPLE	892	REAL-WORLD EXAMPLE
473	Consider the Net::FCP module. It features (among others) the following	893	Consider the Net::FCP module. It features (among others) the following
474	API calls, which are to freenet what HTTP GET requests are to http:	894	API calls, which are to freenet what HTTP GET requests are to http:
475		895
…		…
524	syswrite $txn->{fh}, $txn->{request}	944	syswrite $txn->{fh}, $txn->{request}
525	or die "connection or write error";	945	or die "connection or write error";
526	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });	946	$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
527		947
528	Again, "fh_ready_r" waits till all data has arrived, and then stores the	948	Again, "fh_ready_r" waits till all data has arrived, and then stores the
529	result and signals any possible waiters that the request ahs finished:	949	result and signals any possible waiters that the request has finished:
530		950
531	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};	951	sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
532		952
533	if (end-of-file or data complete) {	953	if (end-of-file or data complete) {
534	$txn->{result} = $txn->{buf};	954	$txn->{result} = $txn->{buf};
535	$txn->{finished}->broadcast;	955	$txn->{finished}->send;
536	$txb->{cb}->($txn) of $txn->{cb}; # also call callback	956	$txb->{cb}->($txn) of $txn->{cb}; # also call callback
537	}	957	}
538		958
539	The "result" method, finally, just waits for the finished signal (if the	959	The "result" method, finally, just waits for the finished signal (if the
540	request was already finished, it doesn't wait, of course, and returns	960	request was already finished, it doesn't wait, of course, and returns
541	the data:	961	the data:
542		962
543	$txn->{finished}->wait;	963	$txn->{finished}->recv;
544	return $txn->{result};	964	return $txn->{result};
545		965
546	The actual code goes further and collects all errors ("die"s,	966	The actual code goes further and collects all errors ("die"s,
547	exceptions) that occured during request processing. The "result" method	967	exceptions) that occurred during request processing. The "result" method
548	detects whether an exception as thrown (it is stored inside the $txn	968	detects whether an exception as thrown (it is stored inside the $txn
549	object) and just throws the exception, which means connection errors and	969	object) and just throws the exception, which means connection errors and
550	other problems get reported tot he code that tries to use the result,	970	other problems get reported tot he code that tries to use the result,
551	not in a random callback.	971	not in a random callback.
552		972
…		…
583		1003
584	my $quit = AnyEvent->condvar;	1004	my $quit = AnyEvent->condvar;
585		1005
586	$fcp->txn_client_get ($url)->cb (sub {	1006	$fcp->txn_client_get ($url)->cb (sub {
587	...	1007	...
588	$quit->broadcast;	1008	$quit->send;
589	});	1009	});
590		1010
591	$quit->wait;	1011	$quit->recv;
		1012
		1013	BENCHMARKS
		1014	To give you an idea of the performance and overheads that AnyEvent adds
		1015	over the event loops themselves and to give you an impression of the
		1016	speed of various event loops I prepared some benchmarks.
		1017
		1018	BENCHMARKING ANYEVENT OVERHEAD
		1019	Here is a benchmark of various supported event models used natively and
		1020	through AnyEvent. The benchmark creates a lot of timers (with a zero
		1021	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
		1022	which it is), lets them fire exactly once and destroys them again.
		1023
		1024	Source code for this benchmark is found as eg/bench in the AnyEvent
		1025	distribution.
		1026
		1027	Explanation of the columns
		1028	watcher is the number of event watchers created/destroyed. Since
		1029	different event models feature vastly different performances, each event
		1030	loop was given a number of watchers so that overall runtime is
		1031	acceptable and similar between tested event loop (and keep them from
		1032	crashing): Glib would probably take thousands of years if asked to
		1033	process the same number of watchers as EV in this benchmark.
		1034
		1035	bytes is the number of bytes (as measured by the resident set size,
		1036	RSS) consumed by each watcher. This method of measuring captures both C
		1037	and Perl-based overheads.
		1038
		1039	create is the time, in microseconds (millionths of seconds), that it
		1040	takes to create a single watcher. The callback is a closure shared
		1041	between all watchers, to avoid adding memory overhead. That means
		1042	closure creation and memory usage is not included in the figures.
		1043
		1044	invoke is the time, in microseconds, used to invoke a simple callback.
		1045	The callback simply counts down a Perl variable and after it was invoked
		1046	"watcher" times, it would "->send" a condvar once to signal the end of
		1047	this phase.
		1048
		1049	destroy is the time, in microseconds, that it takes to destroy a
		1050	single watcher.
		1051
		1052	Results
		1053	name watchers bytes create invoke destroy comment
		1054	EV/EV 400000 244 0.56 0.46 0.31 EV native interface
		1055	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
		1056	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
		1057	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
		1058	Event/Event 16000 516 31.88 31.30 0.85 Event native interface
		1059	Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
		1060	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
		1061	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
		1062	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
		1063	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
		1064
		1065	Discussion
		1066	The benchmark does not measure scalability of the event loop very
		1067	well. For example, a select-based event loop (such as the pure perl one)
		1068	can never compete with an event loop that uses epoll when the number of
		1069	file descriptors grows high. In this benchmark, all events become ready
		1070	at the same time, so select/poll-based implementations get an unnatural
		1071	speed boost.
		1072
		1073	Also, note that the number of watchers usually has a nonlinear effect on
		1074	overall speed, that is, creating twice as many watchers doesn't take
		1075	twice the time - usually it takes longer. This puts event loops tested
		1076	with a higher number of watchers at a disadvantage.
		1077
		1078	To put the range of results into perspective, consider that on the
		1079	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1080	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000
		1081	CPU cycles with POE.
		1082
		1083	"EV" is the sole leader regarding speed and memory use, which are both
		1084	maximal/minimal, respectively. Even when going through AnyEvent, it uses
		1085	far less memory than any other event loop and is still faster than Event
		1086	natively.
		1087
		1088	The pure perl implementation is hit in a few sweet spots (both the
		1089	constant timeout and the use of a single fd hit optimisations in the
		1090	perl interpreter and the backend itself). Nevertheless this shows that
		1091	it adds very little overhead in itself. Like any select-based backend
		1092	its performance becomes really bad with lots of file descriptors (and
		1093	few of them active), of course, but this was not subject of this
		1094	benchmark.
		1095
		1096	The "Event" module has a relatively high setup and callback invocation
		1097	cost, but overall scores in on the third place.
		1098
		1099	"Glib"'s memory usage is quite a bit higher, but it features a faster
		1100	callback invocation and overall ends up in the same class as "Event".
		1101	However, Glib scales extremely badly, doubling the number of watchers
		1102	increases the processing time by more than a factor of four, making it
		1103	completely unusable when using larger numbers of watchers (note that
		1104	only a single file descriptor was used in the benchmark, so
		1105	inefficiencies of "poll" do not account for this).
		1106
		1107	The "Tk" adaptor works relatively well. The fact that it crashes with
		1108	more than 2000 watchers is a big setback, however, as correctness takes
		1109	precedence over speed. Nevertheless, its performance is surprising, as
		1110	the file descriptor is dup()ed for each watcher. This shows that the
		1111	dup() employed by some adaptors is not a big performance issue (it does
		1112	incur a hidden memory cost inside the kernel which is not reflected in
		1113	the figures above).
		1114
		1115	"POE", regardless of underlying event loop (whether using its pure perl
		1116	select-based backend or the Event module, the POE-EV backend couldn't be
		1117	tested because it wasn't working) shows abysmal performance and memory
		1118	usage with AnyEvent: Watchers use almost 30 times as much memory as EV
		1119	watchers, and 10 times as much memory as Event (the high memory
		1120	requirements are caused by requiring a session for each watcher).
		1121	Watcher invocation speed is almost 900 times slower than with AnyEvent's
		1122	pure perl implementation.
		1123
		1124	The design of the POE adaptor class in AnyEvent can not really account
		1125	for the performance issues, though, as session creation overhead is
		1126	small compared to execution of the state machine, which is coded pretty
		1127	optimally within AnyEvent::Impl::POE (and while everybody agrees that
		1128	using multiple sessions is not a good approach, especially regarding
		1129	memory usage, even the author of POE could not come up with a faster
		1130	design).
		1131
		1132	Summary
		1133	* Using EV through AnyEvent is faster than any other event loop (even
		1134	when used without AnyEvent), but most event loops have acceptable
		1135	performance with or without AnyEvent.
		1136
		1137	* The overhead AnyEvent adds is usually much smaller than the overhead
		1138	of the actual event loop, only with extremely fast event loops such
		1139	as EV adds AnyEvent significant overhead.
		1140
		1141	* You should avoid POE like the plague if you want performance or
		1142	reasonable memory usage.
		1143
		1144	BENCHMARKING THE LARGE SERVER CASE
		1145	This benchmark actually benchmarks the event loop itself. It works by
		1146	creating a number of "servers": each server consists of a socket pair, a
		1147	timeout watcher that gets reset on activity (but never fires), and an
		1148	I/O watcher waiting for input on one side of the socket. Each time the
		1149	socket watcher reads a byte it will write that byte to a random other
		1150	"server".
		1151
		1152	The effect is that there will be a lot of I/O watchers, only part of
		1153	which are active at any one point (so there is a constant number of
		1154	active fds for each loop iteration, but which fds these are is random).
		1155	The timeout is reset each time something is read because that reflects
		1156	how most timeouts work (and puts extra pressure on the event loops).
		1157
		1158	In this benchmark, we use 10000 socket pairs (20000 sockets), of which
		1159	100 (1%) are active. This mirrors the activity of large servers with
		1160	many connections, most of which are idle at any one point in time.
		1161
		1162	Source code for this benchmark is found as eg/bench2 in the AnyEvent
		1163	distribution.
		1164
		1165	Explanation of the columns
		1166	sockets is the number of sockets, and twice the number of "servers"
		1167	(as each server has a read and write socket end).
		1168
		1169	create is the time it takes to create a socket pair (which is
		1170	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1171
		1172	request, the most important value, is the time it takes to handle a
		1173	single "request", that is, reading the token from the pipe and
		1174	forwarding it to another server. This includes deleting the old timeout
		1175	and creating a new one that moves the timeout into the future.
		1176
		1177	Results
		1178	name sockets create request
		1179	EV 20000 69.01 11.16
		1180	Perl 20000 73.32 35.87
		1181	Event 20000 212.62 257.32
		1182	Glib 20000 651.16 1896.30
		1183	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1184
		1185	Discussion
		1186	This benchmark does measure scalability and overall performance of the
		1187	particular event loop.
		1188
		1189	EV is again fastest. Since it is using epoll on my system, the setup
		1190	time is relatively high, though.
		1191
		1192	Perl surprisingly comes second. It is much faster than the C-based event
		1193	loops Event and Glib.
		1194
		1195	Event suffers from high setup time as well (look at its code and you
		1196	will understand why). Callback invocation also has a high overhead
		1197	compared to the "$_->() for .."-style loop that the Perl event loop
		1198	uses. Event uses select or poll in basically all documented
		1199	configurations.
		1200
		1201	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1202	clearly fails to perform with many filehandles or in busy servers.
		1203
		1204	POE is still completely out of the picture, taking over 1000 times as
		1205	long as EV, and over 100 times as long as the Perl implementation, even
		1206	though it uses a C-based event loop in this case.
		1207
		1208	Summary
		1209	* The pure perl implementation performs extremely well.
		1210
		1211	* Avoid Glib or POE in large projects where performance matters.
		1212
		1213	BENCHMARKING SMALL SERVERS
		1214	While event loops should scale (and select-based ones do not...) even to
		1215	large servers, most programs we (or I :) actually write have only a few
		1216	I/O watchers.
		1217
		1218	In this benchmark, I use the same benchmark program as in the large
		1219	server case, but it uses only eight "servers", of which three are active
		1220	at any one time. This should reflect performance for a small server
		1221	relatively well.
		1222
		1223	The columns are identical to the previous table.
		1224
		1225	Results
		1226	name sockets create request
		1227	EV 16 20.00 6.54
		1228	Perl 16 25.75 12.62
		1229	Event 16 81.27 35.86
		1230	Glib 16 32.63 15.48
		1231	POE 16 261.87 276.28 uses POE::Loop::Event
		1232
		1233	Discussion
		1234	The benchmark tries to test the performance of a typical small server.
		1235	While knowing how various event loops perform is interesting, keep in
		1236	mind that their overhead in this case is usually not as important, due
		1237	to the small absolute number of watchers (that is, you need efficiency
		1238	and speed most when you have lots of watchers, not when you only have a
		1239	few of them).
		1240
		1241	EV is again fastest.
		1242
		1243	Perl again comes second. It is noticeably faster than the C-based event
		1244	loops Event and Glib, although the difference is too small to really
		1245	matter.
		1246
		1247	POE also performs much better in this case, but is is still far behind
		1248	the others.
		1249
		1250	Summary
		1251	* C-based event loops perform very well with small number of watchers,
		1252	as the management overhead dominates.
		1253
		1254	FORK
		1255	Most event libraries are not fork-safe. The ones who are usually are
		1256	because they rely on inefficient but fork-safe "select" or "poll" calls.
		1257	Only EV is fully fork-aware.
		1258
		1259	If you have to fork, you must either do so before creating your first
		1260	watcher OR you must not use AnyEvent at all in the child.
		1261
		1262	SECURITY CONSIDERATIONS
		1263	AnyEvent can be forced to load any event model via
		1264	$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used
		1265	to execute arbitrary code or directly gain access, it can easily be used
		1266	to make the program hang or malfunction in subtle ways, as AnyEvent
		1267	watchers will not be active when the program uses a different event
		1268	model than specified in the variable.
		1269
		1270	You can make AnyEvent completely ignore this variable by deleting it
		1271	before the first watcher gets created, e.g. with a "BEGIN" block:
		1272
		1273	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
		1274
		1275	use AnyEvent;
		1276
		1277	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		1278	be used to probe what backend is used and gain other information (which
		1279	is probably even less useful to an attacker than PERL_ANYEVENT_MODEL).
592		1280
593	SEE ALSO	1281	SEE ALSO
594	Event modules: Coro::EV, EV, EV::Glib, Glib::EV, Coro::Event, Event,	1282	Utility functions: AnyEvent::Util.
595	Glib::Event, Glib, Coro, Tk.
596		1283
597	Implementations: AnyEvent::Impl::CoroEV, AnyEvent::Impl::EV,	1284	Event modules: EV, EV::Glib, Glib::EV, Event, Glib::Event, Glib, Tk,
598	AnyEvent::Impl::CoroEvent, AnyEvent::Impl::Event, AnyEvent::Impl::Glib,	1285	Event::Lib, Qt, POE.
		1286
		1287	Implementations: AnyEvent::Impl::EV, AnyEvent::Impl::Event,
599	AnyEvent::Impl::Tk, AnyEvent::Impl::Perl.	1288	AnyEvent::Impl::Glib, AnyEvent::Impl::Tk, AnyEvent::Impl::Perl,
		1289	AnyEvent::Impl::EventLib, AnyEvent::Impl::Qt, AnyEvent::Impl::POE.
600		1290
		1291	Non-blocking file handles, sockets, TCP clients and servers:
		1292	AnyEvent::Handle, AnyEvent::Socket.
		1293
		1294	Asynchronous DNS: AnyEvent::DNS.
		1295
		1296	Coroutine support: Coro, Coro::AnyEvent, Coro::EV, Coro::Event,
		1297
601	Nontrivial usage examples: Net::FCP, Net::XMPP2.	1298	Nontrivial usage examples: Net::FCP, Net::XMPP2, AnyEvent::DNS.
602		1299
		1300	AUTHOR
		1301	Marc Lehmann <schmorp@schmorp.de>
		1302	http://home.schmorp.de/
603		1303

Diff Legend

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

Comparing AnyEvent/README (file contents): Revision 1.16 by root, Sat Apr 19 04:58:14 2008 UTC vs. Revision 1.25 by root, Tue Jun 3 09:02:46 2008 UTC

Diff Legend

Comparing AnyEvent/README (file contents):
Revision 1.16 by root, Sat Apr 19 04:58:14 2008 UTC vs.
Revision 1.25 by root, Tue Jun 3 09:02:46 2008 UTC